Since I forgot to post a this week due to continued winter weather, and because the snow battered center of the United States has begun its long awaited warm-up, I thought I’d go ahead and share the statistics and figures relating to February tornadoes. The plots and statistics are similar to those I computed for January that can be found in posts here and here.

Above is a figure depicting the number of February tornadoes per county for the years 1950 to 2009. Most tornadoes tend to occur in the southeast United States, however tornadoes have been reported as far west as California and Hawaii, and as far north as Oregon and western New York!

If we sum up the number of unique tornadoes from the country data and turn it into the number of tornadoes by state, it would look like this:

Florida and Mississippi lead the way with over 180 tornadoes, followed by Texas and then the rest of the southeastern states. Breaking down each states numbers, the tallies look like this:

Now, the number of injuries per state break down as follows (please note the order of the states listed on the left is in descending order of number of tornadoes per state):

The number of fatalities per state break down as follows (please note the order of the states listed on the left is in descending order of number of tornadoes per state):

And putting everything into a single, cluttered figure, we have this:

Next up in the list of Storm Prediction Center (SPC) outlook climatologies is the yearly climatology of slight risks. Between 1990 and 2008 there were 24,455 slight risk polygons issued. This does not mean there were 24,455 slight risk outlooks issued as it is possible to have more than one slight risk polygon drawn on a given outlook. Furthermore, note that each day consists of multiple outlooks and so it is possible for a grid point to receive multiple “hits” for being located in a slight risk outlook on the same day. In other words, if a location was located within 2 slight risk outlooks, this does not guarantee that two days of slight risks. The location might simply be contained within a slight risk from two separate outlooks issued for the same day.

Below is the slight risk climatology that employs the Kernel Density Estimation technique. Notice how it is smoother along the edges than the raw outlook which is located at the end of this post. The black solid lines are contour intervals of outlooks per year in steps of 25. The color fill contours are in increments of 1 outlook per year. As you can see, the central United States is once again the leader in the number of slight risks to expect per year. This has to do with the repeatability and predictability of the ingredients necessary for severe thunderstorm development. In other words, ingredients come together more frequently and more reliability across “Tornado Alley” than anywhere else in the world. This doesn’t mean that other areas aren’t prone to severe thunderstorms. Some of the largest tornado outbreaks in history have actually occurred well east of traditional Tornado Alley (e.g., 21 January 1999 and 3-4 April 1974).

Below is the raw graphic, with now Kernel Density Estimation applied. Because the edges are not smooth line, rather they are intersections from thousands of polygons, I did not plot any black line contours.

If you are coming to this page from an outside source, please see the additional two posts in this series. They can be found by clicking the following: Moderate and High Risk Climatologies and Slight Risk Climatology

Although most of the country is still digging out and cleaning up from a historic winter storm, commonly referred to as the “Groundhog’s Day Storm”, I decided to go ahead and answer the Question of the Week regarding Storm Prediction Center (SPC) Moderate and High Risks.

Before I answer the questions, first let me explain how I arrived at the answers…

I took the polygon outlining the first Moderate [High] risk during a given year, regardless of time of issuance. This means I treated a 12 UTC issuance the same as a 1630 UTC issuance and 2000 UTC issuance. I took the risk polygon, placed it on a 4km grid (specifically grid number 240), and activate all grid points that fell inside the risk polygon. This left me with a grid of 1′s (inside outlook) and 0′s (outside outlook). I created a grid for each year and then summed all the grids together. This gave me a grid containing the number of times each grid point was within the first Moderate [High] risk of the year. I then divided each grid point by the number of years I was examining. This left me with the probability that a grid point would be contained in the first Moderate [High] risk of the year.

The resulting plots were extremely blocky because of the straight edges of the outlook polygons and the relatively small size of the resulting sample size (19 years for Moderate risks and 18 years* for High risks). In an attempt to better retrieve the underlying probability distribution, I employed a technique called Kernel Density Estimation. I used a gaussian kernel with a bandwidth of 120 kilometers (per Brooks et al. 1998), and extended the resulting distributions out to 5 standard deviations.

To determine when the first Moderate [High] risk is typically issued, I converted the day and month of the first Moderate [High] risk into the corresponding day of year (e.g., 1 July is the 182 day of the year). I then took the mean day of the year and converted it back to day and month. This means that the average first issuance of a Moderate [High] risk may not fall on a day that actually had a Moderate [High] risk issued.

For this analysis I only evaluated the years between 1990 and 2008. This is because in the early years of the National Severe Storms Forecast Center (NSSFC) / Storm Prediction Center categorical risks had a different meaning than they do now. Thus, I tried to limit the analysis to relatively current definitions.

So, here are this week’s answers:

In the graphics below, the colored contour intervals are every 1% and the thin black contours are every 5%.

• Climatologically, where is the most likely location to experience the first SPC Moderate Risk of a given year? There are actually two areas when using the raw probabilities: southwest Arkansas and southern Alabama. However, when applying the kernel density estimator, the area in southern Alabama has a slightly higher probability. As for when this first Moderate risk is issued, the average first issuance is 28 January.

• What about High Risk? There is actually more variability (both spatially and temporally) in the issuance of the first High risk than the first Moderate risk. This is most likely the result of the greater variability in time of year, and thus greater variability of the large scale pattern. High risks have been issued as early as 21 January (1999) and as late as 1 July (1997). [Technically it has been issued as late as never, because 2000 did not have a single High risk issued.] For early High risks, the favored location is in the southeast United States where the ingredients that favor severe thunderstorms is more often confined to the Gulf Coast areas. Later in the year, more northern locations are favored as the cyclone track, and ingredients for severe thunderstorms, lifts northward. With this said, the first High Risk is typically issued across portions of southeast Arkansas and northern Mississippi between 27-28 March.

*There were no High risks issued in 2000, thus only 18 years of data were used to calculate averages regarding High risks. Therefore the resulting graphics and statistics should be considered conditional statistics, meaning these statistics assume a High risk will be issued during a given year.

After posting the climatology of where the first moderate and high risks occur, I’ve received a couple of requests for additional graphics. One that was extremely easy to produce, and also one most frequently requested, is a climatology of moderate and high risks. Using the same Kernel Density Estimation technique described in the original post, I’ve calculated the number of moderate and high risk outlooks an area might expect during a given year, based on data from 1990 through 2008.

Edit to add: Between 1990 and 2008 there were 3454 moderate risks and 243 high risks issued.

Please note that each day consists of multiple outlooks and so it is possible for a grid point to receive multiple “hits” for being located in a Moderate or High risk outlook on the same day. In other words, if your location is located within 2 high risk outlooks, this does not guarantee that you will have two days of high risks. The location might simply be contained within a high risk from two separate outlooks issued for the same day.

For moderate risks, central Oklahoma appears to be the clear winner with nearly 30 moderate risks averaged per year. Increase probabilities extend both north and east from here.

As expected, the average number of high risks per years is considerably less than the average number of moderate risks. (In fact, I had to change the color scale!) Northeast Kansas, northern Missouri, and west-central Illinois are the most likely areas to experience a high risk in a given year with slightly more than 2 expected. A minor axis of increased probability extends southward from the eastern edges of this highest probability band, reaching portions of eastern Arkansas and far northern Mississippi.

Notice how in both of these climatologies, the maximum probabilities are centered in the central United States — east of the Rocky Mountains and west of the Appalachian Mountains. We’ll leave discussion as to why this is for another blog post.

A couple of things I will be watching this evening

• How the cold air interacts with the developing cyclone

The million dollar question (literally) is going to be how the cold air interacts with the developing cyclone. If the cold air plunges more south than southeast, then I would expect the ejecting surface low pressure (and associate low-to-mid-level lows) to move with a more northerly trajectory. If the cold air surges southeast, then the lows will eject with a more eastward component. This has huge implications for who will see the heaviest snows in Oklahoma and southwest Missouri.

The reason for watching how the cold air interacts with the cyclone is the result of how and where the low-level cyclones develop — which on the eastern edge of the “cold-dome”. I strongly believe the surface lows (and associated other lows) will ridge along the edge of the cold dome. Thus, if the eastern extent of the cold air has a more north-south orientation, this is how the surface low will move, placing central Oklahoma in a longer period of heavier precipitation. If the cold air orientation is more southwest to northeast, then expect the heaviest snow to shift east of central Oklahoma, as the surface development will be farther east.

• How the upper-level jet streams interact

Nearly all numerical weather prediction models are forecasting extreme precipitation amounts overnight and through the day tomorrow. It’s a bit perplexing considering the dry ambient environment. This dryness is most likely being overcome by the strong upper-level divergence associated with the ageostrophic response to a couple of a southern and northern jet streak. This strong upper-level divergence results in strong low-level convergence across portions of Oklahoma that will most likely rapidly transport moist air currently in place across eastern Texas, southern Arkansas, and Louisiana. When and where these upper-level jet streams do couple will play a major role in deciding where the heaviest precipitation bands are located.

• How does the precipitation shield evolve overnight

The last several runs of the operational GFS and NAM develop rapidly accumulating snows across eastern Oklahoma with strong low-level frontogenetic forcing. This is in spite of a +0.5C to +1.5C warm nose aloft. This would seem to favor more sleet than snow. However, if lift is strong enough, this might quickly be overcome resulting in 3-4″ per hour snow rates that are currently forecast. Farther west, the 12 and 18 UTC NAM are developing a more classic deformation zone across a large part of central Oklahoma during the late morning and early afternoon hours tomorrow. This deformation zone is more typical of substantial Oklahoma snow falls than the strong, low-level frontogenetic forcing described previously. This large deformation zone is the reason why the NAM has a large area of 6-10″ snow accumulations to the west of the 12-18″+ forecast across eastern Oklahoma.

Also, if a large squall line develops across central Texas and races east too quickly, the squall line might use up a lot of the moisture that is poised to be drawn northward into this cyclone. This would decrease snow totals across portions of Oklahoma, Kansas, and southwest Missouri.

• The evolution of the temperature vertical temperature profile

I hinted at this in a couple of the previous bullets, but how the vertical temperature profile evolves will be crucial in determining snowfall amounts. If a strong warm-conveyor belt does develop overnight, warm air might hang on longer than forecast across portions of Oklahoma. This might lead to an extended period of sleet which would cut down on snowfall totals considerably. However, 1-2″ of sleet, coupled with 2-4″ of snow would be just as bad, if not worse, than a pure 8-12″ of snow.