(Updated 29 August 2003. Comments to Harold Brooks)
The ratio of the mean number of days per year with an event within 25 miles of a point reported in 19901994 over the mean number reported in 19801984. The contour interval is .5, with contours greater than 1 colored red. The maximum contour value displayed is 3.
The following maps show when during the year the maximum time of probability of the particular event occurs. The fill levels are labelled by Julian day of the year, but the translation isn't too bad. The lowest nonblack color is associated with 1530 March. The next one starts with 31 March, the next with 15 April, then 30 April, 15 May, 31 May, 15 June, 30 June, 15 July, and 31 July (ending 14 August). Then, it jumps up to 1529 November (the white color). A minimum of 0.4 days per year on average was required to make the cut and the period of record in 19801994.
For tornadoes, you can see the maximum date of threat move northwest through the Plains. For wind, the most striking thing is the very large area with very similar times of peak (early July and, to a lesser extent, late June.)
Ratio of fall peak to spring peak
Another item of interest is how strong the fall peak in tornadoes is compared to the spring peak. Here's a map of the ratio of the maximum daily probability from October through December to the maximum daily probability from January through August. Orange and higher colors are associated with ratios of 0.5 or more. Dark green and white are greater than 1, indicating that the fall peak is stronger than the spring peak over southern Mississippi.
Just to show what "Dixie Alley" looks like, I've also plotted the mean number of days with tornadoes from October through December and from March through July. Note that they have different contour intervals.
From the individual daily probabilities, I've computed the probability of having one or more days with a tornado within 25 miles of a point sometime during a month. The base data for all tornadoes are the reports from 19801994 and for significant and violent tornadoes are the reports from 19211995. The values are probabilities in percent.













































This is a plot attempting to look at the reliability of of the annual cycle. For every point with a mean of at least 0.4 tornado days per year, I've calculated a trimmed standard deviation of the dates of maximum for each year in the annual cycle. (I eliminated the earliest and latest days of the record at each point after attempting to account for the points where the maximum occurs late in the year.) The shaded colors show the trimmed standard deviation for all points where it is less than 30 days, implying that there's a 95% probability that the date of maximum threat will fall in a time window less than 120 days for any particular year. Since even this trimmed standard deviation is not a resistant statistic, values greater than that are not good estimates of the variability in the data set.
Overlaid are the 0.75 and 1.00 mean days per year contours to show where the overall threat is greatest. I've been toying with the idea that the combination of high threat (say, greater than 0.75 days per year) and high repeatability of the annual cycle (standard deviation less than 20 days) is an estimate of what has traditionally been referred to as "Tornado Alley".
An objective Significant Tornado Alley based off of F2 or greater occurrences can also be derived. Here, we're set the criteria for occurrence to be 0.1 days per year and the repeatability of the annual cycle to be having 60% of the dates of the peak threat from each year falling in some 50day window during the year. See Concannon et al. for an example of the window approach from a couple of points.