Toward seamless prediction of severe weather activity

Severe weather indices are functions of atmospheric parameters chosen to identify environments favorable to phenomena such as severe thunderstorms, hail and tornadoes. Typical ingredients in such indices include convective available potential energy and vertical wind shear. These indices are routinely applied to short-range weather forecasts to gauge the potential for severe weather events. Recently we have shown that a tornado index computed from observation-based monthly-averaged parameters from the North American Regional Reanalysis is able to capture aspects of monthly climatological and interannual variability present in the US tornado report database. Additionally we have found that when the index is computed with monthly-averaged forecast environmental parameters from the CFSv2, the index presents some predictive skill on continental and regional scales up to one month in advance. Our previous work has been limited to calendar month averaged quantities and was not able to address intermediate (a few days to a few weeks) lead-times and averaging-windows. In particular, the predictability of severe weather activity versus that of particular events has not been investigated. Here we examine those questions as well as verification issues using 6-hourly CFSv2 forecasts with lead-times of 0-45 days, focusing on forecasts and reports of severe weather during 2013