1. Operational Use of SPC’s Storm Scale Ensemble of Opportunity Bill Martin November 2014

2. NWS Glasgow’s verification comparator, which allows the rapid comparison of forecasts and verification of past events. With this tool, an archive is maintained of various SREF, SSEO, and analysis fields, along with LSR maps. Any two of the available images can be loaded and rapidly compared each day for the last 2 years. By this means, a great deal of insight has been obtained about SSEO performance.

3. This case, chosen at random, shows not bad agreement between the forecast MAX updraft from the SSEO, and the resulting verification. For example, we can view the SSEO MAX-updraft field versus the SPC Day 1 outlook and LSR plot every day for the last 2 seasons in about 5 minutes.

4. Most typically, we compare SSEO forecasts for the 24 hour period from 12Z to 12Z, with the severe weather reports received during that time, and the SPC DAY1 Outlook for that period issued at about 6Z. Below was the forecast from 6/1/2013, which was a tornado outbreak day in Oklahoma, also well-forecast by the SSEO MAX updraft.

5. The Max Updraft (W) plots seem to be the best proxy for all kinds of severe. Theoretically, this makes sense as it can be thought of as realized CAPE. This example is for June 12, 2013.

6. The Max Updraft Helicity (UH) plot is a proxy for supercells/tornadoes. In this case, reported tornadoes in SE Colorado verified area of MAX UH from the SSEO (a forecast with a lead of 36 hours). Tornado reports are problematic as it is common to get small tornadoes reported from otherwise benign systems (as in SE North Dakota here).

7. However, Max UH may not be the best proxy as it is subdued for high-shear/low CAPE situations. This example for the Nov. 17-18 2013 outbreak shows Max-UH on the right and LSRs/Day1 SPC outlook on the left. The case was well anticipated by SPC, despite the modest Max-UH values. The case had 110 kts of deep layer shear, but very marginal instability. There were 73 tornadoes, including 7 EF3, 2 EF4 and 8 fatalities. Traditional tornado parameters were also subdued for this case due to low instability.

8. The Max 10-m Wind Speed is a proxy for severe straight-line winds at the surface. In this case, the SSEO Max 10-m Wind Speed predicts high winds over an area from Wisconsin to Arkansas that was missed by SPC. However, it also shows high winds just east of the Rockies, an area that saw no convection.

9. Most of the large misses by the SSEO are for cases with strong winds. In this case, most severe wind reports were missed by the SSEO Max 10-m wind plot.

10. And by the plot of MAX W (and missed by SPC).

11. Another example where most of the severe wind reports were outside the area outlooked by SPC, and outside the area of strongest MAX-updraft from the SSEO

12. Both MAX and Probability plots can be usefully consulted. In this case, the MAX UH is exceptionally elevated in Oklahoma, which had a tornado outbreak, including a large EF-4 tornado. It was also elevated in Illinois, which also had tornado reports. Notably, the MAX UH was weaker in central Missouri, an area without tornado reports.

13. This is the probability of MAX UH > 25 m^2/s^2 for the same case as the previous slide. This shows a maximum over central Missouri, in contrast to the MAX UH plot. This means most of the models had strong UH over that area. However, areas of strongest MAX UH were a better indicator of the strongest supercells. By analogy with PPT forecasting, probability plots are like POP, and MAX plots are like QPF.

14. The SSEO is also useful for assessing the timing and movement of severe weather. Here we see initiation off the mountains around 21Z, propagating eastward across Montana through 3Z. 21Z 00Z 03Z

15. Same case as previous slide with radar verification at 0Z. This case had a remarkably good agreement between forecast max updrafts and location, pattern, and timing of areas of convection. Below shows 0Z radar (left) versus 21Z-0Z max updraft. This was a 30 hour forecast.

16. Same case as previous 2 slides, but with 24hour MAX W. Storm motion vectors can be accurately deduced from 24 hour plots. This shows mostly northeast storm movement, with one notable right moving supercell.

17. We deduce this was a supercell as the UH was elevated for this cell:

18. Composite parameters can be problematic. They depend on a few parameters deduced from soundings, and do not account for the great variety of storm structure that convection-resolving models can account for. They also can not account for the effects of capping, the influence of clouds, outflow boundary interactions, etc., all of which storm-scale models directly deal with.

19. For this outbreak case, the SIG TOR parameter from the 3-hour SREF forecast looks pretty good, reaching a maximum value in central Oklahoma where tornadoes were the strongest.

20. However, the MAX-UH SSEO 24-hour forecast does a better job:

21. The ability of storm-resolving models to handle vast amounts of complexity potentially makes parameter-based forecasting obsolete. Of what value is CAPE, if actual updraft strength can be directly determined?

22. Most of the time, specific cells from the high-res models do not verify. However, sometimes they do. This case from July 15, 2013 had one ensemble member with a supercell (seen in MAX UH above) that tracked almost exactly where a tornadic cell occurred in northeast Montana.

23. Performance during the 2014 AWOC Severe Forecast Contest 2 Forecasters from WFO Glasgow made heavy use of the SSEO while participating in the 2014 AWOC Severe Weather Forecast Contest. This contest involved forecasting Day 1 severe weather type and precise location; forecast elements that are directly addressed by the SSEO. These 2 forecasters placed third and seventh out of 39 forecasters. Considering the very strong random element involved, this was a pretty good result. These forecasters were making precise severe weather forecasts daily for over a month. They found that using the SSEO alone was not enough to do the forecast. Combining it with the HRRR was found to be very valuable. If the latest and last few HRRR runs agreed in detail (for example, if they both tended to have a severe cell in about the same place) a high level of confidence in the forecast could result. Still, the forecasters found themselves using parameters and the latest satellite and observational data to refine the forecast, especial at times where the HRRR and SSEO did not provide a consistent picture.

24. HRRR - SSEO Agreement Can Be Particularly Powerful. In this example, SSEO Updraft Helicity and HRRR dbZ forecast both indicate a supercellular storm in the same location in Colorado.

25. Resulting Reflectivity from KFTG at 0Z 8/26/2014 Good forecast location, though in this case resulting storm not super-cellular.

26. For winter weather, we still use the SSEO, primarily for the hourly precipitation maps and precipitation types. These are also valuable to compare with the latest HRRR. Hourly plots are very valuable for timing the start and end of events. This shows precipitation amounts and type for a winter case. On the left is the 18 hour forecast from the SSEO and the right is the 4 hour forecast from the HRRR, both for the same synoptic time of 6Z on 11/26/14.

27. Summary  THE SSEO is the state of the art in forecasting convective weather. It is useful for answering questions of initiation, timing, strength, and kind of severe weather.  MAX W is the most useful proxy for all kinds of severe weather. MAX UH was also valuable for tornado forecasting. MAX 10-m wind worked sometimes, but was not as useful. Most of the failures of the SSEO to successfully predict severe weather were cases of predominantly straight line winds.  MAX W was a better proxy for severe weather, by far, than any other forecast method, including composite parameters and things like CAPE and Helicity.  Combined MAX plots are analogous to QPF, while probability plots are analogous to POP. Both can be useful at times.

28. Summary Continued  Individual model realizations do not often verify, that is, the specific storms shown by a model usually do not happen; but the possibilities shown by the overlapped ensemble is very powerful for gaining insight into storm type and probability.  During the forecast process, the SSEO is so good, that it is tempting to go there first, to see what will happen; however, the SSEO is not perfect, and it is important to understand the broader situation first, and use the SSEO to show you the possibilities. SPC’s outlooks are still better than the SSEO.  The future would seem to be bright for storm-scale ensembles, which will make parameter-based forecasting obsolete. Desired improvements to the SSEO include having it run more often, with more ensemble members, and with ensemble members of higher resolution. Also having an updraft rotation parameter instead of updraft helicity might be more sensitive for supercells.