A database containing approximately 4000 operational 0000 UTC soundings that were taken from 1957-1993 in the vicinity of thunderstorms that produced significant severe weather (hail at least 2 inches in diameter, winds of at least 65 kts, and/or an F2 or greater tornado) has been developed. The criteria for proximity are a distance of 100 nm from the sounding location and occurrence from 2100-0300 UTC. Minimal automated quality control involving convective available potential energy and convective inhibition was applied. Over 1000 soundings are associated with tornadoes.

With this large dataset, we have developed conditional probabilities of these events occurring given that significant severe weather occurred focusing on the period, 1973-1993. In particular, we have examined the conditional probabilities for combinations of the height of the mean 100-mb layer lifted condensation level (MLLCL) and wind shear in the lowest 1 km above ground level. The primary finding is that the probability of a tornado occurring increases dramatically as one goes from a high LCL/weak shear environment (~10%) to a low LCL/strong shear environment (~80%), for combinations of LCL and shear that include hundreds of soundings. Given the climatological probability in this dataset of about 28%, the range of values indicates promise as a forecasting tool.

At a more basic level, the dataset allows us to look at changes in the historical record. Using results from a companion study of all 0000 UTC soundings from 1997-1999, we considered the historical performance of the 1997-1999 “best” discriminator between significant tornadic and significant non-tornadic soundings. The performance of the discriminator is relatively consistent from 1973-1993, but a dramatic and sudden fall-off in performance is seen prior to that. The change in performance is associated with a much higher fraction of soundings associated with tornadoes in the part of the parameter space associated with low tornado probability in the years 1973-1993. This implies that either the atmospheric conditions in which strong thunderstorms form changed in 1973 or the reporting changed. Given that significant changes in the reporting system occurred in 1973, it seems that the latter is more likely to be true.