• Banacos, P. C., , and Ekster M. L. , 2010: The association of the elevated mixed layer with significant severe weather events in the northeastern United States. Wea. Forecasting, 25, 10821102, doi:10.1175/2010WAF2222363.1.

    • Search Google Scholar
    • Export Citation
  • Barker, L. J., III, 2006: A potentially valuable WSR-88D severe storm pre-cursor signature in highly dynamic, low CAPE, high shear environments. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 3.6. [Available online at https://ams.confex.com/ams/pdfpapers/115129.pdf.]

  • Bothwell, P. D., , Hart J. A. , , and Thompson R. L. , 2002: An integrated three dimensional objective analysis scheme in use at the Storm Prediction Center. Preprints, 21st Conf. on Severe Local Storms/19th Conf. on Weather Analysis and Forecasting/15th Conf. on Numerical Weather Prediction, San Antonio, TX, Amer. Meteor. Soc., JP3.1. [Available online at https://ams.confex.com/ams/pdfpapers/47482.pdf.]

  • Brooks, H. E., , Doswell C. A. III, , and Kay M. P. , 2003: Climatological estimates of local daily tornado probability for the United States. Wea. Forecasting, 18, 626640, doi:10.1175/1520-0434(2003)018<0626:CEOLDT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Brotzge, J., , Erickson S. , , and Brooks H. , 2011: A 5-yr climatology of tornado false alarms. Wea. Forecasting, 26, 534544, doi:10.1175/WAF-D-10-05004.1.

    • Search Google Scholar
    • Export Citation
  • Burke, P. C., , and Schultz D. M. , 2004: A 4-yr climatology of cold-season bow echoes over the continental United States. Wea. Forecasting, 19, 10611074, doi:10.1175/811.1.

    • Search Google Scholar
    • Export Citation
  • Clark, M. R., 2009: The southern England tornadoes of 30 December 2006: Case study of a tornadic storm in a low CAPE, high shear environment. Atmos. Res., 93, 5065, doi:10.1016/j.atmosres.2008.10.008.

    • Search Google Scholar
    • Export Citation
  • Clark, M. R., 2011: Doppler radar observations of mesovortices within a cool-season tornadic squall line over the UK. Atmos. Res., 100, 749764, doi:10.1016/j.atmosres.2010.09.007.

    • Search Google Scholar
    • Export Citation
  • Coniglio, M. C., 2012: Verification of RUC 0–1-h forecasts and SPC mesoscale analyses using VORTEX2 soundings. Wea. Forecasting, 27, 667683, doi:10.1175/WAF-D-11-00096.1.

    • Search Google Scholar
    • Export Citation
  • Cope, A. M., 2004: An early morning mid-Atlantic severe weather episode: Short-lived tornadoes in a high-shear low-instability environment. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P1.4. [Available online at https://ams.confex.com/ams/11aram22sls/webprogram/Paper81834.html.]

  • Craven, J. P., , and Brooks H. E. , 2004: Baseline climatology of sounding derived parameters associated with deep, moist convection. Natl. Wea. Dig., 28, 1324.

    • Search Google Scholar
    • Export Citation
  • Davies, J. M., 1990: Midget supercell spawns tornadoes. Weatherwise, 43, 260261, doi:10.1080/00431672.1990.9929350.

  • Davis, J. M., , and Parker M. D. , 2012: Radar climatology of tornadoes occurring in high shear/low CAPE environments in the mid-Atlantic and Southeast. Preprints, 26th Conf. on Severe Local Storms, Nashville, TN, Amer. Meteor. Soc., 9.2. [Available online at https://ams.confex.com/ams/26SLS/webprogram/Paper211761.html.]

  • Dean, A. R., , and Schneider R. S. , 2008: Forecast challenges at the NWS Storm Prediction Center relating to the frequency of favorable severe storm environments. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 9A.2. [Available online at https://ams.confex.com/ams/pdfpapers/141743.pdf.]

  • Dean, A. R., , and Schneider R. S. , 2012: An examination of tornado environments, events, and impacts from 2003–2012. Preprints, 26th Conf. on Severe Local Storms, Nashville, TN, Amer. Meteor. Soc., 60. [Available online at https://ams.confex.com/ams/26SLS/webprogram/Paper211580.html.]

  • Doswell, C. A., III, , and Burgess D. W. , 1988: On some issues of United States tornado climatology. Mon. Wea. Rev., 116, 495501, doi:10.1175/1520-0493(1988)116<0495:OSIOUS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, , Davies-Jones R. , , and Keller D. L. , 1990: On summary measures of skill in rare event forecasting based on contingency tables. Wea. Forecasting, 5, 576585, doi:10.1175/1520-0434(1990)005<0576:OSMOSI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, , Brooks H. E. , , and Kay M. P. , 2005: Climatological estimates of daily local nontornadic severe thunderstorm probability for the United States. Wea. Forecasting, 20, 577595, doi:10.1175/WAF866.1.

    • Search Google Scholar
    • Export Citation
  • Edwards, R., , Dean A. R. , , Thompson R. L. , , and Smith B. T. , 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part III: Tropical cyclone tornadoes. Wea. Forecasting, 27, 15071519, doi:10.1175/WAF-D-11-00117.1.

    • Search Google Scholar
    • Export Citation
  • Evans, J. S., , and Doswell C. A. III, 2001: Examination of derecho environments using proximity soundings. Wea. Forecasting, 16, 329342, doi:10.1175/1520-0434(2001)016<0329:EODEUP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Evans, M., 2010: An examination of low CAPE/high shear severe convective events in the Binghamton, New York county warning area. Natl. Wea. Dig., 34, 129144.

    • Search Google Scholar
    • Export Citation
  • Godfrey, E. S., , Trapp R. J. , , and Brooks H. E. , 2004: A study of the pre-storm environment of tornadic quasi-linear convective systems. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc. 3A.5. [Available online at https://ams.confex.com/ams/pdfpapers/81388.pdf.]

  • Grumm, R. H., , and Glazewski M. , 2004: Thunderstorm types associated with the “broken-S” radar signature. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P7.1. [Available online at https://ams.confex.com/ams/pdfpapers/81537.pdf.]

  • Guyer, J. L., , and Dean A. R. , 2010: Tornadoes within weak CAPE environments across the continental United States. Preprints, 25th Conf. on Severe Local Storms, Denver, CO, Amer. Meteor. Soc., 1.5. [Available online at https://ams.confex.com/ams/pdfpapers/175725.pdf.]

  • Guyer, J. L., , Imy D. A. , , Kis A. , , and Venable K. , 2006: Cool season significant (F2-F5) tornadoes in the Gulf Coast states. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 4.2. [Available online at https://ams.confex.com/ams/pdfpapers/115320.pdf.]

  • Johns, R. H., 1993: Meteorological conditions associated with bow echo development in convective storms. Wea. Forecasting, 8, 294299, doi:10.1175/1520-0434(1993)008<0294:MCAWBE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Johns, R. H., , and Hirt W. D. , 1987: Derechos: Widespread convectively induced windstorms. Wea. Forecasting, 2, 3249, doi:10.1175/1520-0434(1987)002<0032:DWCIW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kis, A. K., , and Straka J. M. , 2010: Nocturnal tornado climatology. Wea. Forecasting, 25, 545561, doi:10.1175/2009WAF2222294.1.

  • Kuchera, E. L., , and Parker M. D. , 2006: Severe convective wind environments. Wea. Forecasting, 21, 595612, doi:10.1175/WAF931.1.

  • Lane, J. D., , and Moore P. D. , 2006: Observations of a non-supercell tornadic thunderstorm from terminal Doppler weather radar. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., P4.5. [Available online at https://ams.confex.com/ams/pdfpapers/115102.pdf.]

  • Markowski, P. M., , and Straka J. M. , 2000: Some observations of rotating updrafts in a low-buoyancy, highly sheared environment. Mon. Wea. Rev., 128, 449461, doi:10.1175/1520-0493(2000)128<0449:SOORUI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Marzban, C., 2004: The ROC curve and the area under it as performance measures. Wea. Forecasting, 19, 11061114, doi:10.1175/825.1.

  • Mason, I. B., 1982: A model for assessment of weather forecasts. Aust. Meteor. Mag., 30, 291303.

  • McAvoy, B. P., , Jones W. A. , , and Moore P. D. , 2000: Investigation of an unusual storm structure associated with weak to occasionally strong tornadoes over the eastern United States. Preprints, 20th Conf. on Severe Local Storms, Orlando, FL, Amer. Meteor. Soc., 182185.

  • McCaul, E. W., , and Weisman M. L. , 2001: The sensitivity of simulated supercell structure and intensity to variations in the shapes of environmental buoyancy and shear profiles. Mon. Wea. Rev., 129, 664687, doi:10.1175/1520-0493(2001)129<0664:TSOSSS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Parker, M. D., 2012: Impacts of lapse rates on low-level rotation in idealized storms. J. Atmos. Sci., 69, 538559, doi:10.1175/JAS-D-11-058.1.

    • Search Google Scholar
    • Export Citation
  • Rasmussen, E. N., , and Blanchard D. O. , 1998: A baseline climatology of sounding-derived supercell and tornado forecast parameters. Wea. Forecasting, 13, 11481164, doi:10.1175/1520-0434(1998)013<1148:ABCOSD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Schneider, R. S., , and Dean A. R. , 2008: A comprehensive 5-year severe storm environment climatology for the continental United States. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 16A.4. [Available online at https://ams.confex.com/ams/pdfpapers/141748.pdf.]

  • Schneider, R. S., , Dean A. R. , , Weiss S. J. , , and Bothwell P. D. , 2006: Analysis of estimated environments for 2004 and 2005 severe convective storm reports. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 3.5. [Available online at https://ams.confex.com/ams/pdfpapers/115246.pdf.]

  • Smith, B. T., , Guyer J. L. , , and Dean A. R. , 2008: The climatology, convective mode, and mesoscale environment of cool season severe thunderstorms in the Ohio and Tennessee Valleys, 1995–2006. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 13B.7. [Available online at https://ams.confex.com/ams/pdfpapers/141968.pdf.]

  • Smith, B. T., , Thompson R. L. , , Grams J. S. , , Broyles C. , , and Brooks H. E. , 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part I: Storm classification and climatology. Wea. Forecasting, 27, 11141135, doi:10.1175/WAF-D-11-00115.1.

    • Search Google Scholar
    • Export Citation
  • Smith, B. T., , Castellanos T. E. , , Winters A. C. , , Mead C. M. , , Dean A. R. , , and Thompson R. L. , 2013: Measured severe convective wind climatology and associated convective modes of thunderstorms in the contiguous United States, 2003–09. Wea. Forecasting, 28, 229236, doi:10.1175/WAF-D-12-00096.1.

    • Search Google Scholar
    • Export Citation
  • Thompson, R. L., , Edwards R. , , Hart J. A. , , Elmore K. L. , , and Markowski P. , 2003: Close proximity soundings within supercell environments obtained from the Rapid Update Cycle. Wea. Forecasting, 18, 12431261, doi:10.1175/1520-0434(2003)018<1243:CPSWSE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Thompson, R. L., , Edwards R. , , and Mead C. M. , 2004: An update to the supercell composite and significant tornado parameters. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P8.1. [Available online at https://ams.confex.com/ams/pdfpapers/82100.pdf.]

  • Thompson, R. L., , Mead C. M. , , and Edwards R. , 2007: Effective storm-relative helicity and bulk shear in supercell thunderstorm environments. Wea. Forecasting, 22, 102115, doi:10.1175/WAF969.1.

    • Search Google Scholar
    • Export Citation
  • Thompson, R. L., , Smith B. T. , , Grams J. S. , , Dean A. R. , , and Broyles C. , 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part II: Supercell and QLCS tornado environments. Wea. Forecasting, 27, 11361154, doi:10.1175/WAF-D-11-00116.1.

    • Search Google Scholar
    • Export Citation
  • Trapp, R. J., , Wheatley D. M. , , Atkins N. T. , , Przybylinski R. W. , , and Wolf R. , 2006: Buyer beware: Some words of caution on the use of severe wind reports in postevent assessment and research. Wea. Forecasting, 21, 408415, doi:10.1175/WAF925.1.

    • Search Google Scholar
    • Export Citation
  • van den Broeke, M. S., , Schultz D. M. , , Johns R. H. , , Evans J. S. , , and Hales J. E. , 2005: Cloud-to-ground lightning production in strongly forced, low-instability convective lines associated with damaging wind. Wea. Forecasting, 20, 517530, doi:10.1175/WAF876.1.

    • Search Google Scholar
    • Export Citation
  • Verbout, S. M., , Brooks H. E. , , Leslie L. M. , , and Schultz D. M. , 2006: Evolution of the U.S. tornado database: 1954–2003. Wea. Forecasting, 21, 8693, doi:10.1175/WAF910.1.

    • Search Google Scholar
    • Export Citation
  • Weisman, M. L., , and Trapp R. J. , 2003: Low-level mesovortices within squall lines and bow echoes. Part I: Overview and dependence on environmental shear. Mon. Wea. Rev., 131, 27792803, doi:10.1175/1520-0493(2003)131<2779:LMWSLA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Weiss, S. J., , Hart J. A. , , and Janish P. R. , 2002: An examination of severe thunderstorm wind report climatology. Preprints, 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 11B.2. [Available online at https://ams.confex.com/ams/pdfpapers/47494.pdf.]

  • Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences: An Introduction.Academic Press, 467 pp.

  • View in gallery

    Nationwide distribution of (a) all HSLC significant severe reports without the MUCAPE criterion, (b) all HSLC significant severe reports, (c) all HSLC significant tornado (≥EF2) reports, (d) all HSLC significant wind (≥65 kt) reports, and (e) all HSLC significant hail (≥2 in. in diameter) from 2006 to 2011, defined by SBCAPE ≤ 500 J kg−1, 0–6-km shear ≥ 18 m s−1, and MUCAPE ≤ 1000 J kg−1.

  • View in gallery

    Box-and-whiskers plots showing the distributions of (a) SBLCL, (b) SBCIN, (c) MUCAPE, and (d) LLLR for all HSLC significant wind reports prior to enforcing the MUCAPE criterion by region. The 25th and 75th percentiles are noted by the box, with the median noted by a horizontal line within the box. Whiskers extend to the 10th and 90th percentiles, and additional outliers are plotted as crosses. (e) Subjectively defined regions as referred to in the text. Region labels: NW, Northwest; NR, northern Rockies; NP, northern plains; UM, upper Midwest; EGL, eastern Great Lakes; NE, Northeast; SW, Southwest; FC, Four Corners; SP, southern plains; LMV, lower Mississippi valley; and SA, South Atlantic.

  • View in gallery

    Annual cycle of nationwide HSLC significant severe reports and nulls separated by report type.

  • View in gallery

    Annual cycles of (a) all HSLC significant severe reports, (b) HSLC significant tornado reports, (c) HSLC significant wind reports, and (d) HSLC significant hail reports by regions. Colors in bar graphs correspond to region colors in Fig. 2e.

  • View in gallery

    Diurnal cycle of nationwide HSLC significant severe reports and nulls separated by report type.

  • View in gallery

    (a) TSS for discriminating between HSLC significant severe reports and nulls as a function of parameter threshold value for the given composite parameters in the development dataset. The y axis represents each parameter’s TSS (i.e., difference between POD and POFD) calculated for each possible threshold value along the x axis (i.e., this is a representation of a parameter’s potential skill, not its distribution). Craven–Brooks and VGP values are normalized such that their optimal values of 20 000 and 0.2, respectively, are represented by 1 on the x axis. (b) ROC curve depicting parameter skill for discriminating between HSLC significant severe reports and nulls within the development dataset. The thick dotted line represents no skill (i.e., TSS = 0), and the thin dotted lines represent constant TSS contours, increasing by 0.1 toward the top left. Filled circles represent POFD and POD at max TSS for the given parameter.

  • View in gallery

    As in Fig. 6, but for the nationwide verification dataset.

  • View in gallery

    Max TSS for given composite parameters for discriminating between HSLC significant severe reports and nulls across regions associated with at least 50 HSLC significant severe reports between 2006 and 2011. Dots indicate max TSSs calculated using datasets excluding reports and nulls from development dataset CWAs, ensuring truly independent datasets in LMV and SA.

  • View in gallery

    As in Fig. 6, but for the nationwide verification dataset across all environments.

  • View in gallery

    As in Fig. 6, but for significant tornadoes against nulls.

  • View in gallery

    As in Fig. 6, but for significant tornadoes against nulls within the verification dataset.

  • View in gallery

    As in Fig. 6, but for significant tornadoes against nulls across all environments in the verification dataset.

  • View in gallery

    Box-and-whiskers diagrams for distributions of (a) SHERBS3, (b) SHERBE, (c) STP (effective version), and (d) EHI for (left) HSLC significant severe reports, (middle) HSLC significant tornado reports, and (right) HSLC nulls. Box-and-whiskers diagrams are defined in Fig. 2.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 492 492 110
PDF Downloads 544 544 96

Climatology and Ingredients of Significant Severe Convection in High-Shear, Low-CAPE Environments

View More View Less
  • 1 Department of Marine, Earth, and Atmospheric Sciences, North Carolina State University, Raleigh, North Carolina
© Get Permissions
Full access

Abstract

High-shear, low-CAPE (HSLC) environments, here characterized by surface-based CAPE ≤ 500 J kg−1, most unstable parcel CAPE ≤ 1000 J kg−1, and 0–6-km shear vector magnitude ≥ 18 m s−1, occur at all times of day, across all seasons, and throughout the entire United States. HSLC environments represent a unique challenge for forecasters, as they occur frequently but produce severe weather a relatively low percentage of the time. Recent studies have primarily focused on improving nowcasting and warnings for events through the identification of radar signatures commonly associated with HSLC tornadoes. Few studies have investigated the forecasting of HSLC severe weather, despite the acknowledged poor performance of traditional tools and techniques. A general climatology of HSLC significant severe weather is presented, focusing on regional, diurnal, and annual trends. Through this climatology, it becomes apparent that multiple types of HSLC environments are possible, including surface-based cases with low lifted condensation levels and high-based convection cases. A statistical analysis of HSLC events and nulls from the southeastern and mid-Atlantic states is utilized to assess the performance of conventional composite parameters in HSLC environments. Additionally, a new composite parameter is introduced that utilizes the product of the statistically most skillful parameters in HSLC environments: the 0–3-km lapse rate, the 700–500-hPa lapse rate, and multiple wind and shear metrics. The strengths and weaknesses of these ingredients-based techniques are then reviewed, with an eye toward improving future HSLC severe weather forecasts.

Corresponding author address: Keith Sherburn, North Carolina State University, Campus Box 8208, Raleigh, NC 27695-8208. E-mail: kdsherbu@ncsu.edu

Abstract

High-shear, low-CAPE (HSLC) environments, here characterized by surface-based CAPE ≤ 500 J kg−1, most unstable parcel CAPE ≤ 1000 J kg−1, and 0–6-km shear vector magnitude ≥ 18 m s−1, occur at all times of day, across all seasons, and throughout the entire United States. HSLC environments represent a unique challenge for forecasters, as they occur frequently but produce severe weather a relatively low percentage of the time. Recent studies have primarily focused on improving nowcasting and warnings for events through the identification of radar signatures commonly associated with HSLC tornadoes. Few studies have investigated the forecasting of HSLC severe weather, despite the acknowledged poor performance of traditional tools and techniques. A general climatology of HSLC significant severe weather is presented, focusing on regional, diurnal, and annual trends. Through this climatology, it becomes apparent that multiple types of HSLC environments are possible, including surface-based cases with low lifted condensation levels and high-based convection cases. A statistical analysis of HSLC events and nulls from the southeastern and mid-Atlantic states is utilized to assess the performance of conventional composite parameters in HSLC environments. Additionally, a new composite parameter is introduced that utilizes the product of the statistically most skillful parameters in HSLC environments: the 0–3-km lapse rate, the 700–500-hPa lapse rate, and multiple wind and shear metrics. The strengths and weaknesses of these ingredients-based techniques are then reviewed, with an eye toward improving future HSLC severe weather forecasts.

Corresponding author address: Keith Sherburn, North Carolina State University, Campus Box 8208, Raleigh, NC 27695-8208. E-mail: kdsherbu@ncsu.edu

1. Introduction

A broad spectrum of CAPE and shear combinations is capable of producing severe weather (Schneider and Dean 2008). Severe convective storms in environments with large vertical wind shear1 but meager instability [high-shear, low-CAPE (HSLC) environments] have received only modest attention in the literature compared to their higher-CAPE counterparts. HSLC environments2 are a challenge for operational meteorologists, as reflected by a relatively high number of false alarm hours and relatively low probabilities of detection of associated tornado watches when compared to higher-CAPE environments (Dean and Schneider 2008; Schneider and Dean 2008; Dean and Schneider 2012). HSLC events have been noted typically to occur in the cool season (Burke and Schultz 2004; Guyer et al. 2006; Smith et al. 2008) and overnight (Kis and Straka 2010), corresponding with the most likely time periods for false alarm tornado warnings (Brotzge et al. 2011). Schneider et al. (2006) found that environments with less than 1000 J kg−1 of mixed layer (ML) CAPE (MLCAPE) and over 18 m s−1 of 0–6-km shear often occurred in the southeastern United States coupled with low lifted condensation levels (LCLs). They designated these environments as the second “key subclass” of severe weather in the country, accounting for more significant tornadoes than the conventional central plains environment characterized by MLCAPE over 2000 J kg−1 (Schneider et al. 2006).

To the authors’ knowledge, the only relevant nationwide environmental climatology for HSLC severe weather was developed by Guyer and Dean (2010), who investigated the climatology of tornadoes occurring in environments with low values (i.e., ≤500 J kg−1) of MLCAPE. They found that weak CAPE tornadoes are a nationwide phenomenon, although significant tornadoes were primarily found east of the Rocky Mountains, particularly from the lower Mississippi valley eastward to the Atlantic coast. Compared to tornadoes in higher-CAPE environments, weak CAPE tornadoes were relatively more common during the overnight/morning hours and cool season. However, Guyer and Dean (2010) did not consider the magnitude of deep-layer shear in their climatology, nor did they investigate severe hail or wind reports. Thus, a specific climatology of HSLC severe convection represents a current gap in the knowledge base.

The main goal of the present work is to improve our understanding and forecasting of HSLC severe events. Case studies of severe HSLC events have noted that a dry intrusion aloft may coincide with the onset or strengthening of existing convection, perhaps indicating the release of potential instability (Johns 1993; Lane and Moore 2006; Clark 2009; Evans 2010), while modeling studies (e.g., McCaul and Weisman 2001) have shown that, in weak CAPE environments, robust convection is possible when instability is focused in the lowest levels. Recent studies have largely been focused on cataloging unique radar signatures associated with HSLC severe convection (e.g., McAvoy et al. 2000; Grumm and Glazewski 2004; Barker 2006; Clark 2011), rather than exploring environmental characteristics.

Conventional techniques for forecasting significant severe weather have been noted to perform poorly in HSLC environments (Guyer and Dean 2010; R. Thompson 2012, personal communication). Additionally, the compressed nature of HSLC convection (Davies 1990; Markowski and Straka 2000; McAvoy et al. 2000; Cope 2004) may result in poor radar sampling, as shallow or small-scale circulations may not be adequately resolved by the existing Weather Surveillance Radar-1988 Doppler (WSR-88D) network, particularly at greater distances from the radar (Davis and Parker 2012). HSLC environments can also produce severe weather with little or no cloud-to-ground lightning (Davies 1990; McAvoy et al. 2000; Cope 2004; van den Broeke et al. 2005). The combination of the preceding factors suggests a critical need for a better understanding of the environmental features discriminating between HSLC severe convection and nonsevere convection, both to improve operational forecasts and to advance our understanding.

This article presents a climatology of HSLC significant severe convection [producing tornadoes rated as category 2 or higher on the enhanced Fujita (EF) scale, wind gusts ≥ 65 kt (1 kt = 0.51 m s−1), or hail ≥ 2 in. (1 in. ≈ 2.54 cm) diameter], including an investigation of regional, annual, and diurnal trends. Statistical analysis of HSLC significant severe convection versus nonsevere HSLC convection is utilized to recommend environmental parameters that improve detection and limit false alarms during HSLC events. Section 2 describes the data and methods utilized for the study. The climatology of HSLC significant severe weather is provided in section 3. An overview of the primary discriminating parameters between HSLC significant severe convection and nonsevere convection is given in section 4, including the development of composite parameters specifically designed for HSLC environments. Section 5 discusses the most important findings of the climatology and the caveats associated with HSLC composite parameters. Last, section 6 offers a summary, including recommendations for operational meteorologists and avenues for future research.

2. Data and methods

a. Development dataset

Because HSLC severe weather is one of the most common challenges for forecasters in the southeastern United States (Dean and Schneider 2012; J. Lane and P. Moore 2012, personal communication), we began by creating a development (or “training”) dataset for this region. Collaborators from 11 National Weather Service (NWS) Weather Forecast Offices (WFOs) across the Southeast and mid-Atlantic3 regions subjectively compiled a list of HSLC events in their respective county warning areas (CWAs) between fall 2006 and spring 2011. HSLC events for the purposes of the development dataset were defined as those with surface-based (SB) CAPE ≤ 500 J kg−1 and 0–6-km shear ≥ 18 m s−1. No criterion for most unstable parcel (MU) CAPE was included in the development dataset because a negligible fraction of the events were found to be associated with elevated convection. Events associated with landfalling tropical cyclones were purposefully excluded from this dataset to maintain a focus on midlatitude systems. The Storm Prediction Center (SPC) provided a relational database consisting of well-known convective environmental parameters for the 6245 severe reports [from the National Climatic Data Center’s (NCDC) Storm Data publication] that occurred in the collaborating CWAs during the 107 identified events.

For each report, the relational database included the archived surface objective analysis4 (SFCOA; Bothwell et al. 2002) value for 90 standard parameters (given in the appendix) at the nearest grid point for the preceding hour. Though the SFCOA system ingests surface observations, upper-air features during our period of study were dependent on fields from the Rapid Update Cycle (RUC) model; thus, biases and errors were possible, as documented by Thompson et al. (2003) and more recently by Coniglio (2012). In addition, the SFCOA resolution is less than ideal, with 40-km horizontal grid spacing and archived data available only once every hour. These spatial and temporal resolutions could lead to smoothing or misplacement of significant mesoscale features. Additionally, using the nearest grid point could occasionally result in the sampling of a different air mass if the report occurred close to a boundary (Guyer and Dean 2010). However, through personal communication with forecasters at numerous WFOs, it was clear that SFCOA fields are commonly used during operations, and despite its limitations, Coniglio (2012) showed that the SFCOA scheme markedly improves on raw RUC model analyses. A more thorough discussion on utilizing the SFCOA for research was provided by Thompson et al. (2012).

Because the SPC relational database included all regional reports between the given start and end times for each event, a large percentage of the reports in the database were not actually HSLC reports, as determined by the SFCOA data. For our formal analysis, we only retained a particular CWA’s reports for an event if over half of those reports met the HSLC surface-based CAPE (SBCAPE) and 0–6-km shear criteria defined above. Events fulfilling these criteria in one or more of the CWAs were designated as HSLC events. To prevent dataset bias toward events with particularly widespread reports, only one report was retained per CWA per hour. Tornadoes had the highest priority in this filtering process, followed by wind and hail reports, respectively. In other words, a wind report was used if there were no tornadoes in the CWA during that hour, while a hail report was retained if there were no wind reports or tornadoes in the same CWA for a given hour. Each report type was subsequently sorted by magnitude prior to filtering (i.e., an EF3 tornado would take precedence over an EF2 tornado if both occurred in the same CWA and hour). The final, filtered dataset consisted of 943 reports (one per CWA per HSLC hour), including 80 significant reports (35 significant tornadoes, 44 significant wind reports, and 1 significant hail report). Through discussion with NWS and SPC collaborators, it was decided that the most operationally beneficial comparison would be between significant severe reports (i.e., tornadoes ≥ EF2, wind gusts ≥ 65 kt, and/or hail ≥ 2 in. diameter) and nulls (defined next). This approach helps to avoid some of the potential nonmeteorological factors associated with severe reports, such as incorrect estimations of wind speeds, inflation of severe reports with population increase, the decrease of reports overnight, and variance in WFO verification techniques (e.g., Weiss et al. 2002; Doswell et al. 2005; Trapp et al. 2006; Smith et al. 2012; Smith et al. 2013). The authors acknowledge that not all concerns can be alleviated by only utilizing significant severe reports; for example, only 1 of the 44 significant wind reports had a corresponding measured (i.e., not estimated) wind speed in NCDC’s Storm Data, and, as will be shown later, considerable variability in report numbers across CWA boundaries remains a problem. However, this choice gives us higher confidence in the validity of the reports that remain in our development dataset.

To address the difficulty of false alarms in forecasting HSLC severe convection, we also developed a null dataset to compare with the events dataset. A null was defined by the initial latitude–longitude point of a severe thunderstorm or tornado warning that was issued in an HSLC environment when there were no severe reports from the Storm Data archive in the corresponding CWA throughout that convective day (1200–1200 UTC). This definition allowed us to compare the environments of verified severe HSLC convection (i.e., events) with HSLC environments of convection that appeared to be severe to experienced operational meteorologists (i.e., nulls). Due to some of the ambiguities and biases in the density, timing, and position of storm reports (e.g., Doswell and Burgess 1988, Doswell et al. 2005, Trapp et al. 2006, Verbout et al. 2006), we did not attempt to identify nulls on a warning-by-warning basis; instead, we enforced the rather strong requirement of no severe reports in a CWA for the entire day. In total, 103 null points were identified within collaborating CWAs from October 2006 through April 2011. Environmental data for the nulls were again obtained via archived SFCOA fields.

Environments of HSLC significant severe reports and null points within the development dataset were compared via statistical analyses, focusing on parameters that showed high probability of detection (POD) and a low probability of false detection (POFD; Doswell et al. 1990). The true skill statistic (TSS; Wilks 1995) was utilized in order to determine which environmental parameters in the relational databases were most skillful at discriminating between events and nulls. The TSS is defined by
eq1
where a is a correct forecast of a significant severe report, b is a false alarm, c is a missed significant severe report, and d is a correct nonforecast. The value of the TSS is also equivalent to the difference between the POD and POFD (Doswell et al. 1990), meaning that parameters with high TSSs have an optimal combination of detecting events without misidentifying nulls. The TSS was chosen for this study because it includes all four possibilities (ad) in the standard contingency table, including correct nulls (e.g., a category that is not practical for studies of warning verification). We view the inclusion of correct nulls as particularly important because a forecaster must be able to trust a parameter when it predicts a nonevent in addition to when it predicts an event. Finally, previous parameter-based studies (e.g., Thompson et al. 2003, 2004) have utilized the TSS for determining the optimal value of the composite parameters such as the significant tornado parameter (STP5). Receiver operating characteristic (ROC; Mason 1982) curves, and the subsequent integrated area under each ROC curve (AUC), were calculated as another measure of skill. Values of the AUC range from 0 to 1, with AUC < 0.5 representing negative skill (worse than a chance forecast), AUC = 0.5 representing equivalent skill to a chance forecast, and AUC = 1 signifying a perfect forecast (Marzban 2004). Heidke skill scores (HSSs; Wilks 1995) were also calculated, and trends were generally similar to those of the TSS and AUC. HSS values tended to diverge from the TSS in situations where the sample sizes of events and nulls differed considerably. In these cases, the TSS tended to value high POD, while the HSS preferred low false alarm ratios.

After identifying the individual parameter with the highest TSSs among all 90 environmental parameters in the development dataset, subsequent tests were run in order to decrease the number of false alarms that would be incurred by using just the first environmental parameter alone. This was done because false alarms outnumbered misses when using just the first environmental parameter. All of the reports above the most skillful environmental parameter’s optimal value were used to identify a second conditionally most skillful environmental parameter (having the highest TSS in this round). A third round of tests was then conducted in a similar fashion to determine the third conditionally most skillful environmental parameter. Given the modest size of the development dataset, results were confirmed through a series of Monte Carlo random sampling tests (Wilks 1995). Herein, the TSS for every environmental parameter was recalculated over 10 000 random samples of the original dataset. For the first two rounds of Monte Carlo TSS tests, the sample sizes were 30 reports and 30 nulls; however, due to the small number of nulls in the third round of tests, only 10 reports and nulls were used for the last round. These tests largely confirmed the original findings (i.e., the same environmental parameters had the highest average TSSs across all simulations in the respective round), except in the case of the third conditionally most skillful environmental parameter, for which there were a number of nearly equal options. In developing our composite diagnostic parameters, we considered each of these nearly equal third ingredients in turn, as described and assessed in section 4.

b. Verification dataset

After the initial assessment of the environmental parameters in our southeastern U.S. development dataset, an independent verification dataset of reports and nulls was used to test the robustness of our findings. For this purpose, the SPC provided a second relational database consisting of all significant severe reports (from NCDC’s Storm Data) and null points (as defined in section 2a) from the contiguous United States from 2006 to 2011 (the nulls actually began in October 2006 due to missing data in the archive). As with the development dataset, the nulls were identified on a CWA-by-CWA basis. For this analysis, we used only those reports and nulls meeting our HSLC criteria of SBCAPE ≤ 500 J kg−1 and 0–6-km shear ≥ 18 m s−1. In the course of subsequent analysis it became clear that, in the central United States, many of the reports were associated with elevated convection that had low SBCAPE but high most unstable CAPE (MUCAPE). Therefore, we added the requirement of MUCAPE ≤ 1000 J kg−1 to prevent the inclusion of such elevated convection in the verification dataset (since it was not truly “low CAPE”). Among the development dataset’s significant severe reports, the average ratio between SBCAPE and MUCAPE was approximately 0.39. In other words, our threshold value of SBCAPE = 500 J kg−1 would on average correspond to MUCAPE = 1280 J kg−1, so our criterion of MUCAPE is a somewhat more stringent restriction than what would naturally emerge from the HSLC reports inspiring this work. Additionally, the MUCAPE ≤ 1000 J kg−1 criterion is consistent with the threshold of low MLCAPE utilized by Schneider et al. (2006). The ramifications of the additional MUCAPE restriction are addressed in section 3. Overall, the HSLC verification dataset included 1478 HSLC significant severe reports—200 EF2+ tornadoes, 1046 wind reports ≥ 65 kt, and 232 hail reports ≥ 2 in. diameter—and 1028 HSLC nulls. To test the regional and nationwide applicability of the results from the development dataset, the TSS was again used. Additionally, two-sample t tests (Wilks 1995) were administered and their associated p values were calculated in order to assess the statistical significance of differences in means between the reports and nulls datasets.

3. HSLC climatology

Our original identification method revealed HSLC events in the vast majority of CWAs nationwide (Fig. 1a). Comparing environmental characteristics for significant wind reports6 across subjectively defined regions (see Fig. 2e), it became evident that reports in the northern plains and upper Midwest were associated with higher magnitudes of SB convective inhibition (CIN; Fig. 2b) and MUCAPE (Fig. 2c), suggesting that many of these storms were likely elevated. The impact of utilizing the additional MUCAPE ≤ 1000 J kg−1 criterion in our HSLC definition is reflected by the differences between Figs. 1a and 1b. Much of the signal in the northern plains and upper Midwest (Fig. 1a) was evidently attributable to reports with high MUCAPE. These high-MUCAPE reports were excluded from the subsequent climatology and analysis since they are rather different from the spirit of our “low CAPE” study. When we remade Figs. 2a,b,d using the MUCAPE ≤ 1000 J kg−1 condition (not shown), the distributions showed only subtle changes apart from a noticeable upward shift in SBCIN in the northern plains due to the removal of the presumably elevated cases.

Fig. 1.
Fig. 1.

Nationwide distribution of (a) all HSLC significant severe reports without the MUCAPE criterion, (b) all HSLC significant severe reports, (c) all HSLC significant tornado (≥EF2) reports, (d) all HSLC significant wind (≥65 kt) reports, and (e) all HSLC significant hail (≥2 in. in diameter) from 2006 to 2011, defined by SBCAPE ≤ 500 J kg−1, 0–6-km shear ≥ 18 m s−1, and MUCAPE ≤ 1000 J kg−1.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Fig. 2.
Fig. 2.

Box-and-whiskers plots showing the distributions of (a) SBLCL, (b) SBCIN, (c) MUCAPE, and (d) LLLR for all HSLC significant wind reports prior to enforcing the MUCAPE criterion by region. The 25th and 75th percentiles are noted by the box, with the median noted by a horizontal line within the box. Whiskers extend to the 10th and 90th percentiles, and additional outliers are plotted as crosses. (e) Subjectively defined regions as referred to in the text. Region labels: NW, Northwest; NR, northern Rockies; NP, northern plains; UM, upper Midwest; EGL, eastern Great Lakes; NE, Northeast; SW, Southwest; FC, Four Corners; SP, southern plains; LMV, lower Mississippi valley; and SA, South Atlantic.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Even after the removal of the high-MUCAPE cases, HSLC significant severe reports occurred in nearly every NWS CWA from 2006 to 2011 (Fig. 1b). A corridor of enhanced numbers of HSLC significant severe reports extends from the Midwest and central plains through the Mississippi and Ohio River valleys and into the Southeast. HSLC significant tornadoes (Fig. 1c) had a clear maximum centered in the Tennessee and lower Mississippi valleys, with modest numbers extending into the Midwest, the Ohio valley, and the mid-Atlantic. Significant winds within HSLC environments were somewhat less concentrated, with numerous reports scattered across the lower Ohio valley, the Midwest, the plains, and into the lower Mississippi valley (Fig. 1d). HSLC significant hail reports were maximized across the plains and the upper Midwest (Fig. 1e). Some local maxima and minima were likely attributable to nonmeteorological factors associated with differing local warning verification techniques (e.g., Doswell et al. 2005), but the regional maxima in significant reports were consistent with previous climatologies of HSLC convection (e.g., Schneider et al. 2006).

Annually, both HSLC significant severe reports and nulls have an early spring peak, followed by a decrease through the summer and early fall with only a weak secondary maximum in October (Fig. 3). July–December saw higher numbers of nulls than significant reports, suggesting a decrease in warning skill during those months (although the sample size is rather small in many of these months). A considerable fraction of the nulls during this time period, particularly through the summer and early fall, may be associated with landfalling tropical cyclones prompting unverified tornado warnings, given that the increase in nulls corresponds to the annual cycle of tropical cyclone tornadoes (Edwards et al. 2012). When breaking the annual cycle down by subjectively defined regions (see Fig. 2e), many similarities to previous climatologies of severe weather (e.g., Brooks et al. 2003; Doswell et al. 2005; Schneider et al. 2006) became evident (Fig. 4a). Fall and winter (September–February) HSLC significant severe reports predominantly occur in the South Atlantic and lower Mississippi valley regions. In the spring (March–May), reports became increasingly common in the southern plains and the upper Midwest, though the population remains dominated by the southeastern United States. By summer (June–August), the plains region accounts for a large fraction of HSLC significant severe reports. HSLC significant tornado reports (Fig. 4b) have a February–April maximum in the lower Mississippi valley and South Atlantic, accounting for most of the overall distribution shown in Fig. 1c and consistent with the annual cycle of Guyer et al. (2006). A clear annual minimum in HSLC tornadoes occurs in the summer, presumably because CAPE is typically high and shear is typically low over much of the United States. The HSLC significant wind reports (Fig. 4c) have similar trends to the total HSLC significant reports (cf. Fig. 4a), largely because the majority of HSLC reports are associated with severe winds. The HSLC significant hail threat (Fig. 4d) was largely nonexistent through the winter, with a rapid increase in spring. As with significant winds, by summer, the threat was primarily in the plains and upper Midwest.

Fig. 3.
Fig. 3.

Annual cycle of nationwide HSLC significant severe reports and nulls separated by report type.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Fig. 4.
Fig. 4.

Annual cycles of (a) all HSLC significant severe reports, (b) HSLC significant tornado reports, (c) HSLC significant wind reports, and (d) HSLC significant hail reports by regions. Colors in bar graphs correspond to region colors in Fig. 2e.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

HSLC events can occur at all hours of the day (Fig. 5); however, an evening peak was noted in the diurnal cycle of HSLC significant severe events, with a mid- to late-morning minimum. Notably, false alarms (nulls) were proportionally larger during the overnight and morning (the null counts roughly equal or surpass the significant severe reports from 0300 to 1000 local solar time). This finding is consistent with previous work on tornado false alarms by Brotzge et al. (2011).

Fig. 5.
Fig. 5.

Diurnal cycle of nationwide HSLC significant severe reports and nulls separated by report type.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

As discussed previously, given the regional differences in timing and in predominant HSLC severe weather type, we studied the typical values of environmental parameters in significant wind events by region. In addition to the surface-based, low-LCL cases common to the southeastern United States, and the excluded high-MUCAPE (presumably elevated) convection common to the northern plains and Midwest, it also appeared that some HSLC events were high-based storms in environments with dry boundary layers. There were a considerable number of reports in the dataset with SB LCLs near or above 1 km (Fig. 2a), especially in the northern Rockies (NR) and Northwest (NW) where well-mixed afternoon boundary layers and associated steep low-level lapse rates are common (Fig. 2d). It is likely that environments for high-based convection are considerably different from the moist, surface-based HSLC environments of the southeastern United States, from which our development dataset was primarily drawn. We explore the possibilities and limitations of a “one size fits all” HSLC composite parameter in section 4.

4. An HSLC composite parameter

a. Parameter development and justification

Through the multiple rounds of TSS tests described in section 2a, it was determined that the two conditionally most skillful parameters in discriminating between HSLC significant severe convection (tornadoes, winds, and hail) and nulls in the development dataset were the 0–3-km lapse rate (i.e., low-level lapse rate or LLLR) and the 700–500-hPa lapse rate (LR75). The third conditionally most skillful parameter was less well defined due to a vastly decreased sample size in the third round of TSS tests. The initial TSS tests and subsequent Monte Carlo simulations revealed a number of wind and shear parameters that have comparable skill (listed in Table 1).

Table 1.

Wind and shear magnitude parameters exhibiting skill as the third conditionally most skillful parameter in the development dataset using TSS tests or Monte Carlo simulations. SHERB stands for the severe hazards in environments with reduced buoyancy parameter.

Table 1.

On its own, LLLR had the highest skill among tested parameters with a maximum TSS of 0.342. This value exceeded the TSS of even well-known composite parameters (the STP was the best of them with a TSS of 0.332). Through the procedure in section 2a, it was further determined that by utilizing LR75 as the second parameter in tandem with LLLR, the TSS increased to 0.463.7 Due to the assortment of wind and shear values that were roughly equal as the third-most conditionally skillful ingredient, composite parameters consisting of products of the LLLR, LR75, and each wind–shear metric were created. Each environmental parameter was normalized by its optimal threshold value (i.e., the threshold at which its skill was maximized; the normalization of the wind–shear term was adjusted slightly in order to produce an optimal threshold value of approximately 1 for the final composite parameter within the development dataset).

As shown in Table 2, the most skillful combined parameter within our development dataset (at discriminating between HSLC significant severe reports and nulls) was the version of the severe hazards in environments with reduced buoyancy parameter (SHERB) using the 6-km wind magnitude (or SHERBW6; this and other abbreviations are given in Table 1). This was also reflected in the integrated AUC metric for each parameter (Table 2). However, many other such products of LLLR, LR75, and a shear–wind field were more skillful than conventional parameters. Although they appeared to be conditionally skillful, formulations using storm-relative helicity (SHERBH1 and SHERBH3), the 1-km wind magnitude (SHERBW1), and the 0–1-km shear magnitude (SHERBS1) were less skillful overall within the development dataset when combined with the lapse rates (Table 2). When discriminating between HSLC significant tornadoes and nulls in the development dataset (Table 3), five of the formulations remained especially skillful (SHERBE, SHERBW3, SHERBS3, SHERBW6, and SHERBS6), while the aforementioned low-skill formulations were again poor discriminators; as such, they were removed from consideration for our subsequent analysis.8

Table 2.

Max TSS, optimal threshold, and integrated AUC for given composite parameters for discriminating between HSLC significant severe reports and nulls within the development dataset. Composite parameters include the Craven–Brooks significant severe parameter and STP with effective-layer version (STP-E) and with fixed-layer version (STP-F).

Table 2.
Table 3.

As in Table 2, but for HSLC significant tornado reports and nulls.

Table 3.

By utilizing the nationwide verification dataset, it became apparent that SHERBE and SHERBS3 were the most skillful SHERB parameters. Table 4 shows the maximum TSS in discriminating between HSLC significant severe reports (also divided into tornado, wind, and hail reports) versus nulls in the verification dataset, using the threshold that maximized the TSS for each individual parameter. More telling were the skill scores using the parameters’ conventional thresholds, shown in Table 5. These statistics reveal that SHERBE was the best-performing composite parameter at its designed threshold within the nationwide HSLC verification dataset. In contrast, the thresholds of conventional composite parameters must be adjusted downward to maximize their skill, both in the development dataset (Fig. 6a) and the verification dataset (Fig. 7a). At their designed thresholds, SHERBS3 and SHERBE also outperform all conventional parameters at separating significant tornadoes from nulls (Table 5). As a result, the remainder of section 4 will focus on assessing the skill of these two parameters.

Table 4.

Max TSS using any threshold for all HSLC significant severe reports against nulls (second column), HSLC significant tornadoes against nulls (third column), HSLC significant winds against nulls (fourth column), and HSLC significant hail reports against nulls (fifth column) within the nationwide verification dataset.

Table 4.
Table 5.

As in Table 4, but for conventional parameter threshold values (Craven–Brooks, 20 000; EHI, 1; STP-E, 1; STP-F, 1; VGP, 0.2; SHERB parameters, 1).

Table 5.
Fig. 6.
Fig. 6.

(a) TSS for discriminating between HSLC significant severe reports and nulls as a function of parameter threshold value for the given composite parameters in the development dataset. The y axis represents each parameter’s TSS (i.e., difference between POD and POFD) calculated for each possible threshold value along the x axis (i.e., this is a representation of a parameter’s potential skill, not its distribution). Craven–Brooks and VGP values are normalized such that their optimal values of 20 000 and 0.2, respectively, are represented by 1 on the x axis. (b) ROC curve depicting parameter skill for discriminating between HSLC significant severe reports and nulls within the development dataset. The thick dotted line represents no skill (i.e., TSS = 0), and the thin dotted lines represent constant TSS contours, increasing by 0.1 toward the top left. Filled circles represent POFD and POD at max TSS for the given parameter.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Fig. 7.
Fig. 7.

As in Fig. 6, but for the nationwide verification dataset.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

SHERBE (using effective9 shear or ESHR) is specifically formulated as
eq2
Although SHERBE was the most skillful parameter at its designed threshold, we noted that some HSLC significant severe reports occurred in environments with diagnosed MUCAPE of 0 J kg−1 in the SFCOA. This is problematic because ESHR is by definition 0 m s−1 when there is no CAPE (Thompson et al. 2007), which would in turn cause SHERBE to be 0. Therefore, as a practical consideration, we also tested the most skillful fixed-layer version of SHERB, which was formulated using the 0–3-km shear magnitude (S3MG):
eq3
Within the development dataset, analysis of maximum TSS and ROC curves revealed the evident improvement of skill over existing composite parameters accomplished by utilizing SHERBS3 and SHERBE (Fig. 6). An improvement in skill was also identified in the nationwide verification dataset (Fig. 7), despite some regional variability (Fig. 8). In all of the regions with at least 50 HSLC significant severe reports, either SHERBS3 or the SHERBE was the most skillful parameter at discriminating between HSLC significant severe reports and nulls. The differences between significant severe reports and nulls for the SHERBS3 (SHERBE) distributions are statistically significant at the 99% level for all regions in Fig. 8 except the northern Rockies, where the difference is statistically significant at the 95% level (or, for SHERBE, not statistically significant). To ensure that the skill in the South Atlantic region and the lower Mississippi valley was not merely a consequence of the inclusion of the development dataset within the verification dataset, the statistics were recalculated excluding reports and nulls from our collaborating CWAs. As shown by the dots in Fig. 8, the relative performances of SHERBE and SHERBS3 are at least as good when the development dataset is removed. Finally, Table 6 reveals the practical (as opposed to statistical) significance of SHERBS3 and SHERBE within the verification dataset. For example, at their most skillful thresholds, approximately 300 more events were misclassified by STP than by SHERBS3.
Fig. 8.
Fig. 8.

Max TSS for given composite parameters for discriminating between HSLC significant severe reports and nulls across regions associated with at least 50 HSLC significant severe reports between 2006 and 2011. Dots indicate max TSSs calculated using datasets excluding reports and nulls from development dataset CWAs, ensuring truly independent datasets in LMV and SA.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Table 6.

Shown are the number of hits and correct nulls (second column) and misses and false alarms (third column), identified when utilizing the given composite parameters at their optimal thresholds when discriminating between HSLC significant severe reports and nulls in the nationwide verification dataset.

Table 6.

b. Interpretation of the new composite parameters

Traditionally, composite parameters designed to improve the forecasting of severe hazards—particularly tornadoes—have used CAPE as a constituent ingredient; examples include STP (Thompson et al. 2012), the supercell composite parameter [SCP; effective-layer version from Thompson et al. (2004)], the vorticity generation parameter10 (VGP; Rasmussen and Blanchard 1998), the energy helicity index10 (EHI; Rasmussen and Blanchard 1998), and the Craven–Brooks significant severe parameter (Craven and Brooks 2004). As a result of this reliance on CAPE (and because these parameters were developed using datasets consisting of a wider range of environmental conditions—e.g., the inclusion of a large number of springtime plains tornado events—leading to relatively high normalization values for CAPE components), operational meteorologists have noted a general underestimation of risk in HSLC environments by composite parameters (R. Thompson 2012, personal communication).

Within our verification dataset, the STP, VGP, EHI, SCP, and Craven–Brooks significant severe parameter were all fairly skillful, with a greater than 20% improvement over chance forecasts11 (i.e., a TSS above 0.2; see Table 4). However, the use of lapse rates rather than integrated CAPE is statistically more robust for HSLC environments, likely because lapse rates are not as susceptible to errors in low-level moisture or intralayer details that could cause CAPE (particularly SB or ML) to approach or become zero in environments with marginal instability. Based on the analysis of Coniglio (2012), it is apparent that modeled CAPE fields can be inaccurate due to errors in boundary layer height, low-level moisture content, or lapse rates. These compounding factors often lead to model errors in CAPE that are substantial relative to our 500 J kg−1 threshold, as shown in Coniglio’s (2012) Fig. 10. Although Coniglio (2012) did not specifically address lapse rates, it is worth noting that, in his Fig. 6, the SFCOA lapse rates utilized in the SHERB parameters are generally consistent with observations even where the details of the humidity and temperature profiles are contributing to errors in integrated CAPE.

Importantly, the statistically identified components of SHERBS3 and SHERBE do appear to be physically relevant for use in HSLC environments. McCaul and Weisman (2001) showed that in modeling simulations of marginally unstable environments, steeper LLLRs corresponded to stronger convection when compared to modest LLLRs. Additionally, in the eastern United States, where elevated mixed layers are rare compared to the plains, steep LR75s have shown high correlation with significant severe weather outbreaks (Banacos and Ekster 2010). Other studies have noted that steep lapse rates corresponded to a greater likelihood of tornadogenesis (Godfrey et al. 2004; Parker 2012) and damaging winds (Johns and Hirt 1987), both through intensification of convective updrafts and the facilitation of downward momentum transfer. The possibly different roles of enhanced LLLRs and LR75s may contribute to their improved skill over the surface to 500-hPa lapse rate, when combined with shear and wind parameters. For example, steep LLLRs are correlated to low levels of maximum buoyancy, encouraging more intense low-level updrafts (McCaul and Weisman 2001). It is also likely that our approach leads to “compounding” of the skill of constituent components (i.e., each parameter’s unique combination of hits, false alarms, misses, and correct nulls means that a greater number of components increases the likelihood of properly classifying each case).

Low-level shear magnitudes such as S3MG have been identified as discriminators between severe and nonsevere convection, particularly in cases of QLCS mesovortices and tornadoes (Weisman and Trapp 2003; Godfrey et al. 2004; Lane and Moore 2006), which are common in HSLC environments (Davis and Parker 2012). Meanwhile, the original development of ESHR by Thompson et al. (2007) was partially inspired by the knowledge that, for shallow storms (such as those in HSLC cases), the calculation depth for the most representative vertical shear parameters ought to scale with the depth of the layer possessing CAPE.

c. Performance of SHERB parameters in non-HSLC environments

When a new convective parameter is considered for adoption, it is important to assess its performance across a wider range of environments. In that spirit, the TSS calculations were repeated using the population of all significant severe reports and nulls across the entire contiguous United States from 2006 to 2011 (i.e., regardless of their associated CAPE and shear magnitudes). Despite being developed with an HSLC dataset, SHERBS3 and especially SHERBE showed impressive skill across all environments (Fig. 9). SHERBE had the highest calculated TSS within this dataset, both at any threshold (0.443 at a threshold of 0.95) and at its designed threshold (0.438). A comparison of Figs. 6a, 7a, and 9a displays an additional benefit to utilizing SHERBE to identify environments favorable for significant severe convection: at its designed threshold of 1, the skill is very high, regardless of the CAPE’s magnitude. This is likewise true when comparing significant tornadoes and nulls (Figs. 1012), though other composite parameters are more skillful in the verification dataset (albeit at thresholds below their designed optimal values). Although SHERBE begins to lose skill in vanishing CAPE (due to its inherent dependence on CAPE associated with the definition of effective shear), its remarkable consistency and skill across all environments is noteworthy. SHERBS3 appears to have its skill maximized in HSLC environments (cf. Figs. 6, 7, and 9), suggesting that it would be preferable over SHERBE in cases with forecast CAPE near zero. When the non-HSLC events are included, the other conventional composite parameters see improved skill scores (and at optimal values much more consistent with those conventionally cited), although their skill levels in discriminating between all significant severe reports and nulls fall below that of SHERBE.

Fig. 9.
Fig. 9.

As in Fig. 6, but for the nationwide verification dataset across all environments.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Fig. 10.
Fig. 10.

As in Fig. 6, but for significant tornadoes against nulls.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Fig. 11.
Fig. 11.

As in Fig. 6, but for significant tornadoes against nulls within the verification dataset.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Fig. 12.
Fig. 12.

As in Fig. 6, but for significant tornadoes against nulls across all environments in the verification dataset.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

d. Assessment of other concepts from previous literature

Some previous studies have indicated a relationship between midlevel dry-air intrusions and HSLC severe weather, indicating the possible importance of potential instability (Lane and Moore 2006; Clark 2009). The majority of the reports in the developmental dataset occurred in environments with a modest amount of potential instability. However, only approximately 6% had 850–700-hPa equivalent potential temperature (Θe) differences (ΔΘe) greater than −10 K (i.e., potential instability is modest in the majority of the HSLC cases) when calculated using RUC temperature and height analyses. There is a statistically significant difference in the means of ΔΘe between the events and nulls at the 0.95 level, suggesting that potential instability could play a role in discriminating between events and nulls; however, ΔΘe is not among the most skillful individual parameters in our development dataset. Similar results were found using the 1000–700- and 1000–500-hPa layers. At any rate, we presume that the release of potential instability would already be at least partially represented by the model analysis that underpins the SFCOA dataset we used.

Other studies (e.g., Johns 1993) have found that cool season damaging straight-line wind events may be primarily driven by the strength of the upper-level flow, the momentum of which could potentially be transferred to the surface in convective downdrafts. In our original skill tests using the development dataset, the 6-km wind magnitude was the third conditionally most skillful parameter and, when combined with LLLR and LR75, the product showed considerable skill at discriminating between HSLC significant severe reports (the majority of which were winds) and nulls. This is consistent with Evans and Doswell (2001) and Kuchera and Parker (2006), who found that ground-relative wind velocities were better discriminators than shear magnitudes between convection-producing severe winds and nonsevere convection. No doubt, steep lapse rates (i.e., LLLR and LR75) are also crucial in facilitating the transfer of that momentum to the surface in HSLC environments. While the wind from 6 km is likely not truly being transferred to the ground through downdrafts, this finding does indicate that in highly dynamic environments, the speed of the convective system in addition to the downward momentum transfer may contribute to a considerable fraction of the severe to significantly severe wind events. However, when all reports (including tornadoes and significant hail) are assessed, shear-based SHERBS3 and SHERBE tend to perform best, particularly for separating tornadoes from nulls (Tables 4 and 5).

5. Discussion

The techniques described in this study, including the composite parameters introduced in section 4, were developed using a dataset that included only reports and nulls from collaborating CWAs in the Southeast and mid-Atlantic. Thus, our results are naturally most promising in those regions (Fig. 6). However, the skill levels of SHERBS3 and SHERBE are also high in the nationwide verification dataset (Figs. 7 and 13), suggesting applicability for HSLC environments across the entire country. Despite some regional variance in skill (Fig. 8), the optimal SHERBS3 and SHERBE thresholds are generally consistent nationwide (unlike the conventional composite parameters we tested), suggesting that forecasters can confidently utilize SHERBS3 and SHERBE at their designated threshold of 1. The other composite parameters tested in this study had optimal thresholds well below their conventional values and a relatively high standard deviation in their optimal thresholds (not shown) across different regimes, regions, and seasons when compared with SHERBS3 and SHERBE.

Fig. 13.
Fig. 13.

Box-and-whiskers diagrams for distributions of (a) SHERBS3, (b) SHERBE, (c) STP (effective version), and (d) EHI for (left) HSLC significant severe reports, (middle) HSLC significant tornado reports, and (right) HSLC nulls. Box-and-whiskers diagrams are defined in Fig. 2.

Citation: Weather and Forecasting 29, 4; 10.1175/WAF-D-13-00041.1

Despite the demonstrated skill of SHERBS3 and SHERBE, these parameters were designed as tools to diagnose the likelihood of significant severe HSLC convection, not to explicitly forecast the initial development of convection. All of the data utilized to develop the parameters were associated with confirmed ongoing convection, either via severe weather reports or false alarm warnings. Therefore, these parameters should only be used to forecast the potential severity of convection that has developed, not to forecast whether HSLC convection will be initiated. Forecasters would need to combine the information from SHERBS3 or SHERBE with forecasts of convective development that use other input. We realize that many forecasters will not want to inspect the product unless convection is a viable possibility.

Plan-view plots of modeled and analyzed SHERB data reveal that values above 1 are often found in locations where convection will not occur (e.g., in mountainous terrain, the high plains, cold-air-advection regimes behind synoptic fronts, environments characterized by large CIN). Ongoing research will attempt to identify ways to reduce these false alarm areas, particularly in SHERBS3. For example, although the MU lifted index (MULI) is not formally considered by the SHERB parameters because it does not improve the skill at discriminating between hits and nulls, masking areas where MULI > +6 (alternatively +8 or +10) would explicitly exclude environments where convection is extremely unlikely while retaining over 94% (or 97% or 99%, respectively) of the significant severe reports in the verification dataset.

SHERBE inherently has a mask based on the effective shear’s CAPE reliance, but it is important to reiterate that a nonnegligible number of significant severe reports in our dataset occurred with MUCAPE values of 0 J kg−1 in the SFCOA. Thus, while the inherent masking of SHERBE reduces its false alarm area, SHERBE may miss the threat if the forecast or analyzed CAPE is erroneously 0 J kg−1. Additional masking could also be utilized by forecasters to restrict SHERB analyses and model fields to the environments for which it was specifically developed (our HSLC criteria of SBCAPE ≤ 500 J kg−1, MUCAPE ≤ 1000 J kg−1, and 0–6-km shear ≥ 18 m s−1); on the other hand, section 4c provided evidence that the SHERB parameters (in particular, SHERBE) still exhibit skill in non-HSLC environments.

As noted by Kuchera and Parker (2006) and others, applying any one forecasting technique to all types of severe weather is challenging due to the inherent differences in tornado, wind, and hail environments. Although designed to discriminate between all significant severe reports and nulls within HSLC environments, the SHERB parameters should not be utilized in isolation. As shown by Figs. 6, 7, and 13, composite parameters such as STP and EHI do exhibit impressive skill even in HSLC environments, particularly when separating between significant tornadoes and nonsevere convection (i.e., uses that are more consistent with their original design;12 see Figs. 1012). Thus, these parameters should not be discounted when identifying environments favorable for HSLC severe weather. However, an important caveat underlined by this study is that the most discriminatory thresholds for these conventional parameters are much lower in HSLC cases when comparing significant severe reports (or tornadoes) against nulls. Therefore, it may be worthwhile for forecasters to explicitly contour them at lower levels for HSLC applications.13

Notably, the developmental techniques used in this study do not take into account synoptic-scale features, which are not negligible when determining the risk for significant severe weather, particularly straight-line winds (e.g., Johns 1993; Evans and Doswell 2001). However, this is a limitation shared by all of the composite parameters that are commonly used as convective forecast guidance tools. Based on inspection of many cases, the combination of shear and lapse rates represented by SHERBS3/SHERBE appears to have utility both in situations dominated by convective-scale influences and in situations where synoptic-scale forcing is especially intense. As a step in this direction, in a future manuscript we plan to present maps of SHERBS3 and SHERBE for a number of HSLC event and null cases, showing their spatial and temporal relationships to the larger-scale setting.

6. Conclusions

High-shear, low-CAPE (HSLC) events are a challenge for operational forecasters throughout the United States. Through evaluation of significant severe weather reports between 2006 and 2011, it was determined that HSLC reports—defined by SBCAPE ≤ 500 J kg−1, MUCAPE ≤ 1000 J kg−1, and 0–6-km shear ≥ 18 m s−1—occur in a substantial fraction of the country, at all times of the day and in all seasons. During the cool season, tornado and wind events are favored across the Southeast and Mississippi valley. This threat transitions to primarily a hail and wind threat across the plains and Midwest through the late spring and into the summer. Examination of thermodynamic variables indicates that multiple regimes of HSLC convection are possible, including cases with high low-level moisture content and low LCLs and cases with deep, dry, well-mixed boundary layers.

In our development dataset from the Southeast and mid-Atlantic regions, the 0–3-km lapse rate was the most skillful individual parameter at discriminating between HSLC significant severe convection and nonsevere convection. Combining the 0–3-km lapse rate with the 700–500-hPa lapse rate and multiple wind or shear parameters provided additional skill, raising the true skill statistic to over 0.5 in some formulations. Further analysis revealed that combining the lapse rates with the effective shear magnitude resulted in the most skillful parameter for discriminating between HSLC significant severe reports and nulls nationwide. Because the effective shear trends to 0 for very low values of CAPE, we also investigated the utility of fixed-layer (fixed level) shear (wind) magnitudes. Ultimately, a normalized product of the 0–3-km shear magnitude, 0–3-km lapse rate, and 700–500-hPa lapse rate was found to be the most skillful at discriminating between HSLC significant severe convection and nonsevere convection when using a fixed-layer shear. This product was called the severe hazards in environments with reduced buoyancy parameter (using the 0–3-km shear magnitude or SHERBS3), while the version including the effective shear magnitude was referred to as SHERBE. Both parameters displayed an increase in skill over existing composite parameters within the development dataset.

In the nationwide verification dataset, SHERBS3 and SHERBE (as well as most traditional composite parameters) show skill, though the skill varies somewhat by region. Given that SHERBS3 and SHERBE were created from reports and nulls exclusively from the Southeast and mid-Atlantic, it is logical that they would have the most skill in these regions. However, SHERBS3 and SHERBE are also the most skillful composite parameters at discriminating between HSLC significant severe convection and nonsevere convection across the entire United States. Parameters such as the significant tornado parameter (STP) and energy helicity index (EHI) show skill in discriminating between significant tornadoes (only) and nulls within HSLC environments, but the values at which they are skillful are well below their conventional thresholds (e.g., some of the observed HSLC optimal values are not even contoured on the operational SPC Mesoanalysis, and values this low would likely be common year round). Should forecasters wish to utilize existing composite parameters such as STP or EHI for guidance in identifying environments favorable for HSLC tornadoes, values below commonly accepted thresholds must be used to maximize their skill.

SHERBS3, and especially SHERBE, actually exhibit skill across all environments, with SHERBE having the highest true skill statistic at discriminating between all nationwide significant severe reports and nulls, regardless of the diagnosed environmental CAPE and shear. The skill of SHERBE, coupled with its relatively consistent optimal threshold across all environments, reveals a robustness to the product of lapse rates and shear endorsed here. Based on the statistics presented herein, forecasters can confidently utilize SHERBE (across all environments) and SHERBS3 (particularly in HSLC environments) when attempting to discriminate between environments capable of producing significant severe convection and those associated with nonsevere convection.

The present approach primarily emphasizes local ingredients. Future work is necessary to investigate the influence of synoptic-scale features on HSLC convection (e.g., the relationship between upper-level winds and synoptic boundary orientation). Additionally, it would be helpful to know how well forecast models (and their derivatives, such as SPC’s SFCOA) handle HSLC environments. In marginal CAPE, it is possible that processes such as the release of potential instability become much more important than in high CAPE environments. Finally, the approach in this article is a blunt statistical sifting of available parameters. In our ongoing research, we are trying to understand how these ingredients influence HSLC convection through a series of idealized convective-scale simulations.

Acknowledgments

The authors would like to acknowledge collaborators at NWS WFOs listed within the manuscript for their feedback throughout the study, particularly Justin Lane and Pat Moore from WFO Greenville–Spartanburg, South Carolina. Additionally, the authors would like to acknowledge Andy Dean from SPC for helping with frequent data requests and Rich Thompson for participation in composite parameter discussions. Further, the authors acknowledge the Iowa Environmental Mesonet for access to their NWS product archive for null identification. The authors thank Gary Lackmann and Sandra Yuter for their guidance as committee members for KS’s M.S. research, from which this article is derived. Finally, the authors acknowledge the Convective Storms Group at North Carolina State University for their support and feedback during this research. Funding for this research was provided by National Oceanic and Atmospheric Administration (NOAA) Grant NA10NWS4680007 as part of the Collaborative Science, Technology, and Applied Research Program. Additional funding for KS was provided by an American Meteorological Society Graduate Fellowship sponsored by NOAA’s NWS.

APPENDIX

SFCOA Parameters

Our search for the most skillful environmental ingredients involved the testing of every single parameter in the SFCOA dataset from SPC. Although the main article focuses only on those parameters that are most discriminatory, for completeness we present the full list of tested variables in Table A1. Any parameters that are listed in the table but not discussed in the text were found to be less skillful and/or statistically insignificant in the development dataset (see section 2a).

Table A1.

List of environmental parameters within the SFCOA database subject to TSS tests.

Table A1.

REFERENCES

  • Banacos, P. C., , and Ekster M. L. , 2010: The association of the elevated mixed layer with significant severe weather events in the northeastern United States. Wea. Forecasting, 25, 10821102, doi:10.1175/2010WAF2222363.1.

    • Search Google Scholar
    • Export Citation
  • Barker, L. J., III, 2006: A potentially valuable WSR-88D severe storm pre-cursor signature in highly dynamic, low CAPE, high shear environments. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 3.6. [Available online at https://ams.confex.com/ams/pdfpapers/115129.pdf.]

  • Bothwell, P. D., , Hart J. A. , , and Thompson R. L. , 2002: An integrated three dimensional objective analysis scheme in use at the Storm Prediction Center. Preprints, 21st Conf. on Severe Local Storms/19th Conf. on Weather Analysis and Forecasting/15th Conf. on Numerical Weather Prediction, San Antonio, TX, Amer. Meteor. Soc., JP3.1. [Available online at https://ams.confex.com/ams/pdfpapers/47482.pdf.]

  • Brooks, H. E., , Doswell C. A. III, , and Kay M. P. , 2003: Climatological estimates of local daily tornado probability for the United States. Wea. Forecasting, 18, 626640, doi:10.1175/1520-0434(2003)018<0626:CEOLDT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Brotzge, J., , Erickson S. , , and Brooks H. , 2011: A 5-yr climatology of tornado false alarms. Wea. Forecasting, 26, 534544, doi:10.1175/WAF-D-10-05004.1.

    • Search Google Scholar
    • Export Citation
  • Burke, P. C., , and Schultz D. M. , 2004: A 4-yr climatology of cold-season bow echoes over the continental United States. Wea. Forecasting, 19, 10611074, doi:10.1175/811.1.

    • Search Google Scholar
    • Export Citation
  • Clark, M. R., 2009: The southern England tornadoes of 30 December 2006: Case study of a tornadic storm in a low CAPE, high shear environment. Atmos. Res., 93, 5065, doi:10.1016/j.atmosres.2008.10.008.

    • Search Google Scholar
    • Export Citation
  • Clark, M. R., 2011: Doppler radar observations of mesovortices within a cool-season tornadic squall line over the UK. Atmos. Res., 100, 749764, doi:10.1016/j.atmosres.2010.09.007.

    • Search Google Scholar
    • Export Citation
  • Coniglio, M. C., 2012: Verification of RUC 0–1-h forecasts and SPC mesoscale analyses using VORTEX2 soundings. Wea. Forecasting, 27, 667683, doi:10.1175/WAF-D-11-00096.1.

    • Search Google Scholar
    • Export Citation
  • Cope, A. M., 2004: An early morning mid-Atlantic severe weather episode: Short-lived tornadoes in a high-shear low-instability environment. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P1.4. [Available online at https://ams.confex.com/ams/11aram22sls/webprogram/Paper81834.html.]

  • Craven, J. P., , and Brooks H. E. , 2004: Baseline climatology of sounding derived parameters associated with deep, moist convection. Natl. Wea. Dig., 28, 1324.

    • Search Google Scholar
    • Export Citation
  • Davies, J. M., 1990: Midget supercell spawns tornadoes. Weatherwise, 43, 260261, doi:10.1080/00431672.1990.9929350.

  • Davis, J. M., , and Parker M. D. , 2012: Radar climatology of tornadoes occurring in high shear/low CAPE environments in the mid-Atlantic and Southeast. Preprints, 26th Conf. on Severe Local Storms, Nashville, TN, Amer. Meteor. Soc., 9.2. [Available online at https://ams.confex.com/ams/26SLS/webprogram/Paper211761.html.]

  • Dean, A. R., , and Schneider R. S. , 2008: Forecast challenges at the NWS Storm Prediction Center relating to the frequency of favorable severe storm environments. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 9A.2. [Available online at https://ams.confex.com/ams/pdfpapers/141743.pdf.]

  • Dean, A. R., , and Schneider R. S. , 2012: An examination of tornado environments, events, and impacts from 2003–2012. Preprints, 26th Conf. on Severe Local Storms, Nashville, TN, Amer. Meteor. Soc., 60. [Available online at https://ams.confex.com/ams/26SLS/webprogram/Paper211580.html.]

  • Doswell, C. A., III, , and Burgess D. W. , 1988: On some issues of United States tornado climatology. Mon. Wea. Rev., 116, 495501, doi:10.1175/1520-0493(1988)116<0495:OSIOUS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, , Davies-Jones R. , , and Keller D. L. , 1990: On summary measures of skill in rare event forecasting based on contingency tables. Wea. Forecasting, 5, 576585, doi:10.1175/1520-0434(1990)005<0576:OSMOSI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, , Brooks H. E. , , and Kay M. P. , 2005: Climatological estimates of daily local nontornadic severe thunderstorm probability for the United States. Wea. Forecasting, 20, 577595, doi:10.1175/WAF866.1.

    • Search Google Scholar
    • Export Citation
  • Edwards, R., , Dean A. R. , , Thompson R. L. , , and Smith B. T. , 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part III: Tropical cyclone tornadoes. Wea. Forecasting, 27, 15071519, doi:10.1175/WAF-D-11-00117.1.

    • Search Google Scholar
    • Export Citation
  • Evans, J. S., , and Doswell C. A. III, 2001: Examination of derecho environments using proximity soundings. Wea. Forecasting, 16, 329342, doi:10.1175/1520-0434(2001)016<0329:EODEUP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Evans, M., 2010: An examination of low CAPE/high shear severe convective events in the Binghamton, New York county warning area. Natl. Wea. Dig., 34, 129144.

    • Search Google Scholar
    • Export Citation
  • Godfrey, E. S., , Trapp R. J. , , and Brooks H. E. , 2004: A study of the pre-storm environment of tornadic quasi-linear convective systems. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc. 3A.5. [Available online at https://ams.confex.com/ams/pdfpapers/81388.pdf.]

  • Grumm, R. H., , and Glazewski M. , 2004: Thunderstorm types associated with the “broken-S” radar signature. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P7.1. [Available online at https://ams.confex.com/ams/pdfpapers/81537.pdf.]

  • Guyer, J. L., , and Dean A. R. , 2010: Tornadoes within weak CAPE environments across the continental United States. Preprints, 25th Conf. on Severe Local Storms, Denver, CO, Amer. Meteor. Soc., 1.5. [Available online at https://ams.confex.com/ams/pdfpapers/175725.pdf.]

  • Guyer, J. L., , Imy D. A. , , Kis A. , , and Venable K. , 2006: Cool season significant (F2-F5) tornadoes in the Gulf Coast states. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 4.2. [Available online at https://ams.confex.com/ams/pdfpapers/115320.pdf.]

  • Johns, R. H., 1993: Meteorological conditions associated with bow echo development in convective storms. Wea. Forecasting, 8, 294299, doi:10.1175/1520-0434(1993)008<0294:MCAWBE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Johns, R. H., , and Hirt W. D. , 1987: Derechos: Widespread convectively induced windstorms. Wea. Forecasting, 2, 3249, doi:10.1175/1520-0434(1987)002<0032:DWCIW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kis, A. K., , and Straka J. M. , 2010: Nocturnal tornado climatology. Wea. Forecasting, 25, 545561, doi:10.1175/2009WAF2222294.1.

  • Kuchera, E. L., , and Parker M. D. , 2006: Severe convective wind environments. Wea. Forecasting, 21, 595612, doi:10.1175/WAF931.1.

  • Lane, J. D., , and Moore P. D. , 2006: Observations of a non-supercell tornadic thunderstorm from terminal Doppler weather radar. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., P4.5. [Available online at https://ams.confex.com/ams/pdfpapers/115102.pdf.]

  • Markowski, P. M., , and Straka J. M. , 2000: Some observations of rotating updrafts in a low-buoyancy, highly sheared environment. Mon. Wea. Rev., 128, 449461, doi:10.1175/1520-0493(2000)128<0449:SOORUI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Marzban, C., 2004: The ROC curve and the area under it as performance measures. Wea. Forecasting, 19, 11061114, doi:10.1175/825.1.

  • Mason, I. B., 1982: A model for assessment of weather forecasts. Aust. Meteor. Mag., 30, 291303.

  • McAvoy, B. P., , Jones W. A. , , and Moore P. D. , 2000: Investigation of an unusual storm structure associated with weak to occasionally strong tornadoes over the eastern United States. Preprints, 20th Conf. on Severe Local Storms, Orlando, FL, Amer. Meteor. Soc., 182185.

  • McCaul, E. W., , and Weisman M. L. , 2001: The sensitivity of simulated supercell structure and intensity to variations in the shapes of environmental buoyancy and shear profiles. Mon. Wea. Rev., 129, 664687, doi:10.1175/1520-0493(2001)129<0664:TSOSSS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Parker, M. D., 2012: Impacts of lapse rates on low-level rotation in idealized storms. J. Atmos. Sci., 69, 538559, doi:10.1175/JAS-D-11-058.1.

    • Search Google Scholar
    • Export Citation
  • Rasmussen, E. N., , and Blanchard D. O. , 1998: A baseline climatology of sounding-derived supercell and tornado forecast parameters. Wea. Forecasting, 13, 11481164, doi:10.1175/1520-0434(1998)013<1148:ABCOSD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Schneider, R. S., , and Dean A. R. , 2008: A comprehensive 5-year severe storm environment climatology for the continental United States. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 16A.4. [Available online at https://ams.confex.com/ams/pdfpapers/141748.pdf.]

  • Schneider, R. S., , Dean A. R. , , Weiss S. J. , , and Bothwell P. D. , 2006: Analysis of estimated environments for 2004 and 2005 severe convective storm reports. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 3.5. [Available online at https://ams.confex.com/ams/pdfpapers/115246.pdf.]

  • Smith, B. T., , Guyer J. L. , , and Dean A. R. , 2008: The climatology, convective mode, and mesoscale environment of cool season severe thunderstorms in the Ohio and Tennessee Valleys, 1995–2006. Preprints, 24th Conf. on Severe Local Storms, Savannah, GA, Amer. Meteor. Soc., 13B.7. [Available online at https://ams.confex.com/ams/pdfpapers/141968.pdf.]

  • Smith, B. T., , Thompson R. L. , , Grams J. S. , , Broyles C. , , and Brooks H. E. , 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part I: Storm classification and climatology. Wea. Forecasting, 27, 11141135, doi:10.1175/WAF-D-11-00115.1.

    • Search Google Scholar
    • Export Citation
  • Smith, B. T., , Castellanos T. E. , , Winters A. C. , , Mead C. M. , , Dean A. R. , , and Thompson R. L. , 2013: Measured severe convective wind climatology and associated convective modes of thunderstorms in the contiguous United States, 2003–09. Wea. Forecasting, 28, 229236, doi:10.1175/WAF-D-12-00096.1.

    • Search Google Scholar
    • Export Citation
  • Thompson, R. L., , Edwards R. , , Hart J. A. , , Elmore K. L. , , and Markowski P. , 2003: Close proximity soundings within supercell environments obtained from the Rapid Update Cycle. Wea. Forecasting, 18, 12431261, doi:10.1175/1520-0434(2003)018<1243:CPSWSE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Thompson, R. L., , Edwards R. , , and Mead C. M. , 2004: An update to the supercell composite and significant tornado parameters. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P8.1. [Available online at https://ams.confex.com/ams/pdfpapers/82100.pdf.]

  • Thompson, R. L., , Mead C. M. , , and Edwards R. , 2007: Effective storm-relative helicity and bulk shear in supercell thunderstorm environments. Wea. Forecasting, 22, 102115, doi:10.1175/WAF969.1.

    • Search Google Scholar
    • Export Citation
  • Thompson, R. L., , Smith B. T. , , Grams J. S. , , Dean A. R. , , and Broyles C. , 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part II: Supercell and QLCS tornado environments. Wea. Forecasting, 27, 11361154, doi:10.1175/WAF-D-11-00116.1.

    • Search Google Scholar
    • Export Citation
  • Trapp, R. J., , Wheatley D. M. , , Atkins N. T. , , Przybylinski R. W. , , and Wolf R. , 2006: Buyer beware: Some words of caution on the use of severe wind reports in postevent assessment and research. Wea. Forecasting, 21, 408415, doi:10.1175/WAF925.1.

    • Search Google Scholar
    • Export Citation
  • van den Broeke, M. S., , Schultz D. M. , , Johns R. H. , , Evans J. S. , , and Hales J. E. , 2005: Cloud-to-ground lightning production in strongly forced, low-instability convective lines associated with damaging wind. Wea. Forecasting, 20, 517530, doi:10.1175/WAF876.1.

    • Search Google Scholar
    • Export Citation
  • Verbout, S. M., , Brooks H. E. , , Leslie L. M. , , and Schultz D. M. , 2006: Evolution of the U.S. tornado database: 1954–2003. Wea. Forecasting, 21, 8693, doi:10.1175/WAF910.1.

    • Search Google Scholar
    • Export Citation
  • Weisman, M. L., , and Trapp R. J. , 2003: Low-level mesovortices within squall lines and bow echoes. Part I: Overview and dependence on environmental shear. Mon. Wea. Rev., 131, 27792803, doi:10.1175/1520-0493(2003)131<2779:LMWSLA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Weiss, S. J., , Hart J. A. , , and Janish P. R. , 2002: An examination of severe thunderstorm wind report climatology. Preprints, 21st Conf. on Severe Local Storms, San Antonio, TX, Amer. Meteor. Soc., 11B.2. [Available online at https://ams.confex.com/ams/pdfpapers/47494.pdf.]

  • Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences: An Introduction.Academic Press, 467 pp.

1

Throughout the article, the term shear will be used in reference to the bulk shear vector magnitude over a layer.

2

In this study, high shear is defined as 0–6-km shear ≥ 18 m s−1, and low CAPE is defined as surface-based CAPE ≤ 500 J kg−1. The surface-based parcel criterion was chosen after correspondence with National Weather Service collaborators. An additional criterion of most unstable parcel CAPE ≤ 1000 J kg−1 was introduced during the verification process to eliminate elevated convection cases. Further rationale for these thresholds is presented in section 4.

3

Collaborating WFOs include Wakefield, VA (AKQ); Columbia, SC (CAE); Charleston, SC (CHS); Peachtree City, GA (FFC); Greenville–Spartanburg, SC (GSP); Huntsville, AL (HUN); Wilmington, NC (ILM); Sterling, VA (LWX); Newport–Morehead City, NC (MHX); Raleigh, NC (RAH); and Blacksburg, VA (RNK).

4

This is known operationally as the SPC mesoanalysis.

5

Unless otherwise noted, STP refers to the effective-layer version as defined in Thompson et al. (2012), which has been available in SFCOA since 2005 (A. Dean 2013, personal communication).

6

Here, only wind events were utilized to prevent the confounding factor of comparing tornado, wind, and hail environments.

7

Subsequent tests revealed that the surface to 500-hPa lapse rate showed a marginal improvement in skill individually over the product of LLLR and LR75; however, when combined with wind and shear components, its composite skill was considerably lower than that of the LLLR, LR75, and wind/shear product within the development dataset. The possible reasons for this are discussed later in section 4b.

8

All of the parameters listed in Table 1 were subjected to the same tests as the parameters referenced hereafter. For the sake of clarity and brevity, only the most skillful parameters were included in subsequent figures, tables, and discussion.

9

See Thompson et al. (2007) for description.

10

Throughout the article, VGP and EHI are calculated utilizing storm-relative helicity in the 0–3-km layer.

11

The authors acknowledge that, of these parameters, only the Craven–Brooks significant severe was designed to identify environments favorable for all severe weather, rather than just tornadoes or, more generally, supercells. However, this comparison is purely intended to show the utility of SHERBS3 and SHERBE with regard to existing composite parameters, not to imply that parameters such as STP should be judged on their skill at discriminating between all significant severe reports and nulls.

12

The authors acknowledge that a number of the compared parameters were not originally designed to be used for all convective modes or all types of significant severe weather. Because STP is meant to discriminate between tornadic and nontornadic supercells, the inclusion of nonsupercells in this study could have contributed to the seemingly low optimal value calculated for STP. Nevertheless, we have observed that forecasters still inspect many of these other parameters during HSLC events, so we include them for completeness.

13

In particular, STP-E would benefit from being plotted at lower values due to its use of MLCAPE, which tends to be lower than SBCAPE.

Save