Geocoding precision key to accurately estimating potential flood losses

Precision of insurer location is critically important when it comes to modelling flood risk and estimating potential losses should such an event occur, attendees were told Friday during the Toronto installment of the Aon Benfield Catastrophe Analytics Roadshow.

Detailed information, especially for the flood modelling, is essential, emphasized Vaclav Rara, a flood model developer and hydraulic modeler at Aon Benfield in Prague, who served as part of the team that created the new flood model for Canada. Rara explained to attendees that hazard perimeter “can vary in fewer metres by metre.”

Released in May, the probabilistic flood model is designed to help insurers better underwrite and manage their riverine and off-flood plain exposures. Covering a geographical area representing 98% of the Canadian population, the model provides a view of flood risk, including individual location level underwriting, portfolio-accumulation management, and structuring reinsurance protections.

Rara said that every catastrophe model has four components: geocoding, hazard, vulnerability and loss calculation. When geocoding is not precise, he noted, the calculation of losses will also not be as precise.

Aon Benfield is encouraging “clients to provide as detailed information as possible,” said Paul Cutbush, senior vice president at Aon Benfield, the reinsurance subsidiary of Aon and the largest licensor of vendor catastrophe models in the world.

“Our goal here was to collect the most detailed data and also use the most local or Canadian data as possible,” Rara said of the new flood model. It has a combined hazard resolution of 10 metres for the most exposed areas and 30 metres for remaining parts of Canada, Aon Benfield reports.

The model has three geocoding options: six-digit postal codes (the most detailed), longitude/latitude and three-digit postal codes. Six-digit postal codes “are very detailed,” Rara said, “and there is almost a million of them in Canada.”

Using downtown Calgary to illustrate, he reported that “six-digit postal codes can be very small, a couple of blocks, sometimes only one block.”

Comparing that to three-digit postal codes, which he described as an emergency option, Rara said the latter “can be pretty big, even in the cities. Outside of those areas, they are much, much larger.”

That means calculation of losses when using three-digit postal codes in the model is not as precise as when using six-digit postal codes, he reported.

U.K.-based Adam Podlaha, international head of Impact Forecasting LLC, a catastrophe model development centre of excellence within Aon Benfield, agreed that clients “should not be running from a three-digit postal code.”

Although there are very few incidents of Aon Benfield still receiving three-digit postal codes, Cutbush told attendees “it’s really clear that you’re not going to have very defined model results if you’re giving three-digit postal codes.” Providing more detailed information is “to your benefit,” he suggested.

Rara reported that, simply put, the more rivers and kilometres of rivers contained in a model, the better the possibility of calculating losses. The new model also has a flood plain component, which he explained can help with calculating losses occurring beyond the riverine flood extent.

There are several reasons why losses can occur outside the flood plain or the riverine flood extent, including such things as the groundwater is elevated or the property is flooded by the sewage back-up system, Rara explained.

“There is a limit of water in nature,” Podlaha said, adding that when modelling, “if we see a trend, we will remove it.” Trends over a short period of time are usually not indicators that something is, in fact, becoming more or less extreme, he said.

Factors such as people living in flood plain areas or changes to the regime of the river are much more important, Podlaha explained. “Most of these things will cover the trend, even if there is one, it will be covered, so it becomes unidentifiable,” he told attendees.

Petr Puncochar, a catastrophe model developer at Aon Benfield in Prague, noted that the model focuses on population density distribution, “which is actually an excellent indicator of exposure.” Doing so helped “us a lot to cut down the massive area of the Canadian country” and put more effort into detailed simulation for areas where exposure was higher, Puncochar suggested.

For Canada, “clearly, 2013 was something that we call an eye-opener event, which actually drove the market towards some changes in the insurance market,” he told attendees of the Toronto roadshow.

Although until very recently “there was no full scope of residential cover available” in the Canadian market, Puncochar suggested it appears there is a place for a new insurance product. However, in order to place such a product on the market “with a sufficient level of confidence, you need to have some kind of integrated approach to your risk [and] to quantify your risk.”

The roadshow – which wrapped up in Toronto after visits to Vancouver and Montreal – was held to outline to clients “some important aspects that you should be considering if you’re going to be incorporating any form of flood technology into your workflow process in terms of writing your business,” Cutbush said.