An Optimization Model for Floodplain Systems Considering Inflow Uncertainties

Capturing the uncertainties is of the main concerns in optimal design of floodplain systems. This study deals with risk-based optimal design of flood mitigation measures using a random sampling technique and simulation-optimization approach. An optimization-based inverse approach is developed to map flood discharges, generated by Latin Hyper Cube Sampling (LHS) technique, to the surface hydrographs of sub-watersheds. Provided hydrographs are then imported to a hydrodynamic model to calculate floodplain inundation and to estimate expected damages of generated floods underlying various flood mitigation measures. Models are coupled with NSGA-II optimization algorithm to produce Pareto optimal solutions between two competitive objectives: minimizing i) investment costs and ii) potential flood damages. The proposed approach is finally applied to a small watershed and cost-effective designs of floodplains are derived along their confident intervals. The results give valuable information to decision makers about benefit to cost ratios and the value of robustness for the obtained solutions.