Abstract Studying the impacts of climate change on tornadoes remains challenging, partly because global climate models (GCMs) struggle to resolve the small‐scale atmospheric phenomena relevant to tornadogenesis. To overcome this problem, this study uses self‐organizing maps (SOMs) to identify synoptic‐scale patterns of atmospheric circulation that are both: (a) important to tornado activity and (b) well‐modeled by GCMs. Artificial neural networks (ANNs) are trained to learn relationships between circulation patterns (CPs) and tornado‐days throughout the US, and then simulated on CPs derived from historical and future GCMs. Based upon the reanalysis, the SOM‐based CPs accurately identify patterns of synoptic‐scale circulation that support (E)F2+ (significant) tornado‐days, capturing the historical frequency, magnitude, and spatial patterns of tornadoes. The historical run of the GCM, however, underpredicts total tornado days, models spatial tornado trends poorly, and projects a later‐than‐observed peak tornado season. In the future, some tornado‐favorable CPs are projected to increase in frequency, while others decrease, overall resulting in an earlier start to tornado season, a broadening of the peak tornado season, and an increase in annual US‐wide tornado‐days by +2.7 events per decade by 2100, (increasing from ∼75 in the 2020s to 94 events/year by 2100). Changes in the spatial distribution of tornado activity are also noted, with a general decrease in parts of the Southern Plains and an increase in the Mississippi River Valley, east of traditional Tornado Alley. Future research should employ multiple GCMs and emissions scenarios to offer policymakers and decision‐makers a broad range of plausible future outcomes.
Lee et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: