Interannual Predictions of Landfalling Atlantic Hurricanes

RMS produces software that a large part of the insurance industry uses to set insurance rates in hurricane-prone regions in the US. An important input to our product is the expected number of annually averaged hurricanes making landfall over the next five years. Another important input is the windfield associated with future events. Our approach to the generation of windfields is the use of numerical modeling which we will discuss briefly, while our approach to the hurricane number prediction has two steps. First, we build a large number of models that make this prediction, based on inputs such as historical hurricane numbers in the Atlantic basin and at landfall, historical SSTs, and IPCC climate model output. Second, we ask a group of hurricane experts to weigh the predictions from these models. In this presentation, we discuss some of the models in the suite that has been developed, and welcome comments as to both the validity of the individual models, and as to what additional models we might include. We also very briefly present the results of the expert elicitation that was carried out in 2007. The principal result from that elicitation is that the expected number of hurricanes making landfall in the US is significantly higher than the long-term historical average since 1900.