Development of Administrative District-Based Water Body Monitoring System Using Sentinel-1 SAR and Swin Transformer in South Korea

This study develops an integrated water body monitoring system at the administrative district level using synthetic aperture radar (SAR) images.A semantic segmentation model based on the Swin transformer was constructed and trained on a dataset of 1,043 patches collected from 2018 to 2023.The 5-fold cross-validation results showed an average Intersection over Union (IoU) of 0.9311, precision of 0.9637, and recall of 0.9576, with low variability between folds (IoU standard deviation of 0.0181).The practical applicability of the administrative district-based water body detection system was evaluated through comparison with the normalized difference water index (NDWI) across diverse environmental conditions in Gwangsan-gu, Ulju-gun, and Hapcheon-gun administrative districts.This automated system provides a framework for efficient water resource monitoring that integrates with existing administrative management structures.