Uncertainty Quantification in Flood Inundation Mapping Using Generalized Likelihood Uncertainty Estimate and Sensitivity Analysis

The process of creating flood inundation maps is affected by uncertainties in data, modeling approaches, parameters, and geo-processing tools. Generalized likelihood uncertainty estimation (GLUE) is one of the popular techniques used to represent uncertainty in model predictions through Monte Carlo analysis coupled with Bayesian estimation. The objectives of this study are to (1) compare the uncertainty arising from multiple variables in flood inundation mapping using Monte Carlo simulations and GLUE and (2) investigate the role of subjective selection of the GLUE likelihood measure in quantifying uncertainty in flood inundation mapping. The role of the flow, topography, and roughness coefficient is investigated on the output of a one-dimensional Hydrologic Engineering Center-River Analysis System (HEC-RAS) model and flood inundation map for an observed flood event on East Fork White River near Seymour, Indiana (Seymour reach) and Strouds Creek in Orange County, North Carolina. Performance of GLUE is assessed by selecting three likelihood functions including the sum of absolute error (SAE) in water surface elevation and inundation width, sum of squared error (SSE) in water surface elevation and inundation width, and a statistic (F-statistic) on the basis of the area of observed and simulated flood inundation map. Results show that the uncertainty in topography, roughness and flow information created an uncertainty bound in the inundation area that ranged from 1.4 to 4.6% for Seymour reach and 4 to 29% for Strouds Creek of the base inundation areas. Additionally, flood inundation maps produced by applying GLUE have different uncertainty bounds depending on the selection of the likelihood functions. However, the overall difference in the flood inundation maps on the basis different likelihood functions is less than 2%, suggesting that the subjectivity involved in selecting the likelihood measure in GLUE did not create a significant effect on the overall uncertainty quantification in flood inundation mapping of the selected study areas. DOI:10.1061/(ASCE)HE.1943-5584.0000476. (C) 2012 American Society of Civil Engineers.