On May 8, 2003, several tornadic storms ravaged portions of the central plains. Among those, a tornadic supercell passed through portions of the southern Oklahoma City metro, causing up to F-4 damage at several points along its path. Preliminary information on this event is available from the NWS here. A unique data collection opportunity accompanied this event as the KOUN polarimetric radar was able to collect high resolution data as the storm tracked just north of the radar site. Here, some of the polarimetric radar observations from this event will be presented, with an emphasis on the opportunities for data assimilation and applications to short-term numerical weather prediction to improve warning lead time.

Several case studies have examined the usefulness of polarimetric information for providing insight into precipitation and microphysical characteristics, among other factors. Further, there is also research investigating polarimetric signatures associated with tornadic vs. non-tornadic storms. If such distinctions exist in the data, then it may be possible via data assimilation techniques that use polarimetric information to improve prediction of tornadic storms. Certainly, model studies have shown strong links between microphysical characteristics of a storm and its subsequent behavior, providing optimism that such a technique might prove successful.

There have been a few recent studies that have focused on this event, including reviews of the storm structure, data assimilation, and subsequent forecasting. A quick summary of some of the synoptic and mesoscale environment follows.

At the surface, numerous boundaries were present across the central plains by ~22Z creating a myriad of potential forcing mechanisms for convection - and in fact many convective clusters. Notably, the line of tornadic supercells in central Kansas is along the prefrontal trough, not the dryline as was the case of the convection further south in Oklahoma. Using the 9 May 00Z upper air analyses, a rather uninteresting upper jet with largely active upper jet dynamics across the northern plains with a jet streak sliding up the east side of an upper trough. Meanwhile, a potent shortwave was emerging beneath this at 500 mb. At 850 mb, modest warm air advection was also occurring across much of the plains with deep moisture just ahead of the thermal ridge. Back to the surface, a plot from the Oklahoma Mesonet of the surface storm environment around the time of the first tornado is available here. Clearly, the cell has by this time advanced ahead of the dryline, surface thetaE is nearly constant in the low 360's (ignore the local maxima sw of the storm - note the clearly faulty dewpoint at the maxima) with surface dewpoint depressions of only 8 F in the storm inflow.

Quick analysis of the soundings from OUN show a few noteworthy features. First the 12Z sounding demonstrates a deep moist boundary layer with a capping inversion near 800 mb and fairly steep mid-level lapse rates above. By 18Z the height of the base of the low-level inversion has lowered to near 840 mb, with modest cooling in the layer from 800-700 mb along with warming at mid-levels (centered near 500 mb). Also, the wind profiles at nearly all levels have backed to more southerly and have increased markedly in strength near 600 mb. Finally, by 00Z after the occurrence of the event, the boundary layer depth has continued to decrease, presumably owing to subsidence as the potential temperature at the max of the surface inversion remains the same. Further, an additional inversion starts to take shape between 700-600 mb, with warming above and cooling at the bottom of the layer, and further downward increases in wind speeds along with marked increases in the degree of veering in the low-level wind profile with veering above 800 mb and backing below the low-level inversion.

First echo of the day (on base scan of KOUN) in central Oklahoma appeared at 2022Z (cell A), and only slowly intensified. By 2055Z, several new cells began to develop northward along the dryline in elongated SW-NE bands (new cell B cluster). Note that the boundary layer rolls evident in the reflectivity (NNW-SSE reflectivity bands in lower right corner) are nearly perpendicular to the orientation of these enhanced reflectivity bands. Among the newly developing convective cells is the one that eventually becomes the storm of interest (1), which appears to be south of a small bulge in the dryline. Along the enhanced reflectivity band that the tornadic cell develops on, three initial cells (1,2,3) has a merger of the NE pair (2,3), followed by a split of the southwest-most cell near 2114Z (1(right), 11(left)). The left split continues on and becomes a severe hail producer into the OKC metro. The right mover engulfs numerous small echoes that develop upshear of the main updraft and appear to feed from the SW along the same enhanced reflectivity band, while the cluster as a whole draws progressively east of the dryline boundary. Check here for an QT animation of the sequence below.

Time series of base reflectivity scans of the early evolution of the convective event. Overlain are highlights of the dryline position (gray transparent line), the enhanced convective bands (dashed), and main cell cluster labels. Note that the data used here is raw (no QA or ground clutter algorithms have been used here) - so the data contains considerable noise particularly near the radar site.

Zooming in on the developing tornadic storm, by 2141Z the split cells are gaining separation, and a gust front starts to appear along the upshear side of the storm, with enhanced convective cells along the flanking line along with a separate convergent band of small echoes from the southwest collecting into the right flank of the storm's updraft. By 2153Z, a convergent boundary appears enhanced with perhaps two regions of gust front push. Low-level reflectivity pattern is increasingly wrapping around into a hook appendage shape, but at this time there are no longer feeder convective cells on the southwest flank. Also, the main cell begins to undergo a second split - at this time appearing in reflectivity as a small spike on the NW flank.

1.5 degree elevation reflectivity scan along with base radial velocity scans for 2141Z and 2153Z. Green triangle marks severe hail report location (0.88" at 3:40 pm CST), which appears to have been from the left split now north of the report location shown. Overlain black line marks the apparent location of the near surface gust front or convergence boundary.

Structure continues to rapidly modify as the cell splitting becomes more obvious, and near the surface at least the velocity pattern surprisingly suggests a cyclonic circulation. The first brief tornado report follows shortly thereafter - as well as several new reports of severe hail, one from the left split cell and the other within the forward flank of the right mover. The second tornado report, four minutes later, was also brief. Like the first, this tornado developed on the head of the progressively deformed gust front surge, but the second event finds the circulation has slid to the north side of the gust front head. The second tornado was then followed by a 6 minute duration damaging wind event shortly after the updraft crossed east of the Canadian River. As an aside, early tornado and severe wind reports appear to lead the radar observed positions of features by several minutes.

1.5 scan of reflectivity and base scan velocity at 2159Z and 2205Z, with tornado reports (red squares mark start positions) and a damaging wind report (cyan, end position).

The official storm data indicates two sequential tornado events began at 4:10 CST and another at 4:15, the first producing F3 damage whereas the second produced F4 damage. Interestingly, the storm survey from the NWS accessible from the link above has this as a single long-track tornado. The 2211Z reflectivity and velocity shows the primary tornadic circulation remains on the northeast tip of the gust front surge, which is now wider (N-S extent of bulge) but is not leading the northern edge as much as in the previous sweep. There is perhaps some evidence of misocyclone circulations along the gust front edge which will need to be investigated further later on to check for vertical continuity. The reflectivity shows that the gust front has now undercut a portion of the flanking line, which again is starting to show some small feeder cells. A strong 'S' pattern is now clearly evident in the hook echo reflectivity. Echo intensity in the forward flank of the cell has also increased markedly.

1.5 scan of reflectivity and base scan velocity at 2211Z, with starting tornado report (red square).

Polarimetric observations can be used to retrieve the microphysical characteristics (such as size, shape, phase, and concentration) within each sample volume (e.g., Vivekanandan et al. 1999). Further, observational studies have found some evidence that significant tornadoes are increasingly likely with supercells having weaker cold pools (e.g., Markowski et al. 2002). Also, numerical studies have demonstrated that cold pool strength can be tied to the microphysical characteristics of a storm - in particular when ice number concentrations are large, the smaller particles are carried further downshear before falling as rain, as such less precipitation is present close to the storm's updraft, and hence a weaker cold pool on the backside of the storm (e.g., Gilmore et. al. 2004).

There were some environmental factors to consider that may well relate to the microphysical makeup of this storm event. Relatively few large hail reports accompanied this cell as it tracked over heavily populated areas, with maximum hail size reports of 0.88". The first hail report was 20 minutes prior to the first tornado, with a cluster of reports in the forward flank of the storm just prior to the first touchdown. Further, haze was reported in the pre-storm environment, with visibilities of only around 5 sm. In essence, this polluted air indicates a high concentration of CCN and potentially IN, which would favor warmer cold pools coalescing the arguments above.

Limited observations were made of the cold pool, however, as the tornado passed within 3 miles of the Tinker Air Force base, surface data was collected for at least a portion of the cold pool. A meteogram from KTIC was constructed from the raw METARS. Despite some cooling as the storm approached, modest moisture increase led to a slight increase in equivalent potential temperature in the inflow. The forward flank gust front passed the observing station as the tornado came within 5 sm and then passed just 3 sm southeast of the station as the air pressure spiked and the temperature and dewpoint rapidly dropped, with surface wind gusts to 58 knots. Cloudiness may have played a supplemental role in the environment as well, to be considered later. The visible satellite at the time of the tornadic event shows lots of high level cirrus overlaying the region.

Below is a series of plots of the polarimetric variables during the tornadic phase of the May 8 event. Click for larger versions of the individual plots.

Retrograding to 2153Z (7 minutes before the first tornado touched down), differential reflectivity (Zdr) demonstrates relatively high values throughout the echo (and a lot of ground clutter), suggesting mostly rain with little hail. The large positive Zdr values particularly along the southeast flank suggests large raindrops, though a weaker value is noted along the Oklahoma/Canadian border. This pattern is further corroborated by the specific differential phase (Kdp), which highlights a region of particularly intense rainfall in extreme southwest Canadian county, and high correlation coefficient (Rhv) throughout the echo. The hydrometeor classification algorithm (HCA) output captures this and provides a summary view.

Differential reflectivity, specific differential phase, cross polar correlation coefficient and hydrometeor classification algorithm output at 2153Z (see reflectivity and velocity above).

By 2159Z, the notable field changes is for an expansion in echo size as the cell split begins to take shape, with a shift to heavier rainfall along the western portion of the echo (from predominantly large raindrops) indicated by the lowering of Zdr, however large positive Zdr values sharpen along the southern edge of the echo. Kdp shows an elongation in the heavy rainfall through the center of the echo, with the heaviest rainfall now shifting into southwest Oklahoma county. Further, a few pockets of lower Rhv are evident, particularly between the splitting cells and within the left split - which agrees with the reports of marginal severe hail in those areas. Again, the HCA well summarizes the above.

Differential reflectivity, specific differential phase, cross polar correlation coefficient and hydrometeor classification algorithm output at 2159Z (see reflectivity and velocity above).

At 2205Z, the trend toward lower Zdr values along the western flank of the supercell continues, particularly along the northwest side and in the hook echo region, and further sharpening of the high Zdr along the southern edge. Kdp shows fairly little change, though some decrease in the rain core in the northeast portion of the hook, with the main forward flank rain core showing an upward trend. Rhv values are again fairly high, though with some decreases along the northwest edge of the echo. The HCA shows some upward trend in Hail detection in the western portion of the cell and in the forward flank of both cells, continued decrease in large raindrop detection in the hook echo region, as well as a sharper heavy rainfall boundary along the southern edge of the echo.

Differential reflectivity, specific differential phase, cross polar correlation coefficient and hydrometeor classification algorithm output at 2205Z (see reflectivity and velocity above).

By 2211Z, the southern Zdr axis has rotated dramatically clockwise, along with an increase in Zdr values in the hook echo region for the first time in this series. Kdp demonstrates the main rain core has slid further downshear, and Rhv shows a marked decrease along the southern edge of the forward flank. The HCA then shows increasing trends for hail along the southern edge of the echo, as well as a trend toward less heavy rainfall to large drops within the hook echo region. Further, light to moderate rain associated with a new convective cell along the flanking line at the tip of the hook is also evident.

Differential reflectivity, specific differential phase, cross polar correlation coefficient and hydrometeor classification algorithm output at 2211Z (see reflectivity and velocity above).

Below are the summary composites of the low-level plan view polarimetric features. Some of the more striking features are as follows:

• The axis of divergence is generally along the northwest edge of the maxima in Kdp
• Over the time period, the divergence axis slides upshear closer to the storm's updraft region
• There is a general pattern of the strongest surface divergence in a region with large gradient in Zdr
• A negative Zdr region on the northwest edge of the echo tracks the cell split, as well as appearing with other concave reflectivity patterns on the western edge and is accompanied by lower correlation coefficient
• A weakness in the correlation coefficient appears after tornado development just upshear on the hook echo
• A Kdp and Zdr maxima couplet elongates toward the hook tip and slides toward the southern edge of the reflectivity maxima prior to tornadogenesis

Composite views of the key polarimetric features at low levels for 2153Z-2211Z. Blue line indicates axis of divergence from radial velocity, with D marking regions of maxima. Other variable naming conventions are the same as used above.

Thanks to Kevin Scharfenberg of NSSL for providing the data and Mary Haley of NCAR for helping with a starter NCL script for plotting the data. If you have questions about this work please contact Glen Romine.