Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study

Abstract Differentiable geoscientific modeling has shown promise for leveraging machine learning (ML) to unify physically based and data‐based modeling. Here, we critically analyze this promise in the context of large‐scale parameter optimization with the Noah‐MP land model as an example. The differ...

Full description

Saved in:
Bibliographic Details
Main Authors: Shahryar K. Ahmad, Sujay V. Kumar, Clara Draper, Rolf H. Reichle
Format: Article
Language:English
Published: Wiley 2025-02-01
Series:Geophysical Research Letters
Subjects:
Online Access:https://doi.org/10.1029/2024GL112893
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850136281470730240
author Shahryar K. Ahmad
Sujay V. Kumar
Clara Draper
Rolf H. Reichle
author_facet Shahryar K. Ahmad
Sujay V. Kumar
Clara Draper
Rolf H. Reichle
author_sort Shahryar K. Ahmad
collection DOAJ
description Abstract Differentiable geoscientific modeling has shown promise for leveraging machine learning (ML) to unify physically based and data‐based modeling. Here, we critically analyze this promise in the context of large‐scale parameter optimization with the Noah‐MP land model as an example. The differentiable parameter learning framework is used to calibrate Noah‐MP soil and vegetation parameters such that the simulated surface soil moisture better matches satellite observations over the contiguous US. We found that the optimized parameters only marginally improved soil moisture (average RMSE = 0.092 m3 m−3) upon uncalibrated Noah‐MP (RMSE = 0.10 m3 m−3). Scaling and bias correction factors, often used in ML approaches for enhancing generalizability, were found to limit the transferability of the optimized physical parameters to the land model. The global objective function further compromises the algorithm's ability to simultaneously capture contrasting moisture regimes. Addressing these challenges is necessary to advance ML‐based calibration frameworks to better learn and represent the constraints of the physical model.
format Article
id doaj-art-cc8a87a5e6fc4d90a36eaddfa381c7a4
institution OA Journals
issn 0094-8276
1944-8007
language English
publishDate 2025-02-01
publisher Wiley
record_format Article
series Geophysical Research Letters
spelling doaj-art-cc8a87a5e6fc4d90a36eaddfa381c7a42025-08-20T02:31:10ZengWileyGeophysical Research Letters0094-82761944-80072025-02-01524n/an/a10.1029/2024GL112893Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case StudyShahryar K. Ahmad0Sujay V. Kumar1Clara Draper2Rolf H. Reichle3Hydrological Sciences Lab NASA Goddard Space Flight Center (GSFC) Greenbelt MD USAHydrological Sciences Lab NASA Goddard Space Flight Center (GSFC) Greenbelt MD USANOAA Physical Sciences Laboratory (PSL) Boulder CO USAGlobal Modeling and Assimilation Office (GMAO) NASA Goddard Space Flight Center (GSFC) Greenbelt MD USAAbstract Differentiable geoscientific modeling has shown promise for leveraging machine learning (ML) to unify physically based and data‐based modeling. Here, we critically analyze this promise in the context of large‐scale parameter optimization with the Noah‐MP land model as an example. The differentiable parameter learning framework is used to calibrate Noah‐MP soil and vegetation parameters such that the simulated surface soil moisture better matches satellite observations over the contiguous US. We found that the optimized parameters only marginally improved soil moisture (average RMSE = 0.092 m3 m−3) upon uncalibrated Noah‐MP (RMSE = 0.10 m3 m−3). Scaling and bias correction factors, often used in ML approaches for enhancing generalizability, were found to limit the transferability of the optimized physical parameters to the land model. The global objective function further compromises the algorithm's ability to simultaneously capture contrasting moisture regimes. Addressing these challenges is necessary to advance ML‐based calibration frameworks to better learn and represent the constraints of the physical model.https://doi.org/10.1029/2024GL112893machine learningland surface modelcalibrationsoil moisturegenetic algorithmtransferability
spellingShingle Shahryar K. Ahmad
Sujay V. Kumar
Clara Draper
Rolf H. Reichle
Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study
Geophysical Research Letters
machine learning
land surface model
calibration
soil moisture
genetic algorithm
transferability
title Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study
title_full Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study
title_fullStr Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study
title_full_unstemmed Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study
title_short Challenges in Unifying Physically Based and Machine Learning Simulations Through Differentiable Modeling: A Land Surface Case Study
title_sort challenges in unifying physically based and machine learning simulations through differentiable modeling a land surface case study
topic machine learning
land surface model
calibration
soil moisture
genetic algorithm
transferability
url https://doi.org/10.1029/2024GL112893
work_keys_str_mv AT shahryarkahmad challengesinunifyingphysicallybasedandmachinelearningsimulationsthroughdifferentiablemodelingalandsurfacecasestudy
AT sujayvkumar challengesinunifyingphysicallybasedandmachinelearningsimulationsthroughdifferentiablemodelingalandsurfacecasestudy
AT claradraper challengesinunifyingphysicallybasedandmachinelearningsimulationsthroughdifferentiablemodelingalandsurfacecasestudy
AT rolfhreichle challengesinunifyingphysicallybasedandmachinelearningsimulationsthroughdifferentiablemodelingalandsurfacecasestudy