Influence of the El Niño/Southern Oscillation on tornado and hail frequency in the United States

The El Niño-Southern Oscillation (ENSO) is characterized by changes in sea surface temperature (SST) and atmospheric convection in the tropical Pacific, and modulates global weather and climate1–4. The phase of ENSO influences United States (US) temperature and precipitation and has long been hypothesized to influence severe thunderstorm occurrence over the US5–11. However, limitations12 of the severe thunderstorm observational record, combined with large year-to-year variability12,13, have made it di cult to demonstrate an ENSO influence during the peak spring season. Here we use environmental indices14–16 that are correlated with tornado and hail activity, and show that ENSO modulates tornado and hail occurrence during the winter and spring by altering the large-scale environment. We show that fewer tornadoes and hail events occur over the central US during El Niño and conversely more occur during La Niña conditions. Moreover, winter ENSO conditions often persist into early spring, and consequently the winter ENSO state can be used to predict changes in tornado and hail frequency during the following spring. Combined with our current ability to predict ENSO several months in advance17, our findings provide a basis for long-range seasonal prediction of severe thunderstorm activity. ENSO’s modification of the upstream mid-latitude flow over the Pacific contributes significantly to year-to-year variability of US extratropical cyclones, precipitation, surface temperatures and low-level moisture advection from the Gulf of Mexico1–4. These perturbations suggest that ENSO should influence the large-scale environmental features that produce severe thunderstorms6–11, in a similar manner to the Madden–Julian Oscillation18. On average, severe thunderstorms producing tornadoes, hail and damaging winds cause US$ 1.6 billion of damage per year across the US, and billion dollar events are not uncommon19. Tornado occurrence has previously been related to ENSO phase during the winter months and over small geographical regions6–11. There has been less indication of a relation during the spring or summer months, when the frequency of severe thunderstorms peaks8,9. A recent study related Pacific SSTs and tornadoes using the Trans-Nino Index (TNI; ref. 11), showing a significant correlation to detrended tornado frequency, which had only a weak relationship to other ENSO indices. However, the strength of the relationship to intense tornadoes was not statistically robust, with results weakened by choices of sample stratification, detrending or the influence of outliers. No attempt has thus far been made to show a relationship betweenENSOandhail occurrence.Herewe address this issue using data that are less problematic than the tornado and hail reports themselves, namely proxies derived using environmental data20. Environmental indices are used to relate favourable large-scale atmospheric conditions (such as strong vertical wind shear and thermodynamic potential energy) to severe thunderstorm occurrence for both short-range prediction and climate projection20–24. Here we employ environmental indices for tornado and hail occurrence that have been shown to skillfully represent climatological and interannual variability14–16 (seeMethods). For comparison to the indices, we use the number of hail events in excess of one inch in diameter and the total number of observed tornadoes in a 1× 1 grid box over a season. Seasonal values of the indices, observations and environmental conditions were derived for two seasons: December, January, February (DJF) and March, April, May (MAM). Correlations with observations for National Oceanic and Atmospheric Administration (NOAA) climate regions25, together withmean spatial maps for DJF and MAM (Supplementary Table 1 and Figs 1–3) demonstrate the suitability of the indices to describe the climatology, seasonal cycle and interannual variability of tornado and hail occurrence. Seasonal composites were created for the La Niña and El Niño phases of ENSO, selecting years of moderate to strong signal26 where values of seasonal mean Oceanic Nino Index (ONI) have magnitude greater than 1 for DJF and 0.5 for MAM. Composite DJF anomalies show an asymmetric ENSO signal, with increases of severe weather over the central and southeast US during La Niña and over Florida during El Niño (Fig. 1). For hail, the influence corresponds to a significant increase of 0.5 hail events per season per grid cell over Florida and the Gulf coast during El Niño years, and a small decrease over the southeast US (Fig. 1a,e). This change is large relative to the climatological frequency of 0.25 hail events or less per season. In La Niña years, significant hail increases are identified for the plains to southeast where frequency is low, whereas frequencies over Florida are smaller. For tornadoes, the pattern is similar (Fig. 1b,f), with significant seasonal increases of 0.5 events per grid cell along the Gulf Coast and Florida during El Niño and a northward-displaced decrease in frequency over the southeast and lower Midwest. There is an effective doubling of frequency in this region during El Niño years. During La Niña conditions, significant increases to tornado events are found from the plains eastward (Fig. 1d,h). Significant decreases are also found in tornado frequency for Florida, where ENSO-induced frequency reduction is close to the climatological frequency, implying a total frequency close to zero. The climatological frequencies of both hail and tornadoes are higher in MAM and are reflected in MAM ENSO composites with large amplitudes, despite weakening of the ENSO signal during this period7,8. The largest ENSO composite values are seen over the southern Plains states and the southeast, again with asymmetry between the two phases (Fig. 2). The ENSO signal shifts west relative