Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning

Small reservoirs are increasingly common across Africa. They provide decentralised access to water and support farmer-led irrigation, in addition to contributing towards mitigating the impacts of climate change. Water quality monitoring is essential to ensure the safe use of water and to understand...

Descripción completa

Detalles Bibliográficos
Autores principales: Steinbach, S., Bartels, A., Rienow, A., Kuria, B. T., Zwart, Sander J., Nelson, A.
Formato: Journal Article
Lenguaje:Inglés
Publicado: Elsevier 2025
Materias:
Acceso en línea:https://hdl.handle.net/10568/173072
_version_ 1855528243365412864
author Steinbach, S.
Bartels, A.
Rienow, A.
Kuria, B. T.
Zwart, Sander J.
Nelson, A.
author_browse Bartels, A.
Kuria, B. T.
Nelson, A.
Rienow, A.
Steinbach, S.
Zwart, Sander J.
author_facet Steinbach, S.
Bartels, A.
Rienow, A.
Kuria, B. T.
Zwart, Sander J.
Nelson, A.
author_sort Steinbach, S.
collection Repository of Agricultural Research Outputs (CGSpace)
description Small reservoirs are increasingly common across Africa. They provide decentralised access to water and support farmer-led irrigation, in addition to contributing towards mitigating the impacts of climate change. Water quality monitoring is essential to ensure the safe use of water and to understand the impact of the environment and land use on water quality. However, water quality in small reservoirs is often not monitored continuously, with the interlinkages between weather, land, and water remaining unknown. Turbidity is a prime indicator of water quality that can be assessed with remote sensing techniques. Here we modelled turbidity in 34 small reservoirs in central Kenya with Sentinel-2 data from 2017 to 2023 and predicted turbidity outcomes using primary and secondary Earth observation data, and machine learning. We found distinct monthly turbidity patterns. Random forest and gradient boosting models showed that annual turbidity outcomes depend on meteorological variables, topography, and land cover (R2 = 0.46 and 0.43 respectively), while longer-term turbidity was influenced more strongly by land management and land cover (R2 = 0.88 and 0.72 respectively). Our results suggest that shortand longer-term turbidity prediction can inform reservoir siting and management. However, inter-annual variability prediction could benefit from more knowledge of additional factors that may not be fully captured in commonly available geospatial data. This study contributes to the relatively small body of remote sensing-based research on water quality in small reservoirs and supports improved small-scale water management.
format Journal Article
id CGSpace173072
institution CGIAR Consortium
language Inglés
publishDate 2025
publishDateRange 2025
publishDateSort 2025
publisher Elsevier
publisherStr Elsevier
record_format dspace
spelling CGSpace1730722025-10-26T13:00:44Z Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning Steinbach, S. Bartels, A. Rienow, A. Kuria, B. T. Zwart, Sander J. Nelson, A. turbidity prediction water reservoirs remote sensing machine learning modelling water quality agricultural water management satellite observation Small reservoirs are increasingly common across Africa. They provide decentralised access to water and support farmer-led irrigation, in addition to contributing towards mitigating the impacts of climate change. Water quality monitoring is essential to ensure the safe use of water and to understand the impact of the environment and land use on water quality. However, water quality in small reservoirs is often not monitored continuously, with the interlinkages between weather, land, and water remaining unknown. Turbidity is a prime indicator of water quality that can be assessed with remote sensing techniques. Here we modelled turbidity in 34 small reservoirs in central Kenya with Sentinel-2 data from 2017 to 2023 and predicted turbidity outcomes using primary and secondary Earth observation data, and machine learning. We found distinct monthly turbidity patterns. Random forest and gradient boosting models showed that annual turbidity outcomes depend on meteorological variables, topography, and land cover (R2 = 0.46 and 0.43 respectively), while longer-term turbidity was influenced more strongly by land management and land cover (R2 = 0.88 and 0.72 respectively). Our results suggest that shortand longer-term turbidity prediction can inform reservoir siting and management. However, inter-annual variability prediction could benefit from more knowledge of additional factors that may not be fully captured in commonly available geospatial data. This study contributes to the relatively small body of remote sensing-based research on water quality in small reservoirs and supports improved small-scale water management. 2025-02 2025-02-16T10:10:28Z 2025-02-16T10:10:28Z Journal Article https://hdl.handle.net/10568/173072 en Open Access Elsevier Steinbach, S.; Bartels, A.; Rienow, A.; Kuria, B. T.; Zwart, Sander Jaap; Nelson, A. 2025. Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning. International Journal of Applied Earth Observation and Geoinformation, 136:104390. [doi: https://doi.org/10.1016/j.jag.2025.104390]
spellingShingle turbidity
prediction
water reservoirs
remote sensing
machine learning
modelling
water quality
agricultural water management
satellite observation
Steinbach, S.
Bartels, A.
Rienow, A.
Kuria, B. T.
Zwart, Sander J.
Nelson, A.
Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning
title Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning
title_full Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning
title_fullStr Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning
title_full_unstemmed Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning
title_short Predicting turbidity dynamics in small reservoirs in Central Kenya using remote sensing and machine learning
title_sort predicting turbidity dynamics in small reservoirs in central kenya using remote sensing and machine learning
topic turbidity
prediction
water reservoirs
remote sensing
machine learning
modelling
water quality
agricultural water management
satellite observation
url https://hdl.handle.net/10568/173072
work_keys_str_mv AT steinbachs predictingturbiditydynamicsinsmallreservoirsincentralkenyausingremotesensingandmachinelearning
AT bartelsa predictingturbiditydynamicsinsmallreservoirsincentralkenyausingremotesensingandmachinelearning
AT rienowa predictingturbiditydynamicsinsmallreservoirsincentralkenyausingremotesensingandmachinelearning
AT kuriabt predictingturbiditydynamicsinsmallreservoirsincentralkenyausingremotesensingandmachinelearning
AT zwartsanderj predictingturbiditydynamicsinsmallreservoirsincentralkenyausingremotesensingandmachinelearning
AT nelsona predictingturbiditydynamicsinsmallreservoirsincentralkenyausingremotesensingandmachinelearning