Predictability of US tornado outbreak seasons using ENSO and northern hemisphere geopotential height variability

摘要

The predictability of dangerous atmospheric phenomena such as tornado outbreaks has generally been limited to a week or less. However, recent work has demonstrated the importance of the Rossby wavetrain phasing over the United States in establishing outbreak-favorable environments. The predictability of Rossby wavetrain phasing is strongly related to numerous climate-scale interannual variability indices, which are predictable many months in advance. To formalize the relationship between interannual variability indices and seasonal tornado outbreak frequency, indices derived from monthly mean Northern Hemisphere 500-hPa and 1000-hPa geopotential height fields and Ni?o 3.4 indices for ENSO phase were compared to annual tornado outbreak seasonal frequencies. Statistical models predicting seasonal outbreak frequency were established using linear (stepwise multivariate linear regression–SMLR) and nonlinear (support vector regression–SVR) statistical modeling techniques. The stepwise methodology revealed predictors that are important in establishing outbreak-favorable environments at long lead times. Additionally, the results of the statistical modeling revealed that the nonlinear SVR technique reduced root mean square errors produced by the control SMLR technique by 28% and provided more consistent forecasts. A preliminary physical analysis revealed that years with high outbreak frequencies were associated with the presence of 500-mb troughs over the central and western US during the peak of outbreak season, while lower frequencies were consistent with ridging over the US or northwest flow over the Plains. These patterns support the results of the statistical modeling, which demonstrate the utility of geopotential height variability as a predictability measure of outbreak frequency. Graphical abstract Display Omitted Highlights ? Identified statistical relationships between tornado outbreaks and interannual climate variability. ? Provided speculative physical reasoning for the relationships between outbreaks and climate indices. ? Presents predictive model of annual outbreak frequency.