Gmm Assisted Clustering for Long-Term Meteorological Drought Prediction in Jharkhand Using LSTM

Rainfall is the primary source of precipitation in India’s tropical regions, yet its uneven spatial and temporal distribution frequently leads to meteorological droughts, impacting agriculture and water resources. This study examines precipitation variability during the monsoon months (June to September) over a 20-year period (2003-2023), capturing both short-term and long-term trends, using the Standardized Precipitation Index (SPI) and data from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS). Through the application of an enhanced deep Long Short-Term Memory (LSTM) model in integration with a Gaussian Mixture Model (GMM), the study clusters and forecasts SPI data, providing a more accurate understanding of drought occurrences across the Jharkhand State. The study further evaluates the model’s performance across different precipitation clusters, including challenges in medium SPI predictions, ensuring an in-depth analysis of its forecasting capabilities. The LSTM model demonstrated strong proficiency in capturing significant trends in the SPI data, with a Root Mean Squared Error (RMSE) ranging from 0.095 to 0.08 across clusters, signifying high prediction accuracy. These findings contribute to improving drought prediction and water resource management, emphasizing the study’s significance in mitigating the impacts of climate variability.