• search hit 1 of 91
Back to Result List

High-resolution monitoring of intra-seasonal agricultural drought using sentinel-2 and machine learning across bimodal growing seasons in Kenya

  • Drought presents significant challenges to agriculture, threatening food security and livelihoods, across many regions. In Kenya, recurrent droughts across diverse agro-ecological zones emphasize the urgent need for reliable and scalable drought assessment methods. Although drought assessment with various datasets has been carried out for this region, many of them often use course or moderate resolution data. This study uses high-resolution Sentinel-2 observations and machine learning to monitor intra-seasonal crop conditions and assess drought impacts across bimodal growing seasons. Using pixel-based supervised random forest models trained with multiple vegetation indices as input, we classify croplands into drought-affected and unaffected areas. The phenology information is integrated into the framework to take into account two growing seasons. Our results from 2016 to 2022 highlight an increasing frequency in drought-affected croplands, with 2022 exhibiting widespread crop stress,Drought presents significant challenges to agriculture, threatening food security and livelihoods, across many regions. In Kenya, recurrent droughts across diverse agro-ecological zones emphasize the urgent need for reliable and scalable drought assessment methods. Although drought assessment with various datasets has been carried out for this region, many of them often use course or moderate resolution data. This study uses high-resolution Sentinel-2 observations and machine learning to monitor intra-seasonal crop conditions and assess drought impacts across bimodal growing seasons. Using pixel-based supervised random forest models trained with multiple vegetation indices as input, we classify croplands into drought-affected and unaffected areas. The phenology information is integrated into the framework to take into account two growing seasons. Our results from 2016 to 2022 highlight an increasing frequency in drought-affected croplands, with 2022 exhibiting widespread crop stress, especially during the long rains season. The results have strong agreement to governmental yield statistics and national and international reports as well as regional and global drought monitoring platforms. Despite uncertainties from cloud cover, land cover misclassification, and other stressors, the proposed framework provides high-resolution and operationally relevant agricultural drought monitoring approach with high resolution at national scale. Future enhancements, including radar-based data integration and real-time monitoring and prediction, can further strengthen early warning systems, supporting policymakers in developing targeted drought mitigation strategies.show moreshow less

Download full text files

Export metadata

Statistics

Number of document requests

Additional Services

Share in Twitter Search Google Scholar
Metadaten
Author:S. Mohammad Mirmazloumi, Harison KipkuleiORCiDGND, Rose Waswa, Tobias Landmann, Tom Dienya, Maximilian Schwarz, Fabrizio Ramoino, Clément Albergel, Gohar Ghazaryan
URN:urn:nbn:de:bvb:384-opus4-1294618
Frontdoor URLhttps://opus.bibliothek.uni-augsburg.de/opus4/129461
ISSN:1470-160XOPAC
Parent Title (English):Ecological Indicators
Publisher:Elsevier BV
Place of publication:Amsterdam
Type:Article
Language:English
Year of first Publication:2026
Publishing Institution:Universität Augsburg
Release Date:2026/04/02
Volume:185
First Page:114790
DOI:https://doi.org/10.1016/j.ecolind.2026.114790
Institutes:Fakultät für Angewandte Informatik
Fakultätsübergreifende Institute und Einrichtungen
Fakultät für Angewandte Informatik / Institut für Geographie
Fakultätsübergreifende Institute und Einrichtungen / Zentrum für Klimaresilienz
Fakultät für Angewandte Informatik / Institut für Geographie / Lehrstuhl für Urbane Klimaresilienz
Dewey Decimal Classification:5 Naturwissenschaften und Mathematik / 55 Geowissenschaften, Geologie / 550 Geowissenschaften
Licence (German):CC-BY 4.0: Creative Commons: Namensnennung