Friday, June 6, 2014

Week 4 Summary and Feedback

Below is feedback from our final, week 4 weekly debrief session, along with a photo of the week 4 participants.

This marks the conclusion of the GOES-R portion of the 2014 HWT Spring Experiment!


- Bill Line, SPC/HWT Satellite Liaison



Simulated Satellite Imagery:
- It’s great to use the imagery to evaluate model performance.
- Had issues with burning off low stratus too early, leading to errors with heating during the day and errors of convection timing
- Do you see simulated model imagery being useful in forecast environment; would you like to see this data with other high res models?
    o All say yes, it is useful and would like to see it with other models (hrrr, etc)
    o Higher res models might pick up more discrete features things
 - Easy ways to pick up on errors in the model

Nearcast
- I used it mainly for ted and pw, it almost shows this better then anything I have in my office
- Nice to have in awips to overlay other things
- Helpful to show where boundaries will move, areas that are moistening
- Best as a precursor, ~3 hours prior to ci mostly
- I like the fact that you could really key in on boundaries, and I saw convection go there.
- Even subtle gradients up in high plains, small increases in ted, see showers pop up
- In Raleigh, having a tool that may easily show boundaries, esp in spring and summer, it could help
- I see more utility in theta-e and ted than in the PW field
- Calibration issue, we are used to looking at CAPE and LI. What does the ted values mean? Is there a way to do this?
- Its real-world information, which is good
- Overlayed forcing parameters on it.

Ci
- Didn’t increase my confidence as much as I had hoped
- One cell I tracked for a while in Huntsville area, had pretty high value, that did go up and eventually became strong
- Get rid of lower values

Probsevere
- This is especially useful for more slowly developing convection. Storms that look the same may have different probs, pointing out which storms are actually more threatening,
- Not as useful when convection is rapidly developing. The 1-2 scan lag hurts
- Good as a confirmation tool even for mature convection
- Good for slowly developing storms, detects hail well, poor with wind, not sure with tors. Prob severe hail may be better name. Rapidly growing convection, it wasn’t as useful with lag (1 volume scan), which hurt.
- Broadcaster – I really enjoyed this product. I am doing a lot during a day, especially severe day, it pinpointed where I should look. Same with CI
- Right now, as a regional SA tool, storms coming in from neighboring CWA, would be good to watch. Also for updating warnings.
- Column velocity to determine wind threat,
- Seems like product that has room to do some interesting things.
- with prob severe/hail/wind/tor, have a 4 panel, with total severe and 3 threats. Different parameter underlaid each

Overshooting Tops
- Broadcaster – really useful for me, in busy environment. I can speak for every broadcaster, we’d all love to have this product.
- More utility at night than during the day
- Helpful to CWSU cause I'm looking at a much larger area

Tracking tool
- This trend information is beneficial to have.
- Probably best for warning coordinator, or for research purposes. It is high maintenance, takes a lot of work to use

Lightning
- Broadcaster – lightning is especially useful for me, the chopper folks are constantly asking me where there is lightning to decide whether it is safe for them or not
- Most useful for rapidly developing convection
- Any time a jump occurred on a storm, the dBz’s would spike soon after
- I think it is everybody’s favorite thing, the lightning
- Most helpful for a warning operator, everything else was more SA.
     o Mostly because we had information every minute.
     o Getting additional info while we are waiting for next volume scan
- It impacted my warning decision making process.
- Especially useful in rapidly developing convection
- Broadcaster – lightning is very important even if it isn’t severe
- I wish I had more time to look at it.
- Training for lightning is not that good, to understand more why/how this is working, would be helpful. Why in certain environments and situations would it not work so well. But why better in others
   o Explain what is going in and why it is working

Broadcaster – from this perspective, most of the products demonstrated make it easier for him to do his job, because he can easily spot where a new storm is coming up in busy situations, where lightning is increasing. It helps make things easier for everyone at station.

CWSU – CI and OTD would be most useful for me, covering a broad forecast area. OTD quickly shows me where the strongest updrafts are, and ci highlights where convection is initiating or most likely to initiate in the near future. Also, lightning data is valuable for me.

General – be cautious of data overload. Products have to quickly prove their worth and utility in operations. Some products might be more useful for a warning coordinator, while lightning and maybe prob severe could be helpful for radar operator

Overall
- 1pm briefings from EFP, is there a reason for us to be in on that. A lot of it was not important
- Even from the briefings, I had no idea what was going on. It seemed more like a debrief for themselves. It had no application to us, it was geared to the efp
- I could have better spent my time making myself aware of the meeting.
- Schedule our debriefing during that, and maybe have it earlier (noon)?
- Since a lot is pre convective environment, we should come in at 9 or 10 to start.
- To be able to look at these products with other convective modes would be useful too.
- Broadcaster – recommend others to get with local wfo and get a couple hours in front of awips. Simuawips helped quite a bit.
     o This is most fascinating week of science I've had in a while.
     o Very much against working with just broadcasters in the hwt, found it very beneficial to be here with nws mets.
- All forecasters gained, learned from each others unique perspectives. Against segregated experiments.

Thursday, June 5, 2014

Daily Summary: Week 4, Day 4 (5 June 2014)



Today, we operated in 3 county warning areas:

Team 1 (Fowle & Anderson)
Boulder, CO

Team 2 (Pelczynski & Satterfield)
Pueblo, CO
Huntsville, AL

Given the potential for cirrus-free skies and a lightning mapping array, we elected to operate in the High Plains where another upslope day was expected.  In particular, we chose the Pueblo and Boulder CWAs.  However, it became clear that the Pueblo CWA would wait to convect until later.  Given our time constraints (ended at 7 pm to start surveys), we decided to operate in the Huntsville CWA, where a very fast-moving MCS would move into a lightning mapping array.  Most of our experimental products were used, including the OUN WRF.  Several severe storms formed in both active CWAs, with multiple lightning jumps observed in both.

-G. Garfield
Week 4 Coordinator

Blog 13 June 5 HSV Area

Wind Damage in Sylvania (in Dekalb Cty) AL. OT Top detection correlated very well with this. 2245 OT Top detected damage report at 2311Z 5 June.1 sigma light jump at 2225Z as well!



2320Z D Satterfield

ProbSevere and S. Colorado Severe

The ProbSevere model captured the development of severe weather in southern Colorado nicely during the evening of June 5. With one cell in particular, the first probability over 50% provided 20 min lead time to the first reported severe weather (1″ hail).

The algorithm started tracking the cell at 2136 UTC giving it an initial probability of 9% (to the first severe, Fig 1). At 2210 UTC, the algorithm had its first probability over 50% with a prob of 65%, indicating the cell was more likely than not to produce severe weather (Fig 2). The first severe weather of 1″ hail was reported with this storm 20 minutes later at 2230 UTC (Fig 3). The probability had maxed out at 85% at 2216 UTC.





- Bill Line, SPC/HWT Satellite Liaison

Blog 12 June 5 HSV Area Cherokee County AL

In the last post I mentioned the prob severe as not as helpful today. We now have a cell in Cherokee county ahead of the MCS that has an OT Top detect and a 4 sigma jump in lightning. Prob Severe now at 80% with strong flash density. HUN has warned and with all of these present expect severe reports. In this case it was warned because of history of these storms but all of this increases confidence in the warning and that this storm is still likely getting stronger.. was not as impressive 30 mins ago.

DSatterfield 2255Z


ProbSevere Improvements

Couple quick thoughts on ProbSevere improvement – based on threats:

Wind:

T/TD depressions

Subcloud/Low Level RH

0-3KM Shear

DCAPE

Vertically integrated “core dump” e.g. track descending Z core

Hail:

Areal extent of negative ZDR cores on 0.5 degrees (difficult)

H7-H5 Lapse Rates

CAPE in Hail Growth Zone

500MB Temps

Tornadoes:

0-1km Bulk Shear

0-1 SRH

MLLCL



Fowle

CI tool not showing anything of real value here



The CI tool is really only keying in on the developing cumulus, which is seen easily in the visible satellite imagery. Since CI doesn’t work as well at night when the visible satellite is also not useable, it doesn’t appear to be adding value here. I think that perhaps having a lower limit on the percentages that show up would be best, especially if we are worried about potential for severe storms to develop. All in all I’ve found it somewhat useless today. As far as the location of the highlighted areas with respect to the actual clouds and increased reflectivities, they are slightly displaced to the west of center. A lot of the lower values don’t even result in any enhanced reflectivities.

-KP

-D. Satterfield

Blog 11 June 5 HSV Area Prob Severe Evaluation



Interestingly, today the prob severe with the large MCS that produced lots of storm reports including wind gusts to over 50 knots, and enough lightning to start fires, was not that helpful. I have circled a bow echo moving into GA from AL and also a line near the HTX radar (This storm was quite severe with high winds extreme lightning). OT top detection and flash extent density and light jump all worked well and were valuable members of the severe warn decision tree.

Not the case for prob severe, which could lead the forecaster to hold off on a warning if they overvalued it. I have used prob severe all week and had growing confidence in it for storm trends etc., but this is a lesson learned to consider its limitations.



D Satterfield 2251Z 5 June 2014

CI GOES-W vs E again




This is another example of GOES E vs W seeing CI. In this case, GOES-W showed a fairly large area of 79% while goes east showed little if anything. These clouds did end up developing into echoes well over 35 dBz within 30 minutes after these images were taken.

jca


Blog 10 June 5 HSV Area

Looking at the severe storms over North AL with 1km vis and OT detect overlaid.

The cell which has produced a lot of severe storm reports on the TN AL border is detected Storm near Chattooga Cty. GA (which moved out of Jackson AL) also has OT detected.

I have circled some obvious misses that are visible on the 1km vis. Storm near Boaz in Marshall AL has produced minor damage and high wind gusts and is also missed.

D Satterfield 2240Z 5 June

*** I looked closely at this case in the IR, and am surprised the algorithm didn't pick up on those misses. They were well within the thresholds (OT temp, anvil temp, difference), at least from what sampling of the storms I did. - BL


Prob Severe Tool Fail

Below are numerous images of the Prob Severe tool doing a poor job with severe winds in the Huntsville area and it had inaccurate location placement of the Prob Severe area as well. Perhaps the Prob Severe tool is best utilized for severe hail and possibly tornadoes, but it definitely did a poor job with winds.









-KP

Blog 9 June 5 HSV Area

Image lost do to network issues-

Have spent time tracking a cell ahead of the MCS over NE Al. Mainly in Jackson Cty. Little storm damage reports from it in-spite of lightning jump, and only a so-so reflectivity presentation. At 2130Z we do now have an OT top detection on this cell and it looks rather  impressive on vis imagery.

Did a 64 frame tracking tool analysis on the cell moving through Limestone AL and into Madison. This cluster produced 59 mph winds at KHUN and reports of lightning induced fires in Decatur area. Many reports of damage, trees down, and some building damage from strong winds. While the radar presentation was not impressive the track tool shows overshooting top detection at 2140Z with 5 sigma lightning jump at 2202Z. Flash density peaks at 2145 then diminished but then steadily rises to 60 at 2204.

Prob severe not much above 25- See post by KP for more on this. Good example here of flash extent and OT and lightning jump being a clue that this storm was severe. Radar look alone would be of little initial value, esp with no storm reports (late night hours for instance)

DSatterfield

lightning warning June 5



The lightning jump algorithm for the storm highlighted in purple went from -4 to 6…which is a significant increase over a 1 minute period. This prompted a warning.

jca

Example of cirrus obscuring CI



This loop provides an example of how the thin high cirrus clouds from the MCS to the northwest are obscuring the CI output for the cumulus clouds over Alabama. The edge of the cirrus shield is shown in the zig zagged line within the bubble.

-KP

*** The CI developers are able to track and give probs below thin and moderately opaque cirrus (likely with this case), unfortunately this was not observed/known till the final day so the simple change to the algorithm was not implemented in time for any evaluation. - BL

Pseudo Lightning Data

Today was the first day I had a chance to examine the pseudo lightning data.  We were watching storms just to the south of Denver.  The first lightning jump we saw occurred at 2017Z and jumped 4SD.  Still learning exactly how to use this – but was expecting an increase in storm intensity afterwards (based on the training) and this did occur.  ProbSevere gradually trended up as well – and using a combination of ProbSevere and the lightning jump data – issued a SVR at 2050Z.  Did not really examine the base radar data until after this – and based on dual-pol radar data – there was definitely hail falling by 2055 or 2100.



Thus – I considered the lightning data valuable.  It is another tool assess future storm intensity – i.e. which storm do I need to pay attention to.



Fowle

Blog 7 Day 4 on 5 June

Still watching storm over Jackson AL 4 sigma lightning jump. Reflectivity not all that impressive. but flash extent density and light jump are really popping.

prob Severe 84% but env shear weak.

MESH is at 1 inch. Hoping to get some spotter reports.

D Satterfield







Blog 6 Day 4 on 5 June


Suggestion on light jump usage.
Each to hi sown but I have found that putting light jump below Radar data works well. Spatial extent of Light jump covers larger area in smaller cells. Will not always work but on cells in which it does it seems handy. Example below. Basic suggestion here is to play around with hour you layer and display these tools.
The cell in Jackson cty AL has a 3 sigma jump and strong flash density.
DSatterfield



Blog 5 Day 4 on 5 June

Warning issued by HUN on storm in Jackson County AL. We also issued warning on it just before based on Lightning Jump/Prob Severe and flash extent density. These new products in my opinion really were valuable in knowing that this storm was likely to become severe. One of the best examples I have seen all week.

I suspect we were watching it sooner than even HUN was because of our experimental tools.




Blog 4 Day 4 on 5 June

Monitoring radar and lightning jump -Flash Density over North AL and south TN.

Also using Prob severe. We already issued a warning for NC and NW AL. Having worked as a broadcast Met in this area for 17+ years it is very familiar to me. Unique experience to evaluate these products in an area where I’ve done many hours of severe now-casting.

Lightning jumps are going off on cell in Lincoln (already warned) and prob severe on line appr. NW AL is now at 82%. Also prob severe high on cell now in Jackson Cty. Prob severe rising rapidly

Am finding combo of flash and light jump very helpful in watching these cells which are developing rapidly. Almost seems as if it would be valuable to have one person doing traditional radar analysis and another using these tools over the same storms.

D Satterfield


Blog 3 Day 4 on 5 June Warn for North AL

Have issued a warning for North AL ahead of he MCS in North AL based on past history, flash extent density and MESH indicating large hail NQA Nexrad shows strong inflow winds in mid levels which should reach surface ahead of the line.

ThetaE Difference Loop – Showing Increasing Instability/Convective Potential???

Attached loop shows the Theta E difference values decreasing over the front range over the last 2 hours.  Looks like a minor surge of moisture creeping northward – which if this is correct should increase convective vigor by 21z.  This appears to be happening already – with storm intensity increasing on radar between 2000 and 2015Z.

Fowle


pGLM colormap (comparison)

As the MCS moves towards the Northern Alabama LMA, I thought it was a good time to compare colormaps and get feedback from each of the forecasters.

I added the following image to the large screen of the situational awareness display within the HWT and asked each forecaster which colormap they preferred – the one on the bottom left or the one on the bottom right.

top left: 200 km2 LJA over MRMS lowest refl. Top right: 600 km2 LJA over refl at -10 C. Bottom left: PGLM flash extent density (colormap 1). Bottom Right: PGLM flash extent density (colormap 2)


All forecasters this week preferred the colormap on the bottom left except for one specific case:  when they want to overlay the data on radar.  For that particular occurance, the bottom left is a bit to similar to the radar. However, all agreed that on its own or overlaid on satellite the colormap on the bottom left is much preferred. (note: the colormap on the bottom right is the default colormap for AWIPS2).

-K. Calhoun

Blog 2 Day 4 on 5 June

Have switched to the HUN CWA as of 1930Z.

Discussion of NSSL WRF sim sat with IR and water vapor.

Large MCS to the NW of the CWA moving into it. The NSSL WRF shows the MCS but is far less robust on the eastern end of it than is apparent in the IR imagery. Also strong cirrus shield is present east of the MCS which is not present in the NSSL WRF. This will have obvious impacts on model performance because of insolation issues etc.

Over the Carolinas we have some convection developing ahead of a surface cold front.. the NSSL WRF is to robust on the convection esp in KGSP area. Although it is doing quite well but a bit fast (by 100km) over the outer banks of NC.

Talking with KP about carolinas and the spatial errors may be due to model limitations on depicting the Appalachians.



DSatterfield & KP

CI GOES E vs GOES W




Taking a look at the highlighted areas, the top image is from GOES-E showing CI values of 92% in an area that developed into decent convective towers. The bottom image is from GOES-W with the same area showing CI values of 46%. Why such a difference?

jca

NearCast ThetaE Diff

Examined the ThetaE difference at 19z.  From a qualitative perspective – the ThetaE difference is greater over PUB than BOU.  Thus – you would expect it is more unstable over the PUB CWA – and thus more vigorous updrafts.  We shall see how this verifies this afternoon.


Simulated Satellite – Thursday

Started off the day by examining the Simulated Satellite products.  Overall – they provided a good starting point over CO.  It showed the stratus burning off in eastern CO in about the right time.  Plus – CI initiates around 18-19z which is evident in the actual satellite obs.  Thus – the NSSL WRF should provide an accurate depiction of overall convective evolution.

18z Top – 19z Bottom



EWP Operations Update – Thursday 6/5 – 2:45 pm

Today, we started operations in the Boulder and Pueblo CWAs, anticipating the development of severe storms in upslope flow in the lee of the Rockies.  Our hope is to have the Boulder group (Anderson and Fowle) monitor the Colorado lightning mapping array during the event.

Given the paucity of cumulus clouds in the Pueblo CWA, we believe that convective initiation will hold off until later.  Also, lightning activity is increasing in the Huntsville lightning mapping array as an MCS approaches from the west-northwest.  Thus, we have decided to re-localize the Pueblo group (Pelczynksi and Satterfield) to Hunstville.

-G. Garfield

Week 4 Coordinator

June 4/5 MCS and Overshooting Tops

An MCS originating in eastern Colorado Wednesday evening (06/04) traversed the southern Plains into the southeast by Thursday (06/05) afternoon. This system had many overshooting tops associated with it, highlighting the locations of the strongest updrafts (and associated hazardous weather) within the massive cloud shield. The overshooting top detection algorithm gives forecasters a tool to easily monitor locations of overshooting tops and trends in detection's over time. It is an especially beneficial tool to have at night (as with this case) in the absence of visible imagery. Below is a loop of IR satellite imagery with overshooting tops overlaid from 0345 through 1845 UTC on the 5th.




SPC storm reports for the 4th and 5th show that the occurrences of severe weather matched up with the locations of the overshooting tops.






- Bill Line, SPC/HWT Satellite Liaison



Wednesday, June 4, 2014

Daily Summary: Week 4, Day 3 (4 June 2014)



Today, we operated in 4 county warning areas:

Team 1 (Fowle & Anderson)

Louisville, KY
Springfield, MO
Cheyenne, WY
Team 2 (Pelczynski & Satterfield)

Boulder, CO
Early in the day, it appeared that the greatest severe threat would exist in the MS/OH valley region.  Additionally, there were enough breaks in the high-level clouds to allow for use of the GOES-R satellite products.  However, as the day progressed, it became apparent that the region would not destabilize as much, so we moved Team 1 to Springfield, MO (based on a favorable mesoscale discussion).  As it turns out, the capping inversion would hold over that area.  Thus, we moved Team 1 to Cheyenne, WY – hoping that we could catch some lightning data from the LMA in Denver.  A few marginally-severe storms did develop.

Team 2 remained in Boulder all day.  There, they used the GOES-R products extensively, including the NearCast, UAH CI, Probability of Severe, and lightning data.  Several severe storms developed, though only one severe report was officially logged (for a landspout tornado).

-G. Garfield
Week 4 Coordinator

*** Day 3 Feedback (Bill Line): ***

Simulated Satellite Imagery
- Spot on yesterday. As far as overall timing and development, it was really good. The fact that it picked up on little features early gave me confidence in the rest of the forecast
- Not as good in Missouri area.
- Burning off stratus too quickly seems to be a recurring issue with the model

Nearcast:
- I liked it yesterday, particularly because as we sat down there was convection getting going, related to theta-e diff gradients, you could predict where they would move based on the nearcast forecast.
- We looked at it for a good 2 hours
- Theta-e values were pegging where it finally developed in far northwestern NE

Goes-r CI
- Looked at in with the other products in the 4 panel
- In Kentucky, it was bubbly, weak CIN,  a lot of low values that jumped around, noisy field, didn't see values over 60%
- It triggered on a few but we were already on them.
- Might be better for larger areas of monitoring (larger than CWA)

Prob severe
- Sometimes it is a nice confirmation of what we already concluded
- With rapidly developing convection, storms went severe within a few scans, prob severe lag kind of hurt it
- With gradually developing storms, it may be more of a confidence builder to issue a warning
- In broadcast world, this is valuable in pinpointing a storm to look at
- One went 16 to 60 in one scan, making me look at the storm
- In Louisville, a lot were hanging in low range, it told us what environment would be like that day, very low end, bubbly stuff
- The values it showed us were helpful (predictors) to see
- Satellite ingest made a huge difference (with satellite data compared to without)
- Tells us what is the trend of the day, perculating stuff, or fast growing
- With trends it was somewhat helpful with dissipating storms, you could see that chances of it being severe were becoming less. Caused me to look at it more
- There was a lag between when I issued a warning based on prob severe and when NWS finally issued

overshooting tops
- I can see, with mcs’s, having this info is useful, you can see where stronger storms are in a complex, you have a long line, key in on these areas.

Pglm and LJ:
- I could see reflectivity increase 15 to 20 min after lightning jumped up. Tracking tool helped me see this. I think both of them together is valuable
- The lightning jumps led to me quickly issue a warning with rapidly developing storm in boulder.
- I used just lightning data to issue warnings,
- Using FED in conjunction with LJ was a red flag that storm was rapidly increasing in strength
- I thought it was really good yesterday
- LJ- I find it useful and like being able to see when it jumps,
- Option for max LJ over last 5 mins would be great. Still update every minute though. (Running max). So that I don’t miss an LJ in the rapidly updating product.

Tracking Tool
- I like the idea and concept of it, compare trends in fields and different storms. But there are a lot of problems
- Looks like I have to load all products first, then load  tool
- I crashed my system
- Very handy with lightning and reflectivity
- May be more useful to look back at archive data. Research tool
- In real time, it takes too long to work, you don’t have that much time. Maybe just the warning coordinator
- Its not telling you a lot different than what you can see



One last satellite comparison and forecast

The first two images showing the simulated satellite imagery (left) and the real-time satellite imagery (right) reflect the overly rambunctious coverage in the simulated imagery earlier in the day (1st image), however the latest images (2nd image) shows the simulated imagery underestimating the intensity with the storm in CO. Looking at the last two images, forecast imagery for 03Z and 06Z, they imply that perhaps the simulation is just lagging behind with generating the intense storms in eastern CO. Based on this simulation, and the fact that it has performed well all day, I would expect additional storms to develop in eastern CO/western KS over the next several hours…moving east thereafter.











-KP

Issues with the Tracking Meteogram Tool



The tracking meteogram tool is nice, but there is an issue with the graphs that are plotted. They pop up whenever they feel like it, and while I can understand why some variables are plotted, others seem random.

-KP

ProbSevere Loop CYS

Finally had a stronger storm develop in the CYS CWA.  Used ProbSevere to get a general feel on the severity.  As Z in base data increased – ProbSevere quickly increased from 6% to 60% to 91%.



Fowle

Blog 9 4 June

Just a thought which may be crazy but I overlayed Over shooting top Detect on radar ,5 Z

May have value in diagnosing storm structure? Shear etc.




D Satterfield

CYS warning



A warning was issued based mainly on the prob severe which went from 40 to 93% in 3 volume scans. The warning was issued when the prob severe was 40% after increasing from less than 20%. As a side note, WFO CYS did not warn on this storm.

jca

Blog 8 Track Tool 2345Z 4 June

Used track tool for first time. Much more intuitive than expected from pre-HWT training. First thing I notices was the increase in flash extent density at 2315Z followed by an increase in reflectivity by 2330Z. Quite interesting. Also noticed a lightning jump at 2305 which marked the beginning of the increase in Flash Extent Density.




D Satterfield WBOC