Uncertainty analysis of 100-year flood maps under climate change scenarios

Abstract

Floods are natural disastrous hazards that throughout history have had and still have major adverse impacts on people’s life, economy, and the environment. One of the useful tools for flood management are flood maps, which are developed to identify flood prone areas and can be used by insurance companies, local authorities and land planners for rescue and taking proper actions against flood hazards. Developing flood maps is often carried out by flood inundation modeling tools such as 2D hydrodynamic models. However, often flood maps are generated using a single deterministic model outcome without considering the uncertainty that arises from different sources and propagates through the modeling process. Moreover, the increasing number of flood events in the last decades combined with the effects of global climate change requires developing accurate and safe flood maps in which the uncertainty has been considered. Therefore, in this thesis the uncertainty of 100-year flood maps under 3 scenarios (present and future RCP4.5 and RCP8.5) is assessed through intensive Monte Carlo simulations. The uncertainty introduced by model input data namely, roughness coefficient, runoff coefficient and precipitation intensity (which incorporates three different sources of uncertainty: RCP scenario, climate model, and probability distribution function), is propagated through a surrogate hydrodynamic/hydrologic model developed based on a physical 2D model. The results obtained from this study challenge the use of deterministic flood maps and recommend using probabilistic approaches for developing safe and reliable flood maps. Furthermore, they show that the main source of uncertainty comes from the precipitation, namely the selected probability distribution compared to the selected RCP and climate model.