Predictions of Severe Weather Environments by the Climate Forecast System Version 2 Model Suite

A desire to improve understanding of subseasonal predictions of atmospheric conditions known to be directly correlated to the onset and maintenance of severe thunderstorms is the ongoing motivation for this research.  High interest in extended-range predictions of severe weather continues to provide the catalyst for this assessment of multiple facets of the NCEP Climate Forecast System Version 2 (CFSv2) suite of products.  In particular, predictions from 29 years of Climate Forecast System Reforecast (CFSRR) output have been scrutinized for the spring (AMJ) months, and verified against Climate Forecast System Reanalysis (CFSR) data, with a climatological basis generated from the long-term mean (32-year) of the CFSR.

Our particular focus remains on the predictability of convective environments through the analysis of parameters with well-established correlations to severe weather – specifically convective available potential energy (CAPE) and deep layer vertical wind shear (VWS).  Although inherently necessary for severe thunderstorms to occur, other catalysts, such as outflow boundaries and drylines, dictate whether or not convection initiation occurs.  However, the non-negligible relationship between severe weather and CAPE / VWS can be used to distinguish an atmosphere supportive of rotating storms from one supportive of mainly non-severe convection

Our previous results from operational CFSv2 analyses indicated potential value above and beyond a climatologically-based prediction for specified sub-regions of the central and eastern United States at leads as long as multiple weeks, based upon trends in root-mean-squared difference (RMSD) and Spearman rank correlation coefficients.  A series of methodologies previously utilized for the operational CFSv2 will be applied to the CFSRR output, and will be supplemented by additional techniques incorporating empirical cumulative distribution functions (CDFs), lagged average ensemble forecasting techniques, and complimentary procedures for determining model skill, all with a concerted effort to address inherent biases in the CFSv2 output of the aforementioned parameters.  Furthermore, periods in which the CFSv2 exhibits noteworthy skill are further scrutinized to determine potential connections to the large scale pattern, based upon known skill of the model to predict features associated with El Nino-Southern Oscillation (ENSO) and the Madden Julian Oscillation (MJO) through established teleconnections.