Machine learning-driven rainfall forecasting model for sustainable and adaptive infrastructure planning
Hasan Ahamed Alif, Md. Jisan Mashrafi, Urmi Haldar, Md Mohibur Rahman, Md Alamgir Miah, Mukther Uddin, Sharjil Bin Yousuf, Md Jaowad Hasan
Abstract
Climate variability and extreme rainfall events pose significant challenges to infrastructure development in Bangladesh, where standard statistical models often fail to account for nonlinear anomalies and low-frequency extremes. This study employs two machine learning approaches, Prophet and Long Short-Term Memory (LSTM) networks, to predict long-term yearly rainfall using historical data from 1980 to 2024 across two climatically sensitive locations. Prophet, an additive decomposition model, and LSTM, a recurrent neural network with memory-based learning, were benchmarked using Root Mean Squared Error (RMSE) and the Coefficient of Determination (R2). Results demonstrate that LSTM consistently outperformed Prophet in both Rajshahi and Ishwardi, obtaining lower RMSE values (102.4 mm and 118.7 mm, respectively) and higher R2 scores (0.88 and 0.85). While Prophet generated smoother forecasts, it underfitted severe years, whereas LSTM correctly captured interannual volatility and identified a greater number of high-risk years above the 90th percentile threshold. The estimates were transformed into practical adaptation techniques, including elevated foundation design, decentralised rainwater collection, and the use of water-resistant materials, thereby integrating predictive analytics into civil engineering applications. This integration provides a framework for climate-resilient infrastructure planning aligned with national adaptation plans. Future studies should expand to include multivariate forecasting that incorporates exogenous climate drivers, such as ENSO, soil moisture, and vegetation indices, as well as examine the scalable integration of GIS and BIM platforms for real-time urban resilience design.