Monitoring water quality parameters using multi-source data-driven machine learning models

Systematic monitoring of water quality parameters in aquatic environments was a critical task for environmental protection and water resource management. Remote sensing technology, as an effective monitoring tool, provided real-time water quality data. Currently, most research primarily relied on re...

Full description

Saved in:
Bibliographic Details
Main Authors: Yubo Zhao, Mo Chen, Jinyu He, Yanping Ma
Format: Article
Language:English
Published: Taylor & Francis Group 2025-12-01
Series:Engineering Applications of Computational Fluid Mechanics
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/19942060.2025.2509658
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850238775186161664
author Yubo Zhao
Mo Chen
Jinyu He
Yanping Ma
author_facet Yubo Zhao
Mo Chen
Jinyu He
Yanping Ma
author_sort Yubo Zhao
collection DOAJ
description Systematic monitoring of water quality parameters in aquatic environments was a critical task for environmental protection and water resource management. Remote sensing technology, as an effective monitoring tool, provided real-time water quality data. Currently, most research primarily relied on reflectance analysis of remote sensing data, often overlooking the impact of environmental factors on aquatic environments. This study integrated field data, multispectral imagery, meteorological data, and hydrological data to invert the water quality conditions of aquatic environments. A machine learning model (CNN-RF) was developed to estimate three water quality parameters (TP, DO, COD), and its performance was comprehensively evaluated using four error indices (R², MAE, MSE, MAPE). The results indicated that, compared to models using only spectral reflectance as features, the inclusion of environmental factors significantly enhanced the accuracy of the inversion of the three water quality parameters. For the watershed, the R² values for TP, DO, and COD increased by 0.34, 0.43, and 0.42, respectively. The MAE was less than 0.85, the RMSE was less than 1, and the MAEP was less than 15.65%. For the lake, the R² values for TP across four months were 0.75, 0.66, 0.67, and 0.73; for DO, they were 0.73, 0.79, 0.73, and 0.64; and for COD, they were 0.73, 0.62, 0.66, and 0.61, all exceeding 0.6. The MAE was less than 0.85, the RMSE was less than 0.95, and the MAEP was less than 12.6%. Both the watershed and lake models met the inversion requirements. Additionally, the SHAP method was employed to quantitatively analyze the contribution of environmental factors and spectral bands to the model inversion. Therefore, the findings confirmed the importance of environmental variables in water quality inversion and provided a theoretical basis for the optimization and application of future water quality monitoring systems.
format Article
id doaj-art-c50ae24d8b554f03a01c6c56a0436eb4
institution OA Journals
issn 1994-2060
1997-003X
language English
publishDate 2025-12-01
publisher Taylor & Francis Group
record_format Article
series Engineering Applications of Computational Fluid Mechanics
spelling doaj-art-c50ae24d8b554f03a01c6c56a0436eb42025-08-20T02:01:23ZengTaylor & Francis GroupEngineering Applications of Computational Fluid Mechanics1994-20601997-003X2025-12-0119110.1080/19942060.2025.2509658Monitoring water quality parameters using multi-source data-driven machine learning modelsYubo Zhao0Mo Chen1Jinyu He2Yanping Ma3College of Heilongjiang rive and lake chief, Heilongjiang University, Harbin, People’s Republic of ChinaSchool of Hydraulic and Electric-power, Heilongjiang University, Harbin, People’s Republic of ChinaCollege of Heilongjiang rive and lake chief, Heilongjiang University, Harbin, People’s Republic of ChinaCollege of Heilongjiang rive and lake chief, Heilongjiang University, Harbin, People’s Republic of ChinaSystematic monitoring of water quality parameters in aquatic environments was a critical task for environmental protection and water resource management. Remote sensing technology, as an effective monitoring tool, provided real-time water quality data. Currently, most research primarily relied on reflectance analysis of remote sensing data, often overlooking the impact of environmental factors on aquatic environments. This study integrated field data, multispectral imagery, meteorological data, and hydrological data to invert the water quality conditions of aquatic environments. A machine learning model (CNN-RF) was developed to estimate three water quality parameters (TP, DO, COD), and its performance was comprehensively evaluated using four error indices (R², MAE, MSE, MAPE). The results indicated that, compared to models using only spectral reflectance as features, the inclusion of environmental factors significantly enhanced the accuracy of the inversion of the three water quality parameters. For the watershed, the R² values for TP, DO, and COD increased by 0.34, 0.43, and 0.42, respectively. The MAE was less than 0.85, the RMSE was less than 1, and the MAEP was less than 15.65%. For the lake, the R² values for TP across four months were 0.75, 0.66, 0.67, and 0.73; for DO, they were 0.73, 0.79, 0.73, and 0.64; and for COD, they were 0.73, 0.62, 0.66, and 0.61, all exceeding 0.6. The MAE was less than 0.85, the RMSE was less than 0.95, and the MAEP was less than 12.6%. Both the watershed and lake models met the inversion requirements. Additionally, the SHAP method was employed to quantitatively analyze the contribution of environmental factors and spectral bands to the model inversion. Therefore, the findings confirmed the importance of environmental variables in water quality inversion and provided a theoretical basis for the optimization and application of future water quality monitoring systems.https://www.tandfonline.com/doi/10.1080/19942060.2025.2509658Remote sensingenvironmental factorsmachine learningaquatic environmentwater quality monitoring
spellingShingle Yubo Zhao
Mo Chen
Jinyu He
Yanping Ma
Monitoring water quality parameters using multi-source data-driven machine learning models
Engineering Applications of Computational Fluid Mechanics
Remote sensing
environmental factors
machine learning
aquatic environment
water quality monitoring
title Monitoring water quality parameters using multi-source data-driven machine learning models
title_full Monitoring water quality parameters using multi-source data-driven machine learning models
title_fullStr Monitoring water quality parameters using multi-source data-driven machine learning models
title_full_unstemmed Monitoring water quality parameters using multi-source data-driven machine learning models
title_short Monitoring water quality parameters using multi-source data-driven machine learning models
title_sort monitoring water quality parameters using multi source data driven machine learning models
topic Remote sensing
environmental factors
machine learning
aquatic environment
water quality monitoring
url https://www.tandfonline.com/doi/10.1080/19942060.2025.2509658
work_keys_str_mv AT yubozhao monitoringwaterqualityparametersusingmultisourcedatadrivenmachinelearningmodels
AT mochen monitoringwaterqualityparametersusingmultisourcedatadrivenmachinelearningmodels
AT jinyuhe monitoringwaterqualityparametersusingmultisourcedatadrivenmachinelearningmodels
AT yanpingma monitoringwaterqualityparametersusingmultisourcedatadrivenmachinelearningmodels