Hail or No Hail - Assessing ProbSevere v2 and v3 - Monday Case - 2062 Z

 Today was the first day of Week 2 for the Satellite HWT Experiment.  One of the main applications we are looking at this week is the new ProbSevere Model version 3.  We are able to compare this version to the one currently in use (version 2).  This is the version most operational forecasters are familiar with, and that is available in AWIPS2 and GR2 Analyst.  

The target today was convection occurring in the Shreveport, LA CWA this afternoon.  A QLCS was moving across the region from west to east, as shown below.  The reflectivity was not especially impressive, but velocity scans occasionally showed some stronger winds.

Above: A line of storms moving through Louisiana on June 7, 2021.

The image shows MRMS Vertically Integrated Ice (VII) on the left, a parameter many forecasters are familiar with and may use as perhaps a “sanity check” to see if a cell may be trending towards strong or severe.  On the right is a screenshot of the Shreveport radar (KSHV) at the 0.3 degree reflectivity slice.  Overlaid is the ProbSevere Model with ProbSevere2 and ProbSevere3 parameters both listed for comparison.  

As you can see there are a few areas of concern.  A quick glance at VII indicates there may be some strong updrafts capable of hail over the area of convection over southeastern Caddo County in northwestern Louisiana.  Looking at the output for ProbSevere2 in this same area, there is a 43% chance for hail in this area, and a 14% chance for hail according to ProbSevere3.  The new and not yet operationally available ProbSevere3 uses a newer, more highly-skilled algorithm that should theoretically better match with ground truth.  

During this time there were no hail reports in SPC’s database, severe or otherwise.  There were some areas of surface based CIN noted, and this may have suppressed stronger updrafts and thus large hail. So in this case, it appears that the newer version of ProbSevere was correct about hail not being severe.  

-Dana Scully

Northern VA/Maryland Convective Episode on 6/3/2021

 

Image 1: 1943 UTC GOES East Mesosector Day Cloud Phase Distinction

The modified NUCAPS sounding from 1730 UTC revealed in excess of 2,000 J/kg of CAPE (image 2) and by 1900 UTC thunderstorms had developed in an area less shaded by high level cloud cover. Comparing this to RAP mesoanalysis data, it initially seemed too high as RAP mesoanalysis suggested closer to 1000-1500 J/kg (not shown)  and based on initial lightning activity. SBCAPE from NUCAPS was extremely high and close to 4,000 J/kg which seems very high (image 3). We did note that for some reason the NUCAPS forecast image showed CAPE being “missing” over central MD/northern VA while the CIN fill was more complete. When sampled over the “missing” data in AWIPS in the Cloud,  the readout showed actual values.

Image 2: NUCAPS Sounding near Washington D.C.

Image 3: NUCAPS CAPE and CIN forecast at 2100 UTC

Image 3: 2000 UTC 4 Panel of Day Land Cloud Phase

Initially, storms appeared to struggle with flash counts on the order of 10 to 20 flashes. Over the course of 30 to 45 minutes, lightning flash counts increased by an order of magnitude (closer to 100-150 flashes)

Closer to 2020-2030 UTC the rigor of convective elements increased and we started seeing transient echo overhang/along with some weak echo regions in tandem with an increase in both lightning activity as well as ProbSevere trends (especially ProbSevere3) as seen in Image 4. This increased our confidence in issuing our own warnings with LWX having a couple of warnings already issued.

Image 4: ProSevere Time Series just after 2000 UTC.

Image 5: LWX 0.9 Reflectivity Image near 2000 UTC

Image 6 and 7:(LEFT) ProbSevere Time Series around 20:30 UTC.  (RIGHT) 4 Panel (from top left to bottom right) of Day Land Cloud Phase Distinction, Radar+GLM FED, Day Land Cloud Convection + TOE, ProbSevere3 + GLM MFA

Closer to 21:50 UTC things became a bit more interesting from both a radar and lightning standpoint. Imagery from the KDOX radar (Central Delaware) suggested increasing mesocyclogenesis across Baltimore County (image 8) due north of Baltimore.

Image 8: 20:48Z 0.5 degree cut from KDOX indicating low level mesoscyclone.

Image 9: 4 Panel of ProbSevere, GLM FED, GLM MFA, GLM TOE at 20:45 UTC

Image10: 4 Panel of ProbSevere, GLM FED, GLM MFA, GLM TOE at 20:51 UTC

From a comparison of these images, there was certainly an indication that the updraft was increasing  as FED magnitude increased from 51 flashes/5 min to 99 flashes/5 min. MFA also became more concentrated NW of the City of Baltimore and Baltimore County as seen in images 9 and 10. WFO LWX issued a Tornado Warning around 20:51 UTC.

An animation of the TDWR at Baltimore Washington Airport (image 11) had signs of a possible low-level RFD surge (not shown) in the 0.5 degree rapid scan tilt and increasing low level rotation consistent with either a stronger surge of straight line winds or a QLCS mesovortex/tornado. Aloft, (not shown) there did appear to be healthy reflectivity aloft and the concentrated MFA may suggest that a strengthening updraft.  

Image 12: 21:02 UTC Showing ProbSevere Time Series with KDOX 0.5 data (left panels) and GOES/GLM data (right panels) + ProbSevere

By 21:13Z The storm of interest is looking LESS favorable for tornado although prob severe tor increased significantly to closer to 20% perhaps due to an elongated zone of low level shear.

Image 13: 21:13Z KDOX Radar Imagery

Prob Severe Table Ideas:  If the tables could open in a floating tab in Awips that would be very helpful. This way you can manually dock and move the tab around in awips. This way you can quickly view the tab and keep it open until you want to just close it.  You can also rename the tab to whatever will help you keep track of the storm that it belongs to. This would get around needing to color code multiple tables etc.

A separate tab will also allow room to show additional information (beyond just prob severe).

Prob Severe jumped to nearly 80% with a fast moving bowing segment through Montgomery County.

Storm in question is the southeastern storm here.

– By Miles and Dwight Schrute

Poorly Modified NOAA-20 Sounding

 While reviewing NOAA-20 soundings, I picked a point that may have had convection later in the day.  The regular sounding was fine.  The modified sounding looked crazy.

– Champion

Observations Over the Sterling CWA

 Looked at the modified NUCAPS sounding for and the low levels, below 700mb, were un representative (had an inversion when SPC mesoanalysis had no CINH), however the sky had roughly 80% cloud cover.

NUCAPS Base Sounding

NUCAPS Modified Sounding, Same Location as Above

Looking at the NUCAPS forecast, the holes in the output field due to the cloud cover.  The lack of data was in a bad location, preventing us from seeing the instability potential for a line of storms coming in from the west.  The gridded format was actually better to use in this case as it helped fill in the gap.

Interpolated CAPE NUCAPS Imagery

The interpolated data is easier to visualize gradients in the variables, but our experience was that some important data was filtered out by having this turned on.

The time in the lower left is 19.99z.  A key for the “all” field would be helpful to understand what I am looking at.

Having the CWA borders is handy, however having it as it’s own layer would be more helpful, and separating out the CWA borders from the state borders.

Can storm names be used to correlate the time series (F6, D3) and also have the names plotted in AWIPS for the storms I am looking at a time series of; would be more beneficial than having the lat/lon

-for example, click for a time series of one storm triggers a storm ID to show up in the time series and in AWIPS

-I click another storm and another time series shows up with the storm ID in the time series and in AWIPS allowing me to see which time series goes to which storm

If prob severe and its time series could be put into GR that would greatly improve DSS services when outside the office.  The AWIPS thin client is sloooow, so being able to have the same ability, or similar ability to interrogate storms as in the office would greatly help improve the quality of DSS when deployed.

Prob severe version three continues to look more reasonable for severe wind than version two

Noticed an increasing 5 minute trend in the minimum flash area that was reflected in the FED about 5 minutes later.  Seeing the sustained increase in one minute minimum flash area caused me to pay more attention to that storm than I did earlier due to the sustained growth

While monitoring a storm with FED and minimum flash area, the FED suddenly went down.  The same trend was not seen in the minimum flash area as easily.  Maybe the minimum flash area is more useful tool for monitoring the growth of storm while the FED is better suited for monitoring the overall trend in storm strength and sudden weakening.

Downward trends in FED for one of the storms matched what was being seen on satellite of the storm updraft becoming more ragged as it weakened due to entraining dry air.

-would be great to have lightning data such as FED plotted in a time series as well so trends are more easily seen

The stronger storm we were monitoring (same as in the screenshot below), prob severe version 3 was higher than version 2 for 15 minutes atleast.  Looking closer this was due to the hail category being higher than version two; version three was also higher than version two in the wind category, but not nearly as much.  Toward the end of our time version two was higher due to higher probabilities in the wind category.  Interesting.

We monitored this particular cell on and off throughout the afternoon and tried to gain a better understanding of the minimum flash area. We noticed a close cluster of negative strikes, which really helped as a visual aide for what the GLM was seeing. The GLM MFA was able to isolate the core of the storm really well. We combined this with ProbSevere, and watched the probabilities on this storm increase which was a further confidence booster that the storm was intensifying in addition to what was seen by GLM and ENTLN.  

– Accas and Groot

Optical Flow Winds Versus GOES-East DMW

 The optical flow winds product had much better areal coverage than the GOES-East 000-250 mb DMW product for strong thunderstorms in the San Angelo CWA.  However, it would have been very nice to have both products in virtual AWIPS to compare.

– Champion

Day 2 Thoughts

 NUCAPS…

Can there be a circle (or some reference) around the NUCAPS point that I am currently using for a sounding?  That way I have a reference on the map for where the sounding is that I am looking at.

Can more than two NUCAPS sounding be loaded into an AWIPS pane?  If so, would nice be able to compare soundings/environments more easily.

Having the ability to display one NUCAPS sounding when I have two loaded in Sharppy would be helpful.  Even when loading two soundings and only have one in “focus” the two soundings overwrite each other.  Can this setup be similar to AWIPS that allows us to have multiple soundings loaded and be able to turn one or both of them on when we choose?

Compared a couple NUCAPS soundings surface conditions to the obs for a couple locations and they look reasonable.

Looking at the forecast CAPE/CIN from NUCAPS, the gridded field for CIN is quite splotchy.  Bulls eyes of much higher CIN seem overdone compared to what is expected during the mid afternoon with full sun and with what the SPC mesoanalysis has.  This would make me question how accurate it is.   Looking at the forecast, there is no consistent trend with the CIN bulls eyes, which lowers my confidence in this field. The CAPE field is more uniform, though still splotchy.  The areas of higher CAPE are more consistent, giving me more confidence in this field than the CIN.  Is there a way to average out this field more to make it smoother?  If so, that would greatly increase my confidence in this parameter and my likelihood of using it in the future.

Noticed the surface based CAPE in AWIPS vs. Sharppy was quite a bit higher in Sharppy.

Compared the ML CAPE in a modified NUCAPS sounding in AWIPS and an unmodified NUCAPS sounding in Sharppy and the modified lined up much more closely with the SPC mesoanalysis page.  The the ML CAPE in the unmodified sounding in Sharppy was too low.  Surface based CAPE was actually more representative in the unmodified sounding.

As mentioned earlier, would be nice to compare more than two sounding points for NUCAPS to aid in comparing the environment more easily.

Having the NUCAPS 2m temperature and DP in F instead of C would be much more useable and easier to compare to surface observations.

Noticed the 2m temperature for the gridded NUCAPS was cooler by 5-8 C compared to the observations.  This makes me looks confidence with the CAPE and CIN plots if the surface temperatures are not accurate.  Is there a way to grid the modified NUCAPS data?  When I forecast I like to view parameters in a gridded fashion in the horizontal.  This helps me better understand what is going on with the environment.

Compared the NUCAPS 700-500mb lapse rates to those on SPC’s mesoanalysis page and found the NUCAPS was close, but on the cool side.

In our data sparse CWA, I can see these soundings as being quite useful, as long as forecasters understand the low levels (assuming below 850mb) are less likely to be representative.

Taking a look at the minimum flash area…

Difficult for me to really see any sort of trend with the 1 minute data.  Nothing really catches my eye.  The 5 minute data is much more easy to see trends.  

As mentioned yesterday, am able to see more valuable information with trends in the storm than with flash density. 

Looking at the optical winds…

The background is a bit too dark.  Can the Lat/Lon be put below the imagery?  Having it above seems to detract from what is being displayed.  Adding state borders, cities, would add to the usability of this product, especially if these labels can be turned on and off.

I like being able to pan the image.

I can see this being handy for monitoring for LLWS for aviation, assuming there are clouds to track.  Could this data be merged with NUCAPS to plot shear and helicity?   

Changing the density of the vectors would be handy.  

Color coding the different levels and matching it to the key is easy to determine what level I am looking at.  

Could this track the speed of dust?  If so, could help determine how strong the winds are in dust storms.

Curious why the pressure levels are broken down into 200mb intervals.  Could the winds also be tied to theta levels to help with isentropic analysis?

Having contours for the winds would help limit information overload as far as what is being shown.  Being able to control the density of the number of wind vectors would help, however that could lose some data.  Contours of the wind vectors, say every 5 or 10 kts, could help summarize what the individual vectors are showing.

-Accas

Sampling of Sub-Severe Convection Across the Southeast

 Modified NUCAPS sounding appeared to have a better handle on the environment compared to baseline NUCAPS sounding. However, it also appears it might not be totally representative of the atmosphere given the partly cloudy conditions at the time RTMA data was pulled in. The SPC mesoanalysis page suggested MLCAPE upwards of 2500 J/kg in an uncapped environment.  Using the gridded mid level lapse rate product from NUCAPS we found the data to be representative. It verified well with what was shown in the NUCAPS soundings and matched with the values suggested by the SPC Meso Analysis page.

MODIFIED SOUNDING 1927UTC – WEST HINDS COUNTY, MS

1927UTC BASELINE SOUNDING – WEST HINDS COUNTY, MS

21UTC 700-500MB LAPSE RATES

The NUCAPS mid-level lapse rates were fairly representative when compared to the SPC mesoanalysis page. This was further evidence that large hail was probably not going to be in the cards for the Jackson area today, but marginally severe wind gusts would be something to watch.

16 UTC 700-500MB NUCAPS GRIDDED LAPSE RATES, SAMPLE NEAR THE SOUNDING DATA POINT

Prob Severe version 2 vs version 3, particularly in prob severe wind:

In this event, the prob severe there was a sig wx statement and severe thunderstorm warning put out by the Huntsville office. Around that time, the prob severe was increased specifically for the prob severe wind component. The version 2 had a prob severe value of 3% while the version 3 had a 53%.  Version 3 better captured the significance of the storm with a 40 mph gust reported around the same time.  This is significant since we were also discussing how filtering lower prob severe thresholds would be useful in decluttering the operational screen. We could have missed this event if that was the case (with version 2).

Based on this experience, we can see the vast improvement in the wind component of prob severe version 3.

21:16 UTC ProbSevere Sample (note V2 versus V3 differences in sample).

Local Storm Report of a measured 41mph wind gust in Colbert County AL, just north of the contoured ProbSevere storm.

1 Minute FED overlaid with ENTLN 5 minute (1 minute update) and GM Flash Point

GLM Observation:

Saw a steady lightning jump depicted in the GLM FED correlated with a storm that NWS Huntsville issued a Severe Thunderstorm Warning on. Several mPING reports of wind damage (assuming sub-severe with no LSRs issued as of this time) which raises confidence that storms are intensifying. Perhaps the most interesting thing about this screenshot is seeing the parallax-correction in action when looking at the Flash Points.

Optical Wind:  

Still difficult to tell where you are geographically.  Suggestions to perhaps swap the lime green grid with the state outline colors.

Additionally, the time stamp gets cut off if you zoom in on the product and occasionally gets covered by the wind barbs themselves. Perhaps a floating time stamp would be better for this instance.

Time stamp for the image above.

A note about the timestamps:

The time stamps attached to the wind barbs appear to be formatted incorrectly, showing times like 21.97Z, note the actual time correctly formatted shown in the image below was actually 21:58Z.

– Groot and Dwight Schrute

NOAA-20 Pass Ahead of Severe Convection

 The San Angelo CWA was expecting severe convection in the afternoon.  There was a NOAA-20 pass over their CWA at 1927Z.  A NUCAPS Sounding in the clear air ahead of ongoing convection was chosen.  The approximate point of this sounding is shown by the white circle in the left image.  The sounding in the right image showed an environment very favorable for severe convection, including hail.  Since ongoing severe storms were heading in this direction (left image), the storms could be expected to maintain their intensity or possibly strengthen.

– Champion

GLM Flash Point Product

The GLM Flash Point is a unique addition to the GLM suite of products.  It’s parallax corrected, which is nice.  But the points seem to tell you less data per minute than the FED.  In this example there are eight points for the Sterling/Irion County storm.  However, you need to mouse over each point to get more data (flash duration and area).  By comparison, the FED quickly tells you this is an electrically active storm.  In a warning environment, with limited screen space, and where every second counts, the FED tells you a lot more very quickly than the Flash Points.

 

– Champion

New Mexico Severe Storm

 Radar imagery showed a storm form and rapidly strengthen over Guadalupe County, NM.  ProbSevere matched this intensity increase well.  The storm looked like it was going to become severe.  ProbSevere jumped to 60% at 2226Z, which was about a 40% increase in about 5 minutes.  Surprisingly, the 60% value was the same for ProbSevere Versions 2 and 3.  The modelers mentioned that Version 3 has lower values than what forecasters are used to seeing in Version 2.  Therefore, a 60% value for Version 3 is probably a higher threat than an identical value for Version 2.  That gave me more confidence considering this storm severe.

– Champion