The historical hurricane damage database of Pielke et al. (2008) is analyzed using a Peaks-Over-Threshold approach, in which the Generalized Pareto Distribution is applied to model excesses above a specified threshold for a given damage metric. In addition to the dataset of absolute hurricane damages given in Pielke et al (2008), we define a Damage Index as the ratio of base-year economic damages to the “available” economic value in the affected region. We then incorporate physical covariates at the individual hurricane level into the GPD model, namely maximum wind speed and a simple yet novel measure of the mean bathymetric slope at landfall, and apply our analysis to both the Total Damage and the Damage Index for the purposes of direct comparison.

The parameters of the GPD models with physical covariates are optimized with maximum likelihood estimation. We find that for Total Damage the only useful covariate is maximum wind speed. Meanwhile, for Damage Index both the mean slope and maximum wind speed are found useful, with coefficients that are consistent with the known physics of each covariate in causing damage. These results suggest that damage measured as a fraction of potential damage helps remove this economic signal from the damage database, leaving a dataset that may be better representative of the physical relationship between hurricanes and damage.

Finally, this new methodology is applied to two datasets of approximately 5000 synthetic tracks, each simulated within current and future A1B-scenario climates as modeled by the GFDL CM2.0 model. We demonstrate its potential utility in assessing changes in the upper tail of the distribution of damaging hurricanes as measured by the Damage Index.