Fig 1.
Kansas counties and elevation.
Counties are labeled by the corresponding CWA. Elevation is given at a resolution of 80 m.
Fig 2.
Population changes between 1970 and 2012.
The change is expressed as a percentage difference with 2012 as the base year.
Fig 3.
Population estimates for 2012 by county.
Values are expressed as persons per square km.
Fig 4.
Tornado report frequency by county for Kansas.
Only tornadoes rated EF1 and higher are used. Lines show the tornado track. The shortest tracks are not visible at this scale. Total tornado counts over the period 1970–2013 are listed inside the county and the color scale is from few (blue) to many (red).
Fig 5.
Statewide tornado counts for Kansas from 1970–2013.
The trend line uses a second-order random walk model where the counts are described by a negative binomial distribution. The 90% uncertainty band is shown in gray.
Table 1.
Summary of the data analysis and modeling results.
DIC is the deviance information criterion, AD is the Anderson-Darling test, and r is the Pearson correlation coefficient.
Fig 6.
Correlated random effects from the Kansas tornado model.
Values are the posterior mean and are expressed as the percent difference from the state average. The model includes annual population density and calendar year as fixed effects.
Fig 7.
Standard deviation of the correlated random effects from the Kansas tornado model.
Values have units of percent difference from the state average.
Fig 8.
Posterior density of the elevation standard deviation term.
The 90% CI is shown with the vertical gray lines. The red line indicates no effect.
Fig 9.
Same as Fig 6 except the model has elevation standard deviation as an additional fixed effect.
Fig 10.
Tornado report frequency by county for Illinois, Mississippi, South Dakota, and Ohio.
Fig 11.
Statewide tornado counts.
Fig 12.
Correlated random effects from the state tornado models.
Fig 13.
Posterior density of the elevation standard deviation term from the state tornado models.