Estimation of reference evapotranspiration in South Korea using GK-2A AMI channel data and a tree-based machine learning method

Abstract

Changes in evapotranspiration can affect water availability and climate, leading to extreme weather and severe impact on ecosystems. In particular, increased water stress in farmland, forests, and mountainous areas with limited water resources can result in detrimental impacts such as droughts and wildfires. In this study, we utilized data from the Advanced Meteorological Imager (AMI) sensor on the Geostationary Korea Multi-Purpose Satellite 2A (GK-2A) and employed a tree-based machine learning method to accurately estimate reference evapotranspiration (ET_o) in South Korea. The estimated SAT ET_o was compared to the ASOS ET_o, which was estimated using meteorological variables from the Automated Synoptic Observing System (ASOS) and the Penman–Monteith method. The hourly SAT ET_o demonstrated an estimated accuracy with a relative bias (rBias) of −0.26%, a relative root mean square error (rRMSE) of 34.01%, and a coefficient of determination (R²) of 0.94, whereas the daily SAT ET_o exhibited an estimated accuracy with an rBias of −0.25%, an rRMSE of 8.30%, and an R² of 0.97. In this study, various cases were analyzed in detail, including daytime and nighttime, wet and dry conditions, and varying cloud cover. The highly accurate estimation of ET_o using data from the GK-2A satellite, which have high temporal and spatial resolution, can be effectively utilized as monitoring data for water resource management and natural disaster prevention.