Speaker
Description
Flood disasters are among the most frequent and damaging natural hazards globally, with their impacts exacerbated by urbanization (Rentschler et al., 2023). The precise identification of high-risk flood zones is crucial for effective disaster-resilience planning. Recent advancements in deep learning demonstrate potential in flood prediction, enabling the extraction of complex, nonlinear relationships from large-scale data with high accuracy and efficiency (Zeng, Huang and Chen, 2025). Existing flood prediction models primarily consider factors like climate, topography, and land use (Wang et al., 2020), yet they frequently neglect the significant impact of urban form on flood dynamics (Balaian, Sanders and Qomi, 2024). In general, the present research frameworks overlook the morphological differences and diversity among cities, thereby constraining the understanding of the complex interactions associated with floods.
In this vein, this study proposes a novel framework of flood risk prediction that integrates urban morphological metrics as key predictors. Using Zhengzhou, China, as a case study, we simulated flood risk with historical satellite-derived inundation data and 19 predictor variables, including natural factors, land cover, and urban morphology. We leveraged building footprint data to derive refined urban morphological metrics, providing deeper insights into the influence of urban form on flooding dynamics. Using the R package foot (Jochem and Tatem, 2021), we computed metrics such as mean building shape index, building orientation entropy, mean building compactness and urban porosity, within 100m grid cells applying a 250m focal radius window centered on each template grid cell. These metrics, along with other variables, were incorporated into a hybrid model combining Convolutional Neural Networks (CNNs) and Random Forest. The model was trained on flood inundation data from the extreme rainfall event in July 2021, enabling high-resolution flood risk predictions for any geographic location in Zhengzhou.
Our results revealed a significant correlation between urban form and flood risk, with urban morphological metrics closely associated with flood risk (p < 0.05). Flood areas with medium and high risk account for 11.53% of the total study area (7.75% medium risk and 3.78% high risk). These areas showed substantial overlap with the distribution of critical infrastructure and public service facilities, emphasizing the higher risk of service disruptions in these regions. To address these risks, we recommend prioritizing the resilience of physical and social infrastructure. Additionally, this study provided a robust predictive model for assessing the impacts of climate change and urban spatial development, offering a valuable tool for decision-making in flood risk management.
References
Balaian, S.K., Sanders, B.F. and Qomi, M.J.A. (2024) How urban form impacts flooding. Nature Communications, 15(1), p.6911.
Jochem, W.C. and Tatem, A.J. (2021) Tools for mapping multi-scale settlement patterns of building footprints: An introduction to the R package foot. PLOS ONE, 16(2), pp. e0247535.
Rentschler, J., Avner, P., Marconcini, M., Su, R., Strano, E., Vousdoukas, M. and Hallegatte, S. (2023) Global evidence of rapid urban growth in flood zones since 1985. Nature, 622(7981), pp.87-92.
Wang, Y., Fang, Z., Hong, H. and Peng, L. (2020) Flood susceptibility mapping using convolutional neural network frameworks. Journal of Hydrology, 582, p. 124482.
Zeng, B., Huang, G. and Chen, W. (2025) Research progress and prospects of urban flooding simulation: From traditional numerical models to deep learning approaches. Environmental Modelling & Software, 183, p.106213.
| Keywords | Flood Risk Prediction; Urban Form; Deep Learning; China |
|---|---|
| Best Congress Paper Award | No |
