Estimation of possible extreme droughts for a dam catchment in Korea using a regional-scale weather model and long short-term memory network

Abstract To prepare measures to respond to climate-induced extreme droughts, consideration of various weather conditions is necessary. This study tried to generate extreme drought weather data using the Weather Research and Forecasting (WRF) model and apply it to the Long Short-Term Memory (LSTM), a deep learning artificial intelligence model, to produce the runoff instead of using conventional rainfall–runoff models. Finally, the standardized streamflow index (SSFI), the hydrological drought index, was calculated using the generated runoff to predict extreme droughts. As a result, the sensitivity test of meteorological data to runoff showed that using similar types of meteorological data could not improve runoff simulations with a maximum difference of 0.02 in Nash–Sutcliffe efficiency. During the drought year of 2015, the runoff generated by WRF and LSTM exhibited reduced monthly runoffs and more severe SSFI values below −2 compared to the observed data. This shows the significance of WRF-generated meteorological data in simulating potential extreme droughts based on possible physical atmospheric conditions using numerical representations. Furthermore, LSTM can simulate runoff without requiring specific physical data of the target catchment; therefore, it can simulate runoff in any catchment, including those in developing countries with limited data.