Systematic Discrepancies Should Be Addressed Before Effective Parameter Learning: Insights From Soil Moisture Simulations Using Differentiable Process‐Based Models
- 1Civil and Environmental Engineering The Pennsylvania State University University Park PA USA
- 2Center for Western Weather and Water Extremes Scripps Institution of Oceanography University of California San Diego La Jolla CA USA
Abstract
Differentiable hydrologic models combining neural networks with process‐based backbones—fully implemented or approximated by surrogate models—offer highly competitive accuracy and improved interpretability compared to either model type used alone. However, the impacts of surrogate model fidelity (accurate sensitivity of relationships compared to the original model) and systematic discrepancies between models and observations on parameter learning effectiveness remain underexplored. This study evaluates optimization performance of the differentiable HBV model (δHBV) and its surrogate variants in simulating soil moisture across the contiguous United States. Results show that high‐fidelity surrogate models can support effective differentiable parameter learning once systematic discrepancies are addressed. In contrast, unresolved systematic discrepancies degrade optimization performance, with particularly severe deterioration when low‐fidelity surrogates are used. These findings provide one possible explanation for the modest improvements reported in recent surrogate‐based differentiable parameter learning (dPL) applications. Linear correction partially mitigates these systematic discrepancies, improves optimization performance, and reduces compensatory parameter adjustment, although it does not resolve the underlying model–observation mismatch. Structural refinement remains a physically grounded long‐term direction, while pragmatic correction modules can provide useful partial mitigation when such refinement is not yet available. In addition, fully implementing differentiable process‐based models or using a rigorously trained surrogate can improve gradient fidelity and reduce amplified deterioration associated with surrogate infidelity.
Related
- Advancing global crop simulations with differentiable modelling
- A cost-effective data assimilation system for improving ROMS simulations in the Southwest Atlantic
- Transfer learning reveals large discrepancies between air and land surface temperatures in cities
- Discrepancies between Sentinel-2 and LiDAR-derived LAI in temperate deciduous forests: Influence of apparent optical depth, radiative transfer priors, and horizontal heterogeneity
- PCOD-NN: a differentiable, instrument-independent emulator of monochromatic optical depth for infrared radiative transfer
BibTeX
@article{li2026systematic,
title = {{Systematic Discrepancies Should Be Addressed Before Effective Parameter Learning: Insights From Soil Moisture Simulations Using Differentiable Process‐Based Models}},
author = {Peijun Li and Chaopeng Shen and Jiangtao Liu and Farshid Rahmani and Ming Pan and Kathryn Lawson},
journal = {Water Resources Research},
year = {2026},
month = nov,
doi = {10.1029/2025wr042382},
url = {https://doi.org/10.1029/2025wr042382},
note = {CC BY},
}
RIS
TY - JOUR
TI - Systematic Discrepancies Should Be Addressed Before Effective Parameter Learning: Insights From Soil Moisture Simulations Using Differentiable Process‐Based Models
AU - Peijun Li
AU - Chaopeng Shen
AU - Jiangtao Liu
AU - Farshid Rahmani
AU - Ming Pan
AU - Kathryn Lawson
PY - 2026
DA - 2026/11/11
JO - Water Resources Research
DO - 10.1029/2025wr042382
UR - https://doi.org/10.1029/2025wr042382
AB - Differentiable hydrologic models combining neural networks with process‐based backbones—fully implemented or approximated by surrogate models—offer highly competitive accuracy and improved interpretability compared to either model type used alone. However, the impacts of surrogate model fidelity (accurate sensitivity of relationships compared to the original model) and systematic discrepancies between models and observations on parameter learning effectiveness remain underexplored. This study evaluates optimization performance of the differentiable HBV model (δHBV) and its surrogate variants in simulating soil moisture across the contiguous United States. Results show that high‐fidelity surrogate models can support effective differentiable parameter learning once systematic discrepancies are addressed. In contrast, unresolved systematic discrepancies degrade optimization performance, with particularly severe deterioration when low‐fidelity surrogates are used. These findings provide one possible explanation for the modest improvements reported in recent surrogate‐based differentiable parameter learning (dPL) applications. Linear correction partially mitigates these systematic discrepancies, improves optimization performance, and reduces compensatory parameter adjustment, although it does not resolve the underlying model–observation mismatch. Structural refinement remains a physically grounded long‐term direction, while pragmatic correction modules can provide useful partial mitigation when such refinement is not yet available. In addition, fully implementing differentiable process‐based models or using a rigorously trained surrogate can improve gradient fidelity and reduce amplified deterioration associated with surrogate infidelity.
SN - 1944-7973
SN - 0043-1397
ER -