sign in sign up
<<
Home , Dixie Alley, Storm spotting, Tornado Alley, Tornado climatology

366 WEATHER AND FORECASTING VOLUME 29

An Objective Analysis of Risk in the

TIMOTHY A. COLEMAN University of in Huntsville, Huntsville, Alabama

P. GRADY DIXON State University, Mississippi State, Mississippi

(Manuscript received 10 May 2013, in final form 12 September 2013)

ABSTRACT

In this paper, an objective analysis of spatial tornado risk in the United States is performed, using a somewhat different dataset than in some previous tornado climatologies. The focus is on significant tornadoes because their reporting frequency has remained fairly stable for several decades. Also, data before 1973 are excluded, since those tornadoes were rated after the fact and were often overrated. Tornado pathlength within the vicinity of a grid point is used to show tornado risk, as opposed to tornado days or the total number of reported tornadoes. The possibility that many tornadoes in the were underrated due to the lack of damage indicators, causing a low bias in the number of signifi- cant tornadoes there, is mostly discounted through several analyses. The kernel density analysis of 1973– 2011 significant tornadoes performed herein shows that the area of highest risk for tornadoes in the United States extends roughly from to and northwestern , with the highest risk in the southeastern United States, from central across most of Mississippi and northern Alabama.

1. Introduction Edwards 1999), Kelly et al. (1978) reported that the area of the United States with the highest tornado frequency Debates have been ongoing about the location of the was the Great Plains, or the classic [as most tornado-prone of the United States over mapped by Schaefer et al. (1980) and from McNulty the past 10 yr or so. For many years, ‘‘tornado alley,’’ (1995); Fig. 1a]. Brooks et al. (2003), using tornado days a in the Great Plains extending from northern and limiting the risk areas to those with tornado to , was considered to be the area with the that are consistent by time of year, suggested that the greatest tornado risk in the country. However, the idea primary tornado alley in the United States extends from of another local maximum of tornadoes in the south- Texas to , similar to and slightly northwest eastern United States was first mentioned by Alan of the classic tornado alley (Fig. 1b). Concannon et al. Pearson in 1971 (Gagan et al. 2010), and further shown to (2000) examined days with significant tornadoes [those some degree by Concannon et al. (2000), Ashley (2007), rated on the as category 2 (F2) or higher] Dixon et al. (2011), Smith et al. (2012),andKellogg and from 1921 through 1995, and found an L shaped region Forbes (2013). of maximum tornado risk that included most of the Researchers have also used various ways of defining classic tornado alley from Schaefer et al. (1980), then tornado-prone areas and, by implication, tornado risk. extended eastward through Arkansas, Mississippi, and Using data from what eventually became the Storm Pre- Alabama (Fig. 1c). Smith et al. (2012),usingamore diction Center (SPC) tornado database (e.g., Schaefer and limited dataset (2003–11) including tornadoes of all intensities, suggest that the southern part of the tornado-proneareashownbyConcannon et al. (2000), Corresponding author address: Tim Coleman, Dept. of Atmo- spheric Science, University of Alabama in Huntsville, NSSTC, 320 from Oklahoma to Alabama, has the highest tornado Sparkman Dr., Huntsville, AL 35805. risk, with a maximum over parts of Mississippi and E-mail: [email protected] Alabama. However, this small dataset was probably

DOI: 10.1175/WAF-D-13-00057.1

Ó 2014 American Meteorological Society Unauthenticated | Downloaded 09/30/21 02:21 PM UTC APRIL 2014 C O L E M A N A N D D I X O N 367

FIG. 1. (a) Annual tornado frequency for all tornadoes from Schaefer et al. (1980), shown in McNulty (1995), based on records before 1980. (b) Tornado occurrence based on frequency of all tornadoes, 1980–99, combined with a measure of seasonality of tornadoes (Brooks et al. 2003). (c) Number of days per century with one or more sigtors 1921–95 (Concannon et al. 2000). (d) Count of all tornado events, 2003–11 (Smith et al. 2012). (e) Number of sigtors 1950–2004 (Ashley 2007). (f) Number of days with sigtor per century, 1961–2010 (Carbin et al. 2012). overly influenced by the 27 April 2011 outbreak (Fig. 1d). There are several issues that have not been fully Ashley (2007), using only significant tornadoes during addressed in prior studies. First of the years 1950–2004, and Carbin et al. (2012), examining all, the Fujita scale was not published until 1971 (Fujita the return frequency of significant tornadoes from 1961 1971; Edwards et al. 2010), and was not adopted for through 2010, also showed that the region from Okla- rating tornadoes in their near-immediate aftermath in the homa to Alabama has the highest tornado risk (Figs. 1e United States until 1973. So, tornadoes before this time and 1f). were rated years after the fact in the SPC database, using

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC 368 WEATHER AND FORECASTING VOLUME 29 newspaper stories and photographs that likely over- –related tornadoes, which tend to be emphasized the damage, causing overrating of some weak [EF0 or EF1; e.g., Schultz and Cecil (2009)]. Maps tornadoes prior to 1973 (e.g., Schaefer and Edwards showing the smoothed average annual pathlength of 1999; Brooks and Craven 2002; Schaefer and Schneider tornadoes within 40 km (25 mi) of a point were gener- 2002; Anderson et al. 2007). Also, there has been ated using kernel density estimation (KDE, e.g., a steady increase in annual reported tornadoes in the O’Sullivan and Unwin 2003). United States, even since 1973, due to better spotting KDE is a very common method used in spatial anal- techniques, (NWS) warning ysis and has been applied to tornado risk in several verification procedures, and improved detection by ra- previous papers (e.g., Brooks et al. 2003; Dixon et al. dar. This overall increase has primarily been due to an 2011; Smith et al. 2012; Marsh and Brooks 2012). As increase in weak tornadoes (F0 or F1, hereinafter re- discussed in detail by Dixon et al. (2011), KDE analysis ferred to as weaktors). However, the reporting of sig- implies that spatial patterns have magnitudes at every nificant tornadoes (F2 or greater, hereafter referred to given point, as opposed to only the places directly af- as sigtors) has remained relatively stable since 1973. fected (in this case, by tornadoes) (e.g., O’Sullivan and Additionally, as seen from the results of various re- Unwin 2003). The KDE method used in this study em- searchers in Fig. 1, most tornado climatologies have ploys the Epanechnikov quadratic kernel probability used either tornado days or the number of tornadoes as density function as calculated by ESRI ArcGIS Spatial the basis for their studies, not utilizing pathlength. Analyst software (Silverman 1986; de Smith et al. 2007; However, pathlength seems more desirable because it is Dixon et al. 2011; Marsh and Brooks 2012; Smith et al. more proportional to the area impacted by tornadoes 2012). There are several other types (i.e., shapes) of than tornado numbers or tornado days. kernel functions, but Dixon and Mercer (2012) shows The purpose of this research is to further our under- that the differences in spatial patterns due to kernel standing of the regions of greatest tornado risk in the shape are negligible compared to those related to the United States by addressing the above issues in an - radius of the kernel, which is consistent with previous re- jective analysis. Section 2 will discuss the data and search (Silverman 1986; de Smith et al. 2007). Through- methodology used in the study. Section 3 will examine out this study, a kernel radius of 250 km is used because of some trends in tornado reporting and discuss the issues the relatively short study period of the current analysis. mentioned above, including the importance of path- The steps we used in performing the KDE analysis were length in assessing tornado risk. This information will as follow: 1) The 250-km probability density function help remove some biases that are present in other studies, (KDE) was applied to each tornado path on a 1-km2 producing an objective analysis of tornado risk in the grid. This results in each 1-km2 grid cell within 250 km of United States in section 4. Section 5 contains a summary the tornado path in question receiving a portion of the and conclusions. total tornado path (according to its distance from the path). 2) The sum of each point’s fractional pathlengths (measured in kilometers of tornado path per square kilo- 2. Data collection and methodology meter) for all tornadoes to affect that grid cell was calcu- The primary data source for this study is the SPC lated on a 5-km output grid. So, while only 1 out of every 25 tornado database, which provides the date and time, lo- grids is displayed, this coarser output grid changes the cation, intensity [Fujita or enhanced Fujita (EF) scale], resolution of the map and does not change the numerical pathlength, and other statistics on nearly every tornado results (Dixon et al. 2011). 3) Every cell is then multiplied in the United States for 1950–2011. We focus primarily by the approximate area of a 40-km-radius circle (5024 km2) on data from 1973 through 2011, to eliminate any bias to represent a meaningful risk metric (in kilometers of from tornadoes rated after the fact. We also use only tornado path per 5024 km2, or kilometers of tornado path significant tornadoes (F2 or greater and EF2 or greater within 40 km of the cell), and to be consistent with SPC’s since the was put in place), as tornado risk forecasts (Kay and Brooks 2000). done by many other authors (e.g., Concannon et al. 2000; Meyer et al. 2002; Ashley 2007; Carbin et al. 2012). This is due to the fact that the number of reported sigtors 3. Analysis of trends and selection of dataset has remained fairly steady since 1973, implying that fac- a. Changes in tornado reporting efficiency since 1950 tors improving tornado reporting efficiency over time have not affected the reporting of sigtors nearly as much Since 1950, according to the SPC tornado database, as they have affected reporting of weaktors. As pointed there has been a large increase in the number of reported out by Britt and Glass (2013),thisalsoremovesmost tornadoes per year in the United States (Fig. 2a). The

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC APRIL 2014 C O L E M A N A N D D I X O N 369

FIG. 2. (a) All reported tornadoes and (b) significant tornadoes by year in the United States for 1950–2011. In each chart, the dark curves indicate the actual number of reported tornadoes, blue dashed lines indicate the best linear-curve fit, and the orange curve indicates the best third- order-polynomial fit. In (b), the periods for 1950–73 and 1973–2011 are separated to account for the beginning of the near-real-time rating of tornadoes, as opposed to after the fact rating. increase has been fairly steady (linear correlation co- Most of this overall increase is not due to an increase efficient of 0.85) over the 62-yr reporting period, with an in tornado occurrence but to increases in tornado report- increase of about 16 tornadoes per year reported. Only ing (e.g., Kunkel et al. 2013). This increase in tornado re- slightly improved is the third-order polynomial fit (cor- porting efficiency has likely been due to better spotting relation coefficient of 0.86), which highlights the rapid techniques with time (e.g., Coleman et al. 2011), NWS increase in reported tornadoes from 1950 through 1965, warning verification procedures that prompted more followed by a slower rate of growth in the 1970s and frequent storm surveys (McCarthy and Schaefer 2004), 1980s, and then accelerated growth in reported torna- and the implementation of the Weather Surveillance does after 1990 [during the Next Generation -1988 Doppler (WSR-88D) network in the early Radar (NEXRAD) era]. 1990s (McCarthy and Schaefer 2004). This radar system

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC 370 WEATHER AND FORECASTING VOLUME 29

2 FIG. 3. (a),(b) Plot of annual numbers of weak tornadoes (F0 and F1) per 1000 km of land area in the GP and southeastern United States, respectively. (c),(d) As in (a),(b), but for significant tornadoes (F2–F5). alerts the NWS to areas with possible tornadoes (including , , and Iowa, the traditional tornado alley weaktors), even in unpopulated areas. The NWS then (covering 1.4 3 106 km2). The southeastern United States conducts storm surveys, finding tornadoes they may not includes Arkansas, , Mississippi, Tennessee, have found before the WSR-88D network came into and Alabama (covering 642 000 km2). being. Note that Fig. 3a shows that the number of reported If only sigtors (F2 or greater) are considered, there is weaktors in the GP continued to increase from 1973 practically no overall increase since 1950 (Fig. 2b), and through about 1991, before reaching relative stability. there is actually a decrease in sigtors since 1973. The This is likely due, at least partially, to the large increase number of reported sigtors increased during the 1950s as in research-related and nonprofessional the modern era of tornado forecasting and warning be- in that region starting in the 1980s and growing rapidly in gan (e.g., Schaefer 1986; Doswell et al. 1999). After the 1990s. However, in the southeastern United States, 1973, when tornadoes began to be rated in near-real- storm spotting is more difficult. In many areas, terrain is time poststorm surveys using the Fujita scale, the aver- more rugged, and much more of the area is covered with age annual number of sigtors decreased dramatically, trees than in the GP. In addition, a larger percentage of from 219 during the period 1950–72, to 165 during the southeastern United States tornadoes occur at night. period 1973–2011. Results of our analyses (Fig. 2) sup- Using a sun zenith-angle algorithm to determine sunrise port the assertions of Schaefer and Edwards (1999) that and sunset by date and latitude, it was determined that tornadoes prior to 1973 were likely overrated in many 48.6% of all southeastern U.S. tornadoes between 1973 cases. Therefore, our remaining analyses are focused on and 2011 occurred at night, compared to 39.3% in the tornado data since 1973. GP. Therefore, the number of reported weaktors in the southeastern United States continued along an upward b. Tornado reports 1973–2011: Sigtors contrast trend (Fig. 3b) even after 2000, unlike the GP, as ad- with weaktors vances in technology improved reporting capabilities. During the period from 1973 through 2011, the total It is reasonable to assume that, due to their larger size number of annually reported tornadoes in the United and longer duration, sigtors are less likely to go un- States continued to increase, but the number of sigtors reported (e.g., Concannon et al. 2000; Brooks 2004). has remained fairly steady (Fig. 2). To more specifically According to the SPC tornado database (1973–2011), examine the geographic differences around this issue, the average pathlength of a weaktor was 2.9 km, and the Fig. 3 shows the yearly number of weaktors and sigtors average path area (length 3 width) was only 0.33 km2. per 1000 km2, comparing the Great Plains with the By contrast, the average pathlength of a sigtor was southeastern United States. In this paper, the Great Plains 15.1 km, and the average path area was 6.26 km2,about (GP) is considered to be the states of Texas, Oklahoma, 19 times as large as weaktors. This makes significant

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC APRIL 2014 C O L E M A N A N D D I X O N 371 tornadoes much more likely to be detected, as their damage is more severe and they affect a much larger area. Anderson et al. (2007) found that, in most areas, the probability of detection of sigtors in rural areas is greater than that for weaktors, consistent with the above analysis. Also, the number of sigtors in both regions has been steady or even slowly declining since 1973 (Figs. 3c and 3d). The slight decline might be due to more pro- fessional damage evaluations due to the NWS warning verification program and its growth during the 1980s. However, using the linear fit to each chart, the magni- tude of the change is still much larger with weaktors than it is with sigtors. The number of reported weaktors, combining data from both the GP and the southeastern United States, is increasing at the rate of about 6 tor- 2 nadoes per 106 km2 yr 1, while the number of sigtors is 2 decreasing at a rate of 0.7 tornadoes per 106 km2 yr 1. FIG. 4. KDE analysis showing the average annual pathlength (km) of all significant (F2–F5) tornadoes passing within 40 km In summary, even during the ‘‘modern era’’ (1973– (25 mi) of a point between 1973 and 2011. 2011) the number of reported weaktors has changed by nearly an order of magnitude more than the number of sigtors. Also, there is a greater likelihood for the re- since one single F0 tornado occurring in the vicinity of porting of sigtors as opposed to weaktors due to their a point would have the same input to the climatology as longer pathlengths, larger path areas, greater damage, an outbreak of 25 tornadoes, some of them significant, in and other factors discussed above. Their much more the same area on one day. This study makes use of stable reporting pattern since 1973 and their higher a relatively stable set of data (1973–2011 significant likelihood of being reported consistently over time tornadoes), and Dixon et al. (2011) show only minor make sigtors the best dataset to use when analyzing the variations in the spatial patterns of pathlength analyses areas with the greatest tornado risk in the United States. of tornado days and total tornadoes, so all signifi- cant tornado events will be analyzed. Further, use of all c. Pathlength versus tornado count–tornado days sigtors allows for a clearer understanding of the clima- As illustrated by Dixon and Mercer (2012), large tological differences between the GP and southeastern differences in spatial patterns can occur if one uses point United States, as shown by a quick analysis of the de- analysis [i.e., the point of ; e.g., Brooks struction potential index (DPI; Thompson and Vescio et al. (2003)] instead of pathlength analysis (e.g., Dixon 1998; Doswell et al. 2006). DPI is defined as the area 3 et al. 2011). Given that the pathlength of a tornado is covered by the tornado (pathlength path width) multi- much more proportional to the area it affects, its de- plied by the Fujita or enhanced Fujita rating plus one. structive capability, and its overall impact on the risk of Because it combines the area covered with the intensity, a tornado striking an area within its radius of influence in the DPI for a given tornado provides a reasonable mea- the KDE, pathlength analysis is used in this study unless surement of that tornado’s potential impact on society. otherwise noted. In this study, those events in the da- Since 1973, the median DPI per tornado day is 0.18 in the tabase that lacked a termination point or showed iden- GP, but 0.45 in the southeastern United States. There- tical initiation and termination points were altered by fore, using tornado days greatly lessens the modeled adding a termination point 100 m due north of the ini- impact of tornadoes in the southeastern United States tiation. The ultimate justification for this method is to relative to the GP. reduce underestimation of the spatial risk in areas that often experience short-lived and/or slow-moving tor- 4. Objective analysis of tornado risk nadoes, as even a short path is significantly more im- a. Initial analysis pactful than the unrealistic point locations listed in the database. Given that the intensity data on tornadoes greatly Some authors use tornado days as their measure of improved starting in 1973, and sigtors provide the most tornado risk, to minimize the temporal trends in report- stable dataset for analysis of tornado risk, we first ex- ing frequency that we have already addressed. However, amine sigtors over the United States from 1973 through tornado days are not the best measure of tornado risk, 2011 (Fig. 4). In terms of the average annual sum of the

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC 372 WEATHER AND FORECASTING VOLUME 29

FIG. 5. Plot of all significant tornado tracks (and tornado F-scale intensities) from 0600 UTC 27 Apr to 0600 UTC 28 Apr 2011 over the southeastern United States. (Plot constructed using SPC Severe Plot 3.0.) pathlengths of sigtors passing within 40 km (25 mi) of United States (Fig. 5), Fig. 6 shows a KDE of significant a point, there is a broad area extending from much of tornadoes during 1973–2010, excluding the events of the classic tornado alley in the GP eastward into the 2011. The high risk area (.5 km annual pathlength) still Midwest and southeastern United States that experi- extends from to northwest Georgia, ences .3 km of annual pathlength. A greater risk area with small areas of high risk near the Nebraska–Kansas (.5 km of annual pathlength) extends primarily from border, in central Iowa, and southern . The Oklahoma to northwest Georgia, with isolated areas highest risk area (.9 km of annual pathlength) covers in Iowa and near the Nebraska–Kansas border. The maximum risk area (.7 km of annual pathlength) covers the eastern part of the primary risk area, from central Arkansas into central Mississippi and northern Alabama. A small region of even higher risk (.9 km of annual pathlength) extends from near Jackson, Mississippi, to Huntsville and , Alabama. This distribution is quite consistent with that shown in the much smaller dataset by Smith et al. (2012;seeFig. 1d), and fairly consistent with the larger dataset shown by Carbin et al. (2012;seeFig. 1f). Initially, Fig. 4 indicates that the area of greatest tornado risk in the United States extends from Oklahoma to Alabama and Tennessee, with the greatest risk extending from eastern Arkansas into cen- tral Mississippi and northern Alabama. To determine the impact of the single, very large out- break of long-track significant tornadoes on 27 April 2011 (e.g., Knupp et al. 2014) across parts of the southeastern FIG.6.AsinFig. 4, but between 1973 and 2010.

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC APRIL 2014 C O L E M A N A N D D I X O N 373

FIG.7.AsinFig. 4, but showing the average annual total DPI FIG.8.AsinFig. 4, but showing the average annual number of (‘‘Fujita’’ km2). significant (F2–F5) tornado-initiation locations. eastern Arkansas, central Mississippi, and northern 1) NEXRAD radars went online nationwide in the early Alabama, and it seems that the region of maximum 1990s, making the detection of all tornadoes, including tornadoriskindicatedinFig. 4, from Oklahoma to weak ones, more likely. Also since the early 1990s, Alabama and Tennessee, was already apparent in the with the Verification of the Origins of Rotation in data prior to the 2011 tornadoes. Tornadoes Experiment (VORTEX; Rasmussen et al. If one examines the average annual DPI of torna- 1994)andthemovieTwister, tornado chasing has does, the same general area is once again highlighted, become almost ubiquitous during days with tornado as shown in the KDE analysis for DPI (Fig. 7), except risk, especially in the GP. This has also reduced the there is a more prominent maximum centered over likelihood of unreported tornadoes. Furthermore, eastern Mississippi and western Alabama. Even if one trend data have shown that annual weaktors have only considers the number of sigtors, KDE analysis of become fairly steady over the GP since 1992. A starting points of sigtors (without any input from KDE analysis of 1992–2011 tornado pathlength was pathlength; Fig. 8) shows that the region from Okla- performed (not shown). Even if one assumes that homa to Alabama is also the region with the most 10% of all weaktor pathlength over eastern Kansas individual sigtors, but the maximum over Mississippi and eastern Nebraska was associated with under- and Alabama shown in Fig. 4 is not as apparent. This is reported sigtors, adding that pathlength to the re- likely due to the fact that the average pathlength for ported sigtor pathlength still indicates that those a sigtor in the southeastern United States is 18.3 km areas had less tornado risk than most of Oklahoma, (where many more tornadoes occur during the cool sea- and much lower risk than parts of Alabama, Arkan- son and move quickly, covering more ground), signifi- sas, and Mississippi. cantly longer than the average pathlength for a sigtor in 2) Because the average pathlength of a sigtor is more the GP (14.8 km). than 5 times that of a weaktor, the presence of many long-track weaktors in the GP could indicate an b. Lack of bias due to underrated tornadoes in the GP underrating of some sigtors there. A KDE analysis An argument can be made that some of the torna- of 1973–2011 tornadoes with pathlengths greater than does in the GP do not make it into our calculations 16.1 km (10 mi, close to the average pathlength of because, even though they had the wind speeds of a sigtor) shows a very similar pattern to that of sigtors a sigtor, they were rated as weaktors because there was (Fig. 9). An area of moderate long-track tornado risk simply nothing for them to destroy in the thinly populated does extend through eastern Kansas, southeastern open grasslands and fields in some parts of the GP (e.g., Nebraska, and Iowa. This may indicate that some Doswell and Burgess 1988). However, we present three tornadoes in these states are underrated and, there- points that suggest any such bias does not fundamentally fore, not included in the sigtor database. However, change the spatial patterns of tornado risk in the United many of these areas have population densities not States. that different than those is much of Arkansas and

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC 374 WEATHER AND FORECASTING VOLUME 29

frequency may be calculated, and factors applied to each year to simulate a constant detection frequency (assuming the average annual number of actual weaktors was constant from 1992 through 2011). These calculations were performed for both the southeastern United States and the GP, and factors were applied to all weaktors by year in each region to force the detection frequency to be approximately constant. Using the detrended number of weaktors calculated above, and the reported number of sigtors (since sigtors show almost no trend), 8.8% of all tornadoes in the GP from 1992 through 2011 were sigtors, according to the SPC database. Using similar calculations, 12.8% of all tornadoes in the southeast- ern United States during that time period were sigtors. Making the assumption that the 4% difference is due FIG. 9. KDE analysis of tornadoes (1973–2011) with to underreporting of sigtors in the GP (i.e., each pathlengths . 10 mi. region actually has the same ratio of sigtors to all tornadoes), a number of 1992–2011 tornadoes in the Mississippi (Fig. 10), so it is likely that most sigtors GP that were officially rated as weaktors in the SPC with pathlengths greater than 10 mi would be simi- tornado database were artificially increased to sigtor larly likely to damage something and be classified as status. In addition to sigtors, any weaktor that occurred a sigtor. in the GP with a path width greater than 249 yd 3) The detection frequency for weaktors has increased (1 yd 5 0.9144 m) was included as a sigtor in a new since 1992, especially in the southeastern United database of tornadoes from 1992 through 2011. Only States, due to factors mentioned in section 3a. By officially rated sigtors in the southeastern United applying a linear curve fit to the number of weaktors States and the rest of the country were included in during 1992–2011, the rate of increase in detection the database. This way, both the southeastern United

FIG. 10. Population density by county in the United States, 2011 (U.S. Census Bureau 2013).

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC APRIL 2014 C O L E M A N A N D D I X O N 375

States and the GP had a sigtor ratio near 12.8%. A KDE analysis, similar to Fig. 4, was then performed on this database with the added sigtors in the GP (Fig. 11). Since Fig. 11 is quite similar to Fig. 4,itis apparent that underreporting of sigtors in the GP is not a major factor in the KDE analysis of sigtors.

5. Summary and conclusions

Only data since 1973 were used in this study because many tornadoes were overrated before 1973 (section 3a). Only significant tornadoes were included because they are much more likely to be reported and their frequencies have been fairly steady with time since 1973, unlike weak tornadoes (section 3b). Finally, pathlengths were used in KDE analysis as opposed to tornado- FIG. 11. As in Fig. 4, but between 1992 and 2011 and using the lowered criterion for significant tornadoes in the GP (see text). genesis points or tornado days. Our reasoning for fa- voring the use of tornado pathlengths (representing tornado frequency and severity) as opposed to points risk to people due to a tornado. A KDE analysis for DPI of tornadogenesis or tornado days (representing only was performed (Fig. 7), and the region of highest tor- tornado frequency) is explained in section 3c. Given the nado risk from Oklahoma to Alabama is confirmed, with SPC tornado database of sigtors occurring from 1973 a significant maximum extending from central Mis- through 2011, Fig. 4 shows that the region with the sissippi into northern Alabama. greatest risk of tornadoes in the United States is a The tornado risk area outlined in this paper is con- roughly west–east-oriented area, from central Oklahoma, sistent with a trend in findings by several authors in re- through Arkansas and northern Louisiana, western and cent papers who only examine sigtors, yet use a large middle Tennessee, most of Mississippi, and northern dataset covering at least 30 yr (e.g., Concannon et al. and central Alabama. Even upon examination of tor- 2000; see Fig. 1c; Ashley 2007; see Fig. 1e; Carbin et al. nado count, our analysis of the region with the highest 2012; see Fig. 1f). Therefore, we present it as the part of tornado risk does not change appreciably. There is still the United States with the highest risk of tornadoes. This risk for tornadoes in many other areas, especially across area is so clearly different than the tornado alley dis- parts of Kansas, Nebraska, Iowa, , Indiana, and cussed by the media for many years that many Ameri- . No state is totally immune to tornadoes. This cans are totally unaware of the tornado risk in some of study points out that the greatest risk is in the southern the areas outlined in this study. United States. The seemingly legitimate argument that lower pop- Acknowledgments. This research was funded by the ulation density in the GP causes the underrating of National Science Foundation (NSF Award AGS-1110622). tornadoes, and therefore some sigtors to be rated as Reviews of the manuscript by Jon Davies, Greg Forbes, weaktors, is discussed and mostly discounted in section Kevin Knupp, Ryan Wade, and Todd Murphy were very 4c on the basis of changing some weaktors to sigtors helpful and improved the manuscript greatly. in eastern Kansas and eastern Nebraska, and distribu- tion of moderate- to long-track tornadoes. In addition, REFERENCES when reporting trends are mathematically removed for weaktors and many weaktors in the GP are artificially Anderson, C. J., C. K. Wikle, Q. Zhou, and J. A. Royle, 2007: Population influences on tornado reports in the United States. inflated to sigtor status, forcing the ratio of sigtors to all Wea. Forecasting, 22, 571–579. tornadoes to be the same in the southeastern United Ashley, W. S., 2007: Spatial and temporal analysis of tornado fa- States and the GP, the same general region of risk talities in the United States: 1880–2005. Wea. Forecasting, 22, shown in Fig. 4 is highlighted. Also, the destruction 1214–1228. potential index (DPI) of tornadoes in the southeastern Britt, M. F., and F. H. Glass, cited 2013: The tornado climatology of the St. Louis Weather Forecast Office County Warning United States is much higher, on average. The DPI for Area. National Weather Service Weather Forecast Office, a given tornado provides a reasonable measurement of St. Louis, MO. [Available online at http://www.crh.noaa.gov/ a tornado’s potential impact on society, and the overall lsx/?n5torcli_data.]

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC 376 WEATHER AND FORECASTING VOLUME 29

Brooks, H. E., 2004: On the relationship of tornado path length and Kelly, D. L., J. T. Schaefer, R. P. McNulty, C. A. Doswell III, and width to intensity. Wea. Forecasting, 19, 310–319. R. F. Abbey, 1978: An augmented tornado climatology. Mon. ——, and J. P. Craven, 2002: A database of proximity soundings for Wea. Rev., 106, 1172–1183. significant severe , 1957–1993. Preprints, 21st Knupp, K. R., and Coauthors, 2014: Meteorological overview of Conf. on Severe Local Storms, San Antonio, TX, Amer. Me- the devastating 27 April 2011 . Bull. Amer. teor. Soc., 16.2. [Available online at https://ams.confex.com/ Meteor. Soc., in press. ams/pdfpapers/46680.pdf.] Kunkel, K. E., and Coauthors, 2013: Monitoring and understanding ——, C. A. Doswell III, and M. P. Kay, 2003: Climatological esti- trends in extreme storms: State of knowledge. Bull. Amer. mates of local daily tornado probability for the United States. Meteor. Soc., 94, 499–514. Wea. Forecasting, 18, 626–640. Marsh, P. T., and H. E. Brooks, 2012: Comments on ‘‘Tornado risk Carbin, G., P. Heinselman, and D. Stensrud, 2012: Current chal- analysis: Is Alley an extension of Tornado Alley?’’ Bull. lenges in tornado forecast and warning. Workshop on Weather Amer. Meteor. Soc., 93, 405–407. Ready Nation: Science Imperatives for Severe McCarthy,D.W.,andJ.T.Schaefer, 2004: Tornado trends over the Research, M. K. Lindell and H. Brooks, Eds., NOAA–NSF, past thirty years. Preprints, 14th Conf. on Applied Climatology, A-9–A-16. Seattle, WA, Amer. Meteor. Soc., 3.4. [Available online at https:// Coleman, T. A., K. R. Knupp, J. Spann, J. B. Elliott, and B. E. ams.confex.com/ams/84Annual/techprogram/paper_72089.htm.] Peters, 2011: The history (and future) of McNulty, R. P., 1995: Severe and convective weather: A Central dissemination in the United States. Bull. Amer. Meteor. Soc., Region forecasting challenge. Wea. Forecasting, 10, 187– 92, 567–582. 202. Concannon, P. R., H. E. Brooks, and C. A. Doswell III, 2000: Meyer, C. L., H. E. Brooks, and M. P. Kay, 2002: A hazard model Climatological risk of strong to violent tornadoes in the for tornado occurrence in the United States. Preprints, 16th United States. Preprints, Second Symp. on Environmental Conf. on Probability and Statistics in the Atmospheric Sciences/ Applications, Long Beach, CA, Amer. Meteor. Soc., 212–219. 13th Symp. on Global Change and Climate Variations, Orlando, de Smith, M. J., M. F. Goodchild, and P. A. Longley, 2007: Geo- FL, Amer. Meteor. Soc., J3.6. [Available online at https://ams. spatial Analysis: A Comprehensive Guide to Principles, Tech- confex.com/ams/pdfpapers/27595.pdf.] niques and Software Tools. 3rd ed. Troubador Publishing, 516 pp. O’Sullivan, D., and D. J. Unwin, 2003: Geographic Information Dixon, P. G., and A. E. Mercer, 2012: Reply to ‘‘Comments on Analysis. J. Wiley and Sons, 436 pp. ‘Tornado risk analysis: Is an extension of Tornado Rasmussen, E. N., J. M. Straka, R. Davies-Jones, C. A. Doswell III, Alley?’’’ Bull. Amer. Meteor. Soc., 93, 408–410. F. H. Carr, M. D. Eilts, and D. R. MacGorman, 1994: Verifi- ——, ——, J. Choi, and J. S. Allen, 2011: Tornado risk analysis: cation of the Origins of Rotation in Tornadoes Experiment: Is Dixie Alley an extension of Tornado Alley? Bull. Amer. VORTEX. Bull. Amer. Meteor. Soc., 75, 995–1006. Meteor. Soc., 92, 433–441. Schaefer, J. T., 1986: Severe thunderstorm forecasting: A historical Doswell, C. A., III, and D. W. Burgess, 1988: On some issues of perspective. Wea. Forecasting, 1, 164–189. United States tornado climatology. Mon. Wea. Rev., 116, 495–501. ——, and R. Edwards, 1999: The SPC tornado/severe thunder- ——, A. R. Moller, and H. E. Brooks, 1999: Storm spotting and storm database. Preprints, 11th Conf. on Applied Climatology, public awareness since the first tornado forecasts of 1948. Wea. , TX, Amer. Meteor. Soc., 6.11. [Available online at Forecasting, 14, 544–557. https://ams.confex.com/ams/99annual/abstracts/1360.htm.] ——, R. Edwards, R. L. Thompson, J. A. Hart, and K. C. Crosbie, ——, and R. S. Schneider, 2002: The robustness of tornado hazard 2006: A simple and flexible method for ranking estimates. Preprints, Third Symp. on Environmental Appli- events. Wea. Forecasting, 21, 939–951. cations, Orlando, FL, Amer. Meteor. Soc., 4.2. [Available Edwards, R., J. G. LaDue, J. T. Ferree, K. Scharfenberg, C. Maier, online at https://ams.confex.com/ams/pdfpapers/27694.pdf.] and W. L. Coulbourne, 2010: The enhanced Fujita scale: Past, ——, D. L. Kelly, C. A. Doswell, J. G. Galway, R. J. Williams, R. P. present, and future. Preprints, 25th Conf. on Severe Local McNulty, L. R. Lemon, and B. D. Lambert, 1980: Tornadoes, Storms, Denver, CO, Amer. Meteor. Soc., 4A.1. [Available when, where, how often. Weatherwise, 33, 52–59. online at https://ams.confex.com/ams/pdfpapers/175400.pdf.] Schultz, L. A., and D. J. Cecil, 2009: Tropical cyclone tornadoes, Fujita, T. T., 1971: Proposed characterization of tornadoes and 1950–2007. Mon. Wea. Rev., 137, 3471–3484. hurricanes by area and intensity. SMRP Research Paper 91, Silverman, B. W., 1986: Density Estimation for Statistics and Data University of Chicago, Chicago, IL, 42 pp. [Available from Analysis. Chapman and Hall, 175 pp. Wind Engineering Research Center, Box 41023, Lubbock, Smith, B. T., R. L. Thompson, J. S. Grams, C. Broyles, and H. E. TX 79409.] Brooks, 2012: Convective modes for significant severe thun- Gagan, J. P., A. Gerard, and J. Gordon, 2010: A historical and derstorms in the contiguous United States. Part I: Storm statistical comparison of ‘‘Tornado Alley’’ to ‘‘Dixie Alley.’’ classification and climatology. Wea. Forecasting, 27, 1114– Natl. Wea. Dig., 34 (2), 145–155. 1135. Kay, M. P., and H. E. Brooks, 2000: Verification of probabilistic Thompson, R. L., and M. D. Vescio, 1998: The destruction po- severe storm forecasts at the SPC. Preprints, 20th Conf. on tential index—A method for comparing tornado days. Pre- Severe Local Storms, Orlando, FL, Amer. Meteor. Soc., 9.3. prints, 19th Conf. on Severe Local Storms, Minneapolis, MN, [Available online at https://ams.confex.com/ams/Sept2000/ Amer. Meteor. Soc., 280–282. techprogram/paper_15921.htm.] U.S. Census Bureau, cited 2013: Population density for counties Kellogg, B., and G. S. Forbes, cited 2013: Top 10 tornado cities. The and Puerto Rico municipios: July 1, 2011. [Available online at Weather Channel. [Available online at http://www.weather. https://www.census.gov/popest/data/maps/2011/PopDensity_11. com/news/tornado-central/top-tornado-cities-2013-20130415.] pdf.]

Unauthenticated | Downloaded 09/30/21 02:21 PM UTC