MonoGRU: A theoretical and empirical evaluation of a streamlined gated recurrent unit for rainfall forecasting

Accurate rainfall forecasting is a critical component of climate analysis and disaster management, directly influencing agricultural planning, water resource management, and flood mitigation strategies. In this study, we propose an efficient MonoGRU (GRU: Gated Recurrent Unit) - based model for rainfall prediction, optimized for accuracy and computational efficiency. The model utilizes gated recurrent architectures to efficiently capture temporal dependencies in time-series data, while significantly reducing the number of trainable parameters compared to conventional GRU models. Experimental evaluation using walk-forward cross-validation reveals that MonoGRU consistently outperforms GRU model across key performance metrics, such as Mean Average Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R² Score. The Q-Q plot analysis highlights the robustness of MonoGRU in aligning theoretical and sample quantiles, particularly under high epoch settings, indicating its superior predictive reliability. Further comparison of model complexity demonstrates MonoGRU's reduced memory footprint and computational overhead, making it highly suitable for resource-constrained environments. By integrating lightweight design with high accuracy, this work offers a scalable solution for rainfall prediction, paving the way for more efficient forecasting systems. The findings of this research underline the potential of MonoGRU to be extended to other time-series forecasting domains and encourage further exploration of its capabilities.