Developing reliable urban flood hazard mapping from LiDAR data

Digital surface models (DSMs) are crucial in providing accurate urban flood hazard maps. The ubiquitous availability of LiDAR data (where accessible) makes constructing geometrically sound DSMs feasible. However, little attention has been paid to developing approaches for producing geometrically consistent DSMs. Herein is described an application-driven procedure for creating a geometrically robust DSM (DSM1). Two further DSMs were created, one for portraying streets using breaklines as ancillary information (DSM2) and the other through direct interpolation of LiDAR data (DSM3). The geometrical correctness and vertical accuracy of these DSMs were examined qualitatively and quantitatively by plotting longitudinal profiles and cross-sections onto major runoff pathways and determining statistical error. The effect of these DSMs' geometric consistency on flood hazard maps was also evaluated. For this, hydraulic outputs from DSM1 were used as a benchmark to compare hydraulic outputs from DSM2 and DSM3. This comparison was conducted at two spatial resolutions: i) at the total area flooded using the F statistic; and ii) at the pixel level by employing global indices and category-level indices extracted from a confusion matrix. Our findings revealed that: 1) DSM1 defined the most geometrically coherent configurations for runoff pathways; 2) in urban areas with the most densely packed streets and buildings, DSM2 and particularly DSM3 featured the most unrealistic geometric representations of the urban domain, displaying fake water flow barriers and lower than real runoff pathway cross-sections; and 3) the geometric quality of the DSMs created had a significant impact on flood hazard maps reliability (i.e., the disagreement in flood hazard categories between DSM2 and DSM3 and DSM1 varied from 28% to 82%). These findings can be very valuable in achieving further reductions and better flood risk management.