August 1, 2025Open Access

A highly generalizable data-driven model for spatiotemporal urban flood dynamics real-time forecasting based on coupled CNN and ConvLSTM

View Full Paper

Ask AI

Bookmark

Authors

Discussion

A novel model predicts spatiotemporal urban flood dynamics in Macao, demonstrating robust generalization and accuracy.

The data-driven model effectively captures inundation processes, achieving NSE values greater than 0.80 for most events.
Generalization performance is strong, with mean NSE over 0.70 and low RMSE and MAE values compared to physics-based models.
The model integrates both spatial features and temporal sequences, improving its forecasting capabilities.
Localized discrepancies occur mainly near abrupt terrain changes, indicating areas for potential model refinement.

Lou et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark