Visual and semi-automated interpretation methods for urban flood detection using SAR Sentinel-1A image: study case in North Aceh Regency, Indonesia

Abstract

Flood is a natural disaster when causes damage and loss, especially in urban areas that are inhabited by many people and have many properties. The goal of this study is to map rapidly the flooded areas in North Aceh district on January, 5 2022 using multi-temporal SAR Sentinel-1A images through visual interpretation and semi-automated interpretation methods including thresholding and support vector machine (SVM). Visual interpretation relied on multi-polarization image of VH-VV with median filters of 3x3 and 5x5. This study produced the flooded area maps from three methods used with an area of each method: visual interpretation (422.86 ha), thresholding (791.22 ha) and SVM (1084.76 ha). According to the comparison with the result generated from visual interpretation using Planetscope-3A, SVM method has the similar result in the flooded areas with an area of 1001.52 ha. This study also analysed impact of flood event against land cover referring to Sentinel-2 10-Meter Land Use/Land Cover of ESRI and showed that class of cultivated crops is most affected. Results show that this study can be very helpful in damage and loss assessment of flood event and can be reference for urban planning management in order to face the climate change.