Severe thunderstorm modeling is becoming increasingly routine as insurers realize that using historical claims data alone is not sufficient for managing this complex risk. As with all catastrophe models, severe thunderstorm models are largely based on historical data, in this case, the frequency and intensity of the individual events (hailstorms, tornadoes, and straight-line winds) that comprise a severe thunderstorm outbreak.
The Storm Prediction Center (SPC) at NOAA has collected this data in the form of event reports called in by local authorities, trained weather spotters, and the general public dating back to the 1950s. However, the data contains reporting biases, i.e., underreporting of events in some locales and overreporting in others. A few culprits include non-uniform population distribution (events in sparsely populated regions are more likely to go unnoticed), better radar detection technology in more recent years, and even the increased interest in amateur storm chasing following the 1996 blockbuster hit, Twister.
These biases must be addressed in order to develop a severe thunderstorm model that provides full spatial coverage of simulated events throughout the United States. In other words, relying on data of where events have been known to occur in the past may not provide a complete picture of where they're likely to occur in the future.
One method of addressing these biases is to use statistical smoothing to allow simulated events to occur around the location of the actual SPC report. While this method would provide coverage in areas where actual events have not been been recorded, events in reality are more likely to occur in certain regions around the SPC reported event and less likely in others.
To figure out where conditions (like moisture, instability, rotation and lift) are more favorable for storm formation, scientists can use the Climate Forecast Systems Reanalysis (CFSR) dataset from the National Center for Atmospheric Research (NCAR). Reanalysis data represent a "best estimate" of the state of the atmosphere at the time of a historical event, based on both observational data and numerical weather prediction.
Reanalysis data can be used to smart smooth the historical data in a more physically realistic way by reducing the probability of events occurring in areas unfavorable for severe thunderstorm formation and increasing the likelihood of an event occurring in areas favorable for formation. The approach, pioneered by AIR Worldwide and first introduced in the recently updated AIR Severe Thunderstorm Model for the U.S., results in simulated events that realistically reflect real-world conditions, which insurers can use to analyze the risk to their own exposures. An additional key benefit of smart smoothing is how it enables the model to capture the risk in rural areas, where events historically were more likely to go unreported, but where insured exposure may exist today. Stay tuned to read more about severe thunderstorm reporting biases in athe November issue of Claims.
© Touchpoint Markets, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.