Flexible methodology for hyperlocal flooding risk due to sea level rise

Abstract:

Assessments of hyper-local flood risk as a result of sea level rise are often inhibited by computational cost and complex tail behavior over the domain. In order to compute annual exceedance probabilities over a large domain it was necessary to optimize and automate complex statistical tools. We developed a scalable workflow that can be deployed in parallel and run with AWS Batch. To address the spatial variability in extremes, we test a set of extreme value distributions with and without trends in local sea level rise over time and scenario, and attempt to objectively select the optimal distribution for each point. These distributions are then used to compute return levels with uncertainty estimates, and depending on the application, can be used as inputs for subsequent physical models in the modeling chain or as the final product. Results will be presented in the context of our work in the Houston area.