Implementing Extreme Value Analysis in a Geospatial Workflow for Storm Surge Hazard Assessment

Abstract:

Gridded data of 100-yr (1%) and 500-yr (0.2%) storm surge flood elevations for the United States, Gulf of Mexico, and East Coast are critical to understanding this natural hazard. Storm surge heights were calculated across the study area utilizing SLOSH (Sea, Lake, and Overland Surges from Hurricanes) model data for thousands of synthetic US landfalling hurricanes. Based on the results derived from SLOSH, a series of interpolations were performed using spatial analysis in a geographic information system (GIS) at both the SLOSH basin and the synthetic event levels. The result was a single grid of maximum flood elevations for each synthetic event. This project addresses the need to utilize extreme value theory in a geospatial environment to analyze coincident cells across multiple synthetic events. The results are 100-yr (1%) and 500-yr (0.2%) values for each grid cell in the study area.

This talk details a geospatial approach to move raster data to SciPy’s NumPy Array structure using the Python programming language. The data are then connected through a Python library to an outside statistical package like R to fit cell values to extreme value theory distributions and return values for specified recurrence intervals. While this is not a new process, the value behind this work is the ability to keep this process in a single geospatial environment and be able to easily replicate this process for other natural hazard applications and extreme event modeling.