This study investigates advancements in real-time pluvial flash flood forecasting (PFFF) to enhance flood mitigation capabilities in the context of climate change. It evaluates Malaysian flood prediction models and explores machine learning techniques, including support vector regression (SVR), random forest (RF), and long short-term memory (LSTM) for flash flood (FF) prediction. The literature review highlights progress in understanding weather and hydrological factors contributing to flash floods and use of machine learning models to improve prediction accuracy. The integration of real-time radar rainfall data and river monitoring enhances forecasting and early warning systems. The study compares RF, SVR, and LSTM models, showing RF outperforms SVR in accuracy, precision, recall, and other metrics. RF’s ability to capture complex interactions between weather, hydrology, and landscape is validated through cross-validation. Additionally, combining RF and LSTM models achieves higher accuracy of 0.65 in FF prediction than combining SVR and LSTM, which was 0.47. This study emphasizes the utility of RF for real-time FF forecasting and offers insights into future research in flood management. The novelty of this study is integrating two machine learning models, SVR and RF with a deep learning model, LSTM to improve the accuracy of predicting flash flood events.
Latif et al. (Fri,) studied this question.