Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region

Abstract Estimating spatially distributed hydrological parameters in ungauged catchments poses a challenging regionalization problem and requires imposing spatial constraints given the sparsity of discharge data. A possible approach is to search for a transfer function that quantitatively relates ph...

Full description

Saved in:
Bibliographic Details
Main Authors: Ngo Nghi Truyen Huynh, Pierre‐André Garambois, François Colleoni, Benjamin Renard, Hélène Roux, Julie Demargne, Maxime Jay‐Allemand, Pierre Javelle
Format: Article
Language:English
Published: Wiley 2024-11-01
Series:Water Resources Research
Subjects:
Online Access:https://doi.org/10.1029/2024WR037544
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849227331373105152
author Ngo Nghi Truyen Huynh
Pierre‐André Garambois
François Colleoni
Benjamin Renard
Hélène Roux
Julie Demargne
Maxime Jay‐Allemand
Pierre Javelle
author_facet Ngo Nghi Truyen Huynh
Pierre‐André Garambois
François Colleoni
Benjamin Renard
Hélène Roux
Julie Demargne
Maxime Jay‐Allemand
Pierre Javelle
author_sort Ngo Nghi Truyen Huynh
collection DOAJ
description Abstract Estimating spatially distributed hydrological parameters in ungauged catchments poses a challenging regionalization problem and requires imposing spatial constraints given the sparsity of discharge data. A possible approach is to search for a transfer function that quantitatively relates physical descriptors to conceptual model parameters. This paper introduces a Hybrid Data Assimilation and Parameter Regionalization (HDA‐PR) approach incorporating learnable regionalization mappings, based on either multi‐linear regression or artificial neural networks (ANNs), into a differentiable hydrological model. This approach demonstrates how two differentiable codes can be linked and their gradients chained, enabling the exploitation of heterogeneous data sets across extensive spatio‐temporal computational domains within a high‐dimensional regionalization context, using accurate adjoint‐based gradients. The inverse problem is tackled with a multi‐gauge calibration cost function accounting for information from multiple observation sites. HDA‐PR was tested on high‐resolution, hourly and kilometric regional modeling of 126 flash‐flood‐prone catchments in the French Mediterranean region. The results highlight a strong regionalization performance of HDA‐PR especially in the most challenging upstream‐to‐downstream extrapolation scenario with ANN, achieving median Nash‐Sutcliffe efficiency (NSE) scores from 0.6 to 0.71 for spatial, temporal, spatio‐temporal validations, and improving NSE by up to 30% on average compared to the baseline model calibrated with lumped parameters. Multiple evaluation metrics based on flood‐oriented hydrological signatures also indicate that the use of an ANN leads to better performances than a multi‐linear regression in a validation context. ANN enables to learn a non‐linear descriptors‐to‐parameters mapping which provides better model controllability than a linear mapping for complex calibration cases.
format Article
id doaj-art-f0a4db8eec6d4dcb992fe15e2271cb47
institution Kabale University
issn 0043-1397
1944-7973
language English
publishDate 2024-11-01
publisher Wiley
record_format Article
series Water Resources Research
spelling doaj-art-f0a4db8eec6d4dcb992fe15e2271cb472025-08-23T13:05:51ZengWileyWater Resources Research0043-13971944-79732024-11-016011n/an/a10.1029/2024WR037544Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean RegionNgo Nghi Truyen Huynh0Pierre‐André Garambois1François Colleoni2Benjamin Renard3Hélène Roux4Julie Demargne5Maxime Jay‐Allemand6Pierre Javelle7INRAE RECOVER Aix‐Marseille Université Aix‐en‐Provence FranceINRAE RECOVER Aix‐Marseille Université Aix‐en‐Provence FranceINRAE RECOVER Aix‐Marseille Université Aix‐en‐Provence FranceINRAE RECOVER Aix‐Marseille Université Aix‐en‐Provence FranceInstitut de Mécanique des Fluides de Toulouse (IMFT) CNRS Université de Toulouse Toulouse FranceHYDRIS Hydrologie Parc Scientifique Agropolis II Montferrier‐sur‐Lez FranceHYDRIS Hydrologie Parc Scientifique Agropolis II Montferrier‐sur‐Lez FranceINRAE RECOVER Aix‐Marseille Université Aix‐en‐Provence FranceAbstract Estimating spatially distributed hydrological parameters in ungauged catchments poses a challenging regionalization problem and requires imposing spatial constraints given the sparsity of discharge data. A possible approach is to search for a transfer function that quantitatively relates physical descriptors to conceptual model parameters. This paper introduces a Hybrid Data Assimilation and Parameter Regionalization (HDA‐PR) approach incorporating learnable regionalization mappings, based on either multi‐linear regression or artificial neural networks (ANNs), into a differentiable hydrological model. This approach demonstrates how two differentiable codes can be linked and their gradients chained, enabling the exploitation of heterogeneous data sets across extensive spatio‐temporal computational domains within a high‐dimensional regionalization context, using accurate adjoint‐based gradients. The inverse problem is tackled with a multi‐gauge calibration cost function accounting for information from multiple observation sites. HDA‐PR was tested on high‐resolution, hourly and kilometric regional modeling of 126 flash‐flood‐prone catchments in the French Mediterranean region. The results highlight a strong regionalization performance of HDA‐PR especially in the most challenging upstream‐to‐downstream extrapolation scenario with ANN, achieving median Nash‐Sutcliffe efficiency (NSE) scores from 0.6 to 0.71 for spatial, temporal, spatio‐temporal validations, and improving NSE by up to 30% on average compared to the baseline model calibrated with lumped parameters. Multiple evaluation metrics based on flood‐oriented hydrological signatures also indicate that the use of an ANN leads to better performances than a multi‐linear regression in a validation context. ANN enables to learn a non‐linear descriptors‐to‐parameters mapping which provides better model controllability than a linear mapping for complex calibration cases.https://doi.org/10.1029/2024WR037544parameter regionalizationhybrid AIdistributed hydrological modelingflood forecastingdifferentiable modelsvariational data assimilation
spellingShingle Ngo Nghi Truyen Huynh
Pierre‐André Garambois
François Colleoni
Benjamin Renard
Hélène Roux
Julie Demargne
Maxime Jay‐Allemand
Pierre Javelle
Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region
Water Resources Research
parameter regionalization
hybrid AI
distributed hydrological modeling
flood forecasting
differentiable models
variational data assimilation
title Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region
title_full Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region
title_fullStr Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region
title_full_unstemmed Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region
title_short Learning Regionalization Using Accurate Spatial Cost Gradients Within a Differentiable High‐Resolution Hydrological Model: Application to the French Mediterranean Region
title_sort learning regionalization using accurate spatial cost gradients within a differentiable high resolution hydrological model application to the french mediterranean region
topic parameter regionalization
hybrid AI
distributed hydrological modeling
flood forecasting
differentiable models
variational data assimilation
url https://doi.org/10.1029/2024WR037544
work_keys_str_mv AT ngonghitruyenhuynh learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT pierreandregarambois learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT francoiscolleoni learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT benjaminrenard learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT heleneroux learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT juliedemargne learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT maximejayallemand learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion
AT pierrejavelle learningregionalizationusingaccuratespatialcostgradientswithinadifferentiablehighresolutionhydrologicalmodelapplicationtothefrenchmediterraneanregion