Identifying environmental variables in potential flood hazard areas using machine learning approach at Musi Banyuasin Regency, South Sumatra

Abstract

Abstract Meteorological natural disasters are related to climate. Anomaly conditions in warm sea surface temperatures cause the water vapor to overflow into rain-forming clouds, gradually forming high integrated rainfall in some areas of Indonesia. High or extreme rainfall causes a hydro-meteorological disaster in the form of a flood. Musi Banyuasin Regency, South Sumatra, has a concave to flat topography, a swamp area with abundant large and small rivers prone to flood disasters. Between 2012 and 2022, the National Disaster Management Agency (BNPB) recorded 38 locations had been flooded. This study aimed to identify environmental variables that affect the potential flood hazards and areas with a high flood hazard level. This study used a maximum entropy model approach based on machine learning techniques. The model analyzed all the findings in the sample data to produce predictive information on the contributing environmental variables. The sample data was the 38 flood areas with each preliminary fact and topographic characteristic. Threat components were arranged based on environmental variables (aspect, slope, elevation, land cover, rainfall, and distance from the river). The results indicate that contribution of the average rainfall was 58%, elevation was 26.4%, slope angle was 8.6%, slope aspect was 5.8%, land cover was 1%, and river width was 0.1%. Then, the areas with high flood hazard levels were indicated in eight districts, namely Lais, Sekayu, Babat Supat, Keluang, Sungaililin, Lawang Wetan, Babatoman, and Sangadesa.