Abstract

Three-dimensional composite analyses using 134 soundings from the second Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX2) reveal the nature of near-storm variability in the environments of supercell thunderstorms. Based upon the full analysis, it appears that vertical wind shear increases as one approaches a supercell within the inflow sector, providing favorable conditions for supercell maintenance (and possibly tornado formation) despite small amounts of low-level cooling near the storm. The seven analyzed tornadic supercells have a composite environment that is clearly more impressive (in terms of widely used metrics) than that of the five analyzed nontornadic supercells, including more convective available potential energy (CAPE), more vertical wind shear, higher boundary layer relative humidity, and lower tropospheric horizontal vorticity that is more streamwise in the near-storm inflow. The widely used supercell composite parameter (SCP) and significant tornado parameter (STP) summarize these differences well. Comparison of composite environments from early versus late in supercells' lifetimes reveals only subtle signs of storm-induced environmental modification, but potentially important changes associated with the evening transition toward a cooler and moister boundary layer with enhanced low-level vertical shear. Finally, although this study focused primarily on the composite inflow environment, it is intriguing that the outflows sampled by VORTEX2 soundings were surprisingly shallow (generally ≤500 m deep) and retained considerable CAPE (generally ≥1000 J kg−1). The numerous VORTEX2 near-storm soundings provide an unprecedented observational view of supercell–environment interactions, and the analyses are ripe for use in a variety of future studies.

1. Introduction

Supercell thunderstorms have considerable societal impact through their propensity to produce tornadoes as well as significant severe hail, winds, and heavy precipitation. Based upon a wide variety of studies, it has increasingly become clear that the lower-tropospheric profiles of temperature, humidity, and winds are important to supercells' formation, maintenance, and tornado production [e.g., the climatologies of Rasmussen and Blanchard (1998), Markowski et al. (2003), and Thompson et al. (2003, 2012)]. Accurate characterization of these important fields is challenging because actual measurements are rarely made near active supercells, particularly above the surface. Toward this end, one of the key objectives of the second Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX2; Wurman et al. 2012) was to understand “relationships between supercell storms and their environments”.1

What few prior observations we have suggest that there is likely a large degree of spatial and temporal variability associated with the “environment” near supercells. For example, Markowski et al. (1998b) used a network of soundings from the first VORTEX field campaign (in 1994–95) to show that storm-relative helicity varies regionally (lengths ~100 km and intervals <3 h) on many tornado outbreak days, especially in the vicinity of preexisting mesoscale boundaries. This is troublesome given that many supercell process studies use numerical models with homogeneous initial conditions represented by a single preconvective sounding [e.g., as reviewed by Letkewicz et al. (2013)]. In addition, even within such idealized models, substantial near-storm environmental modifications may be attributable to local, storm-induced perturbations. For example, Brooks et al. (1994) used a simulation to demonstrate that near-storm values of CAPE and helicity might vary by as much as a factor of 2 across spans of <10 km. The true nature of such supercell-induced near-storm variability (especially that linked to temperature and humidity changes above the surface) has not yet been fully constrained by observations.

This observational gap exists largely because it is quite rare for multiple near-storm upper air soundings to be launched simultaneously from different storm-relative positions. As reviewed by Potvin et al. (2010), many historical studies of storm environments have selected one proximity sounding for each case, and then performed statistical analysis of those soundings without concern for each sounding's unique distance from the storm or time separation from key events during the storm's lifetime (e.g., tornado formation). Such a simplification is the understandable result of the rather sparse operational sounding network (standard soundings are made only at 0000 and 1200 UTC at roughly 75 locations in the contiguous United States).

Analysis soundings from models such as the Rapid Update Cycle (RUC; Benjamin et al. 2004) have been invaluable in filling these routine observational gaps, and have been the basis for establishing a number of prominent climatologies for convective storm ingredients (e.g., Markowski et al. 2003; Thompson et al. 2003, 2007). As useful as these analysis soundings have proven, they are probably not reliable for assessing near-storm variability, which would rely heavily on the model's parameterized representation of the storms. Furthermore, Coniglio (2012) compared RUC analyses and 1-h forecasts to preconvective soundings from VORTEX2 and found substantial model errors (“large relative to their potential impact on convective evolution”) even at the analysis time.

Potvin et al. (2010) used what is probably the most elegant approach to assessing near-storm variability with conventional observations, combining approximately 1200 proximity soundings from the vicinity of significant [enhanced Fujita scale 2 (EF2) or stronger] tornadoes and binning them as a function of distance and time from the storms. From this, they found that soundings close to tornadic storms (less than 1 h and less than 40-km separation) on average exhibited parameters less favorable for tornadoes than those from somewhat farther away. They inferred that the soundings from close to the storm were unrepresentative of the far-field environment owing to what they deemed to be “convective feedbacks.” However, Potvin et al. (2010) were still limited by having only one sounding per storm, and they did not attempt to account for differences in azimuth (only distance) from the storms in their dataset. In addition, the Potvin et al. (2010) dataset included only significant tornado-producing storms, so no comparison between tornadic and nontornadic storms was possible.

The aim of the present work is to substantiate the spatial and temporal patterns of environmental variability and storm-induced modification through analysis of numerous sets of contemporaneous near-supercell soundings from VORTEX2. Further details about the VORTEX2 sounding attributes and data processing are reviewed in section 2, after which the resulting composites are presented and interpreted in section 3. The paper concludes with some ideas for extending this work and a summary in section 4.

2. Methods

a. VORTEX2 sounding operations and characteristics

Supercell sampling during VORTEX2 was unique (compared to the studies reviewed in section 1) because four nearly synchronous sounding measurements were regularly made from the near inflow (~30–40 km from the storm's updraft), distant inflow (~70–100 km from the storm's updraft), and forward and rear flanks of active supercells (e.g., Fig. 1; see also Fig. 12 of Wurman et al. 2012). The sounding units were fully mobile, and the sampling strategy was “storm-following” rather than tethered to particular locations. The pattern of four contemporaneous soundings was repeated at an ~45–60-min interval for actively targeted storms. This basic near-storm sampling approach was undertaken for more than 20 supercell cases during VORTEX2 (see, e.g., Wurman et al. 2012, their Fig. 3 and Table 3).

Fig. 1.

Plan view of averaged base scan radar reflectivity (shaded), storm-relative sounding launch points (circles), and storm-relative sounding trajectories for the 12 supercells included in the present analyses. All axis labels are distances (km) from the supercell's updraft position. In this figure (only) the x and y coordinate axes have not been aligned with respect to the deep-layer vertical wind shear (which varies among cases).

Fig. 1.

Plan view of averaged base scan radar reflectivity (shaded), storm-relative sounding launch points (circles), and storm-relative sounding trajectories for the 12 supercells included in the present analyses. All axis labels are distances (km) from the supercell's updraft position. In this figure (only) the x and y coordinate axes have not been aligned with respect to the deep-layer vertical wind shear (which varies among cases).

All of the VORTEX2 mobile soundings were made with Vaisala RS92 radiosondes. Before launching, each sonde's measurements of temperature, humidity, pressure, and GPS location were checked against portable instruments at the launch site. After the field campaign, all of the soundings were quality-controlled by the National Center for Atmospheric Research (NCAR) Earth Observing Laboratory (EOL); the details of these quality control procedures are explained in a “readme” document that is available from the EOL VORTEX2 data archive (http://data.eol.ucar.edu/master_list/?project=VORTEX2; the documentation is also available from the author), and they largely mirror those reported by Loehrer et al. (1996, 1998). The author performed additional subjective quality assessment by reviewing skew T–logp diagrams and sonde trajectories for each sounding in the composite. Soundings that encountered other nearby storms or had unexplained erratic profiles were discarded.

To represent near-storm variability as cleanly as possible, all soundings in the analysis were from periods of active VORTEX2 supercell sampling (not from before storm initiation or after storm demise). Given the nominal ascent rate of roughly 5 m s−1 for soundings during VORTEX2, sondes could travel horizontal distances greater than 50 km during the time that it took them to ascend to the tropopause (e.g., Fig. 1). Therefore, whereas most past studies have treated soundings as if they were measurements of the local vertical column, the present study exploited the true storm-relative positions of each 1-s measurement from each sounding (this process is explained in section 2c). Using these true storm-relative positions in the analysis adds considerable realism to the horizontal structures of the final fields, much as demonstrated for a VORTEX2 squall line by Bryan and Parker (2010).

b. Selection of cases for the composites

From among all VORTEX2 cases, there were 14 days with at least eight near-storm soundings (i.e., two complete launches from the set of four soundings units) made near an active, VORTEX2-targeted supercell.2 Upon close inspection, three of these days were deemed to be unsuitable for inclusion in a composite analysis (e.g., because of widespread intervening convection or sparse sampling on a poor road network). This left 11 days, containing 12 well-sampled supercells. After having removed any dubious soundings (section 2a), for the final analysis an individual storm was required to have at least six remaining nearby soundings so that meaningful perturbations from an averaged base state could be computed. These selection and quality control steps yielded a total of 134 near-supercell soundings. As summarized in Table 1, of the final 12 supercells, 7 were tornadic (with a total of 84 usable soundings) and 5 were nontornadic (with a total of 50 usable soundings). The preponderance of tornadic soundings is attributable to the fact that such storms tended to be sampled for longer periods of time during VORTEX2.

Table 1.

Summary of the cases and soundings used in this study. Only the number of quality soundings made during storm maturity (i.e., only those used for the actual composites) are reported in the fourth column; no preconvective or postdemise soundings are included. The case date in the left-hand column is the calendar date in LST, whereas the sounding times in the right-hand column are launch times in UTC (operations often continued into the next day UTC). (The VORTEX2 field catalog mentioned in the footnotes is available online at http://catalog.eol.ucar.edu/vortex2_2010/.)

Summary of the cases and soundings used in this study. Only the number of quality soundings made during storm maturity (i.e., only those used for the actual composites) are reported in the fourth column; no preconvective or postdemise soundings are included. The case date in the left-hand column is the calendar date in LST, whereas the sounding times in the right-hand column are launch times in UTC (operations often continued into the next day UTC). (The VORTEX2 field catalog mentioned in the footnotes is available online at http://catalog.eol.ucar.edu/vortex2_2010/.)
Summary of the cases and soundings used in this study. Only the number of quality soundings made during storm maturity (i.e., only those used for the actual composites) are reported in the fourth column; no preconvective or postdemise soundings are included. The case date in the left-hand column is the calendar date in LST, whereas the sounding times in the right-hand column are launch times in UTC (operations often continued into the next day UTC). (The VORTEX2 field catalog mentioned in the footnotes is available online at http://catalog.eol.ucar.edu/vortex2_2010/.)

c. Use of adjusted spatial coordinates

To account for differences in surface elevation among soundings (both on a given day, and from case to case), the vertical coordinate transformation of Gal-Chen and Somerville (1975) was used:

 
formula

wherein all heights on the right-hand side are above mean sea level (MSL). For this study, H was taken to be 12 km (roughly the MSL height of the tropopause) and zsfc was the altitude at which the particular sounding was launched. This coordinate has the property of being very nearly equal to height above ground level (AGL) approaching the ground, and very nearly equal to height MSL approaching altitude H. Such an approach is similar to that endorsed by Trier et al. (2000), who noted that the transformation accounts for wide variations in surface elevation while approximately preserving integrated quantities such as convective available potential energy (CAPE) and convective inhibition (CIN). The main reason for using a height-based (instead of pressure based) terrain-following coordinate was that each sonde's GPS altitude at launch was checked against a more reliable instrument than was the pressure. Also, use of z* allows for a cleaner separation of terrain elevation effects from synoptic and mesoscale pressure fluctuations.

To account for different storm orientations on different days, all soundings were then converted into a common horizontal coordinate system (x*, y*) as follows. Updraft-mesocyclone locations were recorded for all target storms based upon manual tracking of the (bounded) weak echo region, hook echo, and rotational velocity signatures in every Weather Surveillance Radar-1988 Doppler (WSR-88D) level-II volume scan from during active VORTEX2 sampling. The updraft-mesocyclone locations were used to compute storm-relative positions for each 1-s sounding record, with (x = 0, y = 0) for the updraft position at each time. Finally, each individual sounding's positions and wind vectors were rotated about the updraft location so that the final x* coordinate axis was aligned with the 0–6-km vector wind difference3 from that storm's “reference sounding” (an average of distant inflow soundings, as explained in the next subsection). The contemporaneous WSR-88D base scan data from each sounding launch were also converted into (x*, y*) coordinates to produce corresponding reflectivity imagery for reference. The effects of reorienting the sounding and radar data can be surmised by comparing the individual cases in Fig. 1 versus the combined population of 134 soundings in Fig. 2.

Fig. 2.

Trajectories for the 134 soundings used in this study, projected onto three Cartesian planes; (lower left) x*–y* plan view plot; (upper left) x*–z* cross-sectional plot; (lower right) y*–z* cross-sectional plot. As explained in the text, all of the positions are storm relative (with each storm updraft position relocated to x* = 0 km, y* = 0 km) and all of the trajectories have been rotated to align the x* axis with each reference sounding's 0–6-km bulk vertical shear vector. The averaged base scan radar reflectivity is shaded on the plan view chart, using one base scan image for each sounding (recentered and rotated into x*, y* space just as the soundings are).

Fig. 2.

Trajectories for the 134 soundings used in this study, projected onto three Cartesian planes; (lower left) x*–y* plan view plot; (upper left) x*–z* cross-sectional plot; (lower right) y*–z* cross-sectional plot. As explained in the text, all of the positions are storm relative (with each storm updraft position relocated to x* = 0 km, y* = 0 km) and all of the trajectories have been rotated to align the x* axis with each reference sounding's 0–6-km bulk vertical shear vector. The averaged base scan radar reflectivity is shaded on the plan view chart, using one base scan image for each sounding (recentered and rotated into x*, y* space just as the soundings are).

d. Objective analysis

The soundings were interpolated to a common grid using the Barnes (1973) analysis technique [following the formalism of Koch et al. (1983)]. Traditionally, the analyzed values of some variable are calculated as a weighted sum of the M total input data via

 
formula

wherein each is a single datum (here, a 1-s sounding measurement), and is that datum's corresponding weighting. In turn, the weighting function is given by

 
formula

wherein is the distance between the location of datum m and the analysis grid point , and κ is a parameter defining the filtering scale of the analysis scheme. Inspection of (3) reveals that the choice of κ defines a “radius of influence” R at which has fallen off to a value of e−1:

For the VORTEX2 soundings, the vertical and horizontal data spacings differed greatly, such that a singular value of R (i.e., an isotropic analysis) was not useful. Instead, distinct vertical and horizontal influence radii (Rυ and Rh) were needed. Pauley and Wu (1990) recommended that R be set to roughly Δn, where Δn is the data spacing. Because the nominal rate of ascent for the VORTEX2 soundings was roughly 5 m s−1, the vertical data spacing of the 1-s records was generally close to 5 m. It is less straightforward to define the horizontal data spacing because the soundings were unevenly spaced (e.g., Figs. 1 and 2). A fairly intuitive approximation is the “random data spacing” Δnr suggested by Koch et al. [1983, their Eq. (12)], which represents “what the average data spacing would be inside a square data area A if M observations were uniformly distributed across the area.” For the entire set of 134 soundings, the computed horizontal Δnr = 11.8 km.

Following the recommendations of Pauley and Wu (1990), this suggests that the data (if well behaved) could tolerate Barnes analysis settings of Rυ ≈ 7 m and Rh ≈ 16 km. Pauley and Wu (1990) noted that, in practice, larger radii than this are acceptable, but will be associated with “an analysis which will appear too smooth” for many applications. However, in the present study a moderate amount of smoothing was ultimately necessary to seamlessly combine each unique case and sounding trajectory.4 The final values of Rυ = 50 m and Rh = 20 km were subjectively determined (through many iterations) to be those that provided the highest effective resolution without producing obvious localized maxima or minima due to individual sounding data records. To prevent extrapolation across large data gaps (and outside of the edges of the well-sampled area), the final values for both the vertical and horizontal analyses were masked at any grid point where there were fewer than two input data within a distance of R, or fewer than five input data within a distance of 2.5R. Details unique to the vertical and horizontal analysis procedures are reviewed in the upcoming subsections.

1) Vertical analysis

As a first step, individual sounding data were analyzed to common z* levels with a vertical spacing of 50 m using a vertical Barnes procedure as described above. The top of the vertical grid was z* = 12 km, because most VORTEX2 soundings were cut off at or below that altitude. The 50-m radius of influence Rυ is large relative to the original data spacing (as reviewed above) but was necessary to filter out noisiness in the data associated with swinging of the sonde (after launch and in regions of turbulence). The theoretical Barnes response in height (Fig. 3, using the bottom x axis) reveals that vertical wavelengths below roughly 100 m should be strongly muted in the analysis. The surface (z* = 0 m) point was treated uniquely by halving Rυ, which yields a more localized analysis in order to compensate for the fact that the interpolation at the surface point is one-sided (since there are no data below the ground). After completing the vertical analysis for each sounding, a reference sounding for each supercell was created by averaging all available “distant inflow” soundings5 from that case. Every sounding was then converted into perturbations in potential temperature θ, mixing ratio qυ, and wind components by subtracting the individual sounding's values from the reference sounding for that supercell. The preceding vertical analysis procedures were applied to each sounding individually prior to the horizontal analysis.

Fig. 3.

Theoretical response function for the vertical and horizontal Barnes analyses. For the vertical analysis, the response refers to the vertical wavelengths shown on the bottom axis (m). For the horizontal analysis, the response curve refers to the horizontal wavelengths shown on the top axis (km). The red, green, and blue lines highlight the wavelengths at which the theoretical response is 0.9, 0.5, and 0.1 respectively.

Fig. 3.

Theoretical response function for the vertical and horizontal Barnes analyses. For the vertical analysis, the response refers to the vertical wavelengths shown on the bottom axis (m). For the horizontal analysis, the response curve refers to the horizontal wavelengths shown on the top axis (km). The red, green, and blue lines highlight the wavelengths at which the theoretical response is 0.9, 0.5, and 0.1 respectively.

2) Horizontal analysis

After each sounding had been converted into perturbation form in the x*, y*, z* coordinate system, the final 2D composite perturbation fields at each level (every 50 m in z*) were created using a horizontal Barnes procedure as described above. The horizontal grid had a spacing of 5 km in x* and y*. The 20-km radius of influence Rh closely approximates Δnr, as reviewed above, and the associated theoretical Barnes response (Fig. 3, using the top x axis) reveals that horizontal wavelengths below roughly 40 km should be strongly muted in the analysis. During an earlier version of the composite analysis it was discovered that the present value of Rh , while suitable for combining and smoothing the different cases, caused low-level perturbations from the outflow sector soundings to bleed into the inflow sector of the composite supercell. To overcome this problem, the analysis grid points from the inflow sector (defined as x* > 0 km and y* < 0 km) were overwritten by values from an analysis in which only inflow sector sondes were used (but all other settings were identical). The full analysis and inflow sector analysis were then blended together by use of a horizontal, nine-point smoother. This procedure resulted in sharper gradients near the outflow boundaries, with minimal contamination of the inflow sector by the outflow soundings. Once the final blended composite perturbation fields were completed, the mean base state sounding (a straight average of the reference soundings from each supercell) was then added back as needed to compute total values of variables.

3) Additional calculations

Quantities that involve either vertical integration (such as CAPE, CIN, and storm-relative helicity) or vertical derivatives (such as vertical wind shear) were computed using the final analysis fields in 1D local columns at each grid point. For each column the CAPE integration was stopped at 12 km (the top of the gridded analysis) even if the lifted parcel had some remaining positive buoyancy. Vertical velocities were computed kinematically6 by integrating the anelastic continuity equation. Therefore, because of the smoothness of the analyzed horizontal wind fields, the vertical velocities are characteristically mesoscale (on the order of 1 m s−1), not convective (on the order of 10 m s−1), in magnitude.

3. Results

a. Full supercell composite

1) General character of the composite

The end-product of the procedures outlined in section 2 is a detailed picture of the environment of VORTEX2 supercells; the perturbation equivalent potential temperature field serves to highlight the general character of the analysis (Fig. 4). The choice of Rh leads to a somewhat smooth analysis, but it is clear from the spatial distribution of soundings (e.g., Fig. 2) that there is a great deal of information available for the analysis within the inflow sector (to the southeast of the storm) at ranges of less than ~75 km. This corresponds with the zone where the standard deviations7 in the variables (exemplified for in Fig. 4, contours) are typically the smallest; in other words, there is strong similarity among cases within the inflow sector.

Fig. 4.

Plan view (x*–y*) plots for all 134 supercell soundings, valid at (a) the surface (z* = 0 km), (b) z* = 1 km, (c) z* = 3 km, and (d) z* = 6 km. The perturbation equivalent potential temperature is shaded (K), with the base state value of for each level reported in the titles. Perturbation wind vectors are plotted using the scale vectors as shown (all vectors are rotated into the x*–y* coordinates). The weighted standard deviation in (K) at each grid cell is contoured in white (the calculation is explained in a footnote to the main text). All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick white contours for reference.

Fig. 4.

Plan view (x*–y*) plots for all 134 supercell soundings, valid at (a) the surface (z* = 0 km), (b) z* = 1 km, (c) z* = 3 km, and (d) z* = 6 km. The perturbation equivalent potential temperature is shaded (K), with the base state value of for each level reported in the titles. Perturbation wind vectors are plotted using the scale vectors as shown (all vectors are rotated into the x*–y* coordinates). The weighted standard deviation in (K) at each grid cell is contoured in white (the calculation is explained in a footnote to the main text). All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick white contours for reference.

The standard deviations become larger toward the northern and western edges of the analysis (Fig. 4) because of both the much wider variation among cases in these areas and the smaller number of soundings constraining the analysis there (e.g., Fig. 2). In particular, it is important to remember that any background mesoscale variability in the larger-scale environment (i.e., not what is induced by the storm itself) will also appear in the analysis; in other words, the largest standard deviations associated with the negative in the domain's northwest corner are at least partly attributable to the existence of various synoptic and/or mesoscale boundaries in some of the cases. Even so, a number of realistic near-storm features clearly appear.

As one would expect, negative values occur in the outflow sectors at the surface and z* = 1 km (Figs. 4a,b) whereas in the midlevels a maximum in appears in the general location of the updraft (Fig. 4c) and tends toward the downshear side of the updraft farther aloft (Fig. 4d). Wind perturbations at the surface (Fig. 4a) are in line with what would be expected in the outflow sectors, and become increasingly rotational in the vicinity of the updraft aloft (Figs. 4b–d), as one might expect for a supercell. Notably, however, the vertical vorticity present in the analysis is an order of magnitude smaller than what is typically observed in radar studies of supercells (due to the diffusiveness of the analysis and the fact that few soundings actually directly sampled the updraft and mesocyclone).

2) Discussion of specific ingredients

Because it is not possible to show every variable on every level, the remainder of the article emphasizes a number of common parameters derived from the raw analysis data (e.g., CAPE, CIN, measures of vertical wind shear; Fig. 5). The prevailing distant inflow environment for the 12-supercell composite has substantial CAPE and modest CIN (roughly 2150 and −25 J kg−1, respectively; Fig. 5b). As one follows the storm-relative surface winds toward the updraft from the southeast (Fig. 5b), a slight increase in CIN occurs (toward values of roughly −65 J kg−1). As shown by composite surface virtual temperature perturbations (Fig. 6), parcels flowing toward the storm in the inflow sector would characteristically cool by roughly 0.5–1.0 K on their way to the updraft. However, this low-level cooling is weak enough and shallow enough so as to almost escape detection in composite gridpoint soundings (Fig. 7) and vertical cross sections (Fig. 8). Associated with the shallow cooling, there are also small decreases in CAPE (from roughly 2150 to 2100 J kg−1, not crossing a shading interval in Fig. 5b) and lifting condensation level (LCL) height (from roughly 1050 to 900 m; Fig. 5d) for the inflow surface parcels.

Fig. 5.

Plan view (x*–y*) plots for all 134 supercell soundings: (a) number of soundings within the 20-km radius of influence at the surface level (shaded); (b) CAPE of the surface parcel (shaded, J kg−1), CIN of the surface parcel (contoured, J kg−1), and storm-relative surface wind vectors (m s−1, scaled as shown); (c) 0–3-km storm-relative helicity (m2 s−2, shaded), magnitude of the 0–6-km bulk wind difference (m s−1, contoured), and perturbation surface wind vectors (m s−1, scaled as shown); and (d) height of the LCL of the surface parcel (m, shaded), 0–1-km storm-relative helicity (m2 s−2, contoured), and 0–1-km vector wind difference (m s−1, scaled as shown). All vectors are rotated into the x*–y* coordinates. All vertical integrals and differences are calculated in a 1D column at the individual x*, y* grid point. All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick black contours for reference.

Fig. 5.

Plan view (x*–y*) plots for all 134 supercell soundings: (a) number of soundings within the 20-km radius of influence at the surface level (shaded); (b) CAPE of the surface parcel (shaded, J kg−1), CIN of the surface parcel (contoured, J kg−1), and storm-relative surface wind vectors (m s−1, scaled as shown); (c) 0–3-km storm-relative helicity (m2 s−2, shaded), magnitude of the 0–6-km bulk wind difference (m s−1, contoured), and perturbation surface wind vectors (m s−1, scaled as shown); and (d) height of the LCL of the surface parcel (m, shaded), 0–1-km storm-relative helicity (m2 s−2, contoured), and 0–1-km vector wind difference (m s−1, scaled as shown). All vectors are rotated into the x*–y* coordinates. All vertical integrals and differences are calculated in a 1D column at the individual x*, y* grid point. All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick black contours for reference.

Fig. 6.

Plan view (x*–y*) plot of surface virtual temperature perturbation (shaded, K) and storm-relative surface wind vectors (m s−1, scaled as shown) for all 134 supercell soundings. The locations of the gridpoint profiles in Figs. 7 and 9 are indicated with black letters (D = distant inflow, N = near inflow, F = forward flank, R = rear flank) and the positions of the two cross sections in Fig. 8 are indicated with dashed white line segments. Vectors are rotated into the x*–y* coordinates. All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick black contours for reference.

Fig. 6.

Plan view (x*–y*) plot of surface virtual temperature perturbation (shaded, K) and storm-relative surface wind vectors (m s−1, scaled as shown) for all 134 supercell soundings. The locations of the gridpoint profiles in Figs. 7 and 9 are indicated with black letters (D = distant inflow, N = near inflow, F = forward flank, R = rear flank) and the positions of the two cross sections in Fig. 8 are indicated with dashed white line segments. Vectors are rotated into the x*–y* coordinates. All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick black contours for reference.

Fig. 7.

(left) Skew T–logp thermodynamic diagram and (right) hodograph for gridpoint profiles at the distant inflow (“D” in Fig. 6, here colored red) and near inflow (“N” in Fig. 6, here colored blue) locations within the full analysis for all 134 supercell soundings. On the hodographs, the surface, 1-, 3-, and 6-km data points are denoted with symbols as shown, with the observed mean storm motion plotted with “M” (the storm motion is the same for both hodographs). For reference, the vertical wind profile associated with the near inflow hodograph is plotted atop the skew-T diagram (half barb = 2.5 m s−1, barb = 5 m s−1, flag = 25 m s−1).

Fig. 7.

(left) Skew T–logp thermodynamic diagram and (right) hodograph for gridpoint profiles at the distant inflow (“D” in Fig. 6, here colored red) and near inflow (“N” in Fig. 6, here colored blue) locations within the full analysis for all 134 supercell soundings. On the hodographs, the surface, 1-, 3-, and 6-km data points are denoted with symbols as shown, with the observed mean storm motion plotted with “M” (the storm motion is the same for both hodographs). For reference, the vertical wind profile associated with the near inflow hodograph is plotted atop the skew-T diagram (half barb = 2.5 m s−1, barb = 5 m s−1, flag = 25 m s−1).

Fig. 8.

Vertical cross sections of virtual temperature perturbation (shaded, K), equivalent potential temperature (contoured, K), and storm-relative winds lying in the plane of the cross section (m s−1, scaled as shown). The cross sections extend from the inflow sector through the (a) rear and (b) forward flank outflow boundary. The orientations of the cross sections are shown in Fig. 6. The vertical velocity component was computed kinematically (as explained in section 2e) and is here multiplied by 40 so that it is visible (i.e., the scale vector length is equivalent to 1 m s−1 in the vertical).

Fig. 8.

Vertical cross sections of virtual temperature perturbation (shaded, K), equivalent potential temperature (contoured, K), and storm-relative winds lying in the plane of the cross section (m s−1, scaled as shown). The cross sections extend from the inflow sector through the (a) rear and (b) forward flank outflow boundary. The orientations of the cross sections are shown in Fig. 6. The vertical velocity component was computed kinematically (as explained in section 2e) and is here multiplied by 40 so that it is visible (i.e., the scale vector length is equivalent to 1 m s−1 in the vertical).

The location of this very shallow cooling in the inflow sector could be attributable to shading by either the parent storm or inflow clouds. Although there are no direct measurements of cloud cover available within the composite dataset to corroborate this, anvil shading effects have been previously shown to influence near-storm inflow air (e.g., Markowski et al. 1998a; Frame and Markowski 2010), and the depth of influence can be quite shallow (e.g., Bryan and Parker 2010; their Fig. 8). Alternatively, the ascending motion and upward-sloping surfaces shown in Fig. 8a (for x* > 0 km) and Fig. 8b (for y* > 0 km), as well as the local enhancement in seen at z* = 1 km in Fig. 4b, are at least consistent with the notion that inflow cloudiness (i.e., “feeder bands”) could have developed in some of the cases. Such an effect was noted in the supercell simulation of Ziegler et al. (2010), who attributed an inflow layer of stratocumuli to the lifting that resulted from flow blocking by the storm's cold pool. Very few other effects of the nearby storm are evident in the vertical temperature profile, although there is a signal of moistening and slight warming throughout the middle and upper troposphere (Fig. 7) that is most likely attributable to net condensation in and near the targeted supercell.

The prevailing deep-layer vertical wind shear [represented by the 0–6-km bulk wind difference (BWD)] is rather uniform in space (Fig. 5c; also see the hodograph in Fig. 7), and it is everywhere above the baseline of 18–20 m s−1 that is commonly thought to be necessary for supercell organization (e.g., Rasmussen and Blanchard 1998; Thompson et al. 2003). Interestingly, however, the lowest 1 km of the near inflow wind profile (hodograph) is dramatically different from the far field (Fig. 7), showing substantial backing and increases in speed. This is almost certainly an effect of the supercell upon its environment via lowered near-surface pressures (associated with latent heat release and dynamic pressure effects, e.g., Klemp and Rotunno 1983; Weisman and Klemp 1984). The net effect is a dramatic increase in the 0–1- and 0–3-km storm-relative helicity (SRH) values compared to the far field (Figs. 5c,d), an effect that has also been reported in high-resolution supercell simulations (e.g., Fig. 7 of Brooks et al. 1994). In addition, although the 0–1-km BWD displayed on the hodograph in Fig. 7 is almost the same for the near and distant inflow profiles, Fig. 5d shows a general tendency for increasing length of the 0–1-km bulk vertical shear vectors as one approaches the updraft from the southeast.

In summary, the prevailing far-field environment of the composite VORTEX2 supercell already is favorable for the development and maintenance of supercell thunderstorms (large CAPE, 0–6-km BWD, and 0–3-km SRH; e.g., Rasmussen and Blanchard 1998; Thompson et al. 2003), and it also has ingredients that are widely regarded as indicating enhanced tornado probabilities (rather low LCL heights and substantial 0–1-km SRH and 0–1-km BWD; e.g., Markowski et al. 2002, 2003; Thompson et al. 2003). Contrasts between the tornadic and nontornadic storms are presented in section 3b. The most obvious difference between the distant and near-inflow environments was a dramatic increase in lower tropospheric vertical wind shear.

An increasing body of work suggests that outflows are important sources of horizontal vorticity (which can be subsequently reoriented into vertical vorticity) for both supercells and tornadoes. Although these outflows have been thoroughly sampled at the surface by mobile instruments (e.g., Markowski et al. 2002; Grzych et al. 2007; Wurman et al. 2012), observations aloft in these parts of storms have generally been lacking. As might be expected, CAPE decreases and the magnitude of CIN increases in the outflow sectors to the north and west of the composite storm (Fig. 5b), although it is noteworthy that the surface parcels within the outflow still possess nonzero CAPE (exceeding 1000 J kg−1 over a broad area). This is important because supercell updrafts regularly reingest evaporatively cooled outflow air (this explains the existence of wall clouds, for example; Rotunno and Klemp 1985). There is also some evidence that the potential buoyancy (i.e., CAPE) of this outflow air may be correlated to the likelihood of tornadogenesis (e.g., Markowski et al. 2002; Grzych et al. 2007).

Both vertical cross sections (Fig. 8) and point soundings (Fig. 9) in the composite's outflow sectors reveal that the layer of low-level cooling is quite shallow (500 m deep or less), with temperature deficits of only 1–3 K relative to the ambient environment (at least until one is well to the northwest of the storm; Figs. 6 and 8a). In contrast, many classically configured model simulations of supercells produce outflows >1 km deep with ~10-K surface temperature deficits8 [e.g., as demonstrated by Morrison and Milbrandt (2011)]. It is likely that the composite underestimates what occurs in nature because very few VORTEX2 soundings were actually launched in the coldest parts of the cold pools (because of the hindrance of strong downdrafts and winds, heavy precipitation, and lightning). Such sampling issues are a limitation given that single-storm measurements of supercell cold pools have shown considerable heterogeneity in outflow temperature (e.g., Markowski et al. 2002) and depth (e.g., Ziegler 2013). Even so, a number of other studies have indeed found weak temperature deficits in the forward flanks of supercells (Shabbott and Markowski 2006; Beck and Weiss 2013), and a few features of the forward and rear flank do seem to be well captured in the composite soundings (warming in the middle troposphere and substantial moistening throughout most of the middle and upper troposphere; Fig. 9). Notwithstanding the preceding caveats, the VORTEX2 supercell cold pools appeared to be surprisingly shallow and weak in the particular locations where outflow soundings were launched.

Fig. 9.

As in Fig. 7, except for the gridpoint profiles at the distant inflow (“D” in Fig. 6, here colored red), rear flank (“R” in Fig. 6, here colored blue), and forward flank (“F” in Fig. 6, here colored green) locations within the full analysis for all 134 supercell soundings. The plotted storm motion is the same for all three hodographs. The reference vertical wind profile is for the rear flank hodograph.

Fig. 9.

As in Fig. 7, except for the gridpoint profiles at the distant inflow (“D” in Fig. 6, here colored red), rear flank (“R” in Fig. 6, here colored blue), and forward flank (“F” in Fig. 6, here colored green) locations within the full analysis for all 134 supercell soundings. The plotted storm motion is the same for all three hodographs. The reference vertical wind profile is for the rear flank hodograph.

A final interesting point about the composite supercell's outflow sectors is that the vertical wind profiles (hodographs) there retain roughly the same shape as the prevailing inflow environment (Fig. 9), such that the values of 0–1-km BWD, 0–1-km SRH, and 0–3-km SRH are not substantially different from those in the inflow sector (Figs. 5c,d). The orientation of the bulk shear vectors in the forward and rear flanks (Figs. 5d and 9) suggests that ambient vortex lines should thread through the storm from the environment with relatively consistent orientations (again, with the caveat that the input soundings were generally not made in the strongest parts of the observed storms' outflows). This is potentially important because the way in which vertical vorticity emanates from downdrafts may be quite sensitive to the low-level wind profile in the outflow sector (e.g., Parker and Dahl 2013).

b. Tornadic versus nontornadic composites

To assess how the near-storm environments varied between tornadic and nontornadic supercells, the soundings were divided using the designations of tornadic versus nontornadic as reported in Table 1, and the composite procedure was then rerun. Because the nontornadic sample size was smaller (as discussed in section 2), the overall coverage of useful information was more limited (cf. Figs. 10a,c). Even so, comparisons between the subsets, and between the tornadic cases and the full composite, are informative.

Fig. 10.

As in Figs. 5a and 5b, but for the (a),(b) 84 tornadic supercell soundings and (c),(d) 50 nontornadic soundings.

Fig. 10.

As in Figs. 5a and 5b, but for the (a),(b) 84 tornadic supercell soundings and (c),(d) 50 nontornadic soundings.

Almost uniformly across the board, the far-field environmental parameters were more impressive in the tornadic composite (Figs. 10 and 11), including higher CAPE (roughly 2200 vs 1800 J kg−1, with values of CIN that were comparable), higher 0–6-km BWD (roughly 27 vs 24 m s−1), and lower LCL heights (roughly 900 vs 1200 m). Gridpoint soundings (Fig. 12) make it apparent that the differences in CAPE and LCL height are almost entirely attributable to lower values of boundary layer mixing ratio in the nontornadic cases. This is one of the most glaring differences between the VORTEX2 tornado and nontornado cases in this study, and it echoes the findings of Markowski et al. (2002) and Thompson et al. (2003).

Fig. 11.

As in Figs. 5c and 5d, but for the (a),(b) 84 tornadic supercell soundings and (c),(d) 50 nontornadic soundings.

Fig. 11.

As in Figs. 5c and 5d, but for the (a),(b) 84 tornadic supercell soundings and (c),(d) 50 nontornadic soundings.

Fig. 12.

As in Fig. 7, but for the gridpoint profiles at the distant inflow position (equivalent to “D” in Fig. 6) in the tornadic supercell analysis (here colored green), the near inflow position (equivalent to “N” in Fig. 6) for the tornadic supercell analysis (here colored red), and the near inflow position (again, equivalent to “N” in Fig. 6) for the nontornadic supercell analysis (here colored blue). The mean storm motion for the tornadic storms is plotted in red and for the nontornadic storms is plotted in blue on the hodograph. The reference vertical wind profile is for the nontornadic near inflow hodograph.

Fig. 12.

As in Fig. 7, but for the gridpoint profiles at the distant inflow position (equivalent to “D” in Fig. 6) in the tornadic supercell analysis (here colored green), the near inflow position (equivalent to “N” in Fig. 6) for the tornadic supercell analysis (here colored red), and the near inflow position (again, equivalent to “N” in Fig. 6) for the nontornadic supercell analysis (here colored blue). The mean storm motion for the tornadic storms is plotted in red and for the nontornadic storms is plotted in blue on the hodograph. The reference vertical wind profile is for the nontornadic near inflow hodograph.

In assessing differences in low-level shear, an interesting contrast is that the distant inflow values of 0–1- and 0–3-km SRH are much lower in the nontornadic cases, but they increase considerably near the supercell (Fig. 11). In contrast, the values in the tornadic supercell composite vary much less across the inflow sector. One speculative interpretation is that, although the near-storm values of vertical wind shear in nontornado cases were within 10%–20% of those in the tornadic cases, these values were primarily storm-generated (as they were not present in the distant inflow environment). The local leftward turning of the 0–1-km bulk shear vectors in the vicinity of the nontornadic supercell's forward flank (Fig. 11d) would be consistent with enhanced baroclinic generation of a westward-pointing horizontal vorticity component associated with comparatively cooler air to the north (unfortunately, the baroclinity in the composites is not sufficiently well resolved to calculate a realistic rate). In turn, this enhanced cooling would be consistent with the dryer environmental boundary layer on nontornado days (Fig. 12). Indeed, such enhanced forward flank cooling has been previously found to be anticorrelated with tornado occurrence (Shabbott and Markowski 2006). Ultimately, however, these are the kinds of sensitivities that can only be properly examined with hypothesis tests in a numerical model.

The curious pairing of similar near-storm 0–3-km SRH with comparatively lower 0–1 km SRH in the nontornadic cases also points to a rather different hodograph shape between the two groupings (Fig. 12). The 3-km winds are quite similar between the tornadic and nontornadic storms, however the winds in the lowest 1–2 km are more strongly backed in the nontornadic cases (Fig. 12; cf. Figs. 10b,d). As noted above, the near-storm 0–1-km bulk shear vector for the nontornadic cases therefore points well to the left of that for the tornadic cases (Fig. 12; cf. Figs. 11b,d). For the tornadic near-inflow point, the 0–1-km storm-relative winds are nearly orthogonal to the 0–1-km bulk shear vector (implying that the horizontal vorticity is primarily streamwise), whereas for the nontornadic near-inflow point, the 0–1-km mean storm-relative winds are more nearly parallel to the 0–1-km bulk shear vector (implying a substantial crosswise component of vorticity). These differences in shear orientation are then compounded by the fact that the tornadic supercells appeared to move more strongly to the right of the hodograph9 (Fig. 12), which entails stronger low-level storm-relative winds and thus an increased flux of streamwise vorticity into the updrafts of the tornadic storms. Even though the direct linkage to tornadoes is not yet fully understood, it has long been established that increased import of streamwise vorticity enhances the low-level vertical vorticity in the updraft (e.g., Davies-Jones 1984). In turn, the enhanced low-level vertical vorticity may promote stronger dynamic lifting in the part of the storm where tornadogenesis typically occurs (Markowski and Richardson 2014).

c. Early-in-life versus late-in-life composites

To this point, the composites discussed in section 3 have incorporated soundings from the full span of individual supercells' lifetimes (e.g., Table 1). To assess how the near-storm environments varied over the course of supercell lifetimes, the soundings were divided into early-in-life and late-in-life subsets, and the composite procedure was then rerun. As shown in Fig. 13, the total soundings for each day were divided in time as evenly as possible without subdividing any set of near-simultaneous launches. This approach yielded 68 early-in-life soundings and 66 late-in-life soundings. As described in section 2, all of the analyzed soundings were launched during sampling of an active supercell, so the time separation between the early-in-life and late-in-life groups ranges by case from roughly 1 to 3 h (Fig. 13).

Fig. 13.

Histogram of sounding launch times (shown with respect to time of local sunset) for each storm in this study. The 68 soundings included in the early-in-life subset are plotted in red. The 66 soundings included in the late-in-life subset are plotted in blue. Soundings used to compute the reference sounding for each storm are denoted by a closed circle. All other soundings are denoted by a symbol. For illustrative purposes, the time of the median launch (for all soundings) relative to sunset is plotted with a green line.

Fig. 13.

Histogram of sounding launch times (shown with respect to time of local sunset) for each storm in this study. The 68 soundings included in the early-in-life subset are plotted in red. The 66 soundings included in the late-in-life subset are plotted in blue. Soundings used to compute the reference sounding for each storm are denoted by a closed circle. All other soundings are denoted by a symbol. For illustrative purposes, the time of the median launch (for all soundings) relative to sunset is plotted with a green line.

The differences between the early-in-life and late-in-life soundings seem to be primarily consistent with the fact that VORTEX2 sampling was undertaken in the late afternoon and early evening (Table 1 and Fig. 13). The majority of the early-in-life soundings were launched >2 h prior to sunset, and the majority of the late-in-life soundings were launched within 2 h of sunset. Not surprisingly, then, the late-in-life soundings had less CAPE, a higher magnitude of CIN, and lower LCL heights (cf. Figs. 14b,d and 15b,d). Gridpoint soundings (Fig. 16) make it clear that this effect is largely due to low-level cooling and moistening as the evening transition of the boundary layer begins (note also the very shallow stable layer at the surface in the late-in-life sounding). Direct thermodynamic modifications of the near-inflow environment by the storm are harder to identify apart from some subtle warming and moistening evident above 400 hPa in the later sounding (Fig. 16).

Fig. 14.

As in Figs. 5a and 5b, but for the (a),(b) 68 early-in-life and (c),(d) 66 late-in-life supercell soundings.

Fig. 14.

As in Figs. 5a and 5b, but for the (a),(b) 68 early-in-life and (c),(d) 66 late-in-life supercell soundings.

Fig. 15.

As in Figs. 5c and 5d, but for the (a),(b) 68 early-in-life and (c),(d) 66 late-in-life supercell soundings.

Fig. 15.

As in Figs. 5c and 5d, but for the (a),(b) 68 early-in-life and (c),(d) 66 late-in-life supercell soundings.

Fig. 16.

As in Fig. 7, but for the gridpoint profiles at the near inflow position (equivalent to “N” in Fig. 6) for the early-in-life supercell analysis (here colored red) and the late-in-life supercell analysis (here colored blue). The mean storm motions for the two samples are plotted in their corresponding colors. The reference vertical wind profile is for the late-in-life hodograph.

Fig. 16.

As in Fig. 7, but for the gridpoint profiles at the near inflow position (equivalent to “N” in Fig. 6) for the early-in-life supercell analysis (here colored red) and the late-in-life supercell analysis (here colored blue). The mean storm motions for the two samples are plotted in their corresponding colors. The reference vertical wind profile is for the late-in-life hodograph.

Although the 0–6-km BWD is almost identical in the early and late-in-life composites (cf. Figs. 15a,c), there is some subtle veering of the flow above z* = 6 km (comparing the hodographs in Fig. 16), which may reflect increasing upper-level divergence over time from the storm to the immediate northwest. More importantly, the winds below roughly 3 km AGL have increased in speed in the late-in-life analysis (Fig. 16). It is possible that this partly represents an accumulation over time of the near-storm accelerations that were evident in the full composite's near-storm profile in Fig. 7. However, given the noticeable low-level cooling and stabilization discussed above, it is also presumed that the lower tropospheric winds accelerated over time in response to declining turbulent mixing during the boundary layer's evening transition [i.e., the classic Blackadar (1957) mechanism, also discussed in the context of supercells by Maddox (1993)]. Although it is impossible to cleanly separate the two effects in the present dataset, the end results are clear: dramatic increases in 0–1-km and 0–3-km SRH, along with modest increases in the 0–1-km BWD (Fig. 15; see also the hodograph in Fig. 16).

At the suggestion of a reviewer, the early and late composites were remade by dividing the soundings evenly based on time relative to local sunset (i.e., separating them at the green line drawn in Fig. 13). Conceptually, this is less than ideal, because several cases are placed almost entirely into one category or the other (meaning that fundamental differences between cases are projected onto the purported diurnal signals). Even so, the end results were between-group differences with the same signs but even larger magnitudes than in the early-in-life versus late-in-life comparisons. This lends confidence that the temporal changes described above are predominantly diurnal in nature.

d. Composites of operational forecasting parameters

So far, many contrasts have been drawn between the distant and near inflow locations, between tornadic and nontornadic supercells, and between the early and late-in-life time periods. Since many of these differences are offsetting in some sense (i.e., unfavorable low-level cooling is paired with a favorable increase in vertical wind shear), it is difficult to assess overall whether the environmental variations have a net impact on the questions of supercell maintenance and tornado production. As a proxy, it can be helpful to refer to the work of Thompson et al. (2003, 2004, 2007), who introduced and refined two skillful operational indices: the supercell composite parameter (SCP) and the significant tornado parameter (STP). These parameters10 represent combinations of thermodynamic and vertical wind shear ingredients that either best separate supercells from nonsupercells (in the case of SCP) or best separate storms producing significant tornadoes (EF2 or greater) from nontornadic supercells (in the case of STP).

Because all of the storms in the composite were supercells, it is probably not surprising that the SCP is well above the threshold of 1 for all four of the supercell subsets (contours in Fig. 17). Even so, the values are highest for the tornadic supercells (and the large values span a much greater fraction of the inflow sector). Fittingly, STP in the distant inflow for the tornado cases is also well above the threshold of 1, whereas STP in the distant inflow for the nontornado cases is below 1 (Figs. 17a,b). It is interesting that, in all four subset composites in Fig. 17, the STP increases noticeably as one travels toward the storm. Assuming that the near-storm enhancements in VORTEX2 storms are common, it is possible that the most discriminatory STP threshold would actually be a value >1 in the near-inflow region. Even so, the substantial differences in distant inflow values (roughly 3.1 in the tornado composite versus 0.8 in the nontornadic composite at x* = 60 km, y* = −40 km) are probably the most representative of those used in climatology studies and by operational forecasters (since the near-storm variability captured in the present composites is not routinely sampled).

Fig. 17.

Plan view (x*–y*) plots of the significant tornado parameter (STP, shaded) and supercell composite parameter (SCP, contoured) for (a) the 84 tornadic supercell soundings, (b) the 50 nontornadic supercell soundings, (c) the 68 early-in-life supercell soundings, and (d) the 66 late-in-life supercell soundings. The STP and SCP are computed as explained in the text using a 1D column at the individual x*, y* grid point. All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick black contours for reference.

Fig. 17.

Plan view (x*–y*) plots of the significant tornado parameter (STP, shaded) and supercell composite parameter (SCP, contoured) for (a) the 84 tornadic supercell soundings, (b) the 50 nontornadic supercell soundings, (c) the 68 early-in-life supercell soundings, and (d) the 66 late-in-life supercell soundings. The STP and SCP are computed as explained in the text using a 1D column at the individual x*, y* grid point. All axis labels are distances (km) from the supercell updraft position. The 30- and 50-dBZ radar reflectivity levels from Fig. 2 are shown as thick black contours for reference.

Just as there is a dependence of the SCP and STP on distance from the storm, there are also modest increases in both parameters later in the supercells' lifetimes (Figs. 17c,d). Apparently, in both cases (closer proximity or later time), increases in vertical wind shear and relative humidity are proportionally greater than the increases in stability. This may be a point that has operational utility, as it implies that marginal environments may become more favorable over time even as the boundary layer stabilizes. However, given the very nonlinear interplay between environmental wind profiles and updraft intensity (e.g., Weisman and Rotunno 2000), these are again the kinds of sensitivities that are best examined with a numerical model. It may well be that the statistical skill of simple multiplicative parameters like SCP and STP does not directly correspond to the actual changes in supercell processes and storm–environment interactions.

4. Conclusions

a. Future work

Section 3 advanced a number of speculative hypotheses to explain the measured environmental heterogeneity as well as differences between tornadic versus nontornadic supercells and early versus late-in-life supercell environments. Unfortunately, this composite is only a starting point: true physical attribution is difficult using a dataset such as this. One could use composite gridpoint soundings from the present subgroupings (tornadic vs nontornadic, early-in-life vs late-in-life) to study sensitivities of simulated storms to the observed environmental differences. One could also potentially use a somewhat coarser analysis (i.e., with larger Rh) as a “synoptic” initial condition within which an idealized simulated supercell is triggered [much as Coniglio and Stensrud (2001) did for derechos]. This might be one avenue toward increasingly realistic incorporation of environmental heterogeneity in idealized supercell studies, a trend that has already begun (e.g., Richardson et al. 2007; Ziegler et al. 2010; Letkewicz et al. 2013). The long-range goal of such modeling work would be to understand the governing physical processes in supercells and tornadoes with due consideration of influences from the near-storm environment. The value of the present VORTEX2 composite supercell environments is in providing better observational constraints for model initial conditions and better quality assurance for the subsequent model output of such studies.

b. Summary

This study combined 134 soundings from VORTEX2 supercells into a 3D composite analysis, accounting for variations in the base state among storm days, downwind drifting of the sondes, and launch-by-launch differences in surface elevation. The dense sounding coverage in the near-storm environment provides an unprecedented observational view of supercell–environment interactions.

The composite analyses make it clear that the use of a single sounding to represent the “environment” of a case is potentially risky: supercell and tornado ingredients (in terms of most widely used metrics) appear to improve as one approaches the storm from the distant inflow region, and also as the storm's lifetime proceeds through the boundary layer transition of late afternoon and early evening. In both cases (moving closer to the storm or later in the storm's lifetime) the trade-off seems to be modest low-level cooling that is offset by increases in the vertical wind shear. It is not clear how much physical similarity is involved in this apparent parallel between the environmental changes linked to proximity and those linked to time.

From the perspective of forecasting indices, the composite environment of the tornadic supercells is clearly more impressive than that of the nontornadic supercells in the far field, with considerably higher values of the operational supercell composite parameter (SCP) and significant tornado parameter (STP). An interesting result is that the near-storm environment of the nontornadic supercells is substantially enhanced compared to the far field (presumably by the storm itself), with parameter values approaching those of the tornadic cases. The observed near-storm backing and lengthening of the nontornadic wind profile may ironically be related to the seemingly unfavorable boundary layer dryness of the nontornadic cases: a speculative interpretation is that low-level evaporative cooling accounts for baroclinic reorientation of the low-level shear vectors in the nontornadic near-storm hodographs. The “enhanced” near-storm nontornadic hodographs also have increasingly crosswise horizontal vorticity, which is presumably less favorable for low-level updraft rotation.

Finally, it is intriguing to see that the composite supercell cold pools are quite shallow (perhaps having depths of only 500 m), at least based on the locations where VORTEX2 soundings were launched. In addition, appreciable surface-based outflow CAPE (exceeding 1000 J kg−1 for all supercell subsets in this study) appears to remain within close range of the updraft. Even though cold pools are known to be spatially heterogeneous, and VORTEX2 sondes were generally not launched in the main downdraft and heavy precipitation zones, these findings are a noteworthy benchmark; almost all historical measurements of supercell outflows have been made at the surface only.

The present results motivate a handful of questions about how supercells modify their surroundings, and how the differences between tornadic and nontornadic supercell environments result in internal changes to the storms themselves. Unfortunately, the answers to such questions are beyond the reach of this simple dataset, and will likely require hypothesis-driven numerical modeling experiments. Hopefully, the VORTEX2 sounding composites provide an updated, robust observational constraint upon what constitutes “realism” in such models.

Acknowledgments

Funding for this research was provided by the National Science Foundation under Grants ATM-0758509 and AGS-1156123. The author gratefully acknowledges the assistance of Brice Coffer and Chris MacIntosh, who tracked and recorded the updraft positions for most of the storms shown in Table 1. The author thanks the Convective Storms Group at North Carolina State University for comments on this paper, as well as the numerous conference and workshop participants who discussed earlier versions of this work. Finally, the formal reviews by Adam Houston, Russ Schumacher, and Conrad Ziegler helped to refine many aspects of the presentation.

REFERENCES

REFERENCES
Barnes
,
S. L.
,
1973
: Mesoscale objective analysis using weighted time-series observations. NOAA Tech. Memo. ERL NSSL-62, 60 pp. [Available from National Severe Storms Laboratory, Norman, OK 73069.]
Beck
,
J.
, and
C.
Weiss
,
2013
:
An assessment of low-level baroclinity and vorticity within a simulated supercell
.
Mon. Wea. Rev.
,
141
,
649
669
.
Benjamin
,
S. G.
,
G. A.
Grell
,
J. M.
Brown
,
T. G.
Smirnova
, and
R.
Bleck
,
2004
:
Mesoscale weather prediction with the RUC hybrid isentropic–terrain-following coordinate model
.
Mon. Wea. Rev.
,
132
,
473
494
.
Blackadar
,
A. K.
,
1957
:
Boundary layer wind maxima and their significance for the growth of nocturnal inversions
.
Bull. Amer. Meteor. Soc.
,
38
,
282
290
.
Brooks
,
H. E.
,
C. A.
Doswell
III
, and
J.
Cooper
,
1994
:
On the environments of tornadic and nontornadic mesocyclones
.
Wea. Forecasting
,
9
,
606
618
.
Bryan
,
G. H.
, and
M. D.
Parker
,
2010
:
Observations of a squall line and its near environment using high-frequency rawinsonde launches during VORTEX2
.
Mon. Wea. Rev.
,
138
,
4076
4097
.
Coniglio
,
M. C.
,
2012
:
Verification of RUC 0–1-h forecasts and SPC mesoscale analyses using VORTEX2 soundings
.
Wea. Forecasting
,
27
,
667
683
.
Coniglio
,
M. C.
, and
D. J.
Stensrud
,
2001
:
Simulation of a progressive derecho using composite initial conditions
.
Mon. Wea. Rev.
,
129
,
1593
1616
.
Davies-Jones
,
R.
,
1984
:
Streamwise vorticity: The origin of updraft rotation in supercell storms
.
J. Atmos. Sci.
,
41
,
2991
3006
.
Dawson
,
D. T.
,
M.
Xue
,
J. A.
Milbrandt
, and
M. K.
Yau
,
2010
:
Comparison of evaporation and cold pool development between single-moment and multimoment bulk microphysics schemes in idealized simulations of tornadic thunderstorms
.
Mon. Wea. Rev.
,
138
,
1152
1171
.
Frame
,
J.
, and
P.
Markowski
,
2010
:
Numerical simulations of radiative cooling beneath the anvils of supercell thunderstorms
.
Mon. Wea. Rev.
,
138
,
3024
3047
.
Gal-Chen
,
T.
, and
R.
Somerville
,
1975
:
On the use of a coordinate transformation for the solution of the Navier–Stokes equations
.
J. Comput. Phys.
,
17
,
209
228
.
Grzych
,
M. L.
,
B. D.
Lee
, and
C. A.
Finley
,
2007
:
Thermodynamic analysis of supercell rear-flank downdrafts from Project ANSWERS
.
Mon. Wea. Rev.
,
135
,
240
246
.
Klemp
,
J. B.
, and
R.
Rotunno
,
1983
:
A study of the tornadic region within a supercell thunderstorm
.
J. Atmos. Sci.
,
40
,
359
377
.
Koch
,
S. E.
,
M.
DesJardins
, and
P. J.
Kocin
,
1983
:
An interactive Barnes objective map analysis scheme for use with satellite and conventional data
.
J. Climate Appl. Meteor.
,
22
,
1487
1503
.
Letkewicz
,
C. E.
,
A. J.
French
, and
M. D.
Parker
,
2013
:
Base-state substitution: An idealized modeling technique for approximating environmental variability
.
Mon. Wea. Rev.
,
141
,
3062
3086
.
Loehrer
,
S. M.
,
T. A.
Edmands
, and
J. A.
Moore
,
1996
:
TOGA COARE upper-air sounding data archive: Development and quality control procedures
.
Bull. Amer. Meteor. Soc.
,
77
,
2651
2671
.
Loehrer
,
S. M.
,
S. F.
Williams
, and
J. A.
Moore
,
1998
: Results from UCAR/JOSS quality control of atmospheric soundings from field projects. Preprints, 10th Symp. on Meteorological Observations and Instrumentation, Phoenix, AZ, Amer. Meteor. Soc., 1–6.
Maddox
,
R. A.
,
1993
: Diurnal low-level wind oscillation and storm-relative helicity. The Tornado: Its Structure, Dynamics, Prediction, and Hazards, Geophys. Monogr., Vol. 79, Amer. Geophys. Union, 591–598.
Markowski
,
P. M.
, and
Y. P.
Richardson
,
2014
:
The influence of environmental low-level shear and cold pools on tornadogenesis: Insights from idealized simulations
.
J. Atmos. Sci.
,
in press
.
Markowski
,
P. M.
,
E. N.
Rasmussen
,
J. M.
Straka
, and
D. C.
Dowell
,
1998a
:
Observations of low-level baroclinity generated by anvil shadows
.
Mon. Wea. Rev.
,
126
,
2942
2958
.
Markowski
,
P. M.
,
J. M.
Straka
,
E. N.
Rasmussen
, and
D. O.
Blanchard
,
1998b
:
Variability of storm-relative helicity during VORTEX
.
Mon. Wea. Rev.
,
126
,
2959
2971
.
Markowski
,
P. M.
,
J. M.
Straka
, and
E. N.
Rasmussen
,
2002
:
Direct surface thermodynamic observations within the rear-flank downdrafts of nontornadic and tornadic supercells
.
Mon. Wea. Rev.
,
130
,
1692
1721
.
Markowski
,
P. M.
,
C.
Hannon
,
J.
Frame
,
E.
Lancaster
,
A.
Pietrycha
,
R.
Edwards
, and
R. L.
Thompson
,
2003
:
Characteristics of vertical wind profiles near supercells obtained from the rapid update cycle
.
Wea. Forecasting
,
18
,
1262
1272
.
Morrison
,
H.
, and
J.
Milbrandt
,
2011
:
Comparison of two-moment bulk microphysics schemes in idealized supercell thunderstorm simulations
.
Mon. Wea. Rev.
,
139
,
1103
1130
.
Parker
,
M. D.
, and
J. M. L.
Dahl
,
2013
: Toy model simulations of surface vorticity generation in downdrafts. Extended Abstracts, 15th Conf. on Mesoscale Processes, Portland, OR, Amer. Meteor. Soc., 12.3. [Available online at https://ams.confex.com/ams/15MESO/webprogram/Paper227679.html.]
Pauley
,
P. M.
, and
X.
Wu
,
1990
:
The theoretical, discrete, and actual response of the Barnes objective analysis scheme for one- and two-dimensional fields
.
Mon. Wea. Rev.
,
118
,
1145
1164
.
Potvin
,
C. K.
,
K. L.
Elmore
, and
S. J.
Weiss
,
2010
:
Assessing the impacts of proximity sounding criteria on the climatology of significant tornado environments
.
Wea. Forecasting
,
25
,
921
930
.
Rasmussen
,
E. N.
, and
D. O.
Blanchard
,
1998
:
A baseline climatology of sounding-derived supercell and tornado forecast parameters
.
Wea. Forecasting
,
13
,
1148
1164
.
Richardson
,
Y. P.
,
K. K.
Droegemeier
, and
R. P.
Davies-Jones
,
2007
:
The influence of horizontal environmental variability on numerically simulated convective storms. Part I: Variations in vertical shear
.
Mon. Wea. Rev.
,
135
,
3429
3455
.
Rotunno
,
R.
, and
J.
Klemp
,
1985
:
On the rotation and propagation of simulated supercell thunderstorms
.
J. Atmos. Sci.
,
42
,
271
292
.
Shabbott
,
C. J.
, and
P. M.
Markowski
,
2006
:
Surface in situ observations within the outflow of forward-flank downdrafts of supercell thunderstorms
.
Mon. Wea. Rev.
,
134
,
1422
1441
.
Snook
,
N.
, and
M.
Xue
,
2008
:
Effects of microphysical drop size distribution on tornadogenesis in supercell thunderstorms
.
Geophys. Res. Lett.
,
35
,
L24803
, doi:10.1029/2008GL035866.
Thompson
,
R. L.
,
R.
Edwards
,
J. A.
Hart
,
K. L.
Elmore
, and
P. M.
Markowski
,
2003
:
Close proximity soundings within supercell environments obtained from the Rapid Update Cycle
.
Wea. Forecasting
,
18
,
1243
1261
.
Thompson
,
R. L.
,
R.
Edwards
, and
C. M.
Mead
,
2004
: An update to the supercell composite and significant tornado parameters. Preprints, 22nd Conf. on Severe Local Storms, Hyannis, MA, Amer. Meteor. Soc., P8.1. [Available online at http://ams.confex.com/ams/pdfpapers/82100.pdf.]
Thompson
,
R. L.
,
C. M.
Mead
, and
R.
Edwards
,
2007
:
Effective storm-relative helicity and bulk shear in supercell thunderstorm environments
.
Wea. Forecasting
,
22
,
102
115
.
Thompson
,
R. L.
,
B. T.
Smith
,
J. S.
Grams
,
A. R.
Dean
, and
C.
Broyles
,
2012
:
Convective modes for significant severe thunderstorms in the contiguous United States. Part II: Supercell and QLCS tornado environments
.
Wea. Forecasting
,
27
,
1136
1154
.
Trier
,
S. B.
,
C. A.
Davis
, and
J. D.
Tuttle
,
2000
:
Long-lived mesoconvective vortices and their environment. Part I: Observations from the central United States during the 1998 warm season
.
Mon. Wea. Rev.
,
128
,
3376
3395
.
Weisman
,
M. L.
, and
J. B.
Klemp
,
1984
:
The structure and classification of numerically simulated convective storms in directionally varying wind shears
.
Mon. Wea. Rev.
,
112
,
2479
2498
.
Weisman
,
M. L.
, and
R.
Rotunno
,
2000
:
The use of vertical wind shear versus helicity in interpreting supercell dynamics
.
J. Atmos. Sci.
,
57
,
1452
1472
.
Wurman
,
J.
,
D.
Dowell
,
Y.
Richardson
,
P.
Markowski
,
E.
Rasmussen
,
D.
Burgess
,
L.
Wicker
, and
H. B.
Bluestein
,
2012
:
The Second Verification of the Origins of Rotation in Tornadoes Experiment: VORTEX2
.
Bull. Amer. Meteor. Soc.
,
93
,
1147
1170
.
Ziegler
,
C. L.
,
2013
:
A diabatic Lagrangian technique for the analysis of convective storms. Part II: Application to a radar-observed storm
.
J. Atmos. Oceanic Technol.
,
30
,
2266
2280
.
Ziegler
,
C. L.
,
E. R.
Mansell
,
J. M.
Straka
,
D. R.
MacGorman
, and
D. W.
Burgess
,
2010
:
The impact of spatial variations of low-level stability on the lifecycle of a simulated supercell storm
.
Mon. Wea. Rev.
,
138
,
1738
1766
.

Footnotes

1

The quoted text appeared in the VORTEX2 Scientific Program Overview that was submitted to the National Science Foundation in 2006.

2

Notably, this threshold eliminated several impressive VORTEX2 cases from consideration (e.g., 9 June 2009, 10 May 2010, and 13 June 2010), but a faithful mapping of near-storm perturbations relies on having a representative mean base state and multiple sounding times from each sector of a mature storm.

3

The 0–6-km vector wind difference (sometimes called the 0–6-km bulk vertical shear vector) was chosen from among several alternatives simply because it most closely aligned the radar reflectivity fields and sonde trajectories (e.g., in the combined view shown in Fig. 2).

4

Many studies also use a second “correction pass” of the Barnes scheme to refine the analysis values (e.g., Koch et al. 1983). In the present study it was found (through trial and error) that there was no obvious benefit to a second pass given the somewhat high degree of smoothness required to blend the different cases.

5

An average of all “distant inflow” soundings was chosen over a single preconvective sounding so that the diurnal cycle in temperature would not dominate the calculated perturbations in longer-lived storms. Interested readers can look ahead to Fig. 13 to see the diurnal distribution of the soundings used to calculate the reference state for each case.

6

Although sondes' GPS rates of ascent were also analyzed, it is unclear how to account for the variations among soundings that occur solely due to the differing amounts of helium used in each balloon. Intense convective updrafts and downdrafts were obvious within the GPS vertical velocity data, but there is low confidence (and visible noise) in the much weaker vertical velocities found outside of storms. Launch-by-launch helium usage was not precisely documented during VORTEX2, so no simple correction could be applied.

7

The standard deviation is typically computed using the simple average of squared deviations from the mean (and then taking the square root of this average). In the present study, given the uneven distribution of nearby soundings, the standard deviation is computed using a weighted average of squared deviations from the mean, with weights determined from the horizontal Barnes analysis.

8

Such strong simulated cold pools may partly reflect uncertainties in the microphysical parameterizations used by many models (a topic that has been widely addressed of late; e.g., Snook and Xue 2008; Dawson et al. 2010; Morrison and Milbrandt 2011).

9

It is not readily apparent from the composites what would explain this more deviant rightward motion in the tornadic cases.

10

In the present work, SCP and STP are calculated using the Thompson et al. (2004, 2007) updates to the original Thompson et al. (2003) formulas, with the exception that, because the present data extend only to 12 km, the surface–6-km bulk wind difference is used in place of the effective bulk wind difference (which approximates the bulk vertical shear over one-half of the storm depth). For the purposes of visualizing spatial variability and simple comparison among various storm subsets (e.g., tornadic versus nontornadic), the effect of this substitution is negligible.