Daily, Weekly and Monthly Rain Forecasting using Deep Learning

This research paper focuses on the extensive exploration and forecasting of daily rainfall, a critical element in weather prediction and environmental planning. The study integrates two-time series datasets spanning from 2013 to 2022, encompassing essential meteorological parameters. Through Exploratory Data Analysis (EDA) visualizations, non-stationary patterns are identified, leading to the application of Seasonal Autoregressive Integrated Moving Average (SARIMA) for capturing temporal dynamics. The study then introduces deep learning, specifically Long Short-Term Memory (LSTM) networks, to enhance rainfall prediction by capturing intricate temporal dependencies. A comparative analysis combines SARIMA with Rolling Forecast, showcasing adaptability to evolving patterns. In this compares we find that SARIMA higher accuracy with a high R2 value. Both are outperformed by the hybrid Rolling Forecast with SARIMA, underscoring the necessity of balancing accuracy and flexibility. The paper highlights SARIMA's ability to capture temporal dynamics, acknowledges LSTM's collaboration with meteorological organizations for improved rain forecasting precision in the face of climate change.