PGLM Assists in Severe Thunderstorm Warning (Repost)

Animation from 12 May 2011 covering from 2211-2226 UTC. The PGLM flash extent density is on the left with the corresponding radar reflectivity on the right. To see the animation, please click on the image.

Today’s afternoon shift started with forecasters working across the Norman, Tulsa, and Little Rock county warning areas. With the some storms beginning to form south and east of Norman, Oklahoma, it was felt this would be a good opportunity to take another look at the PGLM flash extent density observations and focus on total lightning. The PGLM flash extent density was very useful in identifying when the first cloud-to-ground strikes would occur. The PGLM was preceeding the first cloud-to-ground strike by approximately 30 minutes today.
As the the afternoon progressed, the storms began to intensify, both on radar and with the PGLM flash extent density and we shifted from using the PGLM for lightning safety and moved into warning operations. By 2211 UTC on 12 May 2011 (the first image of the loop shown above), three severe thunderstorm warnings were in effect. The area of interest for this post is in between the two existing warnings in the west. At 2211, the PGLM flash extent density was no more than a few flashes per minute. By 2214 UTC the number of PGLM flashes was already approaching 40 per minute. This continued to rapidly increase through 2220 UTC when the PGLM flash extent density observe 82 flashes in a 1 minute interval for a single 8×8 km grid box. This was one of the largest lightning jumps of the day with an increase of 75 flashes per minute in a nine minute time span. With this major lightning jump, along with the forecaster’s interrogation of radar data, a new severe thunderstorm warning was issued at 2226 UTC. This warning was later verified with several severe hail reports.

Total lightning preceeding the first cloud-to-ground strike (Repost)

As we watched the storms move through central Oklahoma today a small, isolated cell developed over Lawton, Oklahoma. This storm conveniently gave us the opportunity to show the effectiveness of total lightning observations in helping gain lead-time ahead of the first cloud-to-ground lightning strike. This small cell turned out to be even more interesting as the PGLM observations gave a 29 minute lead-time over the first cloud-to-ground strike. This was pretty remarkable as the lead time is usually on the order of 5-10 minutes. Below are three images showing the event.

FIGURE 1: A four panel display in AWIPS from 2055 UTC on 11 May 2011. Going clockwise from the upper-left is the radar reflectivity, PGLM flash extent density, PGLM maximum flash density, and NLDN cloud-to-ground lightning strike observations. A single flash just southwest of Lawton (KLAW) can be seen in the PGLM flash extent density and no cloud-ground strikes are observed with the Lawton cell.

FIGURE 2: The same as FIGURE 1, except for the time is 2100 UTC. The PGLM flash extent shows two flashes and the radar reflectivity has strengthened.

FIGURE 3: The same as FIGURE 1, except for the time is 2124 UTC. The radar reflectivity has increased more and the PGLM flash extent density shows several flashes. The NLDN cloud-to-ground lightning observations finally shows a single, negative cloud-to-ground strike just to the northeast of Lawton, Oklahoma. This PGLM gave a tremendous 29 minute lead time on this first strike.

PGLM data and lightning safety (Repost)

Central Oklahoma had several strong thunderstorms move through the region and the forecasters at the Spring Program had the chance to check out the pseudo geostationary lightning products derived from the Oklahoma lightning mapping array. Most of our time was spent investigating the products and discussing the various pros and cons. The figure above shows a good use for these data in a lightning safety perspective. The 1-minute PGLM flash extent density (and the corresponding NLDN cloud-to-ground lightning data) are tightly clustered with the the stronger convective regions, indicated by strong radar reflectivity. However, unlike the NLDN data, the PGLM flash extent density still showed that lightning flashes were extended anywhere from 8-32 km into the stratiform region. This shows the advantage of seeing the spatial extent of lightning activity available from total lightning observations. This is further emphasized with the PGLM maximum flash density in the upper-right which shows the maximum PGLM for each grid box for the past 60 minutes. This shows that most of central Oklahoma has had lightning activity within the past hour, indicating that the threat of a cloud-to-ground strike still exists.

Figure: A four panel display from 2059 UTC on 11 May 2011. The upper-left shows the 1 minute PGLM flash extent density. The upper-right is the 60-minute PGLM maximum flash density while the storm relative velocity is in the lower left and radar reflectivity is in the lower right.

Friday 13 May debrief... finally

Apologies for some lost posts on the blog and the lack of posts since Thursday morning. The Blogspot server went down and we couldn't access it until sometime this weekend. We will be working on restoring some interesting posts about PGLM and SATCAST that got lost. In the meantime, here is a transcript of what we discussed with the EWP forecasters during the end of week debrief...

UWCI / OTTC
- Was not coming in correctly all week, so was not demonstrated within AWIPS.

- Forecasters were not comfortable offering further comments based solely on informal demonstrations with visiting PI since they were not able to use the products within AWIPS.

SATCAST
- I think that does have some utility, but I think the way that it is a yes/no solution doesn’t work. It clearly wasn’t in a position to be used in an operational state with just a yes/no. From visible satellite you can clearly see that there are Cu in the area, but if you have multiple colors for different features of CI it might be very useful.

- From the standpoint of just having more data to compliment the process, I think it would be very useful.

- Looking at “areas” where CI was going to occur versus individual cells was much more useful to me. From the training I was expecting to see individual cells and when I used it I saw way too many false alarms, but once I switch the way I looked at it, it became much more useful.

- When it stayed red, it gave much more confidence than flickering on and off red.

- Having 6 thresholds may have been detrimental because it caused the flickering. Perhaps having some intermediate CI nowcasts would be useful to add.

- I think its strength is going to be its spatial and temporal resolution.

- I think its greatest utility would be in weakly sheared environments… 100’s of Cu over the area and this would help pick out which ones to pay attention to.

- Day/night switch triggered many more false alarms.

- “I think you’re on the right track with what you are doing.”

NEARCAST
- After the training, the forecasters felt comfortable understanding the product and how to use it.

- Color curves seemed to be reversed from NAWIPS/web versus AWIPS, and were confusing, but they worked through it.

- “You can infer convective instability 100 different ways.” … Forecaster was still confused as to how this was showing much more information.

- It is definitely best in the 1-2 hr timeframe, but anywhere outside of that it gets hard to use.

- At one point data did not arrive between 16-21 UTC, made fairly hard to use.

- Issue of clouds… later on in the day there were so many blacked out areas so we couldn’t really see what was happening.

PGLM
- Used the second “jump” as a situational awareness tool and decided to warn on it and it ended up having severe hail (5/12 – Norman). This preceded any radar indicators by 1-2 volume scans.

- I thought it was a good head’s up tool… going into it I didn’t know what to expect, but it ended up getting the storms that had severe just before radar. It was a very good situational awareness tool.

- It did seem that there was a lot more IC than CG yesterday (5/12 - Norman). The CGs were not a good delineator of severe vs. non-severe.

- I would like to see a WES case before getting immersed in the data.

- I definitely saw IC first and then CG, so it definitely provided me with a good heads up tool.

- The 1-minute data was really useful… That was enough resolution for me… 20-30 seconds may be too much. I especially preferred the instantaneous data… I wasn’t really interested in what happened in the past so I didn’t use the max density much.

- An IC-CG ratio product would be very useful.

- “A time-series would be awesome. If I want to interrogate individual cells, I want to dig into it.” … This is in conflict with the overall forecaster feedback last year (and some this year), who did not like the idea… General consensus was that if was not within AWIPS it would not be used… May be something to bring into AWIPS-II

- I don’t know what these values mean yet… If I was to take this back today, I couldn’t even explain what this does. I think you need to explain what a certain value is showing us, otherwise I think you will have a hard time selling this to forecasters.

- Fire weather in the west is obviously a big thing that the PGLM data can help fill the holes by radar and NLDN. There won’t be as much surprise by lightning starts.

OVERALL
- You have to get a large enough group of forecasters comfortable with these products who will go out and spread the word and the spread will become viral.

- “I didn’t find the integration with the CI group helpful, I actually felt more confused with the forecast by going over there. When we come in they have already chosen their domain and I’m not sure how much we can provide to them. The whole time I wanted to go to AWIPS and look at what’s going on. The concept was good, but their mission was different than what we were doing. I could see it being useful for us if we had some time to look at the data and then go over and ask them what they are seeing.”

- Maybe if the morning shift came in at 10 or 11 it might be more useful… The forecasters didn’t seem to find the idea of coming in early useful to adding information for the evening shift.

- Forecasters did not find the large briefing with the EFP useful. It seemed like everyone in the EFP was still working within their own “stovepipe” and that they were just an audience and not participating much.

- Surveys were designed very well and not tedious… having one survey for all products was a great idea.

- Having the PIs around was very helpful because they could ask questions and continue to use the products.

- At the beginning they felt that there were too many projects, but as the week moved on and the PIs sat down and worked with them, it wasn’t too much to handle.

- A WES case at the beginning of the week with all products would help with the exposure to each of the products at least once.

Gravity Waves in the Simulated Water Vapor Imagery

On the morning of 12 May, while discussing the previous day's forecast at the CI desk, we noticed some interesting westward-propagating wave-like features in the observed GOES-13 data over southern Texas. The waves appeared to originate from an MCS in southeast Texas. We then viewed the simulated 6.95 micrometer band from the 11 May model run, and were interested to see that the 30-36 hour forecast also showed similar waves propagating westward. The images above show the simulated and observed WV band at 0600 UTC on 12 May, but the waves are difficult to locate without viewing an entire loop. A few hours later, a storm initiates (in both the observed and the simulated imagery) along the Rio Grande River and might be related to the westward-moving feature. A more thorough analysis of this case is needed.

Day 3 forecaster interactions...

This morning 3 of the visiting forecasters participated in the EFP's new CI desk. The discussion was based on where they expected new convection to occur within a selected domain encompassing most of OK, the northern half of TX, southern half of KS, and eastern halves of CO and NM. The simulated reflectivity from the NSSL-WRf was examined early during the forecast period to help determine the accuracy of the NSSL-WRF model. While it was decided that the NSSL-WRF didn't capture the early MCS over W. TX and OK, it did show some potential useful information regarding CI forming along the dryline later in the day once the MCS moves on and instability returns. We also examined the 10-12 micron band difference product to help determine specific areas where CI would occur (see below).

NSSL-WRF simulated GOES-R 10-12 micron band difference for 1900 UTC on 11 May, 2011. Arrows indicate areas where low-level moisture is favorable for convective development.
While the group decided collectively on a conditional slight/moderate/high contoured area, individual participants were asked to select a point where they expect the first CI to occur within 25 miles, as well as what time they expect it to occur and their confidence in their forecast. This input helps generate a human based PDF that can be compared to the model generated PDF following the forecast period. I picked a point near Lawton, OK at 2000 UTC with a 30 min +/- window and a 70% confidence... we'll see how I do.

NSSL Scientist Mike Coniglio leading the forecast discussion at the EFP CI desk.
Following the CI desk forecast, invited EWP forecasters began working on an initial AFD on their AWIPS stations. Forecasters are using a combination of operational model field, as well as some experimental data from the Nearcast and the OUN-WRF. Ralph Petersen spent some time with the forecasters to explain how the Nearcast output could help increase their confidence of thunderstorm development. Forecasters have seemed very interested in a strictly observation-based forecast out to 9 hours... some have even asked how to get this data back into their AWIPS at their local WFO.
Ralph Petersen of UW-CIMSS explains the Nearcast product to NWS forecasters during real-time forecast operations.

EWP discussion 11 May

Following the joint afternoon discussion period with the EFP, the EWP forecasters met for a discussion about the previous day's experience. The forecasters monitored the SATCAST product during the late afternoon period yesterday and offered these comments...

- Fair amount of CI hits, but it didn't seem like many of them would go on to full convection.

- Displaced a little westward from the visible satellite... probably due to the parallax correction.

- What is it's goal? Should I be watching it to stay red (nowcast) or once it goes red should I expect it to verify? It became unusable because it jumped from red (nowcast) to blue (cloud detection) all the time and it wasn't verifying. It seemed like the false alarm was so high that it deterred me from using it.

- I understand that the satellite scan time is an issue scanning every 15 minutes, but if it shows two back to back scans of nowcast, then that's when I would pay attention.

- How should I verify this? Should I look at a 35 dBZ echo at the surface or anywhere in the column, or the first occurrence of lightning? Probably shouldn't use lightning because we only have CG detections now and that may not show all lightning that may be occurring, so it wouldn't be a fair validation. The EFP CI desk is using a 35 dBZ threshold at the -10 C level and we can display that in AWIPS... this is probably the best way.

UW-Madison CIMSS convective initiation, overshooting-top, and nearcasting update

GOES-R proxy University of Wisconsin convective initiation (UWCI), overshooting-top/enhanced-V, WRF ARW simulated data and nearcasting fields have been flowing in a smooth manner into the EFP via N-AWIPS for forecast discussion integration. UWCI did indicated individual cells developing along north-south boundary in MN yesterday afternoon (10 May 2011).

24-hour UWCI indications, overshooting-tops, and NLDN lightning data from 12 UTC 20110510 - 12 UTC 20110511 below:



Example GOES Sounder nearcasting product within N-AWIPS with radar overlay on top. Red and yellow areas indicate regions of conditional instability.



The usual first week hiccups in GOES-R HWT experiment. Unfortunately the UWCI and Overshooting-top/thermal couplet products are strangely flipped from north to south during GRIB2 to AWIPS netcdf decode at HWT. Jordan Gerth is talking to local AWIPS expertise to resolve the issue however no solution so far. This is preventing evaluation of products within EWP. Nearcasting and WRF simulated data are available.

GOES Hail Probability Product - Day 2

Complementing the blog post below about convective initiation in Minnesota, the image above shows the 21-00 UTC hail probability forecast on 10 May 2011. The units are %, and the small 'a' symbols represent observed hail reports. Note the maximum located in western and central Minnesota; as noted in the other post, storms were just forming at 21 UTC, and proceeded to produce severe hail a little east of their initiation point. Additional hail reports occurred later after 00 UTC. Notice also the maximum in eastern Indiana and western Ohio, and the observed reports there.

First Total Lightning Ops Day

Wednesday is shaping up to be the first ops day to use total lightning data for the Spring Program this year. The main area of focus will be in central Oklahoma using observations from the Oklahoma Lightning Mapping Array.

This year, there will be two additional products added to the lightning product list. In 2010, NASA's Short-term Prediction Research and Transition (SPoRT) program developed the pseudo Geostationary Lightning Mapper (PGLM) product. This was a simple algorithm that could be applied to any ground-based total lightning network. While it is not the official Algorithm Working Group proxy product, it was designed to be used at the Spring Program until the proxy is available. The PGLM is simply a tool to train forecasters about total lightning and the Geostationary Lightning Mapper, and a way to work with forecasters to determine the best way to use the 8 km resolution data. Addtionally, this product gives end users the opportunity to discuss new ways to better visualize these data.

Based on feedback from the 2010 Spring Program and SPoRT's efforts to enhance the visualizations, two new PGLM products have been introduced for 2011. These are the flash initiation density and the maximum flash density (MFD) products. Figure 1 shows the original PGLM (1 minute data, bright color) overlaid with the MFD (60 minute history, faded color). The MFD shows the largest PGLM value for each grid box for either 60 or 120 minutes. It is effective to show a basic trend in lightning activity over time. In Figure 1 below, taken at 1446 UTC on 11 May 2011, the MFD shows a large amount of lightning activity from storms as they entered western Oklahoma. However, the PGLM data show that these storms have greatly diminished in lightning activity. Figure 2 shows the corresponding radar reflectivity from KFDR.

FIGURE 1: The 60 minute maximum flash density (faded color) and the 1 minute pseudo geostationary lightning mapper flash extent density (bright color) taken at 1446 UTC. Note how the PGLM shows that the lightning activity has greatly diminished, both in magnitude and extent.

FIGURE 2: The corresponding radar reflectivity product from 1446 UTC to compare with the lightning products in FIGURE 1.