CLIMATOLOGICAL RELEVANCE OF MAJOR USA WEATHER LOSSES DURING 1991-1994

Two insurance indices, which integrate storm types with their losses f rom 1948 to 1994, were examined to assess the frequency and intensity of damaging weather in the USA from 1991 to 1994. These indices adjust for changing socioeconomic conditions, allowing meaningful temporal a ssessment of 1991-1994 conditions against those in earlier years. Cata strophes, as defined by property loss insurers, were assessed for two groups, those 707 storms with insured losses between $10 and $100 mill ion and those 189 with greater than $100 million. Both categories expe rienced their greatest frequency and caused more loss during the 1991- 1994 period than in any prior period. However, the temporal behaviour of annual frequency and losses for both categories were very different , with steady increases found for the $10 to $100 million catastrophes , while the number of >$100 million catastrophes exhibited a high-low- high distribution and the annual cost distribution was flat with a few isolated peaks. For both categories the temporal behaviour of storm i ntensity was similar, with moderate intensities found in the recent pe riod and the highest intensities in the 1950s. The crop-hail losses we re high in 1992-1994, and the annual loss costs for the 1948-1994 peri od had a U-shaped time distribution similar to that of the > $100 mill ion catastrophes. Weather conditions (USA mean temperature and surface cyclone frequency) were found to largely define: (i) when major, high ly expensive weather events occur (and their annual losses), and (ii) how intense catastrophic storm conditions and crop-damaging hailstorms are during each year. Conversely, the shifting target (population as an index) for damaging (non-hurricane) storms across the USA is the ma jor factor in 'creating' storms with losses reaching into the low ($10 million) to moderate ($100 million) range. (C) 1998 Royal Meteorologi cal Society.