• Ashley, S. T., , and Ashley W. S. , 2008: Flood fatalities in the United States. J. Appl. Meteor. Climatol., 47, 805818, doi:10.1175/2007JAMC1611.1.

    • Search Google Scholar
    • Export Citation
  • Ashley, W. S., 2007: Spatial and temporal analysis of tornado fatalities in the United States: 1880–2005. Wea. Forecasting, 22, 12141228, doi:10.1175/2007WAF2007004.1.

    • Search Google Scholar
    • Export Citation
  • Bluestein, H. B., , and Jain M. H. , 1985: Formation of mesoscale lines of precipitation: Severe squall lines in Oklahoma during the spring. J. Atmos. Sci., 42, 17111732, doi:10.1175/1520-0469(1985)042<1711:FOMLOP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bouwer, L. M., 2011: Have disaster losses increased due to anthropogenic climate change? Bull. Amer. Meteor. Soc., 92, 3946, doi:10.1175/2010BAMS3092.1.

    • Search Google Scholar
    • Export Citation
  • Brooks, H. E., , and Doswell C. A. III, 2002: Deaths in the 3 May 1999 Oklahoma City tornado from a historical perspective. Wea. Forecasting, 17, 354361, doi:10.1175/1520-0434(2002)017<0354:DITMOC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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
  • Changnon, S. A., , Pielke R. A. Jr., , Changnon D. , , Sylves R. T. , , and Pulwarty R. , 2000: Human factors explain the increased losses from weather and climate extremes. Bull. Amer. Meteor. Soc., 81, 437442, doi:10.1175/1520-0477(2000)081<0437:HFETIL>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Clark, R. A., , Gourley J. J. , , Flamig Z. L. , , Hong Y. , , and Clark E. , 2014: CONUS-wide evaluation of National Weather Service flash flood guidance products. Wea. Forecasting, 29, 377392, doi:10.1175/WAF-D-12-00124.1.

    • Search Google Scholar
    • Export Citation
  • Coniglio, M. C., , and Corfidi S. , 2006: Forecasting the speed and longevity of severe mesoscale convective systems. Preprints, Symp. on the Challenges of Severe Convective Storms, Atlanta, GA, Amer. Meteor. Soc., P1.30. [Available online at https://ams.confex.com/ams/pdfpapers/104815.pdf.]

  • Coniglio, M. C., , Hwang J. Y. , , and Stensrud D. J. , 2010: Environmental factors in the upscale growth and longevity of MCSs derived from Rapid Update Cycle analyses. Mon. Wea. Rev., 138, 35143539, doi:10.1175/2010MWR3233.1.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, 2001: Severe convective storms—An overview. Severe Convective Storms, Meteor. Monogr., No. 50, 1–26, doi:10.1175/0065-9401-28.50.1.

  • Doswell, C. A., III, , Brooks H. E. , , and Maddox R. A. , 1996: Flash flood forecasting: An ingredients-based methodology. Wea. Forecasting, 11, 560581, doi:10.1175/1520-0434(1996)011<0560:FFFAIB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ferree, J. T., , Looney J. , , and Waters K. , 2006: NOAA/National Weather Services’ storm-based warnings. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 11.6. [Available online at https://ams.confex.com/ams/pdfpapers/115513.pdf.]

  • Frey, B. J., , and Dueck D. , 2007: Clustering by passing messages between data points. Science, 315, 972976, doi:10.1126/science.1136800.

    • Search Google Scholar
    • Export Citation
  • Fritsch, J. M., , and Carbone R. , 2004: Improving quantitative precipitation forecasts in the warm season: A USWRP research and development strategy. Bull. Amer. Meteor. Soc., 85, 955965, doi:10.1175/BAMS-85-7-955.

    • Search Google Scholar
    • Export Citation
  • Gourley, J. J., and Coauthors, 2013: A unified flash flood database across the United States. Bull. Amer. Meteor. Soc., 94, 799805, doi:10.1175/BAMS-D-12-00198.1.

    • Search Google Scholar
    • Export Citation
  • Hamill, T. M., , Bates G. T. , , Whitaker J. S. , , Murray D. R. , , Fiorino M. , , Galarneau T. J. Jr., , Zhu Y. , , and Lapenta W. , 2013: NOAA’s second-generation global medium-range ensemble reforecast dataset. Bull. Amer. Meteor. Soc., 94, 15531565, doi:10.1175/BAMS-D-12-00014.1.

    • Search Google Scholar
    • Export Citation
  • Hart, R. E., , and Grumm R. H. , 2001: Using normalized climatological anomalies to rank synoptic-scale events objectively. Mon. Wea. Rev., 129, 24262442, doi:10.1175/1520-0493(2001)129<2426:UNCATR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hough, J. R., , and White M. A. , 2003: Environmental dynamism and strategic decision-making rationality: An examination at the decision level. Strategic Manage. J., 24, 481489, doi:10.1002/smj.303.

    • Search Google Scholar
    • Export Citation
  • Houze, R. A., Jr., , Biggerstaff M. , , Rutledge S. , , and Smull B. , 1989: Interpretation of Doppler weather radar displays of midlatitude mesoscale convective systems. Bull. Amer. Meteor. Soc., 70, 608619, doi:10.1175/1520-0477(1989)070<0608:IODWRD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Junker, N. W., , Schneider R. S. , , and Fauver S. L. , 1999: A study of heavy rainfall events during the great Midwest flood of 1993. Wea. Forecasting, 14, 701712, doi:10.1175/1520-0434(1999)014<0701:ASOHRE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Knupp, K. R., and Coauthors, 2014: Meteorological overview of the devastating 27 April 2011 tornado outbreak. Bull. Amer. Meteor. Soc., 95, 10411062, doi:10.1175/BAMS-D-11-00229.1.

    • Search Google Scholar
    • Export Citation
  • KWTV, 2015: Understanding storm shelters could save Oklahomans’ lives. News on 6. [Available online at http://www.newson6.com/story/29010877/understanding-storm-shelters-could-save-oklahomans-lives.]

  • Maddox, R. A., 1980: Mesoscale convective complexes. Bull. Amer. Meteor. Soc., 61, 13741387, doi:10.1175/1520-0477(1980)061<1374:MCC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Maddox, R. A., , Chappell C. F. , , and Hoxit L. R. , 1979: Synoptic and meso-α scale aspects of flash flood events. Bull. Amer. Meteor. Soc., 60, 115123, doi:10.1175/1520-0477-60.2.115.

    • Search Google Scholar
    • Export Citation
  • Markowski, P., , and Richardson Y. , 2010: Mesoscale Meteorology in Midlatitudes. J. Wiley and Sons, 430 pp.

  • Mercer, A. E., , Shafer C. M. , , Doswell C. A. III, , Leslie L. M. , , and Richman M. B. , 2009: Objective classification of tornadic and nontornadic severe weather outbreaks. Mon. Wea. Rev., 137, 43554368, doi:10.1175/2009MWR2897.1.

    • Search Google Scholar
    • Export Citation
  • Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343360, doi:10.1175/BAMS-87-3-343.

    • Search Google Scholar
    • Export Citation
  • Moore, J. T., , Glass F. H. , , Graves C. E. , , Rochette S. M. , , and Singer M. J. , 2003: The environment of warm-season elevated thunderstorms associated with heavy rainfall over the central United States. Wea. Forecasting, 18, 861878, doi:10.1175/1520-0434(2003)018<0861:TEOWET>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Morss, R. E., , Wilhelmi O. V. , , Meehl G. A. , , and Dilling L. , 2011: Improving societal outcomes of extreme weather in a changing climate: An integrated perspective. Annu. Rev. Environ. Resour., 36, 1, doi:10.1146/annurev-environ-060809-100145.

    • Search Google Scholar
    • Export Citation
  • Morss, R. E., , Demuth J. L. , , Bostrom A. , , Lazo J. K. , , and Lazrus H. , 2015: Flash flood risks and warning decisions: A mental models study of forecasters, public officials, and media broadcasters in Boulder, Colorado. Risk Anal., doi:10.1111/risa.12403, in press.

    • Search Google Scholar
    • Export Citation
  • NOAA, 2011: Value of a weather-ready nation. NOAA/NWS, 22 pp. [Available online at http://www.ppi.noaa.gov/wp-content/uploads/PPI-Weather-Econ-Stats-10-13-11.pdf.]

  • Nowotarski, C. J., , Markowski P. M. , , and Richardson Y. P. , 2011: The characteristics of numerically simulated supercell storms situated over statically stable boundary layers. Mon. Wea. Rev., 139, 31393162, doi:10.1175/MWR-D-10-05087.1.

    • Search Google Scholar
    • Export Citation
  • NWS, 2011a: Record floods of greater Nashville: Including flooding in middle Tennessee and western Kentucky, May 1–4, 2010. National Weather Service Service Assessment, 67 pp. + appendixes. [Available online at http://www.nws.noaa.gov/om/assessments/pdfs/Tenn_Flooding.pdf.]

  • NWS, 2011b: The historic tornadoes of April 2011. National Weather Service Service Assessment, 43 pp. + appendixes. [Available online at http://www.nws.noaa.gov/om/assessments/pdfs/historic_tornadoes.pdf.]

  • NWS, 2014: May 2013 Oklahoma tornadoes and flash flooding. National Weather Service Service Assessment, 42 pp. + appendixes. [Available online at http://www.nws.noaa.gov/om/assessments/pdfs/13oklahoma_tornadoes.pdf.]

  • Parker, M. D., , and Johnson R. H. , 2000: Organizational modes of midlatitude mesoscale convective systems. Mon. Wea. Rev., 128, 34133436, doi:10.1175/1520-0493(2001)129<3413:OMOMMC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Pedregosa, F., and Coauthors, 2011: Scikit-learn: Machine learning in python. J. Mach. Learn. Res., 12, 28252830.

  • Peters, J. M., , and Schumacher R. S. , 2014: Objective categorization of heavy-rain-producing MCS synoptic types by rotated principal component analysis. Mon. Wea. Rev., 142, 17161737, doi:10.1175/MWR-D-13-00295.1.

    • Search Google Scholar
    • Export Citation
  • Peters, J. M., , and Schumacher R. S. , 2015: Mechanisms for organization and echo training in a flash-flood-producing mesoscale convective system. Mon. Wea. Rev., 143, 10581095, doi:10.1175/MWR-D-14-00070.1.

    • Search Google Scholar
    • Export Citation
  • Robock, A., , Vinnikov K. Y. , , Srinivasan G. , , Entin J. K. , , Hollinger S. E. , , Speranskaya N. A. , , Liu S. , , and Namkhai A. , 2000: The Global Soil Moisture Data Bank. Bull. Amer. Meteor. Soc., 81, 12811299, doi:10.1175/1520-0477(2000)081<1281:TGSMDB>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rogash, J. A., , and Smith R. D. , 2000: Multiscale overview of a violent tornado outbreak with attendant flash flooding. Wea. Forecasting, 15, 416431, doi:10.1175/1520-0434(2000)015<0416:MOOAVT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rogash, J. A., , and Racy J. , 2002: Some meteorological characteristics of significant tornado events occurring in proximity to flash flooding. Wea. Forecasting, 17, 155159, doi:10.1175/1520-0434(2002)017<0155:SMCOST>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Schumacher, R. S., , and Johnson R. H. , 2005: Organization and environmental properties of extreme-rain-producing mesoscale convective systems. Mon. Wea. Rev., 133, 961976, doi:10.1175/MWR2899.1.

    • Search Google Scholar
    • Export Citation
  • Schwartz, B. E., , Chappell C. F. , , Togstad W. E. , , and Zhong X.-P. , 1990: The Minneapolis flash flood: Meteorological analysis and operational response. Wea. Forecasting, 5, 321, doi:10.1175/1520-0434(1990)005<0003:TMFFMA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, J. A., , Baeck M. L. , , Zhang Y. , , and Doswell C. A. III, 2001: Extreme rainfall and flooding from supercell thunderstorms. J. Hydrometeor., 2, 469489, doi:10.1175/1525-7541(2001)002<0469:ERAFFS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sutter, D., , and Erickson S. , 2010: The time cost of tornado warnings and the savings with storm-based warnings. Wea. Climate Soc., 2, 103112, doi:10.1175/2009WCAS1011.1.

    • 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
  • Trier, S., , Davis C. , , and Ahijevych D. , 2010: Environmental controls on the simulated diurnal cycle of warm-season precipitation in the continental United States. J. Atmos. Sci., 67, 10661090, doi:10.1175/2009JAS3247.1.

    • Search Google Scholar
    • Export Citation
  • Waters, K., and Coauthors, 2005: Polygon weather warnings: A new approach for the national weather service. Preprints, 21st Int. Conf. on Interactive Information Processing Systems (IIPS) for Meteorology, Oceanography, and Hydrology, Phoenix, AZ, Amer. Meteor. Soc., 14.1. [Available online at https://ams.confex.com/ams/pdfpapers/86326.pdf.]

  • Weisman, M. L., , and Klemp J. B. , 1984: The structure and classification of numerically simulated convective storms in directionally varying wind shears. Mon. Wea. Rev., 112, 24792498, doi:10.1175/1520-0493(1984)112<2479:TSACON>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wurman, J., , Robinson P. , , Alexander C. , , and Richardson Y. , 2007: Low-level winds in tornadoes and potential catastrophic tornado impacts in urban areas. Bull. Amer. Meteor. Soc., 88, 3146, doi:10.1175/BAMS-88-1-31.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , Bei N. , , Rotunno R. , , Snyder C. , , and Epifanio C. C. , 2007: Mesoscale predictability of moist baroclinic waves: Convection-permitting experiments and multistage error growth dynamics. J. Atmos. Sci., 64, 35793594, doi:10.1175/JAS4028.1.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Radar image from Dallas–Fort Worth NEXRAD (KFWS), obtained from RadarScope by Weather Decision Technologies, for a recently observed overlapping tornado and flash flood warning scenario valid at 0311 UTC 27 Apr 2015 in north-central Texas (just south of the Dallas–Fort Worth area). Red polygon corresponds to tornado warning, green polygon corresponds to flash flood warning, and yellow polygon corresponds to severe thunderstorm warning.

  • View in gallery

    Adapted from Figs. 2–4 in RR02 showing features associated with each synoptic pattern, identified within the paper, of near-tornado and flash flood events. Presented are (a) frontal-type, (b) meso-high (outflow boundary), and (c) synoptic patterns. Surface low and high centers are represented by L and H, respectively. Surface boundaries are denoted following normal mapping conventions, with the exception of the outflow boundary, which is denoted by the darker, thinner cold front convention. Dashed gray lines represent the 500-hPa trough axis, brown arrows represent the 500-hPa jet location, and gray arrows represent the 850-hPa jet. Orange dashed lines represent 850-hPa potential temperature contours. Shaded green region marks area where the potential for tornadoes and heavy rainfall exists.

  • View in gallery

    Example of the cluster analysis method for observed tornadoes on 27 Apr 2011 during a tornado outbreak. The identified clusters are given a unique color. Each tornado is identified by the small circular markers, while the cluster centroid is demarcated by the triangle markers.

  • View in gallery

    Base radar reflectivity from Goodland, Kansas (KGLD), for a training transitioning TORFF case identified to occur near 0000 UTC 26 May 2010 originating in western Kansas. Images are valid at (a) 2257 UTC 25 May, (b) 2357 UTC 25 May, (c) 0056 UTC 26 May, and (d) 0256 UTC 26 May 2010. A larger geographic area, compared to (a)−(c), is shown in (d), which covers the eastern third of Colorado, the southern quarter of Nebraska, and almost all of Kansas. Flash flood (green) and tornado (red) warnings are overlaid if valid at the time of the radar image. White arrow represents the approximate location of the verified TORFF event. Two observation intersections were associated with this event. The flash flood observation occurred around 0220 UTC 26 May 2010, while the two tornado observations intersecting the flash flood observation occurred at approximately 0026 and 0044 UTC the same day.

  • View in gallery

    Base radar reflectivity from Little Rock, Arkansas (KLZK), for training discrete TORFF case identified to occur near 2200 UTC 30 May 2013 in west‐central Arkansas. Images are valid at (a) 2132 UTC 30 May, (b) 2257 UTC 30 May, (c) 2355 UTC 30 May, and (d) 0059 UTC 31 May 2013. Flash flood (green) and tornado (red) warnings are overlaid if valid at the time of the radar image. White arrow represents the approximate location of the verified TORFF event. The flash flood observation associated with this event occurred around 0000 UTC 31 May 2013 and the tornado observation occurred at approximately 2210 UTC 30 May 2013.

  • View in gallery

    Geographic distribution of concurrent, collocated tornado and flash flood warnings over the period from 2008 to 2014 (colored by month). Black dot represents the geographic mean center, pink ellipse represents one spatial std dev away from mean center, and the black and blue lines represent NWS WFO and RFC boundaries, respectively.

  • View in gallery

    Geographic distribution of identified concurrent, collocated tornado and flash flood events over the period from 2008 to 2013 (colored by month to match Fig. 5). Number near each marker corresponds to event specifics listed in the appendix.

  • View in gallery

    Normalized seasonal frequency of the warning intersections and identified TORFF events in red and blue, respectively.

  • View in gallery

    As in Fig. 8, but for diurnal (LST) frequency.

  • View in gallery

    Histograms of point values of (a) WAA (K s−1), (b) PWAT (mm), (c) MUCAPE (J Kg−1), and (d) 0–6-km bulk shear (m s−1) for TOR (blue lines) and TORFF events (red lines). Mean values for TOR events are blue dashed lines, and TORFF events are red dashed lines. The std devs, and the 95th and 5th percentile differences of the mean of TOR and TORFF events (from the resampling procedure described in the text), are listed; when both percentile values are same signed, the difference of the mean values of TOR and TORFF events is statistically significant (black headers are statistically significant; red headers are not).

  • View in gallery

    (a) The 850-hPa WAA (K h−1; shading with values <1 K h−1 masked), temperature (K; blue dashed lines), geopotential height (m; black contours), winds (kt; gray wind barbs), and the event location (cyan circle) for TORFF events. (b) As in (a), but for the 68-member subsample of TOR events. (c) PWAT (mm; shading with values <25 mm masked), MUCAPE (J Kg−1; blue contours), 0–6-km shear (m s−1; gray wind barbs), and the event location (black circle) for TORFF events. (d) As in (c), but for the 68-member subsample of TOR events.

  • View in gallery

    Synoptic setup at 0900 UTC 27 Apr 2011. Color depicts the (a) GEFS/R control member forecast for precipitable water, (b) meridional component of 80-m winds, (c) 850-hPa vertical velocity (omega), and (d) 850-hPa temperature advection. Contours indicate the computed LSAs for this time and location; contour spacing is 0.5 (top) and 1.0 (bottom).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 229 229 57
PDF Downloads 114 114 30

Double Impact: When Both Tornadoes and Flash Floods Threaten the Same Place at the Same Time

View More View Less
  • 1 Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado
© Get Permissions
Full access

Abstract

While both tornadoes and flash floods individually present public hazards, when the two threats are both concurrent and collocated (referred to here as TORFF events), unique concerns arise. This study aims to evaluate the climatological and meteorological characteristics associated with TORFF events over the continental United States. Two separate datasets, one based on overlapping tornado and flash flood warnings and the other based on observations, were used to arrive at estimations of the instances when a TORFF event was deemed imminent and verified to have occurred, respectively. These datasets were then used to discern the geographical and meteorological characteristics of recent TORFF events. During 2008–14, TORFF events were found to be publicly communicated via overlapping warnings an average of 400 times per year, with a maximum frequency occurring in the lower Mississippi River valley. Additionally, 68 verified TORFF events between 2008 and 2013 were identified and subsequently classified based on synoptic characteristics and radar observations. In general, synoptic conditions associated with TORFF events were found to exhibit similar characteristics of typical tornadic environments, but the TORFF environment tended to be moister and have stronger synoptic-scale forcing for ascent. These results indicate that TORFF events occur with appreciable frequency and in complex meteorological scenarios. Furthermore, despite these identified differences, TORFF scenarios are not easily distinguishable from tornadic events that fail to produce collocated flash flooding, and present difficult challenges both from the perspective of forecasting and public communication.

Corresponding author address: Erik R. Nielsen, Dept. of Atmospheric Science, Colorado State University, 1371 Campus Delivery, Fort Collins, CO 80523. E-mail: erik.nielsen@colostate.edu

A comment/reply has been published regarding this article and can be found at http://journals.ametsoc.org/doi/abs/10.1175/WAF-D-16-0116.1 and http://journals.ametsoc.org/doi/abs/10.1175/WAF-D-16-0151.1

Abstract

While both tornadoes and flash floods individually present public hazards, when the two threats are both concurrent and collocated (referred to here as TORFF events), unique concerns arise. This study aims to evaluate the climatological and meteorological characteristics associated with TORFF events over the continental United States. Two separate datasets, one based on overlapping tornado and flash flood warnings and the other based on observations, were used to arrive at estimations of the instances when a TORFF event was deemed imminent and verified to have occurred, respectively. These datasets were then used to discern the geographical and meteorological characteristics of recent TORFF events. During 2008–14, TORFF events were found to be publicly communicated via overlapping warnings an average of 400 times per year, with a maximum frequency occurring in the lower Mississippi River valley. Additionally, 68 verified TORFF events between 2008 and 2013 were identified and subsequently classified based on synoptic characteristics and radar observations. In general, synoptic conditions associated with TORFF events were found to exhibit similar characteristics of typical tornadic environments, but the TORFF environment tended to be moister and have stronger synoptic-scale forcing for ascent. These results indicate that TORFF events occur with appreciable frequency and in complex meteorological scenarios. Furthermore, despite these identified differences, TORFF scenarios are not easily distinguishable from tornadic events that fail to produce collocated flash flooding, and present difficult challenges both from the perspective of forecasting and public communication.

Corresponding author address: Erik R. Nielsen, Dept. of Atmospheric Science, Colorado State University, 1371 Campus Delivery, Fort Collins, CO 80523. E-mail: erik.nielsen@colostate.edu

A comment/reply has been published regarding this article and can be found at http://journals.ametsoc.org/doi/abs/10.1175/WAF-D-16-0116.1 and http://journals.ametsoc.org/doi/abs/10.1175/WAF-D-16-0151.1

1. Introduction

Losses due to weather-related hazards have continually increased within the United States despite mitigation and advances in predictive science and technology (e.g., Changnon et al. 2000; Bouwer 2011). Among the various atmospheric hazards that can lead to loss of life and property, tornadoes and flash floods are two of the most impactful (e.g., NOAA 2011). Tornadoes occur approximately 1000 times per year in the United States, with almost 20 000 direct fatalities reported between 1880 and 2005 (Ashley 2007). Though the population-normalized yearly fatality rate associated with tornadoes has been steadily decreasing during the twentieth century (e.g., Brooks and Doswell 2002), it is possible that continued urbanization could lead to more catastrophic events in the future (e.g., Wurman et al. 2007). Flooding is responsible for almost 100 fatalities per year with the flash flooding archetype being the most deadly and, unlike tornado-related deaths, did not see any appreciable decrease in fatality rate from 1959 to 2005 (Ashley and Ashley 2008).

Given the considerable threat tornadoes and flash floods pose individually, any simultaneous occurrence of these hazards in the same location is particularly dangerous (Rogash and Smith 2000; Rogash and Racy 2002, hereafter RR02). In the case of multithreat scenarios, effective communication of the most pertinent threat to those in harm’s way can be muddled as a result of differing instructions associated with each hazard. This is of particular concern for tornadoes and flash floods since the lifesaving actions for each hazard are contradictory; tornado safety protocol recommends taking shelter in a low-lying interior room, whereas flood safety protocol recommends retreating to high ground. Throughout this study concurrent, collocated tornado and flash flood events will be referred to as TORFF(s) for the sake of simplicity. “Concurrent” in this study will refer to events where the period of each individual threat overlaps and not necessarily to events that occurred at the exact same time. Furthermore, “collocated” will refer to a particular location experiencing both threats. A visual representation of a recent multithreat tornado and flash flood scenario is presented in Fig. 1 and illustrates the complicated warning scenarios that can arise. The region encompassed by both the tornado and flash flood warnings, including the towns of Keene, Cleburne, and Grandview, Texas (Fig. 1), provides an example of the type of situation in which a TORFF event may occur.

Fig. 1.
Fig. 1.

Radar image from Dallas–Fort Worth NEXRAD (KFWS), obtained from RadarScope by Weather Decision Technologies, for a recently observed overlapping tornado and flash flood warning scenario valid at 0311 UTC 27 Apr 2015 in north-central Texas (just south of the Dallas–Fort Worth area). Red polygon corresponds to tornado warning, green polygon corresponds to flash flood warning, and yellow polygon corresponds to severe thunderstorm warning.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

There are numerous historical examples of the impacts that TORFF events can have on society. On 31 May 2013, a TORFF event in Oklahoma City, Oklahoma, tragically illustrated examples of the additional complexities in warning dissemination and risk perception in a multithreat, collocated event, which further magnified the danger beyond the meteorological hazard alone. Thirteen deaths were associated from the flash flooding whereas eight deaths were directly associated with the tornado. Perhaps most alarming, members of the public interviewed seemed to have no knowledge of the flash flooding threat despite warnings in place and social media dissemination (NWS 2014). More recently, a woman drowned in Oklahoma in May 2015 while seeking refuge from a tornado in a storm shelter (KWTV 2015). These events, and other similar events [e.g., the Nashville, Tennessee, flood in 2010; NWS (2011a)], underscore the exceptionally life-threatening situation that TORFFs can produce. Compared to single-hazard events, it moreover illustrates how such events can increase the public’s vulnerability because of complications affecting their awareness, understanding, sensitivity, and adaptive capacity (e.g., Morss et al. 2011).

TORFF events pose a particularly difficult challenge to operational forecasters (Rogash and Smith 2000), which further compounds the complexity and danger of these situations. Meteorological conditions that are favorable for tornado formation are often not conducive for flash flooding and vice versa (cf. e.g., Maddox et al. 1979; Doswell et al. 1996; Doswell 2001; Markowski and Richardson 2010; Mercer et al. 2009; Smith et al. 2012; Thompson et al. 2012). For instance, tornadoes are associated with surface-based convection (e.g., Nowotarski et al. 2011) and fast convective cell motions, while flash floods can be caused by both surface-based and elevated convection and usually need slow cell motions or “echo training” to cause large rainfall accumulations. Forecasters must be aware of and closely monitor complex situations that can provide the environmental ingredients necessary for both tornadoes and flash floods. Furthermore, the occurrence of one phenomenon—typically tornadoes—before the onset of another hazard can potentially take priority and hamper identification of subsequent weather hazards such as flash floods (e.g., Schwartz et al. 1990). The failure of forecasters to identify the flash flooding event in a timely manner can further increase the danger of TORFF events.

Compared to individual tornado and flash flood events, relatively few studies have examined the meteorology and climatology associated with TORFF events (Rogash and Smith 2000; Smith et al. 2001; RR02). RR02 discussed the meteorological characteristics associated with significant tornado events [at least two tornados rated as category 2 events on the Fujita scale (F2) or one F3 tornado] and flash flooding that occurs within 250 km and 3 h of one another from 1992 to 1998, and found that the meteorological setup was largely indicative of typical tornadic environmental characteristics. However, RR02 identified three main meteorological setups (Fig. 2) that were conducive for the nearby concurrence of tornadoes and flash flooding. Although differences do exist between the scenarios, one commonality is proximity to an approaching, distinct midtropospheric trough with event location in the divergent region of an upper-tropospheric jet streak. While RR02 looked at patterns conducive to both tornadoes and flash flooding from the same synoptic system, these authors did not investigate the enhanced hazards of directly collocated, concurrent tornado and flash flood events, and the associated complicated decision-making aspects of these events.

Fig. 2.
Fig. 2.

Adapted from Figs. 2–4 in RR02 showing features associated with each synoptic pattern, identified within the paper, of near-tornado and flash flood events. Presented are (a) frontal-type, (b) meso-high (outflow boundary), and (c) synoptic patterns. Surface low and high centers are represented by L and H, respectively. Surface boundaries are denoted following normal mapping conventions, with the exception of the outflow boundary, which is denoted by the darker, thinner cold front convention. Dashed gray lines represent the 500-hPa trough axis, brown arrows represent the 500-hPa jet location, and gray arrows represent the 850-hPa jet. Orange dashed lines represent 850-hPa potential temperature contours. Shaded green region marks area where the potential for tornadoes and heavy rainfall exists.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

This study aims to quantify the prevalence and characteristics of TORFF events to help improve future adaptive capacity. The results present a recent climatology of TORFF events, classify each event by storm characteristics, and identify unique environmental characteristics that differentiate TORFF events from non-flash-flood-producing tornado events. Section 2 describes the methodology used in this study, and results are discussed in section 3. Sections 4 and 5 summarize results and present discussion for future work and conclusions, respectively.

2. Methods

a. Case selection

Two different case identification strategies were employed in order to arrive at a representation of TORFF prevalence in both potential and verified meteorological occurrence. Each of these strategies highlights unique aspects of TORFF climatology. The first method was based on overlapping tornado and flash flood warnings, which provides an overestimate for the number of verified events but illustrates the number of times the threat was communicated as imminent to the public. The second set was based upon collocated tornado tracks and flash flood observations, which will give a conservative estimate for the number of cases but is better suited for meteorological examination, as mentioned later in this section.

ArcGIS was used to evaluate and identify cases of both documented events and weather warnings. Archived tornado and flash flood warnings from 2008 to 2014, in shapefile format, were obtained from the Iowa Environmental Mesonet (IEM) Geographic Information System (GIS) archive (available online at https://mesonet.agron.iastate.edu/GIS/). Although the archive goes back beyond 2008, these dates were chosen to coincide with the adoption of storm-based warnings for both tornadoes and flash floods (e.g., Waters et al. 2005; Ferree et al. 2006). The spatial intersections of tornado and flash flood warnings that occurred within 30 min of one another were identified over the continental United States from 2008 to 2014, using the ArcGIS pairwise intersect tool. The 30-min threshold was chosen because it is a reasonable estimate for the amount of time a person could spend sheltering in the case of a tornado warning, since the average warning is approximately 40 min in length (e.g., Sutter and Erickson 2010). The geographic centroid and spatial standard deviations associated with the warning intersections were then calculated. National Weather Service (NWS) Weather Forecast Office (WFO) and River Forecast Center (RFC) boundaries were overlaid to discern any spatial dichotomies across forecast areas. The temporal and spatial distributions of these intersections were then evaluated.

Although warning intersections show where and when the threats were deemed imminent by weather forecasters and public officials, they do not provide definite verification of TORFF events. Tornado tracks and flash flood observations were used to produce a case list of verified TORFF events. Flash flood reports were obtained in GIS shapefile format over the period between 2008 and 2013 from the U.S. Flash Flood Observation Database, which was created as part of the Flooded Locations and Simulated Hydrographs Project (FLASH; Gourley et al. 2013). Tornado track information over the same period was obtained from the Storm Prediction Center’s (SPC) Severe Geographic Information System (SVRGIS) database (available online at http://www.spc.noaa.gov/gis/svrgis/). TORFF observations were identified, using ArcGIS, as locations where both a flash flood observation and tornado track exactly intersected within 3 h of one another, similar to the time period allowed in RR02. This method only identifies exact points that fit the above criteria and will generally underestimate the number of events in the time period, since events that are slightly displaced (by even 1 km) will not be identified. However, based upon the available datasets, the method yields high confidence that a TORFF episode occurred at the identified locations and times. Throughout the rest of this study, events identified via the observation intersection method and clustering analysis discussed below will be referred to as verified TORFF events.

A given severe weather episode may produce many individual flash flood and tornado overlaps. Since the purpose of this analysis is to examine meteorological aspects of the storm system as a whole, a method was used to cluster individual events within a storm system together. Identified TORFF observations from 2008 through 2013 were clustered based upon latitude, longitude, and time using the affinity propagation (AP) technique (Frey and Dueck 2007). The Python programming language library Scikit-learn (Pedregosa et al. 2011) was used to implement this clustering technique. AP clustering considers all points as potential exemplars (cluster centers) with each data point viewed as a node in a network that then communicates recursively across the network to minimize an energy function to arrive at exemplars. In this case, the energy function used was the Euclidean distance. The number of exemplars is dynamic for a given set of points and depends on the user-specified “preference.” Also, a dampening factor is prescribed to set a limit on the number of iterations in the message passing procedure between points that impact the configuration of clusters. Through examination of extreme cases (i.e., large number of tornado event days and days with large geographic and time separation of tornado events), the preference was set to −30 and the dampening factor to 0.8 to ensure reasonable uniformity of meteorological conditions within the cluster. The choices of preference and dampening constant values were made through trial and error to obtain clusters with reasonable uniformity of meteorological conditions through space and time and to create single-point clusters for obviously isolated events. Preference can vary from −1000 to 0 while the dampening factor varies from 0 to 1. These identified cases were then used throughout the rest of the study for storm classification and meteorological characteristic analysis. Figure 3 shows an illustrative case of cluster analysis for the April 2011 tornado outbreak (NWS 2011b) in which clusters are identified outside of the main outbreak in northern Virginia and New York, while the main outbreak is composed of six clusters, in which tornadoes occurred in meteorological conditions that varied through space and time from Louisiana early on 27 April 2011 to later in the day in southwestern Virginia.

Fig. 3.
Fig. 3.

Example of the cluster analysis method for observed tornadoes on 27 Apr 2011 during a tornado outbreak. The identified clusters are given a unique color. Each tornado is identified by the small circular markers, while the cluster centroid is demarcated by the triangle markers.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

b. Radar classification

Verified events were then subjectively classified based on observed radar characteristics and, secondarily, on synoptic analyses. These events were broken down into five main classes describing the meteorological regime in place at the time of the verified TORFF event. The main subdivisions were tropical cyclone, synoptic, discrete cells, mesoscale convective system (MCS), and transitioning from discrete cells to an MCS. Verified TORFF events were characterized as MCSs if the observed radar pattern at the time of the event was indicative of previously identified MCS archetypes (e.g., Maddox 1980; Bluestein and Jain 1985; Houze et al. 1989; Parker and Johnson 2000; Schumacher and Johnson 2005). Tropical events were classified as those clearly occurring within the influence of a landfalling or inland-moving, identifiable tropical storm. Tropical examples in the analyzed period that produced verified TORFF events include Hurricanes Ike in 2008 and Isaac in 2012. Some events that occurred in the cold or warm season were associated with large-scale synoptic forcing (e.g., a cold front) and were characterized as such. However, events featuring the upscale growth of convection from discrete cells to a more organized MCS structure were classified separately. If at any point within 3 h following the verified TORFF event the responsible storm could be identified as one of the MCS archetypes, it was classified as transitioning. This category represents verified TORFF events that occur between cellular thunderstorm and mesoscale convection dominated regimes. The meso-α organization of convection, rather than the evolution of the particular cell or cells associated with the verified TORFF event(s), was assessed to make this distinction between discrete and transitioning classifications. Additionally, the typical warm season classifications (discrete, transitioning, MCS) were further subdivided based on the evaluation of cell motion. Subdivision was made based on whether convective cells repeatedly moved over the same area; events that did were given the label “training” while those that did not were labeled “nontraining.” This distinction was made since training situations typically produce increased flooding risk (e.g., Doswell et al. 1996). This creates simple classifications that describe both the convective motion and organizational characteristics of each identified event. Thus, examples of classifications used to describe verified TORFF events include “training transitioning” and “nontraining” MCS as well as “training” and “nontraining” discrete. The events characterized as nontraining discrete include verified TORFF events that were the result of discrete clusters of individual thunderstorms or in some cases a single thunderstorm. A distinction between the two is made in the appendix but not in the classification results. If no clear radar classification could be identified, the event was classified as “other.” A full list of classifications is provided in Table 1.

Table 1.

Table of classifications used and attributed to the 68 identified TORFF events.

Table 1.

Figure 4 shows a training transitioning example from 26 May 2010. The intersection of the verified TORFF event in western Kansas is annotated in Fig. 4a. Figure 4a shows the initial formation of supercell thunderstorms in western Kansas near the TORFF location. By 2357 UTC, in Fig. 4b, the base reflectivity shows convective activity continuing to affect the TORFF area. An hour later, the initial supercell, with hook echo evident, moved to the east and was replaced by a subsequent storm cell (Fig. 4c). Two hours later in Fig. 4d, at 0256 UTC, the convective activity moved away from the TORFF area and grew upscale into an MCS. Figure 5 shows an example of a training, discrete verified TORFF event (the event in Fig. 5a occurred in western Arkansas). Figure 5b, at 2257 UTC 30 May 2013, shows a situation similar to the training transitioning case with two supercells poised to affect the TORFF area. By 0059 UTC the western convective cell had intensified, but there were no signs of this specific activity organizing upscale into an MCS, although one did form from other convective cells in the area.

Fig. 4.
Fig. 4.

Base radar reflectivity from Goodland, Kansas (KGLD), for a training transitioning TORFF case identified to occur near 0000 UTC 26 May 2010 originating in western Kansas. Images are valid at (a) 2257 UTC 25 May, (b) 2357 UTC 25 May, (c) 0056 UTC 26 May, and (d) 0256 UTC 26 May 2010. A larger geographic area, compared to (a)−(c), is shown in (d), which covers the eastern third of Colorado, the southern quarter of Nebraska, and almost all of Kansas. Flash flood (green) and tornado (red) warnings are overlaid if valid at the time of the radar image. White arrow represents the approximate location of the verified TORFF event. Two observation intersections were associated with this event. The flash flood observation occurred around 0220 UTC 26 May 2010, while the two tornado observations intersecting the flash flood observation occurred at approximately 0026 and 0044 UTC the same day.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

Fig. 5.
Fig. 5.

Base radar reflectivity from Little Rock, Arkansas (KLZK), for training discrete TORFF case identified to occur near 2200 UTC 30 May 2013 in west‐central Arkansas. Images are valid at (a) 2132 UTC 30 May, (b) 2257 UTC 30 May, (c) 2355 UTC 30 May, and (d) 0059 UTC 31 May 2013. Flash flood (green) and tornado (red) warnings are overlaid if valid at the time of the radar image. White arrow represents the approximate location of the verified TORFF event. The flash flood observation associated with this event occurred around 0000 UTC 31 May 2013 and the tornado observation occurred at approximately 2210 UTC 30 May 2013.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

c. Meteorological analysis

The authors employed two separate methods to assess the differences in the meso-α- (~100–1000 km) to synoptic-scale (~1000–2000 km) atmospheric characteristics between TORFF and tornado-only events (TOR). Since most severe weather, including tornadoes, is dependent on absolute environmental conditions being satisfied [e.g., convective available potential energy (CAPE) and low-level shear above certain thresholds], one method looks at the direct full-field values in the assessment of the synoptic pattern. Because flash flood forecasting depends on factors beyond absolute rainfall amount, another examines local climatological anomalies for the determination of flash flood risk and additional synoptic variables. The two complementary analyses will be described in more detail below.

1) Full-field analysis (FFA)

To identify scenarios that produced only tornadoes (i.e., TOR dataset), all tornado events from 2008 through 2013 in the SPC’s SVRGIS were clustered using the same algorithm described for TORFF case identification. From 7886 tornado observations, 1622 event clusters were identified with clusters sizes ranging from 1 to 57 tornadoes. When tornado outbreaks occur, multiple clusters are identified that capture the changes in the environmental variables as the system, for example, transitions from afternoon supercell activity to nighttime MCS activity several hundred kilometers downstream. The verified TORFF events were then removed to form the database of TOR events.

In the first method, wind, temperature, moisture, and geopotential height fields were obtained from the North American Regional Reanalysis (NARR; Mesinger et al. 2006) at the time of each verified TORFF event and 68 (the size of the TORFF case list) randomly selected TOR events. Data were collected within a 45 × 45 gridpoint window (the NARR grid spacing is approximately 32 km) centered at the location of the event.

The authors then generated event-centered composites of 850-hPa horizontal temperature advection, focusing in particular on warm-air advection (WAA), 850-hPa winds, 850-hPa temperature, precipitable water (PWAT), most unstable convective available potential energy (MUCAPE), and 0–6-km bulk shear separately for TORFF events and TOR events. Lifting associated with locally strong low-level WAA (which is identical to isentropic upglide for steady conditions) is often cited as a mechanism for storm initiation and persistence in heavy convective precipitation events (e.g., Maddox et al. 1979; Moore et al. 2003; Schumacher and Johnson 2005; Trier et al. 2010; Peters and Schumacher 2014, 2015). WAA also occurs in conjunction with veering low-level wind profiles, which is an important ingredient in low-level mesocyclogenesis (e.g., Weisman and Klemp 1984; Thompson et al. 2012). Locally (and anomalously) high PWAT values are also a frequent characteristic of flash flood environments (e.g., Junker et al. 1999; Moore et al. 2003; Schumacher and Johnson 2005) and promote high precipitation efficiency (e.g., Doswell et al. 1996). High PWAT may also be a signature of deep-tropospheric lift (which favorably influences convective convergence and intensity by steepening lapse rates), where upward motion transports moisture into the mid- to upper troposphere and increases the column-integrated water vapor. Convective available potential energy is a necessary ingredient for deep moist convection, and high CAPE values promote more vigorous ascent and more intense rainfall rates when compared to low CAPE values (e.g., Doswell et al. 1996). The probability of significant tornadoes is also correlated with high CAPE (e.g., Thompson et al. 2012). Finally, the 0–6-km shear influences storm mode and persistence. Specifically, strong 0–6-km shear environments promote supercell storm mode (Thompson et al. 2012) and long-lived MCSs (Coniglio and Corfidi 2006) when compared to weaker 0–6-km shear environments.

Composites for verified TORFF and TOR events were compared both qualitatively and quantitatively. For the quantitative comparison, values of WAA, PWAT, MUCAPE, and 0–6-km shear were averaged over a 5 × 5 gridpoint box centered at the event location for each case (referred to as local averages). Local averages of 68 element samples, to correspond to the sample size of verified TORFF events, were formed via a bootstrapping procedure from each of the TOR and verified TORFF datasets by sampling uniformly with replacement. This process was repeated 1000 times with a new sample of 68 randomly calculated each time. The differences between the average values of quantities for each pair of resampled TORFF and TOR datasets were then computed. The differences were deemed statistically different from zero if both the 97.5th percentile smallest and the 2.5th percentile smallest were of the same sign. For example, if the 950 largest of 1000 resampled local-average differences in WAA between TORFF and TOR are positive, then the difference between WAA local averages in the TORFF and TOR datasets is statistically significant. Further, probability density functions (PDFs) for the composited fields were created to illustrate any distribution differences and highlight the calculated mean values.

2) Local standardized anomalies (LSAs)

To discern what synoptic-scale anomalies relative to regional climatology distinguish verified TORFF from TOR events, the second meteorological analysis method uses historical model forecasts to compare the typical meteorological conditions associated with each case. Thirty years of reforecasts associated with the National Oceanic and Atmospheric Administration’s (NOAA) second-generation Global Ensemble Forecast System reforecast dataset (GEFS/R) were used to compare these event classes (Hamill et al. 2013). GEFS/R contains reforecasts of the February 2012 version of the NCEP GEFS from December 1984 to the present. A forecaster would have access to real-time guidance, such as GEFS, when attempting to ascertain the threat of a TORFF event. Identified differences between forecast conditions in GEFS/R between verified TORFF and TOR events were thus considered to have potential utility from a forecasting perspective. For the purposes of this study, only data from the control member of GEFS/R was used to compare conditions corresponding to the event types. Since the dataset contains once-daily runs initialized at 0000 UTC, data from the nearest initialization to the event time (0–21-h forecasts) were used. The majority of fields are archived on the model’s native T254L42 (~40-km equivalent horizontal grid spacing at 40° latitude) resolution; however, the fields aloft (on pressure levels) are only stored on a 1° × 1° grid (Hamill et al. 2013).

Making an accurate comparison between the conditions yielding verified TORFF events versus only TOR events requires evaluating sample data that are as comparable as possible. Having one event class disproportionately sampled from one region of the country compared to another may yield results that simply reflect differences in the climatology of the predominantly sampled region or regions. Similarly, one event class being oversampled from one period of the year or day, with respect to the other, may yield results that simply reflect climatological differences between various phases of the seasonal or diurnal cycle, respectively. These differences, while important distinctions that are examined in the FFA, are well established and do not identify all of the unique characteristics associated with verified TORFF events. To alleviate these concerns and ensure a more representative comparison, mean local standardized anomalies (LSAs) were compared between samples following a methodology similar to a previously established procedure (e.g., Hart and Grumm 2001). Conceptually, the differences between TOR and verified TORFF events found in the FFA method [described in section 2c(1)] could come from either regional differences, timing differences, or fundamental environmental differences. The latter part of this decomposition was identified here using the LSA method.

Local standardized anomalies as defined in this study are “local” in both the temporal and spatial senses. For each cluster identified in TORFF and TOR case identification, the nearest data grid point to the cluster centroid was taken as the observation location, and the nearest multiple of 3 h, beginning from 0000 UTC, to the cluster centroid time was taken to be the observation time. Having identified the observation location and time in the GEFS/R record, a 21-day window centered about the date of the event (10 days in either direction) was used for each year over the data record (from December 1984 to May 2014) to gather a local climatology for a specific atmospheric variable at the approximate time of event occurrence. This set of values was then standardized to form a set of LSAs; the LSA corresponding to the date of event occurrence was extracted as the LSA for the cluster and atmospheric field being analyzed. Repeating this for every identified cluster in an event class (TORFF or TOR) yielded a set of LSAs for the given event class and atmospheric variable in question. The mean of these LSA values was computed; these values are interpreted as the typical expected standardized anomaly to anticipate for the atmospheric variable and event type in question, irrespective of time or location. This procedure was performed for all identified verified TORFF and TOR events to determine mean LSAs for both classes of events. The difference in mean LSAs was then calculated to determine what, if any, synoptic differences distinguish cases that produce collocated, concurrent tornadoes and flash floods from those cases, which produce tornadoes but not flash flooding. A difference-of-means test was used to assess the statistical significance of these discrepancies.

This overall procedure was performed for a variety of atmospheric fields, including wind—both zonal and meridional components—at several levels from the surface to 500 hPa, specific humidity also from the surface to 500 hPa, vertical motion at 850 hPa (specifically omega in this analysis), deep-layer (from the surface to 500 hPa) and shallow-layer (from the surface to 850 hPa) meridional wind shear, surface temperature, mean sea level pressure (MSLP), PWAT, and CAPE. Additionally, in order to obtain a very rough gauge of the impact of antecedent land surface conditions at distinguishing these event types, volumetric soil moisture content (SOILW), as output by GEFS/R, was also assessed.

3. Results

a. Climatology

Over the period between 2008 and 2014 in the continental United States, there were approximately 2800 tornado and flash flood warning intersections within 30 min of one another at the same location; the spatial and seasonal distributions of these intersections are presented in Fig. 6. Of these approximately 2800 intersections, the flash flood warning was issued before the tornado warning in about 1300 instances. Furthermore, when the flash flood (tornado) warning preceded the tornado (flash flood) warning, it was issued on average 14.5 (16) minutes earlier; however, the frequency distributions were fairly uniformly distributed. The number of warning intersections increased when allowing a larger time separation between tornado and flash flood warnings (Table 2); however, the spatial distribution was insensitive to the time separation between warnings (not shown). The geographic mean center of the intersections is located in south-central Missouri, with the majority of intersections occurring in the Mississippi River valley from the eastern central plains to the Appalachian Mountains (Fig. 6). Warning intersections, however, are not limited to the geographic area depicted in Fig. 6, with events occurring from coast to coast, although in some areas very few cases were identified (e.g., west of the Continental Divide with only five intersections not present on the map). Seasonally, the warning intersections occurred earlier in the calendar year in the southern United States (March–May), and become more common northward later on in the spring and summer (May–July) (Fig. 6). The effects of the tropical cyclones in the record are apparent along parts of the Gulf Coast and eastern seaboard by the existence of events in August and September. Upon closer examination, significant gradients in warning intersections appear across NWS WFO boundaries (Fig. 6). Some WFOs do not have any intersections over the time period investigated, whereas neighboring WFOs frequently have intersecting tornado and flash flood warnings. Furthermore, differences in warning intersections across NWS RFC boundaries do exist but are not as sudden as those across WFO boundaries. In general, the distribution and frequency of warning overlaps is maximized in the central to lower Mississippi valley from spring to early summer but overlaps have occurred through all seasons and across the contiguous United States.

Fig. 6.
Fig. 6.

Geographic distribution of concurrent, collocated tornado and flash flood warnings over the period from 2008 to 2014 (colored by month). Black dot represents the geographic mean center, pink ellipse represents one spatial std dev away from mean center, and the black and blue lines represent NWS WFO and RFC boundaries, respectively.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

Table 2.

Table depicting the number of collocated tornado and flash flood warnings identified to have been issued within the specified time threshold of one another from 2008 to 2014. During these years, a total 26 108 flash flood warnings and 21 967 tornado warnings were issued. The right column depicts the number of events on average per year over the period analyzed at the specific time threshold.

Table 2.

In total, 68 verified TORFF events were identified in the period from 2008 to 2013 from the clustering of a total of 221 overlapping observations (Fig. 7). The full list of these events is presented in the appendix. The seasonal cycle of warning overlaps versus verified clustered TORFF events is shown in Fig. 8. Both datasets, and particularly the warning dataset, exhibit seasonality that is very similar to the seasonal cycle of tornadoes over the contiguous United States (Brooks et al. 2003), with a broad springtime peak from April through June, and distinct decreases in frequency before and after this period. In the observation record, and to a lesser extent in the warning record, very few events were identified during the late autumn and early winter. This is likely explained by tornadoes and flash floods becoming less frequent during these months, and the subsequent intersections of these events becoming less probable. Figure 9 illustrates the TORFF diurnal cycle for each dataset. The cycle is again in general accord with the U.S. tornado climatology: a peak in frequency in the afternoon and early evening hours (~1400–1900 LST), with a broad minimum occurring during the early morning hours (~0100–0900 LST). There is a slight offset seen when comparing the warning and observation datasets, with the peak frequency of observed TORFF events occurring roughly 1–2 h before the warning-based diurnal maximum; however, this does not imply that the verified TORFF events came, on average, before their warnings. If there were a one-to-one correspondence between observations and warnings (i.e., every warning intersection had a corresponding verified observation, which is not the case here), this would, in fact, imply a lack of (in fact, negative) lead time for events. Here, it could be a number of things, including events being more likely to be verified during daylight (i.e., earlier), an artifact of the identification methods employed, or sampling error, both in true differences in the diurnal cycle based on bulk geographical and seasonal differences between the samples and from random sampling.

Fig. 7.
Fig. 7.

Geographic distribution of identified concurrent, collocated tornado and flash flood events over the period from 2008 to 2013 (colored by month to match Fig. 5). Number near each marker corresponds to event specifics listed in the appendix.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

Fig. 8.
Fig. 8.

Normalized seasonal frequency of the warning intersections and identified TORFF events in red and blue, respectively.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

Fig. 9.
Fig. 9.

As in Fig. 8, but for diurnal (LST) frequency.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

b. Radar classification

The results of the radar-based classification procedure described in section 3b are shown in Table 3. The largest percentage of cool season events fell into the synoptic category, at 20% of total cases (13 events). In the warm season, landfalling and inland tropical cyclones accounted for seven events (11% of the total verified TORFF events). The remainder of the cases (nearly all of which occurred in the warm season) contributed nearly equal numbers to the total: MCS with 15 events (23% of total), transitioning with 16 events (24% of total), and discrete with 15 events (23% of total). Twenty-five nontropical, warm season events were associated with training convective cells, while only 18 were associated with nontraining radar signatures. The discrete category was evenly split among six training and nontraining events. In the case of verified TORFF events occurring from a preexisting MCS, the majority were nontraining events with nine compared to the six training events. Training was present in the majority of the transitioning events, with 13 training transitioning events and 3 nontraining transitioning events present in the data.

Table 3.

Breakdown of radar-based TORFF event classifications for the 68 identified events.

Table 3.

c. Meteorological analysis

1) Full-field analysis

The overall distributions of the local averages for verified TORFF and TOR events were similar for WAA, PWAT, MUCAPE, and 0–6-km shear, with standard deviations of TOR local averages on the order of 10%–15% higher than that of verified TORFF events (Fig. 10). Both TORFF and TOR events typically occurred along an east–west-oriented low-level temperature gradient, within a locally enhanced low-level jet (LLJ), and within a region of locally maximized low-level WAA (Figs. 11a,b). The expanse and magnitude of WAA is larger in the verified TORFF composites than the TOR composites (Fig. 11a), and the verified TORFF local averages are statistically larger than the TOR local averages (Fig. 10a). This result suggests that low-level synoptic-scale forcing for ascent is typically larger, and promotes more widespread storm development in verified TORFF events than TOR events. Further, the increased WAA in the TORFF composite is present not only near the event center, but also upstream (Figs. 11a,b). The patterns of PWAT, MUCAPE, and 0–6-km shear were also compared similarly between the verified TORFF and TOR composites (Figs. 11c,d). Events occurred within east–west-oriented gradients of PWAT and MUCAPE (with highest values to the south), and southwesterly 0–6-km shear of 35–40 knots (kt; 1 kt = 0.51 m s−1). PWAT and MUCAPE values from verified TORFF events near the event location were, on average, larger than in TOR events (Figs. 11c,d), and the local-average differences in these quantities are statistically significant (Figs. 10b,c). This suggests that the environments of verified TORFF events promote more vigorous storms with higher rainfall production and precipitation efficiencies. The 0–6-km bulk shear is slightly higher in the TORFF composites than in the TOR composites (Figs. 11c,d); however, the local averages of this difference are not statistically significant (Fig. 10d). This suggests 0–6-km shear does not strongly influence whether a tornado event will produce a flash flood.

Fig. 10.
Fig. 10.

Histograms of point values of (a) WAA (K s−1), (b) PWAT (mm), (c) MUCAPE (J Kg−1), and (d) 0–6-km bulk shear (m s−1) for TOR (blue lines) and TORFF events (red lines). Mean values for TOR events are blue dashed lines, and TORFF events are red dashed lines. The std devs, and the 95th and 5th percentile differences of the mean of TOR and TORFF events (from the resampling procedure described in the text), are listed; when both percentile values are same signed, the difference of the mean values of TOR and TORFF events is statistically significant (black headers are statistically significant; red headers are not).

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

Fig. 11.
Fig. 11.

(a) The 850-hPa WAA (K h−1; shading with values <1 K h−1 masked), temperature (K; blue dashed lines), geopotential height (m; black contours), winds (kt; gray wind barbs), and the event location (cyan circle) for TORFF events. (b) As in (a), but for the 68-member subsample of TOR events. (c) PWAT (mm; shading with values <25 mm masked), MUCAPE (J Kg−1; blue contours), 0–6-km shear (m s−1; gray wind barbs), and the event location (black circle) for TORFF events. (d) As in (c), but for the 68-member subsample of TOR events.

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

2) Local standardized anomalies

From the filled contours, Figs. 12a–d illustrate the full-field values, which are determined for each event to be an intermediate step in the LSA method, for precipitable water, the meridional component of 80-m wind, and the vertical velocity and horizontal temperature advection at 850 hPa, respectively, for the 27 April 2011 tornado outbreak, which produced a verified TORFF event [case 48; Knupp et al. (2014)]. The full-field variables in Fig. 12 are very similar to the FFA presented in the previous section, but are only an intermediate step in the LSA method. It is apparent that this is a very strongly forced environment, with a large region of warm-air advection exceeding 0.5 K h−1 extending from Arkansas and Louisiana northeast through western Tennessee and Virginia into Indiana. Further evidence is the nearly collocated synoptic vertical rising motion (GEFS/R being too coarse a model to resolve individual updrafts in a convective system) of mostly greater than 1 Pa s−1, in addition to the strong low-level southerlies at 80 m, with sustained wind speeds of up to 15 m s−1 forecast. It is also clear that the environment is conducive to the production of appreciable rainfall, with updrafts in a region of 40–60-mm PWAT. What is less clear, absent additional knowledge of the local climate of the southeastern United States, however, is whether these conditions are sufficient to produce flash flooding, or whether instead these conditions are sufficiently typical for the area that the land surface is able to accommodate the amount of rainfall without appreciable runoff. To gain a better appreciation for how the conditions on this day compare with the local climatology for this time of year, the LSAs for this case are also shown (Fig. 12, contours). Some fields, like vertical motion and temperature advection, have a fairly uniform mean-state value of 0, and the LSAs reduce to a simple rescaling of the full-field values, with the exact rescaling depending on the regional variance. It is evident that the forcing seen here is extremely anomalous, with a large region of anomalies in excess of four standard deviations from typical conditions. However, with other fields such as PWAT, the relationship between full-field values and local anomalies is less clear since the mean value varies with location. This can be seen in Fig. 12a, with a northward displacement seen with the LSA maxima compared with the full-field maxima, indicating the climatologically higher precipitable water to the south. The PWAT anomalies in excess of one standard deviation above normal and maximizing with a PWAT anomaly over two standard deviations above normal over western Tennessee highlight the added risk of flash flooding in this environment, in addition to the identified tornado threat.

Fig. 12.
Fig. 12.

Synoptic setup at 0900 UTC 27 Apr 2011. Color depicts the (a) GEFS/R control member forecast for precipitable water, (b) meridional component of 80-m winds, (c) 850-hPa vertical velocity (omega), and (d) 850-hPa temperature advection. Contours indicate the computed LSAs for this time and location; contour spacing is 0.5 (top) and 1.0 (bottom).

Citation: Weather and Forecasting 30, 6; 10.1175/WAF-D-15-0084.1

Table 4 shows the mean LSAs of verified TORFF events for a variety of atmospheric fields. Overall, the findings largely agree with what one would intuitively expect for a TORFF event with the necessary ingredients for convection of moisture, instability, and lift all being present (e.g., Doswell et al. 1996). In the moisture fields, anomalously high specific humidity is apparent throughout the troposphere, especially in the lower levels where the near-surface mean specific humidity LSA was 1.62 compared with 0.87 at 500 hPa. Anomalous column-integrated moisture is perhaps best illustrated in the precipitable water field, where the mean LSA was 1.54. A stronger signal is observed in the meridional component of the wind, with LSA anomalies larger than one standard deviation at all levels analyzed, and the sign of the mean LSA indicating anomalous southerly flow. The low-level anomalies are consistent with a strong, southerly LLJ signature, while the upper-level anomalous southerlies are also consistent with being on the upstream side of a long wave trough and enhanced moisture flux due to the persistent meridional moisture gradient. A stronger signal still is seen with respect to vertical velocity, with very strong anomalous rising motion observed at the time of the event. Since the vertical motion associated with individual convective cores is not resolved in GEFS/R, this result is a signal of stronger synoptic- to mesoscale ascent forcing in the verified TORFF events. WAA at 850 hPa was also found to be anomalously high with an LSA of 1.27. Results from analysis of the CAPE and SOILW fields should be interpreted with care, since at many grid points, these fields rarely deviate from their mode state (i.e., the most common model value: 0 J kg−1 in the case of CAPE) and rarely have very large values even when the conditions are favorable for convection. Thus, only one or two days may deviate appreciably from the mode state, in some instances producing very large standardized anomalies. Nevertheless, the results indicate that anomalously high forecasted CAPE and anomalously high antecedent soil moisture are present during TORFF events. In the wind fields, the analysis indicates slight easterly anomalies in the low-level flow, with mean LSAs of −0.44 and −0.31 for the zonal component of 10- and 80-m winds, respectively. In the mid- to upper levels, an anomalous westerly component of the wind is typically present, which likely corresponds to a stronger than normal upper-level jet. Further, verified TORFF events also occur with anomalously strong meridional wind shear, which is unsurprising considering shear is a necessary ingredient for tornadogenesis (e.g., Markowski and Richardson 2010). A statistically significant warm anomaly, although of small magnitude compared to other fields in this study, is apparent in surface temperatures on verified TORFF event days, and a strong low pressure anomaly is also observed, with a mean LSA of −1.63 present in the mean sea level pressure field. Overall, this analysis shows that the ingredients for both tornadic and heavily raining storms are in place during verified TORFF events.

Table 4.

Results of the mean LSAs calculated in this study. The TORFF row depicts the mean anomaly from the sample of all identified TORFF cases (68) compared to the climatological environment. The TORFF − TOR row represents the difference in the mean LSAs between identified cases producing simultaneous collocated tornadoes and flash floods and those that produced only tornadoes (1622 cases over 2008–13; positive values indicate that TORFF events were more positively anomalous). Anomaly differences statistically significantly different from zero (α = 0.05) are depicted in boldface; differences significant at 90% but not 95% confidence are italicized. The column headers U10 and V10 depict the zonal and meridional components of the 10-m winds, respectively; U80 and V80 depict the zonal and meridional components at 80 m above the ground, respectively; U500, V500, U850, and V850 correspond to the zonal and meridional components of the wind at 500 and 850 hPa, respectively; ω850 corresponds to omega at 850 hPa (more negative implying stronger ascent); V500,10 and V850,10 indicate the difference between the meridional component of the wind from the 500- and 850-hPa levels and 10 m; T2M indicates the 2-m temperature; Q2M, Q850, and Q500 denote the specific humidity at 2 m, 850 hPa, and 500 hPa, respectively. WAA corresponds to 850-hPa warm-air advection. Data, unless otherwise noted, are taken from the native Gaussian grid of GEFS/R. Variables U850, U500, V850, V500, V850,10, V500,10, Q850, and Q500 use the 1° × 1° grid.

Table 4.

The mean LSA differences observed between verified TORFF and TOR events are also shown in Table 4; positive values indicate the mean LSA was more anomalously positive for the verified TORFF event class. The general finding from this analysis is that cases that produce flash floods in addition to tornadoes are characterized by stronger synoptic-scale forcing for ascent, stronger vertical wind shear, and more low-level moisture. This is shown by the mean LSA differences being approximately one-half standard deviations for the surface, 850‐hPa, and column-integrated moisture fields with 2-m specific humidity, 850-hPa specific humidity, and precipitable water, respectively. Additionally, surface pressure anomalies are, on average, lower in verified TORFF events than TOR events. Forcing for ascent was also substantially larger in verified TORFF events than TOR events at 850 hPa in both the vertical velocity and WAA fields. Also consistent with the enhanced moisture anomalies, though anomalous southerlies are observed at all levels for both verified TORFF and TOR event classes, the anomalous southerly flow is stronger in the verified TORFF cases. This is also true of both shallow- and deep-layer meridional wind shear, which was found to be statistically significantly stronger in verified TORFF cases compared with tornado-only days. In TOR cases, there is no distinguishable signal in the low-level zonal flow anomaly, but because of the anomalous near-surface easterlies observed in verified TORFF cases, there is also an anomalous low-level easterly difference with TORFF events relative to TOR events, perhaps attributable to a difference in the spatial distribution of the two event classes relative to synoptic systems. No statistically significant anomaly differences are observed with respect to surface temperature or upper-level zonal flow.

4. Discussion

This study demonstrates that TORFF warning intersections occur quite frequently with an average of 400 yr−1 at the 30-min overlap threshold. This number is higher than the authors expected and highlights the nonnegligible frequency of these dangerous situations in public dissemination. The distribution of warning intersections changes suddenly, in some cases, near NWS WFO boundaries (i.e., Fig. 6), suggesting warning intersections could be affected by individual WFO issuance tendencies. However, differences in meteorological factors in the zonal and meridional directions could also affect the distribution of TORFF warning intersections. Influences of NWS RFC flash flood guidance (FFG) products could also influence the spatial patterns, especially if FFG is biased low and increases flash flood warning false alarm ratios (e.g., Clark et al. 2014). Additionally, 68 TORFF events were identified from 2008 to 2013, using the conservative method for TORFF verification in this study. This more than doubles the 31 events identified, albeit with different but similar criteria, by RR02 from 1992 to 1998. These results highlight the possible underestimation of the TORFF event frequency in both the scientific data record and the public’s perception of their occurrence. Recent highly publicized TORFF events (e.g., NWS 2014) have brought their existence to the attention of forecasters and the public, but the danger of underestimating their impacts is, nevertheless, present.

The meteorological characteristics of the verified TORFF events are quite complicated and occur across many precipitation modes. The radar classifications (Table 3) produced a near-uniform distribution among the main archetypes (i.e., MCS, transitioning, synoptic, and discrete), with the exception of the seasonally dependent tropical category. The most prevalent classification, training transitioning, highlights the period of upscale growth from discrete convection to organized mesoscale systems as a particularly favorable situation for verified TORFF events to occur. Given the great difficulty associated with warm season quantitative precipitation forecasts (e.g., Fritsch and Carbone 2004) and related flash flood threat, the enhanced TORFF frequency of occurrence in the transitioning phase of MCS development adds to the challenge of TORFF forecasting, especially given that this phase of MCS development is often associated with some of the most rapid error growth (e.g., Zhang et al. 2007). The speed of the transition can also speak to the likelihood of overlapping, concurrent weather threats (e.g., Coniglio et al. 2010). While training transitioning storms make up a large portion of TORFF cases identified in this study, there does not seem to be one dominant mode of TORFF production, which illustrates the forecast complexity modes of classification found in this study would likely change if a spatial buffer between the flash flood and tornado observations is used.

The event-centered composites (Fig. 11) highlight, in general, that verified TORFF event environments exhibit typical lower-level tornadic characteristics, but with stronger lower-level synoptic forcing present in TORFF cases compared to TOR events. The large values of WAA, PWAT, and MUCAPE suggest TORFF environments produce more widespread, vigorous convection. The LSAs identified similar synoptic-scale characteristics indicative of TORFF events even when the effects of local climatology and seasonality are removed. Not only are TORFF environments found to occur in moister environments with a stronger forcing for ascent both upstream and at the event center than the complementary TOR cases, but they also occur in environments that are more anomalously moist relative to the local climatology. Specifically, the verified TORFF events occurred in regions of anomalously high lower-level vertical motion and warm-air advection at 850 hPa, PWAT, and lower-level specific humidity at all levels examined at and below 500 hPa. The LLJ is more amplified in the verified TORFF events compared to TOR cases with enhanced upper-level southerlies, which aids in meridional moisture flux. The LSAs also indicate the importance of antecedent soil moisture conditions on the occurrence of verified TORFF events but is a particularly challenging aspect to forecast and even observe (e.g., Robock et al. 2000). Despite the differences in what these methods diagnose, both the LSAs and full-field composites yielded the same basic results. Given the wide range of the contiguous United States in which verified TORFF events can occur, the application of both full-field thresholds and regional anomalies is advantageous in discerning unique aspects of the differences between the two event classes examined in this study. Though the results of the meteorological analysis are not necessarily surprising, the identified differences may aid awareness and mindfulness when forecasting favorable environmental conditions for tornadogenesis.

Direct comparisons between RR02 and the results in this study can be problematic because of the differences in case identification; however, similarities between the two studies exist. Setting aside this study’s tropical and discrete cases, which do not fit into the classification scheme of RR02, leaves the synoptic, MCS, and transitioning categories. The synoptic class is nearly identical to the RR02 category as a prefrontal, synoptic-scale area of precipitation usually in the form of a squall line. Examining the MCS categories in this study, all 31 cases fit into either the meso-high scenario of RR02, in which an MCS exists in the warm sector, or the frontal category of RR02, which is dominated by activity along the warm front.

TORFF events not only present a complicated meteorological challenge but also, as discussed in the introduction, create intricate communication interactions that can further elevate the danger. Countless hazard identification and communication scenarios are possible depending on the multithreat event specifics, which further complicates the methods of evaluation for the forecaster and public response [e.g., as discussed for flash flooding in Morss et al. (2015)]. It is not unreasonable to expect fluctuations in the length of time spent for multihazard identification depending on the complexity and characteristics of the meteorological setup. Additionally, since a unified threat message associated with the TORFF event would not be delivered, confusion on the appropriate action to take could further increase the time an individual spends deciding what action to take, especially if the lifesaving instructions are contradictory (e.g., Hough and White 2003). The possible notification scenarios and actions are further multiplied with an individual possibly having knowledge or partial knowledge of one threat, both threats, or no knowledge of any threat. The complications due to the communication challenges alone could in turn leave less time to implement an action plan that may or may not account for both threats, potentially leading to hazardous outcomes. Given the amount of uncertainty in the human and meteorological aspects, the final individual personalization and actions taken are unknowable. These potential complications in the communication scenarios imply that different warning communication methods may be needed in concurrent, collocated, multithreat scenarios such as TORFF events. While no definitive answer to these communication challenges can be determined without extensive social science research, this work is an important step in starting the topic of conversation to continually improve and innovate potential lifesaving notifications.

5. Conclusions

TORFF intersections, defined as officially disseminated overlapping warnings, occurred on average 400 times per year between 2008 and 2014. They occurred with maximum frequency in the lower Mississippi River valley with very few occurrences west of the Continental Divide. Between 2008 and 2013, a total of 68 verified TORFF events were identified and classified based on radar and synoptic characteristics into categories of tropical, synoptic, MCS, discrete, and transitioning categories. Through two analyses, the synoptic conditions associated with verified TORFF events were found to be similar to typical tornadic environments; however, the TORFF environment exhibited stronger large-scale forcing for ascent and tended to be moister through the atmospheric column. Furthermore, verified TORFF events are difficult to distinguish from tornadic events that share these characteristics and do not produce a collocated flash flood. The results of this study show that TORFF events occur in complex meteorological scenarios with substantial frequency and present challenges through the entire weather warning process from forecasting to public communication and action. Future work includes examining the storm-scale dynamics of individual TORFF events, including a spatial buffer in case identification, as well as examining the social and communication aspects associated with TORFF and other multihazard weather events.

Acknowledgments

Special thanks go out to the participants of the Studies of Precipitation, Flooding, and Rainfall Extremes Across Disciplines (SPREAD) workshop for inspiration and framing the social context of this research. The authors would also like to thank Stacey Hitchcock, Jamie Rhome, JVD, and three anonymous reviewers for helpful technical suggestions and discussion through the research process. Furthermore, the authors thank the Iowa Environmental Mesonet for making the GIS warning datasets available. This research was supported by National Science Foundation Grant AGS-1157425; NOAA Award NA14OAR4320125, Amendment 26; and by National Science Foundation Graduate Research Fellowship Grant DGE-1321845, Amendment 3.

APPENDIX

Identified TORFF Events

Table A1.

List of verified TORFF events and radar classifications used in the meteorological analysis. Case numbers correspond to those geographically mapped in Fig. 7

Table A1.
Table A2.

Continued list of verified TORFF events and radar classifications used in the meteorological analysis. Case numbers correspond to those geographically mapped in Fig. 7

Table A2.

REFERENCES

  • Ashley, S. T., , and Ashley W. S. , 2008: Flood fatalities in the United States. J. Appl. Meteor. Climatol., 47, 805818, doi:10.1175/2007JAMC1611.1.

    • Search Google Scholar
    • Export Citation
  • Ashley, W. S., 2007: Spatial and temporal analysis of tornado fatalities in the United States: 1880–2005. Wea. Forecasting, 22, 12141228, doi:10.1175/2007WAF2007004.1.

    • Search Google Scholar
    • Export Citation
  • Bluestein, H. B., , and Jain M. H. , 1985: Formation of mesoscale lines of precipitation: Severe squall lines in Oklahoma during the spring. J. Atmos. Sci., 42, 17111732, doi:10.1175/1520-0469(1985)042<1711:FOMLOP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bouwer, L. M., 2011: Have disaster losses increased due to anthropogenic climate change? Bull. Amer. Meteor. Soc., 92, 3946, doi:10.1175/2010BAMS3092.1.

    • Search Google Scholar
    • Export Citation
  • Brooks, H. E., , and Doswell C. A. III, 2002: Deaths in the 3 May 1999 Oklahoma City tornado from a historical perspective. Wea. Forecasting, 17, 354361, doi:10.1175/1520-0434(2002)017<0354:DITMOC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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
  • Changnon, S. A., , Pielke R. A. Jr., , Changnon D. , , Sylves R. T. , , and Pulwarty R. , 2000: Human factors explain the increased losses from weather and climate extremes. Bull. Amer. Meteor. Soc., 81, 437442, doi:10.1175/1520-0477(2000)081<0437:HFETIL>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Clark, R. A., , Gourley J. J. , , Flamig Z. L. , , Hong Y. , , and Clark E. , 2014: CONUS-wide evaluation of National Weather Service flash flood guidance products. Wea. Forecasting, 29, 377392, doi:10.1175/WAF-D-12-00124.1.

    • Search Google Scholar
    • Export Citation
  • Coniglio, M. C., , and Corfidi S. , 2006: Forecasting the speed and longevity of severe mesoscale convective systems. Preprints, Symp. on the Challenges of Severe Convective Storms, Atlanta, GA, Amer. Meteor. Soc., P1.30. [Available online at https://ams.confex.com/ams/pdfpapers/104815.pdf.]

  • Coniglio, M. C., , Hwang J. Y. , , and Stensrud D. J. , 2010: Environmental factors in the upscale growth and longevity of MCSs derived from Rapid Update Cycle analyses. Mon. Wea. Rev., 138, 35143539, doi:10.1175/2010MWR3233.1.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, 2001: Severe convective storms—An overview. Severe Convective Storms, Meteor. Monogr., No. 50, 1–26, doi:10.1175/0065-9401-28.50.1.

  • Doswell, C. A., III, , Brooks H. E. , , and Maddox R. A. , 1996: Flash flood forecasting: An ingredients-based methodology. Wea. Forecasting, 11, 560581, doi:10.1175/1520-0434(1996)011<0560:FFFAIB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Ferree, J. T., , Looney J. , , and Waters K. , 2006: NOAA/National Weather Services’ storm-based warnings. Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, Amer. Meteor. Soc., 11.6. [Available online at https://ams.confex.com/ams/pdfpapers/115513.pdf.]

  • Frey, B. J., , and Dueck D. , 2007: Clustering by passing messages between data points. Science, 315, 972976, doi:10.1126/science.1136800.

    • Search Google Scholar
    • Export Citation
  • Fritsch, J. M., , and Carbone R. , 2004: Improving quantitative precipitation forecasts in the warm season: A USWRP research and development strategy. Bull. Amer. Meteor. Soc., 85, 955965, doi:10.1175/BAMS-85-7-955.

    • Search Google Scholar
    • Export Citation
  • Gourley, J. J., and Coauthors, 2013: A unified flash flood database across the United States. Bull. Amer. Meteor. Soc., 94, 799805, doi:10.1175/BAMS-D-12-00198.1.

    • Search Google Scholar
    • Export Citation
  • Hamill, T. M., , Bates G. T. , , Whitaker J. S. , , Murray D. R. , , Fiorino M. , , Galarneau T. J. Jr., , Zhu Y. , , and Lapenta W. , 2013: NOAA’s second-generation global medium-range ensemble reforecast dataset. Bull. Amer. Meteor. Soc., 94, 15531565, doi:10.1175/BAMS-D-12-00014.1.

    • Search Google Scholar
    • Export Citation
  • Hart, R. E., , and Grumm R. H. , 2001: Using normalized climatological anomalies to rank synoptic-scale events objectively. Mon. Wea. Rev., 129, 24262442, doi:10.1175/1520-0493(2001)129<2426:UNCATR>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Hough, J. R., , and White M. A. , 2003: Environmental dynamism and strategic decision-making rationality: An examination at the decision level. Strategic Manage. J., 24, 481489, doi:10.1002/smj.303.

    • Search Google Scholar
    • Export Citation
  • Houze, R. A., Jr., , Biggerstaff M. , , Rutledge S. , , and Smull B. , 1989: Interpretation of Doppler weather radar displays of midlatitude mesoscale convective systems. Bull. Amer. Meteor. Soc., 70, 608619, doi:10.1175/1520-0477(1989)070<0608:IODWRD>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Junker, N. W., , Schneider R. S. , , and Fauver S. L. , 1999: A study of heavy rainfall events during the great Midwest flood of 1993. Wea. Forecasting, 14, 701712, doi:10.1175/1520-0434(1999)014<0701:ASOHRE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Knupp, K. R., and Coauthors, 2014: Meteorological overview of the devastating 27 April 2011 tornado outbreak. Bull. Amer. Meteor. Soc., 95, 10411062, doi:10.1175/BAMS-D-11-00229.1.

    • Search Google Scholar
    • Export Citation
  • KWTV, 2015: Understanding storm shelters could save Oklahomans’ lives. News on 6. [Available online at http://www.newson6.com/story/29010877/understanding-storm-shelters-could-save-oklahomans-lives.]

  • Maddox, R. A., 1980: Mesoscale convective complexes. Bull. Amer. Meteor. Soc., 61, 13741387, doi:10.1175/1520-0477(1980)061<1374:MCC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Maddox, R. A., , Chappell C. F. , , and Hoxit L. R. , 1979: Synoptic and meso-α scale aspects of flash flood events. Bull. Amer. Meteor. Soc., 60, 115123, doi:10.1175/1520-0477-60.2.115.

    • Search Google Scholar
    • Export Citation
  • Markowski, P., , and Richardson Y. , 2010: Mesoscale Meteorology in Midlatitudes. J. Wiley and Sons, 430 pp.

  • Mercer, A. E., , Shafer C. M. , , Doswell C. A. III, , Leslie L. M. , , and Richman M. B. , 2009: Objective classification of tornadic and nontornadic severe weather outbreaks. Mon. Wea. Rev., 137, 43554368, doi:10.1175/2009MWR2897.1.

    • Search Google Scholar
    • Export Citation
  • Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343360, doi:10.1175/BAMS-87-3-343.

    • Search Google Scholar
    • Export Citation
  • Moore, J. T., , Glass F. H. , , Graves C. E. , , Rochette S. M. , , and Singer M. J. , 2003: The environment of warm-season elevated thunderstorms associated with heavy rainfall over the central United States. Wea. Forecasting, 18, 861878, doi:10.1175/1520-0434(2003)018<0861:TEOWET>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Morss, R. E., , Wilhelmi O. V. , , Meehl G. A. , , and Dilling L. , 2011: Improving societal outcomes of extreme weather in a changing climate: An integrated perspective. Annu. Rev. Environ. Resour., 36, 1, doi:10.1146/annurev-environ-060809-100145.

    • Search Google Scholar
    • Export Citation
  • Morss, R. E., , Demuth J. L. , , Bostrom A. , , Lazo J. K. , , and Lazrus H. , 2015: Flash flood risks and warning decisions: A mental models study of forecasters, public officials, and media broadcasters in Boulder, Colorado. Risk Anal., doi:10.1111/risa.12403, in press.

    • Search Google Scholar
    • Export Citation
  • NOAA, 2011: Value of a weather-ready nation. NOAA/NWS, 22 pp. [Available online at http://www.ppi.noaa.gov/wp-content/uploads/PPI-Weather-Econ-Stats-10-13-11.pdf.]

  • Nowotarski, C. J., , Markowski P. M. , , and Richardson Y. P. , 2011: The characteristics of numerically simulated supercell storms situated over statically stable boundary layers. Mon. Wea. Rev., 139, 31393162, doi:10.1175/MWR-D-10-05087.1.

    • Search Google Scholar
    • Export Citation
  • NWS, 2011a: Record floods of greater Nashville: Including flooding in middle Tennessee and western Kentucky, May 1–4, 2010. National Weather Service Service Assessment, 67 pp. + appendixes. [Available online at http://www.nws.noaa.gov/om/assessments/pdfs/Tenn_Flooding.pdf.]

  • NWS, 2011b: The historic tornadoes of April 2011. National Weather Service Service Assessment, 43 pp. + appendixes. [Available online at http://www.nws.noaa.gov/om/assessments/pdfs/historic_tornadoes.pdf.]

  • NWS, 2014: May 2013 Oklahoma tornadoes and flash flooding. National Weather Service Service Assessment, 42 pp. + appendixes. [Available online at http://www.nws.noaa.gov/om/assessments/pdfs/13oklahoma_tornadoes.pdf.]

  • Parker, M. D., , and Johnson R. H. , 2000: Organizational modes of midlatitude mesoscale convective systems. Mon. Wea. Rev., 128, 34133436, doi:10.1175/1520-0493(2001)129<3413:OMOMMC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Pedregosa, F., and Coauthors, 2011: Scikit-learn: Machine learning in python. J. Mach. Learn. Res., 12, 28252830.

  • Peters, J. M., , and Schumacher R. S. , 2014: Objective categorization of heavy-rain-producing MCS synoptic types by rotated principal component analysis. Mon. Wea. Rev., 142, 17161737, doi:10.1175/MWR-D-13-00295.1.

    • Search Google Scholar
    • Export Citation
  • Peters, J. M., , and Schumacher R. S. , 2015: Mechanisms for organization and echo training in a flash-flood-producing mesoscale convective system. Mon. Wea. Rev., 143, 10581095, doi:10.1175/MWR-D-14-00070.1.

    • Search Google Scholar
    • Export Citation
  • Robock, A., , Vinnikov K. Y. , , Srinivasan G. , , Entin J. K. , , Hollinger S. E. , , Speranskaya N. A. , , Liu S. , , and Namkhai A. , 2000: The Global Soil Moisture Data Bank. Bull. Amer. Meteor. Soc., 81, 12811299, doi:10.1175/1520-0477(2000)081<1281:TGSMDB>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rogash, J. A., , and Smith R. D. , 2000: Multiscale overview of a violent tornado outbreak with attendant flash flooding. Wea. Forecasting, 15, 416431, doi:10.1175/1520-0434(2000)015<0416:MOOAVT>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rogash, J. A., , and Racy J. , 2002: Some meteorological characteristics of significant tornado events occurring in proximity to flash flooding. Wea. Forecasting, 17, 155159, doi:10.1175/1520-0434(2002)017<0155:SMCOST>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Schumacher, R. S., , and Johnson R. H. , 2005: Organization and environmental properties of extreme-rain-producing mesoscale convective systems. Mon. Wea. Rev., 133, 961976, doi:10.1175/MWR2899.1.

    • Search Google Scholar
    • Export Citation
  • Schwartz, B. E., , Chappell C. F. , , Togstad W. E. , , and Zhong X.-P. , 1990: The Minneapolis flash flood: Meteorological analysis and operational response. Wea. Forecasting, 5, 321, doi:10.1175/1520-0434(1990)005<0003:TMFFMA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, J. A., , Baeck M. L. , , Zhang Y. , , and Doswell C. A. III, 2001: Extreme rainfall and flooding from supercell thunderstorms. J. Hydrometeor., 2, 469489, doi:10.1175/1525-7541(2001)002<0469:ERAFFS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sutter, D., , and Erickson S. , 2010: The time cost of tornado warnings and the savings with storm-based warnings. Wea. Climate Soc., 2, 103112, doi:10.1175/2009WCAS1011.1.

    • 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
  • Trier, S., , Davis C. , , and Ahijevych D. , 2010: Environmental controls on the simulated diurnal cycle of warm-season precipitation in the continental United States. J. Atmos. Sci., 67, 10661090, doi:10.1175/2009JAS3247.1.

    • Search Google Scholar
    • Export Citation
  • Waters, K., and Coauthors, 2005: Polygon weather warnings: A new approach for the national weather service. Preprints, 21st Int. Conf. on Interactive Information Processing Systems (IIPS) for Meteorology, Oceanography, and Hydrology, Phoenix, AZ, Amer. Meteor. Soc., 14.1. [Available online at https://ams.confex.com/ams/pdfpapers/86326.pdf.]

  • Weisman, M. L., , and Klemp J. B. , 1984: The structure and classification of numerically simulated convective storms in directionally varying wind shears. Mon. Wea. Rev., 112, 24792498, doi:10.1175/1520-0493(1984)112<2479:TSACON>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wurman, J., , Robinson P. , , Alexander C. , , and Richardson Y. , 2007: Low-level winds in tornadoes and potential catastrophic tornado impacts in urban areas. Bull. Amer. Meteor. Soc., 88, 3146, doi:10.1175/BAMS-88-1-31.

    • Search Google Scholar
    • Export Citation
  • Zhang, F., , Bei N. , , Rotunno R. , , Snyder C. , , and Epifanio C. C. , 2007: Mesoscale predictability of moist baroclinic waves: Convection-permitting experiments and multistage error growth dynamics. J. Atmos. Sci., 64, 35793594, doi:10.1175/JAS4028.1.

    • Search Google Scholar
    • Export Citation
Save