Demonstrative Application
Statistical modeling applied to sinkhole risk management
Challenge:
Sinkholes are common throughout west-central Florida, often causing unexpected catastrophic damage to community infrastructure, degradation of groundwater quality, and adverse economic impacts on residents, business, and industry. As urban development and population growth apply additional stress on Florida’s landscape, it is imperative to mitigate the risk posed by these natural phenomena. Insurers in Florida not only need reliable information on where the sinkholes are before underwriting so they can prevent premium leakage, but also need to know whether a claim presented as a consequence of a sinkhole is fraudulent.
Solution:
This simulation was conducted using real-world data based on information in a sinkhole-intensive county in Florida. An empirically based statistical model of past sinkhole events, geospatial data layers and a bounded area of interest identified at-risk locations based on discerned causal factors. The model analyzed thousands of geospatial data layers to reveal correlations and patterns among otherwise disparate data. The resulting signature was then applied temporally to anticipate locations of future sinkhole events. This improves regional governments’ responsible growth and development initiatives while enhancing risk evaluation for property insurance providers. In this proof of concept, SPADAC validated that 90.9% of sinkholes since 1990 appeared in the top 25% of the area’s initial assessment.