Deep Learning Model for Monthly Rainfall Prediction in Metropolitan Cities in India

Rainfall is very crucial in our day-to-day lives, particularly in areas with seasonal rainfall-dependent livelihoods. In this study, a Deep Learning model that seeks to support water resource planning, agriculture, and urban design by providing monthly rainfall predictions precisely is introduced. A large dataset from January 1981 to December 2022 was compiled, including meteorological factors such as humidity, pressure, wind speed, temperature, and precipitation. The data is divided into training (1981–2018) and test (2019–2022) sets in four metropolitan cities: Delhi, Mumbai, Kolkata, and Chennai. The Long Short-Term Memory (LSTM) model applied in this work was hyperparameter tuned (10–20 units) and trained 50 times. The model consists of a single LSTM layer and one fully connected output layer with Rectified Linear Unit (ReLU) activation. It employs the Adam optimizer with a learning rate of 0.001 and Mean Squared Error (MSE) as the loss. Early stopping for a patience level of 20 was used to prevent overfitting. R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, MAE, RMSE, and NSE are used to measure model performance. Violin plots and the Kruskal-Wallis test are employed for the analysis. The outcomes confirm LSTM’s suitability for extracting rainfall patterns and provide suggestions for enhancing the future process based on real-time meteorological information.