Probabilistic Hurricane Forecasts for Risk Management
Thomas Nehrkorn and Ross N. Hoffman firstname.lastname@example.org email@example.com
Atmospheric and Environmental Research, Inc. 131 Hartwell Ave Lexington Massachusetts 02421-3126 www.aer.com
The US economy is strongly affected by hurricanes making landfall in the US, both directly (e.g., claims paid by insurance companies) and indirectly (e.g., financial performance of companies whose operations are disrupted by hurricanes). Companies seeking to manage their exposure to hurricane-related risks require forecasts of hurricane metrics tailored to their specific needs. We present a method of providing probabilistic forecasts of such a metric from a diverse set of operational forecast products.
The uncertainties conveyed as part of the official hurricane forecasts incorporate the spread in the hurricane forecast products from a number of different operational approaches. However, the various graphical and numerical products issued by the Tropical Prediction Center cannot be directly related to the metrics of hurricane risk of interest to commercial clients. We therefore present a method to combine operational hurricane forecasts from a number of different sources into probabilistic forecasts of specially tailored risk metrics. Individual forecasts are first adjusted by removing known biases, and weighted according to their known error magnitudes using Bayesian Model Averaging. Uncertainties in the predicted characteristics of hurricanes (such as position, strength, size, etc.) of the forecasted hurricanes are combined to provide the overall uncertainty in the customized index.