Untangling hybrid hydrological models with explainable artificial intelligence

Hydrological models are valuable tools for developing streamflow predictions in unmonitored catchments to increase our understanding of hydrological processes. A recent effort has been made in the development of hybrid (conceptual/machine learning) models that can preserve some of the hydrological p...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Daniel Althoff, Helizani Couto Bazame, Jessica Garcia Nascimento
Formato: article
Lenguaje:EN
Publicado: IWA Publishing 2021
Materias:
Acceso en línea:https://doaj.org/article/1fa6d8abbab3463a85b7d216edde8b9f
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:1fa6d8abbab3463a85b7d216edde8b9f
record_format dspace
spelling oai:doaj.org-article:1fa6d8abbab3463a85b7d216edde8b9f2021-11-08T07:59:15ZUntangling hybrid hydrological models with explainable artificial intelligence2616-651810.2166/h2oj.2021.066https://doaj.org/article/1fa6d8abbab3463a85b7d216edde8b9f2021-01-01T00:00:00Zhttp://doi.org/10.2166/h2oj.2021.066https://doaj.org/toc/2616-6518Hydrological models are valuable tools for developing streamflow predictions in unmonitored catchments to increase our understanding of hydrological processes. A recent effort has been made in the development of hybrid (conceptual/machine learning) models that can preserve some of the hydrological processes represented by conceptual models and can improve streamflow predictions. However, these studies have not explored how the data-driven component of hybrid models resolved runoff routing. In this study, explainable artificial intelligence (XAI) techniques are used to turn a ‘black-box’ model into a ‘glass box’ model. The hybrid models reduced the root-mean-square error of the simulated streamflow values by approximately 27, 50, and 24% for stations 17120000, 27380000, and 33680000, respectively, relative to the traditional method. XAI techniques helped unveil the importance of accounting for soil moisture in hydrological models. Differing from purely data-driven hydrological models, the inclusion of the production storage in the proposed hybrid model, which is responsible for estimating the water balance, reduced the short- and long-term dependencies of input variables for streamflow prediction. In addition, soil moisture controlled water percolation, which was the main predictor of streamflow. This finding is because soil moisture controls the underlying mechanisms of groundwater flow into river streams. Highlights A machine learning model is coupled to the GR4J hydrological model.; The hybrid hydrological model consists of a single soil moisture accounting storage.; The performance improvement is significant under low-flow conditions.; Explainable artificial intelligence techniques are used for the global and local interpretation of the data-driven component of the model.;Daniel AlthoffHelizani Couto BazameJessica Garcia NascimentoIWA Publishingarticlegr4jindividual conditional explanationslimepartial dependence profilesregression treesRiver, lake, and water-supply engineering (General)TC401-506Water supply for domestic and industrial purposesTD201-500ENH2Open Journal, Vol 4, Iss 1, Pp 13-28 (2021)
institution DOAJ
collection DOAJ
language EN
topic gr4j
individual conditional explanations
lime
partial dependence profiles
regression trees
River, lake, and water-supply engineering (General)
TC401-506
Water supply for domestic and industrial purposes
TD201-500
spellingShingle gr4j
individual conditional explanations
lime
partial dependence profiles
regression trees
River, lake, and water-supply engineering (General)
TC401-506
Water supply for domestic and industrial purposes
TD201-500
Daniel Althoff
Helizani Couto Bazame
Jessica Garcia Nascimento
Untangling hybrid hydrological models with explainable artificial intelligence
description Hydrological models are valuable tools for developing streamflow predictions in unmonitored catchments to increase our understanding of hydrological processes. A recent effort has been made in the development of hybrid (conceptual/machine learning) models that can preserve some of the hydrological processes represented by conceptual models and can improve streamflow predictions. However, these studies have not explored how the data-driven component of hybrid models resolved runoff routing. In this study, explainable artificial intelligence (XAI) techniques are used to turn a ‘black-box’ model into a ‘glass box’ model. The hybrid models reduced the root-mean-square error of the simulated streamflow values by approximately 27, 50, and 24% for stations 17120000, 27380000, and 33680000, respectively, relative to the traditional method. XAI techniques helped unveil the importance of accounting for soil moisture in hydrological models. Differing from purely data-driven hydrological models, the inclusion of the production storage in the proposed hybrid model, which is responsible for estimating the water balance, reduced the short- and long-term dependencies of input variables for streamflow prediction. In addition, soil moisture controlled water percolation, which was the main predictor of streamflow. This finding is because soil moisture controls the underlying mechanisms of groundwater flow into river streams. Highlights A machine learning model is coupled to the GR4J hydrological model.; The hybrid hydrological model consists of a single soil moisture accounting storage.; The performance improvement is significant under low-flow conditions.; Explainable artificial intelligence techniques are used for the global and local interpretation of the data-driven component of the model.;
format article
author Daniel Althoff
Helizani Couto Bazame
Jessica Garcia Nascimento
author_facet Daniel Althoff
Helizani Couto Bazame
Jessica Garcia Nascimento
author_sort Daniel Althoff
title Untangling hybrid hydrological models with explainable artificial intelligence
title_short Untangling hybrid hydrological models with explainable artificial intelligence
title_full Untangling hybrid hydrological models with explainable artificial intelligence
title_fullStr Untangling hybrid hydrological models with explainable artificial intelligence
title_full_unstemmed Untangling hybrid hydrological models with explainable artificial intelligence
title_sort untangling hybrid hydrological models with explainable artificial intelligence
publisher IWA Publishing
publishDate 2021
url https://doaj.org/article/1fa6d8abbab3463a85b7d216edde8b9f
work_keys_str_mv AT danielalthoff untanglinghybridhydrologicalmodelswithexplainableartificialintelligence
AT helizanicoutobazame untanglinghybridhydrologicalmodelswithexplainableartificialintelligence
AT jessicagarcianascimento untanglinghybridhydrologicalmodelswithexplainableartificialintelligence
_version_ 1718442784500744192