Minutes of the SoilWat meeting 2025

Summary Document SoilWat_2025 (University of Reading, 14-16 July, 2025)

Table of Contents

Workshop Overview

1. Context and Purpose of the Workshop                                                                                                       
2. Key Themes and Insights                                                                                                                           

Soil Hydraulic Properties and Pedotransfer Functions (PTFs)

Uncertainty and Model Evaluation

Dynamic Soil Properties and State Variables

Scaling and Heterogeneity

Plant Hydraulics and Vegetation Dynamics

Mortality and Vegetation Response to Drought

Hybrid Modelling and Machine Learning

Observational Needs and Data Gaps

Collaborative Activities                                                                                                                                               

Unified Hydrothermal Parameterisation

Dynamic Soil Properties

Soil–Surface–Atmosphere Interface

Root Dynamics and Soil–Plant Interactions

Data-Driven Soil Property Estimation

Position Paper

Concluding Remarks                                                                                                                                                   

Working Group Proposals                                                                                                                                   

Working Group 1: Unified Hydrothermal Parameterisation

Working Group 2A: Dynamic Soil Properties and Pore Size Distribution

Working Group 2B: Anthropogenic Impacts and Soil Management Effects

Working Group 3: Soil–Surface–Atmosphere Interface Processes

Working Group 4: Root Dynamics and Soil–Plant Interactions

Working Group 5: Data-Driven Soil Property Estimation

🌍 Context and Purpose of the Workshop

The SoilWat Workshop brought together experts in soil physics, land surface modelling, micrometeorology, hydrology, and related fields to address challenges in modelling soil–vegetation–atmosphere interactions. Session 5 focused on synthesising insights from earlier sessions, identifying research gaps, and proposing collaborative activities and future directions.

🔑 Key Themes and Insights

1. Soil Hydraulic Properties and Pedotransfer Functions (PTFs)

• Challenge: Traditional PTFs are often derived from point-scale measurements and may not scale well to the grid or landscape level.
• Observation: Issues with scaling of soil hydraulic properties, e.g. effective hydraulic conductivity at larger scales (e.g. 100 m) is often lower than predicted by PTFs due to lateral redistribution of water and surface heterogeneity.
• Need: Development of scale-aware or scale-specific PTFs, potentially informed by remote sensing and data assimilation.
• Concern: Many PTFs are outdated (e.g. 30–60 years old), region-specific, or based on limited data, and may not reflect current soil and vegetation/land use conditions, and management approaches.

2. Uncertainty and Model Evaluation

• Uncertainty Sources:
  • Soil maps and their translation into model parameters.
  • PTFs and their applicability across regions.
  • Model structure (e.g. Richards equation assumptions).
• Evaluation Gaps:
  • Lack of robust, scale-appropriate observations.
  • Need for “grey-box” evaluation: assessing internal model states, not just outputs.
• Proposal: Use of data assimilation and machine learning to constrain PTFs using satellite-derived soil moisture dynamics.

3. Dynamic Soil Properties and State Variables

• Concept: Soil properties (e.g. porosity, water retention curve, hydraulic conductivity) should be treated as dynamic ‘state’ variables, not static inputs.
• Drivers of Change:
  • Wetting/drying cycles.
  • Freeze/thaw.
  • Tillage and land management.
  • Organic matter dynamics.
• Implication: Need for models that can simulate feedbacks between structure, hydrology, and biogeochemistry.

4. Scaling and Heterogeneity

• Issue: Models often assume homogeneity within grid cells, but real landscapes are heterogeneous.
• Approach: Use of clustering and tiling schemes (e.g. HydroBlocks) to represent sub-grid variability.
• Insight: Sub-grid heterogeneity can significantly affect land–atmosphere coupling, including boundary layer development and convection, e.g. via secondary circulations.

5. Plant Hydraulics and Vegetation Dynamics

• Discussion:
  • Importance of including (dynamic) plant hydraulic traits in land surface models.
  • Limitations of plant functional types (PFTs) for parameterisation.
  • Need for dynamic rooting depth and root distribution models.
• Challenge: Lack of global, high-quality data on plant hydraulic traits and root systems.
• Proposal: Move towards trait-based or hydraulic functional types, supported by targeted data collection.

6. Mortality and Vegetation Response to Drought

• Observation: Current models often use arbitrary thresholds for drought-induced mortality, which lack empirical support.
• Complexity: Mortality is influenced by multiple interacting factors (e.g. water stress, pests, carbon reserves).
• Need: More mechanistic understanding and better data to inform mortality models.

7. Hybrid Modelling and Machine Learning

• Applications:
  • Parameter estimation and sensitivity analysis.
  • Surrogate modelling (e.g. AI-Land).
  • Data assimilation and observation operators.
• Caution: ML models must be interpretable and physically consistent; risk of overfitting or misrepresentation if used blindly.
• Opportunity: Use ML to emulate complex models, accelerate calibration, and identify structural model errors.

8. Observational Needs and Data Gaps

• Critical Gaps:
  • Unified soil hydraulic and thermal properties across the entire wetness range.
  • Evolution of soil structure/different types of porosity (e.g. macro v. micro) and pore-size distribution (PoSD)
  • Root distribution and dynamics, and its dependence on/interaction with PoSD.
• Suggestions:
  • Leverage underused data sources (e.g. engineering databases, lysimeters, citizen science).
  • Improve standardisation and accessibility of soil and vegetation trait data.

🤝 Proposed Collaborative Activities

Several initiatives, with some of them evolving into working groups were proposed:

1. Unified Hydrothermal Parameterisation:
   • Develop and test consistent parameter sets for hydraulic and thermal properties.
   • Use shared datasets and cross-model comparisons.
2. Dynamic Soil Properties:
   • Explore modelling frameworks where soil properties evolve over time.
   • Link to land management, climate, and vegetation feedbacks.
3. Soil–Surface–Atmosphere Interface:
   • Focus on infiltration, evaporation, and vapour fluxes.
   • Address numerical and conceptual challenges in representing surface processes.
4. Root Dynamics and Soil–Plant Interactions:
   • Develop dynamic root models.
   • Investigate feedbacks between roots and soil structure.
5. Data-Driven Soil Property Estimation:
   • Use flux data and remote sensing to infer soil properties.
   • Combine bottom-up (e.g. mechanistic modelling informed by ML) and top-down (e.g. (proximal) RS) approaches.
6. Position Paper:
   • Draft a community paper outlining key challenges, opportunities, and a roadmap for future research.
   • Target journals like JAMES, Reviews of Geophysics, or Soil.

📌 Concluding Remarks

• The workshop highlighted the need for closer integration between soil scientists and land surface modellers, with a focus on scaling, uncertainty, and dynamic processes.
• There was strong support for collaborative, cross-disciplinary working groups to tackle shared challenges.
• The community recognises the urgency of improving soil representation in Earth system models to better predict climate impacts, water resources, and ecosystem responses.

Working Group Proposals

Here is a comprehensive proposal for the six working groups that emerged from the SoilWat Workshop Session 5. Each proposal includes the group's purpose, objectives, proposed activities, expected outcomes, and potential leads and collaborators.

Working Group 1: Unified Hydro-thermal Parameterisations

Purpose

To develop and test consistent, physically-based parameterisations that unify soil hydraulic and thermal properties across scales and models, improving model realism, consistency, and predictive skill.

Objectives

• Evaluate existing unified models (e.g. Lu et al. (2016), Lu and McCartney (2024), Fu et al.  for hydraulic–thermal coupling.
• Test performance across different soil types, climates, and land uses.
• Develop global parameter maps using remote sensing and soil databases, combined with ML approaches (e.g. following Gupta et al. 2021 type of approaches).

Activities

• Cross-model implementation and benchmarking (e.g. EC-Land, CLM, CABLE, JULES, HydroBlocks, Orchidee, STEMMUS-SCOPE, Noah-MP).
• Use flux tower and lysimeter data for validation, including PLUMBER-2 data.
• Collaborate with data providers (e.g. ISMC, COSMOS-UK, SOPHIE).

Expected Outcomes

• A validated set of unified hydrothermal parameterisations.
• Improved model consistency and reduced parameter redundancy.
• A shared repository of global parameter maps.

Lead Contacts

Rajsekhar Kandala/ Anne Verhoef (EC-Land); Josh Heitman/Yongwei Fu (to provide alternative approaches for testing),

Others Involved

Any interested land modellers, e.g. Filip Kialka (Orchidee), John Edwards (JULES), Nathaniel Chaney (Hydroblocks or NOAA GFDL models), Yijian Zeng (STEMMUS-SCOPE), Matthias Cuntz (CABLE-POP), Mengyuan Mu (LPJ-GUESS & CABLE-GW) etc., Surya Gupta, Tobias Weber, Lutz Weihermueller, Attila Nemes etc.

Working Group 2A: Dynamic Soil Properties and Pore Size Distribution

Purpose

To explore modelling frameworks where soil properties evolve over time (seasons to years) due to e.g. climate, vegetation, land management effects (see Fig. 1).

Figure 1. Interactions between soil texture, soil structure, pore-size distribution (PoSD), and ensuing soil hydraulic, thermal, mechanical, and biochemical properties and their environmental drivers and pressures.

Objectives

• Treat soil hydraulic and thermal properties as dynamic ‘state’ variables (ideally also consider mechanical and key biological properties).
• Link pore size distribution to soil structure, organic matter content, and biophysical processes, including root growth.
• Investigate interactions and feedbacks between soil evolution and land–atmosphere interactions.

Activities

• Develop and test dynamic pore size distribution models, e.g. building on mechanistic approaches already available within the SoilWat community (USSF; Jarvis et al., 2024).
• Integrate with land surface models.
• Use long-term datasets from dedicated observatories (e.g. the Agroscope Soil Structure Observatory) for model testing

Expected Outcomes

• A conceptual and computational framework for dynamic soil properties.
• Improved representation of soil–vegetation feedbacks.
• Guidance for model parameterisation under changing conditions.

Lead Contacts

Filip Kialka, Nick Jarvis, Sara Bonetti/Taiqi Lian, Anne Verhoef

Others Involved

Martine van der Ploeg, Yijian Zeng, Yong Wang….

Working Group 2B: Anthropogenic Impacts and Soil Management Effects

Purpose

To incorporate land use, management, and disturbance effects into soil modelling frameworks.

Objectives

• Investigate how to translate effects of anthropogenic activities (e.g. tillage, grazing, erosion) into model-relevant parameters.
• Develop “Anthropo-Environmental Transfer Functions” for soil models.
• Evaluate long-term impacts on soil structure and function (see WG 2A).

Activities

• Compile literature data (e.g. via meta-analyses) and datasets on land management practices and related soil responses.
• Collaborate with agricultural and forestry practitioners and modellers.
• Develop parameterisation schemes for e.g. compaction, erosion and organic matter changes.

Expected Outcomes

• Improved representation of managed soils in Earth system models.
• Tools for scenario analysis under changing land use.
• Integration with socio-environmental modelling efforts.

Proposed Lead Contacts

E.g. Thomas Keller (colleague of Nick Jarvis),...

Others Involved

Anne Verhoef,....

Working Group 3: Soil–Surface–Atmosphere Interface Processes

Purpose

To improve representation of key interface processes such as infiltration, evaporation, surface soil heat flux and vapour transport

Objectives

• Address numerical and conceptual challenges in modelling surface fluxes.
• Evaluate the role of vapour adsorption and film flow in dry soils.
• Improve soil evaporation modelling, ideally via direct new approaches mentioned in WG1 & WG2
• Improve infiltration modelling under variable surface conditions.

Activities

• Benchmark infiltration and evaporation schemes across models.
• Use high-resolution soil moisture and temperature data.
• Explore dual-porosity and multi-layer approaches.

Expected Outcomes

• Better representation of surface fluxes in land surface models.
• Reduced biases in energy balance closure and soil moisture dynamics.
• Recommendations for standardising interface process modelling.

Proposed Lead Contacts

John Edwards, Matthias Cuntz, Sinikka Paulus, Yunquan Wang, Isaac Towers

Others Involved

Nurit Agam, Anne Verhoef, Yijian Zeng, Jan Vanderborght, Mengyuan Mu........

Working Group 4: Root Dynamics and Soil–Plant Interactions

Purpose

To develop dynamic root models and root water uptake models  and explore their impact on soil structure, water uptake, and vegetation resilience.

Objectives

• Move beyond static root depth and root architecture assumptions.
• Link root growth and other root related processes (compaction, exudation etc.) to soil pore structure and hydraulic-thermal properties.
• Evaluate root–soil feedbacks under drought and land use change.

Potential Activities

• Develop trait-based or optimality-based root models.
• Use root trait databases and field observations (e.g. from lysimeters or rhizotrons).
• Integrate with plant hydraulic models and models of vegetation dynamics.

Expected Outcomes

• Dynamic root modules for land surface and crop models.
• Improved prediction of drought response and mortality.
• Better understanding of root–soil feedbacks.

Lead Contacts

Jan Vanderborght, Manon Sabot, Andrea Carminati….

Others Involved

Shijie Jiang, Yijian Zeng, Mengyuan Mu

Working Group 5: Data-Driven Soil Property Estimation

Purpose

To use flux data, remote sensing, and machine learning to infer soil properties at relevant scales. This includes the Vegetation as a Soil Sensor (VaaSS) approach to be tested by members of the SoilWat community (e.g. the University of Reading-ECMWF, University of Twente AFESP VaaSS project)[8] 

Objectives

• Combine bottom-up (soil texture, lab hydraulic data) and top-down (fluxes, remote sensing) data-model fusion approaches that directly use Earth Observables (rather than derived properties).
• Develop scalable methods for estimating soil properties using nested remoting sensing observations
• Address regional gaps by focused studies (e.g. on the Tibetan Plateau)

Activities

• Use AI and hybrid modelling for parameter estimation.
• Leverage satellite data (e.g. SMAP, Sentinel, VOD).
• Validate against in-situ network data (e.g. ISMN, COSMOS, PLUMBER2, PSInet).

Expected Outcomes

• Scalable, data-driven soil property maps that can inform efforts under other WGs.
• Improved parameter estimation for land surface models.
• Frameworks for uncertainty-aware modelling.

Lead Contacts

Yijian Zeng, Nina Raoult, Mengyuan Mu, Liz Cooper

Others Involved

Surya Gupta, Raj Kandala, Xuelong Chen, Xianhong Meng, Jan Vanderborght, Sinikka Paulus