Medium-term hurricane rates

Medium-term hurricane rates are hurricane frequencies set for the next five years rather than the average observed since the early twentieth century. They reweight a catastrophe model's stochastic catalog without changing the events themselves. The long-term catalog, by contrast, rests on the historical Atlantic hurricane database (HURDAT) going back to 1900. It all goes back to 2005: Katrina, Rita and Wilma hit the United States within weeks of each other, a year after Florida's four hurricanes. In 2006 RMS made a medium-term view, drawn from an expert elicitation, the default setting of its North Atlantic hurricane model. AIR answered with a catalog conditioned on warm sea surface temperatures. The physics is simple enough. The Atlantic runs through warm and cold phases lasting decades (the Atlantic Multidecadal Oscillation, AMO), and a warm phase produces more major hurricanes. So the modeler raises the annual landfall rate by category and by region, then reruns the exceedance curves: average loss goes up, and the return period of any given loss gets shorter. Many reinsurers took this view as their reference. It never won everyone over. The Florida commission that certifies models used in residential ratemaking sticks to long-term historical frequencies. And no major hurricane made US landfall between Wilma (2005) and Harvey (2017), which fed the suspicion that the expert panel had read too much into two extreme seasons. When two PMLs (probable maximum loss) disagree, the first question is which catalog each one runs on: the rate gap often outweighs the model gap.