Bottom-up multilevel flood hazard mapping by integrated inundation modelling in data scarce cities

With increasing urban flood risk due to urbanisation and climate change, flood hazard prediction is ever more crucial for flood risk management and emergency response. However, when sufficient high-quality data are lacking, a standard flood inundation modelling approach has significant uncertainties. Therefore, this study develops a bottom-up approach for urban flood hazard mapping at multiple levels (grid-kilometre-district), built upon the integration of grid-based flood modelling with data acquisition from open sources. This reduces the adverse effects of data scarcity and quality on hazard modelling. In the paper, we first set out an integrated approach for gridded inundation mapping in an urban basin using a hydrodynamic model supported by crowd-sourced social media data. Then, applying the model to flooding in Chengdu in August 2020, we articulate how input data bias and parameter uncertainty both affect urban inundation modelling. The results show that the choice of terrain data and the quantification of urban drainage flows significantly influence the modelled urban inundation extent, depth and duration. This indicates the potential for large variation when using urban inun-dation mapping merely relying on a single-scale flood hydrodynamic modelling even when supported by crowd-sourced data. The multilevel hazard mapping approach developed here presents multi-layered and compre-hensive inundation mapping; thus the effects of data bias or availability are reduced and the coarser hazard mapping shows less sensitivity to the data input quality and model uncertainty, indicating relatively higher reliability at this higher spatial scale. The grid-kilometre-district three level approach provides more reliable flood hazard mapping, which can support rainstorm-induced flood management in data scarce cities.

Automated summary

This study develops a bottom-up approach for urban flood hazard mapping, by integrating grid-based flood modelling with data acquisition from open sources. The approach reduces the adverse effects of data scarcity and quality on hazard modelling.