Enhancing Drought Detection and Visualization with LSTM and SPEI: Addressing Slow-Onset Climate-Induced Water Scarcity

Natural disasters, like droughts, are extraordinarily complex and long-lasting, which highlights how important they are in India because of their frequent occurrences. In an effort to achieve accurate drought prediction, the current study explores the approaches of machine learning (ML), such as support vector machines, neural networks, and deep learning. Effective management of the drought and resource allocation become more important in these situations, especially in regions like Latur district where agriculture is strongly dependent on these resources and water scarcity problems continue.Using historical climatic data that includes variables like temperature and precipitation, the suggested methodology calculates the Standardized Precipitation-Evapotranspiration Index (SPEI) for the Latur region. The study attempts to improve Long Short-Term Memory (LSTM) model predictions by integrating SPEI values in utilizing a high-quality time series dataset obtained from the Indian Meteorological Department (IMD). Interactive insights are provided through visual representations of temperature, precipitation, and SPEI time series data. This model shows a low MSE (0.0187) means small prediction errors. High R^2 (0.99832) shows model explains 99% of data variance accurately.The study aims to improve computational methods and increase the amount of climate variables incorporated, which will help to build stronger models for drought prediction and resilience building. MSC Codes - 68T01, 68T07 JEL Codes - C32