Baseline Climatology of Sounding Derived Parameters Associated with Deep, Moist Convection
 
 

Jeffrey P. Craven*

NOAA/NWS   Jackson, Mississippi
 

Harold E. Brooks

NOAA/National Severe Storms Laboratory, Norman, Oklahoma
 
  

 (Accepted as a National Weather Digest Article, April 2004)
     (Revised June 2004)
 
   
  
*Corresponding author address: Jeffrey P. Craven, NOAA/NWS Jackson MS, 234 Weather Service Drive, Jackson, MS 39232.  E-mail: jeffrey.craven@noaa.gov



          Abstract

     A baseline climatology of several parameters commonly used to forecast deep moist, convection is developed using an extensive sample of upper air observations.  Previous climatologies often contain a limited number of cases or do not include null cases, which limit their forecast utility.  Three years of evening (0000 UTC) rawinsonde data, approximately 60,000 soundings, from the lower 48 United States are evaluated.  Cloud-to-ground lightning data and severe weather reports from Storm Data are used to categorize soundings as representative of conditions for no thunder, general thunder, severe, significant hail/wind, or significant tornado.

     Among the detailed calculations are comparisons between both convective available potential energy (CAPE) and lifted condensation level (LCL) using a most unstable versus a mean lifted 100 mb parcel.    Lapse rates for several different layers are inspected to determine the utility of using static stability versus CAPE to forecast storm severity.  Lastly, low level shear is studied in an attempt to distinguish between severe and significant tornado episodes.

     One of the major findings is a considerable difference between 0-1 km above ground level (AGL) magnitude of vector difference of wind for significant tornado episodes versus the other five categories.  Statistically significant differences are also noted between LCL/mean lifted LCL (MLLCL) heights AGL for significant tornado events and the other convective categories.  In addition, much less seasonal variation is found for 0-1 km shear, 0-6 km shear, and MLLCL heights AGL for significant tornado events compared with the remainder of the data set.



          1. Introduction

      Meteorologists at the Storm Prediction Center (SPC) routinely prepare forecasts of severe thunderstorm potential for the lower 48 states.    Since 1999, SPC has been issuing probabilistic forecasts of tornadoes, damaging winds, and large hail.  In addition, probabilistic forecasts of significant severe weather (i.e. F2+ tornadoes, 65 + knot {120 km h-1} wind gusts, or 2+ inch {5 cm} hail) are composed.  Over the past several years, the availability of gridded model output has made access to explicit forecast parameters such as vertical wind shear and lapse rates possible.   This has allowed forecasters to finally use techniques developed over half a century ago in a real-time operational setting (i.e. Showalter and Fulks, 1943).  Surface to 6 km above ground level (AGL) magnitude of vector difference of wind (hereafter 0-6 km shear) and 700-500 mb lapse rates are used frequently in assessing severe potential, particularly for the prediction of supercells.

     The purpose of this study is to use rawinsonde data to examine several parameters commonly used to forecast severe thunderstorms and tornadoes.  The research complements work by Rasmussen (2003) and Rasmussen and Blanchard (1998), but includes a much larger data set (an order of magnitude larger), nulls cases, and does not attempt to determine convective mode.

          Lapse Rates

     Recent work relating instability to tornado occurrence has focused on convective available potential energy (CAPE - Moncrieff and Miller 1976), with less research devoted to the effects of lapse rates on severe storm/tornado formation.  However, past research has been compiled which studied the effects of elevated mixed layers (steep middle level lapse rates, e.g. 700-500 mb layer) and the associated capping inversion (or lid) on deep moist convection and severe thunderstorm formation.  Although there is no standard definition of what a “steep” lapse rate is, we will arbitrarily classify any lapse rate exceeding 7oC km-1 as steep.  This is derived from values found useful in operational severe thunderstorm forecasting at the SPC.  Carlson et al. (1983) discussed a conceptual model of how the capping inversion associated with the elevated mixed layer focuses the location and even enhances the intensity of severe local storms.  The capping inversion prevents convection from developing in areas of high CAPE, allowing the boundary layer to moisten further and permit the build up of additional potential instability.
 Doswell et al. (1985) discussed the importance of steep 700-500 mb lapse rates for both the creation of strong conditional instability and for enhancing the atmospheric response to quasi-geostrophic forcing.  The superposition of steep lapse rates and low level moisture was shown to be ideal for severe storm/tornado formation.

     Lanicci (1985) and Lanicci and Warner (1991a, b, c) studied the elevated mixed layer over the southern and central Great Plains and the importance of the capping inversion for severe thunderstorm climatology.  Since the capping inversion is normally located between 850-700 mb, steep 700-500 mb lapse rates are typically associated with the elevated mixed layer.  Therefore, this parameter is useful in tracking elevated mixed layer air and capping inversions that have originated over the higher terrain of the western U.S. or northern Mexico.   The importance of mountainous terrain on the creation of steep 700-500 mb lapse rates was shown by Cortinas and Doswell (1998).  A minimum in static stability in the 700-500 mb layer (and thus a maximum in 700-500 mb lapse rate) was found over the Rocky Mountains during much of the year.  Although the elevated mixed layer is more common over the central and southern Great Plains, Farrell and Carlson (1989) found that it played an important role during the major tornado outbreak on 31 May 1985 in Ohio and Pennsylvania.

     Steep lapse rates were found to be associated with most major tornado outbreaks by Craven (2000).  The 700-500 mb lapse rate was greater than or equal to 7oC km-1 during eighty (80%) of tornado outbreaks from 1950 to 1998 that obtained a Destruction Potential Index (DPI) of 100 or more (Thompson and Vescio, 1998).   For reference, a typical moist adiabatic lapse rate is ~ 5.5oC km-1, the standard atmosphere ~ 6.5oC km-1 from 0-6 km AGL, while a dry adiabatic lapse rate is 9.8oC km-1.   The DPI is calculated using the product of the tornado path area (path length multiplied by maximum path width) and the F-scale (F-scale added to 1 so that F0 tornadoes can be assigned a non-zero number).

          Lifted Condensation Level

     Recent research indicates a relationship between tornadic supercells and relatively high boundary layer relative humidity, which can be represented by low lifted condensation levels (LCL).  Rasmussen and Blanchard (1998) found that the parameter that showed the most utility for discriminating between significant tornadoes and supercells with either weak or no tornadoes was the height of the  LCL.  The median LCL height was ~ 500 m lower for the strong or violent tornado cases.   Nearly identical results were found by Edwards and Thompson (2000), with a mean difference in LCL height for significant tornadic versus weak or non-tornadic supercells of ~ 500 m.  During severe weather episodes in the north central United States, Johns et al. (2000) compared the median LCL height near the location of the first intense tornado versus the median LCL height 100 statute miles into the warm sector.  This work highlighted that the median LCL height in the warm sector was nearly 800 m higher than in the area where the tornadoes occurred.

          Vertical Wind Shear

     Much work has been completed that relates deep layer shear to the potential for supercell formation (see Table 1 for a summary and definitions of each shear parameter).   Weisman and Klemp (1984, 1986) and Weisman (1996) performed extensive storm scale modeling which indicates that “shear” of 20 m s-1 over the lowest 4-6 km AGL is sufficient to promote supercell storm formation.   Davies and Johns (1993) calculated the Bulk Richardson Number shear for 260 strong and violent tornadoes and found that the median was  ~ 22 m2 s-2.  Using a years worth of soundings from 1992, Rasmussen and Blanchard (1998) created a climatology of supercell/tornado parameters and found that the median of boundary layer to 6 km shear for supercells was 19  m s-1.   Their results also indicated that there is little difference in this deep layer shear parameter between supercells containing significant (strong or violent) tornadoes (18 m s-1) and those that do not.  A well defined lower threshold of 20 m s-1 in 0-6 km shear was found for a data set of 260 right-moving supercells by Bunkers et al. (2000).  In a study of 65 major tornado outbreaks from 1950-1998, Craven (2000) found that virtually all of the events (97%) were associated with surface to 6 km shear values greater than or equal to 20 m s-1.



Table 1.   Summary of measures of wind shear from previous research

Author                                     Depth                                                      Name/Definition                         Units

Marwitz (1972a,b)                    Surface to 4 km MSL                            Subcloud layer shear                   s-1

Doswell and Lemon                0 - 1524 m AGL                                      Layer average vector shear         s-1
(1979)

Weisman and Klemp              0-5 km AGL shear                                  Magnitude of vector                    m s-1
(1984)                                                                                                         difference

Weisman and Klemp              0-6 km AGL shear                                   Magnitude of vector                    m s-1
(1986)                                                                                                          difference

Davies and Johns (1993)       0-2 km AGL positive                               Hodograph length divided         s-1
Johns and Hart (1993)           mean shear                                                by depth of layer, setting the
                                                                                                                     shear magnitude to zero for
                                                                                                                     those hodograph segments
                                                                                                                     where the ground relative
                                                                                                                     winds back“significantly”
                                                                                                                     with height

Davies and Johns (1993)       0-6 km AGL BRN shear                          magnitude of vector                     (m s-1)2
                                                                                                                    difference between
                                                                                                                     0-500 m AGL mean wind
                                                                                                                    and the 0-6 km AGL
                                                                                                                    mean wind

Rasmussen and Blanchard   Boundary layer to                                  Magnitude of vector                      m s-1
(1998)                                       6 km AGL shear                                      difference between
                                                                                                                    0-500 m AGL mean
                                                                                                                   wind and 6km AGL wind

Craven (2000)                          0-6 km AGL shear                                  Magnitude of vector                       m s-1
Bunkers et al (2000)                                                                                 difference
--------------------------------------------------------------------------------------------------------------------------------------------
Note: “magnitude of vector difference” refers to the difference between the surface wind and the wind at the top of the layer in question.


     These findings have identified a rather simple way to determine the kinematic potential for supercells versus non-supercell thunderstorms from a deep layer magnitude of vector difference of wind.  However, the question of tornado potential appears to be more closely related to low level shear or storm relative helicity.  Davies-Jones et al. (1990) studied 28 tornadoes of various strengths, and found that 0-3 km Storm Relative Helicity (SRH) generally
increases as the intensity of the tornadoes increases.  The Rasmussen and Blanchard (1998) results indicated a statistically significant difference between 0-3 km SRH for ordinary thunderstorms versus supercells producing significant tornadoes.  The SRH values were calculated using observed storm motion in each of these studies.  However, considerable overlap is noted for supercells that did not produce strong and violent tornadoes, indicating a potential false alarm problem.

     Marwitz (1972a) found that the mean subcloud environmental winds for supercells producing hailstorms are strong (greater than 10 m s-1) and veer by more than 60o from the mean  environment winds (and also veer greater than 50o within the subcloud layer), suggesting strong low level shear.  Complementary research of non-supercell storms (Marwitz, 1972b) concluded that the distinguishing characteristic of the environment which produces non-supercell storms (versus supercells) is light winds (and thus weaker low level shear) in the subcloud layer.   In a study of 21 cases of severe thunderstorms including supercells, Doswell and Lemon (1979) found that the most reliable kinematic parameter is the low level shear (surface to 5000 feet AGL {~1500 m AGL}), which seems to be well related to the low-level mean wind speed and to the region of severe convection.  Johns et al. (1990) studied 0-2 km positive shear with a data set of 242 strong and violent tornadoes.  Results indicate that the majority of the tornadoes were associated with low level (0-2 km) positive shear values in excess of 10  x 10 -3 s-1.  A study of severe weather outbreaks involving bow echoes versus those with supercells was completed by Johns and Hart (1993).  In their small sample of cases, they found that tornado outbreaks were associated with 0-3 km SRH in excess of 400 m2 s-2, while the 0-3 km SRH during the bow echo cases was less than 120 m2 s-2.   The 0-2 km positive shear in the supercell outbreaks was also nearly double that found in the bow echo events.  Finally, Edwards and Thompson (2000) used 51  proximity soundings generated by the RUC-2 model to study several forecast supercell parameters.  They found a statistically significant difference between the mean 0-1 km SRH for supercells with significant tornadoes (~150 m2 s-2) versus supercells with either weak or no tornadoes observed (~100 m2 s-2).

          2. Data

     Proximity Criteria

     0000 UTC rawinsonde soundings from 1997-1999 for the lower 48 states are collected.  A total of 60,090 soundings are included.  Proximity is defined as being within 100 nm (185 km) of the sounding release location, and during the period from 2100 UTC to 0300 UTC (6 hour period centered on the 0000 UTC sounding).  The 185 km threshold lies within the range of 80 km (Darkow, 1969; Schaefer and Livingston, 1988; Brooks et al., 1994) and 400 km criteria utilized by Rasmussen and Blanchard (1998).  For a detailed discussion on the difficulty of defining and selecting a proximity sounding, see Brooks et al. (1994).

     Events

     Lightning Data from Global Atmospherics, Inc., (Orville, 1991) and convective severe weather reports (Storm Data, 1997-1999; Hart and Janish, 1999) are utilized to subdivide the data set into 6 categories (Table 2).  Of the more than 60,000 possible events, 32,141 (53%) had non-zero most unstable parcel in lowest 300 mb CAPE (MUCAPE).  Of the 45,508 no thunder events, 27,949 (61%) had no MUCAPE and 17,559 (39%) had non-zero MUCAPE.  The categories are exclusive, and each event was assigned using the most severe report (i.e. a F2 tornado event was only assigned to significant tornadoes, even if 1 inch hail also occurred).

     The lightning strike threshold of 2 or more cloud to ground (CG) strikes is consistent with the criteria established by Reap (1986) and Orvile (2001-personal communication), similar to the 3 or more CG strike threshold used by Hamill and Church (2000), but much less than the 10 or more CG strike criteria used by Rasmussen and Blanchard (1998).



Table 2.  Definitions and number of proximity soundings for the six convective categories

Number                 Category                                          Definition

27949                    No Thunder (noCAPE)                  0-1 CG strikes (and zero MUCAPE)

17559                    No Thunder (CAPE)                       0-1 CG strikes (and non-zero MUCAPE)

11339                    General Thunder                             > 2 CG strikes

2644                      Severe                                               0.75-1.99" hail
                                                                           and/or  50-64 knot gust
                                                                           and/or  wind damage
                                                                           and/or  F0 or F1 tornado

512                         Significant Hail/Wind                    >2.00" hail
                                                                            and/or  >65 knot gust

87                          Significant tornado                         F2-F5 tornado


     Quality Control

 No attempt was made to modify the soundings.  It was anticipated that the effects of unrepresentative, contaminated, or erroneous data would be damped out in the statistical analysis.  A simple objective quality control procedure for the severe, significant hail/wind, and significant tornado soundings removed all soundings with MUCAPE less than 150 J kg-1 (Brooks et al., 1994).  General thunder soundings were removed if no MUCAPE was present.  All CAPE values were calculated using the virtual temperature correction (Doswell and Rasmussen, 1994).
 Subjective quality control was minimal because of the size of the data set.  Lapse Rates greater than 11oC km-1 in the 0-3 km AGL layer and 10.2oC km-1 in the 0-6 km AGL layer, 850-700 mb layer, and 700-500 mb layer were removed.  0-1 km shear greater than 50 m s-1 and 0-6 km shear greater than 100 m s-1 were also excluded.  In addition, all soundings with MUCAPE and/or 100 mb mean layer CAPE (MLCAPE) greater than 5000 J kg-1 were manually inspected and suspect soundings were excluded.

     Parameters

     A list of the parameters computed from the sounding data set is shown in Table 3.  These parameters cover three main groups a) instability/lapse rates, b) LCL heights, and c) vertical wind shear.

Table 3.  Parameters computed from soundings

Parameter                                                                                                                               Units

MUCAPE (most unstable parcel CAPE in lowest 300 mb)                                              J kg-1
MUCIN    (most unstable parcel Convective Inhibition {CIN} in lowest 300 mb)       J kg-1
MLCAPE (100 mb mean layer CAPE)                                                                                  J kg-1
MLCIN    (100 mb mean layer CIN)                                                                                      J kg-1
0-3 km AGL Lapse Rate                                                                                                        oC km-1
3-6 km AGL Lapse Rate                                                                                                        oC km-1
700-500 mb Lapse Rate                                                                                                         oC km-1
850-700 mb Lapse Rate                                                                                                         oC km-1
DCAPE    (Downdraft CAPE)                                                                                                J kg-1
LCL height (lifted condensation level)                                                                                m AGL
MLLCL     (100 mb mean layer LCL height)                                                                         m AGL
0-1 km shear (magnitude of vector difference)                                                                   m s-1
0-6 km shear (magnitude of vector difference)                                                                   m s-1


          3. Results

     Box and whisker plots (Tukey 1977) are used extensively to compare data in each category.  On a single graphic, these plots show information about range, variance, and median values.  The plot shows the 10th (bottom whisker), 25th (bottom of box), 50th (horizontal line within box), 75th (top of box), and 90th percentiles (top whisker) of the particular data.  For example, the 25th percentile, or bottom of the box, indicates that 75 percent of the data is larger than the particular value.  For example, in Figure 1, 75% (bottom of box, or 25th percentile) of all Significant Tornadoes have a MLCAPE value of slightly more than 500 J kg -1.   Comparing box and whisker plots in different categories yields information about the similarity of the data.  For example, in Figure 6, the 75th percentile of Significant Hail/Wind events is less than the 25th percentile of the Significant Tornado events (the boxes don’t overlap).  This lack of overlap suggests a statistically significant difference between the data.

 

 

    

Instability/Lapse Rates

     Recent research suggests that the most accurate estimate of convective cloud base from 0000 UTC rawinsonde data and thus the most accurate representation of parcel path utilizes a mean layer parcel, say from the mean temperature and dewpoint in the lowest 100 mb (Craven et al. 2002).  Thus, MLCAPE was chosen to compare potential instability for the soundings in this database.  Although the median value of MLCAPE tends to increase with increasing intensity of deep convection, there was considerable overlap in the distributions (Fig. 1).  When instability was present, 75 percent of the No Thunder events had less than 250 J kg-1, while more than 50 percent of Thunder soundings had more MLCAPE.  Likewise, 75 percent of the Thunder soundings had less than 1100 J kg-1, while more than 50 percent of Significant Hail/Wind and Significant Tornado events had more MLCAPE.

     Low level lapse rates had a different signal than MLCAPE (Fig. 2).  The 0-3 km AGL layer displayed little difference between thunder, severe, and significant hail/wind with medians near 7.5oC km-1.  However, low level lapse rates were much smaller for significant tornado events, with 75 percent of those events occurring  with values less than 7.5oC km-1.  It is interesting to note that the Significant Tornado distribution looks much like that of the No Thunder (CAPE) distribution.  More evidence for the reasons behind this will be presented later.  However, it is likely related to a moist boundary layer, which reduces the degree of mixing and results in a shallower boundary layer (if strong moisture flux convergence is not occurring).  In addition, capping inversions associated with the elevated mixed layer in the Plains are often associated with tornado events.  The presence of the capping inversion in the 850-700 mb layer would result in smaller 0-3 km AGL lapse rates.

 

     Mid level lapse rates suggest that significant severe weather episodes tend to have steeper values than the rest of the data set (Fig. 3).  While most of the categories show similar medians below 6.5oC km-1, the significant hail/wind and significant tornado events tend to occur when 700-500 mb lapse rates are above 6.5oC km-1.   However, considerable overlap does exist.

 

 

      Downdraft CAPE (DCAPE; Gilmore and Wicker 1998) in this study was calculated by taking the minimum wet bulb temperature in the 700-500 mb layer pseudo-adiabatically to the surface without entrainment.  The area between this line and the ambient temperature is the DCAPE.  Thus, DCAPE is maximized by a combination of steep lapse rates below 700 mb and a very dry layer between 700 and 500 mb.  There was a tendency for DCAPE values to increase during the progression from thunder to significant hail/wind, with median values increasing from 600 J kg-1  to over 900 J  kg-1 (Fig 4).  Although there was considerable overlap, the significant tornado events tend to occur with lower values than significant hail/wind events.  A possible explanation is that higher DCAPE values permit stronger rear flank downdrafts, which could result in an outflow dominated supercell storm that undercuts the mesocyclone and inhibits strong tornadogenesis.  Much like low level lapse rates, the distributions of DCAPE between Significant Tornado events and No Thunder (CAPE) events are quite similar.  Since DCAPE values are proportional to low level lapse rates (the steeper the low level lapse rates, the higher the DCAPE values), these results are consistent with the results from the 0-3 km AGL Lapse Rates data set.

 

 

     Cloud Bases/LCL heights

     The 100 mb mean layer LCL (MLLCL) height AGL shows little difference between most events with median values above 1200 m AGL (Fig. 5).  However, cloud bases tended to be lower during significant tornado events, with 75 percent of the cases containing MLLCL heights less than 1200 m AGL.  The median values of the cloud bases in this category were about 500 m less than the rest of the data set, which is consistent with earlier research (Rasmussen and Blanchard 1998; Edwards and Thompson 2000; Johns et al. 2000; Markowski et al. 2000).  The lower cloud bases likely indicate that less sub cloud evaporation will take place, decreasing the chance that the storm will be dominated by cold outflow that will undercut the mesocyclone.  This is probably also related to smaller surface to 3 km AGL lapse rates (Fig. 2).  The drier the boundary layer is, the deeper the mixed layer can become and the stronger the surface to 3 km AGL lapse rate will likely be (again, if strong moisture flux convergence is not occurring).  Thus, a moist boundary layer with associated low MLLCL heights would also be associated with somewhat smaller low level lapse rates.  In addition, capping inversions are often found between 850 and 700 mb during strong tornado events.  This capping inversion, often the result of an elevated mixed layer from upstream higher terrain (e.g. the Rocky Mountains), would result in smaller surface to 3 km AGL lapse rates.

 

     Vertical Wind Shear

     The most striking results of the study involved the low level shear (Fig 6.).  There was little difference in 0-1 km shear for the first five categories.  However, a very substantial difference is evident between significant tornado events and the rest of the data set.  There is no overlap between the upper quartile of significant hail/wind and the lower quartile of significant tornadoes.   Nearly 75 percent of significant tornado events occurred with values in excess of 10 m s-1.  In contrast, more than 75 percent of the significant hail/wind events had less low level shear.  Thus, much like the lower threshold that has been established for deep layer shear and supercell development (i.e. 20 m s-1; Weisman and Klemp 1982, Davies and Johns 1993; Rasmussen and Blanchard 1998: Bunkers et al. 2000; Craven 2000), it appears than 10 m s-1 (20 kts) may be used as a lower threshold for significant tornadoes events.  Stronger low level shear appears to be associated with a higher frequency of strong and violent tornado events.   These results are consistent with Edwards and Thompson (2000), who found a substantial difference between the mean 0-1 km SRH for supercells with significant tornadoes versus supercells with either week or no tornadoes observed.  While SRH requires an estimated or observed storm motion, using a 0-1 km shear vector does not.

 

 

     The No Thunder (No CAPE) soundings represent almost half of the data set, and are dominated by cold season situations where strong horizontal temperature gradients result in large thermal winds (Fig. 7).  Since vertical wind shear is proportional to the strength of the thermal wind, the 0-6 km AGL shear values can be quite high.  However, this fact is somewhat irrelevant since the lack of instability precludes development of deep convection.  In addition, notice that the distributions of No Thunder (CAPE) and Severe are very similar.   Recall that there was a substantial difference between the MLCAPE for these categories (Fig. 1), with the 25th and 75th percentiles barely overlapping around 250 J kg -1.  Thus, it is possible that updrafts in the No Thunder (CAPE) environment may have trouble sustaining themselves given relatively high vertical wind shear and low potential instability.

 

     Otherwise, deep layer shear appears to increase with increasing severity of deep convection. The 0-6 km shear increases during the progression from thunder events to significant tornado events.   Although there is substantial overlap between severe events and significant hail/wind events, there was no overlap between the upper quartile of severe events and the lower quartile of significant tornado cases.  The expected lower threshold for supercells of 18 to 20 m s-1  is evident in the significant tornado events.  About 75 percent of severe events occur with at least 10 m s-1 of deep layer shear, while almost half of the thunder events had less shear.  Thus, there appears to be some value in 10 m s-1 as a possible lower threshold for severe versus thunder forecasts, although considerable overlap exists in the data set.

     Seasonal Variations

     Due to small sample size, dividing the Significant Hail/Wind and Significant Tornado groups further into seasonal groups and using box and whisker plots was problematic.  Given the degree of overlap within groups using the entire 3 year dataset, showing only the median values to indicate seasonal variation is potentially misleading, especially if one would like to determine useful forecast thresholds.  Thus, the authors advise that caution be used when comparing categories using the median values alone.  The purpose of these figures is to indicate the seasonal variation within each group and show the apparent lack of seasonal variation in a few of the parameters for the Significant Tornado events.

     The data were also partitioned into six two month groups to account for seasonal variability.  The median values of MLCAPE indicated the expected result of higher values during the warm season and lower values during the cold season (Fig. 8).  The higher the category, the higher the median value of MLCAPE tends to be.

 

     The median cloud base heights indicated the opposite annual trend (Fig. 9).  MLLCL heights AGL were higher during the warm season due to deeper mixing.  However, minimal seasonal variation was observed in the significant tornado events, where median values tend to remain below 1000 m AGL.  The other categories increase 400 to 600 meters from the cold season to the warm season.  The discrimination between categories is lost during the cold season, when all median values are near 800 m AGL from November to February.  However, the difference between significant tornadoes and other events increases to 300 to 500 meters during the warm season.

 

 

     Similar to the MLLCL heights, there was little seasonal variability in either 0-1 km shear or 0-6 km shear in the data set for significant tornado cases (Figs. 10, 11).  While the other five categories displayed substantial decrease during the warm season, both low level and deep layer shear remain well above the 10 m s-1 and 20 m s-1 thresholds for significant tornadoes and supercells throughout the year.

 

          4. Parameter Combinations

     0-1 km shear versus MLLCL height

     Examining low level shear and MLLCL height yields a strong signal between significant tornadoes and significant hail/wind events (Fig. 12).  Strong/violent tornadoes tend to occur with relatively high 0-1 km shear (e.g. > 10 m s-1) and relatively low MLLCL height (e.g. < 1200 m AGL).  Storms that produce hail greater than or equal to 2 inches and/or wind gusts greater than or equal to 65 knots but no strong/violent tornadoes tend to have weaker low level shear and higher cloud bases.

 

 

     Significant Severe Parameter

     In general, individual parameters did not discriminate well between Thunder and Severe events.  However, when considering both instability and shear (Davies and Johns 1993; Johns et al. 1993) simultaneously, the results showed a noticeable improvement.  Calculating the product of MLCAPE and 0-6 km shear (defined as Significant Severe Parameter in m3 s -3) yielded a small overlap between upper quartile of Thunder events and the lower quartile of the three severe categories, especially for the significant hail/wind and tornado events (Fig. 13).  Possible lower thresholds of 10,000 m3 s-3 for severe, 20,000 m3 s-3 for significant hail/wind, and 30,000 m3 s-3 for significant tornadoes may be used given the distributions of this instability/shear parameter (see Appendix 1).

     Strong Tornado Parameter

     Five of the individual parameters studied showed some promise in discriminating between significant tornado events and other categories.  The following combination of parameters was examined to see if a parameter could be assembled that would assist in diagnosing the potential for strong/violent tornadoes:



                                                                MLCAPE * 0-1 km shear * 0-6 km shear
Strong Tornado Parameter =        ___________________________________
(Units of m s-2)
                                                                           MLLCL height * DCAPE

 


     This is similar to the Significant Tornado Parameter (Thompson et al. 2003), but uses 0-1 km shear rather than 0-1 km SRH.  Thus, an observed or estimated storm motion is not required.  In addition, the Strong Tornado Parameter includes DCAPE.  Essentially, a combination of high instability and strong vertical wind shear along with low cloud bases and low probability of strong/cold downdrafts will increase the probability of significant tornadogenesis (Fig. 14).  The results clearly indicate a lower threshold of about 0.25 m s-2, with little overlap between the upper quartile of significant hail/wind events and the lower quartile of significant tornado cases (see Appendix 1).

 

          5. Summary

     Inspection of a large data base of soundings from the CONUS from 1997-1999 yielded the following results:

1) MLCAPE discriminates somewhat between No Thunder and Thunder soundings, but there is considerable overlap between Thunder and the three severe categories.

2) Out of about a dozen parameter combinations, the Significant Severe Parameter (product of MLCAPE and 0-6 km shear) appeared to show some discrimination between Thunder and Severe events.

3) 0-1 km shear and MLLCL height both discriminate well between significant tornado events and other severe events.  A combination of the two parameters shows even more skill in distinguishing between categories.

4) There is minimal seasonal variation in 0-1 km shear and MLLCL height for significant tornadoes.  Considerable seasonal variation is noted in the other five categories.  In addition, these parameters are better at discriminating during the warm season.

          6. Acknowledgements

     This paper summarizes a research project completed to partially fulfill the requirements of a M.S. degree in Meteorology from the School of Meteorology, University of Oklahoma.  John Hart’s special programming skills made this project and many others possible.   Joseph Schaefer served as a committee member and reviewer of the manuscript, and provided numerous suggestions for enhancing the paper.  Thanks to committee member Frederick Carr and John Ferree for their reviews.  Special thanks to Richard Thompson, Steve Weiss, and Jeff Evans for many valuable discussions about the data set.

          7. Future Work

     The data set of significant severe events was relatively small in the present study.  Additional 0000 UTC soundings from 1957 to 1996 are being compiled and examined to test the results of the three year study against a much larger data set.  Subsets of the data to determine regional variability are also under inspection.

          8. Authors

     Jeff Craven is the Science and Operations Officer at the National Weather Service Office in Jackson, Mississippi.  He previously has worked at the Storm Prediction Center in Norman, Oklahoma, as well as National Weather Service Offices in Elko, Nevada, Dodge City, Kansas, and Lake Charles, Louisiana.  His primary interests involve all aspects of hazardous weather, but especially thunderstorms.  He received his B.S. in Meteorology (1988) from San Jose State University in California, and his M.S. in Meteorology (2001) from the University of Oklahoma in Norman.

     Harold Brooks is Head of the Mesoscale Applications Group at the National Severe Storms Laboratory in Norman, Oklahoma.  His research involves many aspects of severe weather, but has recently focused on regional and worldwide climatologies of severe weather and especially tornadoes.  He received his B.A. in Physics/Mathematics (1982) from William Jewell College in Liberty, Missouri, and his M.A. in Atmospheric Sciences (1985) from Columbia University in New York, New York.   He received his Ph. D.  in Atmospheric Sciences (1990) from the University of Illinois at Urbana-Champaign.

 


Appendix 1.  Objective Severe Weather Detection

Given Thunder, Frequency of Occurrence (%)   Cape * Shear

SSP    Thunder    All SVR    Svr    Sig    S igTorn     %Total obs     Total

10000    62                38            29        8        1                    13                4147
20000    53                47            43        12      3                    6                  1981
30000    46                54            35        15      4                    3                  1053
50000    41                59            32        21      6                    1                    309
--------------------------------------------------------------------------------------------------------------------
Statistics for detection of any severe event given a sounding with following thresholds

SSP   POD  FAR  BIAS  CSI  HSS

10000   0.48  0.76  1.97  0.194  0.273
20000   0.30  0.70  0.97  0.172  0.255
30000   0.17  0.66  0.51  0.130  0.201
50000   0.06  0.64  0.15  0.028  0.084
--------------------------------------------------------------------------------------------------------------------
Given Thunder, Frequency of Occurrence    (%) Strong Tornado Parameter

STP     Thunder    All SVR     Svr    Sig     SigTorn     %Total obs     Total

0.25           56               44            31       9           4                    2                1301
0.50           52               48            32       9           7                    1                  596
0.75           50               50            33       9           8                 0.5                  314
1.00           52               48            30     10           8                 0.3                  196
--------------------------------------------------------------------------------------------------------------------
Statistics for detection of Strong/Violent Tornadoes given a sounding with following thresholds

STP   POD  FAR  BIAS   CSI     HSS

0.25   0.60    0.96   35.7     0.020  0.072
0.50   0.45    0.93     6.85   0.061  0.112
0.75   0.29    0.92     3.61   0.066  0.123
1.00   0.17    0.92     2.25   0.056  0.104
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Legend: SSP - Significant Severe Parameter (m3 s-3); STP - Strong Tornado Parameter (m s-2); POD - Probability of Detection; FAR - False Alarm Ration; CSI - Critical Success Index; HSS - Heidke Skill Score


     REFERENCES

Brooks, H. E., C. A. Doswell III, and J. Cooper, 1994: On the environments of tornadic and nontornadic mesocyclones.  Wea. Forecasting., 9, 606-618.

Bunkers, M. J., J. W. Zeitler, R. L. Thompson, and M. L. Weisman, 2000: Predicting supercell motion using a new hodograph technique.  Wea. Forecasting, 15, 61-79.

Carlson, T. N., S. G. Benjamin, G. S. Forbes and Y. F. Li, 1983: Elevated mixed layers in the regional severe storm environment: Conceptual model and case studies.  Mon. Wea. Rev., 111, 1453-1473.

Cortinas, J. V. Jr., and C. A. Doswell III, 1998: Climatology of tropospheric static stability across the contiguous united states.  Preprints, 16th Conference on Weather Analysis and Forecasting, 11-16 January 1998, Phoenix, Amer. Meteor. Soc., 409-411.

Craven, J. P., 2000:   A preliminary look at deep layer shear and middle level lapse rates during major tornado outbreaks.   Preprints 20th Conf. on Severe Local Storms, 11-15 September 2000, Orlando, Amer. Meteor. Soc., 547-550.

________________, R. E. Jewell, and H. E. Brooks, 2002: Comparison between observed convective cloud base heights and lifting condensation level for two different lifted parcels.  Wea. Forecasting, 17, 4, 885-890.

Darkow, G. L., 1969: An analysis of over sixty tornado proximity soundings.  Preprints, 6th Conference on Severe Local Storms, Chicago, Amer. Meteor. Soc., 218-221.

Davies, J. M., and R. H. Johns, 1993: Some wind and instability parameters associated with strong and violent tornadoes  1.  Wind shear and helicity.  The Tornado: Its Structure, Dynamics, Prediction, and Hazards, Geophysical Monograph 79, American Geophysical Union.

Davies-Jones, R. P, D. Burgess, and M. Foster, 1990: Test of helicity as a tornado forecast parameter.  Preprints, 16th Conf. on Severe Local Storms, Kananaskis Park, AB, Canada, Amer. Meteor. Soc., 588-592.

Doswell, C. A. III, and L. R. Lemon, 1979: An operational evaluation of certain kinematic and thermodynamic parameters associated with severe thunderstorm environments.  Preprints 11th Conference on Severe Local Storms, 2-5 October 1979, Kansas City, Amer. Meteor. Soc., 397-402.

______________ , F. Caracena and M. Magnano, 1985: Temporal evolution of 700-500-mb lapse rate as a forecasting tool-A case study.  Preprints 14th Conference on Severe Local Storms, 29 October-1 November 1985, Indianapolis, Amer. Meteor. Soc., 398-401.

______________, and E. N. Rasmussen, 1994: The effect of neglecting the virtual temperature correction on CAPE calculations.  Wea. Forecasting, 2, 625-629.

Edwards, R., and R. L. Thompson, 2000: RUC-2 supercell proximity soundings, part II: an independent assessment of supercell forecast parameters.  Preprints 20th Conference on Severe Local Storms, 11-15 September 2000, Orlando, Amer. Meteor. Soc., 435-438.

Farrell, R. J., and T. N. Carlson, 1989: Evidence for the role of the lid and underrunning in an outbreak of tornadic thunderstorms.  Mon. Wea. Rev., 117, 857-871.

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

Hamill, T. M., and A. T. Church, 2000: Conditional probabilities of significant tornadoes from RUC-2 forecasts.  Wea. Forecasting., 15, 461-475.

Hart, J. A., and P. R. Janish., 1999: SeverePlot v2.0.  National Weather Service, National Centers for Environmental Prediction, Storm Prediction Center.

Johns, R. H, J. M. Davies, and P. W. Leftwich, 1990: An examination of the relationship of 0-2 km AGL “positive” wind shear to potential buoyant energy in strong and violent tornado situations.  Preprints, 16th Conf. on Severe Local Storms, Kananaskis Park, AB, Canada, Amer. Meteor. Soc., 593-598.

________ , __________, and ___________ , 1993: Some wind and instability parameters associated with strong and violent tornadoes.  Part II: variations in the combinations of wind and instability parameters.  The tornado: its structure, dynamics, prediction, and hazards.  Geophys. Mongr., 79, Amer. Geophys. Union, 583-590.

________ , and J. A. Hart, 1993: Differentiating between types of severe thunderstorm outbreaks: a preliminary investigation.  Preprints 17th Conference on Severe Local Storms,4-8 October, St. Louis, Amer. Meteor. Soc., 46-50.

________ , C. Broyles, D. Eastlack, H. Guerrero, and K. Harding, 2000: The role of synoptic patterns and temperature and moisture distribution in determining the locations of strong and violent tornado episodes in the north central united states: a preliminary examination, 2000:  Preprints 20th Conf. on Severe Local Storms, 11-15 September 2000, Orlando, Amer. Meteor. Soc., 489-492.

Lanicci, J. M., 1985: An operational procedure using elevated mixed-layer analyses to predict severe-storm outbreaks.  Preprints 14th Conference on Severe Local Storms, 29 October-1 November 1985, Indianapolis, Amer. Meteor. Soc., 406-409.

__________, and T. T. Warner, 1991a: A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in Spring.  Part 1: Structure, dynamics, and seasonal evolution.  Wea. Forecasting, 6, 181-197.

__________ , and _________ , 1991b: A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in Spring.  Part 2: The life cycle of the lid.  Wea. Forecasting, 6, 198-213.

__________ , and _________ , 1991c: A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in Spring.  Part 3: Relationship to Severe-Storms Climatology.  Wea. Forecasting, 6, 214-226.

Markowski, P. M., J. M. Straka, and E. N. Rasmussen, 2000: Surface thermodynamic characteristics of rear flank downdrafts as measured by a mobile mesonet.  Preprints 20th Conference on Severe Local Storms, Orlando, FL, Amer. Meteor. Soc., 251-254.

Marwitz, J. D., 1972a: The structure and motion of severe hailstorms.  part I: supercell storms.  J. Appl. Meteor., 11, 166-179.

__________ , 1972b: The structure and motion of severe hailstorms.  part II: mult-cell storms.  J. Appl. Meteor., 11, 180-188.

Moncrieff, M., and M. J. Miller, 1976: The dynamics and simulation of tropical cumulonimbus and squall lines.  Quart. J. Roy. Meteor. Soc., 102, 373-394.

Orville, R. E., 1991: Lightning ground flash density in the contiguous united states_1989.  Mon. Wea. Rev., 2, 573-577.

_________ , 2001: Personal Communication.

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

______________, 2003: Refined supercell and tornado forecast parameters.  Wea Forecasting, 18, 3, 530-535.

Reap, R. M., 1986: Evaluation of cloud-to-ground lightning data from the western united states for the 1983-1984 summer seasons.  J. Climate Appl. Meteor., 25, 785-799.

Schaefer, J. T., and R. L. Livingston, 1988: The structural characteristics of tornado proximity soundings.  Preprints 15th Conference on Severe Local Storms, 22-26 February 1988, Baltimore, Amer. Meteor. Soc., 537-540.

Showalter, A. K., and J. R. Fulks, 1943: The tornado - an analysis of antecedent meteorological conditions; notes on synoptic situation accompanying the hackleburg, alabama tornado of April 12, 1943.  U.S. Department of Commerce, Weather Bureau, Preliminary report on tornadoes, no. 1151.

Storm Data, 1997-1999: Storm data and unusual weather phenomena.  National Oceanic and Atmospheric Administration, National Environmental Satellite Data, and Information Service, National Climatic Data Center, Asheville, North Carolina.

Thompson, R.L., and M.D. Vescio, 1998: The destruction potential index-a method for comparing tornado days.  Preprints 19th Conference on Severe Local Storms, 14-18 September 1998, Minneapolis, Amer. Meteor. Soc., 280-282.

_____________, R. Edwards, J. A. Hart, K.L. Elmore, and P. Markowski, Paul. 2003: Close proximity soundings within supercell environments obtained from the Rapid Update Cycle. Wea. Forecasting, 18, 6, 1243–1261.

Tukey, J. W., 1977: Exploratory Data Analysis.  Addison-Wesley (Reading MA),  499 pp.

Weisman, M. L., and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy.  Mon. Wea. Rev., 110, 504-520.

____________ , and _______ , 1984: The structure and classification of numerically simulated convective storms in directionally varying wind shears.  Mon. Wea. Rev., 112, 2479-2498.

____________ , and _______ , 1986: Characteristics of isolated convective storms.  Mesoscale Meteorology and Forecasting, P. S. Ray, Ed., Amer. Meteor. Soc., 331-358.

____________,1996: On the use of vertical wind shear versus helicity in interpreting supercell dynamics.  Preprints 18th Conference on Severe Local Storms, 19-23 February 1996, San Francisco, Amer. Meteor. Soc., 200-204.


USA.gov is the U.S. government's official web portal to all federal, state and local government web resources and services.