Reproduce “A Comprehensive Survey on Deep Learning Solutions for 3D Flood Mapping” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wenfeng Jia, Bin Liang, Yuxi Liu, Muhammad Arif Khan, Lihong Zheng
The authors review 39 studies of deep-learning methods for 3D flood mapping, which combines flooded-area extent with estimated water depth. They organize methods into task-decomposition and end-to-end workflows and consider static and dynamic mapping. One surveyed framework ran 19,585 times faster than MIKE 21 with 9.5% relative error, but reliance on simulated data limits real-world applicability.
Key insights
- Deep learning models provide significant computational efficiency, often generating 3D flood maps orders of magnitude faster than traditional hydrodynamic models.
- Higher spatial resolution of input data, particularly Digital Elevation Models (DEMs), correlates directly with improved accuracy in flood depth prediction.
- End-to-end deep learning methods for dynamic scenarios, while often relying on simulated data, are the most common and demonstrate strong performance in predicting both flood extent and depth.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.