Wildfire catastrophe is one of the leading causes of losses in the insurance industry. Wildfire can cause heavy losses to the insured and the insurer. In the event of loss, delay in reaching the Insured can have unintended consequences in terms of an increased expense to the insurer and decreased policy renewals by the customer.
As per the National Interagency Fire Center (NIFC), 50,477 wildfires burned about 4.7 million acres in 2019. The wildfire named Camp Fire in Butte County in 2018 is considered as the most severe fire in California. The Insurance Information Institute (III) estimated the insured loss of the fire to be between $8.5 billion to $10.5 billion at the time of the loss. In 2018, Wildfire resulted in an estimated $24 billion loss with Insurers paying out a total of $18 Billion.
Wildfire can cause huge property losses within a short time frame. Wildfire is extremely hard to predict; the models used by insurers quickly become outdated due to various factors such as climate change, rapidly changing burn probabilities and localized urbanization. Considering the increased risk factors and huge losses, some of the insurers have increased insurance rates for homeowners in counties that are vulnerable to wildfire.
The framework to identify wildfire affected customers and potential losses
Better wildfire detection and quick response to early warning can help avoid some of the losses and overcome a few of the challenges faced by insurers. Insurance companies can use early wildfire detection to estimate losses and to pass on the early warning to the Insurers. Early warning can be used to initiate pre-planned wildfire loss mitigation activities.
Insurance companies can use wildfire data feed and the impacted customers within the wildfire geolocation to build a framework for calculating the potential exposure losses.
A High-Level Approach for Creating a Wildfire Potential Loss Estimate Framework
Here is an outline of the approach that can be used in a wildfire data-based framework:
- Framework will be built using a Microservices-based architecture
- Machine learning models with remote sensing data can detect fire within a span of a couple of hours. The near real-time feed on wildfire, which contains the geolocation boundaries of the wildfire perimeter data, is fetched through API calls or by accessing Keyhole Markup Language (KML) files.
- The KML files can then be parsed using Java Libraries to fetch the wildfire boundary coordinates.
- Latitude and longitude of the insured addresses can be calculated using Geocoding APIs.
- High-level exposure impact can be determined using the exposure risk and the distance between geolocation of the insured and the wildfire boundaries.
- The distance and location can be used to plot the risk on user i
Following are a few benefits of the framework:
- Customers in potential hazard zones are alerted to find a safe place and safeguard the property if possible.
- Actual and potential losses are calculated faster and more accurately.
- The service teams get the lead-time to serve the customers in crisis.
The framework will enable insurance companies to respond in a better way to a wildfire catastrophe. They will be able to proactively predict the impact of the exposure, contact the customers immediately, and prepare for a better customer claims outcome.