Flood Detection and Susceptibility Assessment using Deep Learning Methods: A Case Study of Al-Kut, Wasit Governorate, Iraq

Flooding is a natural hazard that affects the environment, agricultural services, and human settlements. In Iraq, floods are becoming more intense as a result of uncontrolled urban expansion and negative human activities. This research developed deep learning methods for flood detection based on Sentinel-2 satellite imagery for a flood event that happened on November 26, 2018 in Al-Kut city, Wasit province, Iraq. A dataset was prepared based on a flood reference map given by SERTIT for the research area and a total of 144 samples, both negative and positive. Flooding extends in the study area were detected using Normalized Difference Water Index (NDWI) and Convolutional Neural Network (CNN). The results were compared with Support Vector Machine (SVM) and Random Forest (RF). Finally, the overall accuracy (OA), intersection over union (IoU), and F1-score of flood detection models were evaluated. The results showed that NDWI performed poorly, with OA, IoU, and F1-scores of 0.52, 0.48, and 0.51, respectively. CNN outperformed the SVM and RF models. According to OA (0.989), IoU (0.979), and F1-score (0.989), the map was the most accurate.

Flood Detection and Susceptibility Assessment using Deep Learning Methods: A Case Study of Al-Kut, Wasit Governorate, Iraq Keywords:

Flood detection , Deep learning , Sentinel , Wasit province