AI-enabled flood mapping from street photos: the case of 2021-22 floods in U.S. and Canada

Successful flood response and evacuation require timely access to reliable flood depth information in urban areas. However, existing flood depth mapping tools do not provide real-time flood depth information in residential areas. In this paper, a deep convolutional neural network is used to determine flood depth through the analysis of crowdsourced images of submerged stop signs. Model performance in pole length estimation is tested on a test set, achieving root mean squared error (RMSE) of 10.200 in. on pre-flood photos and 6.156 in. on post-flood photos, and an average processing time of 0.05 seconds. The performance of the developed model is tested on two case studies: Hurricane Ian in the U.S. (2022), and the Pacific Northwest floods in the U.S. and Canada (2021), yielding MAE of 4.375 in. and 6.978 in., respectively. The overall MAE for both floods was achieved as 5.807 in., which is on par with previous studies. Additionally, detected flood depths are compared with readings reported by the nearest flood gauge on the same date. The outcome of this study demonstrates the applicability of this approach to low-cost, accurate, scalable, and real-time flood risk mapping in most geographical locations, particularly in places where flood gauge reading is not attainable.