Prob Severe With Line West of D.C.

The image below shows a 4-panel display of the prob severe product for the line of strong to severe storms near the D.C.  area this afternoon. This includes the experimental ProbWind (upper right), ProbHail (lower left) and ProbTor (lower right). Up to this time (2212z) ProbWind had been running 80% or better and ProbHail at 90% or better. ProbTor stayed below 10% and would typically be lower anyway, especially in a QLCS situation. WFO LWX did issue a tornado  warning at 2212z.

Al Cope

Convective Initiation Product

Given the strong instability across the region, I was anticipating a greater signal in the CI product. AllSkyLAP CAPE values sstill exceeded 2500 J/kg in the area and Cu were finally beginning to bubble up, but the CI product showed little in the way of probabilities.

Fig 1.  AllSkyCAPE indicating 2500+ J/kg at 2200 UTC.

Fig 2. CI Probability product at 2237 UTC.

I'm sure there are other limiting factors that I'm not taking into account but I found this discrepancy and departure from my conceptual model to be interesting.

wall_cloud

Severe Weather Over Virginia/Maryland

It was interesting to see the evolution of severe weather to the northwest of Washington, D.C. A line of thunderstorms in southern Pennsylvania and northern West Virginia was moving to the southeast, while isolated cells developed and became severe ahead of the main line near Madison, VA. It was interesting to see how the lightning from the GLM data ramped up quickly shortly before the storm became severe (it produced quarter sized hail).

I was able to note the 1 minute average flash area spread will out ahead of the precipitation area in front on the storm and also extend in the anvil behind the storm as well. As a broadcast meteorologist, it was a great way to explain to viewers the lightning threat away from the precipitation.

Gary Cannalte

Anvil lightning

The GLM data is providing a great example of anvil lightning this afternoon. Per the ENI lightning data, no ground strikes are occurring to the north of the primary cell, but there are Pulse/Flash detections.

Fig 1. Comparison of the GLM and ENI lightning data.

The GLM data shows a large area of low flash density, but the total energy signature and average flash area are relatively high.  I find this to be very useful in the IDSS realm where lightning can occur at various distances from the storms themselves. The GLM provides a good situational awareness whereas the ENI data is somewhat difficult to see given the small size of the data points.

wall cloud

Late Afternoon Convective Trends

Convective development continues to increase near the northwest portion of the forecast area late this afternoon. This broken convective line extends from the Kansas City area into the Texas panhandle along a frontal boundary/dry line.

Day Cloud Phase Distinction RGB/Day Cloud Convection RGB/GLM Flash Extent Density

The airmass to the east of the boundary is becoming very unstable with surface temperatures near 90F and surface dew points near 70F. CAPE values on AllSKy LAP CAPE are around 3000 J/kg. However, a similar NUCAPS sounding suggests CAPE values near 5000 J/kg in the same area. These values are closer to those suggested by some high resolution guidance.

AllSky LAP CAPE

NUCAPS Sounding East of Dry Line

The main limiting factor for severe storms this afternoon is relatively weak shear, with bulk shear values  around 25 kts. However, shear profiles will increase somewhat over the next couple of hours as a weak upper-level disturbance interacts with the dry line. Some of this enhancement can already be seen with the development over the Texas panhandle.

Alex Brown

Watching lightning initiation

I wanted to compare the ENI Total Lightning with the GLM when it come to detecting the onset of lightning within a developing storm. A storm was developing along the OK/KS border between 19-20z.  Overlaying both the ENI and GLM data, they were actually VERY close.

Fig 1. Initial lightning dection on storm near Ponca City,

 OK.

The ENI had the earlier detection in this case, but only by one minute. One thing that jumped out very quickly was the parallax issue which caused the ENI and GLM data to be offset by ~10 miles, each detecting lightning on different sides of the state line.

wall_cloud

Forecast for 5/14/18

Significant instability is currently observed across the Southern Plains. AllSkyLAP CAPE values on the order of 3000-3500 J/kg were detected over north central OK and southern KS. Surface temperatures in the lower 90s were already allowing storms to develop as of 19z with more widespread convection anticipated over the next few hours.

Fig 1. AllSkyLAP CAPE at 1900 UTC.

Deep-layer shear is not optimal for the development of supercells today, but effective bulk wind difference ~30 kts will support mutlicellular convection along the quasi-stationary front and southward in the vicinity of the dryline. Expect multicell clusters to develop this afternoon, providing a significant hail threat given the steep mid-level lapse rates (NUCAPS 700-500 mb lapse rates exceeding 8 C/km) and ample buoyancy. As cold pools begin to develop in a deep mixed boundary layer, storms will likely grow upscale into several loose multicell systems with the severe weather threat transition more toward wind.

Fig 2. NUCAPS 18 UTC sounding near KPNC.

wall_cloud

Week 3 Day 1

Today begins the third week of the Satellite Proving Ground Experiment in the HWT. Today we will be beginning operations in the Sterling, Va and Norman, OK CWAs. These will experience totally different environments as a line of storms is expected to move over the Mid-Atlantic this afternoon and severe storms should develop along the dryline in western Oklahoma later today. It will be a good test to evaluate the products in these two different environments.

-Michael

Week 2 summary and comment round up

As week 2 comes to a close 11 May 2018, we review the comments from our final debrief below. Much of this echos the discovery and opinions from throughout the week.


GLM:

• Training remains an important question and consideration.  Would love to have subject-matter expert available locally for all forecasters (may become frustrated by the data and not use it w/o that person).  
• SOO or focal point can be that local subject-matter expert, however, this will demand evolving training exercises.  Move beyond VLAB, work with WDTD to create hands-on and locally-relevant training opportunities.
• Typical order of use this week:  Total Energy (primary), FED (secondary) and Flash Size (tertiary). 
• Limit the amount of data available on first release (only to the three products listed above???); eliminate point-based or centroid products.
• Integrate GLM data with ground-based lightning systems to help understand and alleviate parallax concerns, particularly when issuing a special weather statement or IDSS for lightning.
•  Want to work more cases and focus areas.  E.g., where radar coverage is poor (western US), fire weather, and aviation -- appears that areal extent and pre-Cloud-to-Ground information will be important.  Context is important and will drive various need/demand for individual GLM products.

ProbSevere (all hazards):

• Did not trust the probabilities in all environments this week.
• Did well with hail and if storms were a bit more steady state.
• Seemed to behave worse in western US and with wind threats.
• Liked using the probabilities as trend information for storm intensity, but not necessarily a particular number.
• Would like to see trend graphs not only of probs, but also internal items in algorithm.
• Preferred the individual hazard products (read-out too long on single ProbSevere)
• Some confusion regarding time scale of probability (minutes vs hour)
• ProbTor could be more difficult to understand reasoning for probs than other hazards; Use a lower threshold for ProbTor and adjust color tables accordingly.
• Would like to see "weak" "strong" indicators on more than just glaciation rate. 
• Difficult to pick out differences in storms between 70-100% probs; color table could use more delineation at higher end (all storms appeared pink).
• Could be useful to communicate impact w/public - would remove numbers and use Low - Mod - High for threats.

Convective Initiation (CI) and Severe-CI:

• Not terribly impressed by algorithm:  completely missed some objects, overdone in some areas, underdone in others.  Hard to understand why.
• Can provide situational awareness, does give some indication of development, but algorithm may need modifications to be more useful.
• Suggest either removing probs <30% or moving to transparent (to know algorithm is tracking these storms).  Large blue confetti-areas are distracting.
• Difficult to utilize in environments where noisy and chaotic. 
• Did not understand differences between algorithms (CI and SVR-CI):  why would CI be low and SVR-CI be high? 

All-Sky LAPS:

• Used in pre-convective environment, showed instability and moisture fields and gradients in those well.
• Could use in social media posts to explain the "WHY" behind the forecast.

NUCAPS:

• Data from JPSS integration was new (or relatively new) to most forecasters.
• Development of best-practices, easier menu-access (not buried in volume browser) and quick-guides would be helpful for increased use.
• Earlier access to data is highly important, but also need better visualization options.
• NSHARP options need to be more user-friendly, cumbersome to pull up multiple soundings in different CAVE displays to compare.  Would like to access shear parameters and get box&whisker plots.
• Difficult to evaluate trade-off of no-data available vs model integration into products.  Would like keep separate from LAPS options, but can be difficult to evaluate with gaps in data coverage in gridded products.

MK-Thursday Prob Severe Review &

Mesoscale 1 on RGB Split Water Vapor, toggle to Day Cloud Convection

The RGB Day Cloud Convection image has become my image of choice to enhance my multi-dimensional thinking of the storm environment due to the color bank and the resolution.
Our focus was Raleigh, but since my broadcast DMA combines 3 different WFOs, I looked at the line of storms developing over the western Piedmont/Foothills.

Prob Severe Wind Over RNK Velocity appears to be running hot with only one or two confirmations of a wind report with a tree down in Yadkin County and Henry county VA. There were several reports of pea to Quarter Size Hail with consistent quarter size hail reports as the storm progressed to the ENE out of the RNK area form Mount Airy into VA. 4 Panel evolution shows a lower percentage of hail possibility of 10 percent after it has already verified hail and then a higher wind prob of about 50% without verification at 19:12 Z.


 Prob Severe Wind Over RNK Velocity  with only one confirmation of a wind report with a tree down in Henry county..., several reports of pea to Quarter Size Hail from Mount Airy into VA. 4 Panel evolution shows a lower percentage of hail possibility of 10 percent after it has already verified hail and then a higher wind prob of about 50% without verification at 19:12 Z.

 This trend is clear as well with the pulse thunderstorm that drops golf ball size hail at 21:25z as seen by this series of 4 panels.

 Notice by adding GLM Total energy here this appears to be a better indicator of storm intensity if not type of severe storms. Typically in this region from a historical perspective using ENI data combined with reflectivity, this is an excellent predictor and indicator of severe hail along with storm heights, and watching developing 'hail' tracks as provided by WSI which is the shear tracks as we track the storms.  Granted, this is a messier presentation than many would like, but I find it interesting that this type of layer comparison with GLM and reflectivity/velocity may be a better indicator of storm intensification than the current ProbSevereProduct. At this time, I would rarely imagine using the ProbSevereAll Product alone and would likely never choose percentages. Hearst Television Meteorologists are using a scale from None, Low, Moderate, High to describe the type of severe impact. The more details we can convey will be helpful as we still hear our viewers in large numbers comment that they were unaware a tornado was possible with "those storms."

For a verification of the ProbWind increase (catch-up) I chose the velocity in a single panel. The Northern section of the storm pushing into Henry County does eventually verify with trained spotter reports of trees down, but most reports are quarter size hail. 

This RGB and ProbModel shows still after verification at the surface the hail model hasn't been responsive-only to 10-20% Prob.

Watching developing storms from GSP to RAH warning area with Severe Prob Hail overlaying the RGB Day Cloud Convection Product.
Severe Prob 4Panel at 22:24Z just before the golf ball size hail /wind reports with lower left panel indicating low level rotation in yellow, blue  hail tracks & GLM in right panel shows higher total Energy growing North. ProbWind Model is performing poorly still as a forecast tool as storm reports verify wind damage & possible microburst indicated.

 Below I used RGB images and a combination of All-Sky Lap like CAPE and Simple Water Vapor in the RGB allow for some gradient perspectives with layering shear tracks and GLM to continue monitoring storm growth and possible intensification that the GLM often seems to match.

 

MK