Application of LSTM and Climate Index for Prediction of Meteorological Drought in South Korea

Climate change has intensified natural hazards, including droughts, which have caused substantial damage in South Korea. Drought, characterized by prolonged moisture deficiency, is typically assessed using drought indices that reflect meteorological conditions. This study examined the influence of various meteorological and climate indices on drought variability in the Yeongsan and Seomjin River basins. The Standardized Precipitation Index (SPI) was used to represent drought conditions, and its monthly variations were predicted using the Long Short-Term Memory (LSTM) algorithm. To assess model performance, four statistical indicators—Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Nash–Sutcliffe Efficiency (NSE), and the Coefficient of Determination (R2)—were employed. The LSTM model that utilized both precipitation and drought indices as input data showed the best performance, achieving an MSE of 0.2, RMSE of 0.477, NSE of 0.77, and R2 of 0.78. Overall predictive performance ranged from 0.298 to 0.347 (MSE), 0.546 to 0.589 (RMSE), 0.578 to 0.628 (NSE), and 0.580 to 0.675 (R2). This study highlights the effectiveness of LSTM in predicting drought conditions and the value of incorporating meteorological and climatic indices. The results can support improved drought hazard assessment and management strategies in South Korea.