Machine learning predicts hydrological drought in the Huaihe River Basin using SPI

A machine learning model accurately predicts hydrological drought with 79.9% accuracy. The Standard Precipitation Index (SPI) is identified as the most influential factor. The XGBoost model, combined with SHAP analysis, was used to predict drought categories in the Huaihe River Basin. While effective for normal and mild droughts, it showed limitations in predicting severe drought events. The study highlights the importance of SPI for drought prediction, with secondary influences varying by season and region, including soil moisture, evapotranspiration, and large-scale climate features.