Quantifying Uncertainty in Flood Predictions in Fixed Cartesian Flood Model Due To Arbitrary Conventions in Grid Alignment

Digital elevation models generated by sampling and interpolating LiDAR data onto a square grid can produce reliable flood predictions. However, the arbitrary conventions in grid alignment that can introduce uncertainty in flood predictions are frequently overlooked. Hence, our research quantified this uncertainty using a Monte Carlo approach and flood model LISFLOOD-FP to generate multiple flood simulations for analysis. The simulations were generated by transforming the alignments of the square grid (North translation, East translation, North-East translation, rotation, and a combination of rotation and translation) with different resolutions (2-, 5-, 10-, and 20-m). We also used different flood scenarios (5-, 10-, 20-, 50-, 80-, and 1,000-year return periods) to observe how the uncertainty changes in a specific event. Results demonstrate that the grid alignment introduces uncertainty in flood predictions, leading to significant variability in flood extent (7%) and the number of flooded buildings (27%). Because the main river aligns with the grid lines, higher variability in water depths, flood extent, and flooded buildings is associated with grid rotation than translation. Finer resolutions have less variability in water depths, flooded areas, and the number of flooded buildings owing to the decreased movement of LiDAR points between pixels. For each flood scenario, if water overtops certain thresholds in only a few simulations, variations in water depths and flooded areas increase. However, if it only fills locations that can be flooded by water volume in smaller flood event, they decrease. The number of flooded buildings depends on if the inundated regions are residential.