Automated flood extent identification using WorldView imagery for the insurance industry

摘要

Flooding is the most common and costly natural disaster around the world, causing the loss of human life and billions in economic and insured losses each year. In 2016, pluvial and fluvial floods caused an estimated 5.69 billion USD in losses worldwide with the most severe events occurring in Germany, France, China, and the United States. While catastrophe modeling has begun to help bridge the knowledge gap about the risk of fluvial flooding, understanding the extent of a flood - pluvial and fluvial - in near real-time allows insurance companies around the world to quantify the loss of property that their clients face during a flooding event and proactively respond. To develop this real-time, global analysis of flooded areas and the associated losses, a new methodology utilizing optical multi-spectral imagery from DigitalGlobe (DGI) WorldView satellite suite is proposed for the extraction of pluvial and fluvial flood extents. This methodology involves identifying flooded areas visible to the sensor, filling in the gaps left by the built environment (i.e. buildings, trees) with a nearest neighbor calculation, and comparing the footprint against an Industry Exposure Database (IE) to calculate a loss estimate. Full-automation of the methodology allows production of flood extents and associated losses anywhere around the world as required. The methodology has been tested and proven effective for the 2016 flood in Louisiana, USA.