Discussing RGB imagery

A forecaster from Austria visited the HWT this week and I asked him to give a short briefing on the use of satellite data over Europe to the EWP forecasters to help them understand some of the capabilities that we will have with GOES-R with respect to the utility of RGB imagery. Over Europe, the use of radar and surface data is relatively limited to the level that we use it here in the US because of the difficulty in obtaining data from the private companies, not to mention other neighboring countries. Scientists and forecasters in Europe have found innovative ways to combine imagery into products that help identify surface features, cloud types, volcanic ash, as well as moisture and jet streaks. These techniques are not commonly used within normal WFO operations within the US, mainly because of the lack of spectral data available from our current operational satellites. EWP forecasters got the chance to have a look at what they use routinely over Europe and found the briefing very interesting and are excited to use this sort of information in the future.

End of week debrief... 27 May

Today we had our end of week debrief for the EWP. We did have a fairly robust discussion yesterday regarding the 24 May event, so we tried to gather some additional information from the forecasters. We covered each of the products and how they performed within warning operations this week.

Convective Initiation

- (26 May event) CI did not so well over AL/TN... high FAR, low POD in morning... Forecaster theorized this may have something to do with the less dramatic temperature differences between the surface and clouds during the morning hours.

- CI did much better during rapid scan.

- The UAH version was much more agressive that the UW version.

- Having a probabilistic approach versus a yes/no would help.

- Forecaster used the UW CTC/CI product to issue a severe weather statement... ended up putting a warning on it afterwards.

- This could be important for not just severe weather... the CI products could be very useful for the onset of lightning as sort of a proxy for the growth of a certain dBZ threshold above say a -10 C level if it had faster updates.

- "I'm assuming that performance should improve pretty dramatically once you get the rapid updates with the next-generation satellites, but now I guess it would work best in a clean environment in the plains. I could also see the probabilities a good way to go."

- Maybe contouring SATCAST probabilities would be a good.

- "Based on what it's designed to do, I can see this working in a typical summer afternoon in Florida, not just over the plains."

- Could help identify waterspout candidates because they're harder to see on radar.

- For non-severe faster moving systems it may be useful in detecting regions of heavy rain.

- Would be a good idea to get west coast offices to look at these things since they rely heavily on satellite data because of a lack of surface observations.

Pseudo-GLM

- When we saw cell mergers there was a rapid increase in flash rate over a 5 minute time period and updraft speed from the 3D-VAR analysis.

- A lot of times we would have flash rate increases over the anvil areas downstream... it could help you focus on the new electrification of the storm as well as where new cells might develop or updrafts cores move... would give 10-15 minutes lead time before it showed on radar.

- 10-15 minutes lead time on the first CG.

- "There's a lot of potential use for these types of products... but there is definitely more room for improvement with additional research, as well as increased temporal or spatial resolution."

- It may be difficult to display the rate of change product, not everyone will be looking at the same storm, so having a gridded history would be really useful that you could click on and get an idea of how that storm has evolved.

- Having polarity information would be very useful.

- "Need more research on the forecasting applications of this data."

- Would be interesting to see the PGLM over mini-supercells as well as some winter cases looking at rainfall rates and updraft strengths.

- Ratio of IC-CG would be very useful... being able to query a cell or cluster for it's trend would be helpful.

- The classic MCCs, it would be interesting to see how that related to severe and heavy rain potential.

- Using the PGLM might be useful for a poor man's microwave information in distinguishing areas of convective and trailing stratiform.

- Would have a lot of utility in mountainous regions where flash flooding is a big issue, especially if there is no radar coverage.

- The sum product was not very useful, mainly because of the color scale... everything becomes white... Forecaster used it as a sort of storm total tool, much like precip.

Overshooting-top / Thermal Couplet

- (26 May event) One couplet was over some leading cells along the PN/MD line... the clouds behind that were masked by cirrus and may have limited the detection. When it did trigger, it well differentiated that cell from the rest of the scene.

Nearcast

- (26 May event) The theta-e and precipitable water differences really indicated the marginality of the storms in the foothills... where the maximum stuff intersected that is where we saw the most sustained convection and highest flash rates. It was definitely a good indicator of flash flooding over the area.

- "A lot of the time to increase my lead time in the morning, I like to take a look at PW and WV... so I found that this was a nice utility because it was indicative of finding areas of greatly deep instability, or moisture source regions."

- "One on storm on tuesday, there was strong theta-e gradient that the storm was moving into and that gave me confidence in that the storm would intensify."

- It's a simple way to identify areas of warm advection and instability... this is why the forecaster found it useful in warning operations.

NSSL-WRF Band Difference

- The band difference has a lot of potential... you can get a head start by looking at the trends in the data that help you anticipate what's going to happen.

Overall

- "I enjoyed seeing the combination of the satellite products interacting with the radar and lightning products... it's a great planning tool."

- Need to come up with some pre-made default procedures for the experimental products... it's tough for forecasters to come in cold and learn how to load all of the products and then go into forecasting operations. The forecasters especially applauded the creation of the "ultimate-CI" procedure from last week and that should be saved.

- Create articulate presentations that forecasters can view beforehand rather than having "powerpoint death" on Mondays... this will also maximize the time forecasters have to look at the products.

- Would also be useful to create very short WES cases to send out beforehand that the forecasters can go through before they arrive.

- Forecaster would like to see the derived sounder products within the HWT AWIPS next year.

Psuedo-GLM Products in the Mid-Atlantic

Forecasters conducted warning operations in the Sterling, VA and State College, PA CWAs, which allowed for analysis of the pseudo-GLM products (from the DCLMA). Although the earliest convection initiated along and just west of the Appalachian Mountains (out of range of the LMA), storms soon formed within range. Isolated storms initiated and developed rapidly ahead of the main line. The pseudo-GLM products detected IC flashes ~10 min prior to the first CG flashes in two of these storms. These storms continued to intensify and eventually merged into a line, further increasing the flash rates.




As the discrete cells merged into a line, flashes remained frequent, and the pseudo-GLM products helped confirm the strength of individual cells within the line. Greater flash rates also helped identify areas of new convection as they formed along the line.



Forecasters also observed that the pseudo-GLM swaths (i.e., 60 min sum of flashes) helped to illustrate persistent cells and identify the most intense portions of the line.



The above screen captures illustrate several four panel displays used to compare individual products in AWIPS. The pseudo-GLM products often were plotted alongside the 3-D Var updraft and vorticity tracks, as well as the multi-radar multi-sensor hail swaths and reflectivity at -10 C.

Scott Rudlosky

EWP forecaster debrief... 26 May

This afternoon we had an extensive debrief session with the EWP forecasters regarding the tornado outbreak from this past Tuesday (24 May). We asked them in detail what they saw for each of the products, and these are the responses/comments we received for the GOES-R products...

Convective Initiation

- We were seeing 10-15 minute lead times from the UAH CI product along the dryline prior to any echoes above 35 dBZ on radar. The UWCI was much more conservative and missed a few instances of CI, but it had less false alarms and similar lead times when it did trigger for CI.

- UAH CI did show some signals after initiation behind the dryline, but nothing really continued to grow. However, forecaster mentioned how this would be very useful in warning operations to help increase situational awareness for future development when you may be focusing primarily on the first storms.

- Forecaster mentioned how he was watching the UAH CI this morning over the SE and it was giving negative lead times and UWCI was not flagging anything at all. However, he did want to emphasize that on Tuesday GOES-E was in rapid-scan operations and was not this morning, so that could be why the lead times were so poor.

Pseudo-GLM

- Very useful as a situational awareness tool during warning ops... Forecasters saw rapid increases in the instantaneous flash rates prior to increases in reflectivity and other products such as the 3D-VAR updraft strength and various MRMS products.

- Forecaster mentioned that it would be nice to have a line graph display for this to help identify jumps better. (Kristin and I did mention that we have heard that forecasters would not like this in the past since it would remove them from the D2D during warning ops... forecaster responded with that you can do this as a percentage change product within the storm interrogator software within AWIPS which could mitigate this.)

- Forecasters found that the track product (that we currently use as a proxy for a jump graphic) was not very useful in detecting 'jumps' mainly because of the resolution of the product... the rapid changes could not be seen because they would generally overlap on the same pixel.

- There were moments that some data outages occurred within the OKLMA network that caused some false instances of 'jumps' and lack of signal. It was theorized that the sensors were becoming attenuated in heavy rain and not transmitting data since they are daisy-chained. Also, one tower was hit by a tornado and caused a large break in the chain, so the OKLMA is down until further notice, possibly throughout the remainder of the experiment.

Overshooting-tops / Thermal Couplet

- Not many were seen, but those that were tended to be near the center of the upper low.

- Those that were seen did coincide with instances of increases in reflectivity aloft.

- Usefulness within warning ops very minimal because of lack of detections.

CI over New York

Another busy day for the EWP forecasters, as severe thunderstorms have erupted over the mid-Atlantic region. As so, screen captures from AWIPS have been tough to come by. Therefore, these next screen shots come from NAWIPS.

SATCAST CI has performed very well over the New York and Pennsylvania areas this afternoon. The following satellite image is from 1945 UTC. SATCAST forecasted CI over south central New York. Approximately 13 minutes later, radar reflectivity exceeded 35 dBZ on the base scan. While not the largest lead time, in another large CAPE environment, SATCAST is performing well.


SATCAST is performing quite well today pinpointing areas where storms will develop in the future. Even in Florida, where the dynamics are not as impressive, storms developed with help from convergence by the sea breeze, and SATCAST provided a very good forecast.

Pseudo-GLM Perspective on Oklahoma Storms

This week started with two very active days within the OUN CWA which allowed EWP forecasters to incorporate pseudo-GLM products into their warning operations. Despite the dense radar coverage, forecasters used the total lightning products and provided some interesting insights.

Day One – Monday 23 May
Two storms developed in Northwest Oklahoma and exhibited very different lightning and radar signatures. The northern storm formed in Major County, exhibited very large IC flash rates, and produced large hail. Conversely, the flash densities were much smaller in the southern storm which produced a brief tornado. This observation illustrated the variability between two storms that occurred in a similar environment and emphasizes the importance of continued research on the relationships between lightning and radar within individual storms.


The greatest pseudo-GLM densities (> 30 flashes/ km/min) accompanied a large storm cluster which followed the merger of two strong storms. The 3D-Var products indicated a strong updraft throughout the length of this storm cluster which helps explain the high flash rates.


Day Two – Tuesday 24 May
The high risk forecast and model-derived products all indicated a very active day was in store. The CI products identified the initial convection 15-20 minutes prior to the first pseudo-GLM signatures. Flash rates increased rapidly in the earliest convection, and the first tornadic storm occurred as two isolated storms merged near Weatherford, OK. Flash rates spiked as the two storms merged, coincident with an increase in mid-level rotation. These combined observations increased forecaster confidence as they issued the first tornado warning of the day.


The EWP forecasters observed that the greatest pseudo-GLM flash densities consistently tracked ahead of the main updraft and actually helped to identify changes in storm motion. On several occasions the pseudo-GLM densities also indicated that the main center of rotation was shifting prior to the identification of new rotation tracks by the multi-radar multi-sensor algorithms.


Although lightning jumps preceded many of the tornados, forecasters commented that they would have liked to examine time trends for individual storms. This has been a common theme during previous spring experiments, and was not fully accounted for by plotting swaths of the pseudo-GLM products.


- Scott Rudlosky

Overshooting tops are being detected with the severe/tor warned storms over Indiana

The UWOTTC algorithm is consistently identifying overshooting tops over the state of Indiana. The above sequence of images (valid from 1902 UTC to 1945 UTC) shows numerous overshooting tops and their associated visible imagery counterparts. No thermal couplets have been detected with any severe storms as of yet, but this is not a surprise with how hard it is to detect the couplet features with the course resolution of GOES imagery. ABI is how long away??? See NWS all hazards image capture (top image) to see associated concurrent warnings.

UWCI consistently capturing best developing convection in central and north central Arkansas

The above sequence of images shows how UWCI nowcasts (upper right panel of images) are consistently picking out the best developing convection. Without the bane of cirrus affecting the algorithm processing, cases such as this show that satellite products can be of good use for situational awareness.

SATCAST CI detected underneath cold core low

Oklahoma was under the gun yesterday, but is under a completely different convective regime today. As the cold core low continues to shift eastward, clear morning skies have created an area of convective instability that has led to convective shower development.

SATCAST has been developed to try and diagnose convection of all types, not necessarily areas of severe weather. SATCAST forecasted CI in northern Oklahoma at 1832 UTC (upper panel of top image). 38 minutes later at 1910 UTC, radar reflectivity exceeded 35 dBZ at -10 C (bottom left panel of bottom image). No severe weather will come of this activity, yet it provides forecasters with the ability to let the public know via short term forecasts that convective showers are developing, and changes to outdoor plans may need to be changed.

Overshooting tops detected over central MO

Overshooting tops are being detected in central MO. The above image (valid 1855 UTC) and an earlier image shown below (valid 1832 UTC) show the OT detections (red spots) superimposed over their associated regions in the visible.

These OTs are associated with storms warranting tornado warnings over the same area. See warning image below. No thermal couplet signatures have been detected as of yet.