2026 Pan-GLASS Conference

In mesoscale models, turbulent kinetic energy (TKE) dissipation is commonly parameterized as a function of bulk TKE, implicitly assuming isotropic turbulence in the convective boundary layer (CBL). In this study, we use long-term Doppler lidar observations at the Land-Atmosphere Feedback Observatory (LAFO), University of Hohenheim, Stuttgart, Germany to evaluate such parametrizations. Two continuously operated Doppler lidars, one in vertical staring mode and one in six-beam scanning mode, provide us high-resolution wind measurements within the CBL (Späth et al., 2023). We have analyzed the statistical relationships between vertical wind variance, TKE (Bonin et al., 2017), and TKE dissipation (Wulfmeyer et al., 2024) under daytime convective conditions. The results reveal a nonlinear relationship between vertical wind variance and TKE. The TKE-based dissipation parametrization from the Mellor-Yamada-Nakanishi-Niino (MYNN) scheme shows a lower correlation with the lidar data (~0.5), whereas the vertical wind variance-based dissipation shows a stronger correlation (~0.8). The turbulent length scales derived from TKE and vertical wind variance exhibit similar characteristics. These findings show limitations of bulk TKE-based parameterizations and demonstrate the value of Doppler-lidar-based diagnostics for improving the turbulence representation in mesoscale models.

Mon & Tue

11:00–12:30

Results from the PLUMBER2 experiment show us that land models (land surface, ecological and hydrological models) are not utilising the information available to them in their driving variables for flux prediction efficiently. They are outperformed in a wide range of metrics by simple empirical models and more complicated machine learning models that predict fluxes entirely out-of-sample, at a wide range of flux tower sites. While this might be confronting, it also means that by examining the precise circumstances in which we know that land models are underperforming relative to the empirical models, we can begin to understand the pathways within models that require improvement: there is a clear pathway forward.

This presentation will argue that the key to the resolution of these issues, and avoiding this happening again, is a comprehensive testbed that incorporates empirical benchmarks of kind used in PLUMBER2, that covers the suite of configurations or use-cases of a land model. We’ll detail efforts to create this testbed as a community tool, the nature of the evaluation results it produces, and the connection of this system to the Fluxnet Shuttle to expand the scope of testing, improve provenance information and reproducibility.

Tue

10:00–10:15

Accurately simulating latent and sensible heat fluxes is a long-standing open challenge in the land modeling community. The recent model intercomparison project PLUMBER 2, spanning 154 flux towers worldwide, showed that simple 1-variable linear regression models can outperform process-based models in simulating latent and sensible heat. PLUMBER 2 simulations were run using default model parameters, leaving the potential performance gains from parameter estimation unquantified.
Identifying optimal parameters in land models has several challenges, including high computational cost and the need to identify parameters that can correctly reproduce temporal dynamics (i.e., good performance across different time epochs) and spatial patterns (i.e., good performance across many sites). To evaluate the ability of different calibration methods to handle these challenges, this study compared the performance of traditional and machine-learning emulator-based calibration methods against multi-frequency Long Short-Term Memory (LSTM) benchmarks, with single-objective experiments (latent heat or sensible heat calibrated individually) and multi-objective experiments (latent and sensible heat calibrated simultaneously). We also tested two ways to train emulators and LSTMs: either considering one site at a time or leveraging information from multiple sites and their attributes simultaneously.
Our results show that the calibrated simulations outperformed the default parameters and the simple benchmarks used in PLUMBER 2, demonstrating the potential to improve process-based models. Moreover, we observed that traditional calibration methods have a tendency to overfit: these traditional calibration methods can achieve high performance during calibration but are unable to achieve similar results during validation. The emulator-based methods achieve more consistent results across both calibration and validation time periods. Additionally, we found that parameter estimation methods that incorporate information from multiple sites simultaneously achieve better spatial consistency than methods that only learn from one site at a time. These results suggest that the performance gap between LSTM and process-based models can be significantly narrowed through calibration.

Tue

09:30–09:45

Seasonally frozen ground (SFG) covers nearly half of the Northern Hemisphere (NH) and plays an important role
in altering water and energy fluxes, thereby influencing atmospheric circulation, hydrology, ecosystems, and the carbon cycle.
However, land surface models (LSMs) differ in simulating winter soil conditions due to the complex freeze-thaw processes
and snow-soil interactions, leading to uncertainties in spring/summer soil moisture and runoff. This study evaluates standalone
simulations from two LSMs, JSBACH and CLM, driven by ERA5 data (1986-2022), compares them with ERA5-Land to study5
model differences across cold regions, and then examines all three models against the reference RIHMI-WDC observational
dataset at 26 sites in order to gain a better understanding of the representation of SFG characteristics. The research aims to
identify biases in simulated SFG characteristics, investigate their underlying drivers, and assess how snow cover errors propa-
gate into frozen-ground biases using site-level evaluation over Russia. The importance of snow parameterization is highlighted
in this study through an improvement to the snow density scheme in JSBACH, which reduced its cold bias in soil temperature10
by up to 10–20°C, and this improved version was used for the comparative analysis. Among the models assessed in this study,
JSBACH reproduced frozen ground extent more realistically but showed a cold soil bias and deeper freezing due to insufficient
snow insulation, whereas CLM simulated soil temperatures closer to observations with excess snow and soil ice, and ERA5-
Land is warm biased, leading to an underestimation of permafrost extent. Longer frozen ground durations are simulated with
contrasting snow cover characteristics, where CLM and JSBACH exhibit different snow insulation effects. All models showed15
notable biases in both SFG and snow cover, and the results reveal that they differ in how control of soil freeze-thaw is shared
between air temperature and snow processes. The study highlights that improving model performance requires better snow
representation and a detailed assessment aimed at enhancing the parameterization of soil thermal and hydraulic properties.

Wed & Thu

11:00–12:30

Compound hot–dry events challenge land–atmosphere modeling because feedbacks among vegetation dynamics, soil moisture depletion, and surface energy partitioning can amplify temperature extremes. This study investigates how satellite-derived biophysical parameters influence regional climate model biases during the extreme summers of 2018 and 2019 over Belgium.
We perform 1.5 km resolution simulations in two complementary configurations: (i) coupled atmospheric–land surface simulations with ALARO–SURFEX, and (ii) standalone offline SURFEX simulations forced by ERA5. In each, a control run using default biophysical parameters is compared with an experiment in which biophysical parameters, such as Leaf Area Index (LAI) and surface albedo, are replaced by spatiotemporally varying satellite observations. The coupled mode quantifies the net effect on simulated weather, while the offline mode isolates the underlying land-surface processes by removing atmospheric feedbacks.
Coupled simulations show that satellite-derived parameters improve mean and minimum temperatures, particularly at night, but introduce systematic daytime warm biases linked to excessive soil moisture depletion, a larger Bowen ratio, and reduced evaporative cooling. This bias structure reflects differences in transpiration regulation under water-limited conditions, where contrasting behavior over different land-cover types governs energy partitioning during heat and drought episodes.
Coupled and offline configurations are extensively evaluated against ICOS eddy covariance observations, LSA-SAF satellite-derived surface flux estimates, and gridded temperature products. By disentangling land-surface biases from land–atmosphere feedbacks, this combined framework links bias diagnosis to process-based understanding, informing land-surface parameterization development for improved simulation of compound extremes.

Mon

14:45–15:00

Soil water content is a key variable controlling hydrological processes, ecosystem functioning, and agricultural water management. Global estimates of water content are commonly obtained either from satellite-based retrievals, such as SMAP, or from model-based reanalysis products, such as ERA5. However, when these products are compared with field- and point-scale measurements, systematic differences emerge in both the magnitude and the dynamic range of soil water content. We hypothesize that these differences are related to an over-simplified representation of the relationship between soil texture, water phase arrangement, and dielectric properties that does not capture the effect of soil structures changing with seasons and extreme events.
To circumvent the use of such relationships and retrieval models based on many covariates, two Long Short-Term Memory (LSTM) models were developed and applied across the United Kingdom. Both models only use horizontally polarized brightness temperature (TBh) from SMAP together with precipitation and evapotranspiration from ERA5-Land as inputs. One model (denoted here as Model A) relies only on these satellite and climatic variables, without any soil-related information. The other model (Model B) extends this approach by additionally incorporating hydrological limits defined by minimum and maximum soil water content, representing the residual and saturated states of the soil. The models were trained on three COSMOS-UK sites and evaluated across 31 independent locations with water content data measured by cosmic-ray method.
Results show that Model A performs better than SMAP and ERA5 in capturing soil moisture dynamics, as indicated by higher temporal correlation with observations, reduced systematic bias, and lower root mean square error (RMSE) and unbiased RMSE (ubRMSE) compared to the global products. This improvement indicates that the satellite signal contains sufficient information to describe soil moisture dynamics when interpreted without the simplified assumptions used in current retrieval and reanalysis approaches.
This also implies that soil processes such as soil structure dynamics, pore connectivity, and hysteresis, which are not explicitly represented in SMAP and ERA5, are implicitly captured by the data-driven model. At the same time, the comparison with Model B shows that brightness temperature and climatic forcing define only the relative position between wet and dry conditions, but do not match the absolute magnitude and range of soil water content. Including hydrological limits in Model B significantly improves the representation of the dynamic range.
These findings show that differences between global products and field-scale observations are not only a consequence of scale mismatch but also reflect limitations in the simplified representation of soil processes. While satellite observations contain sufficient information to describe relative soil moisture dynamics, their ability to define the absolute magnitude of soil water content remains limited. Incorporating basic soil information, such as hydrological limits, provides a direct way to link relative wetness to absolute values and improves field-scale prediction.

Mon & Tue

11:00–12:30

Plant hydraulic traits govern water transport along the soil–plant–atmosphere continuum and regulate the coupling between soil moisture availability, stomatal control, and ecosystem water and carbon fluxes. Incorporating mechanistic plant hydraulics into land surface models such as the Community Land Model (CLM) improves the simulation of vegetation fluxes, particularly under drying soil conditions (Kennedy et al., 2019). However, these formulations introduce additional parameters that are difficult to constrain and can strongly influence model behavior. Global perturbation experiments with CLM have identified plant hydraulic parameters as major controls on evapotranspiration, although their importance varies regionally (Kennedy et al., 2025). Yet, systematic analyses of how parameter sensitivities differ among plant functional types (PFTs) and across seasons remain limited.

Here, we assessed the sensitivity of simulated vegetation water potential and water and carbon fluxes to five key plant hydraulic parameters using eCLM. The parameters represent stomatal regulation (medlyn slope), plant and root hydraulic conductance (kmax and krmax), cavitation resistance (psi50), and root distribution (β). Ensemble simulations were conducted for 13 ICOS sites across four European climate zones and five PFTs for 2009–2018. Parameters were varied within PFT-specific ranges following previous studies (Eloundou et al., 2024; Kennedy et al., 2025), resulting in 336 ensemble members. Variance-based sensitivities (main, interaction, and total effects) were quantified using the GEM-SA framework based on Gaussian process emulation (O’Hagan, 2006), trained on monthly means for each site and output variable.

Across sites, medlyn slope and kmax were the dominant controls on water and carbon fluxes (ET, Tr, GPP, NEE), with main effects explaining more than 60% of the variance, while interaction effects were small. Sensitivities varied substantially among PFTs, with distinct patterns for Mediterranean evergreen broadleaf, temperate deciduous, and evergreen needleleaf forests. Clear seasonal differences also emerged, as parameters such as psi50 became more influential during dry summer periods. At drought-prone Mediterranean sites, we observed seasonal shifts in parameter effects on canopy transpiration: higher Medlyn slope increased transpiration in spring but reduced it in summer, consistent with earlier stomatal limitation under hydraulic stress. Similar shifts were found when comparing dry and wet years.

Overall, model sensitivity to plant hydraulic parameters varies across PFTs, seasons, and hydroclimatic conditions. These results underline the need for improved parameter constraints and continued refinement of plant hydraulic and stomatal representations to ensure robust model performance, particularly during drought.

References

• Kennedy et al. (2019). 10.1029/2018MS001500
• Kennedy et al. (2025). 10.1029/2024MS004715
• Eloundou et al. (2024). 10.5194/egusphere-egu24-16086
• O’Hagan (2006). 10.1016/j.ress.2005.11.025

Tue

15:30–15:45

Soil moisture (SM) is an essential climate variable that regulates energy and water exchanges through land-atmosphere interactions across multiple time scales. To better understand these processes, in-situ SM observations are conducted globally through comprehensive networks such as the International Soil Moisture Network (ISMN). However, in-situ SM observations obtained from dielectric sensors exhibit pronounced temperature sensitivity due to the thermal dependence of soil dielectric permittivity, leading to an unrealistic diurnal cycle characterized by spurious daytime SM peaks that contradict evapotranspiration-driven drying. This study introduces a Long Short-Term Memory (LSTM)-based model to correct temperature-induced measurement errors by reconstructing physically consistent sub-daily SM dynamics. The LSTM model is trained using 1,058 point-scale observations from the ISMN and meteorological reanalysis datasets (ERA5-Land and MERRA-2) as inputs. A reference subset of ISMN observations exhibiting a physically consistent negative diurnal correlation between SM and soil temperature (TS) serves as the training target. Input variables (SM, TS, and latent heat flux) are decomposed into high- and low-frequency components to isolate the diurnal-scale variability, and a stacked LSTM architecture trained with a 24-hour sliding window selectively corrects high-frequency SM fluctuations while preserving low-frequency hydrological variability. The corrected SM time series exhibit a systematic shift of the SM–TS diurnal relationship from predominantly positive to negative and restore physically consistent daytime drying, particularly under warm and dry conditions where these spurious biases are pronounced. Overall, the LSTM-based correction enhances the physical reliability of in-situ soil moisture observations and enables improved sub-daily analyses of soil moisture dynamics and land–atmosphere coupling processes.

Mon & Tue

11:00–12:30

Land surface models used in operational weather forecasting generally handle the water partitioning between land surface storages and fluxes but neglect the water made available to evaporation from inundated and irrigated areas. In this study, we show the importance of considering these processes, as simulated within the ECMWF Integrated Forecasting System. The Catchment-based Macro-scale Floodplain model CaMa-Flood is coupled with the ECMWF land surface model ecLand where the inundated areas are allowed to re-evaporate to the atmosphere. A novel irrigation scheme, based on monthly maps of global irrigation fraction, is used to specify an irrigation flux based on an optimal irrigation strategy. These two processes are evaluated in a set of medium-range forecasts over winter and summer months and show a large, localized impact on temperature, surface pressure, and wind prediction showing a marked positive impact on forecast skill in the area interested. The assumptions and simplifications necessary for considering these processes at the global scale will be illustrated and the steps planned to further increase the realism of dynamic water redistribution, and its land-atmosphere interaction will be discussed.

Thu

09:00–09:15

Freshwater availability is a critical resource for society, ecosystems, and food production, and is expected to be strongly impacted by climate change throughout the 21st century. Increasing temperatures and thus enhanced evaporative demand will alter the continental water cycle, resulting in major challenges for water management, particularly in regions with intensive agricultural activity. A better understanding of both natural and anthropic processes affecting the continental water cycle is therefore essential to support more effective and sustainable water management strategies. In this context, Land Surface Models (LSMs) are key tools for representing the natural hydrological cycle, and ongoing developments aim to better account for human influences. Some LSMs already include partial or full representations of human-induced processes, such as irrigation or reservoir management, enabling the analysis of combined climate and management scenarios.

This work assesses future irrigation water demand in one of the main agricultural regions of the Ebro River basin (northeastern Spain), using climate projections for the period 2006–2099 over the Pyrenees and Pre-Pyrenees regions, developed within the framework of the PIRAGUA project under RCP 4.5 and RCP 8.5 scenarios. Simulations are driven by these projections and the SASER hydrometeorological modelling chain which has SURFEX implemented as its LSM and makes use of its irrigation scheme to estimate future irrigation. For this, a realistic irrigation scenario is defined based on data from a regional survey conducted among farmers, ensuring that current agricultural practices are adequately represented. Next, the impact of the irrigation scenario on drainage and evapotranspiration is evaluated.

The results will enable a comparison between projected irrigation water demand throughout the 21st century and the expected availability of freshwater resources, providing insights into the impacts of climate change and anthropic water use on regional water sustainability. This work is carried out within the framework of the LIFE Pyrenees4clima project.

Wed & Thu

11:00–12:30

Not all droughts are the same. Flash droughts are a subset of drought that rapidly intensify and often yield limited mitigation due, in part, to the quick evolution of environmental conditions. While flash drought events typically evolve at the subseasonal-to-seasonal scale, flash drought onset, intensification, and development are impacted by multiple processes including local terrestrial-atmosphere interactions and feedbacks. This complex set of drivers poses significant challenges to predictability of flash drought events. This study utilizes in situ observations, reanalysis datasets, and numerical simulations to (1) examine the critical land-atmosphere (L-A) pathways and feedbacks that impact the development and evolution of flash droughts events, (2) quantify the impact of L-A interaction on flash drought across varying regional surface conditions, and (3) identify critical L-A variables that impact predictability of flash drought in numerical simulations and forecasts.

Wed & Thu

11:00–12:30

Spatial heterogeneity of land use impacts land-atmosphere feedback and therefore the spatial and temporal variability of latent and sensible heat fluxes within the atmospheric boundary layer. This is especially visible during clear sky days without notable advection.
In spring and summer 2025 at the GEWEX Land Atmosphere Feedback Observatory (GLAFO) site of the University of Hohenheim (Stuttgart, Germany) an extensive field campaign was performed by the research group Land Atmosphere Feedback Initiative (LAFI) funded by the German Research Foundation. During five intensive observation periods (IOPs) the GLAFO equipment, which includes two Eddy-Covariance stations, was extended by Lidar measurements of wind, humidity and temperature.
To study the three-dimensional pattern of the heat fluxes over a heterogeneous surface during the day we applied the Weather Research and Forecasting model (WRF). We used WRF in a nested configuration with resolutions of 1250 m, 250 m and 50 m, forced with ECMWF operational data for a clear sky case study on 11 August 2025. In the two inner domains, WRF was applied in Large-Eddy simulation (LES) mode with switched-off turbulence scheme. The simulated evolution of the planetary boundary layer and the influence of the land surface on its development was compared with the temporal and vertical evolution in data from the lidar systems and eddy-covariance stations.
In addition, we focused on the vertical representation of latent and sensible heat fluxes at the different model resolutions and their dependence on the underlying land surface. This will reveal the so-called blending height, namely the height at which the horizontal distributions of the fluxes are no longer dependent on the underlying surface. The derivation of this important variable paves the way to a more physical coupling of the land surface and the atmosphere in the model.

Wed & Thu

11:00–12:30

Increasing concentrations of atmospheric CO2 alter the Earth’s energy budget not only by interaction with radiation, but also through adjustments of the climate state. One set of such adjustments is mediated by the response of vegetation to an increase in CO2: Changes in vegetation cover and leaf area index (LAI) as well as CO2-induced plant stomatal closure alter surface fluxes of radiation, heat, and water. These changes in surface fluxes in turn affect atmospheric temperature, water vapor, and cloud cover, further affecting the Earth’s energy balance. Past studies have shown that stomatal closure leads to a positive radiative forcing at the top of atmosphere (TOA), largely caused by adjustments of clouds to reduced transpiration. However, the full set of adjustments including the role of LAI changes and the surface energy balance have not been considered in investigations of the Earth’s energy budget.

We aim to provide a detailed analysis of energy budget changes at the land surface and TOA by performing idealized simulations performed with the Max-Planck-Institute Earth System Model (MPI-ESM) and utilize existing simulations made with the Community Earth System Model (CESM). In order to isolate the response of vegetation from direct interactions with radiation in these simulations, an abrupt doubling of CO2 concentration is only seen by the land model, while its atmospheric counterpart continues to experience pre-industrial conditions. To further separate changes originating from an increase in LAI, an additional experiment is run with static LAI prescribed per vegetated area. In order to place results in the context of a larger ensemble, C4MIP data will be analyzed, where transient simulations with a similar coupling are available across a larger group of models.

Preliminary results show persistent decreases in near-surface relative humidity and low cloud cover over land, especially pronounced in the northern extra-tropics. As a result, incident shortwave radiation at the land surface increases by 0.85 W m-2 and 1.06 W m-2 in the global average in MPI-ESM and CESM respectively. Together with a decreased latent heat flux, this is compensated by a greater sensible heat flux and moderate temperature increase, causing more longwave emission. Accordingly, the outgoing radiation at the TOA shows a decrease in the shortwave component, but an increase in longwave radiation. The MPI-ESM simulation with prescribed LAI shows a much higher radiative forcing of 0.33 W m-2 compared to 0.13 W m-2 in the experiment with dynamically evolving LAI, suggesting that adjustments in LAI could compensate significant parts of the forcing through CO2-induced stomatal closure. However, it is unclear which fraction of the LAI response is due to CO2 or temperature and this signal is less robust compared to the persistent changes. Notably, LAI adjustments are implicitly included in current radiative forcing estimates from models and our results indicate, that they could play a significant role in its magnitude that should be explored in additional, dedicated experiments.
These findings show a persistent effect of stomatal closure on energy fluxes that is strongly influenced by atmospheric feedbacks, especially through clouds. They also indicate that changes in LAI affect these energy fluxes and could alter estimates of radiative forcing, highlighting the need for further study.

Mon & Tue

11:00–12:30

We will present the strategy and results of a combination of six scanning lidars to investigate the interplay between daytime surface fluxes, surface layer gradients, convective boundary layer dynamics and development, as well as the characteristics of the interfacial layer and the lower free troposphere. Our observations were made above the agricultural fields of University of Hohenheim [1], Stuttgart, Germany in spring and summer 2025 in the frame of the research unit Land Atmosphere Feedback Initiative (LAFI, https://lafi-dfg.de/).
For this, the automated Raman lidar ARTHUS (Atmospheric Temperature and Humidity Sounder, [2]) built in our institute in recent years, was extended with a scanner for atmospheric measurements in the surface layer just above the canopy. ARTHUS is an eye-safe rotational Raman lidar with five receiver channels. These scanning measurements were performed during intensive observation periods for 50 minutes of each hour while during the remaining 10 minutes of each hour as well as during non-IOP days vertical pointing measurements were made. The surface layer observations of ARTHUS were combined with data measured with two Doppler lidars making simultaneously cross-cutting low-level scans for horizontal wind profiling near the surface. Two more Doppler lidars were measuring vertical wind fluctuations and horizontal wind speed and direction. One of these two Doppler lidars was operated in constant vertical pointing mode while the other was operated in a six-beam scanning mode with an elevation angle of 45°. Our water vapor differential absorption lidar (WVDIAL) made vertical pointing observations of turbulent moisture fluctuations up to the free troposphere. While also the WVDIAL can scan in any direction, it was operated in constant vertical-pointing mode during LAFI.
[1] Späth, F., et al.: DOI: 10.5194/gi-12-25-2023
[2] Lange, D. et al.: DOI: 10.1029/2019GL085774

Wed & Thu

11:00–12:30

The Land surface Interactions with the Atmosphere over the Iberian Semi-arid Environment (LIAISE) field campaign brought together ground-based and airborne measurements along with satellite data, with the aim of improving our understanding of key natural and anthropogenic land processes and boundary layer feedbacks. The study area was a semi-arid region in the Ebro river basin to the North-East of Spain, with a sharp demarcation between a vast intensively irrigated region and a much drier rainfed zone to the East.

The ground-based observations were taken over seven sites within an approximately 30x30km area in the eastern part of the basin, covering the Catalan counties of Urgell and Pla d’Urgell. The land cover varied across these sites, with one of the sites at Mollerussa instrumented over three difference types of vegetation. Super sites were set up at two locations with a similar configuration, one at an irrigated site and one at a rainfed site, allowing a detailed comparison between the conditions for a dry site and one which was no water limitation. These super sites included 50m masts with atmospheric quantities and fluxes measured at multiple heights.

The LIAISE database offers a wealth of observations that can be used to evaluate the current generation of land surface models (LSMs), and further our understanding of semi-arid environments and the impacts irrigation processes. Numerous aspects of these LSMs can be studied, including the following:
• The consistency of model results when driven by atmospheric data at different heights
• The impact of including irrigation in the surface energy balance
• Sensitivity of the surface energy balance components due to the type of irrigation method
• The impact of diurnally varying leaf area index compared to a monthly climatology, or the evaluation of a prognostic phenology scheme
• The heterogeneity of the surface fluxes for numerous sites within a small area (typically just a few gridpoints for an atmospheric model) given the area mean atmospheric forcing versus local atmospheric data.

Results from an initial evaluation of two LSMs will be presented. These results will be used to formulate a model protocol that is designed for a future LSM comparison project to address the suitability of the current generation of LSMs to represent a number of the aspects listed above.

Wed & Thu

11:00–12:30

The turbulent fluxes of heat and water vapor are key land surface – atmosphere interaction process variables in local, regional, and global energy and water cycles. Scintillometry can be considered as the only experimental technique presently available for the operational determination of these fluxes at a horizontal scale of several kilometers. Such measurements are of interest to validate fluxes simulated by regional atmospheric models or derived from satellite images.
Scintillometers measure the fluctuations of electromagnetic radiation propagating horizontally over a path of typically a few hundred meters up to several kilometers in length. These signal intensity fluctuations can be attributed to fluctuations of the refractive index of the air, i.e., of temperature and humidity, caused by turbulent processes in the near-surface atmospheric boundary layer (ABL). Using similarity theory relationships, the turbulent heat fluxes can be derived from the intensity of these refractive index fluctuations. Combining an optical and a microwave scintillometer (OMS), the fluxes of both sensible and latent heat can be obtained.
An OMS system has been operated at the Land-Atmosphere Feedback Observatory site at the University of Hohenheim during the General Observations Period of the Land-Atmosphere Feedback Initiative cluster project over a period of eight months from mid-March to mid-November, 2025. The 1093 m long path covered different types of farmland, including rape, maize, wheat, oat and soybean. These measurements were intended to provide flux values representing a heterogeneous agricultural landscape.
We will describe the data handling and processing of the scintillometer data for this quasi-operational setup. Derived fluxes are compared with local eddy-covariance measurements above two distinct fields below the scintillometer path. Moreover, an attempt will be made to relate the OMS-based fluxes with flux estimates from inside the ABL based on synergetic lidar measurements.

Wed

09:00–09:15

In many regions of the world, seasonal root water uptake (RWU) is an important resilience mechanism that allows vegetation to avoid water stress. One example is the Amazon, where deep roots take up subsurface moisture when moisture from precipitation is scarce, leading to maintenance of transpiration during dry periods. Representation of seasonal deep RWU has not been prioritized in land surface models despite its relevance to the water cycle and large-scale hydroclimate. In this work, we illustrate the use of a dynamic RWU scheme (known as DynaRoot) in the Noah-MP land surface model to simulate deep RWU in South America and the Amazon. We analyze two offline regional simulations of Noah-MP focused on the South American continent, including runs with and without DynaRoot activated. Two main science questions are addressed in this work: Can we capture the appropriate seasonal RWU dynamics for the Amazon and other areas of South America? How does the addition of DynaRoot affect simulation of transpiration and biases in modeled latent heat flux? To answer these questions, we examine time series of modeled soil moisture, transpiration, and latent heat flux and compare them to available observational products. We also compare modeled rooting depth to estimates from previous studies. This study demonstrates the utility of the DynaRoot scheme for the land surface modeling community. Additionally, it highlights progress and areas for future improvement in modeling Amazonian and South American hydroclimate. This study also constitutes foundational work for planned fully coupled simulations of South America using Noah-MP and the Model for Prediction Across Scales (MPAS).

Mon

13:30–13:45

Vegetation is a key component in land-atmosphere coupling, through its dynamic regulation of the water and carbon exchange, yet this control depends critically on access to belowground water resources that cannot be directly observed at large scales. As vegetation greening continues under a changing and increasingly variable hydrological cycle, it remains unclear how ecosystem water use has been realized historically, and whether changes in vegetation structure reflect corresponding changes in water access. To this end, we develop a hybrid machine learning framework that integrates a parsimonious ecohydrological model with data-driven flexibility to infer long-term operated range of belowground water access as a latent, ecosystem-scale state constrained by observations. This framework provides a practical way to diagnose how ecosystems have historically operated within water and energy constraints to sustain evapotranspiration, without reliance on prescribed plant and soil hydraulic parameterizations. By relating this inferred water access to vegetation structure, we characterize the co-evolution of vegetation and water use as an emergent property of land – atmosphere coupling. Applied at continental scale across the U.S., the analysis shows that changes in vegetation structure and changes in ecosystem water access frequently differ in both magnitude and direction. These patterns indicate that vegetation greening has occurred under distinct ecosystem-scale water-use behaviors over time and across regions, suggesting interpretations of ecosystem resilience under hydroclimatic variability. By explicitly diagnosing ecosystem water access and its relation to vegetation change, the framework provides a process-informed, yet observationally driven basis for benchmarking vegetation–soil water coupling and for rethinking how land-atmosphere interactions are represented in land surface models under changing hydroclimate conditions.

Mon

13:45–14:00

Understanding the impact of human activity on the water cycle is essential for climate projections, and it is especially critical in semi-arid regions. For this reason, the Land surface Interactions with the Atmosphere over the Iberian Semi-arid Environment (LIAISE) measurement campaign (Boone et al., 2025) took place in July 2021 in the Ebro basin in North-East Spain, which is bounded to the north by the Pyrenees. Intensive agriculture is major in this area and entirely depends on irrigation. The surface heterogeneity artificially generated by irrigation in response to the intensive agricultural activities induces a very marked discontinuity of surface conditions, with extreme differences in surface temperature and soil moisture.

To understand the effects of this heterogeneity on heat and moisture transfers in the planetary boundary layer (PBL), the two areas (irrigated/non-irrigated) were observed over two different timescales. A long period from April to October, aimed at measuring the seasonal cycle of water supply needed for crops, and a 15-day intensive period in July, when the water requirements are very high and the contrast between the irrigated and rain-fed surfaces is highest. During the intense observation period, the two sites were similarly instrumented with surface energy budget (SEB) stations, ultra-high frequency (UHF) wind radars, high-frequency radiosoundings, wind lidars and a tethered balloon. In addition, fine scale measurements of turbulence and atmospheric variables within the PBL and above were possible with 8 flights of the French ATR-42 research airplane.

Unlike previous studies, this research simultaneously considers all of the available instruments during the 8 flight days, which allows a detailed characterization of the atmosphere above each of the contrasted surfaces. The results highlight the complex structure of the PBL above each site (which varies significantly from theoretical profiles), the value of instruments synergy to study mesoscale processes (such as the local Marinada sea breeze intrusion) and local circulations due to surface heterogeneity. Preliminary results also point to the mesoscale conditions which are favorable for the development of local circulations.

Overall, this study follows up on the significant research already conducted based on the LIAISE campaign, provides insight on the obtained results and further explores the complex surface-atmosphere interactions present in the region.



References:

Aaron Boone, Joaquim Bellvert, Martin Best, Jennifer K. Brooke, Guylaine Canut-Rocafort, Joan Cuxart, Oscar Hartogensis, Patrick Le Moigne, Josep Ramon Miró, Jan Polcher, Jeremy Price, Pere Quintana Seguí, Joan Bech, Yannick Bezombes, Oliver Branch, Jordi Cristóbal, Karin Dassas, Pascal Fanise, Fabien Gibert, Yves Goulas, Jannis Groh, Jan Hanus, Gabriel Hmimina, Lionel Jarlan, Ed Kim, Valérie Le Dantec, Michel Le Page, Fabienne Lohou, Marie Lothon, Mary Rose Mangan, Belén Martí, Daniel Martínez-Villagrasa, James McGregor, Amanda Kerr-Munslow, Nadia Ouaadi, Alban Philibert, Juan Quiros-Vargas, Uwe Rascher, Bastian Siegmann, Mireia Udina, Antoine Vial, Burkhard Wrenger, Volker Wulfmeyer, Mehrez Zribi, The Land Surface Interactions with the Atmosphere over the Iberian Semi-Arid Environment (LIAISE) field campaign, Journal of the European Meteorological Society, Volume 2, 2025, 100007, ISSN 2950-6301, https://doi.org/10.1016/j.jemets.2025.100007. (https://www.sciencedirect.com/science/article/pii/S2950630125000018)

Mon & Tue

11:00–12:30

Reducing Parametric Uncertainty in ICON-XPP-MLe Land-Atmosphere Simulations Using History Matching

Max Bouman1, Mierk Schwabe1, Katie Dagon3, Linnia Hawkins4, Sönke Zaehle5, and Veronika Eyring1,2

1Deutsches Zentrum für Luft- und Raumfahrt (DLR), Institut für Physik der Atmosphäre, Oberpfaffenhofen, Germany.
2University of Bremen, Institute of Environmental Physics (IUP), Bremen, Germany.
3CGD, NSF National Center for Atmospheric Research, Boulder, CO, USA
4Department of Earth and Environmental Engineering, Columbia University, New York, NY, USA
5Max Planck Institute (MPI) for Biogeochemistry, Jena, Germany.

The complexity of land models in Earth system models (ESMs) has increased significantly over recent decades, evolving from simple boundary condition providers to fully interactive components. This increasing complexity introduces structural uncertainty, including parametric uncertainty that stems from poorly constrained model parameters. We extend the emulator-based history-matching approach for the ICON atmosphere component (Bonnet et al., 2025) to the ICON land component (ICON-Land) to diagnose and constrain land-climate feedbacks and uncertainties. In particular, we reduce parametric uncertainty in the JSBACH configuration of ICON-Land, coupled to the ICON model in its XPP configuration, for a coupled coarse spatial-resolution (R2B4) land-atmosphere (AMIP) simulation. Following an initial tuning of the atmospheric component, using the computationally inexpensive Nelder–Mead optimization method (Grundner et al., 2025), we identify influential land parameters via One-At-A-Time sensitivity experiments. These parameters are then constrained via history matching, yielding a Not-Ruled-Out-Yet (NROY) parameter space consistent with observational constraints. From this NROY space, we select plausible parameter sets to conduct high-temporal-frequency climate simulations for assessing land–atmosphere coupling using dedicated coupling diagnostics, such as soil moisture memory, coupling strength, and mixing diagrams. The evaluated coupling behavior is compared against flux tower observations. This study forms an important contribution to better understand, diagnose, and constrain land-atmosphere interactions and uncertainties.

References:

Bonnet, P., Pastori, L., Schwabe, M., Giorgetta, M., Iglesias-Suarez, F., and Eyring, V.: Tuning the ICON-A 2.6.4 climate model with machine-learning-based emulators and history matching, Geosci. Model Dev., 18, 3681–3706, https://doi.org/10.5194/gmd-18-3681-2025, 2025.
Grundner, A., Beucler, T., Savre, J., Lauer, A., Schlund, M., Eyring, V., 2025. Reduced cloud cover
errors in a hybrid AI-climate model through equation discovery and automatic tuning. Scientific
Reports 15, 43836. URL: https://www.nature.com/articles/s41598-025-29155-3,
doi:10.1038/s41598-025-29155-3. publisher: Nature Publishing Group.

Mon & Tue

11:00–12:30

Improving the vegetation representation within the ECMWF ecLand and the Integrated forecasting System

S. Boussetta, C. Rüdiger, G.Arduini, A. Agusti-Panareda, J. Mc Norton, G. Balsamo, S. Garrigues

Getting an accurate estimate of vegetation state is crucial for land-atmosphere coupling through the proper partitioning of energy fluxes, as well as for a correct estimate of the water and carbon cycle components. In this study, we attempt to improve the realism of the phenology representation within the ECMWF land surface model ecLand through the implementation of three modifications of the surface parametrisation: (i) a better consideration of the environmental stress by adding an acclimation function to the Jarvis canopy resistance formulation, (ii) introducing a seasonal varying surface roughness length based on the leaf area index (LAI) variability, (iii) and implementing a prognostic LAI that allows the vegetation to develop based on the environmental forcing. These modifications were tested in standalone and coupled forecast modes, demonstrating potential to improve the vegetation seasonality and the near-surface atmosphere. Details of these implementations will be introduced and discussed in the context of land-atmosphere interaction.
(The conceptual illustration was generated with the assistance of an AI language model )

Wed & Thu

11:00–12:30

Closing the coupled water-energy-carbon budget in land surface models remains one of the central unsolved problems in land-atmosphere science. Parametric uncertainty in stomatal regulation, soil hydraulics, photosynthetic capacity, and soil respiration collectively propagate into biases in surface fluxes, boundary layer dynamics, and the terrestrial carbon sink. Yet most model calibration efforts treat these cycles in isolation.

Here we present ClimaLand (https://github.com/CliMA/ClimaLand.jl), a GPU-native land surface model that simultaneously calibrates water, energy, and carbon fluxes through ensemble Kalman inversion (EKI). Carbon assimilation is represented via an optimality-based photosynthesis framework (P-model), with maximum rate of Rubisco carboxylation and stomatal parameters globally calibrated against ERA5 reanalysis and flux tower observations rather than prescribed from plant functional type look-up tables. Soil respiration follows the DAMM (Dual Arrhenius Michaelis-Menten) formulation driven by SoilGrids soil texture and organic carbon, enabling mechanistic temperature and moisture sensitivity. Together, these components achieve net ecosystem exchange (NEE) closure across biomes.

We evaluate the fully coupled system within the ILAMB benchmarking framework, comparing GPP, NEE, latent heat, and sensible heat against observational benchmarks, CMIP6 and TRENDY model ensembles. Joint calibration of energy, carbon and water cycle parameters via EKI yields improved skill relative to sequentially calibrated configurations, demonstrating that energy-water-carbon coupling is not merely a biophysical detail but a first-order control on model fidelity.

ClimaLand’s architecture — differentiable, GPU-accelerated, and natively coupled to the CliMA Earth System Model — positions it as a platform for the community to revisit fundamental assumptions about land-atmosphere feedbacks at kilometer scale. We discuss implications for evapotranspiration partitioning, carbon-water tradeoffs under drought, and the design of next-generation benchmarking protocols.

Tue

16:15–16:30

Improving the prediction of weather and climate from nowcasting to sub-seasonal to seasonal timescales remains a central challenge for the Earth system science community. Despite major advances in numerical weather prediction and climate modelling, forecast skill is still limited by an incomplete understanding and inadequate parameterization of Land-Atmosphere (L-A) feedbacks. These feedbacks, which link soil, land cover such as vegetation, and the atmospheric boundary layer through exchanges of energy, water, matter, and momentum are fundamental to the scientific evaluation and forecasting of extreme events, regional climate variability, and long-term climate change projections.
This limitation has significant consequences. Reliable representation of L-A processes is essential not only for improving forecasts of heatwaves, droughts, and precipitation, but also for increasing confidence in climate mitigation and adaptation strategies, including land-use change and bio-geoengineering pathways. Addressing this challenge requires a step change in both process understanding and the observational basis underpinning model development.
A key barrier to progress is the lack of comprehensive, high-quality observations spanning the full L-A system. Existing observational networks often do not resolve the coupled processes across all relevant compartments, from groundwater and soil moisture, through vegetation dynamics, to the lower troposphere, nor do they provide the temporal and spatial resolution required to constrain model parameterizations.
The Gewex Land Atmosphere Feedback Observatory (GLAFO) initiative responds to this need through a coordinated, international effort to design and implement a new generation of integrated observatories.

The GLAFO Project Office supports this effort by fostering collaboration across disciplines, facilitating the development of a global network of sites, and enabling the delivery of sustained, high-resolution observations of L-A interactions using a harmonized data format based on the CF and obs4MIPs metadata conventions.
Through this coordinated framework, GLAFO aims to advance process understanding, develop observation-based parameterizations, and provide the benchmark datasets required to improve weather, climate, and Earth system models – ultimately strengthening our ability to predict and respond to climate variability and change.
GLAFOs permits varying tiers of instrumentation to allow flexibility while maintaining the necessary core measurements for observing L-A feedbacks at all sites. The base tier comprises surface energy balance and boundary layer height measurements to capture the coupling between surface energy fluxes and tropospheric entrainment. Higher tiers contain these, and further instrumentation to capture a full picture of the land-atmosphere system, including Doppler lidars and temperature/water vapor profilers, radars and many plant and soil measurements.
GLAFO aims to develop a coordinated global network of observatories across diverse climate regimes. The Land Atmosphere Feedback Observatory (LAFO) at the University of Hohenheim, Stuttgart was the first GLAFO and represents a benchmark for others to follow. Other GLAFOs include the ARM Southern Great Plains (SGP) experimental site in the USA, Cabauw in the Netherlands, Huancayo in Peru, and Lindenberg in Germany. Increasing the size of this network will provide harmonized observations of land–atmosphere feedbacks to advance process understanding and improve model performance.

Wed & Thu

11:00–12:30

The Penman-Monteith equation is probably the most widely used method for determining latent heat fluxes from vegetated land surfaces and is therefore considered the gold standard for calculating evapotranspiration. The original form of the equation was derived by Penman (1948) for the evaporation of open water surfaces or saturated soils. It is based on a simple energy balance approach, for which the Bowen ratio is estimated based on certain assumptions. In 1965, Monteith extended Penman’s equation to make it applicable to vegetated surfaces. To this end, he developed a thermodynamic derivation of the Penman equation, which has since become known as the Penman- Monteith equation.
This study revisits the thermodynamic derivation of the Penman-Monteith equation and presents arguments suggesting an inconsistency in this derivation. Due to this inconsistency, strictly speaking, the derivation is valid only for non-vegetated land surfaces, but not for land surfaces with a vegetation cover. The chain of arguments leading to this conclusion, as well as a thermodynamically consistent derivation for vegetated land surfaces, are presented in detail.

Wed & Thu

11:00–12:30

Surface-atmosphere interactions play a critical role in the development of the boundary
layer and strongly control the diurnal evolution of near-surface temperature and humidity,
important metrics in forecast verification. In semi-arid/arid regions water availability in the landscape can be altered by human management processes including irrigation. However, irrigation processes are typically not represented in general circulation models (GCMs) which can give rise to systematic warm/dry biases.

A simple irrigation scheme has been implemented into the JULES land surface model, coupled to the Met Office Unified Model (UM). Convection-permitting regional simulations of the UM have been run for a study region in north-east Spain, incorporating the Ebro Basin, which has a significant irrigation presence. Simulations are evaluated using observations from the international LIAISE (Land surface Interactions with the Atmosphere over the Iberian Semi-arid Environment) project.

Results show that representing irrigation results in not only improvements to the surface heat and moisture fluxes, but also to the boundary layer depth, influencing the evolution of the near surface atmosphere. The impact of the irrigation is not only seen over the irrigation area, but can also be detected downwind, influencing the boundary layer structure over a wider area than that irrigated.

Analyses of the boundary layer evolution using mixing diagrams identify potential limitations in the mixing diagram hypotheses which not only need to be carefully conceded, but also highlight limitations in our understanding of the near surface in unstable conditions.

Thu

09:15–09:30

Modeling and understanding land-atmospheric (L-A) feedback dynamics is a challenging research area, which is tackled by the Land Atmospheric Feedback Initiative (LAFI). Recent advances in machine learning (ML) and deep learning can support this endeavor. Our goal is to uncover quantitative relationships and causal interactions with the help of ML techniques. Along these lines, we are addressing three challenges.
First, we are aiming at downscaling coarse-resolution data—mainly from geostationary satellites—into higher spatial and temporal resolution estimates to capture local details. Currently, we focus on determining fine resolutions of land-surface temperatures as well as canopy moisture, with promising first results. Second, we study surface fluxes as well as flux gradients in complex terrains. Here, the goal is to surpass Monin-Obukhov similarity theory (MOST) as well as Bulk Richardson Number (BRN)-based approaches, developing equally compact but contextualized similarity equations. Third, we study the energy balance closure, focusing on the prediction of energy balance residuals under different land surface heterogeneity and atmospheric circumstances.
Besides improving resolution and accuracy, we are aiming at identifying key driving factors under different contextual circumstances. These drivers are expected to include terrain properties, atmospheric factors, as well as the recent past of the unfolding L-A feedback dynamics. Explainable machine learning approaches can be used to identify these factors using Shapley-value attribution methods, permutation methods, or layer-wise relevance propagation. Additionally, spatiotemporal dynamic influences will be analyzed by integrating Long-Short-Term-Memory (LSTM) recurrent neural network modules locally in space as well as by feeding in history-compressed data into the ML model. We expect that the identified key driving factors, selective history compression factors, and distinct spatiotemporal dynamics will enable the LAFI consortium to improve current L-A models – either with individual small ML components or with compact equations that will be derived from the key driving factors identified by the explainable ML techniques.
The overall goal is to derive compact representation of core L-A feedback processes to foster our understanding of the underlying core land-atmospheric processes, and to inform models of larger scale weather- and climate dynamics.

Mon

13:30–13:45

The evolution of the stable atmospheric boundary layer depends strongly on surface temperature fluxes and their temporal variability. Yet many idealized numerical studies and parameterizations still assume constant surface cooling with time. We examine how time-dependent surface heat fluxes influence turbulence, stratification, and boundary-layer development using large-eddy simulations in an idealized framework based on the GABLS1 case.

We apply time-varying and constant surface heat fluxes that include a controlled temporal gradient, representing more realistic evening transition conditions over land. The resulting boundary layer structure is analyzed in terms of temperature stratification, wind profiles, turbulent kinetic energy, and boundary layer depth, with particular focus on how the temporal evolution of surface forcing shapes turbulence decay and intermittency.

The results show that the temporal evolution of the surface heat flux is a critical control on stable boundary layer dynamics. Rapidly increasing cooling produces a shallower, more strongly stratified boundary layer with suppressed turbulent mixing and earlier turbulence collapse. More gradual cooling allows turbulence to persist longer, leading to deeper mixing layers and smoother vertical gradients. The cooling history also influences the timing and intensity of intermittent turbulent events, demonstrating that boundary-layer characteristics cannot be explained solely by instantaneous flux magnitude.

These findings directly inform the observation and parameterization of turbulence and surface fluxes at the land atmosphere interface. They highlight that a realistic representation of temporally varying surface forcing is essential for improving process understanding and benchmarking boundary-layer models. The results further advance the modeling of land-atmosphere feedbacks by clarifying how surface energy exchange influences nocturnal boundary-layer structure and evolution.

Overall, this work emphasizes that incorporating temporal variability in surface heat fluxes improves the physical realism of idealized simulations and provides guidance for developing more robust parameterizations of stable boundary layers in weather and climate models, thereby strengthening the representation of land atmosphere coupling across a wide range of conditions.

Mon & Tue

11:00–12:30

The feedback between the surface and the atmosphere is a key component in climate modeling, however, its representation in models still contains uncertainties. It is necessary to better understand the surface-atmosphere interactions and improve the representation of certain processes, as the impact of sub-mesh surface heterogeneities for example. High-resolution models (~100 m), such as those used for Large Eddy Simulation (LES), are commonly used for this purpose, particularly when compared with climate model simulations in single-column configuration (1D).
The use of a surface model common to both types of atmospheric model enables the land-atmosphere coupling to be maintained, while ensuring that the differences observed in LES/1D comparisons are due to the representation of the atmosphere. Complex surface models such as SURFEX (Masson et al., GMD, 2013) or ORCHIDEE (Krinner et al., Global Biogeochem. Cycles, 2005), used in the operational versions of the CNRM and IPSL climate models (Voldoire et al., JAMES, 2019; Boucher at al., JAMES, 2020), represent many processes, but are difficult to implement in other atmospheric models. The use of a simpler surface model, easily implemented in a hierarchy of models and sufficient for LES/1D case studies lasting a few days, can then be considered (Laguë et al., J. Clim., 2019).
This question is addressed here through the development of a simplified continental surface model based on the one used in some configurations of the general circulation model LMDZ (Hourdin, 1992), and its implementation in the surface modeling platform SURFEX used in the large-eddy model Meso-NH (Lac et al., GMD, 2018). This simple surface model is based on heat diffusion in the soil and a prescribed evapotranspiration coefficient. The surface is characterized by six parameters: roughness lengths for momentum and heat, albedo, emissivity, evapotranspiration coefficient and soil thermal inertia. The ability of the simplified surface model to accurately represent the land-atmosphere coupling and its impact on the atmospheric boundary layer, when implemented in LMDZ 1D (Hourdin et al., JAMES, 2020) and Meso-NH, is assessed using a model setup based on the MOSAI (Model and Observation for Surface-Atmosphere Interactions) field campaign (Lohou et al., JEMS, under review) using coupled simulations.
In order to quantify the impact of sub-mesh surface heterogeneities on fluxes and atmospheric variables, a more detailed representation of heterogeneous surfaces based on flux aggregation (Claussen, J. Hydrol., 1995) is also implemented in the simplified surface model. LMDZ 1D simulations are carried out to compare this new representation with the one based on parameter aggregation, and preliminary comparisons with LES simulations are discussed.

Tue

15:45–16:00

Satellite-derived land surface temperature (LST) is widely used to evaluate land surface models (LSMs). However, retrieval processes and the nonlinear nature of thermal emission often obscure the physical drivers of discrepancies between model and observation. This study seeks to explore whether a radiance-based diagnostic framework can provide a more physically transparent assessment of LSM performance.
To investigate this, we develop a forward-simulation pipeline that couples the HydroBlocks LSM with the Community Radiative Transfer Model (CRTM). Our goal is to quantify how subgrid-scale land surface heterogeneity and uncertainties in hydrologic parameters influence top-of-atmosphere (TOA) radiance signals, particularly for GOES-16 Band-13 brightness temperatures (BTs), at the ARM Bankhead National Forest site. Using a 100-member Latin Hypercube ensemble of key parameters, we attempt to define the bounds of a physically plausible uncertainty envelope in radiance space. A key question is how the nonlinear Planck response to surface temperature affects the portrayal of subgrid variability. We hypothesize that the spatial aggregation of heterogeneous patches could lead to scale-dependent biases in BT that are not captured by linear evaluation methods. Additionally, we investigate whether this radiance sensitivity varies with atmospheric conditions such as humidity and cloud cover. The framework aims to determine if we can separate discrepancies caused by parameter choices from those due to fundamental model structural limitations. By incorporating HRRR-derived boundary layer heights and cloud masks, we further examine how surface flux heterogeneity influences land–atmosphere interactions and cloud development. Ultimately, this research aims to establish a mechanism-based approach for model evaluation, offering a scale-aware method to identify process-level drivers of bias and guide future improvements in parameterization.

Mon & Tue

11:00–12:30

As the transition zone in both climate and ecosystem, semi-arid regions are more sensitive to human disturbances and climate change. Semi-arid regions have experienced significant vegetation changes, but the impact of such greening on compound drought-heatwaves (CDHW) remains unexplored. Both East Asia and North America’s semi-arid regions lie at the northern boundary of the summer monsoon and are significantly influenced by interannual and decadal variations in monsoon climates. Investigating the biophysical effect of greening on CDHW in these two regions provide valuable insights for understanding the dynamics of other semi-arid regions. We find that greening can cause cooling effects on the mean surface air temperature (SAT) and more pronounced cooling effects during heat extremes. An attribution analysis suggests that the main driving factor for the additional cooling effect of greening in hot extremes is the enhanced evapotranspiration. Moreover, based on multi-source remote sensing observations, we focus on physiological processes such as vegetation water transport during CDHW to systematically reveal the feedback effects of vegetation on the development of these events. Our results contribute to the understanding of land–atmosphere feedback mechanisms in semi-arid regions on East Asia and North America, which may have implications for climate mitigation in transition zones around world.

Mon & Tue

11:00–12:30

Predicting and understanding the critical soil moisture and water potential at which plants start to downregulate transpiration and photosynthesis is crucial for representing water stress in earth system models, for evaluating drought impact on ecosystems and for disentangling the relative role of vapor pressure deficit and soil drying on transpiration limitation. The underlying mechanisms explaining threshold soil moistures and the controlling ecosystem properties remain elusive.
A soil-plant hydraulic model in which stomatal closure is triggered by a loss in the hydraulic conductance of the soil-plant continuum explained well observed soil moisture thresholds. Observations and our model showed that soil moisture thresholds, as well as their variability, are in large part controlled by soil hydraulic properties. The threshold soil matric potential is not unique (it varies over two orders of magnitude), and it is inversely related to the slope of the soil hydraulic conductivity.
These results demonstrate the importance role of soil hydraulic properties on ecosystem water limitations and call for a better estimation of these properties for their implementation in models at larger scale.

Tue

09:00–09:15

The Earth’s landscape has been profoundly shaped by humans through the exploitation of natural resources, such as dams and irrigation practices. These changes highlight the expanding human footprint on freshwater resources and ecosystem services, raising concerns about the rapid rate of alteration occurring across the Earth. Furthermore, these human-induced impacts significantly alter the water and energy cycles. Therefore, improving representation of these anthropogenic influences is essential for enhancing simulations of land–atmosphere interactions.
In this study, we investigate the combined effect of reservoir regulations and irrigation within the SAFRAN-SURFEX-EauDyssée-RAPID (SASER) modelling chain on the irrigated area of the Canal de Aragón y Cataluña (CAyC), which includes the Barasona and Santa Ana reservoirs, located in the north-eastern part of the Ebro basin in Spain. A reservoir parameterization is implemented as an external module, while irrigation is explicitly represented using the irrigation scheme available in the SURFEX v.9 land surface model. Both activities are integrated within the same modelling framework to understand drought dynamics in the region.
The performance of the reservoir module was evaluated through the Kling-Gupta Efficiency (KGE) metric, and a natural scenario (without any human influence), provided by SASER, is used as a reference to quantify the changes. The results overall indicate a good agreement, particularly for the Barasona reservoir, with KGE values for outflows around 0.75, while Santa Ana exhibited lower values.
Overall, the simulation results indicate that irrigation leads to a notable change in evapotranspiration dynamics and shifts in the river flow regime. Under natural conditions, the system responds directly to drought events, propagating across both variables, river flows and evapotranspiration (ET). In contrast, under human influences, ET deficits are greatly reduced while reservoirs exacerbate the effects of hydrological droughts.
These findings highlight the importance of including human influences toward improving the realism of regional modelling systems and provide new insights into drought assessment and management strategies.

Mon & Tue

11:00–12:30

Humid heatwaves occur when warmer and more humid air conditions combine through large-scale atmospheric dynamics and local/mesoscale processes, involving land-atmosphere interactions. If the roles of land-atmosphere interactions have been largely studied from the perspective of dry heatwaves, much less is known about their effects on humid heatwaves.

Numerical weather simulations of various complexity run at convection-resolving scales (Δx=500m–4km) are used to study the effects of surface heterogeneity on humid heatwaves. A 10-year, pan-African climate simulation run at 4 km grid spacing allowed to demonstrate that humid heat extremes are amplified by ~0.5 C when they occur over locally wetter soils. Soil moisture-induced mesoscale circulations suppress boundary layer growth more efficiently than over larger-scale wet soils, which creates most extreme conditions. A follow-on modelling study using an idealized setup wherein wet patches with diameter λ ∈ 25-150 km are prescribed in an otherwise dry domain shows that a critical soil moisture length scale λc exists, for which the effect on humid heat is maximized. In addition, the background wind and the strength of the soil moisture contrast exert a first-order control on the relationship between humid heat amplification and λc.

Novel understanding about the effects of mesoscale surface heterogeneity on humid heat challenges understanding based on coarser-resolution models. Furthermore, these results could help to predict extreme humid heat at city and county scales in regions of strong land-atmosphere coupling based on land surface information from satellites.

Mon

14:45–15:00

Traditionally, land-atmosphere interactions are evaluated using single-pixel time-series or static spatial snapshots. These methods often fail to capture the critical space-time interactions of surface states that can drive processes such as runoff generation and mesoscale circulations. As models move toward higher resolutions and increased sub-grid complexity, traditional summary statistics become increasingly insufficient for diagnosing model performance.

This study proposes a shift toward evaluating models as “movies” rather than static images, leveraging empirical and parametric space-time covariance to summarize model behavior. We apply this framework to an ensemble of GFDL LM4.2 simulations to analyze how spatial complexity—both at the sub-grid and across-grid levels—influences Land Surface Temperature (LST) patterns. Initial results indicate significant discrepancies between simulated and observed space-time statistics, even in cases where traditional evaluation approaches suggest visual or statistical similarity. These findings establish a benchmarking protocol for future model intercomparisons, aimed at diagnosing and correcting the structural spatial-temporal weaknesses in contemporary land surface models.

Wed & Thu

11:00–12:30

The diurnal cycle of precipitation reflects complex interactions among solar radiative heating, boundary layer processes, and convective processes, and its accurate representation is critical for evaluating atmospheric physical processes at sub-daily timescales as well as for land–atmosphere interactions, and hydrological modeling. Previous studies have identified systematic phase shift of diurnal cycle of precipitation in satellite and reanalysis datasets, yet comprehensive evaluations utilizing extensive ground-based observations to validate multiple datasets simultaneously and to elucidate bias mechanisms of phase shift remain limited.
This study evaluates the boreal summer (June–August) diurnal cycle of precipitation during 2000–2023 across five satellite products (IMERG, TRMM 3B42, CMORPH CDR, GSMaP, and MSWEP) and four reanalysis datasets (ERA5, JRA-3Q, MERRA-2, and NARR) against over 9,000 ground-based observation stations globally. Harmonic analysis and circular statistics are applied to quantify the spatial patterns of diurnal phase and amplitude. Beyond conventional phase and amplitude diagnostics, sub-daily variance analysis is employed to quantify the contribution of sub-daily variability to total precipitation variance, and harmonic power spectrum analysis is used to evaluate the balance among the first, second, and higher-order harmonic components across datasets.
Satellite products successfully reproduce the observed late-afternoon precipitation peak over land, with systematic delays of 1–2 hours attributed to passive microwave sensors detecting hydrometeors at the ice-scattering level. Among satellite products, CMORPH CDR demonstrates the best overall performance in representing both phase and amplitude. Reanalysis datasets systematically underestimate both sub-daily variance and the first harmonic amplitude while overestimating the second harmonic component, and exhibit peaks 3–6 hours earlier than observation over land. The early peaks in reanalysis products are primarily driven by convective precipitation, which peaks several hours earlier than observed, while large-scale precipitation independently peaks during morning hours. Among reanalysis datasets, JRA-3Q shows better performance than other reanalysis datasets, as its DCAPE-based convective trigger depends not only on the magnitude of CAPE but also on its generation rate, preventing premature convective initiation during the CAPE buildup phase, while NARR demonstrates superior representation over North America through precipitation data assimilation.
These results highlight that reanalysis products systematically underrepresent sub-daily precipitation variability and distort the harmonic structure of the diurnal cycle, with the fidelity governed by both the convective triggering mechanism — particularly the representation of CAPE generation rate — and the data assimilation strategy.

Mon

16:45–17:00

Taylor’s hypothesis is the backbone to convert observations done over time to spatial information of the flow while carrying out turbulence measurements on a micrometeorological tower. To address its validity over a highly heterogeneous forest clearcut surface, we utilize an extensive Distributed Temperature Sensing (DTS) and Eddy Covariance (EC) datasets. The DTS measured space-time correlation curves of temperature fluctuations are used to compute the bulk convective speeds of temperature structures in buoyant conditions at a height of 3.1 m above the clearing. These convective speeds are compared with the mean wind speed obtained from the EC system at the middle of the clearcut. In highly turbulent conditions, the convective speeds of the temperature structures are found to be larger than the mean wind speed, which is explained by the random sweeping hypothesis. The presence of sweeping effects violates the frozen turbulence assumption in Taylor’s hypothesis, and they are found to be strongly correlated with the turbulence kinetic energy of the flow. By incorporating these sweeping effects in an elliptical model of space-time correlation curves, one could essentially obtain spatial information from single-point measurements done over time. These results have applications for flux footprint modelling over heterogeneous surface conditions.

Mon

16:30–16:45

Root-zone soil moisture (RZSM) plays a central role in regulating evapotranspiration and plant water availability, thereby influencing land-atmosphere interactions. Satellite missions provide surface soil moisture (SSM), but they do not directly observe deeper layers, and in-situ measurements at root-zone depth remain too sparse for broad spatial coverage. Approaches that relate available surface and environmental information to subsurface moisture are therefore needed to extend monitoring capabilities.

We present a data-driven approach to estimate daily RZSM in Mediterranean vineyards in the Terra Alta county (Catalonia, northeastern Spain), using in-situ observations from eight stations collected between 2020 and 2024. Soil moisture at 25 cm depth is predicted from 5 cm soil moisture, soil moisture change rates at different lags, the Soil Moisture Deficit Index (SMDI), daily precipitation, minimum, mean and maximum temperature, a cyclic encoding of day of year, and static soil descriptors. A multilayer perceptron (MLP) is trained and evaluated using year-block cross-validation to assess temporal generalisation, and leave-station-out experiments to test transferability across sites with contrasting soil properties.

The model achieves strong skill under independent temporal evaluation (median non-parametric Kling-Gupta efficiency around 0.9), while performance across unseen vineyards is more heterogeneous, indicating that further work is needed to improve spatial transferability. Permutation-based feature importance is used to assess the relative contribution of the different variables and support interpretation of the model behaviour.

To extend the approach beyond instrumented sites, future work will explore configurations driven by or trained with satellite-derived surface soil moisture. In parallel, comparisons with alternative methodologies will allow systematic comparison of data-driven and model-based soil moisture estimates at different depths. Overall, this work contributes to ongoing efforts to better understand soil moisture variability across depths while supporting the development of transferable approaches for root-zone soil moisture monitoring in agricultural systems.

Mon & Tue

11:00–12:30

The exchange of carbon, water, and energy fluxes between the land and the atmosphere plays a vital role in shaping global change and extreme events. Yet our understanding of the theory of this surface-atmosphere exchange, represented via land surface models (LSMs), continues to be limited, highlighted by marked biases in model-data benchmarking exercises. Here, we leveraged the PLUMBER2 dataset of observations and model simulations of terrestrial sensible heat, latent heat, and net ecosystem exchange fluxes from 153 international eddy-covariance sites to identify the meteorological conditions under which land surface models are performing worse than independent benchmark expectations. By defining performance relative to three sophisticated out-of-sample empirical models, we generated a lower bound of performance in turbulent flux prediction that can be achieved with the input information available to the land surface models during testing at flux tower sites. We found that land surface model performance relative to empirical models is worse at edge conditions – that is, LSMs underperform in timesteps where the meteorological conditions consist of coinciding relative extreme values. Conversely, LSMs perform much better under “typical” conditions within the centre of the meteorological variable distributions. Constraining analysis to exclude the edge conditions results in the LSMs outperforming strong empirical benchmarks. Encouragingly, we show that refinement of the performance of land surface models in these edge conditions, consisting of only 12 %–31 % of all site-timesteps, would see large improvements (22 %–114 %) in an aggregated performance metric. Better performance in the edge conditions could see mean relative improvements in the aggregated metric of 77 % for the latent heat flux, 48 % for the sensible heat flux, and 36 % for the net ecosystem exchange on average across all LSMs and sites. Precise targeting of model development towards these meteorological edge conditions offers a fruitful avenue to focus model development, ensuring future improvements have the greatest impact.

Tue

09:15–09:30

Understanding land-atmosphere coupling in agricultural landscapes is critical for improving predictions of regional climate and boundary-layer dynamics. Surface energy partitioning between latent and sensible heat fluxes strongly influences near-surface temperature, humidity, and atmospheric boundary layer (ABL) growth. Although biophysical processes strongly influence these processes, current land surface models employed to study land-atmosphere coupling often lack the fidelity to accurately represent differing crop species, limiting understanding of how agricultural landscapes interact with and influence the ABL.
Agricultural systems present unique challenges due to rapid changes in canopy structure and phenology, which alter albedo, roughness, and evapotranspiration rates throughout the growing season, while multi-crop landscapes add further challenges associated with spatial heterogeneity. To address these complexities, we employ large-eddy simulations (LES) using the PArallellized Large eddy siMulation (PALM) model system. Eddy-covariance towers at Iowa State University’s Sustainable Advanced Bioeconomy Research (SABR) farm provide energy flux and environmental observations over four unique agricultural systems under identical climatic conditions, and include both annual and perennial systems. Here we use these observations to drive the development of physics‑constrained machine-learning emulators for three of these agricultural systems (corn, soybean, and miscanthus) designed to reproduce surface energy fluxes under early‑season conditions based on crop phenology and surface climate.
Initial simulations reveal that biophysical differences between annual and perennial crops produce a prolonged period of differential surface heating early in the season when evaluated at daily scales. As canopies mature, contrasts in ABL development become more muted but persist. Ongoing work extends this framework to assess how cumulative effects from varying planting configurations shape surface microclimate and boundary‑layer evolution across the early growing season. These efforts aim to integrate crop‑specific biophysical behavior into a coherent modeling framework, ultimately advancing a more holistic understanding of how planting decisions shape land-atmosphere feedbacks across agricultural landscapes.

Mon

16:00–16:15

Global vegetation provides a wide range of ecosystem services to society. Beyond food and wood production, land vegetation absorbs between 25 % and 33 % of anthropogenic CO2 emissions, helping to stabilize the climate system. Furthermore, vegetation moderates the impact of extreme events due to its role in the exchange of water and energy between the lower atmosphere and the shallow subsurface. Therefore, it is fundamental to have a realistic representation of vegetation dynamics in Earth system models, both at global and regional scales, in order to understand the evolution of the climate. Vegetation processes within land surface models (LSMs) typically depend on simulated physical and biogeochemical processes, with soil hydrology being one of the main physical factors influencing vegetation dynamics. Nevertheless, the implementation of hydrological processes within LSMs hinges, amongst others, on empirical parameterizations of soil hydraulic properties, which introduces uncertainties into the simulated soil water. These uncertainties then propagate, rendering vegetation responses to climate variability and extremes uncertain.

Here, we explore the different vegetation trajectories simulated by the ICON-Land LSM under perturbed soil hydraulics parameter ensembles. The QUINCY configuration of the model is used in order to consider a comprehensive representation of vegetation responses to water limitation and biogeochemical constraints. Alternative pedotransfer functions, selected from the literature, are used to translate the LSM’s default soil texture map into the alternative hydraulic parameter sets used in our perturbation experiment. Different soil hydraulic parameterisations lead to changes to the simulated soil moisture, both at regional and global scales. Differences in soil water levels affect the evolution of vegetation-related variables such as gross primary productivity and leaf area index. For example, a pedotransfer function producing higher field capacity values would lead towards wetter soils, that in turn would lead towards higher values of gross primary productivity and leaf area index. Opting for appropriate soil hydraulic parameterisations – able to realistically represent soil hydrology at different grid resolutions – remains fundamental to ensure the realism of the simulated soil water and vegetation dynamics in land surface models.

Mon

14:30–14:45

Despite exerting a fundamental control on surface energy partitioning and land–atmosphere interactions, the forest floor organic layer (litter and duff) is not explicitly represented in mainstream land surface models (LSMs), including Noah-MP, leading to systematic biases in simulations of surface heat and moisture fluxes. Using observational datasets from the USDA Forest Inventory and Analysis (FIA) program and the National Ecological Observatory Network (NEON), we developed a physics-constrained, data-driven parameterization framework for the contiguous United States (CONUS). Specifically, machine-learning models were used to derive spatially continuous parameter fields of organic layer properties, which were then mapped into a modified Noah-MP architecture to constrain key variables governing surface thermal and evaporative resistances. Preliminary evaluations show that the introduced insulation and surface resistance effects significantly modulate sensible–latent heat partitioning, damp the diurnal amplitude of radiometric surface temperature, and improve the representation of latent heat release during dry-down periods, thereby refining simulations of surface energy balance. This study establishes a practical paradigm for integrating large-scale ecological inventory data into physically based land surface and boundary-layer modeling, substantially reducing parameter uncertainty and enhancing the realism of land–atmosphere coupling in regional climate simulations.

Tue

15:45–16:00

Leaf area index (LAI) is a land surface variable of global significance, describing the density of land surface vegetation and its seasonal variation. It directly affects the interception of rainfall and is used to determine the degree of evapotranspiration in numerical models, influencing surface energy balance and land-atmosphere exchanges of moisture and heat.
Currently, LAI within the Met Office’s models is represented by a mean climatology, derived from five years of MODIS observations (2005–2009), at a resolution of 4 km (0.036°). These five years govern the representation of global vegetation in all model runs at the Met Office, in combination with vegetation fraction data from our land cover datasets. While this LAI dataset was sufficient when introduced, more recent regional modelling activities at 1.5 km and 300 m would benefit from a dataset of higher spatial resolution.
This work presents the development and evaluation of new LAI ancillaries, produced using more modern and higher resolution Earth observation products from Copernicus and NASA’s MODIS programme, and designed as potential replacements for the existing legacy climatology.
New ancillaries were generated both over the legacy climatological period (2005–2009) and an extended observational period (1999–2019). In addition, an experimental ‘blended’ ancillary was produced from the mean of the full-period Copernicus and MODIS datasets. When comparing the new 2005–2009 ancillaries to the existing product, the Copernicus dataset shows reduced winter LAI across north‑west and central Europe. It also exhibits higher, less variable LAI in the tropics – particularly over the Amazon – while summer values remain broadly consistent. By contrast, the updated MODIS dataset shows largely similar global patterns to the legacy climatology, with some minor regional differences.
Initial testing of short 2021 case studies, using the 2005–2009 Copernicus-based ancillary, shows some significant regional impacts, without general degradation in global forecast performance. Some improvements in the tropical and Northern Hemisphere temperature at 500 and 850 hPa have been observed, in addition to an improvement in tropical surface relative humidity.
Further testing of case studies, using the new MODIS and Copernicus ancillaries in the current and next-generation Met Office modelling systems, will be performed across a broader range of random and targeted case dates. The results will be compared against control simulations using the existing LAI data, in addition to evaluation against a range of observational datasets.
Keywords: leaf area index, numerical weather prediction, high-resolution observations, land surface modelling, vegetation

Mon & Tue

11:00–12:30

The El Niño Southern Oscillation (ENSO) exerts control on global weather patterns in the winter hemisphere directly through anomalous heating of the tropical atmosphere over the Pacific. However, the processes through which ENSO influences Northern Hemisphere summer circulation are poorly understood. Here we identify a soil‑moisture pathway linking the phase of the leading boreal summer circulation pattern to the preceding ENSO state. Lagged-maximum covariance analysis is used to reveal a coherent relationship between soil‑moisture anomalies and the leading upper‑tropospheric mode, with expansion coefficients strongly correlated with the winter Nino3.4 index. These soil‑moisture anomalies are generated during winter and spring by ENSO‑related land‑surface forcing and persist into summer, where they modify the atmosphere over North America and Central Asia, regions of strong land–atmosphere coupling, triggering a Rossby wave response. Sensitivity experiments for summers following the 1997/98 El Niño and 2021/22 La Niña confirm this finding. These results demonstrate that soil moisture provides a cross‑seasonal memory that carries ENSO’s influence on the atmosphere from spring to summer.

Wed & Thu

11:00–12:30

Terrestrial vegetation mediates carbon and water fluxes between land and atmosphere via photosynthesis, and can act as a carbon sink if assimilation is higher than respiration. Thus, it is of utmost importance to study vegetation response to changing climatic conditions. Long term measurements of carbon fluxes using eddy covariance (EC) methodology have shown large interannual variability (IAV), and its effect needs to be disentangled to better understand vegetation response. However, current biogeochemical models simulating photosynthesis or gross primary productivity (GPP) often struggle to capture the IAV. Here, we used two models: (1) a light use efficiency (LUE) model, which includes response functions of temperature, vapor pressure deficit (VPD), carbon dioxide concentration, cloudiness, and drought stress, and (2) an optimality-based model, which includes acclimation and drought stress, to study how model parameterization affects the representation of IAV in these models. We parameterized both models from detailed approaches, such as calibrating parameters per site-year, per site, to more generalized approaches, such as calibrating parameters per plant functional type (PFT), and globally for all sites. We found that both models simulated annual average GPP best when parameters were calibrated per site-year among all other parameterization approaches. For example, the site-year calibrated LUE model produced a median normalized Nash-Sutcliffe efficiency of 0.74 between annual average GPP estimated using EC measurements from 198 sites and respective model simulation. Furthermore, we performed sensitivity tests by simultaneously calibrating a group of parameters from a specific environmental response function per site-year, while keeping all other parameters fixed across years within a site. This allowed us to investigate the annual variation of which parameters are important to better capture annual GPP. We found that when only the parameters related to drought stress and hydrology were varied per year, it produced the best annual model performance for arid sites (median NNSE of 0.92 and 0.89 for forests and grasslands, respectively, for the LUE model). Considering all the sites from the dataset, the highest median NNSE achieved was 0.73 when parameters related to VPD or atmospheric dryness were calibrated per year.

Currently, we are interested in investigating whether the yearly variation of model parameters can be statistically explained using annual meteorological conditions, such as annual average temperature, precipitation, or their lag effect from the previous year. We plan to test several non-linear regression approaches, such as random forest, XGBoost or explainable machine learning, such as symbolic regression, to study the variability of parameters inverted per year from our previous experiments.

Overall, our experiments highlight the need for temporally varying, more flexible model parameterization to account for the effects of missing processes or limited understanding in current models. Moreover, the ability to spatiotemporally upscale model parameters, that is, predicting parameters for each year across the globe for locations with missing EC and micrometeorological measurements, can help in resolving the discrepancies between global estimates of carbon fluxes.

Wed

14:30-14:45

Satellite observations from semi-arid regions such as the Central U.S., mid-northern India, and north-eastern China reveal that summer surface soil moisture follows a bimodal distribution. This bimodal distribution features two distinct peaks: a dry state nearing the wilting point and a wet state nearing field capacity. The position of these peaks is crucial for land-atmosphere interaction; as soil moisture increases within the transitional regime, surface fluxes rise until they reach a maximum in the wet regime, where control shifts to incoming radiant energy. Over the last five decades, these semi-arid regions have emerged as global agriculture and irrigation hotspots, raising important questions about how human water management may be reshaping these soil moisture dynamics. This study pursues a two-fold objective. First, we assess how irrigation impacts the bimodal probability distribution of soil moisture. Second, we examine how irrigation-driven shifts in these distributions alter land-atmosphere interactions. To meet these objectives, we analyzed soil moisture data from the MIROC-INTEG-LAND model, sourced from the ISIMIP3a simulation and forced by GSWP3-W5E5 atmospheric conditions. Because observations lack a counterfactual scenario excluding human intervention, we utilized offline simulations to isolate irrigation’s influence. Our analysis compares two distinct scenarios: a “Default” experiment and a “No-Water-Management” experiment, which excludes all irrigation. By keeping all other forcing variables identical between these experiments, we quantify the extent to which soil moisture dynamics and land-atmosphere sensitivities respond to irrigation. Our results reveal a strong link between expanded irrigation and the weakening of bimodal characteristics, particularly where irrigation covers a greater proportion of the grid cell. Irrigation shifts the soil moisture distribution rightward toward a wetter regime. However, this shift does not always push the distribution into a state where radiant energy dominates; instead, soil moisture often remains the active controller of land-atmosphere interactions.

Mon & Tue

11:00–12:30

Eddy covariance (EC) is the primary method to estimate turbulent surface fluxes between surfaces and the atmosphere. However, nearly all EC flux towers fail to observe energy balance closure. Multiple reasons exist, and were explored at a recent AGU Chapman conference in Sept 2025. Several groups have proposed corrections that improve closure based on advances in instrument corrections, raw data processing, accounting of large-scale atmospheric transport, or estimation of unaccounted storage terms. Here, we present initial systematic evaluations from a group formed to evaluate how published and recently developed methods contribute to resolving energy balance closure at a set of consistently processed NEON and ICOS flux tower sites. We compare the corrections’ impact on overall closure, on surface energy balance partitioning, and on diel to seasonal patterns of turbulent fluxes. We also investigate the conditions under which different correction methods perform better or worse. Findings help us scope approaches for reporting turbulent fluxes from eddy covariance with low or no systematic bias and with appropriate metrics of uncertainty and quality for model calibration and validation.

Wed & Thu

11:00–12:30

Soil moisture (SM) regimes, the relationship between SM and evaporative fraction (EF), are important indicators of land–atmosphere interactions. Although Earth system models are commonly used to characterize these regimes, there are significant spatial discrepancies between modeled and observed patterns of SM regimes. In this study, we apply logistic regression to global model output and satellite-derived maps of SM regimes to identify which environmental drivers contribute to geographic mismatches in SM regimes between models and observations. Logistic regression is a supervised learning algorithm used for binary classification tasks to predict the probability that an instance belongs to the “positive” class (typically labeled 1) versus the “negative” class (labeled 0).

To train the logistic regression model, we use multiple predictor variables, including meteorological data, vegetation cover fraction, soil characteristics, and topography. Static fields include gridded soil data provide the percentages of clay and sand across six soil layers, topographic data from the Earth TOPOgraphy (ETOPO1). Daily meteorological data used to calculate seasonal and annual climatologies are derived from ERA5. Mean precipitation and precipitation intensity, as the 99th percentile of daily mean precipitation rate, are from Multi-Source Weighted-Ensemble Precipitation (MSWEP). The aridity index (AI) is computed from both ERA5 and MSWEP. Climatologies from 8-day Leaf Area Index (LAI) data from the Global Land Surface Satellite (GLASS) product and annual land vegetation fractions from the ESA CCI PFT datasets are also used as predictor variables in the logistic model.

In addition to the main predictor variables, the logistic regression model includes interaction terms to capture cases where the effect of one predictor on regime mismatch depends on the level of another. Including these terms allows the model to identify not only relevant individual predictors, but also under what compound environmental conditions mismatches are amplified or suppressed. In this study, the interaction terms are divided into these broad categories: energy-water interactions, atmospheric demand × water availability, vegetation × climate interactions, and soil × climate interactions (how soil texture modulates hydroclimate sensitivity).

The target variables indicate regime matches and mismatches between observationally-derived and model SM regimes. The models are the Community Earth System Model: CESMv2 with the CLMv5 land model, and the Model for Prediction Across Scales framework (MPAS) coupled with the Noah Multi-Parameterization (NoahMP) land surface model. Observational estimates of SM regimes are calculated using multiple sources: soil moisture (Soil Moisture Active Passive mission: SMAP, The Climate Change Initiative: ESA CCI v09.1, and SoMo.ml) and surface heat fluxes (Global Land Evaporation Amsterdam Model: GLEAM v4.1a).

After training the logistic regression model separately for three SM regime types (dry, transitional, and wet), a significance test is applied to the coefficients. Then, odds ratios (multiplicative effect on odds) and confidence intervals are calculated. Each model’s performance is also evaluated using 5-fold stratified cross-validation (CV) and cross-entropy (log-loss). Global maps provide visualizations of dominant SHAP (from SHapley Additive exPlanations) contributors and their direction effects are used to identify where and how climate, vegetation, and soil conditions are linked to the mismatches between model and observational SM regimes. Such information provides an unprecedented look into sources of error in coupled land-atmosphere model systems, providing useful guidance to targeting areas for development and identifying systematic biases in both land surface physics and atmospheric model parameterizations. The results can also improve our understanding of ecosystem dynamics and climate variability, identifying potential implications for model simulations.

Wed

15:00–15:15

Cold pools are regions of cool air near the ground that form in association with convective clouds when cool, dense downdraft air reaches the surface and spreads laterally. This latent cooling combines with hydrometeor loading to increase the local air density. The resulting negatively buoyant pocket of air descends, forming a downdraft. Upon reaching the surface, the dense air collects, forming a laterally expanding outflow consisting of cool air (i.e., the cold pool). Cold pools are often associated with gusty near-surface winds, and they can help to trigger convective initiation along their peripheries while suppressing convection within their stably stratified interiors. As such, they provide a mechanism by which existing convective storms can generate new storms, with important implications for the diurnal cycle of precipitation. These boundary-layer processes occur on horizontal spatial scales too small to be resolved by many earth system models.

It is crucial to understand not only where cold pools will form, but also how cold they will be. The rate of cold pools’ lateral expansion is driven in large part by the density difference between the cold pool air and the ambient air. All else equal, a denser cold pool is likely to expand more rapidly. By extension, a denser cold pool will likely exhibit stronger near-surface winds and provide a more vigorous lifting mechanism for triggering the formation of new convective clouds along its periphery. A colder, denser cold pool can also provide more persistent stable stratification to inhibit cloud formation within its interior.

Our study asks: Where, specifically, does cold pool air become cold, and how? What is the role of land-atmosphere interactions?

It is often asserted that the cooling occurs within rain shafts, by virtue of the phase changes (e.g., evaporation and melting) and consequent latent cooling that take place there. However, latent cooling need not be driven solely by phase changes of airborne hydrometeors. Upon falling onto a land surface, precipitation can face a variety of fates, including interception by vegetation and percolation into the soil. Any canopy-intercepted rainfall exposed to the air above remains available to undergo further phase transitions. The resulting latent cooling can increase the local air density so as to contribute to the cooling of cold pool air.

In this study, we use numerical simulations to assess the extent to which outflow air is cooled by the evaporation of canopy-intercepted rainfall, as opposed to phase changes of airborne hydrometeors within rain shafts. Daytime tropical rainforest convection is simulated at LES resolution using the Regional Atmospheric Modeling System (RAMS), which is coupled to the Land Ecosystem-Atmosphere Feedback, version 3 (LEAF-3) surface scheme. We assess the effects of canopy-intercepted rainfall using a mechanism denial approach, and we test the sensitivity of our results to soil moisture and land cover. Compared to a baseline forest simulation in which land-precipitation interactions are removed, we find that evaporation of intercepted rainwater can create temperature perturbations whose magnitude corresponds to approximately 20-50% of the total cooling. Evaporation of canopy-intercepted precipitation has an especially profound effect on cold pool lifetime: the boundary layer recovers much more slowly when land-precipitation interactions are enabled. These effects are substantially reduced when vegetation cover is eliminated and rain water instead percolates directly into the soil. This presentation will discuss the mechanisms leading to these responses, as well as the broader implications for the diurnal cycle of convection over land.

Mon

16:30–16:45

Tree species are differently affected by drought stress, with responses regulated by plant hydraulic traits. However, it remains challenging to quantify species-specific plant hydraulic traits due to lack of data availability. In this study, we apply the terrestrial biosphere model QUINCY, augmented by a recently developed plant hydraulic architecture module, across three eddy covariance sites in Germany covering broadleaved forest species (Aplern, Hainich and Hartheim). We constrain QUINCY across these species and sites using 30-minute resolution stem water potential measurements collected during the summer and autumn of 2023. Our results show that two groups of model parameters explain most of the simulated plant water potentials: parameters controlling plant water uptake from soil, and parameters regulating stomatal sensitivity to pre-dawn leaf water potential. Based on the optimised parameter sets, the isohydric sensitivity to VPD exhibits species-specific variation. Furthermore, the species-specific optimisation of plant hydraulic improves the model performance of evapotranspiration and gross primary productivity during the drought years. Our study demonstrates that integrating the new generation of in situ plant hydraulic observations into vegetation models can facilitate the quantification of species-specific hydraulic traits and parameters, effectively reducing uncertainty in, and providing robust constraints on, modelled responses to drought.

Mon

13:45–14:00

Vegetation can regulate surface energy and water exchanges that influence boundary layer development and cloud formation. While the impacts of vegetation on albedo, roughness, and evapotranspiration are relatively well known, the indirect influence of trees on clouds are less quantified. Furthermore, the role of the spatial heterogeneity of tree cover is even less understood, in particular within open forests and savannahs that are widely present in African landscapes. This gap limits our ability to represent surface heterogeneity in land atmosphere coupling and to anticipate atmospheric responses to land cover change.

Here we propose a data-driven exploration of how tree cover, in terms both of its absolute coverage and its spatial configuration, are correlated with cloud cover as observed from the European geostationary weather satellites. We use dedicated tree cover maps originating from a dedicated cartographic effort leveraging on very high spatial resolution remote sensing imagery and machine learning. We combine both sources of information, i.e. surface heterogeneity description along with hourly cloud cover data from 2017 to 2021, using a curated space-for-time substitution approach over local moving windows. The result is a detailed description of the spatio-temporal dynamics of the biophysical sensitivity of cloud cover to tree cover and tree cover heterogeneity. Our findings reveal distinct patterns of cloud sensitivity to tree cover changes across climatic zones and elevations, linked to energy partitioning during the day and land surface temperature disparities at night. During the wet season, an increase in spatial heterogeneity amplifies the average reduction in cloud cover associated with forest degradation by about 30%. In contrast, in savannah regions, increased heterogeneity following tree cover gain enhances the positive cloud response by approximately 40%, highlighting the critical role of tree distribution.

These findings highlight how surface heterogeneity, as characterized by the spatial heterogeneity in tree cover, has a considerable role to play is land-atmosphere interactions. The results can potentially provide observation-based guidance for representing subgrid vegetation patterns in models. They can further serve to inspire the design of land restoration strategies in Africa that account for both how much tree cover is added and where it is placed.

Tue

09:30–09:45

Land-surface processes and land-atmosphere interactions are fundamental elements of the Earth’s water, energy and carbon cycles. The land surface exhibits heterogeneity on a wide range of scales and is the location of many complex physical and biological processes. It is tightly coupled to the atmospheric boundary layer, where energy and moisture are transported by turbulent motions also operating on a range of scales.

Improving the representation of land-atmosphere interactions is essential for increasing the accuracy and detail of numerical weather forecasts and climate models, especially as the resolution of these models increases and as they are required to represent more components of the environmental system. This requires a synergistic combination of detailed observations and modelling studies. GLAFO (GEWEX Land-Atmosphere Feedback Observatory) sites provide observations of the boundary layer and land-surface in unprecedented detail and modelling studies based on these data will enable deeper insights into processes operating in the atmospheric boundary layer, its interaction with the land surface and the treatment of these interactions in models.

The GEWEX Atmospheric Boundary Layer Study (GABLS) was a project that ran for two decades and provided a focus for efforts to improve the modelling of the atmospheric boundary layer by formulating a series of intercomparisons of increasing complexity. Each intercomparison was based on observational data, though with some idealization, while the representation of the surface became more important and more detailed through the project. The formulation of a new intercomparison based on GLAFO observations is a natural extension of these activities.

Based on data collected at Hohenheim during the summer of 2025 using enhanced instrumentation, an intercomparison is being formulated, with the intention of formally releasing it at the PanGLASS conference. Following the approach previously used in GABLS, the case will comprise a single-column component and three-dimensional simulations for enhanced process-understanding. Simulations using convection-permitting regional simulations will link these scales. The initial focus of the project will be a detailed comparison between the observational data and the simulated fields, especially focusing on turbulence in the atmospheric boundary layer, but its scope will subsequently be broadened to tackle key issues such as land-surface heterogeneity. Beyond this first case, the aim of the project is to establish a methodology that can be applied to data obtained on other days or to observations made at other GLAFO sites.

Wed & Thu

11:00–12:30

Local land-atmosphere coupling involves the interactions between the land-surface and the atmospheric boundary layer (ABL), and in turn with the free atmosphere above. These interactions consist of a “terrestrial leg”, i.e. the evolution of surface fluxes via sub-surface heat and moisture transport and surface-layer turbulence, and an “atmospheric leg”, i.e. boundary-layer development and the effect of surface fluxes and the role of warm- and dry-air entrainment into the ABL from the free atmosphere above. We examine these interactions using brief analytical developments for both terrestrial (i.e. via “little omega”) and atmospheric (i.e. via RH-tendency at ABL top) legs, followed by an examination of observational data sets and corresponding land-ABL modeling studies including the potential for the initiation of fair-weather cumulus.

Thu

09:00–09:15

Model evaluation typically involves comparing simulated outputs with observational datasets, where deviations are interpreted as errors. However, observational datasets carry uncertainties, and multiple products representing the same variable may offer equally valid representations of reality. As a result, model errors depend on the choice of observational reference.
To address this, we introduce an approach that considers model outputs to be indistinguishable from observations when they fall within the observational range. Errors are only assigned when model values lie outside this range. These adjusted errors can be used to compute Observation Range Adjusted (ORA) statistics, which isolate measurable discrepancies, support more robust model intercomparisons, and help identify areas for targeted model improvement.
We apply ORA statistics to assess the added value (AV) of a Regional Climate Model (RCM) ensemble over the CORDEX-Australasia domain. RCMs show improved performance when evaluated against observational datasets that integrate in-situ measurements with satellite and reanalysis data, and apply precipitation undercatch corrections.
Accounting for observational uncertainty in model evaluation allows model development to be focused on model deficiencies including process representation, and avoids the possibility of a mistaken focus on areas with large observational uncertainty that can be highlighted when using single observational datasets in model evaluation. and AV assessments, offering a more nuanced understanding of model skill and guiding future model development.

Tue

09:45–10:00

The role of soil moisture in land-atmosphere exchange processes is crucial as it impacts water and energy fluxes between the land surface and the atmosphere. Soil moisture plays a significant role in regulating the hydrological cycle, affecting climate and weather processes on various scales.
Point measurements of soil moisture are not necessarily representative for the scales relevant to land surface – atmosphere exchange and feedback. The Cosmic-Ray Neutron Sensing (CRNS) method provides complementary information at the field scale and with temporal smoothing towards daily or subdaily variations. Furthermore, changes in crop biomass can be derived by investigating the ratio of epithermal and thermalized neutrons.
We utilize the CRNS method to observe and compare the soil moisture dynamics of two nearby wheat and maize stands for one growing season at the Land-Atmosphere-Feedback Observatory (LAFO) of the University of Hohenheim. Our measurements are collocated with additional point scale soil moisture sensor profiles and a range of near surface atmospheric observations, like, e.g., eddy covariance towers which enables us to relate differences in field scale soil moisture dynamics with micrometeorological conditions.
We will present the sensor setup and the reference calibration of the CRNS measurements for the two different crops of the 2025 field campaign of the Land-Atmosphere Feedback Initiative (LAFI) at the LAFO and show a first analysis of the field scale soil moisture and biomass variations and their relation with the near surface meteorology.

Wed & Thu

11:00–12:30

The response of terrestrial freshwater storage to anthropogenic climate forcing is a fundamental yet poorly constrained aspect of global hydrological change. Detection and attribution studies have identified human influence in several components of the hydrological cycle, including precipitation and runoff (e.g. Zhang et al., 2007; Marvel et al., 2019; Gudmundsson et al., 2021). However, attribution of observed changes in terrestrial water storage (TWS) has remained elusive due to the short length of observational records, substantial internal climate variability, and the confounding influence of direct human water management.

Here we use observations from NASA’s Gravity Recovery and Climate Experiment (GRACE), providing a uniquely robust, spatially explicit measure of terrestrial water storage change, together with a formal detection and attribution framework (Santer et al., 2013) informed by simulations from the Coupled Model Intercomparison Project Phase 6 (CMIP6). We show that the observed GRACE TWS record contains a spatially coherent signal that exceeds the range of simulated unforced variability, strengthens over time, and is robust across alternative fingerprint constructions and GRACE processing choices.

The detected fingerprint is characterised by large-scale wetting and drying patterns broadly consistent with modelled responses to anthropogenic forcing across many regions. Regional deviations are primarily concentrated in intensively irrigated and groundwater-dependent areas, indicating the superimposed influence of direct human water use and remaining model limitations. Additional analyses using reanalysis products and observationally constrained climate model simulations provide complementary context for interpreting the emergence of this signal. Together, these results provide the first fingerprinting evidence of anthropogenically forced change in global terrestrial water storage and establish continental freshwater storage as a detectable and attributable component of the climate system.


References

Zhang, X. et al. Detection of human influence on twentieth-century precipitation trends. Nature 448, 461–465 (2007).
Marvel, K. et al. Twentieth-century hydroclimate changes consistent with human influence. Nature 569, 59–65 (2019).
Gudmundsson, L. et al. Globally observed trends in mean and extreme river flow attributed to climate change. Science 371, 1159–1162 (2021).
Santer, B. D. et al. Identifying human influences on atmospheric temperature. PNAS 110, 26–33 (2013).

Mon

16:30–16:45

Soil moisture memory is thought to provide a source of temperature predictability on subseasonal-to-seasonal (S2S) timescales. On the order of weeks to months, soil moisture and temperature anomalies persist in ways that impact surface fluxes and the partitioning between heat and moisture, feeding into near surface temperature anomalies and potentially into cloud cover and precipitation. Previous studies have even highlighted the potential for land surface anomalies to enhance predictability of certain extreme weather events.

Recent results from Richter et al. (2024), however, indicate that one of the leading earth system models used globally – the Community Earth System Model (CESM) – fails to produce predictability gains from realistic land surface initialization. Instead, initializing subseasonal forecasts with climatological land surface conditions results in predictions of 2 m temperature that are on par with (if not better than) realistic initialization. The surprising result deserves further investigation to determine if the hypothesis that unrealistic land-atmosphere coupling is one of the culprits behind the lack of system memory. Additional hypotheses included that the influence of land on predictability is highly regional and perhaps only present for certain conditions (i.e., under wet or dry soil moisture anomalies, certain atmospheric conditions, etc.) and that soil moisture anomalies aren’t reasonably captured in CESM.

In this work, we seek to understand (a) the realism of land-atmosphere coupling as represented in CESM; (b) when and where soil moisture anomalies may actually enhance predictability; and (c) compare these results across models to understand how endemic this behavior is. The latter goal will expand our analysis beyond CESM to incorporate both the DOE Energy Exascale Earth System Model (E3SM) and NSF NCAR’s Model for Prediction Across Scales (MPAS). Differences in models may highlight the impacts of different parameterizations, missing processes, etc., while agreement across them points to consistent behavior that is endemic amongst some of the leading models in the world.

Wed

09:15–09:30

This work investigates the atmospheric response to prescribing inter-annually varying Leaf Area Index (LAI) in seasonal hindcasts and assesses its effect on seasonal forecast skill. We focus on Africa, where seasonal forecasts are crucial for agricultural planning and extreme weather preparedness.
In order to investigate the effect of inter-annual variations in LAI, a suite of seasonal hindcasts produced for ECMWF’s CONFESS project are used. Two experiments within the project are compared – a control experiment, which uses climatological LAI and a non-varying land cover map, and an experiment which implements a dataset of inter-annually varying LAI and land cover maps constructed by merging multiple satellite products. Preliminary investigation shows that prescribing inter-annually varying LAI increases African near-surface temperature by up to 0.2K in December-January-February (DJF). We perform a Seasonal-reliant Empirical Orthogonal Function analysis on the driving LAI dataset. This uncovers a mode of variation strongest in DJF that is correlated with the Indian Ocean Dipole (IOD) index for the September-November-December season (correlation coefficient of ~0.75); thus, we view this mode of LAI variation as the vegetation response to increased East African rainfall during active IOD events. The mode of variation in LAI is shown to lead to a temperature response, which is associated with simulated changes in the surface energy balance. We then show that these changes in temperature, seen only when vegetation varies inter-annually, correspond to improvements in bias (calculated as the difference to ERA5 near-surface temperature values). These improvements in temperature forecast skill are largest following extreme IOD events and for areas where the control experiment’s hindcasts’ bias values are largest, with a maximum in temperature bias reduction of 0.6K and an average bias reduction of 0.2K. Thus, our results show that increased complexity in vegetation representation within seasonal forecasts leads to improved forecasts of near-surface temperature, especially through better representation of land-atmosphere feedbacks influenced by the IOD for East Africa.

Mon & Tue

11:00–12:30

Better predictability of monsoonal precipitation in semi-arid environments at a range of time scales from hours to decades is critically needed. At the time scale of weather forecasts, hazardous weather associated with deep convection, including strong wind gusts, wildfires, and flash floods, remains poorly predicted. Climate models offer little reliable guidance on the sustainability of water supplies to the dense populations in these regions, including the Southwestern United States and South-central Asia. CMIP6 models disagree with observations over the last four decades in the Southwestern US: they all predict an increase in water vapor and more precipitation, but it didn’t happen. Key processes controlling precipitation on the northern margins of the monsoon include long-range water vapor transport, land-atmosphere interaction, PBL mixing, moisture venting into the free troposphere, and the off-terrain movement of deep convection. These processes are poorly measured and their feedbacks are poorly understood. Strong diurnal surface forcing, weak synoptic forcing, and terrain all are intrinsic components. This talk will highlight several integrated field campaigns that are being planning in the Southwestern US to deliver an improved understanding of the coupled processes that control the cycling of water vapor between the soil and the free troposphere, through the combination of multi-scale observational networks, data assimilation, and modeling.

Wed

09:30–09:45

High-resolution numerical weather prediction (NWP) models are increasingly being used to study the interactions between the atmosphere and glaciers in complex alpine terrain. However, their performance under these conditions has not been sufficiently confirmed by observations, especially at a dekameter scale. This study comprehensively validates the Large-Eddy Simulation (LES) configuration of the ICOsahedral Nonhydrostatic (ICON) model using observations from the HEFEX II (2023) and HEFEX III (2025) field campaigns. Both campaigns included four weeks of intensive observations at Hintereisferner in the Ötztal Alps and were part of the international TEAMx research program, which studies multi-scale transport and exchange processes in mountainous environments.

HEFEX II focused on characterizing the spatial gradients and temporal variability of surface-layer variables, such as temperature, humidity, and wind. HEFEX III utilized coordinated UAV-based vertical profiling in combination with multiple on-glacier lidar systems to resolve atmospheric flow fields and wind patterns within the valley. Together, the two campaigns provide a unique and unprecedented observational dataset in complex glacierized terrain, offering an exceptional basis for model evaluation.

ICON-LES was applied in a one-way nested configuration, achieving a target horizontal resolution of 51 meters over the study area. We assessed model performance using qualitative and quantitative validation approaches, particularly emphasizing the model’s ability to reproduce the spatio-temporal variability of key atmospheric parameters across surface and boundary-layer scales. The results demonstrate strong agreement between ICON-LES simulations and multi-platform observations, indicating that the model realistically captures flow structures and variability in a high alpine glacier environment.

These findings support the use of ICON-LES as a reliable tool for studying atmosphere-glacier interactions and lay the groundwork for future climate impact and feedback studies in complex terrain. At the same time, the analysis highlights the current limitations of high-resolution numerical modeling and emphasizes the importance of using advanced observational techniques and large-eddy simulations together to improve our understanding of processes in mountainous regions.

Mon & Tue

11:00–12:30

Droughts are a recurring feature of the climate system with large socioeconomic impacts. They occur on multiple scales across time and space and are controlled by different mechanisms. So-called ‘flash droughts’, which rapidly intensify over multi-week time periods, have been linked to land-atmosphere feedbacks mediated by boundary-layer processes that accelerate drying of the land-surface. These interactions may not be captured through local indicators, such that a study of land-atmosphere coupling metrics across an entire drought object that is transient in space and time may help reveal the mechanisms that cause them. Given the transient nature of drought and the need for assessing the models’ ability to represent land-atmosphere coupling and feedback at subseasonal scales, coupling metrics need to be responsive to timescales of drought. We have previously shown that drought intensification is associated with changes in the Convective Triggering Potential and Lower Tropospheric Humidity Index (CTP-HI_L) and Heated Condensation Framework (HCF) metrics.
Here, we quantify CTP-HI_L and HCF metrics for drought objects tracked by the Tracking and Object-Based analysis of Clouds (TOBAC) method to analyze drought and flash drought over the conterminous US during 2017 and 2021 as years with significant drought and flash drought. Coupling metrics are calculated using NLDAS-2 data for surface conditions and ERA5 atmospheric profiles on 137 model levels archived at the NCAR Research Data Archive (doi: 10.5065/XV5R-5344).
Using TOBAC we identified a total of 27 drought objects for 2017 and 23 drought objects for 2021. Using the Flash Drought Intensity Index (FDII) we found drought objects in 2017 to be more likely to be affected by flash drought compared to 2021 as shown by a higher number of drought objects with flash drought areas exceeding 50% (11 out of 27 objects in 2017 compared to 6 out of 23 in 2021).
A preliminary analysis for the 2017 U.S. Northern Great Plains drought that was dominated by flash drought conditions during its expansion from the Dakotas into Montana showed that ERA5 CTP-HI_L compared well with CTP-HI_L calculated from radiosondes, highlighting the applicability of ERA5 data for analyzing land-atmosphere coupling. Analyzing the trend of CTP-HI_L prior to drought expansion, we found that areas of drought expansion that are exposed to flash drought conditions showed steeper increases in CTP-HI_L. This indicates the positive association between flash drought and drought expansion and highlights the potential for land-atmosphere feedback. No such association between coupling and drought expansion was found during periods not dominated by flash drought during 2017 and 2021.
Though preliminary, these results support our hypothesis that drought objects with high area and duration of flash drought will exhibit more active land-atmosphere feedback. Therefore, an accurate representation of land-atmosphere interactions in models may contribute to improved predictability of hydroclimatic extremes at seasonal to subseasonal scales.

Wed & Thu

11:00–12:30

The scaling laws of turbulent transport in the urban roughness sublayer (RSL) remain a formidable challenge toward improved parameterization of surface-atmosphere exchange in urban canopy models. This challenge derives from the heterogeneous form of cities and the complex distribution of scalar sources, which result in the failure of Monin-Obukhov Similarity Theory (MOST) in predicting flux-variance relations. This work uses field measurements from the Boston Urban Roughness Sublayer Tower (BURST), installed above the rooftop of a 20-meter building at Northeastern University campus in Boston, to investigate the failure of MOST in predicting flux-variance similarity. By applying a multi-resolution analysis (MRA) to decompose turbulent signals into discrete frequency bands, we isolate the cumulative contribution of different scales to the total variance of velocity and scalar components. Our results reveal a distinct hierarchy of scale-dependent coupling between the variances and their respective surface fluxes. We find that small-scale, local turbulence (defined by wavenumbers commensurate with the measurement height) accounts for large fractions (eight percent) of vertical velocity and temperature variances, but only fifty percent of the horizontal velocity and humidity variances. These findings suggest that while vertical fluctuations are physically constrained by the roof surface, horizontal motions and moisture fluctuations are dominated by large-scale, anisotropic structures and non-local plumes that do not contribute to the local surface flux. By truncating these large-scale “inactive” motions, we demonstrate that the traditional scaling relations can be largely recovered. This suggests that the apparent breakdown of similarity theory in urban environments is not a failure of the underlying physics, but rather a consequence of energy leakage from non-local scales that are not accounted for in traditional surface-layer frameworks.

Mon

16:45–17:00

This work identifies the fundamental mechanisms controlling the normalized equilibrium surface temperature response to subsurface heating in an idealized subsurface-land-atmosphere configuration. Yet traditional case studies often obscure the underlying physics through site-specific complexity, limiting the ability to identify first-order controls and to generalize results. Using the Large-Eddy Simulation model PALM-4U, we impose constant subsurface temperature perturbations at 3.86 m depth over a 1000-day simulation period and systematically vary soil and atmospheric properties including wind speed, soil type, net radiation, and soil moisture. We derive simplified analytical models to predict the normalized equilibrium surface temperature change based on three heat transfer coefficients: atmospheric convective efficiency, latent heat efficiency, and subsurface conductive efficiency. By introducing a dimensionless framework, we establish universal, independent sensitivity metrics.

Results demonstrate that any process increasing the atmospheric convective efficiency reduces the normalized equilibrium surface temperature response, while the opposite holds for the subsurface conductive efficiency. Specifically, it decreases from 0.042 to 0.015 (65% reduction) as net radiation increases from 1 to 100 W/m². This reduction reflects a shift in the relative atmospheric contribution from 65.9% to 41.4% to the total change. This shift is mainly caused by a reduction in atmospheric potential temperature difference between the perturbed and the control case (from 0.294 K at 1 W/m² to 0.072 K at 100 W/m²) while atmospheric heat transfer efficiency increases by 62.7%. As wind speed increases from 3 to 10 m/s it reduces the normalized equilibrium surface temperature response from 0.015 to 0.012. However, compared to all external factors tested, wind speed shows a relatively minor impact. In contrast, soil moisture exerts the strongest influence: increases from 0.0 to 0.3 m³/m³ enhance the response by over 560% (from 0.015 to 0.1) via an eightfold increase in subsurface conductive efficiency. Yet relative partitioning to the surface temperature change remains stable (~60% subsurface contribution) due to a counterbalancing rise in atmospheric potential temperature difference (0.072 K to 0.523 K). When evaporation is included in the model, latent heat flux provides an additional energy dissipation pathway that dampens the change by effectively increasing the total atmospheric heat transfer capacity.
The analytical models successfully approximate the PALM simulation results. Hence, we provide a method to predict how the surface temperature responds to subsurface thermal perturbations. The strength to encode complex external factors into a scalable, physically consistent framework, enables the extension to future site-specific research across diverse geographic settings, seasonal variations, and diurnal cycles. It allows a first rapid estimation without directly requiring computationally expensive high-resolution simulations and eliminating the need to simulate every individual configuration. Consequently, our findings demonstrate that the surface is most sensitive to subsurface thermal perturbations under conditions of low net radiation such as during winter or nights and with high soil moisture.

Mon

17:00–17:15

The novel Terrestrial Systems Modelling Platform version 2 (TSMP2, github.com/HPSCTerrSys/tsmp2) regional Earth system model (RESM) is a full model makeover of TSMP1 and combines the next-generation atmospheric model ICON (icon-model.org), the eCLM (github.com/HPSCTerrSys/eCLM) land surface model, a fork of the Community Land Model v5.0 featuring an efficient software infrastructure for stand-alone use and model coupling, and the ParFlow (github.com/parflow/parflow) integrated hydrologic model (IHM) coupled via OASIS3-MCT (gitlab.com/cerfacs/oasis3-mct). TSMP2 is highly modular and supports heterogeneous supercomputing with ICON and ParFlow on GPUs of the latest exascale HPC systems. The TSMP model design is scalable and has proven to run across a wide range of spatio-temporal scales, from field to continental spatial and weather to climate time scales. Applications span a wide range of research topics, such as water resources assessments, land-atmosphere coupling, and climate change projections. The coupled model system provides a comprehensive and more realistic process representation of the terrestrial system, enabling a physically-based representation of transport processes and feedbacks across scales and sub-systems, groundwater-to-atmosphere, including human interventions. Aside from eCLM, which can be run in biogeochemistry mode and which also has an urban sub-module, the ParFlow IHM is a unique feature of TSMP2. With an IHM, 3D sub-surface hydrodynamics and surface hydrology are linked. The added value of using the IHM comes from the (lateral) redistribution of water, streamflow- and land surface-aquifer interactions can be simulated more realistically in such a model structure. Explicit simulation of groundwater processes therefore impacts energy budgets and the hydrological cycle. Altered land-atmosphere coupling, land water balance, and hydrometeorology lead to an improved reproduction of heat wave characteristics, for example. Scale-dependent feedback also occurs; km-scale, convection-permitting simulations require 3D hydrodynamics to account for the altered water distribution alongside a more detailed topography representation. Finally, the terrestrial water cycle can be closed, providing novel information for multiple Vulnerability, Impacts, Adaptation and Climate Services (VIACS) applications and water resource investigations, incl. anthropogenic water use. Aside from providing a brief technical overview of this new modelling platform, we show application examples from TSMP v1 and v2 simulations, such as the reproduction of heat waves and the impacts of human water use, land use, and land cover change. We also demonstrate how TSMP data can be used for land-atmosphere coupling studies, as well as for analyzing process representation in high-resolution climate simulations.

Wed

10:00–10:15

Ecosystem complexity emerges from interactions between biotic and abiotic processes operating across a wide range of spatial and temporal scales. While Earth system and ecosystem models represent many of these processes, they typically operate at resolutions that obscure fine-scale heterogeneity in microclimate, soils, and vegetation; heterogeneity that is critical for ecological dynamics. Bridging this scale gap remains a major challenge.

We present the Virtual Ecosystem, a mechanistic ecosystem model developed from scratch to operate at the lower end of spatial resolution (∼100 m) and explicitly integrate microclimate, hydrological, and ecological processes within a single framework. The model is driven by climate-model outputs, providing a structured way to extract large-scale atmospheric information to the scales at which organisms experience their environment.

Using the data-rich SAFE Project landscape in tropical Borneo as an initial test case, we describe the opportunities and challenges of assembling heterogeneous observations and calibrating a unified, process-based model at microclimate scales. This work highlights both the potential of virtual ecosystems to connect land–atmosphere science with ecology and the practical limitations imposed by data availability, scale mismatches, and model complexity.

Tue

09:45–10:00

Mountain regions such as the Pyrenees are characterized by strong surface heterogeneity, nonlinear thermal and hydrological processes, and biases in climate forcing, making them appropriate for testing the next generation of Land Surface Models (LSMs). This study develops and evaluates a novel cluster-based, multiscale modeling framework that rethinks land-atmosphere interactions through advanced representations of subsurface water and energy processes.

The multiscale framework leverages the cluster-based architecture of the HydroBlocks LSM and extends it to explicitly resolve lateral subsurface exchanges of both water and advective heat across three interacting spatial scales: local height band flow, intermediate subsurface hydrological units, and regional cross-domain connectivity. These multiscale fluxes are mapped back to the cluster-level water and energy balances and couple to the Noah-MP land surface physics. This mass and energy transport across heterogeneous surface units represents a novel methodological approach compared to traditional vertically independent tile formulations.

Using high-resolution (~100 m) ensemble simulations over the Pyrenees, we conducted a comprehensive sensitivity analysis to quantify how model performance depends jointly on parameter sets and on choices of cluster configuration, including domain discretization, regional aggregation strategies, and lateral connectivity assumptions. We identify dominant parameter interactions that control snow persistence, soil moisture redistribution, land-atmosphere energy exchange, and emergent sensitivities masked by grid-mean calibration approaches.

Model performance is evaluated against observed discharge from the Global Runoff Data Center, enabling a direct assessment of how heterogeneity representation and subsurface connectivity influence bias and predictive skill. Results indicate that configurations including multiscale lateral flow consistently outperform vertically independent tile formulations, demonstrating that uncertainty in mountainous regions is governed not only by parameter values but by how surface heterogeneity and subsurface processes are structurally represented within tile-based LSMs. This work provides a new benchmark framework for evaluating tile-based LSMs, advances process understanding of coupled water–heat transport in heterogeneous landscapes, and offers guidance for designing more robust, physically grounded models applicable across climates and terrain types.

Mon & Tue

11:00–12:30

Soil moisture (SM) exhibits a pronounced diurnal cycle, typically decreasing during the daytime due to evapotranspiration (ET). However, in-situ SM observations show an opposite pattern, with a daytime maximum that contradicts the expected physical behavior based on the surface water balance. This discrepancy primarily arises from the temperature sensitivity of high-frequency dielectric-based sensors, which leads to artificial positive correlations between SM and soil temperature (TS). In this study, we propose a physically based time-filtering approach to correct diurnal errors in hourly in-situ SM measurements. The method applies the Fast Fourier Transform to separate and remove temperature-induced diurnal components from SM observations in the International Soil Moisture Network (ISMN), using land surface model datasets (ERA5-Land and MERRA-2) as references. The corrected SM time series show improved consistency with the surface water balance and more realistic sub-daily variability, characterized by a morning maximum and a daytime minimum, along with a physically consistent negative correlation between SM and TS. The temperature-sensitivity in SM measurements is pronounced in climatological dry and hot regions, where the diurnal temperature range is exceptionally large. Furthermore, the adjusted SM exhibits more physically meaningful relationships with surface energy fluxes, particularly a negative correlation with latent heat flux (LH) during the daytime, consistent with ET-driven drying processes in energy-limited regions. These results suggest that the proposed filtering method effectively enhances the reliability of in-situ SM observations at sub-daily timescales, providing improved datasets for land-atmosphere interaction studies, model validation, and satellite validation.

Mon & Tue

11:00–12:30

Interactions between land surface, atmospheric boundary layer, and cloud processes impact climate from local to global scales. Over large forested areas, enhanced low-level cloud coverage and local to regional cooling effects have been observed. However, observations of land-mixed-layer-cloud interactions over the vast forested regions of eastern Canada have been lacking. To study these interactions, long-term detailed observations across the soil-vegetation-atmosphere continuum are required. Here, we present results from a new integrated atmospheric boundary layer observatory, which is located at the Acadia Research Forest in New Brunswick, Canada. The observatory integrates measurements of soil water content and potential, surface energy fluxes (using the eddy covariance technique), boundary layer turbulence (using a Halo Photonics StreamLine VS+ Doppler LiDAR), cloud cover (using a Vaisala CL61 ceilometer), and thermodynamic boundary layer profiles (using Sparv Embedded Windsond radiosondes). Additionally, we use the observations to assess the performance of a 1D cloud-topped boundary layer model (CLASS-L) focusing on simulations of mixed layer heights and cloud development. Future work will assess climate change impacts on cloud development using CLASS-L.
Multi-year observations at the Acadia Research Forest demonstrate that snowmelt and vegetation development play a crucial role for seasonal dynamics in land-atmosphere interactions with an observed sharp decrease in Bowen ratios following snowmelt. As a result, maximum mixed layer depths are observed in spring with suppressed mixed layer heights during the summer months, when available energy is mainly partitioned to latent heat flux. Mixed layer clouds develop frequently with a peak in the summer months with frequencies of >55% between June and August. Generally, latent heat fluxes during the summer months are energy limited. However, between August and October 2025, a severe late summer drought led to a substantial drop in soil moisture levels and a sharp increase in soil water potential. As a result, Bowen ratios increased, which was mainly driven by a decrease in latent heat fluxes. Cloud cover during the drought period decreased compared to the non-drought years. The decrease in cloud cover appears to be caused by the increase in lifted condensation levels exceeding the concurrent increase in mixed layer heights. Our results shed light on the effects of enhanced water limitation on cloud cover and boundary layer dynamics during droughts in the otherwise energy-limited humid forests of eastern Canada.

Wed

09:15–09:30

Turbulent exchange fluxes of water, energy, and carbon are simulated reasonably well by land surface models, but considerable random errors and systematic biases remain. For the purposes of water resources management (such as agricultural irrigation), vegetation growth modelling, weather prediction, and climate projection, improved simulation of these fluxes is important. While the representation of processes such as water infiltration into soils, vegetation drought response, and soil respiration in land surface models might be flawed, we focused in this work on the parameterization of processes. We used high-quality time series of latent heat flux, sensible heat flux, and net ecosystem exchange (NEE) from a large number of eddy covariance stations across the European continent to estimate sensitive parameters related to photosynthesis, stomatal control, plant root hydraulics, and (soil) hydrology. In addition, we considered measurements of soil moisture content and, for some sites, sap flow measurements. We estimated sensitive parameters of the Encore Community Land Model (eCLM), a fork of CLM5.0 with its own developments. The parameters were estimated for both the case in which the leaf area index was prescribed using remote sensing information and the case in which vegetation states were prognostically simulated. The parameters were estimated with an iterative ensemble smoother, based on multiple years of past time series, and evaluated using independent measurements over multiple years.

The results show that turbulent exchange fluxes remain difficult to constrain with measurements. Soil moisture measurements have a limited impact on improving the accuracy of exchange flux simulations, except under dry summer conditions. While the assimilation of latent heat flux and net ecosystem exchange measurements has a greater impact and improves the accuracy of turbulent exchange flux modelling also for independent verification periods, the root mean square error reduction is, on average across all sites, smaller than 10% for all simulation scenarios (prescribed LAI and dynamic vegetation states). We argue that the main limitations to further improvement are (i) the random error component of the measurement data themselves and (ii) model structural errors and process representation, whereas the model–data fusion methodology probably has a smaller impact. Sap flow measurements, on the other hand, provide important information on the temporal patterns of transpiration, but the absolute values show a systematic bias. The estimated soil and vegetation parameters were also evaluated at the continental scale and showed a pronounced impact on the spatial patterns of simulated latent and sensible heat fluxes, with values close to those of the GLEAM product.

Thu

09:00–09:15

Urban areas pose unique modelling challenges. The urban micro-climate is characterised by the complex interplay between urban form (land cover, surface/facet materials, roughness element morphology and arrangement, e.g., building geometry, street networks, urban vegetation), function (land- and building use, infrastructure and systems) and human activities (e.g., energy consumption in buildings, transportation). All these aspects have strong spatial (horizontal and vertical) and (in part) temporal variability at both intra- and inter-city levels, making high demands on urban climate and land-surface models with respect to process representation, nature of input data and model configurations.

High-resolution (O(100 m)) urban modelling systems need to consider the multi-scale nature of process interactions from the people scale, over buildings and neighbourhoods to the city-region scale. Human occupancy and activities, for example, determine indoor heat gains, which impact buildings’ heat exchange with the ambient environment in terms of anthropogenic heat emissions, net heat storage, convective and radiative heat fluxes. These directly impact, and are impacted by, the neighbourhood’s surface energy balance. Building and vegetation density and their respective height variability within a neighbourhood, for example, affect sky, wall and ground view factors, which, together with thermal/radiative facet and surface material properties, determine the neighbourhood’s short- and longwave radiative exchange with the urban atmosphere.

We use the land-surface model SUEWS with the Spartacus-Surface radiation model to represent the complexity of the urban energy balance from neighbourhood to building scale, with vertically resolved radiative exchange processes and atmospheric state variables in the roughness sublayer. Coupled to SUEWS, the building energy model STEBBS provides building net heat storage, facet surface temperatures (roofs, walls) and indoor air temperatures. Dynamics of anthropogenic heat emissions are captured by the agent-based model DAVE that represents human activities in indoor/outdoor microenvironments and associated energy consumption. This provides dynamic changes of building occupancy and activity levels and associated internal heat gains in STEBBS, which impact the magnitude and timing of anthropogenic building heat emissions. Based on the example of Greater London, UK, the presentation will characterise the inter- and intra-neighbourhood urban surface heterogeneity with respect to physical and socio-economic markers and discuss the impact based on multi-scale modelling results with the SUEWS modelling system.

Tue

16:30–16:45

Hydrological systems are undergoing rapid change under increasing greenhouse gas concentrations, yet substantial uncertainty remains regarding how precipitation is partitioned into blue (runoff) and green (transpiration) water flows across regions (blue-green water partitioning). This partitioning arises from interactions among climatic conditions, land surface characteristics, and vegetation dynamics that are altered by warming and increasing carbon dioxide concentrations. However, the magnitude, drivers, and management relevance of future shifts in blue and green water shares remain uncertain. Here we investigate how anthropogenic forcing alters blue-green water partitioning from global to regional scales using Coupled Model Intercomparison Project Phase 6 (CMIP6) simulations.

We analyse historical simulations and two contrasting future pathways (SSP1-2.6 and SSP3-7.0). Blue-green water partitioning is quantified from monthly precipitation, runoff, and transpiration and aggregated to multi-decadal means to emphasize robust signals beyond interannual variability. To identify the main factors shaping changes in blue-green water partitioning, we apply statistical learning methods that relate partitioning shifts to candidate controls representing precipitation characteristics, atmospheric demand, soil moisture conditions, and vegetation functioning.

We first characterise global changes and show that increases in extreme five-day precipitation are a primary driver of partitioning shifts, favouring larger blue water shares. This effect is largely independent of mean precipitation change and emerges under both drying and wetting conditions, indicating that changes in event intensity can reshape runoff and transpiration even where annual precipitation trends differ. We then contrast these global relationships with regional responses across selected large scale IPCC reference regions, highlighting where partitioning shifts align with precipitation intensity changes and where other candidate controls dominate. Overall, this global to regional perspective provides process based context for scenario driven changes in blue water availability and ecosystem relevant green water flows, supporting climate impact interpretation and climate service applications.

Mon & Tue

11:00–12:30

Although equation discovery has gained traction in other fields, it remains largely underexplored in land surface modelling, where it offers potential to reveal alternative formulations that may complement or improve existing representations of land-atmosphere fluxes.
This presentation introduces an application of an equation discovery framework, an emerging class of machine learning (ML) methods that identify interpretable functional relationships from data within a predefined space of candidate functions. In particular, we use the Sparse Identification of Nonlinear Dynamics (SINDy) approach to select a parsimonious combination of physically inspired candidate terms, derived from meteorological forcings measured at FLUXNET sites, and infer an interpretable functional equation for sensible heat flux.
Sensible heat flux remains challenging to represent robustly in land surface models. The talk will present both the capabilities and limitations of the ML-based equation discovery approach, drawing on results from multiple sites and examining the physical interpretability, robustness, and generalisability of the derived formulations, and their potential to inform improvements in land surface model representations of sensible heat flux.

Mon

14:30–14:45

Evapotranspiration (ET) regulates the exchange of water and energy between the land surface and the atmosphere and plays a central role in climate and hydrological processes. Soil moisture (SM) variability has long been recognized as a key control on ET, yet changes in SM do not always translate into changes in ET. This nonlinear behavior reflects the existence of distinct ET regimes, under which ET exhibits differing sensitivities to SM depending on whether surface conditions are water-limited or energy-limited. Beyond SM, ET is influenced by a range of forcings, including anthropogenic greenhouse forcing and the associated increase in atmospheric water demand, aerosol–radiative interactions, land-use and land-cover change, etc. These drivers can alter ET either by modifying SM states or by directly affecting surface energy availability and atmospheric demand.

A persistent challenge is to determine whether observed changes in ET induced by these forcings primarily operate through SM-driven pathways or through non-SM mechanisms. Progress has been hindered by the limited availability of long-term observational records and, critically, by the lack of an analytical framework that explicitly accounts for ET regime dependence and its inherent nonlinearity. Here, we introduce an empirical framework that partitions changes in evapotranspiration into soil-moisture-driven and non-soil-moisture-driven pathways. This partitioning framework bypasses explicit SM-regime identification while explicitly accounting for the nonlinear dependence of ET on SM. It requires daily SM and ET data and is applicable to a wide range of research questions.

This approach has been applied to quantify historical regional trends in ET across the globe using observation-constrained datasets, including GLEAM4 and ERA5. The results show that significant regional ET changes over the past few decades can be largely attributed to the SM pathway. Applying the same analysis to CMIP6, AMIP6, and the MIROC large ensemble indicates that this SM-driven pathway is substantially underestimated in Earth system models.

This framework has also been applied to 1pctCO₂ simulations to examine how increasing CO₂ affects ET changes. Motivated by the hypothesis that ET changes under global warming are increasingly attributable to non-SM pathways, driven by enhanced atmospheric water demand, we analyze 100 years of daily SM–ET data, applying the partitioning framework within a moving 10-year window. Trends in the SM-driven and non-SM-driven pathways are then estimated. The results reveal enhanced decadal variability in both pathways, without a clear increase in the dominance of either pathway across most of the globe. In addition, pathways transient responses to CO2 increasing substantial cross-Earth-system-model spread.

As another ongoing effort, this approach is being used to investigate the seasonality of ET using GLEAM4 and ERA5. Preliminary analysis over southwestern United States shows that seasonality of ET is largely governed by the SM pathway, while the non-SM pathway becomes important during summer. However, the relative dominance of these pathways in shaping seasonality differs between GELAM4 and ERA5. Additional observation-constrained datasets, including FLUXCOM and X-BASE, will be incorporated, and the analysis will be extended globally to characterize regional contrasts in ET seasonality and to systematically compare the roles of SM and non-SM pathways across climate regions.

Tue

16:15–16:30

Irrigation represents a major anthropogenic modification of land surface conditions and can substantially alter surface energy and water balances, thereby influencing land–-atmosphere interactions and near-surface hydroclimatic conditions. Accurate representation of irrigation in land surface models remains challenging due to uncertainties in soil moisture availability, vegetation state, and their combined effects on evapotranspiration and surface energy partitioning. This study examines the impacts of irrigation on land surface processes using the Noah-MP land surface model within the NASA Land Information System (LIS), with a focus on surface turbulent heat fluxes and near-surface atmospheric variables, including 2-m air temperature (T2m), specific humidity (Q2m), and vapor pressure deficit (VPD).

An offline (i.e. uncoupled) data assimilation (DA) is run over California to jointly update surface soil moisture (SSM) and leaf area index (LAI) using observations from the Soil Moisture Active Passive (SMAP) mission and the Copernicus Global Land Service (CGLS) respectively. The classical bias-correction methods used for land surface DA are redesigned to preserve the irrigation signals in the SSM satellite observations and feed those into the simulations through DA. On the other hand, the LAI assimilation is bias blind. By simultaneously constraining soil and vegetation states, the DA system aims to reduce uncertainties in latent and sensible heat fluxes and to improve the representation of irrigation-driven variations in the Bowen ratio.

Comparisons between joint DA experiments and control simulations without DA are used to isolate the role of improved soil moisture and vegetation representation relative to control model configurations, in terms of T2m, Q2m, and VPD variations, as well as turbulent fluxes. The latter can drive land-atmosphere interactions in a coupled system in future research. In general, this study presents a physically consistent framework for investigating the impacts of irrigation on surface processes and near-surface hydroclimate. The resulting land surface representation provides a more realistic foundation for land-driven feedbacks in atmospheric modeling applications, highlighting the potential importance of surface process realism for managed agricultural landscapes.

Mon & Tue

11:00–12:30

The terrestrial ecosystem carbon residence time (τe) is a key property of the global carbon cycle by integrating turnover processes throughout vegetation and soil. However, reliable modeling of τe remains challenging due to inconsistent τe trends among land surface models (LSMs) during the historical period (1901–2021). Here, the main causes of the inter-model spread in τe trends during the historical period are examined by using simulations across 15 LSMs in the Trends and drivers of terrestrial sources and sinks of carbon dioxide (TRENDY) project. We quantified the relative contributions from baseline residence time (τb) and three environmental factors: CO2 concentration, climate forcing, and land-use change (LUC). Results indicate that the inter-model variance in τe trends is mostly explained by τb and LUC, 47% and 36%, respectively. Further decomposition into carbon pools reflects that the τb is dominated by soil pools while the LUC response is concentrated on vegetation pools. These findings imply that the main causes of the large inter-model spread are the different representations of soil carbon pools (e.g., structure and parameters) and transitions between natural vegetation and anthropogenic use from LUC. Therefore, achieving more consistent and reliable assessments of changes in the global carbon cycle requires standardized model representations of LUC as well as improved observational constraints on τb.

Acknowledgment: This work is supported by the Korea Institute of Science and Technology (KIST) research programs (26Z9073).

Mon & Tue

11:00–12:30

The terrestrial carbon and water cycles are interconnected through stomatal control, which regulates exchanges between ecosystems and the atmosphere. Water use efficiency (WUE) is a metric that reflects both the competing mechanisms of carbon uptake and water loss in plants. Hence, it is a useful indicator of the sustainable growth of natural ecosystems in a changing climate. Monitoring ecosystem-atmosphere exchanges in India is vital as it houses diverse ecosystems; a sustainable management of land resources depends on understanding these fluxes.
In this context, we studied interannual and seasonal changes in WUE at a subtropical broadleaf deciduous forest (Kaziranga National Park or KNP) in northeast India, which is the most forested region in the country and their hydrometeorological and physiological drivers; we analysed these over two years each containing four seasons: pre-monsoon, monsoon, post-monsoon, and winter, using Eddy Covariance measurements. This study investigated five controlling factors: soil moisture (SM), solar radiation (Rg), air temperature (Tair), vapor pressure deficit (VPD), and Ecosystem respiration (Reco). Partial correlation tests and structural equation modelling (SEM) were used to identify their influence on WUE. Also, the WUE values were compared with the LPJ GUESS model output simulated for the same region.
WUE exhibited a comparable seasonal pattern during the study period, having the highest value recorded in the pre-monsoon 2016 (2.14 g C kg-1 H2O). Rg and VPD were identified as the main factors controlling WUE variability. Onwards, WUE gradually declines through the monsoon due to monsoon-related changes in atmospheric and soil water availability. Statistical analysis indicates that Rg and VPD are the primary factors controlling WUE variability, suggesting that humid KNP is an energy-driven ecosystem rather than a moisture-driven one. The annual WUE derived from the slope of daily gross primary productivity (GPP) and evapotranspiration (ET) for 2016 and 2018 were 1.84 and 1.99 g C kg⁻¹ H₂O, respectively, while the LPJ-GUESS-simulated annual WUE for the recent three decades (1994–2023) was 2.57 g C kg⁻¹ H₂O. The LPJ GUESS output shows a higher simulated WUE than the calculated WUE. This may be because global PFTs don’t accurately represent Indian subtropical forests and instead rely on observations to model them. Our findings will provide insights into managing forest ecosystems in India and other subtropical regions under a warming climate, thereby improving our understanding of how their carbon and water cycles will evolve.
Keywords: Ecohydrology, Eddy Covariance, India, LPJ GUESS, Subtropical Forest, Water Use Efficiency

Mon & Tue

11:00–12:30

Variability in latent and sensible heat fluxes at the surface and their vertical transport within the atmospheric surface layer is key to understanding temperature and moisture fluctuations in the atmosphere. In this study, we quantify how large-scale flow situations and stability conditions modulate daytime surface heat fluxes from day to day under consideration of the local hydrometeorological conditions and the phenological stage. By linking synoptic flow, local land-atmosphere system conditions and surface heat flux variability, we aim to enhance understanding of sub-seasonal temporal variability in land-atmosphere coupling.

We analyze daily ECMWF forecasts (initialized at 00 UTC) for the growing seasons (March–September) between 2017 and 2025 at 0.09° resolution. The analyses focus on Southern Germany and the Czech Republic, which are climatologically and geomorphologically homogeneous (Metzger et al. 2005, 2012). All days were clustered according to their dominant atmospheric flow direction, cyclonicity and the occurrence of instability throughout the day between the surface and 500 m AGL, as well as the phenological stage to account for seasonality.

A descriptive assessment of the meteorological and hydrological conditions within the clusters shows that their characteristics remain seasonally relatively stable in relation to each other. Clusters classified as “cool” or “dry” therefore retain those characteristics compared to other clusters throughout the year. The most frequent dominant flow direction is southwest, which brings warm, relatively moist air with moderate wind speeds into the study area. The least frequent dominant flow direction is southeast, which transports comparatively hot and dry continental air in the study region. Differences in the cluster characteristics were tested for statistical significance, and clusters that were not statistically significant were merged. This reduced classification forms the basis for a joint assessment of how synoptic flow regimes and local atmospheric and hydrological conditions influence daytime latent and sensible heat fluxes in dependence of phenological stages.

Mon & Tue

11:00–12:30

Land surface models and Earth system models often rely on pedotransfer functions (PTFs) to map static soil attributes to hydraulic parameters, yet these mappings are typically derived from laboratory data and applied independently of the coupled land-atmosphere system they are meant to support. This disconnect often leads to parameter sets that are locally plausible but degrade simulated water, carbon, and energy exchange when embedded in land surface or Earth system models. Here we propose a process-guided machine learning framework that learns soil PTFs directly within a differentiable soil-plant-atmosphere continuum (SPAC) model, such that soil hydraulic parameters are constrained by their role in coupled system dynamics rather than in isolation. Hydraulic parameters defining the soil water retention curve are predicted by a neural network as functions of static soil attributes and selected long-term environmental covariates that reflect large-scale controls on soil hydraulic behavior. The fully differentiable SPAC model is driven by meteorological forcing to simulate evapotranspiration (ET) and gross primary productivity (GPP), and eddy covariance observations of ET and GPP are used to constrain the soil PTF through gradient-based optimization across the coupled water and carbon cycles. Using flux tower sites across Europe, we evaluate parameter identifiability, transferability across sites, and consistency with expected soil hydraulic behavior, and benchmark the learned PTFs against conventional formulations. The results indicate that learning PTFs under coupled land-atmosphere constraints can yield more coordinated and flux consistent parameterizations than PTFs derived outside the coupled modeling framework.

Tue

15:30–15:45

Agricultural land management, such as cover cropping, no-till management, or agroforestry, has the potential to substantially modify land–atmosphere interactions, via altering energy and water fluxes, yet many such practices are still poorly represented in Earth system models. Cover cropping is a widely promoted sustainable management practice that alters vegetation characteristics, soil moisture dynamics, and surface energy exchanges, with potential impacts on regional climate and temperature extremes. This study introduces an idealized representation of cover cropping into the coupled climate model ICON to examine its influence on surface fluxes, boundary layer processes, and surface temperature extremes. Cover Crops are implemented via modifications of the leaf area index that mimic a higher vegetation cover after the main crop is harvested. We control for internal-climate variability-induced noise by running an ensemble of perturbed initial condition simulations.

We analyse regional and seasonal changes in near-surface climate, particularly extreme heat events in response to cover cropping in Europe. First results indicate spatially diverse responses: mean surface temperatures tend to decrease in western Europe while they increase in eastern Europe. Both cold and hot extreme temperatures are attenuated, but the mechanisms that drive these changes are region-dependent. These contrasting signals highlight the complex interplay between the regional characteristics in evapotranspiration, radiation, and atmospheric feedbacks with agricultural practices.

In addition to local impacts, the simulations reveal climate effects in remote regions where no land management changes are applied, pointing to possible large-scale adjustments in atmospheric circulation. This emphasizes the importance of accounting for land surface heterogeneity and anthropogenic influences in global modeling frameworks.

By linking land management practices with process-oriented diagnostics in an Earth system model, this work contributes to improved understanding and representation of managed landscapes. The study demonstrates the relevance of agricultural practices for land–atmosphere coupling and provides a foundation for further developments in parameterizations and model evaluation strategies.

Mon

16:45–17:00

Earth system models (ESMs) and their land surface schemes are subject to biases arising from the aggregation of land surface heterogeneity, in particular, with respect to turbulent heat fluxes. At coarse resolutions, ESMs like ICON use simplified representations of the land-atmosphere coupling with fully parameterised schemes to account for convection and turbulence, hence neglecting the non-linear effects of subgrid-scale heterogeneity.
Such a description is used by ICON in its Numerical Weather Prediction (NWP) configuration, typically operating on the kilometre scales. In contrast, ICON’s Large-Eddy Simulation (LES) configuration explicitly resolves large turbulent eddies in the boundary layer, convective dynamics, and surface heterogeneity at decameter scales, reducing the influence of subgrid-scale parameterisations on model performance. This study presents a preliminary comparison of the representation of the turbulent fluxes within these two configurations of ICON. By using the same land surface scheme TERRA across both configurations, we isolate the impact of turbulence treatment and model resolution on the simulated surface energy balance. Our analysis focuses on the effect of switching from 1D- parameterized vertical transport to 3D-parameterised transport and explicitly resolved turbulent structures on the evolution and the diurnal cycle of surface fluxes. We investigate how resolved structural features of the atmospheric boundary layer, such as mesoscale circulations, boundary layer rolls, and cellular convection, differently impact the partitioning and variability of surface fluxes.
In addition to the comparative study, we evaluate model performance by validating these simulations against observational benchmarks from the Lindenberg Meteorological Observatory (MOL-RAO) and the Land-Atmosphere Feedback Observatory (LAFO) at the University of Hohenheim, making use of the comprehensive field measurements made as a part of the Land-Atmosphere Feedback Initiative during the summer of 2025. Thus, this work aims to identify the relevance of resolving land surface heterogeneity and atmospheric circulations for the partitioning and variability of surface fluxes and add to the discussion on the importance of scale-specific parameterisations in unified modelling systems such as ICON.

Mon

13:45–14:00

Understanding the relationships between soil moisture and latent heat flux is essential for identifying the complex mechanisms of land-atmosphere interactions. While many studies have performed linear regression analysis to evaluate these interactions, the actual Earth system exhibits nonlinear dependencies that cannot be fully explained by linear relationships. Despite its ability to detect nonlinearities, Mutual Information (MI) often struggles with non-Gaussian data due to specific estimator constraints. Its primary drawback is a lack of equitability, which causes inconsistent scoring across various relationship types. Unlike MI, the Maximal Information Coefficient (MIC) provides a robust, normalized score that accurately reflects the strength of a relationship regardless of whether it is linear, periodic, or follows a complex functional form. This study utilizes ERA5 reanalysis data from 1940 to 2025 to examine the land-atmosphere interactions over the globe. The total Normalized MI consists of independently derived linear and nonlinear contribution, while the MIC is decomposed into linear, nonlinear, and noise components. A comparison between MI and MIC reveals strong agreement in capturing linear components; however, MI consistently yields higher values for nonlinear components, resulting in low inter-method agreement. This discrepancy suggests that MI tends to overestimate nonlinear dependencies by failing to distinguish stochastic noise from true nonlinear signals. In contrast, MIC provides a more rigorous assessment of authentic nonlinearity by effectively isolating noise from the total information. By overcoming the intrinsic limitations of conventional MI, MIC emerges as a more robust and effective tool for identifying the complex nonlinearities inherent in land-atmosphere interaction research.

Mon & Tue

11:00–12:30

This study uses the regime shift index to detect the interdecadal shift of the summertime intraseasonal oscillation (ISO) in the East Asian monsoon region during July-September from 1979 to 2021. A regime shift index was found to specify three distinct epochs: 1979-1993, 1994-2004, and 2005-2021. The middle epoch yields a westward extension of the subtropical anticyclonic circulation and maximal intensity in the westerly phase of the northward-propagating ISO, and his is apparently different from other epochs. These circulation anomalies intensify southeasterly winds south of Japan, which block tropical cyclones (TCs) from east of this strong wind zone. The circulation anomalies therefore result in more contracented TC tracks between Taiwan and Japan as well as increased TC frequency and duration. The findings yield a concentration of TCs can lead to more rainfall in the westerly phase of the northward propagating ISO or the middle epoch. The present work shows that ISO variability itself, especially the connection between the subtropical anticyclone and ISO cyclonic anomalies, directly influence TC genesis, clustering, and intensity. The resulting changes in TC characteristics, particularly in frequency and intensity, could have significant implications for regional disaster preparedness and climate adaptation strategies. This highlights the critical need to consider the interdecadal ISO variability in long-term climate projections and TC prediction models, ensuring more accurate forecasts and informed decision-making in the face of evolving climate patterns. As we know that global warming is taking place, the present climate pattern change such as decadal variability of the ISO can also be part of the results of the climate change under global warming. This study could be a hint of representing changes in the future. In other words, the climate pattern change in the future might also undergo the decadal change as in the present study. These insights can be useful for improving TC rainfall prediction under changing climate patterns.

Wed & Thu

11:00–12:30

The Arctic is warming at a rate three to four times faster than the global average, threatening to destabilize its permafrost carbon reservoir, which stores about 60% of global soil carbon—an amount three times as large as currently contained in the atmosphere. Accurate estimation of Arctic greenhouse gas (GHG) fluxes is crucial for understanding the feedback processes between the permafrost carbon cycle and climate, as these processes have the potential to transform the region from a carbon sink into a significant carbon source and amplify global climate change.
Quantifying GHG fluxes using conventional eddy covariance (EC) techniques is particularly challenging under stable stratification, where turbulent mixing is suppressed. This study investigates nighttime GHG transport dynamics in the Arctic’s stably stratified boundary layer using Large Eddy Simulation (LES) within the EULAG framework. Site-specific data are incorporated and sophisticated sub-grid scale turbulence models are employed to simulate stable stratification induced by surface cooling.
We analyze the influence of wind speed and sensible heat flux on GHG storage, with the goal of interpreting nighttime fluxes from their measurements. These results provide new insights into stable boundary-layer dynamics and land–atmosphere interactions in the Arctic, and contribute to improving Earth System Model (ESM) representations of Arctic GHG emissions and their role in global climate change.

Mon & Tue

11:00–12:30

Land surface models (LSMs) are a key component of Earth system models, providing the interface through which soil, vegetation, and atmosphere exchange water, energy, and carbon. The realism of these exchanges depends critically on how soil hydraulic and thermal processes are represented. In most contemporary LSMs, soil hydraulic properties and soil thermal properties are parameterised using separate, largely empirical formulations. Their interaction is typically mediated only through soil moisture content, despite growing observational and theoretical evidence that the parameters governing soil water retention and soil thermal behaviour are physically linked. As a result, existing model formulations may inadequately capture coupled soil moisture–temperature dynamics, particularly under dry soil conditions where vapor transport, adsorption processes and strong thermal gradients become important.
In this study, we develop and implement a unified hydro-thermal framework within the ECMWF land surface model ecLand (that sits within the Integrated Forecasting system, IFS), with the aim of explicitly coupling soil hydraulic and thermal properties through shared physical parameters. The new framework is designed to provide a more internally consistent representation of soil moisture and heat transport, and to better capture land–atmosphere feedbacks across a range of hydro-climatic regimes.
The first element of this development concerns soil hydraulic properties. The widely used van Genuchten-Mualem (1980) soil water retention curve and hydraulic conductivity equation were replaced by formulations that explicitly distinguish between adsorbed and capillary water components, following approaches such as those by Peters–Durner–Iden (2024) and Weber (2019). These formulations offer a more physically based description of soil water retention and hydraulic conductivity, particularly at low water contents, where adsorption dominates and traditional van Genuchten-type approaches are known to perform poorly.
The second element addresses soil thermal properties. In the operational ecLand model, soil thermal conductivity is computed using the Peters–Lidard et al. (1998) formulation, which depends on soil moisture but is otherwise disconnected from the hydraulic parameterisation. Here, we replaced this formulation with an approach that directly links soil thermal conductivity to the parameters of the soil water retention curve (building on e.g., Lu & McCartney, 2024). This modification establishes an explicit and physically motivated coupling between soil hydraulic and thermal behaviour, allowing changes in the soil water retention characteristics to consistently influence heat transport.
We first assessed the implications of the unified hydro-thermal framework using a suite of controlled sensitivity experiments. These experiments were designed to isolate the response of simulated soil moisture, soil temperature, and surface energy fluxes to the capillary-only and capillary-adsorption approaches, considering the effects of soil texture, soil depth and vertical discretisation, and climatic forcing. Emphasis was placed on dry and semi-arid conditions, where strong coupling between hydraulic and thermal processes is expected. The sensitivity experiments demonstrated that the unified framework substantially alters moisture–temperature interactions, modifies heat propagation within the soil profile, and leads to pronounced changes in surface energy partitioning.
Next, the performance of the new framework was evaluated against in situ observations at selected field sites. Simulations using the original and modified ecLand configurations were compared with observed soil moisture and soil temperature at multiple depths, as well as latent and sensible heat fluxes. The results revealed systematic differences between the two model configurations, with the unified hydro-thermal framework producing notably different surface fluxes, especially for coarse-textured soils. For loamy sand soils, differences in latent and sensible heat fluxes reached values up to approximately 40 W m⁻² corresponding to relative changes of approximately 65% and 25%, respectively, compared to the original formulation, highlighting the importance of a consistent hydro-thermal treatment for accurately simulating land–atmosphere exchanges.
Overall, this work demonstrates that explicitly linking soil hydraulic and thermal properties within a land surface model can significantly influence simulated soil states and surface fluxes, particularly under dry conditions. The results highlight the potential of this physically consistent framework in offline, point-scale experiments. If these benefits are shown to be robust, the approach will be further tested in global distributed ecLand simulations to examine its behaviour at larger spatial scales. Subject to satisfactory performance at site and global scales, the framework may then be evaluated in coupled IFS simulations.

Mon

14:15–14:30

Soil structure is nearly as important as soil texture in determining the soil hydraulic properties at the core scale. It also mediates a large part of the anthropogenic impact on soil transport properties, resulting from land management or from land cover change. At ecosystem scale, soil structure — and particularly soil macroporosity — strongly affects drainage and the runoff–infiltration partition (Alaoui et al., 2018; Bonetti et al., 2021; Fatichi et al., 2020). At the 100 km scale typical for land surface models, however, the effects of soil macroporosity appear suppressed (Fatichi et al., 2020). This is most likely due to the spreading of precipitation over large model grid cells, which decreases precipitation intensity and results in soils rarely approaching saturation, where the effect of macroporosity on soil hydraulic properties is the strongest. Here, we show that accounting for the subgrid variabilities of soil hydraulic properties or of precipitation counteracts the scale-induced suppression of soil-structural effects in a land surface model. We show this by introducing a physically-motivated representation of soil macroporosity in the ORCHIDEE land surface model and testing to what extent its effects are mediated by subgrid parametrizations. Specifically, we investigate the role of the temporal redistribution of precipitation within the forcing time step and the subgrid variability of soil hydraulic conductivity in the infiltration routine. We find that two-thirds of the effect of soil macroporosity on the terrestrial water balance in ORCHIDEE are mediated by these subgrid parametrizations. With the subgrid parametrizations active, the effect of soil macroporosity on runoff in ORCHIDEE — an average decrease of 44%, with infiltration increasing by a similar fraction — is much larger than previously reported for a similar soil structure parametrization in another land surface model. The additional infiltration increases the total soil moisture by 8% on average, which in turn increases plant transpiration in the more productive half of the grid cells by 28%. The effect of macroporosity on soil moisture is large enough to cause statistically significant regional cooling of up to 2 K when ORCHIDEE is coupled to an atmospheric model. These results show that soil structure can play an important role at large scales, including in the land-atmosphere interaction, but capturing it requires representing the spatial or temporal subgrid variabilities of soil hydraulic properties or of precipitation.

Alaoui, A., Rogger, M., Peth, S., & Blöschl, G. (2018). Does soil compaction increase floods? A review. Journal of Hydrology, 557, 631–642. https://doi.org/10.1016/j.jhydrol.2017.12.052

Bonetti, S., Wei, Z., & Or, D. (2021). A framework for quantifying hydrologic effects of soil structure across scales. Communications Earth & Environment, 2(1), 1–10. https://doi.org/10.1038/s43247-021-00180-0

Fatichi, S., Or, D., Walko, R., Vereecken, H., Young, M. H., Ghezzehei, T. A., Hengl, T., Kollet, S., Agam, N., & Avissar, R. (2020). Soil structure is an important omission in Earth system models. Nature Communications, 11(1), 522. https://doi.org/10.1038/s41467-020-14411-z

Mon & Tue

11:00–12:30

As an essential climate variable, land surface temperature (LST) plays a crucial role in linking the thermodynamics of the Earth’s land surface with the dynamics of the surface proximate atmosphere. LST is a key variable for understanding land-atmosphere interactions, energy balance, and ecosystem dynamics. High-resolution LST information is important for resolving small-scale thermal heterogeneity in agricultural landscapes, forests, and urban environments and is required for applications such as Earth system modelling. A major limitation of optical LST satellite observations is their strong sensitivity to cloud cover, resulting in spatial and temporal data gaps. In contrast, microwave-based LST products are nearly cloud-independent but available at coarser spatial resolutions. This trade-off between spatial detail and atmospheric robustness, as well as between spatial and temporal resolution in general, limits the generation of a consistent all-sky LST product with high spatiotemporal resolution. To address this challenge, the objective of this study is to use a machine learning-based random forest framework to generate an hourly, cloud-free LST product, aiming for sub-kilometer spatial resolution. This study focuses on a regional test domain in Central Europe covering the southwest part of Germany, characterized by heterogeneous land cover including agricultural areas, forests, and urban settlements. The study period extends from 2022 to 2024.
The approach applied here aims to combine multiple satellite-derived datasets. Geostationary LST satellite data (CGLOPS1) with high temporal (1 h) but low spatial resolution (5 km) are integrated as well as low earth orbit satellite LST (e.g. ECOSTRESS) with higher spatial (70 to 1000 m) but lower temporal resolution (1 to 16 days). The coarse spatial resolution (10 km, 0.5 day) microwave LST data (AMSR2) are additionally incorporated to account for cloud-free conditions. Auxiliary predictors include, e.g. surface albedo (MODIS), leaf area index (MODIS), surface soil moisture (Sentinel-1), and a statistical daily cycle of LST (CGLOPS1). The model is trained to learn the relationship between the target and the multi-sensor predictor set. Model training and validation are restricted to clear-sky optical LST observations over land to ensure physically consistent reference data. An 80/20 random train-test split is applied, with 80 % of the pixels used for model fitting and 20 % reserved for independent validation. Missing predictor values are handled through median imputation combined with validity flags, allowing the model to retain information about data availability. After training, the model is applied to all land pixels, including those affected by cloud cover, to generate spatially continuous hourly LST data. The proposed framework is expected to substantially reduce cloud-induced data gaps while maintaining realistic spatial temperature gradients. Such a high resolution consistent all-sky, dataset is crucial as it enables improved monitoring and modeling of heat stress, agricultural drought, land-atmosphere feedbacks, and urban thermal dynamics at spatial scales relevant for local impact assessments.

Mon & Tue

11:00–12:30

Turbulent fluxes remain a major source of uncertainty in weather and climate models, particularly due to the inadequate representation of planetary boundary layer (PBL) processes over heterogeneous land surfaces and a scarcity of high-fidelity observational datasets required for model development. To address these critical gaps, we developed a fast-response meteorological measurement system integrated into uncrewed aircraft systems (UAS), with a primary focus on the measurement of turbulent energy fluxes.
The turbulent fluxes of sensible and latent heat govern the vertical energy distribution within the atmosphere and directly balance the surface energy budget. However, state-of-the-art measurement systems are not able to effectively capture turbulent energy fluxes due to their high spatial and temporal variability. Eddy covariance systems assume a homogeneous surface and provide only local measurements of limited footprint size, while remote sensing instruments lack the resolution required to capture fine-scale PBL structures. Solving this, UAS allow for high-resolution in-situ measurements simultaneously at multiple PBL locations.
We developed the custom sensor package PARASITE (Portable Aircraft Rucksack for Atmospheric Sensing and In-situ Turbulence Estimation), which is mounted on commercially available and automatically operating multi-rotor UAS. This system measures the 3D wind vector, temperature and humidity at high resolution. Field deployments in complex terrain demonstrated the system’s capability to capture the wind velocity vector up to 4 Hz, allowing for the calculation of turbulent momentum fluxes.
Capturing turbulent sensible and latent heat fluxes requires equally highly resolved measurements of tem- perature and humidity. To achieve this, we developed an enhanced sensor package, notably featuring a custom fast-response dew-point mirror hygrometer, which is mounted alongside commercial sensors on a stabilized boom to ensure undisturbed free-stream flow and precise horizontal alignment. Providing previously unattain- able measurements of turbulent fluxes, this system enables an analysis of key thermodynamic PBL processes and contributes critical data to improve PBL parameterizations and reduce uncertainties in atmospheric model- ing.

Mon & Tue

11:00–12:30

Recent European summers have been marked by persistent heatwaves and severe soil moisture deficits, raising questions about how land-atmosphere feedbacks modulate temperature responses to global warming. In this study, we use a nudged storyline framework to quantify the state-dependent temperature response during the summers of 2018-2022 under different global warming levels. Global storylines in which spectral nudging constrains the large-scale atmospheric circulation have been generated with the coupled AWI-CM1 model and dynamically downscaled over Europe using the regional ICON-CLM model. This setup enables a direct day-by-day comparison of the past weather development across different climates ranging from pre-industrial to +4 K, isolating thermodynamic and land surface feedbacks.

We analyze the relationship between evaporative fraction and soil moisture to identify energy-limited and moisture-limited evapotranspiration regimes, and examine how regime transitions influence the temperature response. To quantify this response, we introduce a metric termed warming amplification, defined as the change in daily maximum temperature per degree of global warming. Our results show that warming amplification varies substantially in space and time, with the strongest amplification occurring in late summer and in regions prone to soil moisture depletion. Furthermore, we find that the largest warming amplifications are associated with transitions from energy-limited to moisture-limited regimes. These transitions are characterized by pronounced soil drying and a strong reduction in evaporative fraction, which limits evaporative cooling and enhances sensible heating. In contrast, regions already in a moisture-limited regime under present-day conditions exhibit weaker soil moisture changes and more moderate amplification under future warming.

These findings demonstrate that soil moisture-evapotranspiration feedbacks play a central role in amplifying heat extremes under climate change. The results highlight the importance of accurately representing land surface processes in regional climate projections and provide process-based insight into why late-summer heatwaves are expected to intensify disproportionately in a warmer climate.

Mon

17:00–17:15

Flash droughts, or the rapid onset or intensification of drier-than-normal conditions, have a significant impact on the economy due to agricultural loss, water cycle anomalies, and decreased ecological productivity. These events are challenging to predict because, unlike most droughts, they evolve on a timescale of weeks instead of months. Previous studies indicate that rapid intensification of flash droughts relates to high evaporative demand and, in some cases, land-atmosphere interactions that accelerate soil drying. Irrigation and its associated changes in soil moisture and land cover properties can influence flash droughts by increasing bulk humidity, altering surface energy fluxes, and influencing planetary boundary properties. The importance of these mechanisms in flash drought evolution and predictability is a subject of active research. Here, we examine one of the largest and most damaging flash droughts of recent memory, the US Great Plains drought of 2012, in controlled numerical experiments using the NASA Unified Weather Research and Forecasting (NU-WRF) regional coupled model. We focus specifically on the role that irrigation might play in the regional land-atmosphere feedbacks during drought onset, testing the hypothesis that large-scale irrigation acts as a buffer against flash drought development. Understanding this relationship is crucial for developing drought mitigation strategies and improving the representation of land management in weather and climate models.

Mon & Tue

11:00–12:30

As Earth System Models (ESMs) move toward higher spatial resolutions, the atmospheric and land surface components face contrasting challenges. On the atmospheric side, resolving small-scale processes reduces the need for parameterizations and substantially improves the representation of convection and precipitation. In contrast, the land surface model (LSM) requires more parameterizations at higher resolution, as finer spatial scales reveal heterogeneities that must now be explicitly represented. This shift has driven land surface modelers to develop new features that capture these complexities. Among the underrepresented components in LSMs is groundwater, long modeled in detail by hydrogeologists but rarely integrated into climate-scale simulations. Inadequate soil moisture and groundwater representations affect evaporation, land–atmosphere coupling, and ultimately the atmospheric supply of water as precipitation. These hydrological biases therefore influence not only local surface conditions but also remote moisture transport and recycling. In this work, we improve the representation of subsurface hydrology in the ICON-Land (JSBACH) land surface model, coupled to the ICON atmospheric model. We introduce additional soil layers, implement lateral groundwater flow between grid cells, and connect shallow groundwater to the river network. We evaluate the new developments using standalone kilometer-scale JSBACH simulations over the 1989–2014 period against observations of summer diurnal LST amplitude in the Pyrenees (Spain and France). We then assess their impact on atmospheric variables, specifically 2-m temperature and precipitation, within ICON simulations at 3-km grid spacing over Europe.

Mon

14:30–14:45

Built environments exhibit strong spatial heterogeneity in land use and land cover, producing complex land–atmosphere interactions across scales ranging from meters to kilometers. While the urban heat island is well documented at city scales, the processes governing microclimate variability within individual neighborhoods remain poorly constrained, largely due to the scarcity of spatially dense observations of both meteorological and land-surface conditions. This observational gap limits our ability to manage building energy demand, and optimize urban design.

To address this gap, we conducted the Encanto Neighborhood Experiments in Phoenix, Arizona, as part of the U.S. Department of Energy Southwest Urban Integrated Field Laboratory project. The experiments targeted a 3 × 3 km residential neighborhood surrounding two large, irrigated parks within a semi-arid urban environment. The deployment was designed to support land–atmosphere process studies and model evaluation including the collection of dynamic and static boundary conditions data for the domain. Instrumentation included eddy-covariance flux towers, fixed meteorological stations, lidars, balloon-borne sondes, a temperature profiler, and dense mobile temperature measurements, producing four-dimensional meteorological observations throughout the neighborhood.

We developed a spatiotemporal reconstruction technique to separate temporal evolution from spatial structure in mobile observations allowing us to generate 1-minute air and road-surface temperature fields at 100 × 100 m resolution across the domain. Observations reveal park-induced air-temperature cooling of 2–3 °C relative to the urban background, persisting overnight between 19:00 and 03:00 local time. Under light winds, cooling was systematically observed downwind and strongest near the park edge, suggesting contributions from evaporative cooling and cool-air advection processes. The resulting cool plume extended approximately 180 m downwind, much shorter than the park dimension (~900 m), and remained confined within a shallow stable layer less than 50 m deep.

Efforts are underway to reproduce these conditions using PALM-4U and urbanMicroclimateFOAM. Once validated, such micrometeorological models will enable exploration of alternative scenarios—including irrigation strategies and park configurations—providing insights for urban planning. We invite additional modeling groups to participate in coordinated simulations of this well-constrained urban domain.

Thu

09:30–09:45

Heat extremes are intensifying under global warming, yet major uncertainty remains regarding when, where, and through which physical pathways land–atmosphere interactions actively regulate their life cycle. While large-scale atmospheric circulation is widely recognized as a primary driver of heatwaves, the dynamic role of soil moisture in modulating the extreme heat remains insufficiently resolved. Here, we develop a phase-resolved diagnostic framework that explicitly links soil moisture–temperature coupling regimes to the full life cycle of heat extremes across hydroclimatically distinct global hotspots, including the United States, the African Sahel, Central India, the Brazilian Caatinga, the African Miombo, and Australia. We introduce a globally quantified temperature sensitivity threshold (TST) that delineates the transition into thermally decoupled conditions, and define a thermal buffer (TB) bounded by the TST and a critical threshold separating moisture- and radiation-limited regimes. Building on this structure, we propose a novel Thermal Buffering Index (TBI) that integrates buffering capacity and soil moisture persistence to diagnose the effectiveness and stability of terrestrial water–energy coupling. This unified regime framework enables mechanistic separation of upward (soil moisture–controlled) and downward (atmosphere-driven) coupling pathways across antecedent, peak, and recovery phases of heat extremes. Our results reveal a fundamental hydroclimatic contrast. Transitional climates exhibit pronounced seasonal regime transitions from persistent thermally decoupled spring states to mixed coupling in summer and buffered autumn conditions, where soil moisture most strongly regulates the evolution and decay of heat extremes. In contrast, arid regions remain locked in chronically decoupled states, with extremes governed primarily by atmospheric forcing. Critically, soil moisture exerts its strongest control not at event initiation but during peak evolution and recovery, mediated by its residence within thermally active regimes. By resolving regime structure, thresholds, and persistence, this study advances a process-based understanding of terrestrial feedbacks on extreme heat and provides a transferable diagnostic pathway to improve land–atmosphere coupling representation in Earth system models, thereby strengthening seasonal forecasting and future climate projections.

Wed & Thu

11:00–12:30

East Asia has experienced warming substantially faster than the global average over recent decades, culminating in a record-breaking heatwave during the 2025 summer season. However, the mechanisms sustaining and amplifying this unprecedented event, particularly the role of air-sea interactions, remain insufficiently understood. This study investigates the physical mechanisms underlying the 2025 East Asian heatwave. Under a persistent negative Pacific Decadal Oscillation (PDO) phase, the Kuroshio-Oyashio Extension (KOE) sea surface temperature (SST) front shifted poleward relative to its climatological position, generating pronounced warm SST anomalies over the northwestern Pacific. This anomalous oceanic state relocated the maximum meridional temperature gradient poleward, displacing the East Asian jet stream northward and promoting the expansion of the Western North Pacific Subtropical High (WNPSH) and the Tibetan High. The associated anticyclonic circulation enhanced downward shortwave radiation, reinforcing and intensifying surface warming through positive SST-radiation feedbacks. A moving correlation analysis between an air-sea coupling strength index and 2-m temperature reveals statistically significant relationships over East Asia in recent decades, suggesting that air-sea interactions have become an increasingly dominant driver of regional heatwave variability. These findings provide insight into how anomalous oceanic states contribute to the development and persistence of East Asian heatwaves under continued global warming.

Mon & Tue

11:00–12:30

As the Canadian climate warms, the frequency and severity of wildfires and droughts are rising, underscoring the critical need for accurate prediction of soil moisture across Canada. The Soil, Vegetation, and Snow (SVS) land surface model used at Environment and Climate Change Canada (ECCC) for operational weather and hydrological forecasting has important limitations. It relies on a simplified force-restore scheme for soil temperature and lacks representation of key components, such as soil organic matter and forest litter. A new scheme developed at ECCC, SVS2, addresses these limitations. SVS2 has shown to improve snow process representation in forested regions and adopts a more physically-based, prognostic approach to soil heat and mass transfer. Building on these advances, this work introduces the explicit representation of soil organic matter and forest litter in SVS2, aiming to further refine the simulation of soil moisture and temperature profiles. Model enhancements were first evaluated through site-level comparisons to high-quality field observations at three different sites across Quebec spanning different forest environments, demonstrating improvements in modeled soil moisture and transpiration – especially in forested environments. Subsequently, the upgraded SVS2 was deployed in distributed simulations across North America (10 km resolution), with soil moisture estimates benchmarked against satellite retrievals and in-situ measurements from the International Soil Moisture Network, and evapotranspiration outputs validated against flux tower observations. These improvements support the development of an enhanced operational land surface prediction system for Canada.

Tue

09:15–09:30

Rooting out misrepresentation of water uptake in large-scale models
Dan Li [1, 2, 3], Manon Sabot [3], Jan Vanderborght [4], Daniel Leitner [4], Sufen Wang [1, 2]
[1] Center for Agricultural Water Research in China, China Agricultural University, Beijing 10083, China
[2] State Key Laboratory of Efficient Utilization of Agricultural Water Resources, Beijing 10083, China
[3] Max Planck Institute for Biogeochemistry, Jena 07745, Germany
[4] Agrosphere Institute (IBG-3), Forschungszentrum Jülich GmbH, Jülich 52428, Germany

Abstract: Root water uptake (RWU), influenced by root distribution, soil and root hydraulic properties, soil water status, along with plant and atmospheric demands for water, plays a critical role in determining terrestrial water fluxes. However, most land surface models (LSMs) simplify the representation of root systems and RWU, typically prescribing static vertical root distributions and neglecting soil-to-root hydraulic processes. Such simplifications can lead to biased estimates of evapotranspiration, limiting both LSM predictability and our understanding of ecosystem resilience, especially during drought conditions. In this study, we develop a new RWU scheme that couples dynamic root distribution with soil-to-root hydraulics and embed this new scheme within a plant optimality-based LSM framework (Sabot et al., 2022). The representation of dynamic root distribution allows roots to actively deploy toward water-rich soil zones in response to spatiotemporal moisture heterogeneity. The soil-to-root hydraulics are adapted from Vanderborght et al. (2023, 2024) and consider hydraulic gradients at the soil-root interface and within the perirhizal zone, i.e., the region immediately surrounding the roots. Finally, our new RWU scheme is flexible, water uptake can be integrated over the entire root zone to regulate whole-canopy gas-exchange, or uptake from distinct root-zone regions to affect different parts of the canopy. Using this framework, we ask three questions: (1) does the incorporation of mechanistic RWU improves LSM evapotranspiration estimates? (2) how does root plasticity vary with soil type and moisture conditions? and (3) does belowground hydraulics exert a stronger influence on canopy gas exchange than aboveground plant hydraulics? To answer these questions, we evaluate factorial configurations of our LSM (e.g., static vs. dynamic roots, without vs. with soil-to-root hydraulic gradients, etc.) across a suite of woody observational sites. Preliminary analysis shows that introducing dynamic roots and soil-to-root water potential gradients improves the predictability of carbon and water fluxes, especially during drought. Further exploration will yield new insights into plant water-use strategies and coordinated acclimation to water stress under changing environmental conditions.

Keywords: Root water uptake; Root distribution; Hydraulic gradients; Land surface models

References:
Sabot, M. E. B., De Kauwe, M. G., Pitman, et al. (2022). Predicting resilience through the lens of competing adjustments to vegetation function. Plant, Cell & Environment, 45(9), 2744-2761.
Vanderborght, J., Leitner, D., Schnepf, A., et al. (2024). Combining root and soil hydraulics in macroscopic representations of root water uptake. Vadose Zone Journal, 23(3), e20273.
Vanderborght, J., Couvreur, V., Javaux, M., et al. (2024). Mechanistically derived macroscopic root water uptake functions: The α and ω of root water uptake functions. Vadose Zone Journal, 23(4), e20333.

Mon

14:00–14:15

Wildfires are a major source of black carbon (BC), and their increasing frequency and intensity have amplified BC deposition onto glacier and snow surfaces. This deposition reduces surface albedo, accelerates snowmelt, and may trigger feedbacks that alter surface energy balance, moisture availability, and near-surface thermal gradients, with potential implications for boundary-layer structure, regional circulation, and fire weather. This study examines a largely unexplored feedback loop linking wildfire BC deposition on snow and ice to subsequent surface-driven influences on fire weather. Using the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) integrated with the Snow, Ice, and Aerosol Radiative (SNICAR) model, we explicitly couple wildfire emissions, BC transport and deposition, and snow-ice surface processes to quantify seasonal impacts across western North America. We focus on changes in albedo, snow water equivalent (SWE), and snowmelt. Results indicate that wildfire emissions account for more than 50% of wintertime BC over the study domain. BC deposition substantially reduces surface albedo and enhances snowmelt by ~55%, with melt increasing from 26.8 mm in winter to 41.9 mm in spring across mountain regions including the Alaska Range, the Rockies, and the Sierra Nevada. We further assess wildfire-driven changes in surface energy fluxes, near-surface temperature and humidity, and resulting fire weather indices, providing new insight into wildfire-cryosphere feedbacks relevant for fire preparedness and mitigation.

Wed & Thu

11:00–12:30

The remote influences of springtime Third Pole (TP) snow cover (TPSC) on the Indian Summer Monsoon (ISM) and the East Asian Summer Monsoon (EASM) have been extensively studied. However, a clear mechanism explaining the cross-season links remains not well established. Before we confirm any remote effects, it is essential to first verify local influences. Here, we bear out the enduring local impact of the springtime TPSC according to a numerical experiment together with an observational investigation. By examining the evolution of underlying heat sources, we propose a self-sustaining mechanism elucidating the enduring local impact: considering the case of the springtime TPSC deficit, the excessive precipitation that initially responds to the enhanced surface heat and water fluxes releases extra atmospheric latent heat, which in turn drives an anomalous circulation favouring the next-coming precipitation. This finding adds credit to the cross-season influences of the springtime TPSC remotely on the ISM and the EASM. Furthermore, our work implicates that the TP may get more summer precipitation in a warmer future since there will be an inevitable decrease in springtime TPSC.

Wed & Thu

11:00–12:30

Soil moisture(SM) deficit induced drought is emerging as an increasingly important threat to land ecosystem. On the one hand, soil moisture deficit will directly limit plant photosynthesis and transpiration. On the other hand, soil moisture can also influence atmosphere water vapor deficit (VPD) and heat by modulating water and energy fluxes on land, then impacting land ecosystem in an indirect way.

From the perspective of observations, we combine satellite observations of solar-induced fluorescence to disentangle the relative role of SM and VPD in limiting land ecosystem production on the global scale. We show that SM is the dominant driver of dryness stress on ecosystem production across more than 70% of vegetated land areas with valid data. On the regional scale, we use observational records of global atmospheric CO2 growth rate to represent tropical land carbon sink interannual variation (IAV). We find that the interannual relationship between tropical water availability and CO2 growth rate became increasingly negative during 1989-2018 compared to1960-1989. In other words, droughts increasingly reduce CO2 uptake in the tropical lands. However, most state-of-the-art coupled Earth System models(ESMs) and Land Surface models (LSMs) do not reproduce the intensifying water-carbon coupling.

From the perspective of models, we utilize two generations of factorial ESM experiments to show that SM consistently dominated the IAV of tropical land carbon uptake under both present and future climate in ESMs. The magnitude of this interannual sensitivity of tropical land carbon uptake to water variations (g) under future climate shows a large spread across the latest 16 ESMs (2.3 ± 1.5 PgC/yr/Tt HO). Based on the identified significant emergent relationship between g under future climate and present climate, the mean and spread of future g are reduced by about 41% and 44%, respectively (1.3 ± 0.8 PgC/yr/Tt HO), using observations and the emergent constraint methodology. However, the long-term tropical land carbon-climate feedback uncertainties in the latest 16 ESMs can no longer be directly constrained by land carbon cycle IAV compared with previous ESMs, given that additional important processes are not well represented in IAV but could determine long-term tropical land carbon storage.

Mon & Tue

11:00–12:30

Understanding the spatio-temporal dynamics of turbulent transport over heterogeneous surfaces is essential for improving our understanding and modelling the land-atmosphere feedback, as the near-surface turbulent layer serves as the lower boundary condition that governs momentum, heat, and moisture transport across the atmospheric boundary layer. This study investigates how flow direction relative to the elements of land-surface heterogeneity modulate near-surface turbulence over contrasting agricultural patches (wheat vs. maize) during the LAFI campaign (Land-Atmosphere-feedback-Initiative) in summer 2025 at the Hohenheim Land-Atmosphere-Feedback Observatory (LAFO). Observations were collected using high-resolution fiber optic distributed (FODS) of air temperature and wind 0.125m and 5s resolution across a transect at 2.0/ 2.8m agl and in vertical profiles (0 to 10m agl. Paired heated and unheated fiber-optic cables enable simultaneous measurement of temperature and wind velocity, allowing application of the elliptic model to compute longitudinal space-time correlations for both temperature, streamwise velocity and their cross-covariance under varying atmospheric stability regimes (unstable, near-neutral, stable) The statistical evaluation is posed in terms of flow direction relative to the orientation of patch boundary enabling distinguishing between parallel, perpendicular, and oblique flows modes.
Within the elliptic model framework, 𝑈𝑒 represents the convection velocity the rate at which turbulent eddies are transported downstream while 𝑉𝑒 is the decorrelation velocity, characterizing the temporal decay rate of eddy coherence. Their ratio 𝑉𝑒/𝑈𝑒 serves as a key signature of obliquity, where elevated values reflect enhanced turbulent distortion and non-stationary flux contributions unique to oblique wind regimes. Through systematic comparison across parallel, perpendicular, and oblique flow sectors, we investigate whether oblique flows consistently exhibit higher ratio values and hence variances and covariances compared to canonical cases, and how this signature varies with crop type, stability, and surface heterogeneity. By establishing 𝑉𝑒/𝑈𝑒 as a diagnostic metric for flow-surface misalignment, these findings inform parameterization of heterogeneous surface fluxes in mesoscale and Earth system models.

Mon

16:15–16:30

Land-atmosphere exchange shifts from energy-limited to water-limited regime at a critical soil moisture, which marks a fundamental transition in the Earth system. Estimates of the critical threshold vary a lot across studies despite its importance for the mechanistic understanding of soil moisture limitation on transpiration and plant productivity.
We introduce a novel, model-based diagnostic approach — the Normalized Transpiration Deficit (NTD) method — and demonstrate that it yields results highly consistent with observational methods such as finding breakpoints in the evaporative fraction. Using a hydraulically-enabled version of the CABLE-POP land surface model, we conducted a factorial experiment across various soil textures, climate regimes, and plant hydraulic parameters. It revealed that the critical threshold is at the same soil matric potential ψcrit independent of soil type, which gives a quasi-linear relationship between critical volumetric soil moisture θcrit and sand content, as observed in earlier studies. The dependency of θcrit on soil type vanished when it was normalised by field capacity, which yielded hence also a universal threshold of relative extractible water REWcrit as found empirically for forest ecosystems.
Most of the variance of θcrit, 86%, came from soil texture in the factorial experiment, while the variances of θcrit and REWcrit were dominated by plant hydraulic traits: the P50 value of stomatal conductance and P50 of xylem conductance, accounted for 87% and 77% of total variance, respectively. There was, however, no direct effect of climate on any of the critical thresholds, i.e. the thresholds remained invariant across climates for given soil and vegetation types. This suggests that apparent climate dependencies reported in observational studies may be artifacts due to limited soil moisture ranges at each observational site, or they represent biological adaptation and acclimation that is currently not captured in our static model parameters.
These findings highlight the necessity of incorporating ecosystem-scale hydraulic regulation in biosphere models to reconcile divergent estimates of critical thresholds and to improve predictions of drought impacts on water and carbon fluxes.

Mon & Tue

11:00–12:30

The Land surface Interactions with the Atmosphere over the Iberian Semi-arid Environment (LIAISE, Boone et al., 2025) field campaign took place in 2021 in the eastern Ebro river subbasin (Iberian Peninsula, western Mediterranean Sea). The aim of the project was to explore the surface–atmosphere dynamics in a large irrigated area and its mesoscale interactions with the surrounding rainfed areas.
 
Results from the first mesoscale intercomparison of models in the LIAISE region (Jiménez et al. 2025) showed that models (MesoNH, Unified Model and WRF) are able to reproduce the organization of the flow at lower levels but differences are mainly related to the misrepresentation of the surface processes (such as surface cover variability or soil moisture content, among others). In some models the surface cover was not realistic and in others the surface variability of the region was misrepresented. This case was based on a Marinada event (common wind regime in the region that consists on the arrival of the sea breeze front generated at the coast that surmounts the Catalan Coastal Mountain Range) prior to the LIAISE campaign and the model outputs were validated against surface weather observations from the Catalan Meteorological Service (SMC) and satellite-derived products. On the other hand, the results of Lunel et al (2024) highlight the importance of having an adequate representation of the irrigated area to better capture the temporal and spatial scales of the Marinada.
 
From the experiences and the results of these works, a new intercomparison exercise of models is proposed based on the LIAISE experimental field campaign to take the advantage of having measurements over the different soil covers of the study region. The main objective is to evaluate the importance of taking into account the irrigation processes in the models, especially in such a complex region. Understanding the best parameterizations and their improvement is crucial to increase the complexity of weather prediction and climate models.

Wed & Thu

11:00–12:30

Clouds and vegetation are tightly coupled, linking ecosystem functioning with climate. On the one hand, clouds regulate surface radiation and water availability, thereby influencing vegetation photosynthesis; on the other hand, vegetation modifies land-atmosphere fluxes and feeds back on cloud formation and radiative forcing. Here, we synthesize recent advances from these two complementary perspectives.
The first part examines the impacts of clouds on vegetation photosynthesis. Using observational- and model-based datasets spanning recent decades, we show that the sensitivity of photosynthesis to cloud cover is spatially modulated by hydroclimatic conditions. In water-limited arid regions, clouds tend to enhance photosynthesis by increasing precipitation, typically with a lag of less than one month. Conversely, in energy-limited humid regions, clouds reduce photosynthesis almost instantaneously by diminishing incoming solar radiation. Under global warming, projected changes in cloud cover are expected to reduce gross primary productivity in arid regions while enhancing it in humid regions, thereby amplifying regional disparities in ecosystem functioning.
The second part focuses on the impacts of deforestation on clouds and climate. By combining climate model simulations with data-driven approaches, we find that deforestation leads to reductions in local low-level clouds and tropical high-level clouds. These cloud reductions primarily result from weakened surface turbulent heat fluxes, which constrain atmospheric uplift and moisture supply. The resulting radiative warming from reduced cloud cover partially offsets the cooling effect associated with increased surface albedo following deforestation.
These results highlight the critical role of cloud-vegetation interactions in shaping both ecosystem productivity and climate, underscoring the need for improved constraints on these processes when evaluating future climate and land-use change impacts.

Mon

16:15–16:30

Turbulent surface fluxes are a key component in land-atmosphere interactions and regarded as a crucial factor in the accurate representation of near-surface temperature profiles in numerical weather prediction (NWP) and climate models. As subgrid-scale processes, surface fluxes are parameterized in NWP models and represented as grid cell average. However, unresolved surface heterogeneities and microtopography alter turbulent boundary layer flows in a non-linear manner, introducing uncertainty in this coupling process, calling for a new probabilistic modelling approach. Here, we begin by quantifying the coupling uncertainty using information entropy and demonstrate based on neighboring flux tower sites that their mutual information content is lowest under strongly stable stratification. This shows that the physical process of land-atmosphere decoupling, which typically goes along with a cold bias in NWP models due to run-away cooling, translates to an increased information loss. In a second step, we use a statistical approach to couple surface and atmospheric temperatures vertically. Specifically, we are developing a copula Bayesian network (CBN) for interpolating near-surface temperature profiles, serving as alternative to T2m-diagnostics used in NWP models. The new CBN is trained based on tower observations, tested in a sensitivity study and applied to the HARMONIE-AROME model system. The CBN results in (1) a reduction in the warm bias inherent to NWP forecasts in wintertime stable boundary layers, thereby enabling cold temperature extremes to be better represented, and allows (2) the consideration of uncertainty associated with subgrid-scale spatial variability, and (3) the representation of internal shallow boundary layers below first model level. The use of CBNs combines the advantages of uncertainty propagation inherent to Bayesian networks with the ability to model complex dependence structures between random variables using copulas. By combining insights from copula modelling and information entropy, criteria for the applicability of CBNs in the further development of probabilistic parameterizations in NWP models are derived.

Wed & Thu

11:00–12:30

Understanding soil moisture dynamics and plant-soi-atmosphere interactions is essential for addressing drought, water scarcity and climate-driven ecosystem change. In this contribution, we present two large-scale monitoring applications utilizing METER Group technologies, alongside insights into our next-generation SOLYX 14 sensor for volumetric water content, temperature and electrical conductivity.
The first case study highlights Hungary’s nationwide drought monitoring network, a public-domain system comprising more than 100 stations equipped with temperature and soil moisture sensors at multiple depths. These data are used to calculate the Hungarian Drought Index (HDI), which integrates meteorological drought indicators with a drought stress index considering plant responses to precipitation deficits and heat extremes. The resulting spatially resolved drought information is openly available via the national drought monitoring web platforms, forming a critical operational tool for decision-makers.1
The second case study showcases the ICP Forests intensive monitoring network in the Bavarian Forest region (Germany). Nineteen long-term forest climate stations – ranging from fully equipped meteorological reference plots to easily deployable satellite stations – record meteorological variables and subsurface conditions, including soil moisture, temperature and water potential profiles. The resulting high-resolution data provide essential insights into how changing climate conditions influence physical, chemical and biological processes in forest ecosystems.2
Additionally, we introduce the SOLYX 14, our new multi-parameter soil sensor featuring an advanced CDX measurement methodology that improves accuracy and stability, particularly in saline soils. By independently capturing both the real (related to water content) and the imaginary dielectric permittivity (which is influenced by salts and lossy clays), the sensor provides more reliable data and reduces uncertainty in heterogeneous field conditions.3
Together, these examples demonstrate how advanced sensing technologies and large-scale monitoring deployments contribute to improved understanding of hydrological stress, forest ecosystem processes and climate adaptation strategies.

1https://vizhiany.vizugy.hu/, http://aszalymonitoring.vizugy.hu/, https://vizhiany.vizugy.hu/
2https://www.fovgis.bayern.de/wks/, https://www.waldduerremonitor.de/
3https://metergroup.com/products/solyx-14/

Wed & Thu

11:00–12:30

The recurring devastating floods in KwaZulu-Natal emphasise the urgent need to better understand the drivers of extreme rainfall events in South Africa’s coastal regions. While cut-off low systems are recognised as major contributors to such floods, the role of oceanic forcing—particularly the Agulhas Current System—in shaping boundary layer dynamics and rainfall intensity remains insufficiently explored. This study addresses this gap by employing high-resolution simulations to investigate how sea surface temperature gradients and oceanic heat fluxes from the Agulhas Current influence moisture transport, vertical stability, and convective development during the passage of cut-off lows. The Model for Prediction Across Scales (MPAS) was used to perform a series of multi-simulation sensitivity experiments with both real and idealised sea surface temperature distributions over the Agulhas Current region during a devastating cut-off low event. Results show that increased Agulhas warming enhances latent heat flux, raising near-surface moist static energy that is advected inland within the cut-off low circulation. This additional moisture deepens the boundary layer and strengthens the convergence of low-level moisture flux into the cut-off low. These thermodynamic adjustments elevate convective available potential energy (CAPE) and systematically intensify precipitation associated with the storm. However, the rainfall response to changes in Agulhas warming is nonlinear, varying with the magnitude of sea surface temperature perturbations. These findings highlight the critical role of ocean–atmosphere coupling in shaping extreme rainfall events and emphasise the need to incorporate Agulhas Current variability into regional climate models to improve flood forecasting and disaster preparedness in South Africa.

Mon & Tue

11:00–12:30

Land surface heterogeneity plays a fundamental role in governing exchanges of energy, moisture, and momentum within the coupled land-atmosphere system. Variations in land cover from snow and vegetation to agricultural mosaics and dense urban fabrics, introduce strong spatial contrasts in surface properties such as albedo, roughness length, soil moisture availability, and thermal inertia, thereby shaping boundary-layer processes and local-to-regional climate dynamics. Understanding and modelling these heterogeneous land surfaces is particularly critical in rapidly urbanising regions of the Global South, where anthropogenic land transformation is occurring at unprecedented rates.

This study investigates long-term changes in land surface heterogeneity within the Patna Urban Agglomeration (PUA), India, a densely populated alluvial city highly vulnerable to ecological degradation, hydroclimatic stress, and heat-related risks. Using an observation-driven approach, multi-temporal satellite imagery from 1990, 2010, and 2024 was analysed to characterise spatial and temporal patterns of Land Use-Land Cover (LULC). Major land-cover classes including water bodies, agricultural land, vegetated areas, barren land, and built-up surfaces were mapped using Geographic Information System (GIS) techniques. Supervised classification based on the maximum likelihood algorithm was applied consistently across all time slices to ensure comparability. Classification accuracy was rigorously evaluated using independent reference pixels, yielding Kappa coefficients exceeding 0.90, indicating high reliability of the derived land-cover products.

Change detection analysis reveals substantial modification of the land surface over the past three decades, with built-up areas more than doubling between 1990 and 2024. This rapid urban expansion has occurred largely at the expense of agricultural land, open spaces, and vegetated surfaces, resulting in increased surface sealing and reduced landscape permeability. Such transformations amplify spatial heterogeneity in surface radiative and thermal properties, potentially altering surface energy balance partitioning, near-surface temperature regimes, and atmospheric stability. The observed decline in moisture-retaining land covers further suggests implications for evapotranspiration fluxes and land–atmosphere feedbacks during extreme heat and monsoon periods.

By providing empirically derived evidence of evolving land surface heterogeneity, this study contributes to the Pan-GLASS objective of strengthening observational foundations for coupled land–atmosphere research. The results highlight the need for integrating high-resolution, observation-based LULC datasets into land-surface and regional climate models to better represent heterogeneous urban-rural transitions. While focused on a tropical urban context, the methodological framework and insights are transferable across diverse land-cover regimes from cryospheric to vegetated and urban systems, underscoring the broader relevance of heterogeneous land surfaces in shaping climate processes. The study demonstrates how long-term land-cover observations can inform improved modelling of land–atmosphere interactions and support sustainable urban and regional planning under a changing climate.

Keywords: Land Surface Heterogeneity, Remote Sensing Observations, Land Use-Land Cover Change, Urban Climate and Land-Atmosphere Interactions

Wed & Thu

11:00–12:30

The Pan-Global Land–Atmosphere Conference (GLASS) provides an important platform for enhancing my knowledge and research on the monsoon system, droughts, and extreme weather events, which have direct implications for socio-economic resilience in the South Asia. As the Indian Summer Monsoon is strongly regulated by complex land–ocean–atmosphere interactions, GLASS offers a unique opportunity for Indian researchers like myself to actively engage in global research efforts and contribute to the advancement of climate science. The conference presents an excellent opportunity to interact with internationally leading scientists and gain exposure to advanced research on observational networks, satellite-based remote sensing products, and state-of-the-art coupled land–atmosphere and Earth system modelling approaches. The integration of anthropogenic forcings in these approaches is essential for understanding climate drivers and feedback mechanisms that alter weather and climate systems. The focused sessions on human-induced forcings and climate feedback mechanisms are particularly valuable for improving my understanding of monsoon variability, onset and withdrawal processes, intraseasonal oscillations, and the increasing frequency of extreme events such as heatwaves, prolonged droughts, cloudbursts, and floods. The conference also provides an excellent opportunity to discuss unresolved weather phenomena affecting the monsoon at large, regional, urban, and rural scales with internationally renowned scientists. Furthermore, GLASS fosters international collaboration relevant to India’s diverse climatic regions, where land–atmosphere coupling strongly influences rainfall variability and extremes. Importantly, the conference supports the translation of fundamental research into operational climate services by strengthening linkages between academic institutions and operational agencies involved in weather forecasting, agriculture, water resource management, and disaster risk reduction. By enhancing scientific capacity, promoting interdisciplinary research, and increasing global visibility, GLASS contributes to improved prediction of monsoon variability and extremes, thereby supporting climate adaptation, sustainable development, and evidence-based policy formulation in India. The vast amount of knowledge gained from attending the conference will be indirectly benefited by students and fellow researchers where I am serving as well as collaborating institutes/universities.

Wed & Thu

11:00–12:30

Understanding controls on Mesoscale Convective Systems (MCSs) is critical for predicting rainfall extremes across scales. Spatial variability of soil moisture (SM) presents such a control, with ~200km dry patches in the Sahel observed to intensify mature MCSs. Here we test MCS sensitivity to spatial scales of surface heterogeneity using a framework of 78 Unified Model experiments initialised from scale-filtered SM. We demonstrate the control of SM heterogeneity on MCS populations, and the mechanistic chain via which spatial variability propagates through surface fluxes to convective boundary layer development and storm environments. When all sub-synoptic SM variability is homogenised, peak MCS counts drop by 23%, whereas maintaining small-scale variability maintains primary initiation rates, reducing the drop in MCS totals. In sensitivity experiments, boundary layer development prior to MCSs is similar to that over mesoscale dry SM anomalies, but driven by cloud-free slots of increased shortwave radiation. This reduces storm numbers and potential predictability

Wed & Thu

11:00–12:30

The effects of desert afforestation actions, such as those used for climate change mitigation, during extreme heat events remain an important yet unresolved question. The well-studied, semi-arid Yatir forest provides a unique lens through which we study land surface-atmosphere interactions.
The Yatir forest’s net radiation is higher than in any other eco-region. The massive radiation load is balanced by large sensible heat flux, which can influence the forest microclimate and create a thermal contrast with the surrounding shrubland. These processes, in turn, can affect near-surface atmospheric conditions and boundary-layer dynamics.
Here, we combine in-situ measurements with high-resolution ICON model simulations to offer new insights into the role of local afforestation in shaping surface weather and boundary-layer dynamics during extreme heat events. The in-situ observations not only describe the forest’s physical and physiological properties but also provide essential inputs to the model, enabling an integrated framework that captures known forest-scale processes and demonstrates their upscaling effects across the region.
Our simulations of a heat wave event from May 20 to May 25, 2019, reveal sensible heat flux increases of up to 300 W m−2 within the forest, resulting in surface (skin) cooling of up to 15 °C, while simultaneously producing warming of up to 2 °C in 2-m temperature. These contrasts generate a distinct forest-induced circulation. Remarkably, this circulation produces strong instability even under synoptic conditions dominated by harsh subsidence. Our findings underscore the complex, sometimes counterintuitive role of semi-arid afforestation during extreme heat events, with important implications for land-management strategies under different atmospheric forcing regimes.

Land surface (LSMs) and Earth system models (ESMs) have been shown to underestimate transpiration (T) to evapotranspiration (ET) ratios due to biases in evaporation (E) and T. Underestimated ET and T/ET in models contributing to the Climate Model Intercomparison Project Phase 6 (CMIP6) were found to result in warm biases in 2 m air temperatures in the central United States. The Canadian Land Surface Scheme Including biogeochemical Cycles (CLASSIC), which is the land surface component of the Canadian Earth System Model (CanESM), exhibited low T/ET ratios compared to observation-based estimates and tended to underestimate gross primary productivity (GPP) in arid/semi-arid regions. To address excessive ground E in CLASSIC, we included a dry surface layer (DSL) parameterization to increase surface resistance to water vapour and heat fluxes. The DSL parameterization was combined with modifications to the ET partitioning into canopy E and T, allowing T to occur from the dry fraction of the plant canopy while water is evaporating from the wet fraction, which was previously not the case. These modifications, in offline CLASSIC simulations, reduced ground E and ET during wet periods in arid/semi-arid regions, improving the seasonality of ET, increasing soil moisture availability in the root zone and thus increasing gross primary productivity (GPP) and vegetation biomass in arid regions. Seasonally dry tropical forests, on the other hand, showed a reduction in annual simulated GPP. Despite these larger regional GPP changes, globally, GPP was only slightly different (1.5% lower) due to our modifications. The proportion of T relative to ET globally increased from 25% to 41%, which is an improvement compared to observations (57% ± 7%).

We then implemented these ET-related parameterization changes into the Canadian Earth System Model (CanESM v.5.2). Compared to previous CanESM versions, the land model was updated from CLASS-CTEM to CLASSIC and the number of soil layers was increased from 3 to 20 soil layers going down to a depth of ~61 m (increased from a soil depth of 4.1 m for the 3-layer implementation), which is the default offline CLASSIC setup. Soil permeable depths, however, have been limited to a depth of 4 m. Implementation of the DSL and the canopy ET partitioning changes increased GPP (globally ~8% higher than CanESM baseline), especially in arid regions. These changes also cause the historical (1850-2014) model state to warm by ~0.6°C to a global average screen temperature of 14.1°C. Globally, annual ET is similar to the baseline. Changes in the origin of ET, however, affect P patterns and temperatures. Despite underestimated GPP and T in the Amazon, modified CanESM had increased globally averaged T/ET similar to offline simulations.

Wed

09:15–09:30

The advancement of Earth System Models (ESMs) has reached a pivotal milestone where global coupled simulations at kilometer-scale resolutions are now computationally feasible. This study presents the development and initial results of a global coupled configuration using the ICON-HAMOCC-JSBACH model at a horizontal resolution of 10 km, including full carbon cycle. By explicitly resolving deep convection rather than relying on traditional convective parameterizations, this framework allows for a more physically consistent coupling between the atmosphere and the terrestrial biosphere as well as between water and carbon fluxes.
A central objective of this work is the validation of the simulated land-carbon cycle components. We evaluate the simulated carbon fluxes and key environmental variables by comparing modeled patterns and variability against available observational constraints, including surface radiation, near-surface meteorology, hydrologic states, and ecosystem carbon exchange metrics. Particular attention is given to the model’s ability to capture the spatial and temporal variability of carbon exchange in the Amazon rainforest and potential drifts. First analysis shows that the seasonal and diurnal cycle of gross primary productivity, carbon dioxide flux, and evapotranspiration over the Amazon region is well reproduced compared to FluxNet observations. Furthermore, the simulated precipitation, surface temperature, and incoming radiation are evaluated against satellite-based datasets and reanalysis products to ensure the robustness of the atmospheric forcing. This storm-resolving setup provides a unique platform to investigate the interactions between the hydrological cycle and the terrestrial carbon sink. This work establishes a baseline for next-generation ESMs, moving toward a more process-based representation of the global carbon cycle in a changing climate.

Mon & Tue

11:00–12:30

The CORDEX Flagship Pilot Study “URBan Environments and Regional Climate Change” (FPS-URB-RCC, Langendijk et al. 2024) is established to improve understanding of the interactions between regional climate change and urban environments by conducting coordinated RCM ensemble experiments with refined urban representations. In the second phase of FPS-URB-RCC, 20 institutions worldwide produced an initial 42-member ensemble as a part of the STAGE-0 experiment. These simulations covered a four-month period over the Paris metropolitan area at convection-permitting resolution, with sophisticated representations of urban environments. A big component of this effort was a large, highly coordinated sub-ensemble of 30 sensitivity runs with the Weather Research and Forecasting (WRF) model. We use these WRF simulations to examine how model settings influence the representation of urban features and land–atmosphere (LA) feedback. While 20 runs varied in model physics and static inputs, an additional 10 identical WRF-CTRL configurations were executed across different machines or with slight variations in initial conditions, which allows us to quantify internal variability and distinguish it from true model sensitivity. We focus our analysis on the model representation of urban dry island (UDI), surface energy balance and the effects on the boundary layer height, which represent the atmospheric component of the LA feedback chain. The results are compared with available station observations within the Paris metropolitan area, and the non-urban surroundings. The initial results reveal a significant spread among the WRF-CTRL runs, indicating a strong influence on the internal variability on the results of the STAGE-0 experiment. Representation of the urban features such as UDI is most sensitive to the static input data and to the complexity of the urban physics parameterizations.

Reference: Langendijk, G.S. et al. (2024) “Towards Better Understanding the Urban Environment and Its Interactions with Regional Climate Change – The WCRP CORDEX Flagship Pilot Study URB-RCC.” Urban Climate 58:102165. https://doi.org/10.1016/j.uclim.2024.102165.

Acknowledgments: This work is a part of projects PROTECT (PID2023-149997OA-I00) funded by MICIU/AEI/10.13039/501100011033 and by ERDF/EU, and European Union’s Horizon Europe research and innovation programme IMPETUS4CHANGE (grant agreement No 101081555). JM and AS acknowledge funding by the Ministry for the Ecological Transition and the Demographic Challenge (MITECO) and the European Commission NextGenerationEU (Regulation EU 2020/2094), through CSIC’s Interdisciplinary Thematic Platform Clima (PTI-Clima)

Tue

16:45–17:00

The application of artificial intelligence and machine learning (AI/ML) methods to land–atmosphere (L-A) interaction research is increasingly seen as a key pathway for advancing process understanding and improving models. However, the application of such approaches is constrained by the availability of standardized, interoperable, and well-documented observational and modelling datasets, particularly when data originate from complex three-dimensional observational systems and high-resolution models.
A new generation of Global Energy and Water Exchanges (GEWEX) Land-Atmosphere Feedback Observatories (GLAFOs) is expected to routinely generate such data. An operational prototype at the University of Hohenheim, Stuttgart, is already producing advanced multi-sensor observations and high-resolution model outputs within the DFG Research Unit 5639, Land–Atmosphere Feedback Initiative (LAFI). These datasets provide a unique opportunity to rethink the modelling of L-A interactions through data-driven and hybrid physics-ML approaches, provided they are made FAIR and AI-ready.
In other words, addressing LAFI’s scientific aim of closing key knowledge gaps in L-A feedbacks that limit the accuracy of weather and climate simulations is, therefore, critically dependent on robust research data management. Standardized data structures, interoperable metadata, and reliable data access are essential to enable systematic benchmarking of models, cross-site comparisons, and the application of AI/ML methods across observational and modelling domains.
Within LAFI-GLAFO, ongoing activities focus on converting heterogeneous datasets into Climate and Forecast (CF) convention compliant NetCDF formats and providing them through centralized, version-controlled infrastructures. Reproducible data processing, quality control, and documentation are implemented using Python-based workflows managed via GitLab. This allows us to construct and maintain an up-to-date dataset which can be directly used in different contexts, not only within our group. Initial access is provided to project researchers, with planned expansion through API-based services to facilitate broader use in model development and ML experimentation.
In parallel, collaboration with international initiatives such as obs4MIPs supports the use of LAFI–GLAFO observations for climate model evaluation and benchmarking, including the development of protocols for advanced instrumentation (e.g., Doppler, Raman, and water vapor differential absorption lidars). Training and tutorial activities further lower barriers for uptake by the modelling and ML communities.
We will present relevant lessons learned from ongoing activities in designing FAIR, AI-ready data infrastructures within LAFI–GLAFO, discuss challenges, the role of the importance of early and structured collaboration with AI/ML-focused initiatives and institutional IT services to enable effective, scalable implementation, and outline opportunities for enabling advanced process understanding, benchmarking, and machine learning to advance land–atmosphere modelling. The perspectives presented are relevant for a wide range of researchers working at the interface of data sharing, AI/ML, and land–atmosphere interaction research.

Tue

16:30–16:45

Dynamic Global Vegetation Models (DGVMs) must simulate carbon storage and land–atmosphere exchanges of energy, water, and carbon across diverse landscapes and climates. Few DGVMs have self-consistent approaches to represent canopy structural heterogeneity for three core DGVMs functions that couple with Earth System Models (ESMs): the canopy radiative transfer physics, carbon dynamics, and ecology. We investigate a geometrical optical radiative transfer (GORT) framework in the Ent Terrestrial Biosphere Model, a demographic DGVM coupled to the NASA Goddard Institute for Space Studies ESM, ModelE. The Ent Analytical Clumped Two-Stream (ACTS) model is a GORT model based on multi-cohort canopies with ellipsoidal crowns to account for the effects of vertical foliage profiles (VFPs) and clumping on canopy radiative transfer. We evaluate optimization of allometry and cohort structure to reproduce observed VFP, basal area (a proxy for canopy biomass), and canopy albedo across three contrasting ecosystems: temperate deciduous forest, evergreen conifer forest, and woody savanna. Simulated diurnal and seasonal canopy albedo are evaluated against Fluxnet tower and MODIS satellite observations. In addition, we conduct a sensitivity analysis of how structural heterogeneity (vertical foliage distribution, clumping, crown geometry) and surface optical properties (leaf/stem and soil reflectance) jointly control albedo. We find that the dominant control on VIS–NIR reflectance and seasonal albedo is the seasonality of leaf NIR optical properties. Accurate estimation of VFP and basal area is very sensitive to allometric parameters and vertical structure. We recommend prioritization of measurements of: leaf optical properties to include the seasonality in the NIR and both reflectance and transmittance for the radiative transfer models; and crown allometry for VFPs. This study provides an observationally- constrained framework for identifying the minimum demographic complexity required to capture canopy heterogeneity to accurately and self-consistently predict the RT physics and biomass of DGVMs. Future work will investigate whether the vertical structure that optimizes radiative and carbon dynamics also captures the ecological dynamics of community competition.

Key words: albedo; foliage clumping; vegetation demography; dynamic vegetation modeling.

Mon & Tue

11:00–12:30

Accurate, spatially resolved observations of evapotranspiration are essential for understanding and parameterizing land-atmosphere exchanges, yet they remain difficult to obtain at fine scales. Existing remote sensing methods for measuring evapotranspiration lack the spatial resolution needed to resolve surface heterogeneity in vegetated ecosystems, and their sources of uncertainty are not well-constrained.

Here, we present two approaches for independently quantifying fine-scale transpiration using thermal imagery and a suite of environmental sensors mounted on an unmanned aerial vehicle (UAV) platform. The first is a surface energy balance (SEB) approach adapted for very-high resolution thermal imagery; the second leverages vertical profiles of air temperature and humidity to estimate latent heat flux from the Bowen Ratio. Both approaches derive the energy equivalent of transpiration, latent heat flux (λE), solely using data acquired from the UAV.

We compare the two approaches and their sources of uncertainty using repeat flights over a grassland eddy covariance site under a range of diurnal conditions. The SEB approach produces λE estimates within ~20% of tower measurements, with uncertainty primarily dominated by surface temperature and aerodynamic resistance. λE calculated from the Bowen Ratio approach yields values ~30% higher than tower observations, with uncertainty dominated by inaccuracies in air temperature and humidity measurements. Despite these limitations, the atmospheric profiling approach shows lower theoretical uncertainty and strong potential for improvement over traditional SEB approaches.

These results demonstrate the potential of UAV-based atmospheric profiling to directly observe fine-scale land-atmosphere fluxes, providing new constraints for understanding and parameterizing surface exchange processes.

Mon & Tue

11:00–12:30

Humans and the atmosphere have always interacted intensely in both directions. Through greenhouse gas emissions, anthropogenic activities have directly altered the climate, yet human influence also occurs indirectly through the modification of land surface processes. Irrigation, for example, adds substantial water to systems that are typically moisture-limited, actively modulating local and regional weather patterns worldwide. The most direct impact of this practice is an increase in local soil moisture, which leads to modified surface energy partitioning by significantly increasing latent heat fluxes at the expense of sensible heat fluxes. These irrigation-induced changes in local land-atmosphere coupling can propagate beyond the land surface and influence regional atmospheric circulation and convective organization, suppressing the planetary boundary layer (PBL) height and simultaneously increase the convective available potential energy (CAPE). At regional to continental scales, extensive irrigation has been linked to changes in the initiation, frequency, and duration of mesoscale convective systems (MCSs). Despite growing evidence of these impacts, the specific causal pathways through which soil moisture perturbations influence atmospheric variables remain insufficiently quantified in fully coupled modeling frameworks.
One atmospheric process that might be particularly sensitive to irrigation-driven changes in convection is lightning. Lightning occurrence depends on vertical motion, cloud microphysics, and atmospheric instability, all of which are modulated indirectly by land surface conditions. As lightning is a major natural ignition source for wildfires, understanding how anthropogenic irrigation reshapes these patterns is of increasing relevance for risk assessment and environmental management.
In the work presented here, as part of the Belspo METEORI project, we investigate the influence of large-scale anthropogenic irrigation on land-atmosphere interactions and lightning occurrence over the Contiguous United States (CONUS) between 2015 and 2023 with a spatial resolution of 27 km. We perform fully coupled land-atmosphere simulations using the NASA-Unified Weather Research and Forecasting framework with Noah-MP as the land surface model, conducting multi-season experiments with and without irrigation. The impact of irrigation on land-atmosphere coupling is quantified using the Liang-Kleeman information flow framework, enabling a causal assessment of directional interactions between surface fluxes and atmospheric variables. Lightning occurrence is estimated using multiple lightning parameterization schemes, and the resulting spatiotemporal patterns are compared between simulations to evaluate the expected influence of irrigation on regional lightning activity and potential ignitions.

Mon

16:15–16:30

Rapid warming is reshaping the hydroclimate of Northern Europe, with spring and summer conditions increasingly characterised by warm, dry spells and increased risk of heatwaves and drought. As soils dry, evapotranspiration becomes constrained by water availability rather than atmospheric demand, moving the region toward a moisture-limited regime in which land “memory” and soil moisture anomalies can strongly influence atmospheric variability. Under these conditions, land–atmosphere feedback become critical for the onset, persistence, and intensity of extremes, yet they remain imperfectly represented in many operational climate and sub-seasonal to seasonal forecasting systems. In particular, soil moisture and snow are often initialised from climatology or offline land reanalysis without fully coupled data assimilation, limiting the accuracy of land surface states and weakening their influence on subsequent atmospheric evolution. This work presents a first assessment of the Norwegian Climate Prediction Model version System 2 (NorCPM-S2), a coupled Earth system prediction framework designed to better represent land–atmosphere interactions. NorCPM-S2 combines ensemble-based ocean data assimilation with sequential daily assimilation of ESA CCI soil moisture in a loosely coupled configuration, allowing for more realistic exchanges of water and energy between land, ocean, and atmosphere. Using a large ensemble of reanalysis, we investigate how soil moisture and land memory modulate surface fluxes, near-surface temperature, and the evolution of hydroclimate extremes over Northern Europe. System performance is assessed through comparisons with multiple independent data sources, including reanalysis products, in situ measurements, and satellite-derived indicators of soil moisture and vegetation conditions. Particular emphasis is placed on the 2018 Northern European heatwave, which serves as a representative benchmark event for evaluating the system’s ability to capture extreme land–atmosphere interactions. This event provides a useful test case because of its pronounced soil moisture deficits, persistent high temperatures, and strong land–atmosphere feedbacks across the region. The results demonstrate the added value of incorporating coupled land surface information in the analysis framework for better understanding and predicting regional climate extremes. In particular, the assimilation of soil moisture observations enables the system to more accurately reproduce the lower bounds of soil moisture during drought conditions. This improved representation of land surface states strengthens the simulation of land–atmosphere coupling processes, thereby enhancing the model’s ability to capture feedback mechanisms that can amplify heatwaves and dry spells.

Mon & Tue

11:00–12:30

Land-atmosphere coupling influences critical climate processes from regional drought to seasonal predictability. However, existing methodologies struggle to capture the full spectrum of teleconnection mechanisms. Conventional perturbation approaches—coordinated multi-model experiments, such as GLACE which compares interactive versus prescribed land surface variables like a brute force method, forecast initialization studies, and idealized extreme event simulations—each have limitations in signal detection, often requiring unrealistically strong perturbations that could distort model climatology. We introduce an enhanced system identification framework that injects time-varying zero-mean square-wave forcing into land surface temperature, enabling isolation of atmospheric responses without altering the background state. Using a physics-based AGCM (ICTP AGCM v4.0), the framework successfully captures dynamic atmospheric responses to surface forcing, characterized by jet-guided Rossby wave propagation and coherent vertical baroclinic adjustment. A finite ~10-day linear-response window is identified, beyond which internal variability becomes dominant. To examine the broader applicability of the framework, we conduct an exploratory extension to an AI-based climate model (NeuralGCM). Due to current boundary-condition configurations, analogous zero-mean forcing is applied over sea surface temperature rather than land surface temperature. Differences in teleconnection structure and temporal evolution emerge, suggesting that response characteristics may differ between physics-based and AI-based climate models. This extension highlights the potential of the framework for evaluating AI-based climate models. Overall, the proposed system identification framework provides a process-oriented benchmark for evaluating land-atmosphere coupling fidelity and offers a pathway for assessing teleconnection fidelity in next-generation AI-based climate models as they incorporate more comprehensive surface boundary representations.

Mon & Tue

11:00–12:30

Vegetation is fundamental in regulating land–atmosphere exchanges of water and energy, yet its influence on large-scale hydrological simulations is challenging to quantify. Specifically, the degree to which observational constraints on vegetation states enhance the representation of the global terrestrial water cycle at the basin scale remains insufficiently understood.
This study evaluates the hydrological impact of assimilating satellite-derived Leaf Area Index (LAI) using the global land data assimilation system LDAS-Monde within the ISBA-CTRIP land-river model, both developed at the Centre National de Recherches Météorologiques (CNRM, France). Two global simulations spanning 1982–2017 are compared: an open-loop
experiment with freely evolving vegetation and a data assimilation experiment incorporating LAI observations. Both ISBA-CTRIP configurations are driven by identical ERA5 atmospheric forcing. Model performance is assessed against multiple observation-based datasets and discharge records from 2139 river basins around the world, using anomaly correlation,
unbiased root-mean-square error, and mean bias metrics. To interpret the diverse hydrological responses, basins are classified into hydro-biospheric regimes based on observation-based descriptors of climate, vegetation dynamics, storage variability, and land–atmosphere coupling.
The results indicate that LAI assimilation substantially enhances the realism of vegetation climatology and phenology, and produces robust, spatially coherent improvements in evapotranspiration skill across all hydro-climatic regimes. Additionally, assimilation indirectly improves the simulation of snow water equivalent variations, underscoring the influence of
vegetation on snow interactions through canopy interception and regulation of energy exchanges.
In contrast, improvements in streamflow are more modest and heterogeneous. This outcome reflects the integrative nature of discharge, where compensating errors among evapotranspiration, storage changes, and precipitation forcing constrain the direct translation of land-surface improvements into runoff skill. Nevertheless, LAI assimilation systematically increases discharge correlation and bias in several regimes.
Overall, these findings demonstrate that assimilating vegetation observations offers a practical pathway to improve key components of the terrestrial water cycle at the global scale, particularly evapotranspiration and seasonal snow storage. The analysis of hydro-biospheric similar basins, derived through regime-based clustering, further identifies where
vegetation constraints exert the strongest hydrological influence, providing guidance for the development of next-generation global land surface reanalyses and enhancing hydrological predictions based on Earth system models.

Wed

09:30–09:45

Soil physics models have long relied on simplifying assumptions to represent complex processes, yet such assumptions can strongly bias model predictions. Here, we propose a paradigm-shifting differentiable hybrid modeling (DHM) framework that instead of simplifying the unknown, learns it from data. As a proof of concept, we apply the hybrid approach to the challenge of partitioning the soil water retention curve (SWRC) into capillary and adsorbed water components, a problem where traditional assumptions have led to divergent results. The hybrid framework derives this partitioning directly from data while remaining guided by a few parsimonious and universally accepted physical constraints. Using basic soil physical properties as inputs, the hybrid model couples an analytical formula for the dry end of the SWRC with data-driven physics-informed neural networks that learn the wet end, the transition between the two ends, and key soil-specific parameters. The model was trained on a SWRC dataset from 482 undisturbed soil samples from Central Europe, spanning a broad range of soil texture classes and organic carbon contents. The hybrid model successfully learned both the overall shape and the capillary and adsorbed components of the SWRC. Notably, the model revealed physically meaningful pore-scale features without relying on explicit geometrical assumptions about soil pore shape or its distribution. Moreover, the model revealed a distinctly nonlinear transition between capillary and adsorbed domains, challenging the linear assumptions invoked in previous studies. The methodology introduced here provides a blueprint for learning other hidden or complex soil processes where high-quality datasets are available but mechanistic understanding is incomplete.

Mon

14:15–14:30

The Japan Meteorological Agency (JMA) operates the Coupled Prediction System version 4 (CPS4; Kubo et al., 2025) to support a wide range of seasonal forecasts. CPS4 consists of an atmosphere-land surface-ocean-sea ice coupled model. The atmospheric and land surface model have a horizontal resolution of TL319 (approximately 55 km) with 128 vertical layers up to 0.01 hPa, while the ocean and sea-ice model have a horizontal resolution of 0.25 degrees with 60 vertical layers. Various improvements implemented in January 2026 led to better seasonal prediction performance compared with the previous system. However, there is still room for improvement such as systematic biases in near-surface air temperature over land. We are currently developing global land surface model for the next seasonal prediction system (CPS5) to achieve further improvements in the prediction skills.

With respect to snow-covered regions in the Northern Hemisphere, CPS4 has systematic positive and negative biases in surface albedo over boreal forests and the other regions, respectively. These biases are associated with systematic negative and positive biases in near-surface air temperature over the corresponding regions. To address these issues, vegetation cover fraction (VCF) of tree canopy in the land surface model has been revised. It has been updated from adjusted VCF using the maximum green vegetation cover (Broxton et al., 2014) to the tree cover derived from the Copernicus Global Land Service (CGLS) Land Cover dataset. The revised VCF increases tree canopy coverage over snow-covered boreal forests and reduces positive biases in surface albedo. In addition, snow-related parametrizations in the land surface model have also been revised. The snow aging effect in the snow albedo scheme (Dickinson et al., 1986; Yang et al., 1997) and the formulation of snow cover fraction have been modified to represent higher albedo during the snow accumulation period. These revisions reduce negative biases in surface albedo over snow-covered regions. The combination of the revised VCF and snow-related parametrizations improves prediction skills, especially in near-surface air temperature for winter and spring across time scales. We will present recent developments in the global land surface model at JMA and their impacts on seasonal prediction performance.

Tue

16:15–16:30

Predicting the biophysical impacts of land cover change is a critical challenge for Earth system science and climate policy. This problem is fundamentally causal, as opposed to observational: we seek not only to predict how vegetation affects land surface temperature under observed conditions, but also to estimate what would happen under alternative land management scenarios. Changes in forest cover, agricultural expansion, or urbanization can alter energy and water fluxes, soil moisture dynamics, and local climate, making accurate modeling of these effects essential for both scientific understanding and policy evaluation. Traditional mechanistic models, while physically consistent, suffer from structural simplifications, unquantified uncertainties, and reliance on semi-empirical parameterizations. On the other hand, pure machine learning (ML) approaches excel at pattern recognition but struggle with extrapolation and often learn spurious correlations that break down under distributional shifts, such as those arising when interventional questions explore scenarios outside the training data. Hybrid approaches combining physics and ML have shown promise, yet many fail to preserve causal relationships between input features, limiting their ability to answer policy-relevant counterfactual questions.

We present a causal modeling framework integrating domain knowledge, ML, and causal inference for robust interventional modeling of land-atmosphere processes. Following recent work by Kekić et al. on causal digital twins for evaluating vaccine allocation strategies during COVID-19, we construct a temporal structural causal model encoding mechanistic relationships among 17 biophysical variables including land surface temperature, radiation fluxes, leaf area index (LAI), and atmospheric moisture. In the model, the graph structure is derived from expert knowledge on biophysical processes, while the functional relationships are learned using gradient-boosted trees (XGBoost) from a data cube composed of a mixture of ERA5-Land reanalysis estimations and MSG SEVIRI satellite observations aggregated to 10 km spatial and hourly temporal resolution, with modeling performed at grid-cell scale.

This approach yields a physics-aware model where perturbations propagate through the system according to known causal pathways rather than spurious statistical associations. The resulting causal digital twin supports prediction, forecasting, and crucially, interventional experiments where land cover modifications and their downstream effects on surface temperature and energy fluxes can be estimated.

We validate interventional predictions using two main approaches: comparison with eddy covariance flux tower measurements, and benchmarking against established hybrid and mechanistic models.
This work offers a promising approach for building robust, interpretable hybrid models to tackle interventional and counterfactual questions central to land management policy, climate adaptation planning, and Earth system science. By making causal assumptions explicit in the model structure, we provide a transparent framework for predicting land–atmosphere responses and performing scenario analysis at policy-relevant scale.

Mon & Tue

11:00–12:30

Understanding soil moisture–precipitation (SM-Pr) feedback remains challenging because the atmospheric leg is poorly constrained, and local SM signals are often masked by large-scale circulation and moisture advection. Sparse observational coverage and discontinuous records further yield coupling estimates that are highly sensitive to dataset choice and methodologies. This study investigates boreal-summer SM-Pr feedback over East Asia using the Integrated Global Radiosonde Archive version 2 (IGRA2), a quality-controlled 47-year (1979–2025) archive of vertical soundings. We apply the Convective Triggering Potential (CTP)-Humidity Index (HIlow) framework to characterize lower-tropospheric states in the early morning prior to convection. Afternoon-only rainfall events are classified into atmospherically controlled and land-controlled regimes by comparing cumulative SM distributions at each observational site, and further partition the land-controlled regime into wet (SM-Pr positive correlation), dry (SM-Pr negative correlation), and transition regimes. Results show a higher fraction of land-driven Pr in the north (>35°N; 25%) than in the south (≤35°N; 20%). Notably, the transition regime remains dominant under land-controlled conditions. Composites of afternoon-only rainfall events indicate enhanced northward moisture transport, consistent with stronger atmospherically controlled ratio in the south and more pronounced land-controlled ratio in the north. Furthermore, we evaluate the impact of multi-scale climate modes: ENSO (interannual), Monsoon (seasonal), and BSISO (intraseasonal). Across sites, the disparities in atmospherically controlled ratios between active and non-active climate phases are consistent with the corresponding composites of afternoon-only rainfall events. Meridional contrasts of the atmospherically controlled ratio are particularly pronounced during the Monsoon period, whereas ENSO and BSISO phases exhibit relatively modest latitudinal variations. These observational insights enhance our mechanistic understanding of SM-Pr feedback by delineating the spatial regimes where land-surface versus atmospheric drivers dominate afternoon rainfall. Overall, these findings establish essential observational benchmarks for evaluating and improving the representation of land-atmosphere coupling within climate models.

Mon & Tue

11:00–12:30

The aim of this study is to investigate the effect of land surface conditions on cloud occurrence by quantifying how they modulate the influence of large-scale meteorological conditions.
The land surface impacts clouds through its influence on surface turbulent heat fluxes, local moisture availability and surface roughness. However, quantifying these effects from observations remains challenging: the temporal and spatial variability of cloud occurrence is large, land surface signals are often confounded by meteorological influences, and the relevant drivers interact nonlinearly.
Here, we develop a process-oriented framework to quantify the contribution of land surface conditions to cloud occurrence over Europe (1983–2020) based on satellite observations. We employ a convolutional neural network (CNN) to predict satellite-observed cloud fraction from the CM SAF Cloud Fractional Cover dataset (COMET) using predictors derived from the ERA5 reanalysis, including ERA5-Land variables and atmospheric fields on single and pressure levels. To isolate the contribution of land surface conditions to cloud occurrence predictability, we develop two model configurations: one driven solely by large-scale meteorological conditions, and a second one that additionally incorporates land surface variables such as surface turbulent heat fluxes, near-surface temperature, soil moisture and near-surface winds.
Both models achieve high predictive skill (R² > 0.8), with a slight increase in performance when land surface conditions are included, indicating a small yet measurable contribution to cloud occurrence predictability. Sensitivity analyses based on permutation feature importance and partial dependence indicate that cloud occurrence is primarily controlled by large-scale meteorological drivers. Nevertheless, surface turbulent heat fluxes as well as near-surface temperature and moisture conditions emerge as the most influential land surface predictors. Partial dependence analyses further demonstrate a positive relationship between sensible heat flux and cloud occurrence, highlighting the role of flux-driven boundary-layer processes in promoting cloud formation.
This framework provides an observationally constrained approach to quantify the impact of land–atmosphere interactions on cloud occurrence. It further establishes a transferable methodology to assess how land cover change may modify cloud patterns and to investigate land–atmosphere influences across regions and timescales.

Mon & Tue

11:00–12:30

The phase change of water results in significant exchanges of energy during evaporation and condensation, making it a key process in both the global water and energy cycles. In contrast to rainfall—where droplets condense aloft and return to the surface under gravity—non-rainfall water inputs (NRWI; dew, fog, frost, rime, and vapor absorption/adsorption) form at or near the Earth’s surface. Individual NRWI events tend to result in exchanges of only small quantities of water compared to precipitation events, and therefore, they are not routinely measured or simulated. As a result, the contribution of NRWIs to the net exchange of water and energy remains largely unknown. However, because they occur in the biotic zone of the Earth’s surface, NRWIs are considered key to ‘habitability’, providing a continuous input of liquid water even in periods without rain and thereby potentially influencing soil hydraulic conductivity, microbial activity, and the reaction rates of biogeochemical processes. Thus, while quantitatively small, NRWIs may have a substantial impact on ecosystem processes.

In this work, we aim to stimulate debate about the lack of a macro-scale perspective on NRWI as a coherent class of phase-change phenomena occurring at the Earth’s surface. We synthesize existing challenges of NRWI measurements and modeling and highlight opportunities that arise from them. Our analysis provides a foundation for approaches that allow us to incorporate near-surface condensation into large-scale models more effectively.
First, we demonstrate how adsorptive forces of a wide range of natural materials, including soils, litter, deadwood, bark, stones, and living biomass, decrease the equilibrium vapor pressure by altering the potential of the liquid water retained within these materials under low moisture content. The assumption is that condensation occurs at dew point, which is equivalent to assuming that, e.g., soil and other materials are always at saturation (water potential = 0). Instead, porous materials commonly have water potentials substantially lower than 0, meaning that condensation occurs below saturated vapour pressure, and thus more frequently and over longer durations than is currently estimated. When atmospheric vapor pressure increases, these materials adsorb water until equilibrium with the surrounding vapor phase is established. We show that the effect of this material property is detectable within micrometeorological vapor flux measurements as vapor movement towards the soil, such as negative latent heat fluxes observed by Eddy Covariance. Low vapor pressure in the soil’s pore space, which regularly falls below atmospheric levels, is therefore a realistic boundary condition in many ecosystems during dry conditions. Second, we discuss the relevance of separating dew formation from adsorption, and whether it is possible to obtain this information at large spatial scales.

Based on these findings, we discuss a re-evaluation of recent decades of near-surface condensation research. Specifically, we argue for shifting the focus away from characterising NWRIs as separate processes, and toward synthesizing the observable effects that condensation processes collectively exert on measurements of vapor exchange between the land surface and the atmosphere.

We introduce a new framework grounded in the theoretical understanding of water potential gradients. With this framework, we show that while signals of NRWIs can be detected in observations at the soil column and ecosystem scales (i.e., lysimeters, Eddy Covariance, humidity profiles), they are not, or only incompletely, reproduced in gridded data products (i.e., ERA5 and FLUXCOM), or Land Surface Models (i.e., ecLand and Stemmus-Scope).
Our results show that signals of NRWIs are present in observations of diurnal vapor fluxes between land and atmosphere, but this effect is largely omitted in models. This omission is not only conceptual in nature but has implications for how we allocate water and energy fluxes, interpret ecosystem water use, and represent the coupling between land and atmosphere in different climate zones. When late afternoon and nighttime condensation is systematically neglected, models can skew estimates of evapotranspiration and surface energy cycling, and obscure mechanistic linkages between the atmospheric water availability and carbon cycles. Yet, we believe that the wealth of publicly available environmental data enables us to develop new approaches to bridge the current conceptual gap between definitions of NRWIs at the interface scale and the field or global scale. We would like to stimulate a debate about how atmospheric water vapor flux direction is treated in land surface models, the numerical challenges involved, and the potential consequences of overlooking frequent late afternoon and nighttime condensation across dry and wet conditions.

Mon

17:15–17:30

Terrestrial energy, water and carbon exchanges are regulated by the strength and sign of the coupling between the shallow land subsurface and the lower atmosphere. This coupling is essential for understanding weather patterns and long-term climate dynamics. Therefore, an accurate representation of these interactions in climate models is essential for obtaining reliable climate projections. This is particularly important for anticipating extreme events, such as droughts and heat waves. In the literature, various metrics and datasets are employed to study the land-atmosphere coupling across several spatial and temporal scales, including in-situ observations, remote sensing products, and Earth System Model outputs. Here, we evaluate different products and metrics showing that all data sources include large uncertainties that should be considered before drawing conclusions about the represented coupling in such datasets. In-situ measurements have limited spatial coverage and long-term consistency, remote sensing products are affected by retrieval uncertainties, and Earth system models contain structural uncertainties relative to the representation of coupling processes. By evaluating a set of CMIP6 models, we demonstrate that the structural difference surpasses the internal climate variability in terms of land-atmosphere coupling. More observational constraints in models would be necessary to restrict uncertainties. Meanwhile, to reduce uncertainties in observational products of land-atmosphere coupling, it is critical to collect long-term measurements at the land surface and implement more observational and physical constraints in the algorithms used to derive Earth observation (EO) products. More accurate, physically consistent EO products that accurately represent land-atmosphere coupling will, in turn, help develop future generations of Earth system models and extreme projections.

Wed & Thu

11:00–12:30

The Tropical Andes provide critical water supply for millions, yet their hydrological stability is
threatened by warming climate shifts in freezing level and precipitation phase balance. In these
high-mountain regions, precipitation quantification remains challenging due to sparse
monitoring networks and instrumentation limitations in resolving liquid, solid, and mixed phases.
This study evaluates a PARSIVEL 2 optical disdrometer at 4,709 m a.s.l. at CEMGEM, near
Huaytapallana glacier, Peruvian Andes. Despite measuring diameter and velocity to calculate
precipitation intensity at 1-min, high-altitude applications reveal significant microphysical
misinterpretations compromising regional climate records.
To diagnose these observational biases, we analyzed a comprehensive dataset spanning one year
of observations from November 2023 to November 2024. The instrument’s default manufacturer
algorithm originally identified seven precipitation types according to the SYNOP 4860 code: rain,
drizzle, drizzle with rain, snow, hail, soft hail, and mixed rain-drizzle with snow. Initial results
suggested that “drizzle with rain” was the most frequent state (30.4%), followed by snow 26.2%.
However, validation against an OTT Pluvio2 weighing rain gauge (used as the reference
instrument with a >0.25 mm threshold) revealed that these classifications were physically
inconsistent with the mass measured by the gauge. A total of 70 precipitation events ≥10 min
duration) were analyzed, uncovering a systematic and massive overestimation by the PARSIVEL
2. Specifically, the instrument exhibited a 98.6% overestimation for mixed-phase events (bias of
3.92 mm) and an 83.9% overestimation for solid precipitation events (bias of 7.09 mm), while
liquid-only events showed a minimal bias of 15.7% (0 mm). The greatest discrepancies occurred
during extreme events (>11.5 mm and >1 hour) dominated by snow, soft hail and hail. These
errors stem from a fundamental physical misinterpretation: the default internal algorithms
mistakenly identify large, low-density melting hydrometeors as high-velocity liquid drops,
leading to an exponential propagation of error in the mass-rate calculation.
To address these biases, we implemented a “Back to the Drawing Board” approach, discarding
the manufacturer’s classification in favor of a physics-based post-processing framework
optimized for high-altitude tropical microphysics. This methodology involved two critical steps.
First, we developed custom segmentation masks based on empirical V-D relationships (Locatelli
and Hobbs, 1974; Heymsfield and Wright, 2014) adapted for atmospheric pressure at 4,700 m
a.s.l. This filtered spurious detections artifacts with D > 10 mm but V < 1 m s⁻¹ and reclassified
particles into five categories: rain, snow, wet snow, graupel, and soft hail, enabling robust
apparent density estimation. Second, we recalculated precipitation intensity replacing the
default density assumption (near 1.0 g cm⁻³) with optimized bulk densities from reference gauge
data: 0.05 g cm⁻³ for dry snow and 0.12 g cm⁻³ for wet snow. We adopted the comprehensive
snow density range of 0.05–0.35 g cm⁻³ from literature rather than distinguishing individual
crystal unique, which would require supplemental imaging instrumentation.
The results of this correction show two key findings that bridge the gap between technical
observation and glaciological impact. First, the post-processing framework drastically improved
accuracy, reducing the solid precipitation Root Mean Square Error (RMSE) from 13.04 to 4.12
mm and refining the mixed-phase regression slope from 1.936 to 0.976, nearly matching the
reference gauge. Second, under the corrected classification, wet snow (29.7%) and graupel
(15.3%) emerged as the dominant precipitation types, whereas pure dry snow accounted for
only 3.8%. This shift is not merely a semantic adjustment; it represents a critical distinction for
physical process modeling in the Andes, where freezing levels frequently oscillate near the
surface. The dominance of wet snow, previously misidentified as rain or dry snow, indicates that
the region is experiencing a higher frequency of “near-melting” events than previously reported.
These findings profoundly impact land-atmosphere interactions and tropical glacier energy
balance. Wet snow dominance suggests significantly lower albedo (0.5–0.7) versus dry snow
(0.8–0.9), triggering positive feedback that accelerates glacier retreat. Uncorrected models
failing to distinguish wet snow likely underestimate melt rates and mischaracterize glacier
sensitivity to temperature fluctuations. By providing the first physics-based V-D calibration for
the Peruvian Andes, this study proves microphysical corrections are prerequisites for accurate
hydrological forecasting and glaciological predictions. This framework provides essential ground
truth for climate model parameterizations and monitoring network expansion across the Tropical
Andes vulnerable-ecosystems.

Mon & Tue

11:00–12:30

Earth system and weather forecasting models are moving to km-scale resolutions to provide more pertinent information to society on extreme events or the impacts of climate change. As some parametrized processes can be explicitly represented, increasing spatial resolution is expected to be beneficial for the atmosphere and oceanic components. It is not obvious that the same benefits will be achieved for land surface models (LSMs), as landscape organising processes start to play a role.

To evaluate the consequences of increasing resolution in LSMs, a set of atmospheric forcing data at 3km resolution is developed over a region covering all catchments flowing off the Pyrenees. It is shown that the high-resolution forcings can
capture the contrasts in atmospheric conditions between mountainous areas and valleys absent for coarser grids.
Six LSMs driven by these forcings and compared to their reference simulation at 50km resolution, displayed reduced evaporation over the semi-arid catchments within the domain. This cannot be explained by differences in the atmospheric forcing between resolutions. The cause has to be sought in the lack of spatial redistribution of water within the catchments. The observed diurnal amplitude of land surface temperature shows that the models do not reproduce the local minima along rivers and in irrigated areas caused by increased evaporation.

We conclude that at km-scale resolution, lateral transfers of water that organize landscapes play an important role to correctly predict evaporation. These flows can be neglected at few deca-kilometres resolutions. However at higher resolutions groundwater flows, riparian recharge and human water management for irrigation need to be simulated to represent realistic spatial contrasts in the surface fluxes which drive the atmosphere. We call upon the community to invest into the development of representation of these processes in our LSM so that they are ready for higher resolution applications.

Tue

15:30–15:45

Surface heterogeneity exerts a first-order control on land-atmosphere exchange processes by modulating the partitioning of energy and water fluxes across spatial scales. However, its representation in regional Earth system models remains a central challenge due to scale mismatches, nonlinear feedbacks, and subsurface-atmosphere connectivity. Here, we investigate the impact of explicitly resolved terrestrial heterogeneity on coupled system dynamics using the Terrestrial Systems Modeling Platform (TSMP2).

TSMP2 integrates the nonhydrostatic atmospheric model ICON, the land surface model eCLM (a fork of Community Land Model v5.0), and the integrated hydrological model ParFlow through the coupling framework OASIS3-MCT. This fully integrated configuration enables physically consistent simulation of water and energy transport from groundwater to the atmosphere, allowing heterogeneity in soil hydraulic properties, topography, vegetation structure, and land cover to dynamically influence atmospheric states. To isolate the contribution of surface and subsurface heterogeneity to coupled dynamics, simulations are performed with progressively increasing process complexity, ranging from atmosphere-only configurations to fully integrated groundwater-land-atmosphere simulations.

Results demonstrate that inclusion of lateral subsurface flow systematically reorganizes soil moisture patterns, amplifies spatial contrasts in latent and sensible heat fluxes, and alters boundary layer stability. We analyze differences in surface flux partitioning, soil moisture dynamics, and boundary layer characteristics across configurations to identify the sensitivity of coupled system behavior to representations of heterogeneity and cross-compartment connectivity.

Wed & Thu

11:00–12:30

Soil respiration generates CO2 that is at least partially deposited into the atmosphere. We briefly review the temporal and spatial scales of the currently available measuring techniques to measure the CO2 flux into the atmosphere: concentration sensors buried in the soil at multiple depths, gas capturing chambers placed on the soil surface, and eddy covariance towers. During periods of limited change in soil water content, diffusion drives the CO2 exchange at the soil surface. Especially during heavy rainfall events, rapidly infiltrating water forces CO₂ enriched soil air into the atmosphere, creating a convective flux. Evidence in the literature suggests that the amount of CO2 forced into the atmosphere can be substantial. However, it is difficult to measure this convective flux with the currently available measuring techniques: concentration sensors can only estimate the diffusive component, and chambers can create preferential conduits for soil gas by keeping the soil below them dry. We also found that a widely used numerical model for soil water flow includes diffusive CO2 flow, but ignores the convective flux. We will demonstrate how postprocessing of the model output can correct this without modifying the model code. We will also introduce an analytical model of the soil gas balance that incorporates production, diffusion, and convection. The model is parsimonious, computationally efficient, applicable to a range of spatial scales, and can easily be coupled with soil water balance models. Model results for several years will be compared to those of the more elaborate numerical model.

Mon & Tue

11:00–12:30

Fog remains a challenge for numerical weather prediction (NWP) as accurate forecasts rely on the representation of many interacting physical processes, especially over complex terrain where, for instance, radiative cooling varies sharply with slope and shading due to the heterogeneity of the terrain. Radiation fog, primarily generated by radiative cooling, is generally considered the most common type of fog over land, especially in valleys or basins where cold air can settle.
Land surface properties—including land use, thermal roughness, albedo, snow depth, and soil characteristics, together with the land surface model—play a critical role in controlling the onset and duration of radiation fog over land. Motivated by this, the present study investigates the influence of soil moisture on radiation fog formation using high resolution mesoscale simulations (dx = 500 m) for the Southern European Alps. Soil moisture strongly influences land–atmosphere exchange by modifying soil thermal conductivity and turbulent boundary-layer mixing, two key parameters that govern near-surface cooling and saturation. Wet soils exhibit higher thermal conductivity and enhanced evapotranspiration, allowing heat to be transferred more efficiently into deeper layers while diverting more energy into latent rather than sensible heat flux. This combination cools the surface, weakens turbulent mixing, and limits near surface warming, thereby favouring saturation. Conversely, dry soils have low thermal conductivity and restrict evaporation, causing stronger sensible heat fluxes and deeper daytime mixing, while simultaneously reducing atmospheric moisture availability needed to reach saturation. At night, dry soils cool rapidly and reinforce stability, but the limited humidity still inhibits fog formation. Through these combined effects, soil moisture emerges as a dominant control on how surface energy is partitioned, how vigorously the lower atmosphere mixes, and ultimately how readily radiation fog forms and persists in basin environments.
A set of numerical experiments have been designed to reproduce a wintertime stable radiation fog event observed in the frame of the TEAMx observational campaign (TOC). The simulations were implemented in the ICON model using two one-way nested domains. The sensitivity of the model to soil moisture in reproducing the fog layer is assessed through three numerical experiments. The first experiment serves as a control simulation, in which the model is initialized using the ICON Data Assimilation Coding Environment (DACE) dataset to drive the model, followed by one drier and wetter sensitivity experiments.
Observations of terrestrial water storage suggest that soil moisture variability across Europe is on the order of ±40%. However, models do not capture this variability, showing only ±10% variability in high-resolution climate studies. Even at kilometre scales, soil moisture is still a poorly constrained variable, often underestimated or oversmoothed because soil-moisture analyses are coarse, meaning the variability is inherited from a much lower resolution product.
The sensitivity of the model to an initial soil-moisture anomaly, defined as an intentional modification of the starting soil-moisture field in the simulation, is investigated. The initial soil-moisture content is perturbed by ±30% across the entire model domain, following previous investigations of the effect of soil moisture in the development of valley winds. One experiment is initialised with 30% wetter soil than in the control simulation, while a second experiment is initialised with 30% drier soil. This perturbation is applied uniformly at all grid points and throughout the entire soil column. If the +30% anomaly causes soil moisture to exceed the soil’s water-holding capacity, the model automatically converts the excess water into runoff, thereby preventing unrealistic soil saturation.
The analysis will therefore provide insight into the role of soil-moisture conditions in modulating the surface energy balance and boundary-layer cooling that control radiation fog formation and persistence in complex terrain.

Thu

09:45–10:00

Accurate monitoring of water availability in the root zone is a prerequisite for many applications such as understanding land-atmosphere feedback. This work evaluates the performance and physical consistency of two distinct modelling paradigms to retrieve Root-Zone Soil Moisture (RZSM) in Mediterranean vineyards (Catalonia, Spain), serving as a benchmark for hybrid modelling strategies.

We contrast a purely data-driven method (Multilayer Perceptron, MLP) against a process-based approach coupling a parsimonious multilayer soil model (multilayer FAO-56) with an Ensemble Kalman Filter (EnKF) for the assimilation of Surface Soil Moisture (SSM).

Initial evaluations at the plot scale, using in-situ observations and standard meteorological forcing, highlight a critical trade-off. Both the MLP and the process-based model demonstrate similar predictive performance in capturing local moisture dynamics. However, while the process-based model ensures physical consistency, its domain-wide application is often hindered by the need for complex parameter calibration in unmonitored areas.

To address this limitation, this study explores the upscaling of RZSM retrieval from a few monitored plots to broader vineyard areas by integrating high-resolution remote sensing SSM. We hypothesize that the data-driven approach offers greater generalizability across different vineyards where specific physical parameters are poorly constrained. Ultimately, we aim to explore how data-driven flexibility and physical constraints could be combined to achieve robust water accounting across unmonitored areas.

Wed & Thu

11:00–12:30

Persistent non closure of the surface energy balance from eddy covariance (EC) measurements remains a major challenge for quantifying land–atmosphere exchange. The soil water balance method estimates evapotranspiration (ET) by measuring changes in soil water storage and seepage or capillary rise rates. This method provides an independent, process based benchmark for EC estimates. We applied the soil water balance method to a maize stand at the land atmosphere feedback observatory (LAFO) in Stuttgart, Germany. We compare 30 min EC latent heat fluxes with ET derived from soil moisture and matric potential dynamics during selected rain free periods in 2025. Observations include volumetric soil moisture profiles, matric potential time series, mini-lysimeter measurements of soil evaporation, and site energy budget variables, including soil heat flux, measured within the Land–Atmosphere Feedback Initiative (LAFI; FOR 5639). Rain free intervals were chosen in April (bare soil) and June/July (vegetated) with multi-day durations to reduce the noise in the water balance terms. For each interval, we compute cumulative ET from the observed water balance terms and compare these totals with ET derived from EC data. First results will be presented, showing the comparability of soil balance ET and EC derived ET, and highlighting methodological considerations for using the soil water balance method as an independent reference at LAFO.

Mon & Tue

11:00–12:30

Land surface models play a crucial role in weather and climate predictions by providing key prognostic states that influence atmospheric dynamics. In this study, we train a multilayer perceptron (MLP) to emulate the key prognostic states of the Integrated Forecasting System (IFS) land surface scheme, ecLand, from the European Centre for Medium-Range Weather Forecasts (ECMWF). The MPL model, named aiLand, is evaluated at both continental and global scales. As a single-column model, aiLand can be trained and tested across regions with different climate conditions, allowing us to assess its response to “out-of-sample” climatic conditions. Additionally, aiLand is resolution-agnostic, allowing it to be applicable at various spatial scales. To further enhance its predictive capability, we fine-tune aiLand using observational data, exploring whether this approach can uncover structural deficiencies in the physics-based ecLand model. This work is part of the Destination Earth programme, which aims to build machine-learned components for Earth system modelling.

Tue

15:45–16:00

NASA recently released Version 3 of its Subseasonal-to-Seasonal (S2S-3) ocean-atmosphere-land prediction system. S2S-3 is based on the Goddard Earth Observing System (GEOS) modeling and data assimilation framework. The initial conditions of the S2S-3 forecasts are informed by ocean observations (via direct assimilation) and by atmospheric observations (via “replay” of the atmospheric analysis from the GEOS weather prediction system). Additionally, precipitation observations are used to drive the land surface model in the S2S-3 analysis, as in the GEOS “MERRA-2” reanalysis. The observational precipitation products used in MERRA-2 and S2S-3 are primarily based on gauge observations. Neither MERRA-2 nor S2S-3 assimilates land surface observations.

The objective of this study is to quantify the potential of satellite observations of soil moisture and precipitation to improve S2S-3 land forecasts at subseasonal (2-week to 2-month) lead times. Given the intractability of running multiple ensemble experiments with the full S2S-3 system, our historic ensemble forecast experiments instead utilize an offline, land-only modeling system. It is important to note that each of the offline forecasts is fair, making no use of observations after the forecast start date. The meteorological forcing used to drive the offline land model in our experiments comes directly from forecasts produced by the full S2S-3 system, but with forecast meteorological biases corrected, as a function of lead, using established procedures.

The study exploits the advances in the land-only Soil Moisture Active Passive (SMAP) Level 4 Soil Moisture (L4_SM) system, which assimilates SMAP L-band (1.4 GHz) brightness temperature observations and drives the land surface model with state-of-the-art precipitation from the Integrated Multi-satellitE Retrievals for the Global Precipitation Measurement (IMERG) products. Here, the L4_SM system is adapted to generate four sets of historical, land-only, subseasonal ensemble forecasts, each utilizing a distinct set of initial conditions (ICs): (i) Forecasts produced with land ICs mimicking those of the full S2S-3 system, serving as the control; the ICs reflect antecedent precipitation forcing informed by gauge measurements and by surface meteorological forcing (air temperatures, winds, etc.) from MERRA-2. (ii) As in (i), but with the precipitation underlying the ICs instead derived from satellite- and gauge-based IMERG precipitation data. (iii) As in (i), but with the ICs taking advantage of SMAP data assimilation. (iv) As in (i), but with the ICs making use of both IMERG data and SMAP data assimilation as in the L4_SM algorithm. The ICs used in these four forecast experiments are derived from long-term, land-only simulations configured to utilize the ancillary information indicated in the list above.

By analyzing the skill of the forecasts initialized with information from IMERG and/or SMAP relative to the skill of the control forecasts, we isolate and quantify the impact of the IMERG and SMAP data on the S2S forecast skill of soil moisture, streamflow, and ET. Results indicate that the skill of the subseasonal soil moisture forecasts, when verified against independent satellite observations, indeed increases when the forecasts are initialized with SMAP and IMERG information. As expected, the positive contribution to the forecast skill diminishes with increasing lead time. Geographically, the improvements in soil moisture forecast skill are largest in otherwise data-sparse regions, particularly in the Southern Hemisphere. In South America and Australia, the use of SMAP and IMERG information increases the anomaly correlation coefficient of soil moisture forecasts by 0.05-0.1 at 1–2-week lead time and by 0.02-0.05 at 5–6-week lead time, relative to the control experiment. In Africa, smaller but consistently positive skill improvements from using SMAP and IMERG information are seen at 1-2- and 3-4-week lead times. In North America and Eurasia, skill improvements are seen only at 1–2-week lead time, owing to the generally higher baseline skill of the control experiment owing to the denser precipitation gauge coverage. For the same reason, streamflow forecast skill measured against in situ measurements in the continental United States is only marginally improved by using SMAP and IMERG for forecast initialization. Finally, verification of the subseasonal forecasts against analysis estimates from the SMAP L4_SM product shows consistent improvements in the soil moisture, streamflow, and evapotranspiration skill out to lead times of 5-6 weeks.

Ongoing work aims to determine if SMAP and IMERG information also enhances the forecast skill for carbon-related fluxes such as gross primary productivity.

Wed & Thu

11:00–12:30

Accurately representing land–atmosphere energy partitioning remains a central challenge in land surface modeling. The division of net radiation into latent (LE) and sensible heat (H) governs boundary-layer development, atmospheric heating over land, and land–climate feedbacks. Yet widely used observational benchmarks rely heavily on eddy-covariance measurements that suffer from systematic energy imbalance, limiting their suitability for model evaluation.

Here we develop a multi-scale, process-oriented benchmark for land energy partitioning by integrating catchment-based water and energy balance constraints, satellite radiation products, and eddy-covariance–based flux data for 2001–2020. The resulting constraint indicates substantially higher sensible heating, and thus higher Bowen ratios, than represented in most CMIP6 models. Only a subset of models reproduces present-day partitioning within observational uncertainty.

To disentangle forcing biases from land-surface process representation, we employ machine-learning emulators trained separately on observations and model output. These diagnostics demonstrate that discrepancies primarily arise from how models translate radiative forcing into surface fluxes, rather than from biases in precipitation or net radiation. Models consistent with present-day constraints exhibit a stronger increase in Bowen ratio under future forcing, implying amplified routing of additional energy into sensible heating over land.

This framework illustrates how combining independent hydrological constraints with machine-learning diagnostics can provide a process-level benchmark for land–atmosphere coupling. Such constraints are critical for advancing next-generation land surface models and for improving confidence in simulations of land atmospheric heating and feedback dynamics under climate change.

Mon

14:45–15:00

Since 1972, KNMI has been carrying out a continuous measurement program for atmospheric research at Cabauw. Differentiating aspects include the unique completeness of the program, the high quality of the measurements, and the continuous nature of the time series of the data. The measurement program has evolved with increasingly advanced measurement techniques and instruments, often in collaboration with other institutes and universities. This has created a research station that is unique (in Europe and the world) where the state of the atmosphere is recorded in great detail. The Cabauw station is managed by KNMI and the observation programme is carried out with its partners in the consortium Ruisdael Observatory.

Since 2025 an investment programme has started called FTO Duurzaamheid which amongst other aims at future-proofing the Cabauw station so that the facility can be used sustainably to address societal developments and policy themes. Within the framework of FTO Duurzaamheid, it is anticipated that new measuring equipment and upgrades and appointment of associated personnel to commission and deploy the equipment are necessary. We shall also invest in providing access to data via FAIR and OpenData policies. Finally, investments are needed to upgrade the infrastructure such as electrical power supply and network cabling for future-proof deployment of equipment, as well as use of the facility by third parties and in measurement campaigns.

In the presentation an overview will be given of the investments that are being done within the framework of FTO Duurzaamheid with a focus on the role that the Cabauw facility can play in international networks such as ACTRIS, ICOS, GRUAN, GSRN, and research initiatives such as GLASS and other projects that are performed within the framework of GEWEX.

The current remote sensing capabilities for aerosols, clouds and water vapourwill be enhanced to improve the temporal and spatial resolutions and to include temperature profiling in the boundary layer and beyond. By combining these remote sensing capabilities with the suite of in-situ observations that focus on the land-atmosphere exchange we will work towards an enabling of the use of Cabauw data for use in the GLAFO concept.

Wed & Thu

11:00–12:30

Recent years have seen an uptick in the occurrence of “weather whiplash” events, or rapid shifts between opposing weather conditions, that have led to detrimental outcomes for agriculture, water resources, and human lives. Weather whiplash often co-occurs with rapid shifts in vegetation states that seed the conditions for catastrophic outcomes. “Dry-to-wet” transitions (i.e., severe droughts followed by intense rainfall) may result in flash flooding, while “wet-to-dry” transitions (i.e., wet conditions quickly giving way to extreme drought) may provide ample dry fuel for wildfires. As sudden changes in vegetation state can amplify the land-surface response to extreme weather events, there is a need to characterize vegetation state transitions towards better predictions of flash flood and wildfire. Recent research investigated the event characteristics of rapid transitions in vegetative states through sudden shifts in phenology from vegetation die-back to leaf-out or visa-versa (i.e. “vegetation whiplash”). Rapid transitions in phenologic state are important to understand as they are directly incorporated into Land-Atmosphere (L-A) coupling through the partitioning of surface heat fluxes and estimations of surface temperature. Our hypothesis is that vegetation plays a key role in the L-A coupling that shapes the evolution and severity of extreme events, particularly related to pulse changes in atmospheric forcings that are currently not captured in forecast models. In this study, we compare observations of surface temperature, latent and sensible heat fluxes (FLUXNET) to modeled values during vegetation whiplash events. Preliminary results demonstrate that surface fluxes from land surface models behave linearly and lack the hysteresis response seen in flux tower observations during rapid transition of vegetation. Adding dynamic vegetation did not change the linear response of surface fluxes in Noah-MP land surface models for some locations. This indicates that current land surface parameterizations cannot replicate these dynamic transitions in the land surface. The extent that these results impact the global energy balance, as well as their role in improving land surface processes in climate models towards improved predictions of extreme events is discussed.

Wed

15:00–15:15

Irrigation is the targeted application of water to the land with the aim of sustaining plant productivity, particularly for agricultural crops, during periods of limited water availability. Additionally, water application can be regulated by human interventions through management practices and government licensing. This results in huge spatial and temporal heterogeneity which is an ongoing challenge to simulate in land surface models.
Over 324 million hectares of land are equipped for irrigation worldwide. 42% of this is in only two countries: India and China (FAO 2014). These areas of high irrigation also coincide with hotspots in land atmosphere coupling strength (Koster et al. 2004), which highlights the importance of irrigation for land atmosphere coupling through the sensitivity of the atmosphere to soil moisture.
The aim of this work is to demonstrate and understand the impact of irrigation on the terrestrial water fluxes, surface energy fluxes and atmospheric fields. This presentation outlines the work that has been done to develop a new irrigation scheme for JULES which is suitable for future requirements. The current fraction-based irrigation scheme does not give us the necessary flexibility to enable future developments, in particular the ability to irrigate specific surface types, distinguish between different irrigation methods and run with and without tiled soils. Exploiting soil tiling will allow us to have the correct physical representation of irrigation. However, due to technical reasons, the full soil tiling functionality is not available yet. This presentation will therefore focus on an interim solution that uses a single soil water profile with irrigated C3 and C4 grass tiles. Initial results from single point experiments show that the impact of including irrigation is an increase latent heat flux and screen level humidity, which cools land surface temperatures. Results and evaluation from global and regional simulations coupled to the Unified Model will be shown and future priorities for irrigation will be presented.

Thu

09:45–10:00

Characterising land surface-atmosphere interaction processes and monitoring the physical state of the lower atmospheric boundary layer (ABL) and the underlying surface is the main focus of the measurement program at the Falkenberg boundary layer field site (in German: Grenzschichtmessfeld, GM) of the Meteorological Observatory Lindenberg – Richard – Aßmann – Observatory (MOL- RAO, DWD). The site is embedded in rather slightly structured orography, it is surrounded by a mixture of (mainly pine) forests and agricultural farmland with small lakes and villages, typical for a rural area in North-East Germany. The extensive long-term measurement program at GM Falkenberg started in 1998, it comprises measurements of atmospheric profiles at towers up to 99 m height, soil parameters under grass and fallow surfaces down to 1.5 m depth, short-/longwave radiation budget components, and of the turbulent energy and momentum fluxes at several heights. Large-aperture optical and a microwave scintillometer are operated for the estimation of area-averaged sensible and latent heat fluxes over the heterogeneous landscape around GM Falkenberg. Beyond the tower height the vertical profiles of selected atmospheric state and process quantities are complemented by sodar / RASS (till July 2023) and infrared Doppler Lidar remote sensing measurements (since 2014). Over the whole time period, measurements were carried out with a minimum of technical changes trying to achieve a maximum of temporal homogeneity concerning sensors, setup, and data acquisition. Complemented by free-troposphere information from radar wind profiler / RASS measurements and from radiosoundings four times a day at the Lindenberg observatory site (5 km to the North of GM Falkenberg), a long-term quality-controlled data set covering a wide range of meteorological and environmental situations has been generated, including extremes such as heat waves, storms, cold winter periods, dry and wet soil conditions. The data is used in DWD for routine and process-oriented validation of numerical weather prediction models.
The contribution will give an overview of the measurement program and its data coverage. We will demonstrate statistical aspects of various long-term time series of the measured variables. Some of the efforts undertaken to correct individual time series to account for changing conditions or technical discontinuities will also be presented.

Wed & Thu

11:00–12:30

From heightened canopy dieback to tree die-off, many forest ecosystems are showing signs of poorly coping with more severe, more frequent, or hotter droughts. Understanding forest resilience to drought has become paramount, and eco‐physiological optimisation approaches that test behavioural hypotheses have been proposed as a means to build this understanding in global terrestrial models. Here, we used a land-surface modelling framework that considers competing optimality principles to simulate root water uptake dynamics, canopy gas exchange, and leaf nitrogen investments into the photosynthetic apparatus, whilst also accounting for sustained impairments to both soil-to-root and plant hydraulics.

We applied this model to a pristine observational site of the Caatinga, Brazil’s drought-hardy, seasonally deciduous, and exceptionally diverse dry tropical forest. Six woody species dominate 80% of the study area whilst displaying contrasting hydraulic and photosynthetic functional strategies. For example, their respective P50s (the water potential at which 50% of a plant’s hydraulic conductivity is lost) range between ca. -1 MPa and -5 MPa. Model predictions were first compared with species-specific leaf-level observations of water potential, stomatal conductance, and photosynthetic uptake to assess functional realism. They were further probed against eddy covariance measurements of ecosystem carbon, water, and energy fluxes spanning a period with high interannual rainfall variability (and including a severe multi-year regional drought).

We found that none of the six species could, in isolation, explain the magnitude and dynamics of the observed surface ecosystem fluxes. However, taken together and accounting for their relative contribution to total ecosystem fluxes, they did. Further, our analysis emphasises the vital role of phenology in mitigating seasonal and inter-annual hydraulic risks, with foliage reductions triggered by a 10 to 20% loss of hydraulic conductivity in the canopy. On the whole, accounting for diverging species-level responses and their relative influence at the ecosystem-scale appears key to improving model predictions in functionally diverse forests.

Tue

09:00–09:15

The increasing frequency and severity of heatwaves over India demand improved understanding of their governing mechanisms and precursors. Although numerous studies have documented heatwaves across the Indo-Gangetic Plains (IGP), process-oriented analyses distinguishing between dry heatwaves (DHWs) and moist heatwaves (MHWs) remain limited. In particular, the relative roles of remote heat advection versus local land-atmosphere interactions, and the processes that constrain heatwaves to localized spatial extents despite large-scale anticyclonic circulation, needs to be studied more. This study addresses these gaps by examining (i) the contribution of heat advection to heatwave development, (ii) the mechanisms responsible for their spatial confinement, and (iii) the antecedent land and atmospheric conditions that precondition heatwave onset and persistence.
We analyze ten major pre-monsoon heatwave events using Eulerian temperature decomposition and anomaly diagnostics relative to climatology. By comparing heatwave-affected pixels with adjacent non-heatwave regions prior to onset, we isolate the processes that differentiate DHWs and MHWs. Our results show that horizontal heat advection in the lower free troposphere, near-surface layer, and boundary layer is relatively weak during heatwave buildup. This finding challenges the conventional paradigm that emphasizes remote northwesterly heat transport as the primary driver. Instead, quasi-local diabatic and adiabatic processes dominate temperature tendencies, operating through persistent diurnal amplification within the boundary layer.
MHWs are typically preconditioned by antecedent pre-monsoon precipitation that moistens the land surface and lower troposphere. Enhanced soil moisture increases daytime latent heat flux, elevates atmospheric humidity, and supports shallow convection and cloud formation. At night, low-level shallow clouds reduce outgoing longwave radiation, promoting radiative warming and sustaining high wet-bulb temperatures. Concurrent upper-level subsidence associated with anticyclonic circulation stabilizes the column, suppresses deep convection, and reinforces moisture accumulation within the boundary layer. The repetition of these coupled processes leads to sustained moist heat stress.
In contrast, DHWs arise under dry antecedent conditions characterized by suppressed precipitation, reduced cloud cover, and enhanced solar insolation. Limited soil moisture constrains latent cooling and shifts surface flux partitioning toward sensible heat flux, promoting rapid daytime warming and the development of a deep, dry boundary layer. The absence of significant moisture advection and cloud feedbacks allows efficient heat accumulation through repeated diurnal cycles.
Overall, our findings demonstrate that Indian heatwaves are primarily regulated by local thermodynamic feedbacks involving soil moisture, boundary layer evolution, clouds, and precipitation rather than by strong remote heat advection. The distinction between DHWs and MHWs emerges from differences in antecedent moisture states and their modulation of land-atmosphere coupling, which ultimately control heatwave intensity, persistence, and spatial structure.

Mon & Tue

11:00–12:30

The Mediterranean region, characterized by its water-limited climate, is particularly vulnerable to hot and dry summers. As climate change progresses, rising temperatures and decreasing soil moisture contribute to an increasing frequency of compound hot and dry extremes. The drying of root-zone soil moisture leads to a decrease in the rates of latent heat flux, causing an increase in sensible heat flux, which creates a robust positive feedback loop wherein warmer maximum air temperatures further exacerbate soil dryness. This feedback mechanism reinforces the existing water-limited regime of the land-atmosphere system and accelerates the onset of hot and dry extremes. Prior to these extreme events, the increase in both temperature and soil drying intensifies land-atmosphere coupling. This heightened coupling serves as a critical threshold, where the sensitivity of atmospheric temperature to declining soil moisture increases significantly. When soil moisture drops below this critical point, the atmosphere becomes increasingly susceptible to heatwave conditions.
This study investigates the temporal dynamics of land-atmosphere coupling and its role in amplifying hot and dry extremes. We analyze daily data from the ERA5-Land reanalysis, focusing on daily maximum air temperature and root-zone soil moisture (0–100 cm depth) from 1950 to 2025, specifically during the extended summer season spanning May to September. To elucidate the mechanisms driving this intensification, we employ segmented regression analysis applied to maximum temperature and root-zone soil moisture to identify the critical threshold at which rapid intensification of land-atmosphere coupling is triggered. Furthermore, we evaluate this using well-known past extreme events to gain a deeper understanding of how this spatiotemporal evolution occurs across the Mediterranean region.
Our analysis highlights that a significant portion of the Mediterranean region is currently experiencing rapid intensification of land-atmosphere coupling. The examination of various case events supports our findings, suggesting that this intensification poses substantial implications for the frequency and intensity of future heatwaves. Understanding these dynamics is crucial for predicting and mitigating the impacts of climate change on both ecological and human systems in this vulnerable region.

Mon & Tue

11:00–12:30

Spatially and temporally continuous observations of atmospheric boundary layer (ABL) dynamics and thermodynamics are key for a better understanding of land surface atmosphere coupling. State-of-the-art observational approaches are either limited by vertical resolution (satellites), spatial coverage (ground-based profilers) or spatial and temporal coverage (radiosondes). Thus, spatial-temporal ABL patterns will only be resolvable through sensor synergy. The VITAL II campaign (Vertical Profiling of the Troposphere: Innovation, Optimization and Application, Löhnert et al., 2025) aims at taking a first step in closing this observational gap. VITAL II is part of HErZ – the Hans Ertel Center for Weather Research, a cooperation on fundamental weather and climate research between the German Meteorological Service (DWD) and German universities.
VITAL II will take place from June 1 – August 31, 2026 in the Cologne Bay region between the west German cities of Cologne, Bonn and Aachen employing and installing seven profiling sites over a wide variety of land surface types. At the profiling sites, ground-based remote sensing systems such as water vapor lidar, Doppler lidar and microwave radiometers will be operated which yield high potential to reliably profile ABL temperature, humidity, winds and turbulence in a nearly continuous manner. Uncrewed aircraft systems (UAS) will additionally be operated at selected sites during an intensive observation period. A large number of radiosondes will complement the profiling observations.
During VITAL II, first data from the Meteosat Third Generation Sounder (MTG-S1) satellite instrument IRS (Infrared Sounder) will become available providing continuous 3D observations of temperature and humidity over large parts of Europe and Africa with a temporal resolution of ~30 min. However, IRS deficits will remain in observing the ABL, especially in cloudy conditions. VITAL II will leverage the use of these novel, hyperspectral IRS observations by combining them with the VITAL II surface-based in-situ and remote sensing observations. Novel machine learning algorithms which synthesize MTG-S1 data with surface-based observations will be applied to and assessed by the multitude of additional VITAL II profiling observations. The objective is to significantly enhance the observed information content in the ABL. The profiling observations will be used for data assimilation studies with the DWD numerical weather prediction model ICON and for evaluating and improving ICON land surface and ABL turbulence parameterization schemes.
Next to the extensive profiling, VITAL II will install a dense near-surface observation network on the meso-beta-scale (20–200km = regional scale). Up to 50 surface stations of the updated autonomous cold pool logger (APOLLO 2.0) will be installed within Cologne Bay. This observational setup, together with the synergistic profiling approaches, will be used for enhancing the understanding of the evolution of the stable and convective ABL over an urban – rural transition zone, as well as on convective cold pools. For the latter, the main focus is on merging dense spatial, near surface data with vertical temperature and humidity patterns.
This contribution will present in detail the VITAL II concept and ideally, first measurements and results of VITAL II highlighting the importance of sensor synergy for ABL coverage.

References
Löhnert, U., Ament, F., Platis, A., Sakradzija, M. et al. (2025). VITAL II Concept Paper: Vertical Profiling of the Troposphere: Innovation, Optimization and Application II. Zenodo. https://doi.org/10.5281/zenodo.17424652

Wed & Thu

11:00–12:30

Hydrological modelling is increasingly adopting hyper-resolution approaches in an attempt to provide more accurate and high-resolution representations of local hydrological dynamics. However, increasing the resolution of a distributed regional hydrological model requires to establish new parametrisation as well as revisit current calibration strategies as they often neglect to capture localised dynamics. This is particularly the case in human-modified catchments where calibration of hydrological dynamics is often not feasible due to the interruption of natural flows and water cycles by reservoir storage and releases as well as irrigation. Introducing additional parametrisation therefore poses an additional risk of overfitting the model when calibrating stations downstream of reservoirs.
In these basins the availability of water resources is not only closely linked to ecosystem functions but also agriculture and electricity production and a shift in water resources will have a substantial socio-economic impact. Therefore, established canals, dams and reservoirs regulate the local water cycle. Current common calibration strategies for fully distributed high-resolution hydrological models ( ≤1km grid cells) such as the Community Water Model (CWatM) can lead to overfitting, and potentially to erroneous assumptions about water resource availability for future climate scenarios within these heavily modified catchments. For understanding the impacts of climate change on anthropogenic catchments and to adequately simulate water availability under climate change pressure, new calibration strategies need to be found without overfitting the model.
Addressing this question is crucial, as these catchments rely heavily on water resources as socioeconomic driver and therefore more accurate modelling will impact the adoption of long-term, future-proof land use policy decisions and communities’ livelihoods.
Here we explore and compare the effects of running and calibrating a hydrological model for the Ebro basin in Spain with high resolution meteorological data under different strategies: a traditional single-station discharge approach, based on an evolutionary algorithm, a weighted calibration which takes into account the scores both at the outlet and at subbasin stations, as well as a regionalization approach that ties the model parameters to the physiography in natural sub-catchments and applies the relationship everywhere, as tested by Cenobio-Cruz et al. (2023), avoiding that human processes affect the calibration of physical processes. Finally, the metrics used to calibrate will be compared: the composite KGE will be compared to its non-parametric counterpart (Pool et al., 2018)
The results provide insights to the current debate on additional parametrisation and calibration strategies in hyper-resolution hydrological modelling. It will address the complexities associated with calibration strategies for heavily human impacted basins and the uncertainty of the adequacy of calibration results to represent the local water cycle. Insights from this research can enable water-allocation decision makers understand shortcomings of traditional calibration strategies and find trade-offs for an adequate depiction of future water availability under climate change scenarios with the least possible complexity.

Mon & Tue

11:00–12:30

Mountain glaciers are vital components of the global climate system, playing a crucial role in regional hydrology, energy balance, and atmospheric dynamics. These systems are highly sensitive to climate change, and small-scale processes, such as localisedfin thermodynamic adjustments, can trigger rapid feedback mechanisms that significantly alter large-scale atmospheric conditions. Observing and directly interpreting these adjustments is challenging due to nonlinear and often obscured cause-and-effect relationships mediated by intermediary steps. This complexity limits the predictability of meteorological and cryospheric phenomena in mountainous regions. Addressing these challenges requires a holistic analysis approach, without relying on assumptions of linearity or simple correlations. To overcome these obstacles, we employ high-resolution numerical atmospheric simulations to study interactions between glacier microclimates and the free atmosphere, as well as the feedback effects that emerge across scales. Using transfer entropy, we uncover the causal relationships driving these feedbacks, identify directional influences between mass and energy fluxes, and analyse how localised processes propagate across micro-, meso-, and synoptic scales. In addition, we perform a systematic glacier-area sensitivity experiment using simulations with different prescribed glacier extents. This allows to quantify how variations in glacier coverage modify the valley-scale atmospheric energy balance and the associated turbulent and radiative fluxes. For example, our analysis reveals how changing glacier geometries affect the microclimates and regional energy balances, driving mesoscale atmospheric circulation patterns. By combining causal inference from transfer entropy with the glacier-area sensitivity analysis, we identify the mechanisms through which glacier retreat alters energy exchanges between the surface, the valley atmosphere, and the overlying free atmosphere. This presentation highlights key insights from these simulations, including the quantified sensitivity of the valley energy budget to glacier area, and discusses the role of glacier–atmosphere interactions in shaping elevation-dependent warming and energy flux dynamics.

Wed

14:45–15:00

Numerical weather prediction (NWV) as well as climate models require external parameter data sets which prescribe the state of the underlying land surface, such as e.g. soil types or land-use cover. In recent years, a range of new high-resolution datasets derived from satellite data together with machine learning has been released. In this study we investigate the impact of replacing the outdated coarse FAO dataset to a fine-scale contemporary dataset based on SoilGrids. In addition to changes of the soil texture, the soil hydraulic model is updated. Both global simulations and regional high-resolution simulations within an Alpine domain in an NWP setup using the ICON model are done. While the global impact of the described changes is moderate, the influence locally can be considerably larger. In addition, local adaptions to the parameters display a strong sensitivity

Wed & Thu

11:00–12:30

Eddy covariance (EC) measurement sites are often located in heterogeneous terrain where aggregated ecosystem-exchange fluxes are observed originating from a mosaic of structured patches of different land cover types and mixed ecosystems, which may even exhibit sources and sinks simultaneously. This complex spatial heterogeneity makes it challenging to identify controls and drivers governing carbon cycle processes of homogeneous sub-units surrounding the tower.

We present FLUGS, a novel framework that infers land-cover-specific ecosystem-exchange fluxes provided the EC time series of aggregated fluxes and the land cover map of the ecosystem surrounding the EC tower. The FLUGS framework combines a multitask machine learning approach with high-resolution flux footprints computed with the Boundary Layer Dispersion and Footprint Model (BLDFM), a fast numerical solver of the advection-diffusion equation. The framework learns the environmental response functions (ERFs) from EC data for each land cover class simultaneously. FLUGS is validated against synthetic and real data experiments. The latter uses data from a twin tower site in Northeast Siberia and the STORDALENX25 campaign. The approach is versatile, robust to multicollinearity and yields interpretable ERFs with a unique global optimum. Machine learned patch-level ERFs from FLUGS may be used directly for upscaling. Applying spatial flux decomposition with FLUGS to a single tower effectively multiplies its scientific value, providing land-cover-specific insights equivalent to operating two or more or X conventional towers, one for each patch type individually. By offering a fast, transparent workflow for spatially decomposing ecosystem fluxes, FLUGS opens new opportunities to attribute EC fluxes to ecological processes, benchmark land-surface models and improve our understanding of land-atmosphere interaction.

Wed

09:45–10:00

Human interactions with the water cycle are increasingly recognised as important drivers of land–climate feedbacks, yet they remain under‑represented in climate models.
With ongoing climate change and global population growth, rising water demand leads is accelerating the overexploitation of already limited water resources. Consequently, water management strategies and irrigation practices are becoming more important across many parts of the world. Because these activities can significantly alter surface energy and water fluxes, and thus local and regional climate, it is important to study these processes in more detail.

Although some Earth system models and regional climate models have started to incorporate irrigation routines, most still lack a representation of water availability from different sources, such as groundwater pumping and reservoir operations, as well as competing demands from other sectors. To address this gap, we are developing the flexible water modelling tool C-CWatM that uses land-surface model outputs as inputs and can be easily coupled with existing (regional) climate models. Based on the socio-hydrological model CWatM, it simulates river discharge, groundwater, reservoirs and lakes, as well as water demand and consumption from industry, households and agriculture.

In this contribution, we evaluate the performance of C-CWatM and highlight the improved process representation through a fully coupled hydrology–climate modelling system. We present initial results from coupled simulations using C-CWatM and the regional climate model REMO to study the impact of large-scale irrigation on regional climate conditions in Europe. The coupling is implemented via the OASIS3-MCT coupler, which manages synchronised data exchange and regridding of coupling fields. REMO provides the forcing fields required by C-CWatM and receives irrigation water amounts from C-CWatM, which are then applied within REMO’s irrigation scheme.

The development and coupling of C-CWatM allows climate models to realistically represent
irrigation constraints, which is particularly important in water-scarce regions and under the
increasing risk of droughts driven by climate change. With its flexible, open-source, and accessible design, C-CWatM marks an important step towards fully coupled modelling of climate–water–human interactions. Incorporating C‑CWatM into modelling systems closes critical gaps between water management, hydrology, and atmospheric processes and thus helps to better understand the implications for land-atmosphere interactions and associated surface fluxes.

Wed & Thu

11:00–12:30

Recent improvements in the description of the surface-atmosphere interactions in the ICON atmospheric model of the German Meteorological Service are presented. The continental surfaces, including vegetation cover, represent an important component of the earth’s climate system. On the one hand, they are the habitat of humanity, which makes it important to try to understand the governing processes and living conditions at the land surface and how they may evolve in the future. On the other hand, from the point of view of atmospheric sciences, the land surface and biosphere interact with the lower atmosphere, and they have a significant impact on near-surface meteorological and climatological phenomena. During the recent years, the formulations of several processes in the ICON multi-layer land surface scheme TERRA were either deeply revised, or replaced by more realistic versions, for instance the formulation of the bare soil evaporation. A very persistent problem in atmospheric models is the computation of the surface temperature, in particular over vegetation. In TERRA, this was addressed by a simplified canopy approach, inspired by ECMWF’s IFS skin temperature formulation, embedded in the tile approach of ICON. Furthermore, with the trend of an increasing resolution of atmospheric models for numerical weather prediction or climate simulations, more fine-scale processes at the land surface can be resolved. Consequently, an urban canopy parameterization was developed and implemented in ICON, heading for km-scale or even hectometric scale applications. Beside further developing the physical parameterizations and using them with well-tuned, but static, parameter values, another possibility to improve the model skill is to apply an adaptive parameter tuning, based on the information from the data assimilation system. Meanwhile, this method is applied to several ICON land surface parameters. It is demonstrated that all these measures, and others, improve the interactions and feedbacks between land surface and atmosphere in ICON.

Tue

16:00–16:15

In the tropics, the land-ocean precipitation partitioning χ is skewed toward land. We analyze how CO2- and uniform sea surface temperature increase affect this partitioning of precipitation. To do so, we use 15 years of global atmosphere-land simulations conducted with the ICON model at 10 km horizontal grid spacing and explicitly resolved convection, unlike previous studies that parameterized convection. ICON produces a precipitation partitioning that is more consistent with observations compared to the AMIP6 ensemble. Under 4xCO2, precipitation partitioning toward land increases, whereas it decreases in +4K. We develop a framework based on energy and moisture budgets to decompose the response of the precipitation partitioning into contributions from the land column-integrated atmospheric heating, circulation efficiency, moisture cycling, and tropical radiative cooling of the atmosphere. In ICON and the AMIP6 ensemble, the land’s column-integrated atmospheric heating is identified as the primary driver of changes in precipitation partitioning. This is a result of the change in land moisture convergence and land precipitation in response to circulation adjustments driven by land-sea asymmetries in atmospheric heating. These circulation adjustments drive changes in land moisture convergence, that decide the sign of change in precipitation partitioning. The response of the controlling factors are similar in ICON and in the AMIP6 ensemble, apart from two qualitative differences. First, the land’s circulation efficiency is more stable in ICON than in AMIP6, which we attribute to a stronger coupling of precipitation to surface heat fluxes in AMIP6. Secondly, ICON shows a larger increase in precipitation in response to CO2 that the AMIP6 enxemble. This leads the opposing response in χ upon 4xCO2 and +4K to be virtually equal in magnitude in ICON, whereas in AMIP6 χ decreases more in +4K than it increases in 4xCO2.These findings suggest that coarse-resolution GCMs may overestimate the predicted decrease in land precipitation under global warming.

Mon & Tue

11:00–12:30

During Intensive Operations Periods (IOPs, three days monthly, May-August 2025) of the Land-Atmosphere Feedback Initiative (LAFI, see https://lafi-dfg.de) field campaign, we operated our high performance water-vapor differential absorption lidar (WVDIAL) at the Land-Atmosphere Feedback Observatory (LAFO) of University of Hohenheim, Stuttgart, Germany. Vertical profiles of absolute humidity were measured continuously from sunrise to sunset, whenever weather conditions permitted. With temporal and spatial averaging of 1 s and 70 m, respectively, the datasets reveal detailed insights into the diurnal evolution of water-vapor and turbulent transport processes from the lower planetary boundary layer (PBL) to the lower free troposphere with high precision and resolution. In particular, the transition from nocturnal residual layers to the convective boundary layer can be studied and quantified in detail, as well as small-scale processes such as updrafts and downdrafts and entrainment at the PBL top.

For LAFI, the WVDIAL was upgraded with a new diode-pumped Nd:YAG pump laser. It provides a stable nearly Gaussian profile at a repetition frequency of 200 Hz and up to 29 W at 532 nm, reliably pumping a Titanium:Sapphire crystal, which provides laser operation on suitable water-vapor absorption lines. Two diode lasers stabilize emission via injection seeding near 818 nm: one locked to a wing of the water-vapor absorption line (online), the other in the adjacent spectral valley where absorption due to water-vapor is negligible (offline). A laser power of 1.8 W is extracted from the resonator and transmitted through an optical fiber. The beam is directed into the atmosphere off-axially through a transmitter telescope mounted next to the 80-cm receiving telescope. The transmitter-receiver unit is motorized in azimuth and elevation, allowing for 3-dimensional atmospheric measurements. The water-vapor cross section is well known so that the profiling of absolute humidity is self-calibrating with an outstanding low error of < 2 %.

The WVDIAL measurements contribute directly to the objectives of the GEWEX Global Land–Atmosphere System Studies (GLASS) Panel by providing high-resolution profiles to better quantify land–atmosphere feedback processes, to evaluate model simulations of the planetary boundary layer, and to develop new parametrizations of atmospheric turbulence. Highlights of the LAFI measurements with corresponding derivations of profiles of turbulent variables will be presented. Typically, we found a water-vapor entrainment flux of 100  Wm⁻² and a peak of the water-vapor variance profile of 1.4 g²m⁻⁶ in the interfacial layer. Further results will be presented at the conference.

Mon & Tue

11:00–12:30

Land-atmosphere feedback is caused by mutual links between fluxes at the surface and atmospheric conditions. Part of this feedback concerns the water balance: evaporation at the surface causes addition of water vapour to the atmosphere, but is at the same time inhibited by high atmospheric humidity, constituting a negative feedback. However, the strength of this feedback could be diminished or even reversed into a positive feedback in the presence of plants, which are known to open their stomata when the air is humid, and close stomata when air is dry. This behaviour is commonly expressed as stomatal response to air vapour pressure deficit (VPD), all else being held constant. However, stomata also respond to light, air temperature, wind speed, atmospheric carbon dioxide (CO2) concentration, and soil water status. Considering that most of the vapour added to the atmosphere over land passes through plants and is controlled by stomata, a detailed understanding of stomatal behaviour appears crucial for our understanding of land-atmosphere feedback.
A myriad of models simulating stomatal response to environmental variables exist in the literature, ranging from entirely empirical models (deduced from multivariate regression of past observations) to mechanistic models considering detailed hydraulics and hormone signalling within a plant. An other approach, formulated by Cowan and Farquhar in 1977, assumes that plants optimise stomatal conductance dynamically in a way to maximise the time-integral of CO2 uptake given limited soil moisture availability. Mathematically, this translates into maintaining a constant slope (dE/dA) between transpiration (E) and CO2 assimilation (A) during a day and between leaves of the same plant (dE/dA=const.). In contrast to the empirical/mechanistic models, such optimality-based approaches do not rely on the assumption that stomatal behaviour and plant properties observed in the past will remain unaltered under new environmental conditions, e.g. under elevated atmospheric CO2 concentrations.
Here, we report observed leaf-scale stomatal responses to both natural environmental fluctuations and experimental manipulations of atmospheric conditions and soil moisture in the field and/or in the lab. Using a new method to determine dE/dA, based on experimental data alone (without the need for photosynthesis model parameterisation) we investigated if and under what conditions lambda stays indeed constant over the course of a day and between leaves of the same plant, and how it depends on soil moisture conditions. We found that on cloudy days, stomata of wheat plants adjusted stomatal conductance rapidly to fluctuations in light availability (by an order of magnitude within minutes!), as clouds passed in front of the sun. Hereby, stomata appeared to indeed converge towards a constant dE/dA value within the same plant. Keeping dE/dA constant is consistent with closing stomata at low atmospheric humidity and opening at high humidity, as observed empirically. However, the optimality-based approach allows predicting the magnitude of this response from fundamental principles, along with responses to light, air temperature, and wind speed.
Under very high air temperature and/or very limited soil water supply, dE/dA is not maintained constant, as evaporative cooling of the leaves and maintenance of plant hydraulic integrity becomes more important than maximising leaf carbon gain. The shift from carbon gain maximisation to “survival mode” was evident in our measurements as a tendency of stomata to maintain constant transpiration flux rather than dE/dA as atmospheric vapour pressure deficit was varied. This change in behaviour also became apparent in a change in vertical within-canopy profiles of CO2 and water vapour, which showed the imprint of stomatal behaviour observed at the leaf scale, indicating the relevance of leaf-scale stomatal control for larger scale atmospheric conditions and fluxes.
Our experimental results pave the way to a more nuanced understanding of stomatal behaviour under environmental fluctuations, including shifts between carbon gain maximisation and survival strategies, and the effect of stomatal behaviour on land-atmosphere feedback.

Wed & Thu

11:00–12:30

Uncrewed Aircraft Systems (UAS) are invaluable tools for atmospheric profiling, offering the mobility and capability to operate within the planetary boundary layer (PBL). Here, an observational gap exists between accurate but spatially restricted ground-based stations and remote sensing instruments or satellites. To fill this gap while prioritising usability and versatility, we have developed a custom meteorological sensor suite integrated into commercially available multicopter UAS.
The custom sensor package PARASITE (Portable Aircraft Rucksack for Atmospheric Sensing and In-situ Turbulence Estimation), integrates data from the aircraft’s positioning system and external meteorological sen- sors, including fast measurements of temperature, relative humidity and barometric pressure. We demonstrated the capabilities of this sensor package in flight on two commercially available platforms: a smaller multicopter (1 kg take-off weight) and a larger industrial multicopter (7 kg take-off weight).
The three-dimensional wind vector is estimated using only data from the aircraft’s internal avionics, elimi- nating the need for external flow sensors or expensive calibration infrastructure, by combining a physics-based model with a neural network residual correction. Validation against ultrasonic anemometers on a 99 m meteo- rological mast and 19 radiosonde ascents up to 2000 m a.s.l. yields a horizontal wind speed RMSE as low as 0.30 m s−1 for 1-minute averages. The method resolves atmospheric turbulence comparable to the reference, though performance varies by airframe: the industrial platform produces turbulence spectra closely following reference instrumentation, while the smaller UAS exhibits low-frequency oscillations in the variance spectra, likely introduced by its flight control algorithm.
During the VITAL campaign at Forschungszentrum Ju ̈lich in 2024, the system demonstrated compliance with WMO requirements by delivering processed data products in BUFR (Binary Universal Form for the Rep- resentation of Meteorological Data) format within minutes of each flight. During the TEAMx measurement campaign in the Inn valley (Austria) in summer 2025, the systems were used to measure wind shear and tur- bulent fluxes at four positions simultaneously. In the WINSENT-Valid project, the system resolved turbulent wakes of onshore wind turbines in complex terrain. Across these deployments, over 1000 measurement flights with five UAS demonstrate that commercially available multicopter UAS can serve as viable meteorological tools.

Mon & Tue

11:00–12:30

Understanding soil-vegetation-atmosphere interactions and feedback requires the characterization of the canopy air space as the interface between plant physiological processes and the atmosphere. Vegetation-atmosphere coupling at this interface is driven by source-sink dynamics. During daytime, photosynthesis and transpiration modify the canopy air space as they humidify the air, and deplete carbon dioxide (CO2) through stomatal pores. Stomata commonly open at high humidity and low CO2 concentration in the air. Under weak mixing of the canopy air space, this could lead to a positive feedback between stomatal control, humidification, and CO2 depletion. At the same time, weak vertical mixing allows the development of strong vertical gradients in air temperature, humidity, and CO2 concentrations, potentially decreasing photosynthetic efficiency due to depletion of within-canopy CO2.

Here, we investigate how stomata respond to canopy microclimate and how leaf-air feedback influences water vapor (H2O) and CO2 concentrations in the canopy air space. We measured within-canopy vertical profiles of H2O and CO2 concentrations at sub-hourly time scale in a wheat and a maize field between sunrise and sunset and put them into relation with canopy-scale and leaf-scale measurements of photosynthesis and transpiration to assess the stomata-atmosphere coupling strength. Additionally, we subjected individual leaves to varying air exchange rates while measuring their stomatal responses at different coupling strengths.

We found systematic humidification and simultaneous depletion of CO2 of the canopy air in the mornings and a reversal in the afternoons that correspond with stomatal conductance and light availability. These dynamics reveal the canopy air space buffer function that varies through the day and determines the vegetation-atmosphere coupling strength. Vertical profiles of H2O and CO2 concentrations illustrate the feedback strength and the vertical position of the dominating H2O sources and CO2 sinks in the canopy, which correspond to leaf area and light intensity distribution.

In leaf-scale experiments, we simulated feedback strength by controlling the exchange rate between leaf chambers and ambient air under changing light conditions. Under strong positive feedback conditions (low air exchange), the leaf gas exchange had a strong influence on the leaf chamber air and stomatal conductance continued to increase, releasing additional H2O while CO2 uptake became exceedingly limited by the declining CO2 concentration. Under low feedback strength (high air exchange), such positive feedback was not observed.

The canopy air space is an important component in soil-vegetation-atmosphere transfer (SVAT), as it determines the influence of vertical mixing on the feedback strength between stomatal control and atmospheric conditions. We characterize the interface between stomata and the atmosphere and thereby contribute to the mechanistic understanding of how vertical mixing and stomatal control regulate SVAT of H2O, CO2, and energy. Up-scaling from leaf to canopy through process-oriented measurements of the canopy air space could contribute to improved understanding of the vegetation-atmosphere coupling and how SVAT responds to environmental change.

Mon

15:00–15:15

Vegetation plays a crucial role during heatwaves by altering surface energy partitioning and influencing local to regional climate. In addition to the thermodynamic response of vegetation, the differential heating caused by sensible heat gradients across adjacent regions of vegetation and dry, bare soil can generate a mesoscale circulation akin to sea breeze-like circulation, known as a ‘vegetation breeze’ [1], which redistributes heat and moisture and affects downwind regions. While the impacts of large‑scale heterogeneities such as land-sea contrasts and topography are well established, the influence of finer‑scale vegetation heterogeneity remains uncertain. This gap is critical because semi‑arid forests, covering nearly 18% of Earth’s land surface, are highly sensitive to heat extremes. Differences in their Bowen ratios can substantially alter surface energy budgets, producing varying levels of hydroclimatic stress under similar atmospheric forcing. Yet, their potential to amplify or mitigate the impacts of extreme heat events is still poorly understood.
This study focuses on the semi-arid deciduous forests of the Eastern Ghats in Peninsular India, which are part of the Nagarjulam Srisilam Tiger Reserve and neighbouring protected areas located along the ecotone between the dry Deccan Plateau and the Eastern coast. It is spread over 5 districts in Andhra Pradesh and Telangana which are known to experience extreme heatwaves. Our previous observational analyses show that these transitional forests are highly sensitive to climatic stressors, particularly through their land surface temperature (LST) and evapotranspiration responses. During heatwave events, we observed pronounced LST gradients between forested and adjacent non-forested areas, indicating strong surface thermal contrasts arising from vegetation-atmosphere interactions. Given the heightened climate sensitivity of these transitional ecosystems, it is essential to understand not only how these ecosystems respond to extreme heat but also how they may influence local atmospheric dynamics.
To address this, we investigate how vegetation driven circulations such as the ‘vegetation breeze’ and the canopy convector effect [2] emerge from land surface heterogeneity, and how these processes affect boundary layer processes and downwind thermal anomalies during heatwaves. Our approach combines atmospheric reanalysis data for large‑scale boundary conditions, satellite observations to characterize land surface and vegetation, and high‑resolution WRF simulations to resolve fine‑scale forest-atmosphere feedbacks. Through a series of forest‑configuration experiments, we assess the capacity of semi‑arid forests to alter boundary layer processes and explore the implications for local and regional modification of extreme events as well as downwind impacts. By isolating the role of semi‑arid forests during heatwaves, these experiments contribute to the mechanistic understanding of semi-arid forest-atmosphere interactions and their role in shaping hydroclimatic extremes under a changing climate.

References
[1] McPherson, R. A. (2007). A review of vegetation—atmosphere interactions and their influences on mesoscale phenomena. Progress in Physical Geography, 31(3), 261-285.
[2] Banerjee, T., De Roo, F., and Mauder, M.: Explaining the convector effect in canopy turbulence by means of large-eddy simulation, Hydrol. Earth Syst. Sci., 21, 2987–3000, https://doi.org/10.5194/hess-21-2987-2017, 2017.

Mon & Tue

11:00–12:30

The evaluation of climate models is crucial for assessing their reliability and predictive capabilities. This study introduces an integrated model evaluation metric, which is a mathematically complete, three-dimensional representation of mean square error (MSE) that is decomposed into mean error (ME), variance error (VE), and correlation error (CE). This metric offers a more thorough evaluation of model performance by pinpointing sources of error. By normalizing each variable’s error using observed variability, our approach enables intercomparison across different climate variables. To demonstrate this framework, we evaluate surface air temperature and precipitation simulations from CMIP models. The demonstration highlights that significant improvements in CMIP6 temperature simulations over CMIP5 are primarily due to reductions in ME, while precipitation simulations exhibit only marginal improvements. A comparison of normalized errors between the two variables indicates superior performance in temperature, primarily driven by the significantly lower CE. Long-term error analysis reveals decreased actual errors in both the CMIP5 and CMIP6 models in recent periods, primarily due to the expansion and advancement of satellite-based observations, which inherently enhance the model performance in this period. However, the normalized errors increase due to the decrease in observational variability. These results indicate that while climate model performance has shown improvement recently, the gains in model fidelity are limited considering the reduced variance in the observation. Our proposed metric offers a flexible approach for model evaluation, allowing for the assessment of different temporal scales and application to forecast models as well.

Tue

09:00–09:15

Surface temperature over land is governed by the balance between radiative fluxes and non-radiative turbulent exchanges at the land-atmosphere interface. Rapid land-use and land-cover changes in India, particularly cropland expansion and afforestation, modify surface albedo, roughness, and evapotranspiration, thereby altering surface energy partitioning and near-surface climate. However, whether contrasting land-cover transitions produce symmetric or distinct temperature responses, and which physical processes dominate these responses, remains poorly understood in India.
This study examines surface and near-surface air temperature responses over India using long-term simulations from the ICON Earth system model spanning 1850-2014. A control simulation (CTRL) represents the historical evolution of land use and land
cover during this period. Two idealized experiments are analysed relative to CTRL: cropland-to-forest (C2F) and forest-to-cropland (F2C) (Figure 1). Surface temperature responses are interpreted through changes in shortwave radiation (SW), longwave
radiation (LW), latent heat flux (LHF), and sensible heat flux (SHF). In the C2F experiment, increased forest cover enhances turbulent exchanges, with higher LHF and SHF (Figure 3a and Figure 4a), promoting evaporative and turbulent cooling. However, forests exhibit lower albedo (Figure 5a), increasing SW absorption, and a warmer, more humid lower atmosphere that enhances downward LW radiation (Figure 6a). These radiative warming effects outweigh turbulent cooling, resulting in a weak surface warming of about 0.07 degrees Celsius relative to CTRL (Figure 2a). Increased surface roughness and displacement height strengthen surface-atmosphere coupling, leading to warmer near-surface air temperature despite enhanced evaporation (Figure 7a). In contrast, the F2C experiment shows a much stronger surface warming of about 0.26 degrees Celsius (Figure 2b). Replacing forests with cropland suppresses turbulent cooling, with reduced LHF and SHF (Figure 3b and Figure 4b). Although surface albedo increases slightly (Figure 5b), the associated SW cooling is weak and insufficient to compensate for the loss of turbulent fluxes. Reduced SHF also limits heat transfer from the surface to the atmosphere, resulting in cooler near-surface air temperature despite surface warming (Figure 7b).
Overall, the results demonstrate that surface temperature responses to land-cover change are nonlinear and asymmetric. Radiative processes primarily control surface warming, while turbulent fluxes regulate the efficiency of heat exchange with the atmosphere. Neither complete afforestation nor extensive cropification maximizes surface cooling; instead, mixed land-cover configurations balance competing biogeophysical processes more effectively.

Tue

09:15–09:30

Urban expansion modifies land surface properties and alters boundary-layer processes, with implications for heat and hydrometeorological extremes. We developed the first multi-decadal Local Climate Zone dataset for Greater Sydney (1990–2020 at five-year intervals), documenting substantial post-2005 densification marked by expansion of mid- and high-rise classes and reduction of open low-rise areas. These LCZ maps were implemented in the Weather Research and Forecasting model using the BEP/BEM urban canopy scheme, replacing the single bulk urban category with morphology-resolved classes. We performed a sequence of long-term simulations for 1990, 1995, 2000, 2005, 2010, 2015, and 2020 under two configurations: a control experiment applying fixed 1990 land cover across all years, and a dynamic experiment using the corresponding LCZ map for each simulation year. Both configurations reproduce the observed spatial distribution and totals of rainfall and near-surface temperature with comparable skill. The dynamic simulations exhibit a systematic reduction in surface wind speed relative to the control, particularly in locations that transitioned from open to compact urban forms, consistent with enhanced aerodynamic roughness under densification. The growth in urban structures produces stronger warming signals, with local increases exceeding 1 °C in minimum and mean temperature, potentially indicating amplified nocturnal heat retention. In contrast, rainfall differences between the two experiments do not show robust statistical significance; any differences are subtle and spatially heterogeneous, and far less pronounced than the wind and temperature responses. To assess whether urban growth influences storm characteristics beyond bulk rainfall totals, we are applying storm-object tracking to examine potential changes in convective intensity, morphology, and lifecycle behaviour, including merging and splitting. Collectively, these results provide process-based insights into how progressive urban densification reshapes Sydney’s local climate.

Mon

14:15–14:30

Accurate and spatially explicit soil information is a fundamental prerequisite for Earth system modelling, land surface simulations, and global environmental assessments. Although existing global soil datasets, such as GSDEv1 (The first version of this study, DOI:10.1002/2013ms000293), HWSD 2.0 and SoilGrids 2.0, have substantially advanced large-scale soil representation, they still exhibit limitations in spatial resolution, vertical consistency, and physical realism.
Here we present GSDEv2, a next-generation global soil physical and chemical property dataset developed to meet the increasing demand for high-resolution Earth system modelling. GSDEv2 provides seamless global predictions at 90 m spatial resolution for nearly 30 static soil properties, including soil organic carbon, texture fractions, bulk density, porosity, and related variables, across six standard depth intervals (0–200 cm).
The dataset is built upon an unprecedented compilation of approximately 23 million soil profiles, primarily sourced from the World Soil Information Service (WoSIS) and complemented by high-quality regional and national datasets. All profiles were subjected to rigorous, pedologically informed quality control procedures to remove implausible or inconsistent observations that can bias machine-learning predictions.
To better capture pedological heterogeneity, soil profiles and environmental covariates were stratified into desert, non-desert mineral, and organic soil domains. Separate Random Forest models were trained for each domain using a comprehensive set of covariates representing climate, topography, vegetation, and parent material, including AlphaEarth Foundations data.
Model predictions were validated using both internal cross-validation and independent reference datasets, demonstrating clear improvements in spatial detail and physical realism compared with GSDEv1, SoilGrids 2.0, and HWSD-based products. In addition, GSDEv2 adopts a data fusion framework, allowing high-quality regional soil maps to be integrated into the global predictions while preserving global consistency.
GSDEv2 represents a substantial step forward in global digital soil mapping, providing a physically consistent, high-resolution soil dataset that is better suited for hydrological, biogeochemical, and land–atmosphere modelling applications. The dataset is intended to support the GlobalSoilMap initiative and next-generation Earth system simulations.

Mon & Tue

11:00–12:30

Terrestrial water and carbon exchanges are intrinsically linked through plant physiological processes, yet this coupling is frequently obscured in satellite-based diagnostics where gross primary productivity (GPP) and evapotranspiration (ET) are estimated independently. While foundational theories establish the primary relationships governing photosynthesis, their realised expression is strongly conditioned by environmental heterogeneity, management, and scale. Consequently, it remains unclear whether increases in water use—arising from vegetation greening, irrigation, or enhanced atmospheric demand—translate into proportional gains in carbon uptake, limiting the interpretability of ecosystem water use efficiency (WUE).

We develop a hybrid diagnostic framework that preserves the core physiological structure of carbon–water exchange, while inferring its state-dependent realisation from data. The approach reconciles two complementary productivity perspectives: an energy-limited view in which GPP is constrained by absorbed radiation, and a water-limited view in which GPP is linked to ET through effective water–carbon trade-offs. Rather than prescribing this coupling through fixed closures, neural networks are used to infer the nonlinear mapping required for internal consistency between these perspectives. Under this formulation, WUE emerges as a field-scale diagnostic of carbon–water interaction rather than an imposed empirical parameter.

Weakly supervised learning addresses scale mismatches in satellite observations by using coarse-resolution light-use efficiency (LUE)-based GPP estimates as constraints to guide finer-resolution WUE-based representations. Transfer learning further improves model portability and mitigates data limitations. The framework is trained using the global FLUXNET 2015 dataset and fine-tuned with eddy covariance observations from India, a region characterised by strong anthropogenic regulation of ET through irrigation and rapid agricultural intensification.

Site-scale results show improved consistency between carbon and water fluxes relative to independent global products and reveal distinct temporal regimes in the co-evolution of dynamically inferred LUE scalars and WUE. The nonlinear behaviours captured by the neural networks provide insight into the dominant environmental forcing and regulatory states. By treating coupling as an inferable constraint rather than a prescribed mechanism, this work establishes a process-consistent foundation for field-scale diagnostics and future upscaling, advancing remote-sensing interpretations of land–atmosphere carbon–water dynamics.

Mon

14:00–14:15

With advancing climate change and mounting scarcity of hydrological resources, the interaction between land and atmosphere (L-A) is becoming increasingly important for weather and climate development. Since the resulting L-A feedback mechanisms depend on the water pathways between land and atmosphere, a sound understanding of the associated processes is necessary in order to represent them realistically in models. For this purpose stable water isotopes can be used as natural tracers, since their signature changes constantly due to various influences along the water pathways. By comparing observed isotope data with modeled data within the DFG-project “LAFI” (Land-Atmosphere-Feedback-Initiative) the origin of the modeled behavior can be traced back and the model can be further improved.

To this regard, we employ the appropriate fully coupled atmospheric-hydrological modeling system, WRF-Hydro-iso, which features the novel “-iso” implementation. We aim to improve our understanding of weather- and season-dependent interactions between groundwater, soil moisture, plant transpiration, soil evaporation, isotope signatures, and land-atmosphere (LA) feedback in agricultural areas at the convection-permitting scale. Therefore, a setup spanning multiple months with a resolution of 1.25 km was chosen based on ERA5 reanalysis forcing data and iCESM isotope climatology.

Benchmarking modeled outputs with all compartments of the water cycle requires an observatory with multiple instruments, such as the Land-Atmosphere Feedback Observatory (LAFO) near Hohenheim, Stuttgart, where various variables were measured within defined observation periods during the 2025 growing season.

In this session, we present modeled results that have been evaluated employing statistical metrics against lysimeter, sap flow, isotope chamber, CRNS (cosmic ray neutron sensing) soil moisture, and precipitation measurements. Due to the chosen resolution at convection permitting scale, the results provide detailed insights into the interactions between L and A, with a particular focus on the distribution of evapotranspiration (ET) compartments.

Mon & Tue

11:00–12:30

Shallow cumulus clouds are key contributors to vertical heat and moisture transport, yet their variability and sensitivity to surface conditions remain difficult to represent in atmospheric models. This study examines how soil moisture influences shallow cumulus mass flux and ensemble statistics using the Icosahedral Nonhydrostatic (ICON) model in large-eddy simulation (ICON-LES) mode under realistic large-scale forcing for three convective cases from the FESSTVaL campaign. Cloud objects are identified and tracked using a feature-based framework, allowing consistent extraction of cloud-base properties and associated thermodynamic and dynamical fields. Clouds are classified as active or passive based on excess in-cloud virtual potential temperature relative to the environment. The resulting cloud-base mass flux distributions are well described by Weibull functions. Active clouds exhibit a heavy-tailed distribution with a shape parameter (k ~ 0.8), indicating lifecycle-related variability, whereas passive clouds approach (k ~ 1), consistent with short-lived, stochastic behavior. The ensemble-mean cloud-base mass flux scales robustly with the thermodynamic efficiency of a moist heat cycle, linking surface flux partitioning to convective intensity. Sensitivity experiments demonstrate that soil moisture strongly modulates cloud number, updraft strength, cloud size, and the mass flux distribution. Drier soils promote high mass flux, and larger number of the total clouds, while wetter soils suppress convective intensity. These results provide physically grounded constraints for improving stochastic shallow cumulus parameterizations.

Mon

16:00–16:15

The aim of this presentation is to stimulate the GEWEX GLASS community to support and contribute to the development of a Synthetic Aperture Radar (SAR) mission concept to provide sub-daily global, fine resolution observations of soil moisture and vegetation water storage
The SLAINTE mission concept was initially developed in response to ESA’s 12th call for Earth Explorers. Though not selected, this first attempt highlighted the significant potential for sub-daily observations[1,2].
Though current satellite missions meet the requirements of many meteorological applications of surface soil moisture, some significant and poorly-quantified land-atmosphere exchanges are associated with rapid changes in surface soil moisture at fine spatial scales. For example, rapid changes in soil moisture and its spatial distribution influence the development and propagation of mesoscale convective storms. Estimates of irrigation amounts are essential to quantify the impact of agricultural water use on local, regional and large-scale atmospheric processes, but are characterized by short-lived wetting/drydown events.
The primary motivation for SLAINTE was to provide sub-daily observations of vegetation water storage. These are needed to study the vegetation response to the daily cycle in vapour pressure deficit (VPD), the impact of stomatal regulation, and the rate at which vegetation is able to recharge water lost during the day. Sub-daily variations in vegetation water storage reveal how ecosystems respond to biotic and abiotic stress (e.g. changing temperature and vapour pressure deficit, soil moisture, insects, disease) and disturbances (e.g. drought, fire). Observing these processes is critical to understand the resilience of terrestrial ecosystems, to parameterize their response to stress to predict how they will respond to increasing climate variability and extremes, as well as pressures from human land and water use.
In this presentation, we will briefly review key aspects of the original mission concept, and use results from recent studies to illustrate the unique perspective offered by sub-daily observations in applications relevant to the GEWEX community. We will highlight current research activities exploiting ground-based observations, radiative transfer modeling and a combination of satellite and aircraft data to consolidate requirements and strengthen the science case in preparation for the next opportunity. Finally, the audience will be invited to contribute their perspective on the potential value of sub-daily observations to address crtitical gaps in their capacity to observe, understand and parameterize turbulence and surface fluxes at the land-atmosphere interface.

[1] Steele-Dunne, Susan, et al. “SLAINTE: A SAR mission concept for sub-daily microwave remote sensing of vegetation.” EUSAR 2024; 15th European Conference on Synthetic Aperture Radar. VDE, 2024.
[2] Matar, Jalal, et al. “A Concept for an Interferometric SAR Mission with Sub-daily Revisit.” EUSAR 2024; 15th European Conference on Synthetic Aperture Radar. VDE, 2024.

Wed

09:45–10:00

Wheat production in dryland regions, such as the Eastern Mediterranean, is highly vulnerable to climatic variability. This study utilizes a high-resolution (3 km²) coupled modeling framework to investigate the bidirectional interactions between the land surface and the atmosphere. Using the Weather Research and Forecasting (WRF) model coupled with the Noah-MP-Crop model, we simulate the complex feedbacks between wheat crop, soil moisture dynamics, and atmospheric conditions.

Following calibration against commercial spring wheat data in Israel, the coupled model was executed over a 30-year period to evaluate crop development under a fully interactive system. Our results reveal a critical non-linear relationship between soil moisture, surface fluxes, and water use efficiency (WUE). Specifically, we identified a process-based threshold where water stress exceeding 30% triggers a rapid decline in potential yield and shifts in evapotranspiration rates. Soil texture analysis further indicates that while clayey soils offer greater resilience to moisture variability, sandy soils exhibit distinct feedback behaviors under severe water stress.

By emphasizing the importance of representing crop-specific phenology and soil heterogeneity, our findings highlight how high-resolution coupled crop-climate modeling provides a pathway for improving land-surface representations. This approach can enhance our ability to predict the impacts of climate change on regional energy and water cycles.

Wed & Thu

11:00–12:30

Soil moisture serves as a fundamental regulator of the land-atmosphere interface, controlling the surface water and energy balance through evapotranspiration. This interaction plays a crucial of initiation, intensification, and persistence of extreme heat events, such as heatwaves and droughts. Variations in soil moisture not only influence local thermodynamic processes but also trigger hemispheric changes in upper-level atmospheric circulation, which subsequently modulate surface temperature anomalies.
From a local process perspective, the sensitivity of maximum temperatures to soil moisture exhibits a non-linear increase, particularly under moisture-limited conditions. This relationship is characterized by a conceptual soil moisture threshold, known as the breakpoint, at which temperature sensitivity increases significantly. Within the dry regime below this threshold, reduced soil moisture leads to a substantial decline in latent heat flux and suppressed cloud formation. Consequently, sensible heat flux becomes the dominant driver of surface heating, amplifying land-atmosphere feedback. This mechanism serves as a critical factor in sustaining the duration and increasing the intensity of extreme heat events across the Northern Hemisphere.
Despite its importance, quantitative assessments of how these interactions dictate the frequency and duration of heatwaves remain insufficient. To address this, we utilized large-ensemble simulations from the EC-Earth3 climate model, comparing an interactive soil moisture experiment with a prescribed (non-interactive) soil moisture experiment.
Our findings demonstrate that interactive land-atmosphere coupling significantly amplifies both mean summer temperatures and the magnitude of extreme heat events compared to non-interactive runs. Specifically, the progressive desiccation of soil in interactive simulations triggers a shift in the surface energy budget, where the marked reduction in latent heat flux is compensated by intensified sensible heat flux, directly heating the lower troposphere. Furthermore, this soil moisture variability induces significant teleconnections in the Northern Hemisphere’s upper-level atmospheric circulation, specifically modulating jet stream. These shifts promote the formation of stagnant high-pressure systems, creating a self-reinforcing cycle that exacerbates the intensity and duration of heatwave conditions.
Moreover, the inclusion of land-atmosphere interactions substantially modulated the soil moisture-temperature sensitivity and its critical breakpoint. The interactive soil moisture experiments demonstrated prevalent soil drying and amplified warming, most notably over the Northern Hemisphere mid-latitudes, which in turn strengthened the land-atmosphere coupling and temperature sensitivity. Consistent with this overall drying, the critical breakpoint also exhibited a systematic shift toward drier soil moisture conditions. By integrating these localized surface flux anomalies with large-scale dynamical changes, this study provides a comprehensive quantitative evaluation of land-atmosphere coupling. Our results emphasize that accurately representing these interactive feedbacks is essential for improving the fidelity of future climate projections and the skill of seasonal heatwave forecasting.

Mon & Tue

11:00–12:30

Monsoons are characterized by distinct seasonal wind shifts accompanied by alternating wet and dry phases. While land surfaces play a critical role in shaping weather and climate during monsoons, our understanding of these interactions remains limited. This study investigates the influence of soil moisture on surface temperature at sub-monthly timescales, with a particular focus on monsoon regions. Using the novel causal analysis technique, Liang–Kleeman information flow, together with ERA5 reanalysis data, we identify a significant influence of soil moisture on surface temperature that is closely linked to the seasonal monsoon cycle. In monsoon regions, stronger effects generally occur during the onset and withdrawal phases of the monsoon. The seasonal progression of regions sensitive to soil moisture closely follows the evolution of monsoon onset. In contrast, during the peak monsoon season, the influence of soil moisture on surface temperature weakens due to an energy-limited regime associated with reduced insolation. During the withdrawal phase, favorable radiative and soil moisture conditions again enhance the influence of soil moisture on surface temperature. Notably, our results show that soil moisture can significantly influence surface temperature even under wet land conditions, challenging the conventional understanding of soil moisture–temperature coupling. Additionally, we find a clear relationship between soil moisture and the occurrence of hot days (surface temperatures exceeding the 95th percentile of daily temperature) during the pre-monsoon period in South Asia, highlighting the important role of soil moisture in modulating heatwaves in monsoon regions. These findings advance our understanding of land–atmosphere interactions in monsoon systems and provide new insights into the subseasonal predictability of surface temperatures driven by soil moisture variability and Earth system modeling.

Thu

09:15–09:30

How soil moisture (SM) influences convective precipitation remains somewhat controversial. Many studies consider the problem from the perspective of a horizontally-uniform vertical profile. On the other hand, landscapes are heterogeneous, and satellite observations of convective storms tend to consistently show an important role for SM spatial variability and its impact on convection via mesoscale circulations. Here we shed new light on this debate using a high resolution dataset of over 2 million convective initiations across Sub-Saharan Africa based on 21 years of Meteosat Second Generation (MSG) imagery.
Defining convective initiation (CI) as the first afternoon appearance of cloud-tops of -40 degrees or colder, we characterise the pre-initiation surface using soil moisture from ASCAT and MSG land surface temperature. In both datasets we find CI is consistently favoured over locally drier/warmer surfaces close to cooler and wetter conditions. The signal strengthens in climatologically drier regions, and only in the most complex terrain do we not see this signal of CI organised by heterogeneity. The most striking new findings however are when considering the influence of vertical wind shear on the interaction between CI and SM. Alongside Convective Available Potential Energy (CAPE), wind shear between low and mid-levels is well-known to impact storm growth, and indeed we find the most rapidly-cooling cloud-tops are affected by these factors. Critically, we also find that local SM gradients influence cloud growth rates, with the greatest vertical storm growth occurring where soil moisture-driven circulations oppose the direction of shear-induced cloud displacement. This means that within a heterogeneous SM field, CI is strongly-favoured on gradients with specific orientations relative to the direction of low and mid-level winds. Once initiated, the cloud moves with the mid-level flow. When low and mid-level winds are aligned, this favours storm rainfall over wetter soils (positive feedback). On the other hand, if the winds are in broadly opposing directions, as is typically the case over much of Tropical North Africa, a strong negative feedback emerges. In fact we can detect an impact of this interaction between directional shear and SM heterogeneity on spatial SM-precipitation feedbacks globally. Moreover, the combination of soil moisture heterogeneity and wind shear provides a potentially important source of predictability for where deep convection develops, particularly for the most rapidly-developing thunderstorms.

Mon

17:15–17:30

Accurate representation of vegetation type and spatial distribution is fundamental for accurate land surface modelling, as vegetation strongly regulates momentum transfer, surface albedo, evapotranspiration, and soil moisture dynamics. Mischaracterisation of land cover or vegetation properties propagates through surface energy and water flux calculations, hydrological responses, and land–atmosphere feedbacks. This study investigates how land cover and dynamic vegetation properties influence surface energy and water fluxes across Australia using the Joint UK Land Environment Simulator (JULES).

We compare two widely used global land cover (LC) datasets, IGBP and ESA-CCI, against a high-resolution national LC dataset developed by the Australian Bureau of Meteorology (BoM) to assess JULES sensitivity to differing vegetation classifications and spatial patterns. Australia’s uniquely sparse, heterogeneous, and rainfall-driven ecosystems, present a rigorous testbed for assessing the impacts of LC-driven uncertainties. In addition, we examine the impacts of introducing annually and monthly varying vegetation fractions relative to static climatological maps. We also evaluate the effects of replacing climatological leaf area index (LAI) with satellite-derived monthly LAI.

Results show that improved characterisation of vegetation types and their distribution produces more realistic simulations of latent and sensible heat fluxes, validated against flux tower measurements. Both IGBP and CCI configurations exhibit a cool LST bias exceeding 2 K over tropical and subtropical regions, whereas simulations using BoM land cover fractions substantially reduce this bias to approximately 1 K and 0.5 K respectively for the same regions. Incorporating monthly varying vegetation fraction and observed LAI enhances the model’s capacity to capture rapid ecosystem changes. This is illustrated using a major bushfire event, where static climatological fractions fail to represent abrupt reductions in canopy cover and associated shifts in surface flux partitioning.

Overall, this study demonstrates that both accurate land cover mapping and temporally varying vegetation information are essential for reducing uncertainty in land surface modelling. Incorporating dynamic vegetation properties substantially enhances model performance for Australia’s highly variable and disturbance‑prone landscapes. These findings underscore the importance of accurately representing vegetation spatial distribution and temporal dynamics to improve land surface modelling under current and future climate conditions.

Wed & Thu

11:00–12:30

The recent years have shown increasing interest and effort to include representations of irrigation in climate models to better account for the effects of this anthropogenic process on climate. In semi-arid climates, irrigation can create strong heterogeneities of soil moisture and turbulent fluxes, affecting the structure of the atmospheric boundary layer over irrigated areas, and over neighbouring zones depending on the context. Such heterogeneities can be explicitly represented in high resolution simulations but must be parameterized within ESMs.
We present here results of simulations with the atmosphere and land surface components of the IPSL Climate Model. ICOLMDZ, coupling the physics of LMDZ to the recent icosahedral dynamical core DYNAMICO, is run in a Limited Area Model (LAM) configuration to conduct a regional study over North-Eastern Spain, at 25-km resolution. A new representation of irrigation, based on a soil moisture deficit approach, has recently been developed in the ORCHIDEE land surface model and simulations are run with and without it to assess the impacts of simulated irrigation in the model.
The simulations are compared to observations from the Land-surface Interactions with the Atmosphere In Semi-Arid Environment (LIAISE) field campaign, conducted in 2021 in the Ebro Valley. This campaign was specifically designed to provide a better understanding of the local and regional impacts of irrigation and the surface heterogeneities it creates. Surface measurements over 2 weeks and radiosoundings on two IOP days, over an irrigated site and a rainfed one, are confronted to simulated variables on corresponding grid cells.
LAM simulation outputs are also compared to higher-resolution simulations (2km) that have been conducted using the Meso-NH model in the context of the LIAISE project. These simulations serve as a bridge between point-based measurements and 25-km grid cells of the LAM to assess the importance of subgrid variability in land-atmosphere coupling variables and ABL structure. Using this novel approach for model evaluation, the LAM is found to simulate correct grid-cell mean surface fluxes, but shows limitations in the representation of the ABL structure depending on low-level wind speed and direction. Based on this analysis, possible improvements for surface and ABL parameterizations in the IPSL Climate Model are identified.

Mon & Tue

11:00–12:30

Wetlands regulate land–atmosphere exchanges by sustaining high evaporative fractions and enabling strong carbon–climate feedbacks. Yet most Earth System Models (ESMs) represent wetness primarily through soil moisture or water-table depth, omitting a defining wetland behavior: shallow surface-water ponding that seasonally expands, stores water above ground, and laterally redistributes it across microtopographic structures. Because ponding alters albedo, heat storage, surface roughness, and aerodynamic and surface resistances, it can shift surface energy partitioning and boundary-layer development while reorganizing the hydrologic connectivity of wet-landscapes.
We present a new Surface–Soil Exchange and Emergent Ponding (SEEP) framework implemented within the NOAA GFDL land model (LM4.1) to represent dynamic surface ponding inside an ESM. SEEP introduces (i) an explicit surface-water column bidirectionally coupled to the soil (infiltration/exfiltration), (ii) revised surface energy-flux calculations over ponded tiles, (iii) freeze–thaw-aware vertical re-layering of surface water, and (iv) a topography-based lateral overflow parameterization. To resolve the heterogeneity that governs where ponding emerges within coarse ESM grid cells, we pair SEEP with a machine-learning-informed subgrid tiling strategy derived from high-resolution environmental datasets.
We benchmark simulated ponding states and dynamics using the Everglades Depth Estimation Network (EDEN) daily water-level products. Initial applications reproduce realistic seasonal inundation patterns and associated increases in evapotranspiration and reductions in skin temperature over ponded land units, demonstrating how emergent surface water modulates coupled water–energy exchange. Ongoing work is generalizing SEEP across multiple wet-landscapes by constraining key parameters controlling infiltration resistance, overflow thresholds, and clogging-layer effects. In parallel, we are incorporating a sub-grid-scale river-network representation to enhance hydrologic connectivity among ponded tiles, enabling two-way exchanges among surface ponding, lateral runoff, and channel flow within ESM grid cells. This integrated tiling–river framework allows us to investigate how fine-scale surface-water redistribution influences land–atmosphere coupling and surface flux heterogeneity from sub-seasonal to decadal timescales.
By advancing the physical representation of surface water and hydrologic connectivity in ESMs, grounded in field observations, this work provides a pathway to represent better surface heterogeneity and its impacts on boundary-layer processes and climate feedbacks across wetland-dominated landscapes.

Mon & Tue

11:00–12:30

Extreme precipitation is among the most societally relevant manifestations of land-atmosphere variability and climate change. Yet, the short-term evolution of convective storms remains difficult to model temporally consistently and computationally efficiently. We are developing a machine learning framework for object-centric tracking of intense precipitation systems in high-resolution radar sequences. Building on recent advances in object-centric foundation models and compact weather-data representations, the approach encodes local radar observations into compressed latent codes. Focusing on storm cells, the approach learns storm-predictive latent codes as well as 2D Gaussian parameters that specify its position and spatial extent. Given a history of these latent encodings, a temporal predictive model is trained to predict the short-term evolution of the storm cells.

The study is planned for high-resolution radar composites of intense precipitation events over Germany, with particular emphasis on coherent convective cells and rapidly evolving storm structures. The workflow first compresses the radar field locally into compact latent representations that preserve storm-scale information relevant for forecasting. An object-centric encoder then maintains one latent state per identified storm cell and disentangles compact cell morphology and intensity information from spatial location. Based on the sequence of past latent codes and Gaussian parameters, an attention-based dynamics model learns temporal propagation patterns and predicts the future trajectories of the tracked cells for lead times of a few tens of minutes. Model evaluation will focus on track continuity, localization accuracy, lead-time-dependent displacement error, and the robustness of the learned trajectories during severe events, including comparisons with persistence and optical-flow-based baselines. In this way, we aim to obtain an interpretable and computationally efficient framework for following extreme precipitation systems in time. More broadly, the approach investigates how compact latent representations and object-centric learning can support storm-focused diagnostics and provide a useful foundation for future data-driven analyses and forecasts of precipitation extremes and their dynamics.

Mon & Tue

11:00–12:30

How to properly leverage data sparsity in space and time remains the key challenge of data assimilation. In practice, this data sparsity realizes itself as spatially inconsistent in-situ observations alongside remote sensing approaches that as a whole are not able to provide continuous coverage at a consistent fine spatial and temporal resolution. Unsurprisingly, the gap in understanding this creates for multiscale land-atmosphere processes hampers the accuracy and generalizability of our models to better inform policy-making and prepare vulnerable communities for incoming risks.

Traditionally, recovering missing state variables from limited data has relied on standard numerical data assimilation (Carrassi et al., 2018) or, more recently, deep learning approaches (Reichstein et al., 2019). Here we present a framework built on a physics-informed latent diffusion model to fundamentally investigate whether underlying physical dynamics can be learned from sparse observations. In doing so, we can not only achieve high-fidelity reconstructions but also lay the groundwork for probabilistic downscaling of boundary layer and land surface fields in data-sparse regimes. Our approach involves an inverse problem motivated by PDE-constrained reconstruction. We compare four inference strategies: a physics Maximum a Posteriori (MAP) estimator that optimizes directly over PDE latents, a Multilayer Perceptron (MLP) inference encoder, and two posterior samplers built on a latent diffusion model initialized from either MAP or the encoder. We demonstrate that this generative framework successfully generalizes to structured, out-of-distribution spatial observation patterns and report reconstruction quality via metrics like RMSE, SSIM, and spectral fidelity. Probabilistic calibration scores like CRPS, energy score, and coverage are also computed to quantify both the accuracy and uncertainty under these conditions.

Wed

14:45–15:00

Accurate field-scale estimation of surface and root-zone soil moisture (SM) is critical for irrigation management and agricultural decision-making, particularly in regions dominated by smallholder farms. Conventional Land Surface Models (LSMs), which typically operate at coarse spatial resolutions (>10 km), often fail to represent sub-grid heterogeneity in soil, topography, vegetation, and precipitation, limiting their usefulness for farm-scale applications. Hyper-resolution modeling frameworks provide an opportunity to overcome these limitations, but their performance depends on how well these controlling factors are represented. In this study, we evaluated how the representation of these controlling factors influences high-resolution soil moisture estimates. HydroBlocks is a hyper-resolution LSM that captures fine-scale topographic heterogeneity through a Hydrologic Response Unit (HRU) framework coupled with subsurface lateral flow connectivity, thereby improving the representation of sub-grid hydrologic dynamics. In addition to topography, we further assessed how the remaining three controlling factors (soil, precipitation, and vegetation) contribute to improving hyper-resolution soil moisture estimates.

In this study, we employ the HydroBlocks framework with Noah-MP at its core to simulate 3-hourly soil moisture at ~30 m resolution over the Upper Bhima Basin, India, and systematically evaluate the influence of soil, precipitation, and vegetation representation on SM simulations. To assess the role of soil structure, simulations using vertically homogeneous (VHom) and vertically heterogeneous (VHet) soil properties derived from the SoilGrids database were compared. Both configurations show strong agreement with satellite, reanalysis, and in situ observations, while VHet reduces subsurface bias and enhances spatial variability. A season-wise global sensitivity analysis across soil layers indicates that soil porosity, the Brooks–Corey parameter, and wilting-point soil moisture are dominant controls, with stronger influence near the surface and during the monsoon season.

Mon & Tue

11:00–12:30

Land-atmosphere interactions play a key role in the Earth’s climate. The surface temperature is a key parameter in calculating the latent and sensible heat flux and thus important for the closure of the surface energy balance. Yet vegetated surfaces have different properties compared to bare soil and thus behave differently. Grass-vegetated surfaces are by far the most common type of land cover, covering over 40 % of all land area. Therefore, accurate modelling of soil and grass temperatures is essential for improving numerical weather prediction models.

Surface temperatures over grass are often modeled by using the grass -‘thermal roughness length’ and the grass ‘skin-layer conductance’. However, both concepts are highly empirical and poorly rooted in physics. Parameters are uncertain which leads to large errors in the prediction of surface temperature. We therefore need more physics-based approaches with better constrained parameters. For tall vegetation such approaches are available, but for grass progress was hampered by our inability to probe the microclimate of it. Here we show novel data obtained with Distributed Temperature Sensing. By coiling up optical fibers, temperature-profiles can be probed with mm resolution. This reveals that temperatures in grass obey nontrivial dynamics that cannot be captured accurately by the aforementioned concepts.

Here, we introduce alternative diffusion-type of concepts as to describe in-and above-grass temperature dynamics in agreement with the DTS observations. In the future, model parameterizations based on those concepts will be compared with performance of traditional parameterizations that use skin-layer conductance and surface roughness lengths, as to benchmark the results in a climatological sense using larger datasets.

Wed & Thu

11:00–12:30

Soil water retention curves (SWRCs) and their associated hydraulic conductivity curves (HCCs) form the fundamental constitutive relationships governing unsaturated flow in soils. Together, these soil hydraulic properties determine moisture availability, hydraulic connectivity, and the redistribution of water under drying and wetting cycles. Through their role in Richards’ equation and conductivity closures, these relationships directly influence soil moisture persistence, evaporative resistance, and surface energy partitioning. Despite their central importance, most operational land surface models rely on classical capillarity-based formulations such as van Genuchten, Brooks & Corey, and Kosugi. While these models are effective across moderate-to-full relative saturation values, they primarily represent capillary-controlled processes and do not explicitly account for adsorption-dominated behaviour in the dry regime.
Under dry conditions, water is retained primarily by adsorption-induced thin films coating particle surfaces rather than by capillary mechanisms operating within connected pores. Neglecting these two distinct forms of soil water alters the shape of both the retention and conductivity functions in the dry range, affecting hydraulic connectivity and moisture availability. Such model structural differences can propagate into biases in simulated soil moisture, soil temperature, and surface heat fluxes (latent, sensible, and ground heat flux). Addressing this limitation requires formulations that explicitly incorporate adsorption processes and provide a continuous representation across capillary and adsorption regimes.
In this work, we present advances toward adsorption-explicit SWRC formulations embedded within a unified hydro-thermal framework. We examine and compare several physically based adsorption-capillary formulations for soil hydraulic properties, including those proposed by Weber et al. (2019), Lebeau & Konrad (2010), Lu et al. (2016)/Gou et al. (2023), among others. These formulations explicitly incorporate thin-film and adsorption contributions, enabling a continuous description of soil water retention and hydraulic conductivity across capillary and adsorption regimes. We analyse the mathematical structure and assumptions of these models, with particular focus on their behaviour in the dry range and their implications for hydraulic connectivity.
A key component of this effort is ensuring thermodynamic and structural consistency between hydraulic and thermal processes. Soil heat transport is strongly influenced by moisture (re)distribution, phase connectivity, and soil structure, yet in the vast majority of modelling frameworks thermal conductivity parameterizations are treated independently of hydraulic descriptions. To address this inconsistency, we evaluate formulations that link soil hydraulic parameters (SHPs) directly to thermal properties. Specifically, we test the relationships proposed by Lu & McCartney (2024) and Fu et al. (2026), which connect soil hydraulic property equations with thermal conductivity curves through shared structural descriptors. In addition, we develop and explore alternative relationships between SHPs and soil thermal parameters designed to preserve physical coherence across hydro-thermal processes.
By coupling retention, hydraulic conductivity, and thermal conductivity within a consistent structural framework, we aim to reduce fragmentation between soil water and heat representations. The resulting unified hydro-thermal approach allows moisture-dependent thermal behaviour to emerge naturally from the same structural parameters that govern retention and conductivity. Particular attention is given to the behaviour in dry soils, where adsorption processes influence not only hydraulic transport but also thermal connectivity through changes in phase distribution and film continuity.
Collectively, this work provides a systematic examination of adsorption-explicit retention formulations and their implications for hydro-thermal consistency. By comparing multiple adsorption-based models and evaluating their coupling to thermal parameterizations, including via parameter parsimony and parameter sensitivity analyses, we identify pathways toward reducing structural uncertainty in medium-to-dry-soil physics. These developments lay the foundation for more physically grounded representations of soil water-heat interactions and provide a framework that can be integrated into next-generation land surface modelling systems.

Wed & Thu

11:00–12:30

Forests and clouds are central to Earth’s carbon and water cycles, yet they are rarely studied as a coupled system. Recent observations reveal concurrent shifts in forest CO₂ uptake and cloud regimes across tropical, temperate, and boreal biomes, signaling changes in forest–atmosphere coupling with profound implications for cloud cycling and climate feedbacks. While rising CO₂ may enhance forest assimilation, declining trends in low cloud cover alters radiative fluxes and amplifies warming, potentially modifying forest photosynthesis, turbulence, and biogenic volatile organic compound emissions. In turn, these processes influence clear/cloud boundary layer dynamics by controlling the partitioning of canopy turbulent fluxes, influence boundary-layer dynamics and cloud formation. Yet current Earth system models largely overlook these cross-scale interactions.

To advance our understanding on the forest-cloud coupling, we focus on the Amazon basin as a proof-of-concept where we integrate field observations from the CloudRoots-Amazon22 campaign with new multi-layer canopy large-eddy simulations that explicitly resolve interactions between the forest canopy and the clear/cloudy boundary layer. The CloudRoots-Amazon22 experiment, conducted at the ATTO and Campina supersites during the August 2022 dry season, investigated the sub-diurnal evolution of the common clear-to-cloudy transition in the Amazon.

High-frequency observations reveal that stomatal conductance responds to variations in cloud optical thickness, demonstrating that canopy–cloud radiative perturbations regulate sub-diurnal canopy carbon and water exchange. Turbulent fluxes and vertical transport adjust within minutes to cloud passages, highlighting rapid land–atmosphere coupling. Collocated surface fluxes, profiles of thermodynamic variables, and CO₂ concentrations, further establish causal links between biophysical canopy processes and cloud dynamical development.

Building on these insights, we present an integrated framework that combines high-frequency observations with turbulence-resolving simulations embedded in global storm-resolving models to quantify shifts in cloud–forest coupling under climate change. This coupled approach advances our understanding of how cloud-radiative perturbations, turbulent transport, and photosynthesis co-evolve, bridging leaf-level processes and cloud-scale dynamics, and provides a pathway to constrain key uncertainties in Earth system models.

Mon

17:00–17:15

Accurate partitioning of evapotranspiration (ET) into crop transpiration (T) and soil evaporation (E) is essential for understanding water fluxes in agricultural systems and improving the representation of land-atmosphere interactions in land surface models. Despite its importance, methodological challenges limit the accurate measurement of T/ET ratios on field scale. Here, we quantify the diurnal and seasonal dynamics of ET, T and E in winter wheat and maize using high-resolution in-situ water isotope flux chamber measurements, complemented by micro-lysimeters, sap flow sensors and eddy-covariance (EC) measurements. Our study aims to (i) evaluate the performance of water isotope flux chambers relative to complementary methods, (ii) assess crop stand-specific differences in the fluxes ET, E, and T and their temporal variations, and (iii) provide robust ET-flux data for land surface model calibration.
Our method comparison was conducted based on data collected during 2-5 consecutive days per month during the 2025 growing season at the agricultural experimental site “Land-Atmosphere-Feedback Observatory” (University of Hohenheim, Germany), covering key crop phenological stages of winter wheat (April-June) and maize (June-August). Actual ET was measured using an on-site EC system installed on a grass strip between two fields with identical crops. Isotope-based chamber measurements were performed in the vicinity of the EC tower from sunrise, after morning dew evaporation, until sunset. E, T and ET chambers were measured consecutively, resulting in 8-12 measurements per flux type per day. The T/ET ratio was estimated from water isotope compositions of E, T and ET using an isotope mass balance approach (Keeling method). Additionally, E, T and ET were calculated from temporal changes in the chamber water vapor concentration during measurements. Complementary estimates of E and T were derived from five micro-lysimeters installed in a star-like pattern around the EC tower, and five sap flow micro sensors installed on five individual plants.
In 2025, the climate was generally warmer and drier than long-term average, with extreme drought from February to mid-April. Rainfall was mostly limited to late April, late May/early June, and July. During the study period, temperatures increased from about 10ºC in April to 20ºC in June and remained stable through August, while relative humidity ranged from 50–60% but increased to about 70% in July. EC-derived ET ranged from 0.5 to 5 mm d-1 and generally increased over the growing season for both crop species. Sap flow-derived T varied between 0.5 and 4 mm d-1, whereas lysimeter-derived ranged from 0.3 to 1.2 mm d-1 but increased up to 5.5 mm d-1 after rain events. Day-to-day flux variability follows patterns of solar radiation and vapor pressure deficit. Diurnal cycles showed typical bell-shaped patterns, modified by meteorological conditions. Chamber-based estimate of E, T and ET reproduced the temporal dynamics observed with EC, lysimeter, and sap-flow measurements, although scattering was larger, likely due to spatial heterogeneity and differences in observational scales from single plants to the EC footprint. Across measurement approaches, T/ET ratios mostly ranged between 0.6 and 1.0, indicating that T dominated ET. Lower T/ET ratios observed after precipitation events are attributed to enhanced soil evaporation.
We will identify the environmental and plant physiological drivers of crop-specific E, T and ET dynamics and quantify uncertainties across measurement approaches to derive best-estimate fluxes and partitioning. Our findings advance the mechanistic understanding of agricultural water fluxes and provide benchmark observations for improving land-surface model performance.

Mon & Tue

11:00–12:30

Surface solar irradiance (SSI) drives the surface energy balance. Its partitioning into sensible and latent heat fluxes fuels boundary layer turbulence, which in turn shapes the clouds that modulate the radiation in the first place. Under broken cloud conditions, SSI is far from uniform: cloud shadows can reduce irradiance by several hundred watts per square meter within seconds, while scattering and reflection of photons around cloud edges can push irradiance above clear-sky values. Despite the importance of this variability for land-atmosphere coupling, weather and climate models rely on one-dimensional (1D) radiative transfer, which by design cannot represent the horizontal transport of photons that creates the observed patterns of shadows and enhancements. Getting the surface radiation field right is a prerequisite for getting land-atmosphere feedbacks right, which makes this problem directly relevant to the PAN-GLASS goal of improving coupled land-atmosphere models from the fundamentals up.

The SLOCS project tackled this problem on three fronts: observations, modeling, and bridging tools. On the observational side, we developed FROST, a low-cost instrument that measures shortwave radiation at 10 Hz across spectral bands from UV to near-infrared. Networks of FROST instruments were deployed during the FESSTVaL and LIAISE field campaigns, complemented by a ten-year record of 1 Hz SSI observations at the Cabauw measurement site in the Netherlands. For modeling, we developed a GPU-accelerated Monte Carlo ray tracer coupled to our LES model MicroHH, creating a virtual laboratory with fully interactive 3D radiative transfer. To connect these idealized simulations to real-world conditions, we built (LS)2D, a tool that derives initial and boundary conditions from ERA5 reanalysis for doubly-periodic LES domains.

Analysis of the field campaign data revealed that the characteristics of cloud shadows and enhancements differ strongly among cloud types: cumulus, altocumulus, and cirrus each produce distinct patterns of variability with different amplitudes, durations, and spectral signatures. Using the Cabauw long-term dataset, we demonstrated that SSI variability follows power-law distributions linked to cloud size distributions, providing a statistical framework that connects cloud geometry to surface radiation. Building on both observations and ray tracing of simulated clouds, we developed a conceptual model that explains all observed variability patterns from four mechanisms: forward escape, downward escape, side escape, and albedo enhancement. This provides a physical basis for parameterizing SSI variability in models.

The modeling results revealed that 3D cloud-radiation interactions have significant effects on cumulus convection over land. Simulations with coupled 3D radiative transfer produced deeper and larger clouds with more liquid water compared to those using conventional 1D radiative transfer. A systematic follow-up study confirmed this pattern and uncovered a surprising result: despite the substantially different cloud fields, the domain-mean surface irradiance was nearly identical between 1D and 3D simulations. This suggests the existence of a negative feedback mechanism through which the altered cloud field compensates for the redistribution of radiation by 3D effects. However, while domain-mean irradiance may be similar, the spatial distribution of surface irradiance is fundamentally different between 1D and 3D. This spatial redistribution drives differential surface heating, heterogeneous evapotranspiration, and mesoscale circulations that feed back into boundary layer growth and cloud formation. On the computational side, we showed that spectral resolution in radiative transfer can be reduced by a factor of three to four without significant loss of quality, and that machine-learning-based denoising enables an order-of-magnitude reduction in the number of required Monte Carlo samples. Together, these advances bring 3D radiative transfer within reach of routine use in cloud-resolving simulations.

We argue that incorporating 3D radiative transfer in next-generation coupled land-atmosphere models is both necessary and increasingly feasible. As the community returns to the drawing board to improve coupled land-atmosphere models, in our view, the treatment of three-dimensional cloud-radiation interactions at the surface deserves a prominent place on the agenda.

Wed

09:45–10:00

We present work of the Land-Atmosphere Feedback Initiative (LAFI) [1], which aims to quantify and understand land-atmosphere feedback by utilizing synergetic observations and simulations. Our specific focus here within LAFI lies on the quantification of entrainment, i.e., the exchange of energy and humidity between the boundary layer and the free atmosphere above.

As a first step, we investigated a proposed theoretical description of the latent heat entrainment flux at the top of a convective boundary layer. This relation is expressed as the vertical velocity scale times the ratio of the water vapor mixing ratio gradient to the Brunt-Väisälä frequency (summarized in [2]). We studied this equation using observations of Raman and Doppler lidar as well as radiosonde measurements from the US. The analysis, conducted at the lower boundary of the interfacial layer, did not support the proposed formulation. In the search for potential driving variables of the latent heat entrainment flux at this height level, we found a high correlation with the square root of the variances of vertical wind and water vapor mixing ratio.

For these first results, we utilized observations from several months between 2017 and 2019 from the Atmospheric Radiation Measurement Program’s Southern Great Plains site. Water vapor mixing ratio and vertical wind data were retrieved from turbulence-resolving lidar systems, while temperature data were obtained by interpolating radiosonde profiles on a height grid normalized with the boundary layer height [3]. The available data were screened for convective boundary layer situations, resulting in 81 selected profiles that we used for the conducted analysis [4].

Further exploration of this connection led to the conclusion that the currently most accurate description of the latent heat entrainment flux is the product of the vertical wind variance and the ratio of the water vapor mixing ratio gradient to the Brunt-Väisälä frequency, confirming recent results over Europe [5].

Within LAFI, these results will be validated through new measurements conducted during a specific field campaign by LAFI in 2025. Here, the profiling of sensible heat fluxes is also possible. First results will be presented at the conference. Furthermore, we will investigate in greater depth potential parameterizations of the vertical wind variance.

References:
[1] https://www.lafi-dfg.de/
[2] Wulfmeyer, Volker et al. (2016): Determination of Convective Boundary Layer Entrainment Fluxes, Dissipation Rates, and the Molecular Destruction of Variances: Theoretical Description and a Strategy for Its Confirmation with a Novel Lidar System Synergy. In Journal of the Atmospheric Sciences 73 (2), pp. 667–692. DOI: 10.1175/JAS-D-14-0392.1
[3] von Klitzing, Linus; Turner, David D.; Lange, Diego; Wulfmeyer, Volker (2026): Improved method for temporally interpolating radiosonde profiles in the convective boundary layer. In Atmos. Meas. Tech. 19 (1), pp. 359–370. DOI: 10.5194/amt-19-359-2026.
[4] von Klitzing, Linus; Turner, David D.; Lange, Diego; Wulfmeyer, Volker; Senff, Christoph J. (2026): How can the Latent Heat Entrainment Flux in the Convective Boundary Layer be Characterized? In JGR Atmospheres. Submitted.
[5] Gibert, Fabien; Edouart, Dimitri; Monnier, Paul; Collignan, Julie; Lopez, Julio; Cénac, Claire (2025): Scalar turbulent fluxes and variances in the interfacial layer from lidar observations and assessment of Lagrangian Stochastic Models. In Quart J Royal Meteoro Soc, Article e70016. DOI: 10.1002/qj.70016.

Mon & Tue

11:00–12:30

Recent research has shown that a considerable share of the energy balance gap is caused by energy transport through mesoscale secondary circulations. The contribution of this transport to the surface energy balance in the form of dispersive sensible and latent heat fluxes is influenced by various factors, including thermal surface heterogeneity. Previous approaches to characterize thermal surface heterogeneity have been developed only for checker-board type structures. We want to overcome their limited applicability by developing a Machine Learning method that can capture the complexity of real-world surface structures as well. To address this goal, we train a spatiotemporal artificial neural network (ANN) that first analyzes surface heterogeneity and next predicts dispersive sensible and latent heat fluxes in a spatiotemporally distributed manner. We train the network on an idealized LES dataset, representing randomized quadratically shaped heterogeneity distributions, as well as real-world complex surface heterogeneity with a broad spectrum of patch sizes. The dataset was derived from 30 daytime simulations using PALM, where we combined four different heterogeneity length scales, a realistic landcover distribution, two different Bowen ratios, and three different geostrophic wind speeds. In this way, we are able to cover a broad range of scales of traditional variables such as the heterogeneity parameter, temperature or humidity gradients, boundary layer height and atmospheric stability measures. By simulating a 12 hour daytime interval, the model covers the full daytime variability. The surface fluxes are modelled by the PALM land surface model in combination with a clear-sky radiation model. To investigate the role of the different input variables, we train the ANN on different combinations of input variables and compute feature importance weightings afterwards. Aside from the afore mentioned traditional variables, we consider the incorporation of raw input features, such as horizontal and vertical wind speed, temperatures, and humidity. Finally, we incorporate spatial temperature maps. Using ML techniques, we expect to provide a parameterization of dispersive sensible and latent heat fluxes that can be applied to correct field measurements and improve energy balance closure. Furthermore, we hope to identify key drivers of the dispersive fluxes.

Mon

13:30–13:45

Urbanization represents one of the most profound anthropogenic modifications to the land surface, which fundamentally alters surface energy and water budgets and reshapes boundary-layer dynamics. Despite their importance, urban processes remain incompletely represented in many land–atmosphere modeling systems. This largely limits our ability to quantify anthropogenic influences on surface fluxes, land–atmosphere coupling, and predictability across weather and climate timescales. This presentation will introduce recent advances in the Arizona State University Single-Layer Urban Canopy Model (ASLUM) to illustrate how explicitly resolving the interactions between buildings and soil–plant–atmosphere continuum alters surface flux partitioning and hydrometeorological conditions across climates and modeling scales. This includes the incorporation of urban hydrological modules through a multi-parameterization approach in ASLUM-Hydro, which enables the physically consistent representation of surface water partitioning. A comprehensive global evaluation of our model across over 20 urban flux tower sites spanning diverse climates demonstrates improved performance in simulated sensible and latent heat fluxes compared to previous versions. We will also introduce the recent implementation of ASLUM within the High-Resolution Land Data Assimilation System (HRLDAS), its coupling with Noah-Multiparameterization Land Surface Model (Noah-MP), and integration within the Weather Research & Forecasting Model (WRF) modeling system. Initial regional experiments over the New York City metropolitan area illustrate how improved urban representation influences near-surface hydrometeorology under heat wave conditions. The presentation will conclude with perspectives on how physically based urban canopy models can improve the representation of anthropogenic influences in land–atmosphere interactions, enhance benchmarking across observational networks, and inform future community model development. These advances are particularly relevant for understanding urban impacts on boundary-layer structure, extreme heat, and land–atmosphere feedbacks across diurnal to subseasonal timescales.

Mon & Tue

11:00–12:30

The northern slope of the Tianshan Mountains in arid northwestern China is a strongly heterogeneous mountain–oasis–desert system, where land use is tightly constrained by hydroclimatic variability and water availability. Over the past four decades, this region has undergone pronounced climatic change, with a clear abrupt warming shift after 1997, accompanied by increasing precipitation, glacier retreat, rising snowlines, and changing runoff regimes. In parallel, rapid agricultural expansion, urbanization, and energy development have substantially reshaped regional land-use patterns. However, the mechanisms through which long-term climate change has influenced land-use change across this dryland landscape remain insufficiently understood.
Here we diagnose the impact of climate change on land-use change across the northern slope of the Tianshan Mountains over the past 40 years, with particular emphasis on surface heterogeneity within the mountain–oasis–desert system. We examine changes in oasis extent, land-use structure, land-use intensity, and major transitions among cropland, grassland, woodland, built-up land, and unused land. Our analysis integrates multi-source land-use datasets, long-term meteorological observations, hydrological information, and regional statistical data within a diagnostic framework linking climate trends, water-resource change, and land-use responses.
The results show that climate change has had both enabling and constraining effects on regional land-use dynamics. Following the abrupt warming shift after 1997, improved thermal conditions and, in some basins, increased effective water availability promoted the expansion and restructuring of agricultural land. Over the past 40 years, cropland area increased by 40.65%, reflecting the combined effects of climatic warming, changing hydrological conditions, and intensified human land development. In several irrigated oasis areas, warming and wetting supported the expansion of cropland, orchards, and shelterbelts, and favored shifts toward higher-value cash crops. At the same time, climate change also increased long-term uncertainty in water supply through glacier shrinkage, reduced snow storage, and enhanced hydrological variability. In water-limited areas, these changes, together with increasing evapotranspiration and more frequent hydroclimatic extremes, constrained further land reclamation, intensified competition among agricultural, urban, industrial, and ecological water uses, and contributed to grassland degradation, soil salinization, and localized desertification risk.
These impacts are spatially heterogeneous across the region. The central piedmont oasis belt exhibits rapid growth of both cropland and built-up land, but also rising water stress. The Yili Valley remains a hotspot of agricultural expansion under relatively favorable hydroclimatic conditions, whereas the eastern Tianshan and Hami regions show weaker agricultural expansion and stronger resource constraints. Overall, our findings suggest that climate-driven land-use transitions in arid regions are mediated by surface heterogeneity and coupled hydroclimatic–human processes. This study provides new evidence on how climate change reshapes land-use trajectories in drylands, with implications for land–atmosphere interactions, sustainable land and water management, and climate adaptation.

Mon & Tue

11:00–12:30

Recent research has identified mesoscale secondary circulations as a significant contributor to the energy balance gap, as they transport energy in ways that cannot be captured by single-tower eddy covariance measurements using 30-minute averaging intervals. These circulations contribute to so- called dispersive fluxes, which depend on factors such as atmospheric stability and thermal surface heterogeneity but also the vertical distribution of the scalar of interest. Secondary circulations likely transport not only energy but also gases and particles and therefore also affect CO2 fluxes measured at flux monitoring stations worldwide. However, the lack of independent ecosystem-scale measurements of net surface CO2 fluxes makes it difficult to quantify the magnitude of this effect. Energy balance corrections of sensible and latent heat fluxes are often used as proxies to estimate errors in CO2 flux measurements, which may not be sufficient due to the differences in vertical distribution of heat, H2O, and CO2. Large-eddy simulations (LES) provide valuable insights into these processes by offering detailed information on CO2 fluxes at single points (virtual towers), surface fluxes, dispersive fluxes, and flux profiles. To advance understanding, we present an LES dataset of clear-sky daytime simulations designed to model CO2 transport through secondary circulations. This dataset includes scenarios with varying atmospheric stabilities and thermal surface heterogeneities, such as quadratic patches of different sizes, a more complex land cover pattern and a homogeneous surface. Using this dataset, an artificial neural network (ANN) is trained to predict dispersive CO2 flux contributions. This approach bridges the gap between field measurements and mesoscale processes, offering a promising path for more accurate ecosystem-scale CO2 flux assessments.

Mon

14:00–14:15

Freshwater scarcity is a pressing issue driven by over-extraction of natural resources and worsened by land use change and global warming. To address this issue, a novel approach has been proposed: Technology-Enhanced Atmospheric Moistening (TEAM), e.g. large-scale evaporation of seawater to moisten the atmosphere with the goal of triggering precipitation downwind over coastal dry-lands (Warnau et al., 2026). Two different approaches of TEAM have been suggested: spray evaporation and interfacial solar evaporation. While the evaporation rates of these methods are known for laboratory or other small-scale settings, a critical knowledge gap remains in understanding how small-scale technological evaporation capacities translate to large scale moistening potential in the atmospheric boundary layer (ABL).
Here, we explore for each evaporation technology – spraying and solar evaporation – 1) How efficient can moisture be added to the ABL under different atmospheric conditions? 2) How will downwind feedbacks impact the effective moistening capacity of the technology? To quantify and understand the relevant interactions between the technologies and the ABL we use a dual modelling approach: On the one hand, we do high-resolution large-eddy simulations (LES) to explicitly resolve turbulent ABL processes. Idealized atmospheric conditions representative of the Mediterranean basin are used to conduct a sensitivity analysis to several atmospheric parameters. Complementary to the LES, we use a slab model of the mixed ABL coupled to a surface model with implementations of the two moistening methods. Using a conceptual model can help with the interpretation of LES results and can be used as a tool for further computationally cheap sensitivity studies.
Preliminary results show that spray evaporation is highly limited by the atmospheric conditions, since the maximum moistening capacity is capped by the wet bulb temperature of the sprayed layer. Also, since the sprayed layer cools, it behaves as a stable internal boundary layer, limiting downwind evaporation from the sea surface (negative feedback). On the other hand, interfacial solar evaporation is mainly limited by the net incoming radiation. If this is large enough, enhanced buoyancy from the technology surface causes convective growth of the ABL, entraining warm and dry air from the free troposphere. Under most tested atmospheric conditions, the effect of entrainment causes a downwind enhancement of evaporation from the sea surface (positive feedback).
These findings provide insights for further development of evaporation technologies and regional implementation design to optimize the atmospheric moistening effect, as well as deepen our understanding of evaporation feedbacks in the ABL.

References
Warnau, S. N.; Theeuwen, J. J. E.; Sadeghi, G.; Benedict, I.; Hamelers, B. H. V. M.; van Heerwaarden, C. C. Technology-Enhanced Atmospheric Moistening (TEAM) for More Precipitation: A Perspective. Environ. Sci. Technol. 2026, 60 (2), 1612–1620. https://doi.org/10.1021/acs.est.5c06428.

Mon & Tue

11:00–12:30

Land–atmosphere exchange in tall and short canopies is shaped by turbulence within and above the canopy and in the roughness sublayer (RSL). Recent developments in the Noah-MP® community land surface model (LSM) include a unified turbulence parameterization intended to provide a consistent treatment of turbulence from within the canopy through the RSL to the surface layer. While this scheme has been primarily evaluated under snow dominated conditions, its impact on simulated fluxes for non snow, multi canopy environments over long time periods is still not well quantified.
In this study, we assess the representation of sensible and latent heat fluxes in Noah-MP (version 5.1.1) using multi year, multi level observations from the Lindenberg observatory of the German Weather Service (DWD). We focus on two contrasting sites: (i) Kehrigk, a tall evergreen needleleaf forest where RSL effects are expected to be pronounced, and (ii) Falkenberg, a short grassland site that more closely satisfies Monin–Obukhov similarity theory (MOST) assumptions. Both sites provide continuous 30 min data since 2005, including eddy covariance fluxes of sensible and latent heat, radiation components, soil heat flux at 5 cm depth, skin temperature, and multi level profiles of air temperature, humidity, and wind speed. All forcing and flux data undergo standard DWD quality control and energy balance closure procedures.
Noah-MP is run offline at both sites with identical land and soil parameterizations and is driven by observed meteorology at the flux height. We compare a suite of configurations. Besides standard flux evaluation, we analyse friction velocity, Monin–Obukhov length, bulk transfer coefficients for heat and moisture, and the vertical structure of wind and temperature as a function of season, canopy type, and atmospheric stability.

Wed & Thu

11:00–12:30

Monin Obukhov Similarity Theory (MOST) has long served as the basis for parameterizations of turbulence exchange between the surface and the atmospheric boundary layer (i.e. for determination of surface fluxes) in weather prediction and long term variability. Decades of research, however, has illuminated some of the limitations of MOST based surface layer parameterizations, particularly when MOST’s foundational assumptions of flat and horizontally homogeneous terrain are violated as they frequently are in our increasingly complex regional and global modeling systems. This problem is particularly salient in high resolution land surface models, where the increasing resolution further weakens assumptions of homogeneity. In addition to a wide number of traditional forms of the flux-gradient relations, modern work in surface layer parameterizations has explored ways to modify the flux-gradient relations to capture heterogeneity, roughness-sublayer effects and other deviations from MOST. In particular, recent work has leveraged the anisotropy of turbulence (yb) as an additional non-dimensional term to extend and generalize MOST to complex terrain (Stiperski 2023, Mosso 2024, Waterman 2025) and address ways that observations deviate from traditional theory.
In this presentation we explore 7 years of data from 47 sites in the National Ecological Observation Network (NEON) to explore existing surface layer schemes and generate novel surface layer parameterizations. These schemes are then implemented in single point runs of NCAR’s CTSM model at a selection of the NEON tower sites. The results show that 1) anisotropy generalized MOST can capture a large portion of the deviation from surface layer theory in observations, covering the range of many different traditional functions using a single function, 2) that the anisotropy generalized MOST relations cause a sensitivity on the order of 10-20% on the surface fluxes when compared to traditional surface layer models and 3) when the anisotropy generalized MOST, as well as other MOST forms, are implemented within a large scale model (CTSM), the expected sensitivity decreases significantly. This indicates that many modern surface layer schemes, as implemented within large scale earth system models and weather prediction schemes, may not be able to capture sensitivity to surface layer physics expected.

Mon

16:00–16:15

Observational data from a field site in the UK are used to identify conditions for the formation of radiation fog and how these subtlety differ from the conditions for the formation of dew and frost. Seven years of high-resolution atmospheric profiles were studied, along with dew meter measurements for radiation nights with stable conditions. The data show that in the absence of fog, dewfall does not correlate with negative values of the surface moisture flux derived from eddy covariance data, as would be expected by our classical surface exchange theory. Moreover, these dewfall observations occur when the surface is warmer than the immediate overlying air due to a lifted temperature minimum (LTM) at about 0.15m. LTMs were observed in 65% of cases of dew without fog and >90% of cases of dew under radiation fog.

Observations of aerosol number density and average hydrated radii show that aerosol optical extinction (and hence their radiative effect) is weakly but significantly correlated with the intensity of LTMs. Low wind speed on stable nights allows settling of aerosols which radiatively cool the air near the ground more quickly than the surface cools – thus creating LTMs. In the presence of LTMs typically dew and frost form not by condensation, but by occult deposition of water droplets onto the canopy and ground. Moreover, when the rate of removal of suspended water droplets by occult deposition is slower than the creation by radiative cooling, then the build-up of droplets in the air just above the surface results in the formation of radiation fog, leading to a potential positive feedback though the radiative cooling of the air.

These features do not follow our well established theories for surface exchange processes, and hence the current generation of Land Surface Models (LSMs) are not able to represent these observed physical phenomena. However, without including these near-surface processes the forecasting of the onset of radiation fog will remain poor. Hence developments are required to represent the settling of aerosols, their radiative cooling of the atmosphere, and any subsequent occult deposition of water droplets in light wind, stable conditions. In addition, new surface exchange theories are required to sustain the observed LTMs, which are currently unachievable in these LSMs.

Thu

09:45–10:00

As land surface models (LSMs) continue to grow in complexity and spatial resolution, there is an increasing need for comprehensive evaluation systems capable of rigorously assessing their performance. This paper presents the Open Source Land Surface Model Benchmarking System (OpenBench), a cross-platform, open-source benchmarking framework designed to evaluate state-of-the-art LSMs. OpenBench overcomes key limitations of existing evaluation frameworks by incorporating the assessment of human activity processes, supporting arbitrary spatiotemporal resolutions, and providing robust visualization capabilities. The system employs a diverse suite of statistical metrics and normalized scoring indices to enable multi-faceted evaluation of model performance. Core features include automated management of multiple reference datasets, advanced data preprocessing and quality control capabilities, and unified support for both station-based and gridded data evaluations. Through case studies encompassing river discharge, urban anthropogenic heat flux, and crop yield modeling, we demonstrate OpenBench’s capacity to diagnose model strengths and deficiencies across a wide range of spatiotemporal scales and land surface processes. The system’s modular architecture allows for the seamless integration of new models, variables, datasets, and evaluation metrics, ensuring its adaptability to evolving research requirements. OpenBench offers the research community a standardized yet extensible framework for systematic model assessment and iterative improvement. Its comprehensive diagnostic capabilities, coupled with an efficient parallel computational architecture, make it a valuable tool for both model development and operational applications across diverse domains in Earth system science.

Wed

09:00–09:15

Timely estimates of land surface temperature (LST) are critical for weather and climate prediction, particularly for modelling transport processes in the atmospheric boundary layer and assessing the effects of extreme heat and drought on the biosphere. Yet forecasting the spatiotemporal variability of LST remains challenging: the surface skin responds to forcing instantaneously and temperature is governed by multi-scale thermodynamic processes that act over heterogeneous landscapes. Existing approaches follow two paradigms: A) numerical weather prediction and its AI-driven emulators that both simulate skin temperature (SKT) large-scale at coarse resolution with reduced subgrid complexity, which can introduce region-specific biases. B) Regional studies at finer resolution use satellite-retrieved LST to extrapolate in time or space but rarely incorporate synoptic-scale atmospheric forcing. In this study, we bridge these two paradigms and produce short-term LST forecasts at geostationary resolution (5 km) covering 3–5 day horizons that resolve the diurnal cycle. Combining Meteosat SEVIRI LST retrievals with ERA5 reanalysis, we train autoregressive neural networks and evaluate temporal and spatial generalisation of their forecast skill across climatically distinct regions in Africa. Our region selection is motivated by two complementary needs: On one hand we address biases in both day- and nighttime SKT, that are especially present over arid and semiarid regions. On the other hand, LST as a proxy for soil moisture deficit can be used to initialise convective storms nowcasts. We additionally characterise systematic divergences of SEVIRI-based forecasts from SKT estimates as a regional diagnostic.

Tue

16:00–16:15

The vertical fluxes of sensible and latent heat represent a major contribution to the exchange of energy between the land surface and the atmosphere. Their adequate description in numerical weather prediction and climate models is essential to realistically simulate near-surface weather conditions. Traditionally, these heat fluxes are parameterized relying on the Monin-Obukhov Similarity Theory (MOST) or the use of the Bulk-Richardson number. These parameterizations are based on differences in wind speed, air temperature, and humidity between adjacent measurement or model levels.

Wulfmeyer et al. (2023) estimated the heat fluxes with machine learning approaches and achieved a higher accuracy compared to MOST. Additionally, the analysis revealed radiation as a key predictor. However, their analysis is based on a rather short data period in August 2017 at three nearby locations in Oklahoma, USA, which limits the generalizability of the results.

In our study we replicate and expand the findings from Wulfmeyer et al. (2023) using a multilayer perceptron model (MLP) on a dataset from the boundary layer field site (GM) Falkenberg of the German Meteorological Service. The dataset consists of soil and meteorological variables over a period of twenty years, covering various seasons, synoptic weather situations and extreme weather events.

Our preliminary findings support the role of radiation as a dominant predictor for both the latent and sensible heat fluxes. We further studied the performance of the MLP for datasets of different lengths (e.g., one month as in Wulfmeyer et al., 2023, the same month over twenty years, or complete twenty-year data sets). Additionally, we tested the impact of removing redundancy in the selection of the predictor variables and the performance of the model under extreme conditions.

Tue

16:00–16:15

Land surface phenology–the seasonal rhythm of leaf emergence and senescence–is shifting in response to climatic changes. These shifts modify how the land and atmosphere exchange energy, water, and carbon, feeding back on the climate system. This presentation explores the question: What is the overall phenological feedback emerging from land-atmosphere interactions and how strong is it? Using a fully coupled Earth system model with prescribed realistic satellite-derived land-surface phenological trends (∼2.1 days/decade earlier spring, ∼1.8 days/decade later autumn), we find: A 10-day growing-season extension triggers a global surface cooling of -0.10 ± 0.03 °C, strongest in northern high latitudes, suggesting a net negative feedback at surface level. This presentation investigates various biogeophysical and biogeochemical pathways, including changes in surface albedo, turbulent heat fluxes, carbon cycling, and cloud formation, which regulate the emergent negative feedback to climate change. This first single-model global quantification provides a prior for hypothesis testing and motivates a forward looking research agenda. Overall, phenology emerges not only as a climate responder but as an active regulator of the Earth system.

Mon & Tue

11:00–12:30

Recently, the German Research Foundation (DFG) funded the Collaborative Research Unit 5639 “Land-Atmosphere Feedback Initiative (LAFI, https://www.lafi-dfg.de)”. LAFI is an interdisciplinary consortium of researchers from atmospheric, agricultural, and soil sciences as well as from biogeophysics, hydrology, and neuroinformatics applying a novel combination of advanced research methods. The overarching goal of LAFI is to understand and quantify L-A feedbacks via unique synergistic observations and model simulations from the micro-gamma (approx. 2 m) to the meso-gamma (approx. 2 km) scales from diurnal to seasonal time scales.
LAFI addresses six research objectives and hypotheses on the understanding of L-A feedbacks: 1) alternative surface layer similarity theories, 2) the impact of land-surface heterogeneity, 3) partitioning evapotranspiration, 4) entrainment in the convective boundary layer, 5) synergistic observations of L-A feedbacks, and 6) droughts or heatwaves potentially investigated by adapted field measurements. Three Cross Cutting Working Groups on Deep Learning, Sensor Synergy and Upscaling, as well as the LAFI Multi-model Experiment foster collaboration across our twelve projects.
We will present first results of LAFI dealing with the observation of L-A system processes and feedbacks at two sites dominated by agricultural land use with unprecedented resolutions and accuracy such as 3D distributions of temperature and moisture from the soil to the canopy to the lower atmosphere, the validity of Monin-Obukhov similarity theory (MOST), and the study of turbulent transport processes in the convective boundary layer. These observations a compared with dedicated large eddy simulations optimized with respect to the representation of land-surface heterogeneity in the LAFI region close to the University of Hohenheim in Stuttgart, Germany. Based on this combination of research components, we will characterize the multi-dimensional phase space of L-A system variables with various process-based metrics over an entire vegetation period in order to reach the overarching LAFI goal.

Wed

14:30–14:45

Lake-atmosphere interactions have significant impacts on local climate, acting as indicators of climate changes. Due to low albedo, high heat capacity and low surface roughness, lakes perform differently from other land surfaces in momentum, water vapor exchange and energy budgets. Based on eddy covariance measurements over Lake Erhai in southwest China, characteristics of lake-atmosphere interactions and effects of local circulation on energy exchanges were investigated. The lake acted as a heat sink in spring and summer and a heat source in winter. The latent heat flux reached its minimum value in the morning and peaked in the afternoon. The diurnal variation of sensible heat flux was opposite to that of latent heat flux. Impact factors for the sensible heat flux were mainly the lake-air temperature difference and the product of lake-air temperature difference and wind speed. The latent heat flux was mainly affected by the vapor pressure deficit and the product of vapor pressure deficit and wind speed. The annual mean values of bulk transfer coefficients for momentum, heat, and water vapor were not equal. This result indicates that the parameterization of energy exchange in numerical models, which typically assumes equal coefficients for heat and water vapor, requires improvement. Local circulation plays a crucial role in lake-atmosphere exchange. Lake breezes dominate during the daytime while, due to different prevailing circulations at night, there are two types of nighttime breeze. The mountain breeze from the Cangshan Mountain range leads to N1 type nighttime breeze events. When a cyclonic circulation forms and maintains in the south of Erhai Lake at night, its north branch contributes to the formation of N2 type nighttime breeze events. The prevailing wind directions for daytime, N1 and N2 breeze events are southeast, west and southeast, respectively. During daytime breeze events, the lake breeze decreases the sensible and carbon dioxide fluxes, while increasing the latent heat flux. During N1 breeze events, the mountain breeze suppresses the sensible and latent heat fluxes, but enhances the carbon dioxide flux. For N2 breeze events, the southeast wind from the lake surface increases the sensible and latent heat fluxes, but decreases the carbon dioxide flux.

Wed & Thu

11:00–12:30

Over the past few decades, irrigated farmland and urban areas in the arid oasis regions of Central Asia have expanded substantially. Combined with ongoing climate change, this expansion threatens to intensify water scarcity in the coming decades, a pressing concern in a region where water resources depend predominantly on snowmelt from the Tianshan Mountains. In response, mitigation measures such as the transition from flood to drip irrigation and the widespread application of plastic mulch have already been adopted to reduce water losses. However, it remains uncertain whether these measures are sufficient to prevent extreme water scarcity under present-day and future climate conditions. This study uses the SURFEX land-surface model, enhanced with detailed representations of diverse irrigation methods (flood, drip, and sprinkler) and a novel implementation of agricultural plastic mulching, to investigate how different mitigation scenarios affect the surface energy balance and atmospheric water losses during compound hot-dry events in the region. Through a series of controlled experiments varying irrigation type, volume, and timing, we quantify the effectiveness of current mitigation strategies in reducing water losses to the atmosphere. Future work will extend this analysis through coupled simulations with a regional climate model, examining water losses under moderate and extreme climate change scenarios as well as projected expansion of irrigated farmland.

Thu

09:15–09:30

In the central and eastern Tibetan Plateau (TP), where snow cover is relatively extensive, alpine grassland is the dominant vegetation type. During the non-growing season, grasses wither and form withered grass stems (WGS). Owing to the generally shallow snow depth over the TP, WGS are often not completely buried by snow. The exposure of WGS above the snow surface can substantially influence the surface energy budget as well as the thermal and hydrological processes of the snowpack. Using a combination of remote sensing data, in situ observations, and model simulations, we demonstrate that an increase in WGS leads to a rise in ground temperature by approximately 0.4 °C, thereby accelerating snowmelt and sublimation. Consequently, the mean snow depth decreases by about 6 mm, and the snow cover fraction (SCF) is reduced by approximately 5%. Furthermore, by explicitly accounting for the effects of WGS and topographic relief, we parameterize the snow accumulation probability factor rather than prescribing it as a constant. In addition, a revised factor is introduced to modify the shape parameter of the snow depletion curve, thereby improving the SCF parameterization scheme in Community Land Model version 5 (CLM5). Preliminary validation results indicate that the optimized scheme substantially reduces positive wintertime SCF biases over barren land and grassland by 34%–88%. It also improves surface albedo simulations, which in turn alleviates cold surface temperature biases by approximately 1–2 °C in snow-affected regions.

Wed & Thu

11:00–12:30

In irrigated areas, water is artificially managed to achieve optimal crop growth. Large volumes of water are withdrawn from various sources, such as rivers, lakes, ponds, and groundwater. Throughout the water control by irrigation, soil moisture is maintained at higher levels, leading to increased evapotranspiration and latent heat flux alongside reduced sensible heat flux. Consequently, accounting for these irrigation effects is essential for modeling the distinct water and energy exchanges between the land surface and the atmosphere. While the global extent of irrigation is relatively small, the irrigated fraction exceeds 20% in certain regions. This underscores the importance of considering irrigation’s impact on atmospheric states, particularly at the local scale.
The Simple Biosphere including Urban Canopy (SiBUC) land surface model is among the few models capable of treating irrigated areas as a distinct land-use category. Based on the original SiB model, SiBUC was developed by incorporating urbanized areas, water bodies, and irrigated regions. The MRI-AGCM is one of the climate models contributing to the Coupled Model Intercomparison Project Phase 6 (CMIP6) and has been utilized to generate the d4PDF large-ensemble global climate projection dataset. Its spatial resolution is among the highest of current atmospheric general circulation models (AGCMs). While the MRI-AGCM originally incorporates a SiB-based land surface model, it does not account for the irrigated land-use category. Therefore, a derivative version of the MRI-AGCM has been developed by integrating SiBUC into the model framework.
For climate simulations accounting for irrigation, it is essential to prepare a global crop type map and a corresponding cropping calendar. In this study, we utilized the SAtellite-derived CRop calendar for Agricultural simulations (SACRA) dataset. Additionally, information on irrigated areas was derived from the Global Map of Irrigation Areas version 5 (GMIAv5). Regarding boundary conditions, sea surface temperature (SST) and sea-ice concentration (SIC) were prescribed using the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) dataset. Initial conditions were established based on the fifth-generation ECMWF atmospheric reanalysis (ERA5). Other boundary conditions and model parameters remained consistent with the original MRI-AGCM settings. Two experiments were conducted at a TL319L64 resolution: the first utilized the MRI-AGCM coupled with the SiBUC model without irrigation control (control run), while the second explicitly incorporated irrigation activities (irrigation run). Both simulations spanned from July 1979 to December 2008. To eliminate the influence of initial conditions, the first 30 months were treated as a spin-up period and excluded from the analysis; climatological values were subsequently calculated from the remaining years of data.
First, the basic performance of the MRI-AGCM coupled with SiBUC was validated against the Global Precipitation Climatology Centre (GPCC) monthly dataset, confirming it performs at a comparable level to the original MRI-AGCM. Subsequently, the differences between the two experiments were analyzed. Generally, significant deviations emerge during the boreal summer, as irrigated areas are extensive in the Northern Hemisphere and water withdrawal peaks during this season. Specifically, in the irrigation experiment, annual evapotranspiration increased by over 200 mm in the central United States and by more than 100 mm in Eastern Europe. These increases are attributed to substantial irrigation water withdrawal, which maintained soil moisture at higher levels. Consequently, the irrigation experiment showed an increase in both precipitation and runoff in these regions.
In contrast, some irrigated regions in Asia showed negligible changes in evapotranspiration. Furthermore, both precipitation and runoff decreased in the irrigation experiment for these areas. This suggests that incorporating irrigation altered the surface energy balance, which in turn modified surface wind patterns. Based on the experiments in this study, considering irrigation in global climate simulations produces both positive and negative effects on evapotranspiration, precipitation, and runoff depending on the region.

Wed & Thu

11:00–12:30

Irrigation and reservoir regulation are two dominant anthropogenic modifications of the terrestrial water cycle, profoundly influencing land–atmosphere interactions. While irrigation has been incorporated into land surface models (LSMs), estimating water withdrawals remains challenging due to the complex interplay between hydrological, agricultural, and socio-economic systems. Critically, surface water is not used locally in isolation. It is withdrawn and transferred through spatially coordinated infrastructures such as reservoirs and river channels, with return flows further modulating downstream hydrology. However, existing LSMs simulate irrigation without representing these water transfers and reservoir dynamics.
Here, we present the Conjunctive Surface-Subsurface Process Model version 3 (CSSPv3), a global LSM that innovatively couples a water-balanced, regional surface-subsurface irrigation scheme with a multi-objective, coordinated reservoir operation model. The primary innovation lies in its hierarchical, regionally coordinated surface water withdrawal mechanism. Unlike previous models that restrict withdrawals to the irrigation grid cell, CSSPv3 operates in two steps: (1) local river storage is first used; (2) if insufficient, a regional redistribution algorithm activates, sourcing surplus water from neighboring grid cells based on relative water abundance. This effectively simulates remote water transfers and coordinated surface water allocation. Only after exhausting both local and regional surface water sources is groundwater extracted from the local aquifer. Moreover, CSSPv3 considers multiple functions of reservoirs, including hydropower, irrigation water supply, and flood control.
Validated against extensive observations from 1991–2020 at a 0.25° resolution globally, CSSPv3 demonstrates marked improvements over its predecessor (CSSPv2) and 26 state-of-the-art LSM/global hydrological model products. It significantly reduces simulation errors for irrigation amounts, streamflow, surface soil moisture, and evapotranspiration, particularly in human-dominated basins and irrigated regions. By realistically representing the spatial coordination of water infrastructure, CSSPv3 establishes a robust platform for simulating coupled ecohydrological processes under both natural and anthropogenic influences, offering a powerful tool to advance Earth system modeling in the Anthropocene.

Thu

09:30–09:45

Environmental change in high latitudes is progressing faster and more intensely than in many other regions of the Earth. A warmer climate has already triggered unprecedented changes in terrestrial cryosphere, hydrosphere, and vegetation shifts, which feed back to the atmosphere and the hydrologic cycle. Yet, many studies still rely on one-way coupling between the atmosphere and the land surface, thereby neglecting important interactions and feedback mechanisms. To advance our knowledge on climate change in northern environments, LATICE (Land–ATmosphere Interactions in Cold Environments) was established as a strategic research initiative at the University of Oslo in 2015. Since then, it has developed into a collaborative network of researchers from atmospheric and terrestrial research groups. We combine novel observations, process-based model development, and integrated Earth observation frameworks to improve our understanding of high latitude environmental processes and their representation in Earth System Models – particularly the Norwegian Earth System Model (NorESM).

The current LATICE infrastructure including flux towers, drone-based measurements, and a cold climate container contributes to expanding our observational data records in cold environments complemented by regionally calibrated satellite products tailored to boreal and Arctic conditions. These coordinated in situ, airborne, and spaceborne datasets provide important constraints on surface fluxes, snow dynamics, and vegetation processes. On the modelling side, LATICE researchers have been contributing to land surface and hydrological models to better represent cold region processes by including process parameterizations, new plant functional types, and vegetation demography. The integration of observations and models through data assimilation methods, and further exploration of hybrid approaches with machine learning algorithms support improved understanding of land-atmosphere coupling and uncertainty quantification. This presentation highlights key activities and scientific advances within the LATICE community, presents example studies, and discusses future directions for scientific collaboration with the Pan-GLASS community.

Mon & Tue

11:00–12:30

Global terrestrial ecosystems are showing a troubling decline in carbon sequestration—driven by nutrient shortages and intensifying drought stress—that current land-surface models struggle to reproduce. Satellite observations reveal a widening gap between real-world CO₂ uptake and model projections, highlighting deficiencies in how we represent nutrient dynamics, water limitations, and long-term vegetation acclimation. Closing this gap demands an integrated, physics-based framework that fully leverages Earth Observation—from VNIR, SWIR, TIR, and microwave sensors—to trace water–energy–carbon–nutrient interactions across scales.
To this end, we propose SPACEO (Soil-Plant-Atmosphere Continuum enhanced by Earth Observation), a digital twin framework that couples the STEMMUS-SCOPE soil-plant process model with the radiative-transfer models. By simulating signals from VNIR, SWIR, to TIR end-to-end, SPACEO uniquely links satellite measurements to the underlying ecophysiology of the soil-plant system.
SPACEO’s research is organised into three interlocking science cases:
1. Unified Forward Simulation: With STEMMUS-SCOPE, we build a single simulator that coherently predicts multi-frequency EO signals, establishing the scientific foundations for synergy of multi-satellite multi-frequency EO data.
2. Advanced Retrievals & Reference Dataset: Drawing on extensive field campaigns, we will build a FAIR-compliant Reference Dataset and develop data-driven, hybrid, and physics-informed machine-learning algorithms to estimate key ecosystem metrics—Nitrogen Use Efficiency (NUE), Water Use Efficiency (WUE), canopy and soil temperatures, and stress indicators—from optical, fluorescence, thermal, and microwave data.
3. Digital-Twin Data Assimilation: By assimilating multi-mission observations into our digital twin, we will generate self-consistent soil and plant states and fluxes across diverse biomes. We will quantify how multi-frequency data tighten model constraints versus single sensors and translate these insights into design blueprints for next-generation, multi-sensor monitoring systems.

Wed & Thu

11:00–12:30

The increasing frequency of severe drought events poses a growing threat to terrestrial gross primary productivity, with nitrogen availability emerging as a critical modulating factor. This study looks at how nitrogen cycling affects how ecosystems respond to drought by comparing Earth System Models from CMIP6 that include both carbon and nitrogen processes with those that only consider carbon. To assess drought sensitivity, we quantified average productivity deviations during extreme dry periods and employed a decomposition approach that separates total GPP into contributions from canopy structure and physiological activity per unit leaf area. Additionally, we examined ecosystem water-use efficiency and its constituent components to unravel underlying carbon-water coupling mechanisms.
Our analysis reveals contrasting regional patterns: models with a nitrogen cycling project amplified GPP losses in water-limited environments (-0.25 PgC yr⁻¹) but comparatively muted declines in energy-limited humid zones (-0.37 PgC yr⁻¹). These divergent responses stem from systematic alterations in water-use strategies driven by nitrogen constraints. In arid regions, elevated water-use efficiency intensifies LAI sensitivity to moisture stress, exacerbating productivity reductions. Conversely, humid regions exhibit diminished water-use efficiency that buffers leaf-level photosynthetic suppression during dry periods. These findings show that nitrogen cycling significantly changes how plants adjust their physiology and structure when there is not enough water, leading to different responses to drought in different climates. Our results underscore the necessity of integrating carbon-nitrogen interactions for reliable carbon cycle projections under climate extremes and highlight critical knowledge gaps requiring targeted field experimentation to validate modeled mechanisms.

Wed & Thu

11:00–12:30

Atmospheric Radiation Measurement (ARM) program (supported by US DOE) hosts longterm ground-based observations at various locations worldwide, which provide high resolution comprehensive data from bedrock to top of atmosphere. We use these data to help characterize the interactive processes among land surface, planetary boundary layer (PBL), clouds and precipitation and focus on both the terrestrial leg between soil moisture and surface fluxes and the atmospheric leg between surface fluxes, PBL turbulence and convective triggering. Global earth system models have approached kilometer scale to resolve convective systems in recent years, which significantly bridged the gap between models and ground-based site observations for model validation and improvement of their representation on physical processes. In this study, we adopt a hierarchy of modeling tools developed by Energy Exascale Earth System Model (E3SM) including decadal-long global simulations by Simple Cloud-Resolving E3SM Atmosphere Model (SCREAM) and golden-case-wise regional simulations by a stand-alone doubly periodic SCREAM (DP-SCREAM). We focus on several ARM sites to systematically evaluate SCREAM’s performance on representing land-atmosphere interactive relationships under local-coupling regimes including clear-sky PBL, surface-forced fair-weather shallow cumuli and their transition into deep convection. By doing so, we aim to dissect model biases and attribute model error sources to advance mechanistic understanding of physics and to provide insights on possible pathways of model improvement.
This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract DE‐AC52‐07NA27344, Lawrence Livermore National Security, LLC.

Wed

09:00–09:15

Bryophytes and lichens in permafrost regions form a porous, air-filled ground cover that directly regulates surface water and energy exchange. The retained water content of the bryophyte and lichen layer varies in response to thermal exchange between atmosphere and soil systems, thereby modulating evaporation. Climate change is driving alterations in functional diversity of these highly sensitive non-vascular vegetation communities. Shifts in functional traits will likely affect water retention and transport capacity, in addition to the thermal properties of the bryophyte and lichen layer. However, it is largely unclear how changes in functional diversity will affect overall water exchange and the surface energy balance. Yet this gap may be addressed by trait-based models that simulate the mutual interaction between biodiversity and evaporation.

This study focuses on bryophyte and lichen vegetation in high-latitude permafrost ecosystems, aiming to: (1) quantify their interception effects on surface water and energy exchange under climate change, and (2) clarify the underlying mechanisms by which functional diversity modulates the surface energy balance. To this end, we refine the permafrost processes within the trait- and process- based LiBry model to accurately capture the coupled states of thermal-water exchange and diversity. Model experiments to isolate effects of bryophyte and lichen layer are implemented to determine their contribution to variations. We further drive the model with a gradient of climate and diversity scenarios to reveal the relationships between distribution of functional traits and water and energy exchange. Our findings contribute to a more comprehensive understanding of the impacts of functional diversity on key permafrost processes in data-scarce contexts.

Mon & Tue

11:00–12:30