Scottsbluff storm

ProbSevere v3 (PSv3) has shown a consistently higher signal than ProbSevere v2 (PSv2) on storm near Scottsbluff, NE. PSv3 was 20-40% higher during the development stage and as it matured. PSv3 also had consistently (and correctly) lower tornado probabilities than PSv2. The elevated tornado probabilities was likely a result of very high MRMS azimuthal shear, which was a result of sidelobe contamination. In this case, ProbTor v3 correctly didn't emphasize the azimuthal shear as much as PTv2. ProbHail v3 has been the highest hazard probability, whereas ProbWind v2 was the highest for version 2. One forecaster noted that v3 is definitely picking up on the hail threat better than v2. Hail up to the size of 1.5"-diameter has been reported thus far. Forecasters have noted that the higher PSv3 values seems to correlate better with their expectations of the storm.

Figure 1: ProbSevere v3, MRMS MergedRef, and NWS severe weather warnings for a storm near Scottsbluff, NE. 

ProbSevere Ingredient Distributions

Many of the NWS forecasters appreciated seeing ProbSevere 1-d and 2-d distributions of the data fields used to calculate probabilities.   Forecasters wanted to see these because they wanted to understand why ProbTor values may be higher one day versus lower the next.  Or why one day may have more 90%+ ProbHail values versus another.

One topic discussed was a 2-d NWP-based predictor in ProbHail (figure below).  The red box indicates a very favorable environment for severe hail (ratios near 10, which act to increase ProbHail values, representative of OUN/SJT CWAs on May 1, 2019) and the yellow box indicates a more marginal environment for severe hail (ratios near 1, which are neutral (do not significantly increase or decrease ProbHail values), representative of OUN/TSA CWAs on April 30, 2019).  This was one reason forecasters were seeing more 90%+ ProbHail values on May 1 compared to April 30, when similar values of MESH were observed.

Forecasters really appreciated seeing this level of detail as a mechanism to better understand ProbSevere and how to interpret and use probabilities from day to day.

ProbSevere-ProbHail 2-d predictor of hail cape (CAPE -10C to -30C) vs precipitable water (PWAT).  The red box represented conditions in the OUN/SJT CWAs on May 1 2019 and the yellow box represents the conditions over OUN/TSA CWAs on April 30 2019.

-J. Sieglaff

A Quick Start to the Convective Day

After our day 2 debrief, forecasters began operations as severe thunderstorms were already developing over the OUN and SJT.  A common theme from the day 2 debrief, was forecasters really liked the trends captured by ProbSevere.  While work continues on getting ProbSevere time series plotting capabilities in AWIPS-II, forecasters are utilizing time series capabilities from the ProbSevere website.  Forecasters mentioned they appreciated being able to quickly see how the first storms evolved by using the ProbSevere time series.  Also noted by forecasters were strong satellite growth rates and very high CAPE in the hail growth region lending to very rapidly increasing ProbHail values, even before MESH reached 1.00".

ProbSevere time series capability from the ProbSevere website for the first severe thunderstorm of the day in the Norman WFO, near the Red River.

AWIPS-II 0.5 degree reflectivity and NOAA/CIMSS ProbSevere for the evolution of a rapidly developing thunderstorm near the Red River on May 1 2019.

-J. Sieglaff

Underwhelming ProbHail

ProbHail on the storm in southwest MO seemed to be too low based on other derived parameters (MESH/Hail Index) and storm structure. ProbHail was consistently 70-75%. It was explained that the hail cape (~750 J/kg), the EBShear (47kt) and lightning (29 flash/min) were relatively lower than one would expect given the environment. Justin showed the parameter space for each of the ingredients and how they were weighting down the overall ProbHail due to the ratios being <1.



-- SCoulomb

UNR - GLM/ENTLN lighnting data on a severe storm in SD

ProbSevere indicated a severe thunderstorm that explosively developed along the SD/NE border. This storm initiated in a favorable environment & I anticipated it to further develop, so the rapidly increasing values in the ProbSevere values along with an increase in CG activity from the ENTLN were plenty for me to feel confident in issuing a warning for 2" hail.20 minutes after my warning, ProbHail had increased to 99% with a MESH value of 2.25".

For comparison, I looked at the GLM data along with observed lightning data from the ENTLN. The ENTLN showed an obvious increase in CG activity beginning around ~0015Z from the ENTLN data. However, there never was an obvious increase in GLM flash extent density or total energy.

GLM/ENTLN loop:

Peter Sunday

UNR - Warning isssued for storm moving into SW SD. ProbSevere.

2310Z: After a lengthy internal struggle I decided to issue a SVR on a strong storm moving from NW NE into SW SD (2206Z issuance time). Primary reason for issuing was ProbHail which had values of 92% & as MESH of 1.62" right before the storm crossed into the CWA and a warning decision had to be made. This storm had rapidly intensified in the prior 20 minutes with a drastic increase in ProbSevere & MESH values & was moving into the area of our CWA I was most concerned about for severe weather this evening.

However I had reservations based on the near-storm environment the storm was moving into from SPC Mesoanalysis, which showed mixed-layer capping still in place. Furthermore the all-sky CAPE showed very low values less than 500 J/kg across SW SD & NW NE. Initially I dismissed this product as having values unrealistically low but I did notice an increasing rend in CAPE advecting into our CWA. Based off of a conventional near-storm environment analysis I was not a fan of the environment the storm was moving into but it's really hard to ignore a storm with ProbHail > 90% & MESH > 1.5 moving into your CWA, while under a severe thunderstorm watch, with satellite data showing an increasing trend in instability. Looking back at the storm after the fact, the MRMS reflectivity at -20C & 50 dBZ height were right on the line for issuing but at a threshold where I would have been willing to wait a bit & see what happened as the storms moved deeper into the questionable environment in our CWA. Without the tools we're testing this week I likely would have issued a SPS given the lack of reports from this storm & waited a few scans to see how it reacted to the environment across our CWA. In this case the storm began to weaken almost immediately after I issued the warning.

GLM data also was not that impressive on the storm even showing the weakening trend while reaching peak ProbSevere & MESH values I've witnessed with other storms in the west this week. I think this even reverse trend in GLM lightning activity added to my decision to issue a warning based on seeing it with other severe storms in ABQ/MAF so maybe I was anticipating a trend that doesn't actually exist.

Warning decision making relies on confidence in your understanding of the near-storm environment, & that the tools you are using are accurately depicting what is going on in the storm/environment. With this week being the first time we have seen GLM & all-sky products, & with limited use of ProbHail/Wind/Tor, I found it hard to decide which ones to trust when faced with conflicting warn/don't warn information between them and alongside more familiar conventional  products. Obviously this will change as we get more familiarization & training with the products which is one of the primary benefits of these HWTs!

GLM/ProbSevere loop of event:

All-sky CAPE/MRMS RALA loop of event:

Peter Sunday

UNR - SVR issued off of ProbSevere. Not a great idea!!

2125Z: Strong/Severe storms continue to move from SE WY into the UNR CWA as expected so took after quickly glancing at GLM/ProbSevere after a blog post, decided to issue a SVR based on ProbSevere & MESH values. WHen I first noticed the storm it was nearing the CYS/UNR CWA boundary & had a ProbHail of 77%, ProbWind of 53%, & MESH of 1.68". Loopping through the last 10 minutes of data I noticed an increasing trend for all values so issuing a quick warning for golf-ball sized hail based on MESH & ProbHail values seemed like a no-brainier to me. Furthermore GLM total energy showed relatively high values compared to other storms in the area so far this afternoon.

Unfortunately... after spending time on a blog post I committed the cardinal sin of warning operations & lost my situational awareness before issuing the warning. The storm began to weaken almost immediately after I issued my warning ProbHail values quickly dropping to less than 10% & GLM lightning data showing a clear decreasing trend as well. SPC mesoanalysis clearly showed these storms moving into a poor environment in NE WY with less than 500 J/kg MLCAPE, less than 30 kts effective bulk shear, & less steep lapse rates as well. I looked at the all-sky CAPE post warning as well but it was not useful in depicting the CAPE gradient from south-north in eastern WY.

Obviously it goes without saying but situational awareness & a proper analysis of the near-storm environment is essential in a warning environment. ProbSevere cannot be used by itself in a vacuum for warning decision making but rather as part of a package with other storm interrogation techniques.

ProbSevere/GLM 4 panel animation:

2030Z all-sky CAPE:

Peter Sunday

MAF - SVR issued based on increase in lightning activity

2300Z: I decided to try & get a jump on the storm east of the already warned one in eastern Jeff Davis county. GLM was my primary tool used in deciding to issue a warning based on an increasing trend I began to notice around 2230Z. between 2230 & 2250 Flash density increased from 10 to 30 & total energy from 50 to over 100. The animation below I could have issued a warning even earlier when I first noticed this increase in lightning activity but waited until I saw more of an increasing trend in prob severe. ProbHail was at 6% when I first noticed the increased lightning trend & I waited until I saw an increase to 14% & MESH increasing from 0.5" to 0.74". At the same time I noticed a cooling trend on cloud tops on the GOES 16 clean window IR which added even more confidence to the lightning data that this storm would continue to intensify. Unfortunately I fumbled around with WarnGen a bit so my warning wasn't actually out until 2051Z but I made up my mind on issuance & began the warning process around 2040Z

GLM 4 Panel animation:

Around 2300Z the storm became more obviously severe with ProbHail of 84%, ProbWind of 68%, & MESH of 1.38".  I'm not anticipating receiving any reports out of this storm given the sparse population in that part of TX but given those numbers I would certainly consider this storm a severe candidate. Taking a look only at "conventional" storm interrogation products such as 50 dBZ height for 1" hail & isothermal reflectivity at -20C I don't think I would have made a decision on warning the storm until the 2255Z scan. Shortly after 2300Z the storm even took on supercell characteristics & began a turn to the right which certainly would have resulted in a warning. GLM products & ProbHail/ProbSevere without a doubt added value to my warning decision making in this case.

Of note is how the GLM fields actually decreased towards the end of the above loop despite the storm seemingly reaching it's peak intensity. I noticed similar trends yesterday watching storms in ABQ so this may be unique to GLM & storms out west.

UPDATE (2045Z): decided to reissue on this storm given  a steady trend in high GLM Flash Density Values, Prob Hail over 60%, & MESH near 1.5". Below is a 4 panel animation of the GLM fields/ProbSevere/& a MRMS 1km RALA over the whole period I was actively watching it.

Peter Sunday

ABQ - 2300Z Update

2240Z: Heading into 23Z, ProbSevere indicated a likely severe storm SE of ABQ with ProbWind 88%, ProbHail 76%, & even ProbTor of 21%. MESH values are slightly higher than 1", & the prescence of a mid-level mesocyclone along with -20C reflectivity greater than 50 dBZ all indicating the potential for severe hail.
Interestingly enough, there continues to be little correlation between increasing trends in the GLM fields & increasing storm severity (gif below). This has been noted in past HWTs with western storms which may be a limitation with using GLM products in this region.

Zooming out, GOES-16 IR/VIS data shows some cu development & cooling cloud tops across southern NM.

Looking at the CI fields, not too much is expected with this development with values generally 50% or lower & severe CI probs only around 10%

And finally, little to no increasing trend in GLM Flash Density makes me even more confident in anticipating little to no development of this convection over southern NM

Peter Sunday

CWA TomClemmons RAH 05-10-2018

Synopsis: Well removed from the better upper dynamics/forcing and shear
that will move through the NE US this afternoon and evening,
forcing across central NC will be rather weak, limited to mainly
weak near sfc convergence, as the surface trough crosses the area.
However, strong daytime heating, featuring afternoon temps in the
mid 80s, will result in steep low-level lapse rates. These steep low-
level rates when coupled with with BL dewpoints in the lower 60s,
will fuel moderate instability of 1500-2000 J/Kg that extends well
into the mixed phased region. Despite the deep westerly flow in
place, majority of the CAMS indicate a fairly congealed multi-cell
line segment crossing the area between 18z to 03z. Given 20 to 25
kts of deep layer shear, some of the strongest updrafts will have
the potential to produce severe weather, including, hail and
damaging wind gusts.

Hampered by slow computer systems Tuesday, Wednesday, and Thursday:(

Notice low correlation between  -20 C Isothermal Reflectivity & GLM Total Energy 5min-1min update. Notice storm near the cursor.

In this example the storm is on the northern RAH CWA border, the updraft has pushed 50-60dbz echoes up through the -20 C isotherm, notice this time the GLM Total Energy and the Extent Density (# of Flashes that have crossed a pixel) are both showing high values. This would be a good situation awareness tool during an ongoing severe weather event. This storm was currently producing damaging hail and strong winds.

ProbSevere Wind and Hail:

1.75" hail was reported at 5:25 pm near Bahama.

ProbWind:

3NW of Stovall 2 Trees Down

Splitting of Supercells and Prob Severe/Tor/Wind

A supercell that had been warned on split with the northern storm moving into the more favorable area of instability and the updraft rapidly strengthened to severe. Noticed that Prob Tor has increased to 9% on the northern storm and Prob Hail/Wind are 99%/89% respectively.  Issued a severe thunderstorm warning shortly after the last scan on this animation. Will be monitoring that northern storm for ProbTor. Have seen some upstream storms with strengthening with rotation.

More to come...

Supercell

Severe Storm with All Sky CAPE Prob Severe

Interrogated a storm entering the gradient in instability shown by the All Sky CAPE products and also increasing Prob Hail and Wind Products (cell with SVR box on it). The cell is moving into Garden County Nebraska:

More to come...

Supercell

ProbSevereWindTor with severe storm

Looked at Prob Severe Tor/Wind/Hail with MRMS reflectivity. The Hail probabilities actually increased to 90% prior to warning issuance. MESH values were around 1.56 inches. Prob Wind was a lower (78%) value. Did see where there was a TBSS in the dual pol and reflectivity products indicative of hail. Issued severe thunderstorm warning shortly after the probs showed this. Did notice that Prob Tor was around 1% but monitoring the velocities as the cell is located near the warm front and wind shear is sufficient to support it.

More to come...

Supercell

Prob Severe & Wind Flurry of readouts

Looking at the Probabilistic products during the latest warning operations, if there are 2 or more cells within warnings, it is definitely useful to compare the parameters in the readout and probabilities. It would helpful if the color table for the Prob Severe, Hail, and Wind could have a better gradient in colors from 70% and above. Also, did not find it useful to have MESH in the Prob Wind product readout, nor did I find it useful to have the flash rate especially with the GLM data available for us to use. This image also displays why having so much on the readout for Prob Severe is too much on the screen.

Supercell

ProbHail Overestimating Severe Hail in Montana.

MRMS Composite Reflectivity over southeast Montana on Monday afternoon 2022 to 2136Z with ProbHail overlaid. Convection firing off near the Billings area with individual cells displaying high ProbHail values of 50 to 80 percent in the initial and mature phase of development. This was due to MESH values of 1" to 1.7". Issued 3 warnings for hail size of 1" but storm reports of only 0.25" to 0.88" were received before the storms weakened as they worked east of the Billings area.

- Jack Swigert

Severe storm outbreak in the Mississippi and Ohio River Valleys

A potent shortwave forced severe storms across the middle of the U.S. yesterday, wreaking havoc from Texas to Ohio. Below is the 850mb analysis from 00Z 04/04/2018 (Figure 1), showing well-defined warm and cold fronts emanating from the low, centered on Lake Michigan. We also see that moisture was well-established in the warm sector and the flow at 850mb was quite strong (30-40 kts).

Figure 1: 850mb analysis from 00Z 04/04/2018.

The NOAA/CIMSS ProbSevere model is experimenting with hazard-specific models: ProbHail, ProbWind, and ProbTor. The ProbHail model did fairly well to capture the initial hail threat along the cold front, with numerous storms exhibiting strong satellite growth from GOES-16.

These storms produced numerous hail reports between 1" and 1.5" in and around the Mark Twain National Forest.

Figure 2: ProbHail for storms in eastern MO, with MRMS MergedReflectivity and NWS severe weather warnings.

With sampling turned on in AWIPS2 (Figure 3), users of ProbHail can see the constituent predictors. Here, the strong satellite growth rate, high MRMS MESH (1.5"), and high Earth Networks® total lightning flash rate (40 fl/min) contributed to a ProbHail of 96%. NWP predictors of effective bulk shear, CAPE between -10C and -30C ("hail cape"), precipitable water (PWAT), and the lowest height of the wetbulb 0C isotherm are also used as inputs in ProbHail.

Figure 3: ProbHail readout for storms in eastern MO, with MRMS MergedReflectivity and NWS severe weather warnings.

Due to very favorable kinematics, the tornado threat was rather high, with the Storm Prediction Center issuing a wide swath of 10% probabilities for tornadoes within 25 mi of a location for their 2000Z outlook (Figure 4).

Figure 4: SPC tornado outlook and verification for 20Z and 12Z, respectively.

A storm that produced numerous reports of significant wind damage in the Dayton, OH metro area also produced a tornado east of the metro. ProbTor captured it's development well (Figure 5).

Figure 5: ProbSevere, MRMS MergedReflectivity, and NWS severe weather warnings for southwest OH.

The time series of this storm's attributes demonstrate that very high MRMS 0-2km AzShear (>20 * 0.001 /s), along with SRH of 300-400 m2/s2 in the 0-1km layer helped spike the ProbTor value to over 90%. In this case, the tornado report came slightly after the time of peak ProbTor. Several wind reports in the Dayton area were significant (e.g., "large debris cloud", "structural damage to homes"), around 20Z when ProbTor was in the 30-50% range. Thus, it's possible there may have been a tornado earlier (a NWS survey should confirm the presence or lack thereof).

Figure 6: Time series of ProbTor (red curve) for this storm with its predictors. The bottom axes show NWS warnings and reports.

The "ProbSevere" value simply uses the maximum of ProbHail, ProbWind, and ProbTor. For this long-lived storm, each model was the maximum at different times. Regardless, the ProbSevere values were mostly over 70% when there was reported severe weather. The MESH and lightning (pink and green curves) show the cycling of this storm nicely. The storm produced hail and wind reports in the Columbus, OH metro and later east of the metro.

Figure 7: Time series of ProbSevere (red curve) for this storm with its predictors. The bottom axes show NWS warnings and reports.

Storms later evolved into big wind-producers, evidenced by these storms in the Nashville, TN area. There were many reports of trees and power lines down in the metro area around 0200-0230Z.

Figure 8: ProbWind, MRMS MergedReflectivity, and NWS severe weather warnings for storms affecting the Nashville metro.

 ProbWind was in the 50-70% range for these storms, owing to a strong MeanWind in the 1-3km layer and moderate 0-2km MRMS AzShear. Note that the ProbWind readout in AWIPS2 now has a "weak", "moderate", and "strong" designation tag for the MRMS AzShear fields (e.g., Figure 9)

Figure 9: ProbWind readout for a storm south of Nashville.

At the end of each day, we make an "accumulation" product, plotting the centroids of storms colored by their ProbSevere value (or ProbTor value), along with the NWS severe weather warnings and SPC preliminary reports. From Figure 10, we see that ProbSevere corresponded with severe weather warnings and reports quite well in MO, AR, IL, IN, OH, KY, PA, and TN, and parts of TX and LA. However, many wind reports were missed (at the 50% threshold) in MS, AL, and GA. Many of these storms had meager reflectivity and lightning flash rates, and no discernible satellite growth. ProbWind was largely 10-30% for these storms, a small improvement over the legacy "all-in-one" ProbSevere product, which had < 10% for most of these storms. We are still working to improve upon probabilistic prediction of storms in this regime, and are open to ideas from the field.