1. Is there a demographic bias in the Kansas tornado record? Dr. John Heinrichs Samuel Lane Department of Geosciences Fort Hays State University

2. Overarching research goal Identify the climatological factors that influence tornado occurrence in Kansas Tornadoes cause many deaths and injuries each year, predictions can assist disaster managers Simulations of global climate change suggest increases of extreme weather Argonia, KS May 29, 2004 (http://www.mesoscale.ws/pic2004/040529-33.jpg)

3. Background Previous research indicated a clear relationship between the El Niño/Southern Oscillation and tornado frequency (in particular, tornado activity increases in the western portion of the state after La Nina events) Higher frequency of tornadoes in counties containing major metropolitan areas was noted

4. Background (continued) During that research, observed that Kansas tornado report occurrence shows dramatic positive trend with time Raised question about whether the observed increase was due to more observers (i.e., tornadoes were just as common in the past, we just missed them) Average number of tornadoes per year: 56 Standard deviation: 30 Note “lull” in tornadoes in 1970s, steep increase afterwards

5. Previous research on the topic Divided opinions: –Aguirre et al., 1994: population and income not related to detectability of weak (F0 and F1) tornadoes across the continental US – excellent paper but needs to be confirmed and updated (also county level may be too fine) –Anderson et al., 2005: spatial variability of tornado reports may be modeled by a measure of human population density

6. Data Tornado data obtained from the Severe Storms database at the National Climatic Data Center (NOAA/NESDIS) for 1950-2004 3087 total tornado reports All data assigned counties (post 1993, data set has more precise locations) Population data for 1950, 1960, 1970, 1980, 1990, 2000 from US Census Bureau

7. Dramatic increase in F0 tornadoes since 1975 (note that in the 70s, more F1s and F2s then F0s, though!); F4s and F5s have been stable over time Perhaps the weakest tornadoes were missed previously?

8. Methodology Spatial unit of analysis: climate division Calculate tornado trends (all tornadoes and F0s only), population trends for each division over 1950-2004, 1975-2004 Test null hypothesis that the tornado and population trends are correlated across the divisions

9. Results: All tornadoes 1950-2006 Divisions 3,7,8 have same trend sign, rest are opposite Division 6 (largest increase in population) had a decrease in tornadoes!

10. Statistical results for all tornadoes, 1950-2004 Pearson’s correlation coefficient (assumes normality) over all divisions: -0.41 Spearman’s rank correlation (nonparametric): -0.23 Critical value for 7 degrees of freedom: 0.521 Can reject null hypothesis at 95% level of significance

11. Results for all tornadoes 1975-2004 Larger tornado trends, all positive 5 divisions with same trend sign Division 6: greatest population growth but lowest tornado trend

12. All tornadoes 1975-2004 Pearson’s correlation: -0.19 Spearman’s rank correlation: -0.12 Null hypothesis can be rejected

13. Results for F0 tornadoes 1950-2004 Every division has positive tornado trend Signs of tornado and population trends opposite for 5 divisions

14. F0 tornadoes 1950-2004 Pearson’s correlation: -0.12 Spearman’s rank correlation: 0.02 Null hypothesis can be rejected

15. Results for F0 tornadoes 1975-2004 Tornado trends greater then full time period 4 divisions with opposite sign

16. F0 tornadoes 1975-2004 Pearson’s correlation: -0.23 Spearman’s rank correlation: -0.02 Null hypothesis can be rejected

17. Graphical representation by decade Left graphs show tornado totals by decade, right show population by decade

18. Another interesting result If multiple tornado reports on the same day in the same county are combined (as done for the ENSO study), trend is largely eliminated Difference between reported tornadoes and tornado events/days is greatest after 1990

19. Conclusions Can reject hypothesis that increases in tornado report frequency are related to population change Results consistent with Aguirre et al. (1994) In fact, correlations surprisingly suggest very weak inverse relationship Divisions containing KC/Topeka, Wichita (rapid growth areas) have very different trend patterns Much of apparent growth of weak tornadoes is in multiple events

20. Discussion Is the observed increase in tornadoes a real one? Not necessarily! –Changes in reporting process? –More attentiveness/awareness? –More spotters/chasers? Could it be that the observed increase is because we are better at seeing what goes on in large mesocyclones with tornado swarms? Future work needed to test these hypotheses

21. References Aguirre, B. E., Rogelio Saenz, John Edmiston, Nan Yang, Elsa Agramonte, Dietra L. Stuart (1994). "Population and the Detection of Weak Tornadoes." International Journal of Mass Emergencies and Disasters 12 (3): 261-277. Anderson, C.J., Wikle, C.K., Zhou, Q., and J.A. Royle (2005). "Population Influences on Tornado Reports in the United States" In Review.