Answers are grounded in this paper and cite their sources.
A workflow diagram illustrating flood susceptibility mapping in Adama city, showing the processing of flood conditioning factors, model training via twenty ensemble machine learning scenarios, and final susceptibility prediction maps. Flooding is a natural disaster that impacts socioeconomic development in Ethiopia. Advancements in Earth observation and machine learning (ML) have improved flood susceptibility mapping. However, there is no universally agreed model for better flood prediction. The objective of this study was to develop multiple ensemble ML models for flood susceptibility mapping. Factor maps, namely, elevation, topographic wetness index, drainage density, aspect, slope, stream power index, flow accumulation, soil type, land use/land cover, Sentinel 1 (vertical transmit-vertical receive (VV) and very high backscatter), rainfall, normalized difference water index, and normalized difference vegetation index were derived from satellite imageries and ground truth datasets for model development, training, and validation. Twenty ensemble scenario models were developed from random forest (RF), gradient boosting (GB), support vector machine, long short-term memory, and convolutional neural network (CNN) ML models. The performances of the models were evaluated using statistical metrics, namely, accuracy, precision, and ROC. The results showed that 15 of 20 ensemble models predicted flooding with accuracy >90%, where RF + GB and RF + CNN show superior performance with 97 and 95% accuracy, respectively. The result further showed that RF + GB and RF + CNN predicted approximately 38.82–39.82% of Adama city as highly to very highly susceptible to flooding, respectively. The findings of this study can be used to develop efficient flood risk mitigation plans for Adama and other similar cities.
No takes yet. Share an insight, caveat, or question.
Tiky et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: