ENSO: El Nino Southern Oscillation
SST: Sea-surface temperature
With the violent tornadoes that happened Sunday in Alabama, a lot of headlines have been showing climate change may be responsible for the violent tornado outbreak. AccuWeather even put out a very specific tornado forecast for this upcoming tornado season, which has also sparked controversy. Questions such as how well can we predict tornadoes quickly arise. Seasonal tornado forecasts is nothing new. Several papers have been published on the subject using statistical analysis. Scroll down below for a paper I wrote on the correlation between ENSO phases and tornado outbreaks in the United States a few years ago. The latest paper by Cook, A.R., L.M. Leslie, D.B. Parsons, and J.T. Schaefer (2017) strongly reinforces the fact that tornado occurrences across the United States is correlated by ENSO phases. According to Cook et al. (2017), La Nina conditions produce more intense tornado activity across the US. This study focused on tornado outbreaks and ENSO, but I wanted to look at violent tornadoes and ENSO.
Above is the plotted EF5 tornadoes and ENSO SST anomalies from 1950 to 2017. The black line represents the ENSO phase averaged out for that specific year. Negative number being assigned to La Nina phase, zero being a neutral phase and positive number being El Nino. The red line shows the EF5 tornado counts for each year with peaks of 15 in 1974 and 9 in 2011. The blue line depicts the trend line, shows a weak downwards trend in the number of EF5 tornadoes. The tornado data was taken by National Weather Service's Storm Prediction Center and the ENSO data from the Climate Prediction Center.
While a lot can be inferred from the above graph, several patterns are noted. Looking at the two significant EF5 peak activity of 1974 and 2011, a strong El Nino event, followed by a strong La Nina is observed. This is confirmed by the raw data table depicting ENSO phases (see below).
While data is still preliminary, both Figure 1 and Figure 2 depict 2018 being a moderately strong El Nino year (depicted as +0.7, +0.9, +0.8, +0.8 going into 2019). An El Nino year is declared when a minimum of 5 consecutive overlapping seasons are observed (NOAA predicts 55% chance of weak El Nino to continue). Figure 3 below depicts the same as Figure 1, but with added boxes with focus areas. Here you can clearly see a somewhat repeating pattern. We are currently coming out of La Nina (indicated by a downward trend) and in an El Nino. This tornado season could be similar to:
1977 (one F5 in AL, 13 F4's);
1987 (three F4's in MS/TX/WY);
2005 (only one EF4 in Kentucky) and
2007 (one EF5 in Kansas and one F5 in Canada, four EF4's in MS/AL/ND).
Figure 3 is important to note that it is quite incomplete. 2017-2018 spawned a La Nina event and therefore caused a downward trend in our data from the previous 2015-2016 years, which had a strong El Nino. The last box in Figure 3 with the question mark is where the prediction comes in. ENSO data from the end of 2018 going into 2019 shows a positive SST anomaly and therefore the curve should start moving upwards, but the 2019 data isn’t available yet.
While nothing can be concluded from this brief analysis, it is clear that further studies could be conducted for higher-quality trend analysis such as using monthly data instead of yearly data. This data shows us that strong El Nino years followed by strong La Nina years tend to favor more violent tornadoes and tornado outbreaks in the United States. Furthermore, years with negative SST anomalies tend to favor violent tornadoes such as EF5’s, simply based on the graphs. Based on similar years that have generated EF5 and/or EF4 tornadoes in the past, this year should see some EF4 tornadoes (already had one on Sunday) and possibly an EF5 or two. Considering last year there was only one EF4 tornado in Manitoba, Canada across North America, it is safe to say this season should see more violent tornadoes than last year. One thing to note is that the trend line does not depict a correlation between increasing temperatures, (which has been shown by other research) and an increasing amount of EF5 tornadoes.
Cook, A.R., L.M. Leslie, D.B. Parsons, and J.T. Schaefer, 2017: The Impact of El Niño–Southern Oscillation (ENSO) on Winter and Early Spring U.S. Tornado Outbreaks. J. Appl. Meteor. Climatol., 56, 2455–2478, https://doi.org/10.1175/JAMC-D-16-0249.1
ENSO Effects On Tornado Outbreaks and Violent Tornadoes
by: Francis Lavigne-Theriault
The everlasting debate on whether the El Niño Southern Oscillation (ENSO) effects tornadoes in the United States is a hard question to answer. Relatively poor instrumental and observational record and high variability of tornado reports each year make it even harder to identify a correlation between ENSO and tornadic activity in the United States (Allen, Tippett & Sobel, 2015). However, a general trend is found when studying older and newer data. El Niño years are found to have less violent tornadoes as well as less tornado outbreaks, while La Niña years are found to have more violent tornadoes as well as having a higher probability of tornado outbreaks (Knowles & Pielke, 1993). This paper will focus on tornado outbreaks occurring in the United States during both warm (El Niño) and cold (La Niña) ENSO phases. The probability of tornado outbreak occurrences, their intensity and when they are most likely to occur is also discussed. Identification of the cause/consequence relationship of changes in sea surface temperature (SST) and its modification of weather and climate throughout the continental United States will be discussed. These findings are presented as a possible long-range seasonal prediction method to severe thunderstorms (Allen et al., 2015).
2.1. ENSO-outbreak classification for cold-season tornado outbreaks
In order to classify their results, Nunn & DeGaetanno (1998) established a basis for ENSO-outbreak relationships. Outbreaks were classified on a regional scale such as Deep South (Texas, Louisiana, Mississippi, Alabama, Georgia and Florida), Mississippi Valley (Wisconsin, Iowa, Missouri, Arkansas, Oklahoma and Kansas) and the Ohio Valley (Tennessee, Kentucky, Ohio, Indiana and Illinois). The regions were separated based on the similarities between climates. After obtaining the total amount of cold-season outbreaks that occurred in each region such as 90, 47 and 43 respectively, the next step was to classify the events by ENSO phases such as El Niño, La Niña and Neutral. After being classified, chi-squared tests were performed for each region. Expected results (numbers) were put in brackets, while actual results were also described and classified in two categories such as above mean and at or below mean of the number of tornado outbreaks for each region regardless of ENSO phase. Chi-squared tests were then run based on ENSO phase. X2 values are seen to be a function of phase.
2.1. Tornado outbreaks by ENSO phase
Knowles & Pielke’s (1993) study included data collected from tornado reports between 1953 and 1989 where 1953 was selected as a starting point for the study since this was the first year the weather bureau started issuing tornado watches. Tornadoes occurring during downburst thunderstorms, squall lines, towering cumulus and the supercell were included in the study. The seven strongest El Niño and five strongest La Niña events were compared and classified into categories such as total number of tornadoes per year, median tornado track length in miles for violent tornadoes and number of violent tornadoes per year. A comparison of all El Niño and La Niña years is also included. 14 tornado outbreaks between 1953 and 1989 were also classified by strongest and all El Niño/La Niña phases, whereas a tornado outbreak is defined by 39 or more tornadoes during one event.
3.1. Defining the El Niño Southern Oscillation (ENSO)
According to Allen et al. (2015), ENSO is characterized by changes in sea surface temperatures (SST) and atmospheric convection in the tropical Pacific. ENSO phases regulates global weather patterns and climate. Warmer than average Pacific sea surface temperatures is called the El Niño phase. Conversely, colder than average Pacific sea surface temperatures is called La Niña. ENSO influences precipitation and temperatures across the continental United States and Canada (Allen et al., 2015). El Niño usually brings colder than average temperatures to the Southeast United States and warmer than average temperatures to western United States as well as higher rainfall to both regions. Conversely, La Niña usually brings warmer than average temperatures for the Southeast and colder than average temperatures to western United States (Barnston, Livezey & Halpert, 1991). The oscillation between the warm and cold phases are referred to as the Southern Oscillation (Knowles & Pielke, 1993).
3.2. ENSO effects on continental weather and climate
Studies of sea surface temperature changes and the thermal response of the atmosphere suggest a “lag time” between SST change and atmospheric changes. According to Knowles & Pielke (1993), a lag of about three to five months is generally observed between the maximum SST in the Pacific and the atmosphere in the continental United States. Rasmusson et al. (1982) further adds that since the usual SST maxima occurs from January through June in the Pacific, the effects of El Niño and La Niña would therefore be felt from March through November in the United States. This time frame coincides with the most active period for tornadic activity in the United States, which is from April to July.
3.3. Defining modification of environmental factors during ENSO
ENSO modification of extratropical cyclones, precipitation, jet stream position and strength, surface temperatures and low-level moisture advection from the Gulf of Mexico can influence the environmental factors needed for tornadogenesis (Allen et al., 2015). Below are environmental factors, which are described to be less favorable for cyclogenesis and tornadoes during El Niño and more favorable during La Niña.
3.2.1. El Niño modifications
According to (Allen et al., 2015), El Niño years favor less low pressure systems over the Plains while increasing low pressure system development and impacts in the southeast United States. Surface winds and warm moist air convergence from the Gulf of Mexico are weakened during El Niño years, which in turn allows cold arctic air to surge further south due to the southward shift of cyclogenesis. Significant decrease in moisture advection therefore decreases atmospheric thermodynamic energy needed for severe thunderstorms in the southern Plains. A more specific decrease in mixed-layer convective available potential energy (MLCAPE) and 0-3km storm relative helicity (SRH) is noted during El Niño years, both of which are key ingredients in tornadogenesis. Surface temperatures in the southern United States are cooler than -1°C and warmer further north, which opposes climatological temperature gradients (north-to-south) reducing probabilities of cyclogenesis east of the Rockies. Reduction of resistance to convection also reduces vertical instability. El Niño favors strong/deep ridges and troughs across the central United States, which in turns favors cooler than average temperatures for the southern US and warmer than average temperatures in the Pacific Northwest (Nunn & DeGaetanno, 1998).
3.2.1. La Niña modifications
According to Allen et al. (2015), lesser jet stream flow modification during La Niña favor the development of high pressure systems over the southwest due to stronger flow above the continent and reduced flow further south. La Niña favors greater northward moisture advection from the Gulf of Mexico, especially over eastern Texas and the southern United States, therefore increasing thermodynamic energy needed for severe thunderstorms over the continent. Increased southeasterly surface flow results in larger 0-3km storm relative helicity (SRH), giving way to a more favorable environment for severe thunderstorm development and tornadoes. La Niña surface temperature increase of greater than +1°C enhances the climatological north-to-south temperature gradient, which favors cyclogenesis. This increased surface temperature in turn increases MLCAPE and makes for steeper lapse rate, potentially increasing the resistance to convection. This in turn gives for a more favorable environment for tornadogenesis.
3.4.ENSO and tornado outbreaks
Grazulis (1993) defines a tornado outbreak as a series of tornadoes from the same storm system that occur with no more than six hours between each event. According to Nunn & DeGaetanno (1998), there were 180 major tornado outbreaks between 1950 and 1995 during the “cold season” in the Deep South (90 outbreaks), in the Mississippi Valley (47 outbreaks) and the Ohio Valley (43 outbreaks). Only outbreaks with F2 to F5 tornadoes were utilized in this specific study. Table 1 divides the aforementioned cold-season outbreaks based on their ENSO phase. Counting the number of seasons from 1950 to 1995, 15 were found to be El Niño, 8 La Niña and 22 Neutral. Each region’s outbreak was then separated by the phase of each season.
Furthermore, statistical analysis of below, normal and above average tornado outbreaks per ENSO phase per year was calculated. Table 2 shows X2 values as a function of phase. All regions indicating an X2 less than one indicates a “normal” tornado outbreak season while X2 values of 5.07 indicates an above-normal tornado outbreak season. For all regions, the La Niña phase shows a strong correlation between a cold ENSO phase and an increase in tornado outbreaks while El Niño and Neutral remains around “normal” with the exception of the Ohio Valley, which sees an impressive increase in tornado outbreaks during Neutral phases.
In addition, Knowles & Pielke (1993), an average of 750 tornadoes per year occurred between 1953 and 1989, most of which occurred between March and August. However, one must count in the bias of poor reporting methods during this time period and the general “increase” in tornado reports from that time until today. Table 3 depicts the overall tornado summary classified by ENSO phases from strong El Niño to strong La Niña as well as all El Niño and La Niña events between 1953-1989. Tornado outbreaks are also depicted. Tornado outbreaks are hereby defined as 39 tornadoes or more occurring during one event.
As noted by Nunn & DeGaetanno (1998), El Niño’s effect on the polar and subtropical jet and the increase in convective activity offshore of Western/Central America leading to the formation of upper-level lows lead to tighter pressure gradients between Central American lows and strong Alaskan high pressures that tend to form during a northward retreating polar jet with a sharp baroclinic zone between the warm tropical Pacific waters and the cool subtropical waters, which were found to increase the speed of the subtropical jet. This, combined with increased moisture transport in the southern United States and northern Mexico has been found to increase storm intensity for northern Mexico and southern California and lead to wetter than normal conditions for the Gulf States. During La Niña years, the polar jet becomes stronger while a generally weaker subtropical jet is observed. The polar jet therefore dominates the synoptic pattern in North America. Warmer and drier conditions are observed in the southern US. Increased baroclinity between warm and cold air masses favor an increase in both strength and numbers of synoptic-scale development.
According to Knowles & Pielke (1993), the year-to-year average of all tornado counts are generally constant, but when studying the violent tornadoes and tornado outbreak criteria, the numbers become statistically significant. Violent tornadoes (F4-F5) and tornado outbreaks were much lower during El Niño years and much higher than “normal” during La Niña years between 1953 and 1989. El Niño years average around 7.8 violent tornadoes while La Niña years average around 12.6 violent tornadoes. During the 14 tornado outbreaks of 39 tornadoes or more, zero occurred during strong El Niño years and three have occurred during strong La Niña years. Out of the 14 outbreaks, 11 occurred during La Niña events while only three occurred during El Niño events regardless of strength of the ENSO phase.
The method for long-range tornado prediction consists of analyzing winter ENSO phases and creating a probabilistic forecast for the March to May severe thunderstorm season. According to Allen et al. (2015), there is a clear statistical correlation between March to May tornadoes associated with moderate to strong December to February ENSO phases. This information is then used to create a forecast for tornadoes based on past similar ENSO conditions. This suggests that for moderate to strong ENSO phases, above average or below average tornado activity across the United States can be predictable based on the ENSO pattern into the following spring.
Statistical evidence seems to support the fact that ENSO cold phases (La Niña) contribute to environmental conditions that are most favorable for violent tornadoes (F4-F5) as well as significant tornado outbreaks in the United States, while warm (El Niño) and Neutral ENSO phases seem less favorable for violent tornadoes as well as tornado outbreaks besides some exceptions such as in the Ohio Valley during Neutral phases. Better environmental ingredients conductive to tornado outbreak evolution exist during La Niña events. Environmental conditions such as the weakening of the subtropical jet and domination of polar jet favoring stronger cyclonic development, stronger pressure gradients, better vertical thermodynamic energy such as MLCAPE and 0-3km storm relative helicity (SRH) and better northward moisture transport from the Gulf of Mexico are most favorable for tornado outbreaks during La Niña phases. Statistical analysis of both cold-season and spring-season tornado outbreaks found that La Niña years favor more violent tornadoes as well as more significant tornado outbreaks of 39 or more tornadoes. Aforementioned causes of sea surface temperature (SST) modifications and its consequential effects on continental atmospheric conditions in the United States can lead to long-range predictions for severe thunderstorms and tornadoes as proposed by Allen et al. (2015). More research would need to be done on ENSO phases and effects on tornadic activity for one to provide any solid conclusions. Our reporting methods and technological advancements in the pass 20 years have drastically improved our understanding of tornadoes and ENSO phases and new research such as the one by Allen et al. in 2015 would need to be repeated. Research focusing on spring tornadoes such as the one currently being done by the VORTEX projects in the United States can provide more insights as to how SST changes affect the peak of the tornado season as well as cold-season tornado intensity and outbreaks. I would have liked to also see more research done on ENSO and tornadoes in the northern Plains and Canada. I found almost no research being conducted on the subject in these areas. This would be an interesting subject of future research.
Grazulis, T. P. (1993). Significant Tornadoes 1680-1991. Tornado Project of Environmental Films.
Knowles, J. B. & Pielke Sr, R. A. (1993). The Southern Oscillation and Its Effect On Tornadic Activity in the United States. Colorado State University: Department of Atmospheric Science, Paper No. 755.
Nunn, K. H. & DeGaetano, A. T. (1998). The El Nino Southern Oscillation and Its Role in Cold-Season Tornado Outbreak Climatology. Cornell University, Ithaca, NY, USA. Paper presented to the American Meteorological Society, 1998.
Rasmusson, E. M. & Carpenter, T. H. (1982). Variations in Tropical Sea Surface Temperature and Surface Wind Fields Associated with the Southern Oscillation El Niño. Monthly Weather Review., 110, 354-384