Mid-Level Dry Intrusions as a Factor in Tornado Outbreaks Associated with Landfalling Tropical Cyclones from the Atlantic and Gulf of Mexico

Lon Curtis
KWTX-TV
Waco, Texas

(as published in Weather and Forecasting, Vol. 19, No. 2)



AMS Copyright Notice

© Copyright 2004 American Meteorological Society (AMS). Permission to use figures, tables, and brief excerpts from this work in scientific and educational works is hereby granted provided that the source is acknowledged. Any use of material in this work that is determined to be "fair use" under Section 107 or that satisfies the conditions specified in Section 108 of the U.S. Copyright Law (17 USC, as revised by P.L. 94-553) does not require the Society's permission. Republication, systematic reproduction, posting in electronic form on servers, or other uses of this material, except as exempted by the above statements, requires written permission or license from the AMS. Additional details are provided in the AMS Copyright Policies, available from AMS at 617-227-2425 or amspubs@ametsoc.org. ### Permission to place a copy of this work on this server has been provided by the AMS. The AMS does not guarantee that the copy provided here is an accurate copy of the published work.


ABSTRACT

    Mid-level dry intrusions have been mentioned as a factor in tornado outbreaks associated with landfalling tropical cyclones for more than three decades, but a systematic analysis of historical outbreak cases with respect to this pattern has been missing.  Herein, following an examination of all named tropical cyclone landfalls from the Atlantic and Gulf of Mexico for the period 1960-1999, thirteen “outbreak” cases (defined as storms producing twenty or more tornadoes) have been identified and eleven of the cases offering clear evidence of a dry intrusion at mid-levels over the outbreak area. The outbreaks all occurred when the favored area for tornadogenesis (the right-front or northeast quadrant of the storm) coincided with a pronounced gradient of relative humidity, with the gradient best reflected at the 700 hPa and/or 500 hPa levels. Two distinct patterns were identified with respect to the origin of the mid-level dry air involved in these cases: in one, a mass of dry air that impinged on much of the northern or northwestern semi-circle of the storm's outer circulation gradually divided into two separate areas (one to the northwest and the other to the northeast of the storm center) as the storm advanced; the other involved ingestion of a lobe of dry air from a reservoir most often located in the eastern semi-circle of the storm. The outbreak cases were determined to have a strong diurnal signal detected in the temporal distribution of the tornadoes, as almost two-thirds (65%) of the tornadoes in the eleven "positive" cases occurred between 15 UTC and 24 UTC (daylight hours in the region examined). A review of all tropical cyclone landfall events since 1960 revealed that there are only three cases with clear evidence of a mid-level dry intrusion which failed to produce an outbreak as defined herein.


Notes: (a) Minor editorial variations may exist between the published and this on-line version. In case of any such variations, the published version is deemed to be the official source. (b) In the on-line version, footnotes appear at the bottom of the section in which they occur.


1. INTRODUCTION

    The potential for landfalling tropical cyclones (hereafter, TCs) to produce tornadoes has been previously documented. Some storms produce tornado outbreaks while others produce few or none. Vescio et al. (1996) noted that mid-level dry intrusions have the potential to substantially alter the thermodynamic structure of the TC environment (with significant enhancement of CAPE and surface-based instability), and suggested it would be beneficial to document future tornado outbreaks to better understand the role of dry intrusions in the outbreaks. Examination of historical outbreak cases yields important information and has the advantage of providing improved understanding of the role of mid-level dry intrusions without awaiting future storms.

    Thirteen “outbreak” cases, defined herein as storms producing twenty or more tornadoes, for TCs making landfall from the Atlantic and Gulf of Mexico have been examined for evidence of mid-level dry intrusions. This definition is consistent with the “large outbreak” definition found in Galway (1975, 1977) as applied to all tornado outbreaks.¹  Eleven cases were found to offer clear evidence of a dry intrusion at mid-levels over the outbreak area. The outbreaks all occurred when the favored area for tornadogenesis (the right-front or northeast quadrant of the storm) coincided with a pronounced gradient of relative humidity, with the gradient best reflected at the 700 hPa and/or 500 hPa levels.

    Two distinct patterns were identified with respect to the origin of the mid-level dry air involved in these cases. In one, a mass of dry air that impinged on much of the northern or northwestern semi-circle of the storm's outer circulation became divided into two lobes as the storm advanced, with one to the northwest and the other to the northeast of the storm center. In the other, dry air from a reservoir most often located in the eastern semi-circle of the storm was ingested into the storm’s circulation. The eleven "positive" outbreak cases had a strong diurnal signal detected in the temporal distribution of the tornadoes; almost two-thirds (65%) of the tornadoes occurred between 15 UTC and 24 UTC (daylight in the region examined).

¹ McCaul (1991) classified tropical cyclone tornado outbreaks as minor (not more than 8), major (9 to 23 tornadoes) or severe (more than 24 tornadoes). A review of McCaul’s Table 1 (not shown) reveals that 124 tornadoes were produced by the eleven storms producing 10 to 19 tornadoes; 278 tornadoes were produced by the seven storms producing 20 or more tornadoes.



2. BACKGROUND

2.1 Factors in Tornado Location:
    Two areas within TCs produce most of the reported tornadoes: near the core of the storm and in the outer rainbands (Gentry 1983; Weiss 1987). Although Gentry said that most tornadoes occur in areas close to the coastline and Novlan and Gray (1974) asserted that nearly all hurricane-induced tornadoes occur within 200 km of the coast, it is now clear that the threat can extend well inland.

    Novlan and Gray (1974) found that the centroid of tornadogenesis is 240 km northeast (i.e. 50 degrees east of due north) from the storm center. Gentry (1983) noted that the tipping term² of the vorticity equation is twice as large in the northeast quadrant of a landfalling storm. McCaul (1991) examined 1,296 proximity soundings from TCs producing tornadoes and found that ambient helicity is maximized in the right-front quadrant. Bogner et al. (2000) examined dropwindsonde data from six Atlantic storms while over open water and found that the right-front quadrant (relative to storm motion) contained the greatest speed and directional shear.

***

2.2 Factors Leading to Tornadogenesis:

    Various factors have been proposed to explain tornadogenesis in landfalling TCs. Novlan and Gray (1974) noted that speed shear is pronounced in areas within 200 km of the coast as a storm makes landfall. They estimated a 20 m s-1 increase in wind velocity from the surface to 850 hPa. Gentry (1983) noted that surface wind speeds over land areas were reduced by friction while the flow at slightly higher levels remained strong and was not affected by friction.

    Studies since the deployment of the WSR-88D radars indicate that many of the tornadoes are related to intense, persistent cells within the outer rainbands (Spratt et al. 1997). These cells develop mesocyclones, although identifying the mechanism by which this occurs has proven elusive. McCaul and Weisman (1996) simulated shallow supercell storms in landfalling hurricane environments and found that while the wind shear profiles often favor the development of storms with rotating updrafts, low-level mesocyclogenesis may be hampered by a relative lack of strong storm-induced baroclinity in the simulated storms.

    Meso-beta scale boundaries can be an important factor in tornadogenesis associated with supercellular storms (Markowski et al. 1998). TC outer rainband cells that develop mesocyclones have been described as mini-supercells (Spratt et al. 1997; McCaul and Weisman 1996) and tornadoes have been documented as the cells in TC outer rainbands and other peripheral mini-supercells interact with baroclinic boundaries (Edwards et al. 2000).

***

2.3 Significance of Mid-Level Dry Intrusions:
    The traditional TC paradigm is a system dominated by barotropic influences. The barotropic structure of the typical TC appears to limit CAPE to levels well below those normally associated with deep convective storms on the Great Plains (McCaul 1991). In apparent contrast to McCaul’s findings, Bogner et al. (2000) examined dropwindsonde data from six Atlantic hurricanes over open ocean, and found CAPE of around 1500–1700 J kg-1 at 300 km or more from the storm center and found little variation in CAPE by storm quadrant.

    The origin of baroclinity in an otherwise barotropic system has been the subject of investigation, informed speculation, and some consternation for at least the past four decades. Discussing convective instability as a “concomitant ingredient” for tornado formation in hurricanes, Hill et al. (1966) noted that the notion of relatively dry air at intermediate levels aloft is difficult to reconcile with the generally held view that the TC consists of a roughly uniform air mass both vertically and horizontally that is at or near saturation.

    Novlan and Gray (1974) specified “significant dry air intrusions in the right rear quadrant … at 850-700 mb” as an indicator for potential tornado “family” outbreaks. McCaul (1987) noted that dry air advected into a landfalling storm’s mid-levels may influence the structure of the storm, and that mid-level dry air intrusions are fed by a   large-scale reservoir of dry air outside the hurricane that may eventually penetrate quite close to the center of a storm, causing local increases in the lapse rate, thereby contributing to an increase in convective instability. Thus, the notion that mid-level dry intrusions may occur in TCs is not of recent origin.

    Cione et al. (2000) examined the "conventional wisdom" that differences between sea surface temperatures (SSTs) and surface air temperatures (TAs) within the hurricane environment are small and do not vary much as a function of distance from the storm center. Using 7,800 individual observations from the National Data Buoy Center's moored and drifting buoys, they developed a database of 153 time series from 37 hurricanes over a 23-year period (1975-1998). They found a number of events where TAs were significantly cooler than SSTs, and concluded that these appeared to be related to concurrent episodes of low-level drying.

***
² The preferred terminology is “twisting” term, but tilting term and tipping term are also used. The process describes a term of the vorticity equation that represents generation of vertical vorticity by the twisting of horizontal vorticity into the vertical through the agency of shear in the vertical velocity. (Glickman, 2000)
³ These over-water values are of unknown relevance for TCs that have made landfall, with the possibility of several factors altering the kinematic and thermodynamic structure of the TC post-landfall.


3. EXAMINATION OF OUTBREAK CASES

    For purposes of this study, outbreak cases are defined as those occurring subsequent to 1960 and producing 20 or more tornadoes with landfall on the Atlantic or Gulf coast (including the northern Mexican coast because of its proximity to Texas). The storms meeting these criteria are: Agnes (1972), Alicia (1983), Allen (1980), Andrew (1992), Beryl (1994), Beulah (1967), Carla (1961), David (1979), Danny (1985), Georges (1998), Gilbert (1988), Josephine (1996), and Opal (1995). Beryl and Josephine were tropical storms; all others were hurricanes. Table 1 contains a list of the storms, the number of significant tornadoes produced by each, and the casualties which resulted.

    The location of each tornado was ascertained utilizing SVRPLOT2 (Hart and Janish 1999), Significant Tornadoes 1880–1989 (Grazulis 1993) and Storm Data (National Climatic Data Center 1960–99). All thirteen storms were examined in the same manner. Rawinsonde data at 500hPa, 700 hPa and 850 hPa for the period commencing 36 hours prior to landfall through 36 hours following landfall was examined. Analysis involved a smoothed plot of relative humidity at those levels for evidence of a mid-level dry intrusion. This was accomplished using Digital Atmosphere software (Vasquez 2000) applying a standard Barnes analysis scheme utilizing a 2-pass filter. Skew-T/Log-P diagrams were created using RAOB for Windows (Skewchuk 2002).


TABLE 1.Tropical cyclones since 1960 producing outbreaks of 20 or more tornadoes with landfall on the Atlantic or Gulf coast (including the northern Gulf coast of Mexico).
Storm (Year)
Number of tornadoes
Tornadoes with F2 damage
Tornadoes with F3 or greater damage
Deaths
Injuries
Agnes (1972)
30
7
2
0
84
Alicia (1983)
22
1
0
0
0
Allen (1980)
35
10
1
0
31
Andrew (1992)
56
1
1
2
49
Beryl (1994)
35
8
3
0
58
Beulah (1967)
117
5
6
5
41
Carla (1961)
20
6
7*
14*
335*
Danny (1985)
42
9
5
1
63
David (1979)
34
13
2
2
31
Georges (1998)
42
1
0
0
36
Gilbert (1988)
41
3
0
1
10
Josephine (1996)
30
2
0
0
1
Opal (1995)
32
2
0
1
6
Totals
536
68
27
26
745
*Includes one tornado that produced F4 damage at Galveston as a large waterspout moved onshore; most deaths and injuries were associated with that event.


3.1 Hurricane Agnes (1972):
    Hurricane Agnes moved northward across the eastern Gulf of Mexico and made initial landfall over the Florida panhandle on 19 June as a Category 1 storm4. The remnants moved northeastward and back over the Atlantic on 21 June, after which the storm made a second landfall on 22 June near New York City.

    Prior to the initial landfall, Agnes produced many tornadoes as the right front quadrant of the storm moved northward over the Florida peninsula. Reports contemporaneous with the event attributed 15 tornadoes to Agnes but recent reanalysis of the historical data (applying current Storm Data standards) indicates more than 20 tornadoes occurred in Florida (Hagemeyer and Spratt 2002). No additional tornadoes were documented with the second Agnes landfall in southeastern New York.

    At least 28 tornadoes occurred as Agnes tracked northward off the west coast of Florida on 18 June. Following six pre-dawn tornadoes in the Florida Keys, there was a lull until the afternoon, then 19 tornadoes occurred between 18UTC and 03 UTC over the central peninsula, suggesting a strong diurnal influence. A pronounced relative humidity gradient is seen with dry air to the north and moist air to the south in an analysis of 500 hPa relative humidity data at 12 UTC on the morning of 18 June within the outbreak area (Fig. 1). During the ensuing twelve hours, the midtroposphere moistened considerably as the strong southerly flow advected moist air northward.

    This event can be described as an extended-landfall situation in that the track of the storm produced an extended period in which conditions favored tornadogenesis (in this case, across the central Florida peninsula) as the favored quadrant remained over land for an extended period without the storm center making landfall. The 00 UTC upper air data from both Tampa and Miami (not shown) exhibited sufficient storm-relative helicity (hereafter, SRH) for rotating updrafts, with 0-3 km values at both Tampa and Miami exceeding 280 m2 s-2 (Droegemeier et al. 1993). It is unlikely that the dry air ahead of Agnes is responsible for all of the tornadoes, but the occurrence of daytime tornadoes in the gradient between dry air and moist air will be seen in the ten additional qualifying cases reviewed below.  Based on the evidence at hand, this appears to be a case where the outbreak can be sufficiently attributed to a mid-level dry intrusion to justify inclusion of this case.

4All references to storm category are based on Saffir and Simpson (1974).



3.2 Hurricane Allen (1980):
   Hurricane Allen passed near the Yucatan peninsula on 7 August and made landfall on the lower Texas coast near Brownsville about 06 UTC on 10 August, then continued moving to the west-northwest. Allen was a Category 5 storm 400 km east-southeast of Brownsville but had weakened to Category 3 at the time of landfall.

    The storm produced numerous tornadoes across south central Texas on 10  August. Stiegler and Fujita (1982) surveyed the San Marcos, Texas tornado (one of several in the Austin-San Marcos area) and found a 12.5 km long damage path with evidence of multiple vortices and some F3 damage5. Another strong tornado (damage rated F2) caused $50 million in damage at the Austin Municipal Airport about the same time. At least 16 of Allen’s 29 tornadoes occurred in central Texas, in both the right-front quadrant and the northeast quadrant, some 250 to 400 km from the storm center.

    Examination of the sounding data revealed a pronounced dry intrusion that appears to have been entrained into Allen’s circulation from the east and southeast. Best depicted at the 700 hPa level (Fig. 2), the intrusion rotated westward over a period of 24 to 36 hours, producing a very sharp north-south gradient of relative humidity above the Austin-San Antonio area during the day on 10 August. The intense mid-level gradient  coincides with the outbreak area.

5 All F-scale references are based on Fujita, 1981.



3.3 Hurricane Andrew (1992):
    Hurricane Andrew made landfall twice. On 24 August, the storm crossed southern Florida as a Category 5 storm (Landsea et al., 2003, manuscript submitted to Bull. Amer. Meteor. Soc.), then turned northwest and made a second landfall as a Category 1 storm on the Louisiana coast before sunrise on 26 August. The storm turned to the north and then northeast and passed near Tupelo, MS around 17 UTC on the 27 August.

    Andrew may have been the most prolific producer of tornadoes since Hurricane Beulah (1967). All of the documented tornadoes occurred with the second landfall, when numerous tornadoes occurred across Louisiana, Mississippi and Alabama from the evening of 25 August through the afternoon of 27 August. The most damaging tornado occurred prior to the second landfall when one moved to the west-northwest across an area just west of New Orleans, LA on the evening of 25 August. This tornado produced F3 damage along a path 14 km in length, and caused two deaths and 32 injuries. A rawinsonde observation made within three hours of the event at nearby Slidell, LA revealed an environment with CAPE around 700 J kg-1 and 0-3 km SRH in excess of 300 m2 s-2.

    Most of the tornadoes occurred on 26 August: nine between 12 UTC and 18 UTC and thirteen from 18 UTC to 00 UTC. Ten more were reported in the next three hours. The sounding data from 12 UTC on 26 August revealed evidence of an extensive area of dry air north and east of the storm center. The Centreville, AL sounding (not shown) found a dew point depression of 20° C at 700 hPa and the profile was quite dry above ~850 hPa. The extent of the dry air intrusion is best depicted at 700 hPa  (Fig. 3).

    Close to the area of tornadogenesis, the vertical extent of the dry intrusion was sampled by the 12 UTC sounding from Jackson, MS, which found relative humidity averaging less than 65 percent through most of the layer between 850 hPa and 500 hPa, with several points below 50 percent around the 725 hPa level. A comparative plot6 of both the 12 UTC and 00 UTC soundings (Fig. 4) shows that the atmospheric column moistened during the day, but without much change in the low-level lapse rate.

    By the 00 UTC, the obvious signs of the dry intrusion at mandatory levels had been reduced, perhaps by the strong, deep southerly flow in the eastern semi-circle of the storm. However, 00 UTC sounding at Centreville, AL (not shown) revealed that relatively dry air persisted in the layer between 800 hPa and 600 hPa where the average relative humidity was around 60%. Dry air persisted at mid and upper levels over Georgia and northern Florida, where mid-level dew point depressions exceeded 20° C.

6 All comparative sounding plots have been truncated at 400 hPa for better resolution of the lower and middle levels.



3.4 Tropical Storm Beryl (1994):
    Beryl developed close to the central Gulf coast southeast of Mobile, AL on 14 August and moved northeast, making landfall near Panama City, FL late on the afternoon of 15 August.  Although no tornadoes accompanied the landfall, Beryl’s remnants moved northward into Georgia and an outbreak of 28 tornadoes occurred on 16 August, mainly in South Carolina.

    This case was examined by Vescio et al. (1996), who found that dry air in a deep layer above the surface was entrained into the northeast quadrant of the system over portions of the Carolinas. The dry intrusion resulted in steeper lapse rates and greater instability. A comparison of the 00 UTC and the 12 UTC soundings on 16 August at Charleston, SC (Fig. 5) reflected a drastic change. Although the winds had backed through a deep layer, providing more of an onshore flow, significant drying had occurred through a deep layer from just above the surface to near 500 hPa. The sounding profile at 00 UTC does reflect a substantial increase in the low-level lapse rate through approximately the lowest 200 hPa of the sounding.

    The tornadoes reflected a strong diurnal influence, as all but two occurred between 17 UTC and 00 UTC. Several of the tornadoes had long tracks (>15 km) and two had path lengths exceeding 40 km. Two of the tornadoes produced damage rated F3 and eight produced damage rated F2. The outbreak was centered in the sharpest portion of the relative humidity gradient at 700 hPa (Fig. 6), centered near Columbia, SC.



3.5 Hurricane Beulah (1967):
    Hurricane Beulah crossed the Yucatan peninsula on the 17 September and was briefly rated a Category 5 storm over the Gulf of Mexico prior to landfall. However, Beulah weakened to Category 3 just prior to landfall on the morning of 20 September, which occurred over Padre Island northeast of Brownsville, TX.  Beulah continued toward the north-northwest for 18 hours, then turned toward the west, and finally toward the southwest, moving into northern Mexico.

    Beulah’s tornado productivity was highest on the day of landfall (20 September), but the storm produced at least 20 tornadoes on 21 September (after the westward turn) and an additional 21 tornadoes on the 22 September (while moving to the southwest across the Rio Grande and into northern Mexico). Examination of the sounding data reveals a dry intrusion, again best seen at 700 hPa, which appears to be entrained into the storm’s circulation from the east. The analysis reveals a significant gradient of relative humidity analyzed across the upper Texas coast and into central Texas on the morning of the 20 September. This gradient remained in place through the subsequent 48 hours, gradually shifting to the southwest. The dry layer became increasingly apparent on the Lake Charles, LA soundings at 00 UTC on 19 September and 12 UTC on 20 September (neither shown).

    The dry layer became faintly evident on the Victoria, TX sounding 00 UTC on 21 September and was pronounced by 12 UTC on 22 September, as seen in a comparison of those soundings (Fig. 7). The diagram depicts a significant change in the environmental lapse rate commencing at about the 850 hPa level and extending upward. The change coincides with a large drop in relative humidity that commences at 900 hPa and extends upward through most of the sounding. Surface-based CAPE (SBCAPE) had increased to almost 2000 J kg-1 at that time.

    An analysis of 700 hPa relative humidity at both 12 UTC and 00 UTC (not shown) on 20 September revealed a significant dry to moist (east to west) gradient that coincided with the area of tornadogenesis. Tornadoes began occurring well before sunrise in the coastal plains between Victoria and Houston as the right-front (northeast quadrant) moved onshore, contemporaneous with the center’s landfall north of Brownsville.

    The exact number of tornadoes on 20 September is a matter of some conjecture because the recorded locations and times for the events occasionally seem to overlap. The State Climatologist reported a total of 115 tornadoes for Beulah, and 67 tornadoes on 20 September, both totals that were reached by reviewing "all storm reports collected by ... Weather Bureau Offices in the areas affected ..., by examining 156 Hurricane Report questionnaires (prepared) by persons in the affected area, and by scanning more than 700 newspaper clippings pertaining to ... descriptive evidence of tornado occurrence" (Orton 1970).

    Tornadogenesis on 21 September and 22 September displayed a distinctly diurnal pattern, with most of the tornadoes occurring during daylight hours. Analysis of 700 hPa relative humidity at 00 UTC on 21 September (Fig. 8) found a significant relative humidity gradient across south central and southeastern Texas, over the area in which the outbreak occurred on 21 September. At 12 UTC on 22 September, the gradient was centered near Alice, west of Corpus Christi (not shown). All of the tornadoes on 21 and 22 September occurred in the "rear" quadrants of the storm relative to forward motion, but in the northeast quadrant of the storm relative to the storm’s location.



3.6 Hurricane Danny (1985):
    Hurricane Danny made landfall as a Category 1 storm on the southwest Louisiana coast around 17 UTC on the 15 August. The storm continued northward, passing near Monroe, LA around 12 UTC on 16 August, then turned more to the northeast, passing near Muscle Shoals, AL around 06 UTC on 17 August. Danny produced numerous tornadoes on 16 August, the day after landfall. A strong diurnal trend was noted, as most of the tornadoes (20) occurred in a six-hour period between 15 UTC and 21 UTC across portions of northern Alabama and southern Tennessee.

    McCaul (1987) provided a detailed analysis of this tornado outbreak. He noted that some of the tornadoes occurred in “tornado families”, produced relatively long damage tracks, and in some instances, produced F3 damage. He concluded that one of the factors in the outbreak may have been a mid-and upper-level dry intrusion, noting that satellite imagery depicted a narrow but well-defined mid- and upper-level dry intrusion in western Alabama that appeared to be continuously connected to a large source of mid-level dry air over Mississippi, Louisiana and Arkansas. However, analysis of relative humidity at 500 hPa at 12 UTC (Fig. 9) suggests that dry air was already very close to, but just south of, the outbreak area.

    This source may have aided in initiation of the outbreak with the dry intrusion noted by McCaul aiding in its continuation. Significant drying through a substantial layer of the midtroposphere (800 to 500 hPa) is seen in the comparative plot of the Jackson, MS soundings at 12 UTC and 00 UTC on 16 August (Fig. 10).  The evening (00 UTC) sounding revealed a significant increase in the low-level lapse rate (surface to 950 hPa) but this was in the westerly flow in Danny's wake.



3.7 Hurricane David (1979):
    Hurricane David passed south of Puerto Rico and struck the island of Hispaniola as a Category 5 hurricane on 31 August. Weakened by interaction with the island, the storm continued northwest and then north, never regaining its prior intensity.

    David came very close to making continental landfall twice. On 3 and 4 September, the storm was just offshore from the east coast of Florida moving northward. Ten brief, mostly weak tornadoes occurred along the immediate coastal area. The storm continued northward to a decisive landfall near Charleston, SC on the night of 4 September as a strong Category 1 storm. Four additional brief, weak tornadoes occurred near and north of Myrtle Beach coincident with landfall. There is no evidence of a dry intrusion with any of these tornadoes.

    David continued northward and produced an outbreak of 19 tornadoes in the District of Columbia, Maryland, Delaware and Pennsylvania on 5 September. The tornadoes occurred in a strong gradient of relative humidity at 500 hPa between mid-level dry air to the south and moist air to the north (Fig. 11). These tornadoes displayed a strong diurnal trend, with 14 of the 19 tornadoes occurring between 18 UTC and 01 UTC. Several tornadoes had remarkably long tracks (>25 km) and were quite intense (two produced F3 and eight produced F2 damage). A comparison plot of soundings for Washington, DC at 12 UTC and 00 UTC on 5 September (Fig. 12) shows drying through the lower troposphere despite significant moistening in the layers above 700 hPa. SBCAPE at Washington was approximately 900 J kg-1 on both the 12 UTC and 00 UTC  soundings. Perhaps of greater importance, the SRH at Washington was very high during the outbreak (from ~280 m2 s-2 in the 0-2 km layer7  to more than 500 m2 s-2 in the 0-3 km layer). There was only a minor increase in the low-level lapse rate between the 12 UTC and 00 UTC soundings.

7Johns et al. (1993) examined the relationship between CAPE and 0-2 km helicity for tornadoes rated strong and violent. In four TC cases involving strong tornadoes, the authors found relatively low values of both CAPE and helicity.



3.8 Hurricane Georges (1998):
    Hurricane Georges crossed Puerto Rico, Hispaniola and Cuba. The storm attained Category 4 status before reaching Puerto Rico, but was weakened by interaction with the islands. Georges departed Cuba heading northwest, then turned northward across the central Gulf of Mexico to a continental landfall shortly after 06 UTC on 28 September just west of Gulfport, MS. A Category 1 storm at landfall, Georges moved a short distance inland, then slowed and turned abruptly eastward, moving parallel to the Gulf coast, a motion that permitted Georges to maintain tropical storm status for an extended period.

    The unusual motion of Georges following landfall resulted in an extended landfall event as the storm remained within ~100 km of the coast as it moved slowly eastward, producing an extended period of strong onshore flow from the very warm waters of the Gulf of Mexico. Twenty tornadoes occurred on 28 September, mainly over southeastern Alabama.

    The 12 UTC analysis for 28 September at 700 hPa  (Fig. 13) and 500 hPa (not shown) demonstrated a significant relative humidity gradient across the area where the tornadoes subsequently occurred, with dew point depressions of 8° to 10° C at 700 hPa in the zone from Tampa to Tallahassee to Atlanta. The timing of the outbreak (afternoon and evening) on the 28 September suggests the involvement of diurnal influences.

    The Tallahassee sounding at 00 UTC on 28 September) (not shown) found a distinct dry intrusion in the layer between 650 hPa and 450 hPa, where the average relative humidity was less than 35 percent. Twelve hours later, the 700 hPa relative humidity analysis for 12 UTC on 29 September (Fig. 14) found the area of very dry air was still present over northwestern Florida and adjacent areas of Alabama and Georgia. The dew point depression at 700 hPa had increased to 6° C at Tallahassee, 22° C at Tampa and 17° C at Jacksonville. At 500 hPa (not shown), the dew point depression at Tallahassee (21?C) continued to reflect a very dry pocket of air. The remnants of Georges produced 17 more tornadoes on the 29 September over southeastern Alabama, southern Georgia and northwestern Florida, with slightly more than two-thirds of the tornadoes occurring during the afternoon and early evening hours.



3.9 Hurricane Gilbert (1988):
        Hurricane Gilbert crossed Yucatan as a Category 5 storm on 14 September and made landfall as a Category 4 storm on the northeast coast of Mexico about 210 km south of Brownsville, TX on 16 September.

    Gilbert, a very large hurricane at the time of landfall, produced 12 tornadoes in the lower Rio Grande Valley of Texas during and immediately after landfall. The larger outbreak (28 tornadoes) occurred the day after landfall on 17 September. These tornadoes occurred at distances of 375 to 750 km from the storm center, which was then about 350 km inland and moving over the increasingly mountainous highlands south of Monterrey, Nuevo Leon. Examination of the sounding data from 17 September revealed a similar pattern to that seen in the other cases but the dry intrusion is slightly better depicted at the 500 hPa level than at 700 hPa. The 12 UTC analysis at 500 hPa on 17 September (not shown) found a pronounced relative humidity gradient extending inland toward central Texas from the upper Texas coast from a relative humidity minimum just south of Lake Charles. Twelve hours later, the 500 hPa analysis (Fig. 15) found that the gradient had intensified and advected westward. Analysis of the 00 UTC sounding at Stephenville, TX (not shown) yielded SBCAPE of almost 1400 J kg-1 and sufficient low-level support for updraft rotation (SRH of 150 m2 s-2 in the 0-1km layer and 210 m2 s-2 in the 0-2 km layer).

    An interesting subset of Gilbert's tornadoes occurred at a remarkably extreme range (~750 km) from the remnants of the storm center. Ten tornadoes occurred in the counties just east of San Angelo, TX along a baroclinic surface boundary that appears to have developed in place as an indirect result of the mid-level dry intrusion to the east. The outer rain shield of Gilbert cooled the lower troposphere overnight on the 16th/17th as it rotated east to west across Texas. Then significant insolation in the areas affected by the dry intrusion warmed the lower troposphere during the morning and early afternoon of the 17th. The result was a boundary oriented northeast-southwest with cool, moist air to the west and warm, moist air to the east (Fig. 16). All of the tornadoes occurred between 17 UTC and 22 UTC close to this low-level boundary. As noted in Sec. 2.2, low-level meso-beta scale baroclinic boundaries have been implicated in tornadogenesis in both TC and mid-latitude supercells (Pietrycha and Hannon 2002; Edwards et al. 2000; Markowski et al. 1998).



3.10 Tropical Storm Josephine (1996):
    Josephine developed in the Gulf of Mexico on 4 October and moved to the northeast along a stalled baroclinic zone. The storm made landfall on the evening of  7 October at Apalachee Bay (FL) shortly before 06 UTC. Never a very robust TC, the system quickly lost its tropical characteristics as it moved rapidly northeastward into Georgia on 8 October and continued moving to the northeast.

    Prior to landfall, Josephine spawned an outbreak of 18 tornadoes over the central Florida peninsula on 7 October. At 12 UTC, the sounding profiles at both Tampa and Tallahassee were saturated from the surface to well above 500 hPa; at Miami, the profile hinted at the beginning of a dry intrusion above 700 hPa; and at Key West, there was evidence of significant drying from 850 hPa upward. The flow at Key West was nearly unidirectional from the south-southwest through almost 500 hPa. Twelve hours later, evidence of a mid-level dry intrusion was profound at both Tampa and Miami. The dry intrusion is best depicted by the analysis at 500 hPa (Fig. 17). The tornadoes over the Florida peninsula presented a distinct diurnal pattern, as fifteen of the eighteen occurred between 15 UTC and 00 UTC, all prior to the actual landfall.

    At 00 UTC, the Tampa surface wind was from 215° at 7.5 m s-1, but very strong winds (~30 m s-1) were detected just above the surface. Not far inland from Tampa, surface winds were backed, with the flow from approximately 155° at around 7 m s-1. Most of the tornadoes occurred between 18 UTC and 00UTC in the area where the backed flow was located. Adjustment of the 00 UTC Tampa wind profile to a surface flow from 155° at 7 m s-1 yielded a very potent SRH value of 268 m2 s-2 in the 0-1 km layer and 385 m2 s-2 in the 0-2 km layer.

    As the remnants moved into the Carolinas, a brief flurry of eight tornadoes occurred along the coast on the morning of the 8 October. Hudgins and Frederick (1997) examined some of the 8 October tornadoes and found that they occurred along a strong coastal front located just inland of the coast. Objective analysis of surface data at 12 UTC confirms their analysis (Fig. 18).

    The tornadoes were relatively weak and brief, although one did produce F2 damage. All of the tornadoes occurred between 0930 UTC and 1400 UTC, just ahead of the surface cyclone, in an area where winds of 14 to 23 m s-1impinged on the warm front. The Newport, NC (KMHX) 12 UTC rawinsonde reflected 0-2km SRH exceeding 600 m2 s-2. These tornadoes, produced by the remnants of a TC about to leave the continental landmass and move back over the ocean, are consistent with the pattern described by Edwards (1998). Hudgins and Frederick (1997) mention that mid-level drying is shown above the 700 hPa level.

    The tornadoes occurred prior to 15 UTC, so diurnal influences did not play a role, but the mid-level dry air could have contributed to this part of the outbreak through other mechanisms, such an entrainment into mini-supercells and downward transport to the near-surface environment, as discussed hereafter in Sec. 5.2. As was the case with Agnes, it is unlikely that the dry intrusion was responsible for all of the tornadoes, but this appears to be another case where the outbreak can be sufficiently attributed to a mid-level dry intrusion to justify inclusion of this case.



3.11 Hurricane Opal (1995):
    Hurricane Opal developed over the Yucatan and became a tropical storm on 30 September as it moved into the northern Bay of Campeche. Opal became a hurricane on 2 October and began a north-northeast motion at increasing forward speed.  Landfall was near Pensacola, FL on the evening of 4 October. Opal will be remembered not only for its tornado outbreak but also for its sudden fluctuation in intensity (Shay et al. 2000). Between late evening on 3 September and sunrise on 4 September, the storm rapidly intensified to a Category 4 hurricane, but then just as quickly weakened again to Category 1 prior to landfall. The remnants of Opal moved rapidly northeast across Alabama, Tennessee and Kentucky on 5 October.

    Opal produced numerous tornadoes on 4 October in a twelve-hour period centered around landfall. Inspection of the locations and times of tornado occurrence suggests two episodes of tornado activity: between 15 UTC and 19 UTC, and again between 22 UTC and 02 UTC. The initial burst of tornadoes (9) can be classified as an outer rainband event. The subsequent burst of tornadoes (13) was related to landfall of the core of the storm. In both events, the tornadoes occurred within 120 km of the coast.

    Visible satellite imagery from the morning of 4 October depicted a clear slot (a zone of cloud erosion) in the eastern semi-circle of Opal’s circulation. The GOES-8 visible image from 1602 UTC (Fig. 19) shows that the clear slot extended to a point very close to Pensacola, FL, and a zone of intense convection was located just on the north and northeast side of this area. The initial outbreak of tornadoes was underway within the band of intense convection coincident with this image.

    Analysis of the 12 UTC sounding data at both 700 hPa and 500 hPa, as well as water vapor satellite imagery (not shown) suggests that the dry intrusion occurred in a narrow channel that was west of both Tampa and Tallahassee, perhaps coinciding with the clear slot feature seen in the satellite image.  Analysis of relative humidity at 500 hPa at 00 UTC (Fig. 20) indicated that the drying had spread eastward to Tallahassee, FL.  Although drying was present between 700 hPa and 500 hPa in both soundings, the full extent of the dry intrusion was best detected in the evening sounding from Tallahassee (not shown), where the dew point depression at 500 hPa increased by 15° C from the 12 UTC sounding.

    The dry intrusion in the eastern semicircle appears to have originated well to the northwest of Opal several days prior to the landfall. Rodgers et al. (1998) found that the dry region intruded cyclonically around the western and southern regions of Opal. They attributed the origin of the dry air to vertical motions within a mid-latitude trough that crossed the southern plains, eventually producing Opal’s northward acceleration and concluded that dry air intruded quite close to Opal’s center by 12 UTC on 4 October.



4. DISCUSSION OF NON-QUALIFYING CASES

    Two of the thirteen outbreak cases do not appear to have been related to mid-level dry intrusions. It is worthy of note that although these events appear to be contrary to the pattern seen in the eleven qualifying cases, they did produce a number of significant tornadoes.

4.1 Hurricane Alicia (1983)
    Hurricane Alicia made landfall on the upper Texas coast near Galveston before sunrise on 18 August as a Category 3 storm, and then continued toward the north-northwest. Examination of the location and times of the tornadoes attributed to Alicia reveals that virtually all were associated with the core of the storm as opposed to the outer rainbands. Most occurred in the immediate area of landfall, in the greater Houston-Galveston area. Inspection of the pertinent upper air data revealed no evidence of a dry intrusion at either 700 hPa or 500 hPa, although at 500 hPa very dry air existed in all quadrants beyond ~350 km. Examination of satellite imagery from the morning of landfall (although primitive compared to today’s imagery) does not suggest evidence of a dry intrusion into the storm’s main circulation. Coupled with the fact that most of the tornadoes were associated with the core of the storm, Alicia was determined to be non-qualifying.

***

4.2 Hurricane Carla (1961)
    Hurricane Carla made landfall on the middle Texas coast on the evening of 11 September. Rated a Category 5 storm over the western Gulf of Mexico, Carla had weakened to a Category 3 storm at landfall. The storm continued northwest into central Texas on 12 September. Tornado production began on the 10th when the storm was 275 to 300 km south of Galveston. All of the tornadoes on 10 September and all but three on 11 September occurred in Louisiana. Relative to storm motion at the time, all but one of the Louisiana tornadoes would be classified as having occurred in the right-rear quadrant. However, relative to true north, they were in the northeast quadrant of the storm.

    But for the Louisiana tornadoes, Carla would not qualify for inclusion in this study. All of these tornadoes (on 10 and 11 September) occurred at rather extreme range from the storm (400 to 700 km) and well prior to landfall. Examination of mid-level relative humidity reveals little evidence of a dry intrusion, and these tornadoes (antecedent to landfall by 18 to 36 hours) suggest the operation of factors other than a mid-level dry intrusion involving outer rainbands. The subsequent tornadoes in Texas on 11 September and the tornadoes in both states on the 12 September were associated either with the core of the storm or with outer rainbands, but the number of these tornado falls well below the "outbreak" threshold.8

***
8Included in the 12 September events is the early morning (0915 UTC) tornado that struck Galveston Island after initially developing offshore as an intense waterspout  (Grazulis 1993). This tornado caused damage rated F4 and according to Grazulis, may have been the most intense tornado ever to strike Galveston. This event stands for the proposition that TC tornadoes need not be related to a dry intrusion to produce very significant damage.


5. DISCUSSION: THE ROLE OF MID-LEVEL DRY INTRUSIONS

    As suggested in Sec. 2.3, there are a number of issues involved in tornado outbreaks associated with landfalling TCs and the involvement of mid-level dry intrusions. Various explanations have been offered, with the most frequent being an increase in convective instability due to an increase in the environmental lapse rate, other alterations of the thermodynamic structure of the tropical cyclone environment with substantial enhancement of CAPE and surface-based instability, and enhanced evaporative cooling within the rear-flank downdraft (RFD) with modulation of the development of mesocyclones in the outer rainband as a result.

***

5.1 Sources of the Mid-Level Dry Air:
    McCaul (1987) raised a crucial issue, noting that if the enhancement of convective instability due to dry intrusions aloft is indeed an important factor in triggering large outbreaks of TC tornadoes, it is important to discern the critical factors that dictate which hurricanes will ingest dry air in the most destabilizing manner. Based on the storms examined herein, it appears that tornado outbreaks are more likely when the mid-level dry intrusion produces a relatively sharp gradient of relative humidity in the storm quadrant where parameters support the development of mesocyclones. However, the source of the dry air that was ingested by the storms was found to vary.

    To examine this issue, 6-hourly data from the Climate Diagnostics Center's NCEP/NCAR Reanalysis Project (Kalnay et al. 1996) was used to examine relative humidity distribution in the mid-levels for an extended period and over a broad area prior to storm landfall. Animated loops of these graphics suggest two primary patterns. One pattern (arbitrarily designated Type 1) involved a mass of dry air that impinged on much of the northern or northwestern semi-circle of the storm's outer circulation. In these cases, the dry air appeared to gradually divide into two lobes as the storm advanced, with one to the northwest and the other to the northeast of the storm center. In the other pattern (arbitrarily designated Type 2), the dry air was ingested into the storms from a reservoir of mid-level dry air, most often (but not exclusively) located in the eastern semi-circle of the storm.

    Agnes, Beulah, David and Gilbert were found to be Type 1 storms. Allen, Andrew, Beryl, Danny, Georges, and Josephine were found to be Type 2 storms. Opal was found to exhibit both patterns. The time-series loop suggested that Opal's envelope of tropical air gradually displaced the pre-existing dry air over the southeastern U.S., but that the more significant dry intrusion originated with the air that wrapped into the circulation in the strong cyclonic flow from the south and southwest.

***

5.2 Convective Instability and the Diurnal Signal:
    The qualifying cases have a strong diurnal signal detected in the temporal distribution of the tornadoes. Almost two-thirds (65%) of the tornadoes in the eleven qualifying cases occurred between 15 and 24 UTC (daylight hours in the region examined). By contrast, in the two cases excluded because of the absence of evidence of a dry intrusion, only 37% of the tornadoes occurred in the same period.  Some of the previously published research has addressed this issue. McCaul (1991) found that the tornadoes in his proximity database featured a "distinct maximum in early and mid afternoon", with 54% of the tornadoes occurring between 9am and 6pm local solar time.  He concluded that "this finding suggests that solar radiational heating might be important even in the mostly cloudy environments characteristic of landfalling hurricanes." Novlan and Gray (1974) noted a "weak frequency peak around 1100LST" but did not attribute any significance to this finding. Hill et al. (1966) noted that other researchers had examined and rejected the notion of a temporal distribution because they were examining only one or two storms. They found a general preference for the period from local noon to local midnight, but declined to attribute appreciable significance to the finding. Orton (1970) found a significant peak for Beulah’s tornadoes between 9am and 6pm local time.9

    The manner in which the mid-level dry intrusion enhances tornadogenesis has been the subject of some research. McCaul (1991) found that overall there was a negative correlation between CAPE and tornado potential on average for all of his cases. He noted that CAPE was typically at a minimum near and to the left of the center’s track, and maximized well to the right of the center’s track. He found that in the vertical, most of the tropical cyclone’s CAPE is found a levels below 650 hPa, because of warming above that level. As noted previously, Vescio et al. (1996) suggested that mid-level dry intrusions have the potential to substantially alter the thermodynamic structure of the tropical cyclone environment (with substantial enhancement of CAPE and surface-based instability).

    Proximity soundings (defined in McCaul {1991} as those occurring within 3 h and 185 km of a tornado event) for the closest one or two stations in the outbreak cases herein were examined for evidence that mid-level dry intrusions might produce a consistent pattern with respect to temperature lapse rates. Despite the use of a variety of levels (950-700 hPa, 850-500 hPa, 700-500 hPa), no consistent pattern was discernible. In each case, the morning and evening soundings on the outbreak day were examined. On average, the lapse rates between in the 950-700 hPa layer increased by less than 0.2° C, the lapse rates in the 850-500 hPa layer were unchanged, and the lapse rates in the 700-500 hPa layer decreased by slightly more than 0.2° C.  This finding may reflect that the sounding sites were too far removed from the tornadic cells to detect the alteration of the near-storm environment by the mid-level drying, or it may suggest that other factors were involved.

    As noted previously in Sec. 2.3, Cione et al. (2000) examined changes in air temperature and sea surface temperature changes in 37 hurricanes between 1975 and 1998, and concluded that the change in the spread between sea surface temperature and TA was primarily due to a reduction in TA (and not the result of adiabatic expansion), and proposed that one explanation “may be that outside the hurricane inner core, unsaturated convective downdrafts act to dry and evaporatively cool the near-surface environment”. They also reviewed previous research indicating that convective downdrafts are capable of bringing cooler air to the surface in tropical systems, provided the air aloft is sufficiently dry.

    Spratt et al. (2000) documented a convective downdraft in an outer rainband cell of Hurricane Bonnie (1998) utilizing data from a dropsonde launched from a NOAA Hurricane Research Division (HRD) aircraft as the hurricane neared the North Carolina coast. The dropsonde fortuitously advected into a mini-supercell soon after its release, and remained entrained in the cell until just prior to reaching the surface. The dropsonde documented a steep lapse rate during descent, indicative of a layer of surface-based instability. Vertical velocity data from the aircraft was retrieved and it confirmed downward motion associated with the low-level theta-e minima, thus confirming the origin of the parcel (aloft, descending within downdraft core)10.

    The transport of mid-level dry air to or near the surface in TCs does not directly explain the diurnal signal seen in the outbreak cases. Another possible mechanism involves the erosion of mid and upper tropospheric cloudiness by the intruding dry air, allowing full insolation to penetrate to near (if not all the way to) the surface, thereby enhancing convective instability on a relatively local scale. But ultimately, the answer to how mid-level dry intrusions are related to tornadogenesis in landfalling TCs may depend upon model simulations.

    One such effort has produced interesting (if preliminary) results. Romine and Wilhelmson (2002) have conducted numerical model simulations of the Hurricane Opal (1995) tornado outbreak, one of the qualifying cases herein. They have successfully simulated the dry intrusion in the storm’s eastern semi-circle, and are presently using the NCOMMAS model, a derivative11  of the Collaborative Model for Multiscale Atmospheric Simulation (COMMAS) (Gilmore and Wicker 1998) in simulations focusing on the role that the dry air may play with respect to the evolution of mini-supercells in the area affected by the dry intrusion.

***
9Note that seven of the thirteen cases examined herein were part of McCaul’s dataset; the other six  occurred subsequent to the last case included in his dataset.
10 Spratt 2000, personal communication.
11 Romine 2003, personal communication.


6. DISCUSSION: NULL CASES

    Another significant issue with respect to the primary findings discussed in Sec. 3 is whether there have been landfall events with detectable mid-level dry intrusions that did not produce tornado outbreaks as defined herein. To examine this issue, the 4-times daily data from the Climate Diagnostics Center's NCEP/NCAR Reanalysis Project was used to examine relative humidity in the mid-levels for all TCs making landfall in the study area during the specified period. All landfalling cyclones were examined in the same manner: soundings were examined at 500 hPa, 700 hPa and 850 hPa for the period commencing 36 hours prior to landfall through 36 hours following landfall. There were three events with definitive evidence of a mid-level dry intrusion in which the number of tornadoes failed to reach at least twenty (the minimum established for inclusion as “outbreak cases” herein), but each of those cases did involve a significant number of tornadoes.

    Hurricane Earl (1998) made landfall around 06 UTC on 3 September near Panama City, FL. The remnants of the storm moved northeast to near Wilmington, NC at 06 UTC on 4 September. Edwards et al. (2000) described the tornado production by Earl as bimodal because the tornadoes occurred in two phases: on the day prior to landfall and on the following day as the remnant circulation was about to move back over the open ocean. A total of 15 tornadoes occurred, about half in the pre-landfall stage and the others in the “exit” phase (Edwards, 1998). Examination of the sounding data for 00 UTC on 03 September reveals a pronounced relative humidity gradient at 700 hPa overlying the area of the pre-landfall tornadoes, with drier air to the southwest of Tampa and moist air to the northeast. The mid-level dry air continued advecting to the northeast and produced a pronounced relative humidity gradient at 700 hPa over the area of the “exit” phase tornadoes at 00 UTC on 4 September.

    Hurricane Edith (1971) made landfall on the Louisiana coast on 16 September.  Edith continued to the northeast through Louisiana and Mississippi and produced sixteen tornadoes on the day of landfall. A strong diurnal trend was noted, with most of the tornadoes (11) in a nine-hour period between 15 UTC and 00 UTC and with peak occurrence shortly after 18 UTC. Examination of the sounding data revealed a pronounced dry intrusion that appears to have been entrained into the Edith's circulation from the east and southeast (Type 2). Best defined at the 700 hPa level, the intrusion remained east of the center of the storm following landfall. This non-qualifying outbreak of tornadoes occurred directly beneath the intense relative humidity gradient.

    Hurricane Floyd (1999) made landfall before sunrise on 16 September near Wilmington, NC. Floyd produced eighteen tornadoes across mostly coastal sections of North Carolina on 15 September (prior to landfall). Although there was a significant diurnal trend in this outbreak, suggesting the possibility of a mid-level dry intrusion, it appears that other factors were involved. Pietrycha and Hannon (2002) found that 16 of the 18 tornadoes developed immediately along and/or within the warm side of a coastal front that developed and moved onshore ahead of the storm as it approached the coast. Analysis of the 00 UTC sounding data (on the evening of 14 September) revealed a pronounced relative humidity gradient at 500 hPa (less so at 700 hPa) with very dry air to the northwest and very moist air to the southeast (Type 1). By 12 UTC on 15 September, the intense gradient at 500 hPa remained but the 700 hPa gradient had relaxed, perhaps due to the arrival of the moist envelope associated with Floyd.

    By 00 UTC the gradient at 500 hPa was arranged in a more west to east (dry to moist) configuration and the 700 hPa gradient was beginning to suggest some wrapping of dry air from the southwest around the south side of Floyd’s circulation. The intense 500 hPa gradient overlay the zone of tornadogenesis during the daylight hours on 15 September.



7. DISCUSSION AND CONCLUSIONS

    Table 2 contains a comparison of the three classes of storms (qualifying, non-qualifying and null) including the number of tornadoes produced by storms with mid-level dry intrusions that produced outbreaks (qualifying), storms without mid-level dry intrusions that produced outbreaks (non-qualifying), and storms with mid-level dry intrusions that failed to produce outbreaks (null) by number of tornadoes, significant F-scale damage, and casualties.


TABLE 2. Comparison of tornadoes produced by storms with mid-level dry intrusions that produced outbreaks (qualifying), storms without mid-level dry intrusions that produced outbreaks (non-qualifying), and storms with mid-level dry intrusions that failed to produce outbreaks (null) by number of tornadoes, F-scale damage, and casualties.
Class
Number of tornadoes
Tornadoes with F2 damage
Tornadoes with F3 or greater damage
Deaths
Injuries
Qualifying
494
61
20
12
410
(Avg. per storm)
(45)
(6)
(2)
(1)
(41)
Non-qualifying
42
7
8
14*
335*
(Avg. per storm)
(22)
(4)
(3)
(5)*
(168)*
Null
51
10
2
2
33
(Avg. per storm)
(17)
(3)
(1)
(0)
(11)
*This classification (and its averages) appear somewhat distorted by the single event with unusually high casualty figures (F4 at Galveston during Hurricane Carla).

    In an attempt to discern significant differences in the environments of storms that had dry intrusions and produced outbreaks, those that had no dry intrusion and yet produced outbreaks, and those that had dry intrusions but failed to produce outbreaks, composite soundings were generated utilizing the closest one or two stations for which soundings were made in or close to the favorable right front (northeast) quadrant closest to the time of the tornado events or the outbreak. The results are presented in Table 3.



TABLE 3. Comparison of average LCL height and of dew point depression at various levels in storms with mid-level dry intrusions that produced outbreaks (qualifying), storms without mid-level dry intrusions that produced outbreaks (non-qualifying), and storms with mid-level dry intrusions that failed to produce outbreaks (null) .
Class
LCL (in meters)
Surface (Tdd in °C)
925
(Tdd in °C)
850
(Tdd in °C)
800
(Tdd in °C)
700
(Tdd in °C)
600
(Tdd in °C)
500
(Tdd in °C)
Qualifying
141
1.5
1.8
3.0
3.4
4.4
6.6
7.0
Non-qualifying*
325
3.0
2.4
3.0
3.4
4.2
0.5
1.5
Null*
215
1.2
4.7
3.9
6.5
4.0
2.5
1.6
*Note that these two classifications represent very small samples (2 cases in the non-qualifying class and 3 cases in the null class).

    Three findings are worthy of note: (a) the composite for storms that had mid-level dry intrusions (i.e. qualifying outbreak storms) had a lower average LCL compared to the other two classes of storms; (b) the composite for storms that had mid-level dry intrusions and produced outbreaks (qualifying storms) were more moist in the layer from just above the surface through 900 hPa than the other two classes of storms; and (c) the qualifying storms were much drier at levels above 700 hPa than were either the non-qualifying storms or those that had evidence of a dry intrusion but did not produce outbreaks.

    It is important to keep in mind that two of the storm classes contain very small samples. Nevertheless, it is appropriate to observe that findings (a) and (b) are suggestive of variations between storm classes in the depth and quality of moisture in the lowest 100 hPa of the atmosphere. It is possible that the “qualifying” storms were more productive because the lowest layer was more uniformly moist than in the other two storm classes. This appears to be consistent with the findings of Rasmussen and Blanchard (1998)12, who (in a study of various sounding-derived supercell and tornado parameters across the United States) noted that the LCL for soundings associated with tornadoes was significantly lower than for soundings associated only with supercells or even with nonsupercells. Thompson et al. (2002) report verification of the Rasmussen and Blanchard research. Using close proximity soundings derived from RUC-2 model hourly analyses, the authors found that substantially lower LCLs were associated with significant tornadic supercells as compared to nontornadic supercells. McCaul and Weisman (2002) performed simulations of storm structure as related to variations in the depths of the mixed layer and the moist layer by specifying that LCL=LFC in all simulations, then varying the height of the LCL, CAPE and shear between simulations. They found that a lowered LCL (a frequent characteristic of TC environments) reduced the strength of surface outflow, prevented outflow dominance, and promoted storm persistence. This potential relationship merits further research.

    Thirteen tornado outbreak cases associated with landfalling TC along the coast of the Atlantic and of the Gulf of Mexico occurring since 1960 have been analyzed for evidence of mid-level dry intrusions as a factor in the outbreaks. Eleven of the thirteen cases offer clear evidence of a dry intrusion at mid-levels over the outbreak area. Tornadogenesis appears to be enhanced when the favorable northeast (or right front quadrant) of the storm coincides with a pronounced gradient in layer relative humidity,  generally best reflected at the 700 hPa or 500 hPa level.

    Two distinct patterns were identified with respect to the source of the mid-level dry air involved in these cases: in one, a mass of dry air that impinged on much of the northern or northwestern semi-circle of the storm's outer circulation became divided into two lobes as the storm advanced, with one lobe to the northwest and the other to the northeast of the storm center; the other involved ingestion of a lobe of dry air from a reservoir most often (but not exclusively) located in the eastern semi-circle of the storm.

    The qualifying cases were determined to have a strong diurnal signal detected in the temporal distribution of the tornadoes, as almost two-thirds (65%) of the tornadoes in the eleven "positive" cases occurred between 15 UTC and 00 UTC (daylight hours in the region examined). This characteristic has been noted by other authors, but an explanation of its relationship to tornadogenesis remains somewhat elusive, at least in part because of the sparseness (both geographically and temporally) of rawinsonde observations. A  denser network of observation sites coupled with more frequent sonde releases (both dropwindsonde and rawinsonde) would be an important step for both research and operational purposes.     While modern remote sensing platforms are invaluable aids in ascertaining the existence and extent of mid-level dry intrusions, data from rawinsondes and dropwindsondes are critical.

    Issues deserving of additional research include determining the origin of the dry air reservoirs that contribute to these events. Based upon the findings reported in Sec. 5.1, it is likely that there are at least two mechanisms at work, but there may be others. As has been demonstrated herein, baroclinic surface boundaries may play a role in enhancing the  potential for tornadogenesis in some situations. Additional research will pursue this issue through objective analysis of surface data.

    The importance of the relationship between mid-level dry intrusions and tornado outbreaks associated with landfalling TCs cannot be underestimated from an operational standpoint. The dry intrusions may be detected utilizing observational data, as demonstrated herein, but may also (at least in some cases) be discerned from operational NCEP model output regularly available to forecasters (Vescio et al. 1996). While not all tornado outbreaks associated with tropical cyclone landfalls involve dry intrusions, it is clear that the recurrence of the distinctive pattern seen in these cases should heighten forecaster vigilance when seen operationally as a storm approaches and makes landfall.

***
12 The findings are also consistent with updated findings reported by Rasmussen (2003).


REFERENCES

Bogner, P.B., G.M. Barnes and J.L. Franklin, 2000: Conditional instability and shear for six hurricanes over the Atlantic Ocean. Wea. Forecasting, 15, 192-207.

Cione, J.J., P.G. Black and S.H. Houston, 2000: Surface observations in the  hurricane environment. Mon. Wea. Rev., 128, 1550-1561.

Droegemeier, K. K., S. M. Lazarus, and R. Davies-Jones, 1993: The influence of helicity on numerically simulated convective storms. Mon. Wea. Rev., 121, 2005–2029.

Edwards, R., 1998: Tornado production by exiting tropical cyclones. Preprints, 23rd Conf. on Hurricanes and Tropical Meteorology, Dallas, Amer.Meteor. Soc., 485- 488.

_______, G.V. Rao and J.W. Scheck, 2000: Examination of tornadic supercells in tropical cyclone Earl (1998) using conventional and WSR-88D data suites. Preprints, 20th Conf. on Severe Local Storms, Orlando, FL, 97-100.

Fujita, T.T., 1981: Tornadoes and downbursts in the context of generalized planetary scales. J. Atmos. Sci., 38, 1511-1533.

Galway, J.G., 1975: Relationship of tornado deaths to severe weather watch areas. Mon. Wea. Rev., 103, 737-741.

________, 1977: Some climatological aspects of tornado outbreaks. Mon. Wea. Rev., 105, 477-484.

Gentry, R.C. 1983: Genesis of tornadoes associated with hurricanes. Mon. Wea. Rev.111, 1793-1805.

Gilmore, M. S., and L. J. Wicker, 1998: The influence of midtropospheric dryness on supercell morphology and evolution. Mon. Wea. Rev., 126, 943–958.

Glickman, T. Ed.,  2000: Glossary of Meteorology. 2d ed. Amer. Meteor. Soc., 855 pp.

Grazulis, T.P.,1993: Significant Tornadoes 1680–1991. Environmental Films, 1326 pp.

Hagemeyer, B.C. and S.M. Spratt, 2002: Thirty years after Hurricane Agnes – the forgotten Florida tornado disaster. Preprints, 25th Conference on Hurricanes and Tropical Meteorology, Amer. Meteor. Soc., San Diego, CA, 422-423.

Hart, J. A., and P. R. Janish, 1999: SeverePlot: Historical Severe Weather Report Database. Version 2.0. Storm Prediction Center, Norman, OK. [Available from http://www.spc.noaa.gov/software/svrplot2/]

Hill, E.L., W. Malkin and W.A. Schulz, Jr., 1966: Tornadoes associated with cyclones of tropical origin-practical features. J. Appl. Meteor., 5, 745-763.

Hudgins, J.E. and R.J. Frederick, 1997: Analysis of coastal front tornadoes  associated with Tropical Storm Josephine over eastern North Carolina on 8 October 1996. Eastern Region WSR-88D Oper.Note. No.11, 11 pp.

Johns, R.H., J.M. Davies, and P.W. Leftwich, 1993: Some wind and instability parameters associated with strong and violent tornadoes. Part II: Variations in the combinations of wind and instability parameters. The Tornado: Its Structure, Dynamics, Prediction and Hazards. Geophys. Mongr., 79, Amer. Geophys. Union, 583-590.

Kalnay, E., et al., 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.

Markowski, P.M., E.N. Rasmussen, and J.M. Straka, 1998: The occurrence of tornadoes in supercells interacting with boundaries during VORTEX-95. Wea. Forecasting, 13: 852-859.

McCaul, E.W. Jr., 1987: Observations of Hurricane Danny tornado outbreak of 16 August 1985. Mon. Wea. Rev., 115, 1206-1223.

_______, 1991: Buoyancy and shear characteristics of hurricane-tornado environments. Mon. Wea. Rev., 119, 1954-1978.

_______, and M.L. Weisman, 1996: Simulations of shallow supercell storms in landfalling hurricane environments. Mon. Wea. Rev., 124, 408-429.

_______, and C. Cohen, 2002: The impact on simulated storm structure and intensity of variations in the mixed layer and moist layer depths. Mon. Wea. Rev., 130,  1722-1748.

National Climate Data Center, 1961-1999: Storm Data. Vols. 3-41. National Oceanic and Atmospheric Administration.

Novlan, D.J. and W.M. Gray, 1974: Hurricane-spawned tornadoes. Mon. Wea. Rev., 102, 476-488.

Orton, R., 1970: Tornadoes associated with Hurricane Beulah on September 19–23, 1967. Mon. Wea. Rev., 98, 541–547.

Pietrycha, A. E., and C. D. Hannon, 2002: Tornado distribution associated with Hurricane Floyd 1999. Preprints, 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 619-622.

Rasmussen, E.N., 2003: Refined supercell and tornado forecast parameters. Wea. Forecasting, 18, 530-535.

________ and D.O. Blanchard, 1998: A baseline climatology of sounding-derived supercell and tornado forecast parameters. Wea. Forecasting, 13, 1148-1164.

Rodgers, E.B., W. S. Olson, V. M. Karyampudi, and H. F. Pierce, 1998: Satellite-derived latent heating distribution and environmental influences in Hurricane Opal (1995). Mon. Wea. Rev., 126, 1229-1247.

Romine, G. and R. Wilhelmson, 2002: Numerical investigation of the role of mid-level dryness on tropical mini-supercell behavior. Preprints, 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 631-634.

Saffir, H., and R. Simpson, 1974: The hurricane disaster-potential scale. Weatherwise, 27, 169 and 186.

Shay, L.K., G. J. Goni, P. G. Black, 2000: Effects of a warm oceanic feature on Hurricane Opal. Mon. Wea. Rev., 128, 1366–1383.

Shewchuk, J., 2002: RAOB (the Rawinsonde Observation Program) for Windows Ver. 5.1 Environmental Research Services. [Available from Environmental Research Services, 1134 Delaware Drive, Matamoros, PA 18336]

Spratt, S.M., D.W. Sharp, P. Welsh, A. Sandrik, F. Alsheimer, and C. Paxton, 1997: A WSR-88D assessment of tropical cyclone outer rainband tornadoes. Wea. Forecasting, 12, 479-501.

_______, F.D. Marks, Jr., P.P. Dodge, and D.W. Sharp, 2000: Examining the pre-landfall environment of mesovortices within a Hurricane Bonnie (1998) outer rainband. Preprints, 24th Conf. on Hurricanes and Tropical Meteorology, Amer. Meteor. Soc.,  Fort Lauderdale, FL, 300-301.

Stiegler, D.J. and T. T. Fujita, 1982: A detailed analysis of the San Marcos, Texas tornado induced by Hurricane Allen on 10 August, 1980". Preprints, 12th Conf. On Severe Local Storms, Amer. Meteor. Soc., San Antonio, TX, 371-374.

Thompson, R.L., R. Edwards and J.A. Hart, An assessment of supercell and tornado forecast parameters with RUC-2 model close proximity soundings, 2002: Preprints, 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 595-598.

Vescio, M. D., S.J. Weiss and F.P. Ostby, 1996: Tornadoes associated with Tropical Storm Beryl. N.W.A. Digest, 1, No. 21, Nat’l. Weather Assoc., 2-10.

Vasquez, T., 2000: Digital Atmosphere Ver. 2000 V1.0. Weather Graphics Technologies. [Available from Weather Graphics Technologies, 2110 Slaughter Lane, Ste. 115-101, Austin, TX 78748]

Weiss, S.J., 1987: Some climatological aspects of forecasting tornadoes associated with tropical cyclones, Preprints, 17th Conf. On Hurricanes and Tropical Meteorology, Amer. Meteor. Soc., Miami, FL, 160-163.
=====================================