Asfaw et al. (2026) Calibration using downscaled and bias-corrected satellite soil-moisture data can improve watershed model representation of soil-moisture variability

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-09-24
• Authors: Binyam Workeye Asfaw, Siam Maksud, Daniel R. Fuka, Amy S. Collick, Robin R. White, Zachary M. Easton
• DOI: 10.5194/hess-30-5999-2026

Research Groups

• Department of Biological Systems Engineering, Virginia Tech, Blacksburg, VA, USA
• Department of Agricultural Science, Morehead State University, Morehead, KY, USA
• School of Animal Sciences, Virginia Tech, Blacksburg, VA, USA

Short Summary

This research evaluated whether calibration using downscaled and bias-corrected satellite soil moisture improves prediction of field-scale soil moisture relative to conventional streamflow-based calibration. The study found that leveraging downscaled satellite soil moisture data substantially improved estimation of temporal soil moisture variability without affecting model streamflow performance, especially with multi-objective calibration.

Objective

• To evaluate whether downscaled and empirically bias-corrected satellite soil moisture products can improve internal hydrologic state estimation in a small (approximately 14.5 km²), saturation-excess dominated watershed using the SWAT-VSA model.
• To test whether satellite soil moisture-informed calibration improves temporal soil moisture variability without degrading streamflow performance.
• To determine if multi-objective calibration can balance predictive accuracy with hydrologic realism.

Study Configuration

• Spatial Scale:
  • Watershed: Upper Stroubles Creek watershed, Montgomery County, Virginia, USA (14.5 km²).
  • Monitoring Field: 4.2 ha mixed-grass pasture with 25 in-situ soil moisture measurement locations.
  • Satellite Soil Moisture Resolution: Downscaled to 500 m.
  • Model Soil Layers: 0–50 mm and 50–120 mm.
• Temporal Scale:
  • Downscaled Satellite Soil Moisture Data: April 2015 to August 2022.
  • In-situ Soil Moisture Measurements: 20 sampling dates between March 2023 and January 2024.
  • Streamflow Data: January 2013–December 2021 (calibration), January 2022–May 2024 (evaluation).
  • Model Time Step: Daily.

Methodology and Data

• Models used:
  • Soil and Water Assessment Tool – Variable Source Area model (SWAT-VSA).
  • ArcSWAT version 2012.10_8.26 (for model initialization).
  • TopoSWAT (ArcGIS plugin for VSA initialization).
  • R packages: "mlhrsm" (machine learning for high-resolution soil moisture downscaling), "DEoptim" (differential evolutionary algorithm for single-objective calibration), "EcoHydRology" (baseflow separation), "zoo" (streamflow aggregation), "FedData" (NLCD data acquisition).
  • Python library: "pymoo" (NSGAII for multi-objective calibration).
• Data sources:
  • Satellite Soil Moisture: Enhanced SMAP 10 km resolution product, downscaled to 500 m using the "mlhrsm" R package. Downscaling predictors included Sentinel-1 backscatter, MODIS land surface temperature, Landsat-derived vegetation indices, 10 m USGS digital elevation model (DEM), POLARIS soil properties, and NLCD land-cover classes. An event-based bias-correction was applied.
  • In-situ Soil Moisture: Time Domain Reflectometry (TDR) measurements at 25 locations (0–120 mm depth) within a 4.2 ha pasture, aggregated using Topographic Index (TI) class weighting.
  • Streamflow: Daily mean streamflow from the Virginia Tech StREAM Lab monitoring station.
  • Land Use: USGS National Land Cover Database (NLDC) 2019.
  • Topography: USGS 3DEP program 1 m resolution elevation data.
  • Soils: FAO-UNESCO Digital Soil Map of the World.
  • Weather: Global Historical Climatology Network (GHCN) data.

Main Results

• The downscaled and bias-corrected satellite soil moisture product exhibited increased peak responses following precipitation events and a greater dynamic range compared to the uncorrected product.
• Streamflow estimation performance was comparable across all three calibration approaches (streamflow-only, soil-moisture-only, multi-objective) during both calibration and evaluation periods, with Nash-Sutcliffe Efficiency (NSE) values ranging from 0.54 to 0.58.
• Multi-objective calibration achieved identical streamflow NSE (0.58) and Root Mean Squared Error (RMSE) (0.32 m³/s) as streamflow-only calibration, but with a more negative Percent Bias (PBIAS) (−10.7 %).
• Soil-moisture-only calibration resulted in a positive streamflow bias (28.7 %) and consistently overestimated low flows.
• Calibration directly targeting soil moisture (soil-moisture-only and multi-objective) significantly improved field-scale soil moisture estimation (R² = 0.88, RMSE = 0.03 m³/m³) compared to streamflow-only calibration (R² = 0.5, RMSE = 0.05 m³/m³). The multi-objective approach reduced PBIAS for field-scale soil moisture to −0.8 %.
• Multi-objective calibration reduced parameter uncertainty and equifinality, particularly for soil-water and evapotranspiration-related parameters (Available Water Content, Saturated Hydraulic Conductivity, Soil Evaporation Compensation Factor, Plant Evaporation Compensation Factor), indicating more robust parameter sets.
• Water balance partitioning showed similar surface runoff fractions across models, but the soil-moisture-only model produced lower evapotranspiration and higher total streamflow, linked to reduced soil moisture variability and Available Water Content.

Contributions

• Demonstrates that downscaled and empirically bias-corrected satellite soil moisture data can effectively improve internal hydrologic state estimation, specifically field-scale soil moisture variability, in small, saturation-excess dominated watersheds when integrated with a terrain-informed model structure like SWAT-VSA.
• Highlights that multi-objective calibration, combining both streamflow and satellite soil moisture, offers a more balanced and robust approach, enhancing soil moisture performance while maintaining streamflow accuracy and significantly reducing parameter uncertainty and equifinality compared to single-objective calibration.
• Reinforces the critical importance of evaluating hydrologic realism (e.g., water balance consistency, independent field observations) in addition to statistical performance metrics when incorporating satellite data for model calibration, to avoid parameter compensation effects.
• Advances the understanding of how remotely sensed soil moisture can meaningfully constrain watershed model parameters in small, terrain-controlled catchments, providing a pathway for improved hydrological modeling in data-limited regions.

Funding

• USDA CPS program
• National Institute of Food and Agriculture (grant no. 2021-67021-34769)

Citation

@article{Asfaw2026Calibration,
  author = {Asfaw, Binyam Workeye and Maksud, Siam and Fuka, Daniel R. and Collick, Amy S. and White, Robin R. and Easton, Zachary M.},
  title = {Calibration using downscaled and bias-corrected satellite soil-moisture data can improve watershed model representation of soil-moisture variability},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-5999-2026},
  url = {https://doi.org/10.5194/hess-30-5999-2026}
}