Day 4 wrap up and weekly feedback

Forecasters finished up yesterday operating in the Topeka and Mount Holly CWAs. Once again it was a pretty marginal day severe weather wise, but there were a few storm throughout the area. Weekly feedback is posted below:

GLM

The 9km grid resolution would probably be ideal to reduce the dropouts

It should just fit the grid resolution of the sensor

If it could be gridded before hand and just sent out after, that would be easier and would simplify the menu

Might be helpful to have its own menu tab to make it easier to find

The parallax issue has to be fixed before going out to the public for sure

Interpolating loses some resolution on the event counts

Think the data will be very helpful for use in operations

Keep the menu simple, instead of having so many options

ABI

Need to cut back on the amount of RGBs, some don’t really have much use aside from just imagery

Need to know which ones are useful and for what

Forecasters need to know why this should be used and why it’s here anyway

Some are very good and useful, but some just seem unnecessary

Label the ABI channels with what it is, like “Red” visible, low-level water vapor, etc…

Stability indices were really good and useful with the limitations it has in clouds

Did very good with the TPW values

Don’t really need it filled in

Just overlaying with model data is probably sufficient

ProbSevere

Like the separating out of probabilities to see what the threat is.

The wind probs seem pretty sensitive to certain characteristics

Good as an SA tool, especially in a busy environment and in the junky convection days

Helps bring attention to the storms that are different

Monitoring trends are important for ProbTor

Not good with lines and object tracking

Overall, fairly good

Provide numerous training events and scenarios showing the variability in the model based on different weather situations

Nice to have the ProbTor breakout

Would like to be able to change the width of the contour

NUCAPS

If latency can get fixed, I can see it being pretty useful

The experimental correction was consistently better throughout the week

Change the colors to be easier seen and to not necessarily think that green means good

Might need some better QC to examine some of those green points which are not good soundings

MetOp soundings seemed to have coding bugs and weren’t as good as the NPP soundings

Most forecasters are going to discredit the sounding when the soundings don’t deal with the boundary layer very well

Moisture in boundary layer, even in the experimental soundings didn’t do to well on the coastal areas assuming the well mixed boundary layer

Getting it faster and having the experimental correction would make it much more useful in operations

If latency can get down to an hour or so, it would be very useful

If MetOp soundings can be improved it would be useful to have that many more passes showing the evolution of the atmosphere

Good for use in areas void of upper air data

Plan view needs some improvements, needs to be more like the GOES plan views to be very useful

Need to have better layers to calculate with and be able to show SBCAPE, MLCAPE, etc…

Plan views need to be more continuous with some better interpolation

-Michael Bowlan

GOES Sandwich - Temporal Comparison

With conditions quiet in the Mt Holly CWA, it was interesting to compare the temporal resolutions of the GOES-13/15 (15-min), GOES-16 CONUS (5-min), and GOES-16 Meso (1-min).  In this case, there were thunderstorms rapidly developing in southern Illinois and the Meso domain provided a great view of the evolution of the storms and the overshooting tops (note the waves moving away from the overshooting tops in the 1-min data). 

3 Panel comparison of GOES data: Top - GOES-16 Meso, Middle - CONUS, Bottom - GOES-13/15

-GOB

NUCAPS near Topeka

We had good clearing over Topeka during the 1842Z pass for NUCAPS today.  The original NUCAPS sounding showed plenty of dry air in the mid/upper levels.  Adding in the experimental adjusted NUCAPS for the lower levels, it did much better again today.  The adjusted lower levels made a great improvement to the sounding and brought surface CAPE values way up as well to nearly 3400J/kg.  Outside of the latency issue, the experimental NUCAPS made some nice adjustments and would add value assuming we could get these soundings quicker.

NUCAPS 1842Z

Experiemental NUCAPS 1842Z

-ISU2004

GLM practical use with storms and outflow boundaries

Here's a very practical use for the GLM and not just in terms of forecasting or determining severe weather potential in thunderstorms.  During summertime multicell mode with many outflow boundaries from collapsing storms... the GLM was able to pinpoint rapidly developing updrafts behind the outflow boundary.

In the radar loop above... additional storms developed behind the outflow boundary, but it took a few radar scans for them to show up.  This could be used with converging boundaries too along with GOES-16 data... like sea breeze induced convection.  While the storms were not severe at the time, there was a dramatic lightning jump with 2 individual cells.

The interpolated 5 min GLM events with 1 min updates on a 9km grid along with 5 min ENTLN overlay did show the increase in lightning in the cells, but honestly I prefer the non-interpolated data IF I'm zoomed into the CWA, since it does pick up the higher GLM event totals.  Warning were also issued from the dramatic lightning jumps.

-Dan S.

Day Cloud Phase RGB - New Mexico/Colorado/Kansas

The Day Cloud Phase RGB has become a decent tool for observing rapidly developing cumulus, it offers a little more insight outside of of traditional Red Vis, with the noticeable change from white to yellow for increasing vertical convection. Note near the end of the loop, the storms developing in far west central Kansas and eastern Colorado. You can see the agitated cumulus develop and then rapid upward growth.

Dale Doback

New Warning with GLM increase

We had a big jump in GLM with the storm in TOP's area and we opted to warn on the lightning increase.  In fact as we were getting the warning out, we had 60dbz to 37kft.  Lightning increase was rather quick too.

-ISU2004

GLM Uniform Grid Issue

The grided GLM data appears to be reliable near the nadir, but issues appear when you move farther to the edges of the GOES-16 domain. The below image is from New Jersey, this afternoon. The individual boxes or plots for events as you go clockwise around the image, 5km, 8km, 10km, 9km, resolutions, remain the same as they should. However, the grid that is derived for each resolution does not change with grid projection of GOES-16, they remain uniform. This results in individual events being counted twice in one grid box or data drop outs. you can see examples of both as you look at the different projections. The grid resolution should mimic that of the GOES-16 grid resolution, which is dependent on the distance/location from its nadir.

Dale Doback

ProbWind: Highly Sensitive To Seemingly Subtle Changes in Values

When watching a thunderstorm to the north of the KDIX radar, I noticed rapid changes in the ProbWind values.  Radar interrogation didn't indicate anything severe and generally weak outflow from the storm.  It appears that these values were tied to the subtle changes in the LLAzShear that only changed by around 0.01 at times.  Here are a few screenshots:

2142Z

2144Z - Group splits, which is reasonable based on the multicell nature on reflectivity.

ProbWind on the previous scan was around 57% after the split.

With no significant change in the values, the ProbWind increases to 71% and likely due to the slight increase in LLAzShear.

2146Z

2150Z - The ProbSevere data is grouped back together and increases to 85%.

2152Z - Storms separate again.

This transition highlights some concern in the ProbWind values with the rapid fluctuations seen.  Today's storms have been generally lower topped and we have seen brief and rapid increases and decreases in the ProbWind values.

-GOB

First Warning for TOP (GLM lightning jump)

We switched to TOP (Topeka, KS) upon seeing rapid intensification of storms.  We have just issued our first warning, mainly due to a notable jump in lightning data from the GLM.  Earth networks is also plotted with the GLM data

Storms are strengthening to the southwest and lightning has actually decreased while making this post.  But storm has intensified.

-ISU2004

Operations Update

WFO Pueblo is now WFO Topeka for the rest of the operations today.

-Michael

Another Warning Issued for Mt Holly

Have seen increases in ProbWind for a couple other storms and issued experimental warnings based on their increases.  One interesting item of note on the latest warning is the presence of the TBSS while ProbHail was only 1%.

2120Z - Note the ProbHail of 1% on the storm with a TBSS

This is probably due to the low tops on the storm (50dBZ is below 26kft) and generally small hail.


ProbHail did increase to 11% on the next scan and about 5 minutes after the warning was issued.

2124Z after the warning was issued.

-GOB

AWIPS Issue with GLM over GOES-16 Data

One item I've noticed with my 4-Panel procedure (time matched to radar) is that if GLM data is placed over GOES-16 data, it will be time matched to the GOES-16 data instead of the Radar.  Thus, you end up with GLM data that only appears every 5 minutes (lower left screen).

GLM Data over GOES - Note how the GLM is only every 5 minutes.

If you remove the GOES-16 data from the lower left panel, the GLM data will update at the same frequency as the radar that is time matched.

Loop with the GOES data removed and now the GLM data updates at the same frequency as the radar in the upper left corner, which was time matched.

-GOB

First Warning for Mt. Holly CWA

A storm in the northern part of the CWA continued to show increasing ProbWind values and reached 91% at 2052Z.

2052Z - ProbWind up to 91%

Since the storm was showing some indications of an increasing wind threat along the leading edge of the storm (up to 40kts around 3kft above the surface), decided to test out issuing a warning on the ProbWind data.

2058Z - Warning Issued

In addition, the GLM data steadily increased in the 30 minutes leading up to the warning.

-GOB

GLM, ProbSevere, and GOES-16 Sandwich Product as SA Tools

As the title says, the GLM, ProbSevere, and GOES-16 Sandwich Product have been useful as SA tools in identifying the storms to take a closer look at.

2047Z

The storms have been pulsing up and down lately, but the northern storm continues to show the greatest lightning and has had the highest ProbSevere for wind so far (78%).  The radar data shows some increase in winds and will be montioring for a warning.

-GOB

SNPP NUCAPS vs Experimental NUCAPS over Eastern Seaboard

NUCAPS soundings for New Jersey, Delaware, eastern Maryland, and the rest of New England became available around 19:30z for the 17z observation (technically approximately 1730z). The vast majority of our CWA (Philly/Mt. Holly) remains socked in with strato-cumulus. I chose to look up stream across southern Delaware and far eastern Maryland for sites that offered minor data quality issues. I chose a point near KSBY or Salisbury, MD and looked at several other points.

Here is the 17z sounding from NUCAPS.

The surface conditions on this sounding were off from surface conditions observed at KSBY at 17z. KSBY had an observation of 94/76, while the NUCAPS sounding started at 83/69.

I made these adjustments as well as some modifications to the boundary layer points to make it more realistic, this improved it's instability calculations.

Here is the 17z initialized RAP sounding for the same latitude and longitude as the NUCAPS sounding above. Comparing the two, the mixed layer and surface based CAPE calculations are now more similar.

 
The SNPP NUCAPS has trouble resolving the boundary layer, this is a known issue. Therefore, my first move was to modify it using the surface observation as a starting point. With that said, after working with the SNPP NUCAPS, my co-worker, GOB, loaded up the Experimental NUCAPS sounding which became available around 20z.

 The Experimental NUCAPS did much better resolving the surface conditions, with a surface observation of about 95/75. It's Instability values are bit more, especially the 3800 J/Kg of mixed layer CAPE. This is likely due to the sharp dew point increase off the surface through 900 mb. With modification, Instability values were closer to the RAP/HRRR/3km NAM soundings.

In short, the experimental NUCAPS soundings were more representative of the environment than the SNPP NUCAPS, primarily due to their better representation of the surface/boundary layer.

However, the timeliness of the soundings remains an issue, 1.5 to 2 hours after the observation. If this can be resolved, these could be better used by forecasters. With that said, it still took upwards of 15 to 30 minutes to interrogate the validity of the soundings, comparing surface observations and mesoscale guidance. I still worry this could be sticking point for forecasters, the level of work put in to quality control may not be worth the effort, especially in a faced changing convective environment.

If I had to chose between the two suites of guidance, I would lean toward the Experimental NUCAPS.

Dale Doback

GLM: Event Increase Preceeds GOES-16 Overshooting Top

While watching the storms enter our far western CWA, I noticed the GLM Events on the southern storm increased some (near the tip of the western part of the cwa in yellow) around 2010Z. At the same time, the GOES-16 data (upper right) didn't show any stronger areas or overshooting near that storm.

2010Z (UL - All Tilts Radar and ProbSevere, UR - GOES Sandwich, LL - GLM/NLDN, LR - ENTLN)

But when the next GOES-16 scan arrived, you can see the colder cloud tops with the overshooting top, matching the strengthening updraft indication from the GLM.

2017Z (UL - All Tilts Radar and ProbSevere, UR - GOES Sandwich, LL - GLM/NLDN, LR - ENTLN)

At 2022Z, the GLM data increased again, but the overshooting top is diminishing and dissipated on the next GOES-16 imagery.

2022Z (UL - All Tilts Radar and ProbSevere, UR - GOES Sandwich, LL - GLM/NLDN, LR - ENTLN)

In addition, I decided to try out the Stano colortable for a while and this seems to have increased the contrast on the moderate GLM values.

-GOB

GLM Grid Spacing for Events

When selecting the resolution for the GLM 5 min events with 1 min updates, I compared the different sized grids that are being tested in this HWT EWP.  There are definitely double count issues with 10km grid (169 events for nearly the same location).  The 8km grid had zero event areas.  When you zoom out to get the bigger picture for a wider area for storms, the 8km grid became a little fuzzy and you could tell the higher event areas.  The best grid from what I looked at seemed to be the 9km grid (5 min events with 1 min updates), even though the native resolution of the GLM is 8km.

-Dan S.

MetopA issue

Noticed a slight bug or issue with the MetopA this morning.  Most of the yellow dots (microwave only) returned back erroneous soundings that did not look realistic.  Here is an example of one of the yellow dots from MetopA:

Here is a yellow dot example from NUCAPS yesterday to show how NUCAPS is not being impacted by the same issue:

-ISU2004

Storms Nearing Mt Holly CWA from the West

The storms that have been showing a consistently "high" ProbTor value (above 20% earlier) have started to weaken as they near the Mt. Holly CWA.  This weakening can be seen on several of the experimental products.

1952Z (UL - All Tilts Radar and ProbSevere, UR - GOES Sandwich, LL - GLM/NLDN, LR - ENTLN)

In the image above, the ProbSevere values were around 66% due to Wind, the sandwich product had colder cloud tops, and the GLM data was more intense (171 Flashes).

2008Z (UL - All Tilts Radar and ProbSevere, UR - GOES Sandwich, LL - GLM/NLDN, LR - ENTLN)

Moving to 2008Z, the ProbSevere values fell between 30-50% due to wind.  The southern storm in the Sandwich product showed slightly warmer and fuzzy/glaciated cloud tops and the GLM flashes fell to 113. Also notice how the event grid on the southern storm (lower left image) has focused more on the northern storm and matches the trends in the radar and satellite data.

At this time, no plans to issue any warnings on these storms as they enter the CWA. 


-GOB

GLM - Reason for Empty/Zero Pixels in Gridded Event Data

Since I've mentioned the empty/zero pixels in the GLM Gridded Event data over the last couple of days, I thought I would post the reason for it.

In the image below, the blue dots are the GLM Events and the gridded data is the GLM Events that are gridded.  Due to the way that the data is AWIPS grid is verses the GLM data (slightly tilted) it will lead to occasional grids that don't have a GLM event in them.  The 9km data is improved from the 8km data that we have been looking at, but still contains these zero pixels.

Would it be possible to deliver this data at its native resolution and in a gridded form, so we don't run into this issue? Similar to how we get the gridded MRMS data instead of having it calculated on the fly.

-GOB

Mount Holly, NJ CWA - Starting to See Development

Over the last 13 minutes, we are starting to see development in the Mt. Holly CWA.  At 1918Z, there were a few showers but at that time, no lightning present.

1918Z - Little to no lightning in the Mt. Holly CWA

Just a short 13 minutes later (1931Z), two storms have developed in the unstable air mass and are producing a decent number of GLM Flashes.

1931Z - Two storms have intensified in the CWA

Unfortunately, the existing cloud cover has limited the use of the Derived Stability products today.

1917Z - GOES-16 Derived Parameters (UL - CAPE, UR - LI, LL - PWAT, LR - Sandwich)

Looking at the SPC Meso page, there is decent MLCAPE and Effective Bulk Shear over the CWA.

19Z MLCAPE & Effective Bulk Shear

-GOB

ENTLN Cloud Flash Increase Much Greater than GLM: Pennsylvania

Tracking a stout thunderstorms across central Pennsylvania this afternoon, a noticeable trend in the ENTLN detecting a much greater number of cloud flashes, vs the GLM Flashes, granted, these are not a 1 to 1 comparison, this is the most significant and persistent difference I have observed this week.

Near the start of the loop, 18:29z: 23 GLM Flashes vs 28 ENTLN Cloud Flashes

At 18:32z, the ENLTN begins to distance itself from the GLM.

 At 18:46, the ENTLN maintains it's large spread on the GLM, 182 to 104.

At 18:53, the ENLTN reaches it's greatest spread from the GLM, 353 to 134.

Of note, during this time period, the NLDN and ENTLN had nearly identical strike counts, for both positive and negative strikes.

Dale Doback

GLM Flash Density

Comparing the flash density for the 5, 8, 9, 10 km grids.  It looked the the 9 km grid had the least amount of data drop out areas to make the lightning fields look more continuous.  Here is a loop of lightning data from southern Ohio/Northern Kentucky showing the differences in the resolution.

5 km has way too many drop off areas and just looks strange, while the 9 km looked the best overall.

Pueblo convergence zone

Pretty good surface convergence zone setting up over Pueblo this afternoon with convection firing up along the boundary.  GLM has been active along some of the line.  Prob severe has been handing out in the single digits and storms are having a tough time sustaining a good depth.  However, storms will continue to move into an area with high CAPE values this afternoon, so some storm strengthening with time seems very likely. 

Convergence zone (noted with white lline) with GLM overlaid.  Shading is events, and plotted is flashes.  This is one of the experimental color curves begin tested this week.

-ISU2004

ProbTor and GLM Comparison

Here is a loop of the ProbTor overlayed on the MRMS Rotation Tracks (Upper Left), along with GLM (Lower Left), ENTLN (Lower Right) and Sandwich GOES-16 (Upper Right).

~80 minute loop over central Pennsylvania

The thing to take notice of in the lower left loop is how the GLM lightning flashes/events quickly ramped up just ahead of the highest ProbTor values (74%).  That GLM increase highlighted the intensifying updraft in the storm.  In addition, when taking a close look at the MRMS Rotation Tracks, you can see increases around the same time or just after the GLM events/flashes increase (possible lightning jumps).

One item that I did notice with the 9km GLM data with this storm is a few pixels of no data (notice it during the first half of the loop and again near the end as the storm moves directly over one of the spots where the lightning data doesn't match with the AWIPS grid.

Example of zero data due to the grid in AWIPS at 1809Z.

 -GOB