Severe convection is responsible for a significant fraction of weather-related loss of life and property in extratropical latitudes (Pielke and Klein 2001; Fritsch and Carbone 2004). The most important precondition for convection is an unstable environment, usually associated with tropospheric cooling as a result of ascent driven by the synoptic flow (e.g. Roberts 2000). This indicates that an accurate forecast of the large-scale weather systems can provide valuable information about the occurrence of convection, however, the convective systems themselves are typically tens of kilometres in size with lifetimes measured in hours . A precise forecast at that scale would require observations at impracticably high density and even then would lose accuracy in less than a day. It is generally conceded that convection is forecast badly at present, but the limits of predictability for convection are not known (Ebert et al. 2003). Significant gains have been obtained in recent years from high resolution forecasting systems, but these will be ultimately limited by chaotic nature of the storms on relatively short timescales (Walser et al. 2004). Ultimately, the most valuable forecast is likely to be probabilistic, taking account of the uncertainties in both the large and small scale processes, but progress in designing an optimal forecasting system is limited by a lack of understanding of the roles and interactions of the various scales of atmospheric motion involved in the initiation of a convective event (Fritsch and Carbone 2004).
Severe convection on 11 June 2007 during the COPS (Convective and Orographically-induced Precipitation Study) experiment. Left: A thunderstorm in its development stage. Right: The approach from the right of a severe thunderstorm with heavy precipitation.