Sarah Lawrence

Major: Geography

Faculty Advisor: Sam Brody

Project Title:

Delineating the reality of flood risk: a comparison of physics based and machine learning models

Abstract

Flood incidents are increasingly not aligning with the traditional indicator of flood hazard: The Federal Emergency Management Agency's (FEMA) 100-year floodplain. To stay resilient, homeowners need precise and reliable information concerning their home's flood risk. New models of flood risk have been proposed to help fill this gap, but validation of these models is difficult due to a lack of residential flood exposure information. In particular, a recently developed flood hazard model was created using historic flood insurance claims that can provide important flood hazard information to residents at the parcel level. This study is the first to validate such a model using survey results from residents in Southeast Texas. It is hypothesized that this new delineation of flood risk will show marked improvement in capturing residents feedback regarding flood frequency as compared to the 100-year FEMA floodplain. The number of times a resident has been flooded in their current home was used as ground-truth information to test the accuracy of the new flood hazard model and FEMA's floodplain. This paper used exploratory data analyses of where flooding has occurred relative to the delineations, t-tests to cross-validate differences in model predictor variables, and a receiver operating characteristic (ROC) curve to test the accuracy of the newly developed flood hazard model. Survey responses are still being received. However, initial results illustrate that the newly developed flood hazard model adequately captures respondents experiences with floods and exhibits decent overall accuracy. Unlike FEMA’s binary 100-year floodplain, the flood hazard model uses a continuous spectrum to classify levels of flood risk further enhancing its effectiveness. Making this flood risk communication tool available in addition to FEMA’s floodplain information can help residents make better informed choices for disaster resilience. Furthermore, it can assist decision makers in determining areas where help is most crucial.

Funding: National Science Foundation (NSF) under Grant No. 1950910. REU: OCEANUS at Texas A&M, Galveston Campus and the Institute for a Disaster Resilient Texas