Hail Modeling Explained

With the aim of producing more accurate loss information as well as pinpointing emerging areas of exposure, models are being used to try and control the multi-billion dollar losses caused by hail.

During a podcast interview with Claims Journal, Dr. Arindam Samanta, director of product management and innovation for Verisk Analytics, said that hail models are invaluable in that they can review past hail exposure and predict future risk.

“It’s really unique because you can get claims because of events that have happened in the past… You can also have claims due to future exposure to damaging hail events,” said Dr. Samanta. “Our research shows that roughly one in three claims tends to have erroneous dates of loss. Claims tend to come in several months and years after hail events. In many situations people had not noticed that there was a problem on the roof.”

In addition, Dr. Samanta said that while certain states are more prone to hail events, that’s changing.

“Having the ability to model hail that looks at both the past exposure as well as the future risk is a very critical piece of information that can be useful in a range of insurance applications, especially when combined with information on roof age, roof type, and other construction characteristics,” Dr. Samanta explained.

Modeling hail exposure can aid underwriting during the inspection process, in writing applications and in setting deductibles, he said.

“It can also help inform the process of segmentation and rating, as well as in setting up exposure management guidelines,” said Dr. Samanta.

Hail modeling incorporates state of the art data and techniques in weather and climate research and machine learning, Dr. Samanta said. Real-time data feeds and ground-based radar collect data on fast moving storms every two to five minutes.

“This data stream is fully enriched with host of data coming from weather sensor networks, from ground observations, as well as from weather models. This kind of framework is essential for accurate records of hail property locations. This guides our proprietary algorithms that create hail models,” said Dr. Samanta.

A hail model will show what the actual exposure to damaging hail was in the past as well as whether certain areas have a higher exposure to damaging hail, he said.

At a property level, it should tell us what was the actual exposure to damaging hail was in the past, for example,” he said. “‘What was the exposure in the past few months, in the past few years?’ That’s one piece of information which is very useful for the inspection and underwriting uses. The other piece of information that it should help us with is whether certain areas, certain properties, have a higher risk of future exposure to damaging hail events.”

Dr. Samanta said there are significant differences in modeling hail versus hurricanes and tornadoes. Hail tends to occur in narrow swathes versus hurricanes, which tend to occur over broad areas. The characteristics of how the hail was distributed is important, he said.

“What areas in particular were under the influence of certain hail sizes? What areas were under the influence of, for example, less three quarters of an inch? What areas were under the influence of greater than triple of an inch, or greater than one inch, or greater than two inch? Because that kind of information, that kind of granularity, is essential to understanding at the property level where the maximum scope for damage is,” he explained.