Comparative Evaluation of Machine Learning Algorithms for Irrigated Area Mapping in the Upper Blue Nile Basin

Abstract Reliable mapping of irrigated agriculture is important for sustainable water resource management, allocation and agricultural productivity, particularly in data-scarce river basins such as Upper Blue Nile Basin. Despite the basin’s significant irrigation potential, accurate and spatially explicit irrigated area maps remain limited. This study evaluated and compared machine learning classifiers including Random Forest (RF), Support Vector Machine (SVM), Gradient Tree Boosting (GTB), and Bidirectional Long Short-Term Memory (BiLSTM) for mapping irrigated areas in the Upper Blue Nile Basin, Ethiopia during the December 2025 to April 2026 irrigation season using multi source remote sensing datasets. The classifiers used two complementary data groups: multi-source remote-sensing and ancillary predictor variables, including Sentinel-2 multispectral imagery, Sentinel-1 synthetic aperture radar backscatter, CHIRPS rainfall estimates, and NASADEM-derived topographic variables; and ground-reference samples of irrigated and rainfed fields for model training and validation. The model performance was evaluated using the performance metrics overall accuracy, kappa coefficient, F1-score, precision, recall, and AUC-ROC, supported by nested spatial cross-validation with five outer spatial folds. Among the four classifiers, GTB scored the highest overall performance with accuracy, Kappa coefficient, F1-score, with values of 0.96, 0.91, and 0.96, respectively. RF ranked second and scored the highest AUC-ROC value of 0.99, followed by BiLSTM and SVM. Analysis showed that February Sentinel-2 spectral indices were the most influential predictors and February was the most important acquisition month for GTB and RF. Multi model ensemble was used to produce a conservative consensus map. The consensus map classified 588,545 ha as irrigated agriculture in the Upper Blue Nile Basin.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-13
DOI
https://doi.org/10.21203/rs.3.rs-11524145/v1
Primary Topic
Remote Sensing in Agriculture
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Comparative Evaluation of Machine Learning Algorithms for Irrigated Area Mapping in the Upper Blue Nile Basin

Haileyesus Belay Lakew, Zeleke Agide Dejen, Abdurezak Arage Hamza, Sirak Tekleab Gebrekristos et al.
3 links
Research Square
Remote Sensing in Agriculture
preprint

Comparative Evaluation of Machine Learning Algorithms for Irrigated Area Mapping in the Upper Blue Nile Basin

Haileyesus Belay Lakew, Zeleke Agide Dejen, Abdurezak Arage Hamza, Sirak Tekleab Gebrekristos, Tewele Gebretsadkan Haile
preprint en

Abstract

Abstract Reliable mapping of irrigated agriculture is important for sustainable water resource management, allocation and agricultural productivity, particularly in data-scarce river basins such as Upper Blue Nile Basin. Despite the basin’s significant irrigation potential, accurate and spatially explicit irrigated area maps remain limited. This study evaluated and compared machine learning classifiers including Random Forest (RF), Support Vector Machine (SVM), Gradient Tree Boosting (GTB), and Bidirectional Long Short-Term Memory (BiLSTM) for mapping irrigated areas in the Upper Blue Nile Basin, Ethiopia during the December 2025 to April 2026 irrigation season using multi source remote sensing datasets. The classifiers used two complementary data groups: multi-source remote-sensing and ancillary predictor variables, including Sentinel-2 multispectral imagery, Sentinel-1 synthetic aperture radar backscatter, CHIRPS rainfall estimates, and NASADEM-derived topographic variables; and ground-reference samples of irrigated and rainfed fields for model training and validation. The model performance was evaluated using the performance metrics overall accuracy, kappa coefficient, F1-score, precision, recall, and AUC-ROC, supported by nested spatial cross-validation with five outer spatial folds. Among the four classifiers, GTB scored the highest overall performance with accuracy, Kappa coefficient, F1-score, with values of 0.96, 0.91, and 0.96, respectively. RF ranked second and scored the highest AUC-ROC value of 0.99, followed by BiLSTM and SVM. Analysis showed that February Sentinel-2 spectral indices were the most influential predictors and February was the most important acquisition month for GTB and RF. Multi model ensemble was used to produce a conservative consensus map. The consensus map classified 588,545 ha as irrigated agriculture in the Upper Blue Nile Basin.

Research Square
Adigrat University (ET), Addis Ababa University (ET)
European Commission
Zero hunger, Clean water and sanitation
Remote Sensing in Agriculture
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon