The frequency of strong tornadoes

Given the recent EF5 Moore tornado, the 2011 Joplin EF5, the 2011 Tuscaloosa EF4 and other strong tornadoes in the last few years, a lot of people have been commenting on whether or not we're seeing more violent tornadoes now than in the past.  A few have even tried ascribing recent violent weather to climate change.  I find this particularly amusing, particularly consider that these last 12 months have been the least active 12 months of tornado activity on record.  But what do the trends say--are we seeing more strong or violent tornadoes more frequently?

A lot of people have looked at this particular question, including some excellent blog posts like those from  Jeff Masters.  A timely article in the Bulletin of the American Meteorological Society by Kunkel et al. shows that the frequency of strong (EF1+) tornado reports is not increasing, though weaker (EF0) tornado reports have increased.

Number of tornadoes per year (From Kunkel et al. 2013)

However, they note that these observations are complicated by many different factors, most notably the expansion of population areas over time and the increased prevalence of automated observations.  Because of these factors they looked at changes in the frequency of weather conditions that favor strong tornadoes instead.

Let's dive into these counts a little deeper to illustrate why it's difficult to pull any trend out of these numbers. I pulled a Patrick Marsh and grabbed the tornado count data from the SPC website and made the same sort of plot as the Kunkel et al. plot above, but for all of the different F/EF ratings.  On the left we see the counts of the total tornadoes reported each year in each strength category and on the right we see these counts as a fraction of the annual total.

In terms of raw numbers of tornadoes reported, the number definitely has been increasing over the last 53 years. However, we can see that it really has been those F/EF0 reports that have been driving the increasing trend in tornado reports.  If we look at the percentage plot on the right, we only see an increasing trend for F/EF0 reports--there is virtually no trend in the fraction of F/EF 1, 4 or 5 with slight downward trends (in terms of total fraction in F/EF2 and F/EF3 reports.  This is an interesting breakdown, particularly in the divergence between the reports of F/EF1 and F/EF 2 tornadoes.  This seems to be the breaking point--the total number of F/EF 1 and F/EF 0 tornadoes reported has been increasing while other levels haven't shown a whole lot of change.

A lot of this increase in reporting weaker tornadoes has to do with population growth as we urbanize more areas and a general increase in public awareness.  If you have strong tornadoes (here, F/EF 2 or greater), they're typically more likely to be reported; the damage is a lot more obvious and more easily attributable to tornadoes.  As we increasingly build up our urban areas and become more interconnected, we've become more sensitive to even small disruptions in our infrastructure and as such even damage caused by weak tornadoes gets reported.

I'm also curious as to the sudden jump in the number of  F/EF0 tornadoes that began in the late 1980s and has continued through this day.  I was wondering if this coincided with the roll-out of Doppler radars nationwide that took place in the late 80s and early 90s.  It turns out that a 2005 study by Simmons and Sutter looked at the impact of our Doppler radars on tornado warnings.  They broke down (on a forecast office-by-forecast office basis) the number of tornado reports by F-scale before and after the WSR-88D implementation:

From Simmons and Sutter (2005)

We can use these numbers to plot out the percent change in total number of tornado reports of each F scale category before and after Doppler radar implementation.

The percentage of reports of F0 tornadoes increased by about 11 percent--the only category where this happened.  I really think that the Doppler radar implementation has had a significant role in increasing the number of weak tornadoes reported, as this has allowed meteorologists to see small-scale spinups that might otherwise have been lost in larger thunderstorm wind damage.  This increased detection of weaker tornadoes helps focus storm survey efforts and attributes more damage reports to weak tornadoes than we may have suspected in the past.

So, after that brief look, it's true that the number of tornadoes reported has been increasing over time, but that doesn't necessarily mean that the total number of tornadoes has been increasing over time--we're just getting better and detecting weaker tornadoes and our population is more sensitive to the effects of weaker tornadoes.  In focusing on stronger tornadoes, there's not enough evidence to suggest that the number of reports are increasing.  In fact, as a fraction of total reports, the amount has remained somewhat steady if not decreasing slightly as weaker tornado reports have become far more common.

Quick Comparison of This Event with the January 1999 Blizzard

In reading a lot of the forecast discussions from WFOs across the upper midwest, I noted several references comparing this coming winter weather event to the blizzard of January 2-4th, 1999.  Though I apparently lived right through the middle of it, I honestly didn't remember this particular event.  So, I thought I'd see what all this hubub was about and quickly, qualitatively compare the two events.

The initial synoptic setup is remarkably similar.  Take this morning's 12Z 500mb analysis from the GFS:

Fig 1 -- GFS 500mb geopotential height and wind analysis at 12Z, Monday, January 31, 2011.

Note the two features I noted in my last post that will combine to produce this powerful storm--a shortwave digging out of southern Alberta into nothern Montana and another in the desert Southwest over Arizona.  Compare this to an archived 500mb analysis from 12Z on Friday, January 1st, 1999.

Fig 2 -- NOAA Difax 500mb N. America geopotential height analysis from 12Z, Jan. 1, 1999. 

There are indeed some eerie similarities between the two analyses.  There were also two shortwaves analyzed before the 1999 storm, though these were slightly further east--one was coming out of southern Saskatchewan while the axis of the other was in New Mexico.  Though these features are slightly further east, the forecast for the current event will quickly catch this week's shortwaves up and even speed past the timing of the 1999 event.  But more on that later...

Back to this current event, twenty-four hours later (12Z on Tuesday morning), the GFS forecast has a well-defined, if somewhat east-west expansive, trough across the central part of the country with embedded short waves.

Fig 3 -- GFS 24-hour forecast of 500mb geopotential height and winds valid 12Z, Tuesday, February 1, 2011.

Comparing this to the 1999 event 24 hours later, we see that the analyzed 500mb trough didn't seem to have these embedded shortwaves and was more north-south oriented.  This is an important difference--the lack of tilt in the 1999 event slows down the progression of the trough and associated surface features.

Fig 4 -- NOAA Difax 500mb N. American analysis of geopotential height valid 12Z, Saturday, Jan. 2, 2011.

Even with these differences, however, both analyses feature at least some form of trough axis moving through central or eastern Texas with an attendant jet streak around the base of the trough.  Both of the exit regions of the jets would lie in the mid-Mississippi valley, enhancing the divergence aloft there.  But the timing and orientation of these two features is still different.  Here's the surface pressure forecast for the GFS at 12 hours later--00Z on this Tuesday night.

Fig 5 -- GFS 36-hour forecast of 500mb geopotential height and winds valid 00Z, Wednesday, February 2, 2011

Compare that to the surface analysis from the 1999 event at 24 hours later--12Z, Sunday, Jan. 3 1999.

Fig 6 -- NOAA Difax surface analysis valid 12Z, Sunday, Jan. 3, 1999.

BOTH analyses have a deep surface low located in the Memphis/southern Illinois region.  However, the GFS (and most of our other) model forecasts bring that low there a full 12 hours faster than the January 1999 event.  By the time we move the full 48 hours out in the GFS forecast for this current event, the low pressure center at the surface has already moved into Indiana and begun to occlude.

Fig 7 -- GFS 48 hour forecast of surface pressure and temperature valid 12Z, Wednesday, Feb. 2, 2011.

We don't see the occlusion on the 1999 storm until sometime before 72 hours out:

Fig 8 -- NOAA Difax analysis of surface pressure (and fronts) for 12Z, Jan. 4, 1999.

So by the time this low occludes, our current forecasts for the storm this week move things along 12-24 hours faster than the 1999 storm.

There are cold air intrusions in both events, however the observed 20-25 degree Fahrenheit morning lows in Oklahoma by the time of occlusion in the 1999 event are outstripped by the single-digit lows being forecast by the GFS for this week (NOT that we trust the GFS surface temperature forecasts...).  This hints at a stronger baroclinic zone and much colder air behind the front.  This becomes important because if we look at snowfall totals from the 1999 event--

Fig 9 -- NOAA Difax analysis of observed snowfall on the ground at 12Z, Jan.3, 1999.

--we see that while the upper midwest saw large amounts of snow (17 inches at O'Hare in Chicago on that list to the right...), the snow amounts taper off to the southwest.  Only a few inches were reported at places in Missouri and none in Oklahoma.  Contrast this with the blizzard warnings extending all the way into central Oklahoma with the event tomorrow--12-18 inches of snow are expected in parts of Oklahoma.  From this, we have to conclude that there is more moisture being forecast with the current event and/or the temperatures were colder further south, supporting more snow all the way into Oklahoma.

Interestingly, according to records at Wiley Post Airport in Oklahoma City, on January 2-3, 1999, snow and rain were both reported in the observations, but no measurable accumulation of either seems to be recorded.  The high temperature also hovered in the low to mid 30's with overnight lows in the low teens after the event. Not near the cold weather we're looking at according to current forecasts...

So, in conclusion, there are some general similarities in the evolution of both the 1999 blizzard and this week's event.  However, this week's storm will move much more quickly, have much colder air behind it, and bring wintry precipitation further south.  Up north, however, the proxy seems to be pretty good--if a storm of that intensity (and perhaps even more limited moisture) in 1999 could bring 17 inches of snow to Chicago, projections in that range would not seem to be out of the question for the upper midwest with this week's storm.

Finally (on a slightly technical note), I just want to leave with a cross-section image generated on the College of DuPage website from this evening's soundings.  This cross section runs from Albuquerque in the west to Nashville in the east.  We can see that "warm conveyor belt" starting to ramp up with that bulls-eye of moisture (the green contours) in an area of southerly winds to the east.

Fig 10 -- Cross section from Albuquerque, NM (left) to Nashville, TN (right) showing contours of potential temperature (red), mixing ratio (green), theta-e (yellow) and wind barbs valid at 00Z, Tuesday, Feb. 1, 2011.  From the College of DuPage website.

And so it begins...

Also, for all of you reading this at the University of Washington--I will be giving tomorrow's weather briefing in 310C at 12:30 PM and will be going into a fair amount of detail about this event during the presentation.  Feel free to come by...