Statistics of Hurricane Damage in the US

US damage from hurricanes has remained sensibly constant since 1900, if one corrects for changes in population, wealth, and value of the currency, as have US landfalls. Seasonally aggregated damage and mortality are reasonably well approximated by combinations of log-normal distributions. Despite the absence of a century-scale increase attributable to climate change, damage is significantly correlated with warm ENSO events and the Atlantic Multidecadal Oscillation (AMO).

Present results show that accumulated cyclone energy and TC counts in a US-Waters domain within 500 km of the shore were also constant since 1900, and also exhibited significant correlation with ENSO and the AMO. These observations contrast with reports of increases in TC energy dissipation and numbers in the Atlantic Basin as a whole. Two possible explanations for this apparent discrepancy exist: a shift of Atlantic TC tracks eastward away from US shores or lacunae in the record of historical hurricanes. Downscaled Global Climate Model simulations support the idea that on a warmer globe the lifetime maximum intensities most intense hurricanes (those limited solely by thermodynamics) will increase as numbers of TCs remains constant, decrease slightly, or increase slightly. Examination of the most intense Atlantic Hurricanes since 1974, when geostationary satellite imagery first became available, reveals an abrupt increase in 1995, coincident with a cool-to-warm AMO transition. Other basins also show increases in extreme TCs, but given the relatively short record, the signal is less well defined. The most intense Atlantic hurricanes generally reach maximum intensity in the Caribbean, Gulf of Mexico, or over tropical North Atlantic. A significant fraction of these hurricanes reach US shores when they are 2-3 Saffir- Simpson categories weaker than their maximum.

Approximately 2/3 of US damage occurred during the most damaging 10% of hurricane seasons. The “fat tails” of the cumulative distributions for both seasonally aggregated and individual-hurricane damage are well approximated by Pareto distributions: Pr{d > D} = P_0(D_0/D)^á, where d and D are damage values, D_0 and P_0 are the baseline damage and probability at the lower end of the tail such that Pr{d > D_0} = P_0., and á is the Pareto exponent. The Pareto exponents for seasonally aggregated and individual hurricane damage are 1.37 and 1.14, respectively. If the full range of damage obeys log-normal distributions, the Pareto distributions on the tails derive from Tailor expansions of the complementary cumulative distributions. The variances of Pareto distributions diverge when á < 2, and the means diverge when á < 1, but the underlying log-normal distributions are well behaved. Thus, extrapolation of the Pareto approximation to extreme values may be an artifact of the approximation.

Intensities of TCs are thermodynamically limited, and their size is limited by the requirement that the Rossby number must be significantly larger than unity. Thus, they lack the self-similarity with increasing intensity that characterizes many other geophysical threats. It is reasonable to suppose that the fat tails of the TC damage distributions stem from the distributions of assets at peril. Memorable disasters happen with an intense, but not extreme, hurricane strikes a large concentration of vulnerable property. The Galveston Hurricane of 1900, Great Miami Hurricane of 1926, Katrina of 2005, and Sandy of 2012, are examples.