An Assessment of the Subseasonal Predictability of Severe Thunderstorm Environments and Activity using the Climate Forecast System Version 2
The prospect for skillful long-term predictions of atmospheric conditions known to directly contribute to the onset and maintenance of severe convective storms remains unclear. A thorough assessment of the capability for a global climate model such as the Climate Forecast System Version 2 (CFSv2) to skillfully represent parameters related to severe weather has the potential to significantly improve medium- to long-range outlooks vital to risk managers. Environmental convective available potential energy (CAPE) and deep-layer vertical wind shear (DLS) can be used to distinguish an atmosphere conducive to severe storms from one supportive of primarily non-severe ‘ordinary’ convection. As such, this research concentrates on the predictability of CAPE, DLS, and a product of the two parameters (CAPEDLS) by the CFSv2 with a specific focus on the subseasonal timescale. Individual month-long verification periods from the Climate Forecast System reanalysis (CFSR) dataset are measured against a climatological standard using cumulative distribution function (CDF) and area-under-the-CDF (AUCDF) techniques designed mitigate inherent model biases while concurrently assessing the entire distribution of a given parameter in lieu of a threshold-based approach. Similar methods imposed upon the CFS reforecast (CFSRef) and operational CFSv2 allow for comparisons elucidating both spatial and temporal trends in skill using correlation coefficients, proportion correct metrics, Heidke skill score (HSS), and root-mean-square-error (RMSE) statistics. Key results show the CFSv2-based output often demonstrates skill beyond a climatologically-based threshold when the forecast is notably anomalous from the 29-year (1982-2010) mean CFSRef prediction (exceeding one standard deviation at grid point level). CFSRef analysis indicates enhanced skill during the months of April and June (relative to May) and for predictions of DLS. Furthermore, years exhibiting skill in terms of RMSE are shown to possess certain correlations with El Niño-Southern Oscillation conditions from the preceding winter and concurrent Madden Julian Oscillation activity. Applying results gleaned from the CFSRef analysis to the operational CFSv2 (2011-16) indicates predictive skill can be increased by isolating forecasts meeting multiple parameter-based relationships.
Baldwin, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our