Short-term prediction of hourly reference evapotranspiration in Gangwon State, South Korea, based on numerical weather prediction data

Abstract Accurate prediction of reference evapotranspiration (ET o ) is vital for water resource management, especially in water-limited regions that are difficult to monitor in situ. In this study, we performed short-term ET o prediction for Gangwon State, South Korea, using the Local Data Assimilation and Prediction System (LDAPS) data from the Unified Model and the Penman–Monteith (PM) evapotranspiration approximation method. ET o was predicted at 1-h intervals over 48 h for each LDAPS run time (00, 06, 12, and 18 UTC). ET o prediction using the PM method varied in sensitivity to meteorological variables, particularly daytime downward shortwave radiation and air temperature and nighttime relative humidity. The predicted ET o was compared with meteorological observation data from the Automated Synoptic Observing System (ASOS) located in Gangwon State and hourly ET o estimated through the PM method. To improve prediction accuracy, the bias of the predicted ET o was linearly corrected using 3-year prediction data. In the evaluation of prediction accuracy, LDAPS ET o showed a mean absolute error of 0.04 mm h −1 , root mean square error of 0.07 mm h −1 , and correlation coefficient of 0.91, compared with ASOS ET o . Cases with low-to-moderate ET o (0–0.6 mm h −1 ) had an equitable threat score above 0.5. The ET o prediction method of this study can be used for the installation and maintenance of observation equipment for evapotranspiration as well as for the management and monitoring of water resources in areas where it is difficult to observe meteorological variables, thereby contributing to the prevention of meteorological disasters.