5218 J O U R N A L O F C L I M A T E VOLUME 24 An Assessment of the Climatology of Florida Hurricane-Induced Tornadoes (HITs): Technology versus Meteorology ERNEST M. AGEE AND ALYSSA HENDRICKS Department of Earth and Atmospheric Sciences, Purdue University, West Lafayette, Indiana (Manuscript received 3 May 2011, in final form 11 August 2011) ABSTRACT The placement of operational Doppler radar at National Weather Service offices in Florida in 1994 and 1995 has produced a distinct climatological discontinuity in archived records of hurricane-induced tornado (HIT) events. During the period of this study (1979 2010), Florida experienced 91 named tropical cyclones (omitting years 1994 and 1995) that produced a total of 300 HITs, as recorded in the NOAA Storm Data climatological publication. The 30 years of records examined for 1979 2010 consisted of 15 years as the pre- Doppler period (1979 93) and 15 years as the Doppler period (1996 2010) and produced an average of 1.92 HITs per tropical cyclone for the pre-doppler period and 3.85 HITs per tropical cyclone for the Doppler period. Additionally, the HIT statistics were further compiled as the number of HITs per potential HIT day, which yielded values of 0.52 and 1.14, respectively, for the pre-doppler and Doppler periods. Normalized statistics are based on the total accumulated number of potential HIT days for each period (96 for pre- Doppler and 219 for Doppler). Further, it has been determined that (a) no HITs were recorded in the pre- Doppler period for hurricanes greater than category-2 intensity, ostensibly because of the inability to separate hurricane wind damage from tornado damage, whereas (b) the Doppler period averaged 5.0 and 11.7 HITs per category-3 and category-4 storms, respectively. Finally, this study has clearly documented the magnitude (and effect) of Doppler technology on HITs reported in Florida Storm Data archives. The Doppler era is much more accurate, whereas the pre-doppler era HIT occurrences are severely underestimated. 1. Introduction This study has been undertaken in part based on the prior work by Belanger et al. (2009), which gave a nice treatment of variability in tornado frequency associated with landfalling tropical cyclones (TCs) along the Gulf Coast region of the United States. One aspect of their study pertained to the role in the implementation of the National Weather Service (NWS) Doppler radar network, and its effect on tornado frequency archives associated with landfalling TCs. The study reported on here has focused more specifically on the role of Doppler technology and associated effects of such on hurricaneinduced tornadoes (HITs) 1 found in the archived records. 1 HITs are defined to include tornadoes produced by both tropical storms (category 0) and hurricanes (categories 1 5). Corresponding author address: Ernest M. Agee, Department of Earth and Atmospheric Sciences, Purdue University, 550 Stadium Mall Drive, West Lafayette, IN 47907 2051. E-mail: eagee@purdue.edu Any study that addresses HIT climatology must determine how to define the count of events (e.g., landfalling TCs or not), and how far inland from the coast to count HITrelated tornado events [e.g., see Storm Data reports for Tropical Storm Hermine (2010) over the Dallas area]. This latter point (of how far inland) is further illustrated in the study by Schultz and Cecil (2009). The study reported on here has made a more definitive analysis of the role of technology by choosing the state of Florida, and therefore counting not only all landfalling TCs and HITs, but also all potential HIT days that occurred over Florida, as well as HIT opportunities when only a portion of the TC affected the state (even if the storm center remained offshore). Further, being an elongated north south peninsula state located in the hurricane alley of North Atlantic TCs, there were ample data for this study, and there were no issues regarding distance from the coastline (since all of Florida was considered). Two Doppler radars were installed in Florida in 1994, and four additional radars were installed in 1995; for purposes of this study 30 years of data have been collected and assembled into two data periods (15 pre-doppler DOI: 10.1175/JCLI-D-11-00235.1 Ó 2011 American Meteorological Society
15 OCTOBER 2011 A G E E A N D H E N D R I C K S 5219 FIG. 1. Annual trend in the number of named tropical cyclones for both the North Atlantic tropical ocean and Florida, for the period of this study (1979 2010). The 15 pre-doppler years for Florida are 1979 93 (period I), and the 15 Doppler years are 1996 2010 (period II). years, 1979 93, in period I and 15 Doppler years, 1996 2010, in period II). TCs that affected Florida were considered to be both landfall events and storm centers within 200 n mi of the Florida coast (when a portion of the storm was over land, including spiral feeder bands). This study began by examining the trend in the annual occurrences of named tropical cyclones for both the tropical North Atlantic and Florida from 1979 to 2010 (see Fig. 1). As seen, Florida is a state that is a reasonable mirror image of the North Atlantic tropical storm/hurricane seasonal pattern. It is further evident that the number of tropical cyclones has increased between periods I and II; however, the data have been normalized to validate comparative statistics. This normalization was achieved FIG. 2. Annual variations in TC activity for the tropical North Atlantic, parsed into weaker events (categories 0, 1, and 2) vs stronger hurricanes (categories 3, 4, and 5).
5220 J O U R N A L O F C L I M A T E VOLUME 24 TABLE 1. Tabulation of pertinent TC and HIT statistics for pre-doppler (I) and Doppler (II) periods. Potential HIT days is denoted as PHD. Note that the difference between the means of the two periods is large, although no test of statistical significance has been performed. Florida TCs and HIT data I II Potential HIT days 96 219 No. of TCs 26 65 No. of HITs 50 250 HITs per TC 1.92 3.85 HITs per PHD 0.52 1.14 by determining the exact number of potential HIT days for periods I and II, and thus statistics could be presented for both HITs per TC (for all storm categories 0 5) as well as HITs per potential HIT day per TC (as presented and discussed in the next section). Further, Fig. 2 is presented to show a longer background trend of North Atlantic tropical cyclone activity, from 1950 to 2010. The events in Fig. 2 were separated into two classes (the three lowest scale intensities and the three highest scale intensities, the same grouping used in this study), which shows an increasing trend for the weaker tropical cyclone events and a more cyclic pattern to the intense hurricanes. This study takes place within a time period of rising trends, as seen in the longer record. 2. HITs per TC and HITs per potential HIT day As noted in Table 1 (derived from the raw dataset presented in Table 2), the samples for both data periods (I and II) are sufficiently large to render valid statistics. Accordingly, there are 96 and 219 potential hit days, respectively, for periods I and II. The potential HIT day is determined by the presence of a TC, as defined and discussed in the introduction, which may or may not have a recorded HIT event. The improvement in HIT archives (due to Doppler radar technology) is most evident when comparing HITs per TC or HITs per potential HIT day, TABLE 2. Summary of collected statistics for the pre-doppler period (1979 93) and the Doppler period (1996 2010); 1994 and 1995 are listed but are not included in the calculation, and the boldface for those years indicates that they are the years during which the Doppler radars were installed. Year TC name(s) Potential HIT days HITs 1979 Bob, David, Frederick 12 17 1980 0 0 1981 Dennis 5 2 1982 Alberto 12 2 1983 Barry 2 0 1984 Diana, Isidore 8 0 1985 Bob, Elena, Isabel, Juan, Kate 25 18 1986 Andrew, Charley 4 1 1987 Arlene, Floyd 4 2 1988 Beryl, Chris, Florence, Keith 9 6 1989 Hugo 3 0 1990 Marco 2 2 1991 Fabian 2 0 1992 Andrew, Earl 8 0 1993 0 0 1994 Alberto, Beryl, Gordon 12 7 1995 Allison, Erin, Jerry, Opal 12 27 1996 Arthur, Bertha, Josephine, Lili 11 23 1997 Danny 4 0 1998 Earl, Georges, Hermine, Mitch 12 29 1999 Dennis, Floyd, Harvey, Irene 14 9 2000 Gordon, Helene, Leslie 6 17 2001 Allison, Barry, Gabrielle, Michelle 14 28 2002 Bertha, Edouard, Hanna, Isodore, Kyle 20 7 2003 Bill, Erika, Henri 8 1 2004 Alex, Bonnie, Charley, Frances, Gaston, Ivan, Jeanne, Matthew 32 74 2005 Arlene, Cindy, Dennis, Franklin, Katrina, Ophelia, Rita, Tammy, Wilma 35 27 2006 Alberto, Ernesto 8 9 2007 Andrea, Barry, Humberto, Noel, Olga 11 2 2008 Cristobal, Edouard, Fay, Gustav, Hanna, Ike, Paloma 29 23 2009 Claudette, Danny, Ida 8 1 2010 Bonnie, Nicole, Paula 7 0
15 OCTOBER 2011 A G E E A N D H E N D R I C K S 5221 FIG. 3. Annual distribution of HITs per potential HIT days (and per TC) for pre-doppler (I) and Doppler (II) periods. The two vertical lines capture the events for 1994 and 1995, the installation years (which were omitted in this study). The discontinuity between the two periods is remarkable, illustrating the importance of Doppler radar in confirming HIT events. for period I versus period II. Specifically, the numbers are 1.92 and 3.85 HITs per TC, respectively, for periods I and II. Further refinement in efforts to better normalize the statistics, based on HITs per potential HIT day (0.52 for period I and 1.14 for period II), continue to show the increase and improvement in the HIT archives in Storm Data in the presence of NWS Doppler radar. Figure 3 is presented to show the annual distribution of HITs per TC and HITs per potential HIT day for periods I and II. The heterogeneity in the record is remarkable, and the magnitude of the effect that Doppler radar has for period II clearly shows the failure to archive HIT events in the pre- Doppler period. Even the installation years, especially 1995 when all six Doppler radars were in place, show the increased detection (see Fig. 3). 3. HITs in major hurricanes Another intriguing result found in this study pertains to HITsperTCbycategory,forperiodsIandII.Figure4 shows a bar graph distribution of these results, which reveals no HITs from major hurricanes (. category 2) for period I and averages of 5.0 and 11.7, respectively, for category-3 and -4 events in period II. There were a total of 5 potential HIT days in period I and 26 potential HIT days in period II for all category-3 and -4 hurricanes. It is reasonable to assume that in the absence of Doppler radar, it is extremely difficult to identify tornado damage within the damage field of major hurricanes. It is further noted that the Doppler data period shows an increase in the number of HITs per TC when moving from the weaker tropical cyclones to the stronger hurricanes. More intense convection in the spiral feeder bands, as well as increased shear vortices along the hurricane eyewall, would be possible explanations for the increased HIT activity with higher-category hurricanes. 4. Summary and conclusions This study has clearly shown the magnitude (and effect) of Doppler radar on the record of HITs in Florida Storm Data archives. Specifically, the pre-doppler-era archives are seen to grossly underestimate the actual HIT occurrences, whereas the Doppler-era records are perceived as more accurate. The 30 years of records examined for 1979 2010 (omitting 1994 and 1995) consisted of 15 years as the pre-doppler period (1979 93) and 15 years as the Doppler period (1996 2010) and produced an average of 1.92 HITs per tropical cyclone for the pre-doppler period and 3.85 HITs per tropical
5222 J O U R N A L O F C L I M A T E VOLUME 24 FIG. 4. Histogram distribution of HITs per TC category for pre-doppler (I) and Doppler (II) periods. For major hurricanes, no HITs were recorded in period I during 5 potential HIT days, yet 26 potential HIT days produced 26 HITs for category 3 and 35 HITs for category 4. This produced an average of 5.0 and 11.7 HITs per category 3 and category 4, respectively. The Doppler era also shows a trend of increasing HITs with increasing hurricane intensity. cyclone for the Doppler period. Additionally, the HIT statistics were further compiled as the number of HITs per potential HIT day, which yielded values of 0.52 and 1.14, respectively, for the pre-doppler and Doppler periods. Normalized statistics account for all named tropical storms that affected Florida and its coastal environment, and the HIT averages are based on the total accumulated number of potential HIT days for each period (96 for pre- Doppler and 219 for Doppler). Further, it has been shown that major hurricanes in the pre-doppler era have no recorded HITs whereas the Doppler era shows a trend of increasing HITs with increasing hurricane intensity. The latter is expected since more intense hurricanes should have more intense convective feeder bands, as well as a greater opportunity for shear vortices associated with strong annulus shear region inside the eyewall. REFERENCES Belanger,J.I.,J.A.Curry,andC.D.Hoyos,2009:Variabilityin tornado frequency associated with U.S. landfalling tropical cyclones. Geophys. Res. Lett., 36, L17805, doi:10.1029/ 2009GL040013. Schultz, L. A., and D. J. Cecil, 2009: Tropical cyclone tornadoes, 1950 2007. Mon. Wea. Rev., 137, 3471 3484.