Martini et al. (2026) From near surface to root zone soil water losses: a new model validated with field TDR and remotely sensed data

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-03-24
• Authors: Tommaso Martini, Aurora Olivero, Alessio Gentile, Davide Gisolo, Stefano Ferraris
• DOI: 10.1016/j.jhydrol.2026.135355

Research Groups

• Interuniversity Department of Regional and Urban Studies and Planning (DIST), Polytechnic and University of Turin, Italy.

Short Summary

This study introduces muSEC, a new multilayer soil moisture model based on the surface evaporative capacitor (SEC), designed to link surface and root zone soil moisture drydown dynamics. Validated against three other models using field TDR and satellite SMAP data across contrasting soils, muSEC demonstrates superior transferability from surface to root zone, supporting its potential for agricultural water management.

Objective

• To evaluate the performance of four models (LAIO, DES, muSEC, HYDRUS-1D) against multiyear in-situ TDR soil moisture measurements in a sandy grassland soil profile at multiple depths (0–15 cm, 0–30 cm, 0–60 cm) to reproduce observed drydown events.
• To test the four models against NASA SMAP Level-4 soil moisture retrievals in a nearby loam agricultural region, assessing their applicability for predicting drydowns at satellite sensing scales and under different land-use conditions, considering effective sensing depths of 0–5 cm and 0–10 cm.
• To propose muSEC as a physically based simple multilayer model for extending remotely sensed soil moisture to the whole root zone during drydowns, investigating whether parameters calibrated at surface depth (0–15 cm) can predict root zone dynamics (0–60 cm).
• To explore the possibility of applying a single parameter set consistently across depths within each soil type for all four models and two sites.

Study Configuration

• Spatial Scale:
  • Grugliasco experimental site (NW Italy): Permanent grassland, sandy soil profile (80% sand, 14% silt, 6% clay).
  • Villanova Solaro (NW Italy): SMAP pixel (9 km resolution), homogeneous maize cultivation, loamy soil (43% sand, 46% silt, 11% clay).
  • Depths: 0–15 cm, 0–30 cm, 0–60 cm (TDR measurements); 0–5 cm, 0–10 cm (SMAP effective sensing depth).
• Temporal Scale:
  • Grugliasco: Multiyear (2003–2008) daily mean TDR measurements.
  • Villanova (SMAP): Multiyear (2022–2024) daily SMAP Level-4 surface soil moisture retrievals.
  • Model simulations: Hourly time step, aggregated to daily for comparison.
  • Seasonal analysis: Growing season (1 April–30 September) and dormant season (1 October–31 March).

Methodology and Data

• Models used:
  • muSEC (multilayer uptake Surface Evaporative Capacitor model): New multilayer model extending the SEC framework with field-capacity threshold, explicit transpiration, and vertical redistribution.
  • LAIO (Laio bucket scheme): Simplified single-layer model with lumped evapotranspiration and exponential hydraulic conductivity.
  • DES (Desorptivity-based Evaporation Model): Single-layer model with explicit evaporation and transpiration partitioning, power-law hydraulic conductivity, and desorptivity-limited evaporation.
  • HYDRUS-1D: Mechanistic Richards equation numerical solver for unsaturated flow.
• Data sources:
  • In-situ TDR measurements: Volumetric soil moisture at 0–15 cm, 0–30 cm, 0–60 cm depths (Grugliasco).
  • Satellite remote sensing: NASA SMAP Level-4 surface soil moisture retrievals (0–5 cm top layer, 9 km resolution) (Villanova Solaro).
  • Meteorological data: Automated weather stations (Grugliasco, Villanova Solaro) providing air temperature, relative humidity, wind speed, precipitation, and solar radiation.
  • Leaf Area Index (LAI): MODIS (MOD15A2H, 500 m resolution, 8-day composites) for Villanova; constant LAI for Grugliasco.

Main Results

• All four models reproduced soil moisture drydowns with good accuracy across sites and depths, with Root Mean Squared Error (RMSE) consistently lower than 0.06 m³ m⁻³.
• Predictive skill improved with depth at the Grugliasco site, with RMSE values ranging from 0.003–0.010 m³ m⁻³ at 0–60 cm compared to 0.005–0.016 m³ m⁻³ at 0–30 cm.
• Model performance was generally better in the dormant season than in the growing season, especially at shallower depths.
• For SMAP data at Villanova, simulations assuming a 0–10 cm effective sensing depth yielded slightly lower RMSE values than 0–5 cm, suggesting SMAP retrievals may integrate signals over a deeper layer.
• In the surface-to-root zone transfer experiment (calibration at 0–15 cm, validation at 0–60 cm at Grugliasco), muSEC demonstrated superior transferability among simplified models, achieving an RMSE of 0.021 m³ m⁻³ at 0–60 cm validation. This indicates that surface-calibrated parameters can predict root zone dynamics in relatively homogeneous sandy soils.
• The LAIO model provided a robust baseline, while the DES model often overestimated early drying. muSEC offered the most balanced performance and strong cross-depth generalization. HYDRUS-1D provided accurate calibration but did not systematically outperform muSEC in validation, despite higher computational cost.
• Global calibration (single parameter set per site across all depths and years) showed stable performance, with RMSE values around 0.008–0.012 m³ m⁻³ for Grugliasco and 0.016–0.020 m³ m⁻³ for Villanova.

Contributions

• Development and validation of muSEC, a novel multilayer soil moisture model that extends the Surface Evaporative Capacitor (SEC) framework by incorporating explicit transpiration, a field-capacity threshold, and a multilayer structure, enhancing its ability to represent vertical soil moisture dynamics.
• Comprehensive benchmarking of muSEC against three other models (LAIO, DES, HYDRUS-1D) using diverse datasets, including multiyear in-situ TDR measurements and satellite SMAP retrievals, across contrasting soil types and depths.
• Demonstration of the practical utility of surface-calibrated parameters for predicting root zone soil moisture dynamics, particularly highlighting muSEC's superior performance in transferring predictive skill from surface to root zone, which is crucial for integrating remote sensing data into agricultural water management.
• Quantification of model performance across various spatial and temporal scales, including seasonal variability and the assessment of effective sensing depths for satellite products, providing valuable insights into the trade-offs between model complexity and predictive performance in soil moisture modeling.
• Reinforcement of the "loss function estimation" approach for drydown modeling, emphasizing its relevance for predictive applications in irrigation management and drought monitoring.

Funding

• Italian Ministry of University and Research (MUR) under the PNRR – M4C2, Investment 1.5 (grant no. ECS00000036) for the NODES project.
• PRIN 2022 Project SUNSET (grant no. 202295PFKP).
• 2021 Funding Programme of Fondazione CRT (grants no. 2022.0998, 2023.0369, and 2025.0780).

Citation

@article{Martini2026From,
  author = {Martini, Tommaso and Olivero, Aurora and Gentile, Alessio and Gisolo, Davide and Ferraris, Stefano},
  title = {From near surface to root zone soil water losses: a new model validated with field TDR and remotely sensed data},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135355},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135355}
}