Reduced Arctic Sea Ice May Be Suppressing July Tornadoes

A notable drop since the 1990s in the number of July days with significant U.S. tornadoes may be related to the dramatic loss of midsummer Arctic sea ice, according to a study published this month in the open-access journal npj Climate and Atmospheric Science. The new work may help explain a more general drop in U.S. tornado days, part of a recent clustering of tornado activity into highly active periods interspersed with distinct quiet spells.

The new study, by Robert Trapp (University of Illinois) and Kimberly Hoogewind (Purdue University), is the first to delve into potential connections between tornado activity and Arctic sea ice. The two phenomena appear to be more closely correlated in July than in other months, noted the authors, who provide several possible reasons for the connection and its timing.

July does not tend to produce the most severe U.S. tornado outbreaks, but on average, July contributes more tornadoes rated at least EF1 to the annual total than April does (12.7% vs. 11.4%), and it produces many weaker tornadoes (EF0). “The month of July is in fact a significant contributor to the annual tornado occurrence, especially given that low-end (F/EF0-1) tornadoes comprise the majority of the U.S. tornado population,” note the authors. On July 19, 2018, Iowa experienced 21 tornadoes, including two rated EF3; one caused 13 injuries near Pella, while the other tore through the heart of Marshalltown, injuring 22 people and causing more than $200 million in damage.

The average number of U.S. days with at least one EF1-or-stronger tornado has dropped more dramatically in July than in any other month, decreasing from 11 to 20 tornado days per July in the 1990s to 6-13 tornado days per July in the early 2010s. Beyond the end of the study period (2015), there were 13 tornado days each in July 2016 and 2017, according to Patrick Marsh (NOAA/NWS Storm Prediction Center); the numbers for July 2018 are still being finalized, said Marsh.

An analysis published in 2014 in Science and led by Harold Brooks (National Severe Storms Laboratory) found that the number of EF1-or-stronger tornadoes had changed little, but the number of days with such tornadoes has dropped overall since the 1970s, leading to more such tornadoes per tornado day as well as an overall increase in monthly and annual variability.

Figure 1.  Arctic sea ice extent averaged through the month of July has dropped by more than 20% from 1979 to 2018. Image credit: National Snow and Ice Data Center.

As Earth’s climate has warmed over the last 20 years, Arctic sea ice extent has dropped substantially, especially during summer and autumn. Several studies—including one in 2017 led by Michael Mann (Pennsylvania State University)—have found a weakening of summer circulation across the midlatitudes of the Northern Hemisphere, with weather patterns more likely to get “stuck” and produce extremes in temperature and precipitation.

In line with these studies, Trapp and Hoogewind found that the Julys with fewer tornado days tended to have stronger upper-level ridges extending across the northwest and north-central states. The ridges lead to weaker upper-level winds and less vertical wind shear in the lowest 6 kilometers (3.7 miles) of the atmosphere, which in turn tend to allow fewer days with tornado-producing storms across these areas (which are typically some of the most common locations for July tornadoes). Changes in atmospheric stability don’t seem to be as much of a factor as wind shear, said the authors. This makes sense: a warm, moist, unstable atmosphere is more common in July than it is in the spring, so wind shear becomes the main determining factor in tornado production.

Midsummer tornadic storms in the northern and central Great Plains often morph into huge, rainfall-dumping mesoscale convective systems (MCSs). The authors found that tornado-sparse Julys also tended to see drier conditions across a large part of the central Great Plains and Midwest. In other words, both tornado days and average rainfall have been on the decrease in this area since the 1990s.

In asking “Why July?” the authors note that the apparent link between sea ice and tornadoes may be more evident in July than in other months because large-scale factors known to affect tornado frequency across the Great Plains in springtime become less influential by midsummer. These include the El Niño/Southern Oscillation, the Madden-Julian Oscillation, and sea surface temperatures in the Gulf of Mexico.

Figure 2. Scatterplots of monthly mean Arctic sea-ice extent (SIE) versus monthly E/EF1+ tornado days (TOR) over the period 1990–2015. Solid lines show linear fit. The Pearson correlation coefficient (Rp), Spearman rank correlation coefficient (Rs), and p-value for each month are indicated near the top of each plot. Shading shows 95% confidence intervals. July shows the greatest correlation between SIE and TOR, with lower values of Arctic sea ice extent correlated with fewer tornado days and greater ice extent co-occurring with more tornado days. Image credit: From Figure 1c, Trapp and Hoogewind, “Exploring a possible connection between U.S. tornado activity and Arctic sea ice,” https://www.nature.com/articles/s41612-018-0025-9, courtesy Nature Publishing Group.

“I am generally supportive of the idea that summertime sea-ice can alter jet patterns that in turn have forcing on severe local storms,” said Victor Gensini (Northern Illinois University). He leads a project called Extended Range Tornado Activity Forecasts (ERTAF) that’s been using long-range computer model forecasts to issue weekly outlooks for tornado activity up to three weeks in advance. “This is just one study,” Gensini added, “but it pushes the severe storms community to think more about how climate and global change (including changes in large-scale circulation patterns) may impact severe weather frequency.”

The study does not attempt to directly attribute the observed changes in tornado days to the changes in sea ice, or or to assess how atmospheric dynamics might have interacted with sea-ice loss. “Future global modeling experiments will help resolve such questions,” the authors said.

We'll have a full update on the Atlantic and Pacific tropics in our next post later today.