Cat models must be better at recognizing uncertainty

Catastrophe modelers must do a better job of expecting the unexpected, according to speakers at a European conference, as reported by A.M. Best.
I think we need to move to this new age the Age of Recognized Uncertainty, A.M. Best quotes Peter Taylor, a project director at Lighthill Risk Network, as saying.
He was speaking at the conference, “Catastrophe Modeling from a European Perspective,” sponsored by the London-based International Underwriting Association and the Reinsurance Association of America.
Even for the areas that we know, were not very good at looking just outside our range of vision, Taylor said. Taylors Lighthill Risk Network operates within University College London. It is described as a bridge between academic theory and industry practice.
Taylor noted better co-ordination and quality of information from brokers, insurers and risk managers might help identify catastrophe risks that lie just beyond the vision of the models.
For example, Taylor said, modelers need current information to help them understand that a hotel on a coast might face a higher storm risk than one inland.
Modelers have also been weak at assessing collateral factors such as storm surge or event timing, A.M. Best cites Taylor as saying. As an example of event timing, Taylor presented the following example: What would happen if a quake occurred at a time of day when there were lots of people in their houses with the gas on?
Taylor suggested there should be more research into the possibility of event clustering (a close proximity between seemingly unrelated disasters). Examples would include a consideration of a series of earthquakes of the kind that destroyed San Francisco in 1906, or a scenario in which a major earthquake occurs one month after a terrorist event on the order of magnitude of 9-11.
Modelers might also try to get a better sense of new areas of risk, such as tsunamis. While tsunamis are rare, A,M. Best paraphrases Taylor as saying, they can be severe.