SAR imagery and deep learning techniques for reservoir monitoring in Korea

Abstract

Agricultural reservoirs are key structures for water supply on the Korean Peninsula, where the water resources are concentrated seasonally. Monitoring of agricultural reservoirs is essential for efficient management of available water resources. However, in the case of the Korea, there are many unmeasured reservoirs without observation facilities, so it is difficult to monitor available water at a regional scale. Remote sensing-based reservoir monitoring that can observe the water surface in a wide area is essential. In the case of Synthetic Aperture Radar (SAR) image, continuous water body detection is possible regardless of weather conditions. Recently, water body detection research using AI techniques has been actively conducted to improve accuracy. In this study, water body detection was performed on an agricultural reservoir using Sentinel-1 SAR image and AI-based U-net, HR-Net, and Swin-Transformer techniques. The water/non-water binary classification images from the Sentinel-2 satellite were used for validation. In addition, time series validation was performed using in-situ reservoir storage and evaluated the performance of each deep learning techniques. If SAR image with high spatial and temporal resolution can be utilized in the future, it is expected that more efficient management of available water resources will be possible. Keywords: Sentinel-1, SAR, Deep learning, Water body detection, Reservoir Acknowlegment: This work was supported by the “Development of Application Technologies and Supporting System for Microsatellite Constellation”project through the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. 2021M1A3A4A11032019). This research was supported by the BK21 FOUR (Fostering Outstanding Universities for Research) funded by the Ministry of Education (MOE, Korea) and National Research Foundation of Korea (NRF). This work is financially supported by Korea Ministry of Land, Infrastructure and Transport (MOLIT) as 「Innovative Talent Education Program for Smart City」