Flood mapping under uncertainty: a case study in the Canadian prairies

Abstract

In Canada, approaches for hazard mapping involve using a hydraulic model to generate a flood extent map that distinguishes between inundated and non-inundated areas in a deterministic way. The authors adapt a probabilistic approach to obtain flood hazard maps by accounting for uncertainty in boundary conditions and calibration parameters of a hydrodynamic model, while considering extent, probability, and depth of inundation as important flood indicators. The Qu’Appelle River in the Canadian prairies is considered as a case study to demonstrate the benefits of a probabilistic approach to obtain flooding indicators. The probability and depth of inundation are obtained for all locations in the floodplain, and their uncertainties are evaluated. A sensitivity analysis to determine factors affecting the extent of flooding at multiple locations along the river is also carried out. Further, the criterion to distinguish between flooded and non-flooded areas is modified and its subsequent effect on the flooding indicators is also evaluated. Results show low variability in depth and probability of inundation at most locations along the floodplain. It is also observed that the influence of the roughness parameters on the flooding extents is higher in the steeper stretches of the river, compared to the flatter stretches. Another component of the presented work focusses on evaluating an existing topography-based hazard map for Canada that classifies the country into different levels based on severity of flooding, by comparing it with the probabilistic hazard map obtained in this study. The comparison indicates a good level of agreement between both maps and highlights the reliability of using the topography-based hazard map where hydraulic modeling is infeasible.