Brooks et al. (2003b) used data from the NCAR/NCEP reanalysis to create pseudo-radiosoundings and then used them to develop relationships between reanalysis environments and severe thunderstorm events. Using reanalysis data from points over land from around the world, Brooks et al. (2003b) estimated the distribution of severe thunderstorms around the world. Regions east of elevated terrain and poleward of sources of lower-tropospheric moisture were particularly favored for severe thunderstorms. The central part of the United States was the region with the most frequent number of days with favourable severe thunderstorm environments.
Global climate models are unable to simulate severe thunderstorms and tornados because their spatial resolution is too coarse to be able to simulate such mesoscale events. However, they should be able to simulate the environmental conditions under which such severe weather develops, including abundant lower tropospheric moisture, steep mid-tropospheric lapse rates, and strong tropospheric wind shear. High space and time resolution data from control simulations with global climate models archived at NCAR is being used to estimate the frequency of favorable conditions for severe weather, as simulated by the models. The basic methodology follows Brooks et al. (2003b). Soundings are created from the temperature, wind and specific humidity data at each of the model grid points. Two convective parameters, CAPE and shear over various layers of the troposphere, which have previously been shown to be relevant to identifying severe thunderstorm environments and for discriminating between tornadic and non-tornadic severe thunderstorms, are derived from those soundings.
Analysis of the distribution of parameters is carried out in a variety of ways on different scales, including selected points and regions. Parameters are investigated both in single and multiple dimensions. Since there are no observed severe weather events to associate with individual soundings from the model output, the study is limited to the distribution of the parameters. Comparisons between the distribution seen in the reanalysis and in the control runs of the GCMs will be shown.