Published Paper Details:

Accurate rainfall forecasting is vital for India's agricultural sustainability, water resource management, and climate resilience. This study conducts an exhaustive evaluation of four advanced time-series forecasting models--Long Short-Term Memory (LSTM) neural networks, Seasonal Autoregressive Integrated Moving Average (SARIMA), Facebook Prophet, and Extreme Gradient Boosting (XGBoost)--to predict annual rainfall from 2016 to 2030, leveraging India's area-weighted rainfall dataset from 1901 to 2015. Each model is meticulously assessed for its ability to capture temporal trends, seasonal patterns, and nonlinear dependencies. The findings reveal distinct predictive behaviors: LSTM projects a steady increase in rainfall, reaching 1207.36 mm by 2030, indicating potential climatic shifts; SARIMA predicts stable oscillations between 1070-1135 mm, reflecting historical variability; Prophet forecasts a gradual decline from 1121.83 mm to 1100.96 mm, emphasizing long-term trends; and XGBoost yields a constant 1151.65 mm, highlighting its limitations without temporal feature engineering.