Flood Risk Modelling Based on Machine Learning Using Google Earth Engine in Hulu Sungai Utara Regency
Krisna Aditya, Ichsan Ridwan, Nurlina Nurlina · Nature Environment and Pollution Technology · 2025
Flood risk modeling is essential for effective disaster mitigation, particularly in flood-prone areas such as Hulu Sungai Utara Regency, Indonesia. This study leverages Google Earth Engine (GEE) to integrate multi-source satellite data and machine learning techniques for flood susceptibility mapping. Key geospatial variables, including the Normalized Difference Vegetation Index (NDVI), elevation, distance from rivers, and the Topographic Position Index (TPI), were analyzed using a weighted overlay method within GEE. A supervised classification approach was employed to enhance accuracy, and validation was performed using historical flood event data. The results indicate that 51.66% (47,875.86 ha) of the study area falls into the low-risk category, 42.90% (39,763.08 ha) is at moderate risk, and 5.44% (5,040.36 ha) is highly susceptible to flooding. This study highlights the advantages of GEE in large-scale flood risk assessments by enabling real-time processing, high computational efficiency, and seamless integration of geospatial datasets. The findings provide critical insights for local governments and disaster management agencies to develop proactive flood mitigation strategies.