Update on HSLC event-relative composites

Note: For ease of viewing, figures referenced herein are provided in the attached powerpoint (at bottom).

As discussed during the February 2015 CSTAR conference call, Dr. Parker and I are in the process of developing event-relative composite maps for HSLC environments associated with significant severe reports (i.e., EF2+ tornadoes, 65+ kt wind gusts, and 2”+ hail) for comparison to those associated with unverified warnings (i.e., nulls, as defined in previous work). These composites are created with North American Regional Reanalysis (NARR) data, which are available every 3 h at a 32-km horizontal resolution with 29 vertical levels. Data are temporally averaged over a 20° latitude by 20° longitude domain centered on each respective report or null. The 3D nature of the NARR offers the opportunity to calculate nearly any variable desired and represents continuity throughout the time period of our report and null datasets. Archived SPC mesoanalysis data were used in this work only to determine which reports were HSLC using our previous criteria of SBCAPE ≤ 500 J kg-1, MUCAPE ≤ 1000 J kg-1, and 0-6 km bulk wind difference ≥ 18 m s-1.

The distribution of HSLC significant severe reports (2006-2011) and nulls (October 2006-April 2011) used in this study is shown in Figure 1. Nearly every state had at least one significant severe report occur within an HSLC environment over the eight-year period of record. However, the significant severe report distribution is notably skewed towards the Ohio, Tennessee, and Mississippi Valleys (Figure 2), particularly when considering only significant tornado reports (Figure 3). Nulls, despite having a distinct maximum in the Lower Mississippi Valley, are much more evenly distributed across the CONUS (Figure 4).

The differences of typical synoptic and mesoscale features between events and nulls are striking, as shown in Figure 5. Events tend to be associated with deeper troughs and surface lows than nulls, leading to correspondingly stronger differential divergence, mid-level negative omega, low-level temperature advection, and a more pronounced upstream vorticity maximum. Buoyancy differences are relatively modest, with a mean difference in MLCAPE of 100-200 J kg-1 centered to the south of the composite report/null location.

Seasonal subsets reveal similarities to the entire dataset; however, the key feature appears to vary by season. For example, wintertime differences appear to be dominated by differential divergence near the composite center, likely influenced by a coupled-jet feature aloft and a stronger low-level front (Figure 6), while summertime differences are primarily found in low-level warm air advection and buoyancy southeast of the composite center (Figure 7). Variability in the importance of certain features is likely attributable to the annual cycle in report location (cf. Figures 8-11) and general climatology.

To assess regional variability in HSLC features, the dataset was split into three subsections (northeast, southeast, and west) based upon the latitude and longitude of each respective report and null (Figure 12). The southeast region composites, which encompass the majority of reports within our collaborating CWAs, reveal similar difference fields to the nationwide composites (Figure 13). The northeast composites, not shown, represent the weakest differences in upper-level and lower-level divergence fields and suggest reports tend to occur just south of a warm front, based upon positioning of the MSLP and MLCAPE differences. Composites in the western region (Figure 14) again suggest more of a warm front (or perhaps triple point) structure, though other fields depict differences of similar magnitude (or larger, such as low-level warm air advection) compared with the southeast composites.

As discussed during the conference call, nationwide SHERBS3 comparisons revealed that the 0-3 km shear vector magnitude was the primary constituent in discriminating between events and nulls, while the 700-500 hPa lapse rate was similar or even lower in the report composite. This is consistent with differences in the west, as shown in Figure 15 and, to a lesser extent, with those in the southeast (Figure 16). Similar differences are noted when using the 3-6 km lapse rate rather than the 700-500 hPa lapse rate (not shown). The effective shear term within the SHERBE is comparable in apparent importance to the 0-3 km shear vector magnitude in the SHERBS3. These findings suggest that the SHERBS3 and SHERBE can likely be improved upon in the future through modifications of existing terms and/or an inclusion of other terms.

Additional composites and variables continue to be investigated. For example, potential instability fields and 0-3 km CAPE do show some clear differences between events and nulls (Figure 17) in the vicinity of the composite center. Further, the 0-3 km CAPE differences line up quite well with the 1000-850 hPa theta-E differences, suggesting a conversion from potential instability to CAPE in the lowest 1-3 km. Finally, composite soundings are being generated at the composite report and null locations. These soundings have noted shear magnitude and CIN differences in all regions, the latter of which potentially for opposing reasons (e.g., lapse rates in southeast and low-level relative humidity in west; cf. Figures 18-19). Moreover, respectable differences in low-level relative humidity and hodograph shape were noted when comparing the significant tornado and significant wind composites (Figure 20), though given the spatial climatology of HSLC significant tornado and wind reports, regional analyses of these findings are necessary before any conclusions can be reached.

So far, we are confident in the following:

  • The composite environment associated with an HSLC significant severe report is characterized by considerably stronger forcing aloft and in low levels than the composite environment associated with an unverified warning.
  • On the mean, CAPE differences between events and nulls are modest, with primary dissimilarities noted to the south of the composite report/null location.
  • Similar findings were noted in seasonal and regional subsets, though the degree to which each variable is important fluctuates.
  • The SHERBS3 exhibits skill in all regions, but its signal may be swamped by the 0-3 km shear vector magnitude, particularly in the west. The 700-500 hPa lapse rate shows weak discrimination, suggesting an adjusted parameter may yield more skill in discriminating between events and nulls.

Since last month’s conference call, I have been in the process of re-running the composites to ensure accuracy and allow for the calculation of other variables. Initial steps are underway to develop these findings into a journal article to be submitted later in the year. Future work will assess additional composite fields and subsets along with the comparison of high-shear, high-CAPE environments to HSLC. Furthermore, skill tests utilizing the NARR composites will allow us to determine the utility of previously uninvestigated variables at discriminating between HSLC events and nulls, potentially improving existing forecasting parameters.


This entry was posted in CIMMSE, Convection, CSTAR, High Shear Low Cape Severe Wx. Bookmark the permalink.

3 Responses to Update on HSLC event-relative composites

  1. Matt Parker says:

    posting a non-message so I can get email alerts of any new comments

  2. Jonathan Blaes @ WFO RAH says:


    Thanks for putting together and sharing these composite charts. For me at least, there are several unique results for me so I thought I would pass along a couple of comments.

    Figure 3 which shows the HSLC significant tornado reports with shading corresponding to the kernel density highlights the unique threat of HSLC EF2 tornadoes in the Southeast especially across TN, AL, and GA. The composite difference charts shown in figure 5 nicely highlight the synoptic and mesoscale differences between HSLC significant events and nulls with the significant events containing stronger troughs and surface lows. We’ve heard from many of our forecasters about the perceived requirement for stronger forcing with HSLC events and figures 5, 6, 13, and 14 note this.

    I would mention though that the composite MLCAPE differences may be more than “modest.” Given the HSLC requirement of SB/ML CAPE values less than 500 J/kg, a difference of 200 J/Kg is a pretty sizable fraction while the magnitude remains modest. While the location of the CAPE difference is up shear of the event location, we have seen several events in central NC where convection fires in a region with more favorable instability (SC for example) and then moves into a more stable environment across NC. The storm’s more mature structure and supercell characteristics appear to allow it to maintain itself in an otherwise less supportive thermodynamic environment.

    Regarding the SHERB components, the fact that the 0-3km shear vectors was the primary discriminator is also consistent with the prejudice of these events being associated with stronger forcing.

    Still with all of this said, your first bullet is the most important… “The composite environment associated with an HSLC significant severe report is characterized by considerably stronger forcing aloft and in low levels than the composite environment associated with an unverified warning.”

    Thanks, Jonathan

  3. Keith Sherburn says:

    Hi JB,

    Thanks for your comments.

    You make a good point about the seemingly modest instability differences actually being quite substantial given our parameter space. This could certainly be associated with the high CAPE to low CAPE transition, and it may be possible to explore these sensitivities via composites taken 3 h or 6 h prior to the reports (which are a current work in progress).

    In addition to creating composite maps and soundings with the NARR data, we can examine the distribution of a given parameter (e.g., a given metric for forcing) over the entire population of events and nulls. Moreover, we can use these distributions to perform skills tests comparable to our previous parameter-based work in order to determine what other parameters (in addition to those determined using the Mesoanalysis fields) exhibit skill in discriminating between events and nulls. Initial results to this end are promising, and I hope to share some preliminary ideas at our next conference call.

    I’ll touch more on the SHERB and its components in an upcoming listserv response.

    Thanks again,

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s