[Ed. – David’s post is taken from his daily journal notes during his week at the HWT.]

2010_0517

I’m in Norman, Oklahoma, for EWP2010 to forecast severe weather scenarios using new technologies and software. This should be both exciting and challenging and I’m looking forward to the experience.

Much of the day was spent with an overview of the various products we will be using and testing and there is much that is new and potentially very useful. By early evening we switched into forecast mode and loaded up real-time data of the ongoing convection and severe weather in southwest Texas and southeast New Mexico. Not surprisingly, there were a few software glitches but there always are with these type of programs and we just worked our way through it. Eventually we were able to view the multi-radar data fields and the new satellite tools. These included convective initiation products and overshooting convective cloud tops. Because these are new tools that we have not used before, it takes some time to learn how to use them and how they can be used to improve severe weather warnings.

2010_0518

We started with a debriefing of yesterdays events over the southern High Plains. Next we briefed on todays expected weather which should include supercells with tornadoes possible over the southern and central High Plains.

We received a brief overview of some of the new satellite products including the simulated satellite data generated from model output. It uses NSSL 4km WRF data to simulate all IR bands and produces results that are very similar to true radar data. It is, however, very compute intensive and requires many hours to generate. It’s unlikely that we will see this product on any operational workstation in the near future.

After the EFP weather briefing, we began forecasting operations for the day. Our group of four broke into 2 groups of two with our group forecasting for AMA and the other for PUB. Within a short time, we had convective initiation but were unable to use the satellite CI products — or any others — because of excess cirrus cloud obscuring the low cloud. We switched to MRMS products and began the forecast. These new products are a challenge to use at first — as with any new product — but have great potential value. It requires that we load and these tools alongside the more conventional tools as we prepare our warnings.

There are simply too many products to attempt to use and view all.  One needs to judiciously choose a few and work with these during the forecast and warning session. To select too many will result in information overload. I suspect that the products that I selected today and that worked for me today may not be the same on another event. For today, I found the MESH and ROTATIONAL products to be useful as well as the REFLECTIVITY -20C for warning on large hail and tornadoes.

2010_0519

The debriefing today included comparisons between the warnings we issued and those issued by the NWS offices. Our team was warning for AMA and the other team was warning for PUX. There were only a few supercells and these quickly became severe and then tornadic so that warning was fairly easy. I’m not sure that any conclusions can be drawn from this event since we were competing against forecasters familiar with the area and consequently our warnings were usually a few minutes behind theirs in issuance time.

Today poses a Moderate to High Risk across portions of the central Plains and both teams will be forecasting for the OUN warning area. We sectorized by storms as necessary. The strongest storms of the day were generally north of the I-40 corridor and the other team handled most of these. We did warn on one storm in that area but our warning was almost 30 minutes later than that issued by the WFO and I believe ours was more timely. MRMS parameters suggested that this storm was not severe and not tornadic when initially warned. It’s possible that their upgrade to a warning was predicated upon the evolution of an earlier storm that quickly became tornadic. All or most of the storms in this area eventually became tornadic and VORTEX2 was operating in this area which allowed us to receive timely reports of large hail and tornadoes.

We warned on a storm that began near Lawton then moved towards Chickasha and eventually moved across south Norman. It was very slow to evolve and we delayed warnings until MRMS and base state data convinced us that the storm had finally evolved into a severe storm. First warnings were SVR, then extended, then finally TOR, extended, then dropped back to SVR. No tornadoes were reported by experienced chasers and spotters but large hail was reported 3S of NWC. I think holding back on the warning was an improvemnt on FAR for individual counties and cities early in the life cycle of the storm.

2010_0520

With the front now stalled across Texas we focus on that section of the country using KFTW as a localization. Severe storms form on the front and we warn on them using the new radar products. Not too different from the previous days. But there is one interesting feature and that is an outflow boundary and fine line moving westward. The AzShear and RotationTracks products both show this feature well and it could serve as an initition point for convection later in the afternoon. By the time we break, however, it has not yet done so.

In the evening we switch to KHSV so that we can use the pseudo-GLM (global lightning mapper). These are tools that we have not yet used and it maps all lightning channels including CG, CC, and IC.  The most interesting thing we notice is that there is substantial lightning being detected in the trailing stratiform region. The echo line is oriented north-south and during the evening a few low reflectivity notches develop on the rearward side followed by very strong winds on the leading edge of convection. It appears that rear inflow jets (RIJ) are developing under the mesoscale anvil. Reminds me of some Pre-STORM events.

2010_0521

Today we discuss and debrief on all the weeks activities. There is general consensus that the radar products are good tools and become more useful with use. We’re not yet certain of the value of the satellite products since they aren’t telling us much more that we can see with other products but it may be that there will be events in which they outperform the radar. So, other groups may see some value that we didn’t get to experience.

David Blanchard (Lead Forecaster, NWS Flagstaff AZ – 2010 Week 6 Evaluator)

I was privileged to be invited as a participant in the GOES-R/MRMS portion of the 2010 Experimental Warning Program, week 6 (mid-May). We had a fairly busy week with tornadoes and other severe weather occurring in the Amarillo and Pueblo WFO’s along with the hometown Norman WFO (the Amarillo and Norman experiment days happened to feature Vortex II providing live on-site data of the tornadoes). Going into the program, I was expecting to be more interested in the GOES-R side of the experiment; however, as the week progressed, it became apparent that the MRMS system was a more promising innovation at the present time. The GOES-R tools proved to be difficult to fully analyze because their namesake satellite with its quicker routine scan intervals had yet to be launched. GOES product issues and other technical glitches aside, the week was successful and I very much enjoyed my stay. Here is a brief collection of my thoughts on each of the experimental tools:

MRMS

This system was by far the most impressive tool premiered during the week. The technology has been around for several years, but this was the first I had heard of it. The concept is simple: avoid radar-data overload and simplify warning operations by combining ancillary radars (TDWR, CASA) and neighboring WFO radars into one streamlined product. This provides an excellent way to analyze multiple radars at the same time, provided a WFO has overlapping radar coverage. Traditional cell diagnostics such as POSH (probability of severe hail) can then be constructed from this base data, along with new products that take advantage of the isothermal plotting capabilities (i.e. reflectivity plotted on an isothermal surface).

One readily apparent drawback of the MRMS system we previewed was the sheer number of different analyses available. It was nearly impossible in the time allotted to adequately test or even plot all of the fifty or so products listed for us. I settled into using about 5 of the products and was able to try about 15 over the course of the experiment. Continued experimentation (such as the EWP program) should help whittle this down to a more manageable list when MRMS products are made available to WFO’s.

I found several of the MRMS products to be very helpful in forecasting severe hail. The traditional MESH and POSH algorithms available through MRMS performed well, both in highlighting the onset of severe hail and following its track. Curiously, the bias-corrected MESH performed the worst, being off-track by an appreciable margin for many of the storms. Reflectivity was available on isothermal surfaces; the reflectivity at the 0C and -20C surfaces in particular were handy for issuing warnings for severe hail. For tornado warnings, the 0-1 km azimuthal shear and 30-minute rotation tracks both provided valuable information of low-level rotation and tornadic history. I found that the rotation track tool gave a good first guess for shaping the path of the warning polygon in situations where the track forecast was more difficult.

One drawback of relying extensively on MRMS products is the slight data latency of approximately 2 minutes. Warning decisions which require up-to-the-minute radar data would be hampered by waiting on the next available data, which might be around 2 minutes late in comparison to the WFO radar itself. I suppose in this regard, MRMS data is probably more useful in tracking and updating existing warnings than in issuing new ones.  [Note:  The latency is a result of the experimental nature of the AWIPS set up.  An operational system, and hopefully our future EWP system, should have reduced latency.  -Stumpf]

Additionally, I suspect that WFO’s which lack overlapping radar coverage probably wouldn’t experience the full benefit of the MRMS system. In particular, it seems that the low-level shear products will suffer since some of required elevation scans might not be available at greater distances from the radar.  [Note:  The 0-2 km AGL azimuthal shear and rotation tracks products always use data from the 0.5 degree elevation scan even if it is above the 0-2 km AGL layer.  -Stumpf]

GOES Overshooting Top/Enhanced-V Algorithm & U. Wisc Convective Initiation Product

We were provided with an overshooting top algorithm which located the colder clouds of an overshooting cloud top along with the associated “enhanced-v” signature. We were also given a convective initiation product which provided four discrete values indicating the likelihood of convection over a given area. I’ve lumped the two tools together in this review because it was difficult to gauge the usefulness of either in warning operations, due to the current GOES scan interval of 15 minutes. New convection and even overshoots were often easily diagnosed by radar within the time required for a new scan. To make matters worse, the scheduled afternoon calibration and full disc scan created occasional 30 minute gaps in the imagery, further hampering the tools. Needless to say, these wide gaps in the imagery updates rendered the products difficult to evaluate. However, when the GOES-R satellite is launched, the algorithms will receive 5-minute imagery at all times of the day (up to 30-second imagery with rapid scan mode), which should greatly enhance the utility of these products. Until that time, I would say the jury is still out.

GOES-R Geostationary Lightning Mapper (GLM)

The GLM was one tool which was not actually available for the EWP, and was instead mimicked with other data to give a rough estimate of how it might behave. In the future, GLM data will give forecasters a unique look at storm activity by providing the total flash rate via a visible channel on the GOES-R satellite. This provides much more information than the current cloud-to-ground lightning data provided by Vaisala (NLDN), not to mention the benefits of public-use lightning data instead of Vaisala’s proprietary data. As previously mentioned, for the purposes of our experiment it was intended to use a pseudo-GLM (GLM output being imitated by real total lightning data) in warning operations. Unfortunately, this also required the operations to take place in locations which featured the total (3D) lightning-mapping instrumentation, which was rarely the case for our week of operations. On the one day we did have pseudo-GLM data available, the storms were sub-severe. Other weeks of operation probably worked better for analyzing the GLM, so I would defer to participants of those weeks for more information on this tool.

Darren Van Cleave (Meteorologist Intern, NWS Rapid City SD – 2010 Week 6 Evaluator)

In March  2010, I was asked to participate in the CASA (Collaborative Adaptive Sensing of the Atmosphere) portion of the 2010 Spring Experimental Warning Program (EWP) in the Hazardous Weather Testbed (HWT) at the National Weather Center (NWC) in Norman, OK (Figure 1).  CASA operates a dense radar network of four X-band 3cm radars between Oklahoma City and Lawton, OK.  These radars only have a 30nm effective range but overlap to provide multiple-radar analyses of reflectivity and velocity. For more information on CASA, see http://www.casa.umass.edu/ or the CASA IP1 Wiki at http://casa.forwarn.org/wiki/. The purpose of the CASA EWP experiment is have experienced forecasters evaluate real-time and case studies of CASA radar data.

Beginning on the week before my arrival at the NWC, I became increasingly excited because of consistent model forecasts of severe weather in Oklahoma.  I even stayed up the night before my departure to view the SPC outlook for May 10 – High Risk of severe thunderstorms and large tornadoes in Oklahoma!  When I arrived at OKC airport at noon on Monday, I immediately began coordinating my arrival with Jerry Brotzge and Brenda Philips (CASA Principal Investigators) via phone and text messages.  Brenda’s flight had also landed at OKC around noon, so we drove down together.  We had just enough time to grab a quick lunch to go and arrived at the HWT around 130pm where we immediately began reviewing the latest information.  Central Oklahoma was still under the gun and storms were developing along the dryline to the northwest of the CASA testbed area.  I don’t think I had finished my lunch yet when Brenda told me it was time to make a forecast!  We used twitter and NWSChat as our primary mediums for disseminating our forecasts, warnings, and updates.  After pegging a time of 5pm for activity to reach the testbed area, I watched the event unfold with one supercell after another developing along and ahead of the dryline.  Unfortunately, all of them seemed to develop just outside of the testbed area.  Around 515pm, one left-moving supercell storm split off to the NE and moved inside the network.  This storm contained an unusually strong anticyclonic mesocyclone (mesoanticyclone?) and hook configuration (Figure 2).  When asked if I would issue a tornado warning on that storm, I replied “No, because anticyclonic mesocyclones rarely produce tornadoes.”  At 525pm, a Tornado Warning was issued by WFO OUN for this storm.  It turned out that a six mile long tracked EF1 anticyclonic tornado had touched down at 518pm near Bray, OK and another pair of tornadoes (one anticyclonic and the other cyclonic) near I-35 and Wayne, OK.  Around this time, two LP-like supercells were approaching Moore and Norman.   As the Norman storm approached, I saw SPC forecasters run to the west windows. Being a conscientious, safety-minded NWS meteorologist, I also ran to the window and observed a rapidly rotating funnel nearly over the National Weather Center (Figure 3).  The tornado grew in size as it tracked east along Highway 9 and even damaged a few of the NWC employees’ homes.  As storms moved east and away from the network that evening, we closed operations for the day.  Between 2 and 8 pm, 31 tornadoes were confirmed across the state [http://www.srh.noaa.gov/oun/?n=events-20100510-tornadotable].  An exciting start to the experiment to say the least!  In the following days, there were several close calls to observing severe storms within the network during and after operations, but none as significant as the May 10 event.

The orientation planned for Monday took place on Tuesday.  During the rest of the week, I went through several displaced real-time simulations using 88D data only, then repeated using 88D and CASA radar data, multi-radar wind analyses and high-resolution model forecasts. The simulations included the Anadarko, OK tornado of 14 May 2009 and the Rush Springs, OK tornado during the early morning of 2 April 2010.  Without knowing any details of either case, I was challenged trying to issue timely warnings based on CASA radar data.  Without going into detail, CASA radars use adaptive scanning strategies that depend on the coverage and intensity of storms.  Data at any one elevation angle can be as frequent as every 30 seconds.  Trying to mentally process data from four CASA radars in the same way we do one 88D data was an exercise in futility.  I do not believe manual interrogation of such high-resolution radar data is a realistic option for warning forecasters of the future.

The Rush Springs, OK tornado case was very eye-opening and showed the tremendous potential of CASA radar technology to detect smaller tornadoes.  Figure 4 shows a side by side comparison of KTLX and the KRSP CASA reflectivity at the time of the tornado.  Many areas east of the Mississippi river are prone to these smaller tornadoes that develop more rapidly than those from supercells.  Trapp and Weismann (2005) more recently showed how tornadoes spin-up in the comma head portion and along the leading edge of quasi-linear convective systems (QLCS).  Tornado warning lead time and accuracy is lower for both QLCS and tropical cyclone storms than that of supercells.  A local study done in the Peachtree City WFO in 2009 showed that 13 out of 16 unwarned F2 or greater tornado events resulted from QLCS storms.  A Hollings Scholar is also studying  QLCS tornado climatology and warning accuracy this summer in the Peachtree City WFO.  So far we have determined that the initial lead time of tornadoes from QLCS storms across the mid-south and southeast averages about 25% of those from supercells (3-5 minutes vs 20 minutes).

During the week, I was able to pick the brains of some scientists. I shared presentations and concerns from the 14-15 March 2008 tornado and 21 September 2009 flash flood events in north Georgia.  I discussed research on unwarned tornadoes recently published in WAF with Jerry Brotzge, who showed how such missed events can be correlated to smaller tornadoes (as just mentioned).  I plan on collaborating further in the future.

I will certainly remember the experience I had at the EWP this year and look forward to the day when technology like this is deployed operationally.

Steve Nelson (Science and Operations Officer, NWS Peachtree City/Atlanta GA – 2010 Week 5 Evaluator)