Tuesday, September 4, 2018

If a tornado blows through the Maine woods but no one’s there to see it, did it really happen?

The title of this post is exactly what the Bangor Daily News asked after a supercell snapped trees in the Maine northwoods on August 29th. The average radar beam height of the lowest tilt was over 2 km for much of this storm's lifetime, but the storm exhibited a very strong increase in rotation in the low-levels and mid-levels at around 19:54Z, a few minutes before the first tornado warning was issued. The storm also had markedly weak lightning activity from the ENTLN.

The NWS in Caribou and the Maine Forest Service did a fly over and looked for ground damage but apparently found none. So in this case, it probably didn't happen, but at least any lumberjacks in the area were warned for this potent storm.

Figure 1: ProbTornado contours, MRMS MergedComposite reflectivity, and NWS severe weather warnings. 
Figure 2: Time series of the probability of tornado and constituent predictors for the tornado-warned storm in Figure 1.



Thursday, June 14, 2018

Wilkes-Barre tornado

A pronounced shortwave traversed the eastern Great Lakes with a trailing cold front spawning severe storms in New York and Pennsylvania. The Storm Prediction Center forecasted a 2% outlook for tornadoes within 25 miles of a given point (Figure 1).

Figure 1: SPC tornado outlook from 06/13/2018 2000Z. 

Wednesday evening, an embedded supercell emerged from a linear storm segment over north-central PA. Strong low-level rotation, adequate effective bulk shear and meanwind in the 1-3km layer (both 35-40 kts), as well as very high 0-1km storm-relative helicity (> 200 J/kg) produced ProbTor model output over 40% when the storm was first tornado-warned by the NWS. As 0-2km MRMS AzShear decreased, so did the probability of tornado. Then, from 0152Z to 0202Z, the ProbTor value jumped from 20% to over 80% as both low-level and mid-level rotation increased in this storm. A wind report (but likely tornado damage) was received from Wilkes-Barre at 0215Z, with multiple injuries and cars flipped over. See Figure 2 and Figure 3 for a depiction of the evolution of this storm.

Figure 2: ProbTor contours, NWS warnings, and MRMS MergedReflectivity from 0100Z to 0230Z.
Figure 3: Time series of ProbTor and constituent predictors. NWS warnings and preliminary LSRs are plotted as well.

An accumulation of the 0-2km MRMS AzShear nicely shows the cycling nature of the strong low-level rotation in the storm, with white pixels exceeding 0.015 s^-1 over Nordmont and Pennsylvania state lands, and then over the city of Wilkes-Barre (Figure 4).

Figure 4: MRMS low-level rotation track over the Wilkes-Barre, PA region for the evening of June 13, 2018.

The GLM and ABI instruments from GOES-16 also captured the evolution of the storm (Figures 5 and 6).

Figure 5 shows a 4-panel of new GLM products (produced via Eric Bruning's GOES-R Geostationary Lightning Mapper Tools), along with MRMS MergedReflectivity. Here, the FlashExtentDensity is the count of flashes in 10-km boxes over 3 min, updated every min; the TotalEnergy is the accumulated energy of all GLM flashes in each box over 3 min; and the FlashAvgArea is the average area per flash over 3 min in each 10-km box. We see increases in the FlashExtentDensity and the TotalEnergy fields from 01:15Z to 01:20Z, 01:40Z to 02:00Z, and then a smaller increase from about 02:10Z to 02:20Z. The second jump in total lightning activity corresponds well with increased rotation in the storm and increased probability of tornado.




In Figure 6, note the cooling cloud tops as the storm enters Luzerne county. The cooling starting at about 01:40Z corroborates the increased lightning activity from GLM. Evolution of products from infrared and optical imagers (i.e., ABI and GLM) with high temporal resolution data is an active area of research for severe storm nowcasting at CIMSS.

Figure 6: 10.35µm channel from GOES-16 mesoscale sector for 0130Z to 0230Z. Note the cooling cloud-tops prior to and during tornado occurrence.
EDIT:

The National Weather Service in Binghamton, NY has confirmed an EF2 tornado, beginning approximately at 10:00pm EDT.

Location...Wilkes-Barre Township in Luzerne County Pennsylvania
Date...June 13 2018
Estimated Time...1000 pm EDT
Maximum EF-Scale Rating...EF2
Estimated Maximum Wind Speed...130 mph
Maximum Path Width...200 yards
Path Length...0.75 mile
Beginning Lat/Lon...41.2436/-75.8467
Ending Lat/Lon...41.2390/-75.8392
* Fatalities...0
* Injuries...6

Thursday, May 24, 2018

GLM parallax issues?

We've been watching t-storms across central/southern SD and northeast ND today. A few storms were severe with reports of hail 1" or greater. I was kind of
expecting to see some GLM FED lightning jumps with these storms as they became severe, but that just wasn't the case.

Could this be a parallax issue?

This far N (roughly a similar latitude to Toronto, only farther W, so that there are even MORE potential parallax problems), the GLM is likely not sampling storms very well (at more of a side angle), so perhaps it can't actually "see" all of the lightning occurring within a storm.

If that is the case, watching for lightning jumps within a storm might not be a good "warning determination method" for a forecaster across the Northern High Plains and Northern Mississippi Valley...

Or...

Maybe we can still see lightning jumps in more northern/western storms, but due to parallax, we should expect more storms to have lower lightning values, hence lower lightning jump thresholds?

Other parallax-related questions I have include:

How does parallax affect other GLM products like the Total Energy and Avg Flash/Group Areas?

If the GLM is sampling more of the side of a storm across the northern US, can we expect storms to look brighter or darker?

Should we expect the areal extent of lightning flashes within a storm appear to increase or decrease as parallax increases?

How parallax affects GLM data is a mystery to me at this point and certainly another area of research that GLM developers should explore.

- Thomas Bell

DLH - GFS CAPE forecast degrading all-sky CAPE

During my mesoscale discussion I blogged about how the all-sky CAPE did not seem to match up well at all compared to SPC mesoanalysis. I compared the all-sky CAPE to the clear-sky only CAPE & noticed that the most questionable areas were blacked out on the clear-sky product, & thus filled in with GFS data. Looking at the 12Z GFS CAPE data it was obviously too high with 2500 J/kg SBCAPE over NW MN & over 3000 J/kg over SE MN. It took me a while to identify that the GFS CAPE was too high & thus the all-sky LAP CAPE would also be too high because I really do not use the GFS much when doing a short-term mesoscale analysis. I think it would be much more useful to fill in the missing clear-sky data with RAP/HRRR/or some other hourly updating field vs the every 6-hour GFS.

22Z LAP all-sky CAPE:

22Z clear-sky CAPE:
Peter Sunday

Highest Flash Event Densities Collocated With Updraft Location

[23:00 UTC] Have tried to overlay semi-transparent GLM data over GOES-East meso-sector visible imagery and the results are pretty useful. For one, the highest flash event densities seem to be collocated well with perceived updraft location (overshooting tops are being used as a proxy), which fits the conceptual model of what one would expect. It should be noted that the values themselves were not quite as useful as seeing the change in the values and the "peaks" in values which were highlighted a bit more by altering the color scale. This certainly is useful in an operational setting in terms of isolating the most important (rapidly developing) updrafts in a quick and timely fashion.

Fig. 1: GLM FED data overlaid on GOES-East meso-sector Ch. 2 data

Rosie Red

CI With Radar



Couldn't tell if there was something weird with the CI probs...it looked like the storms had already developed.  Unless it picked up that more towering CU was developing right next to the new storms.  Which in this case it seems possible as the storms were building off each other and morphing into a line.



-Penny Gardens

*Later note. I tried using this to overlap outflow boundaries with the CI or CI Severe to see if any CU developed along the boundary.  Didn't see anything, but that may be because of the cirrus.
And is there a probability tool for a gust front? I'm always reminded of the Indiana State Fair situation.  I kept seeing gust fronts ahead of our storms today...however the peak wind seemed to be only about 30 mph.  

GLM Average Flash Area - Updraft vs. Non-updraft/Anvil

[22:00 UTC] After an idea was shared with me from a colleague of mine of increasing the transparency of GLM data, I decided to explore its usability overlaid on visible satellite.

More specifically, I took a look at average flash area to see if there was any notable difference between the perceived updraft and anvil areas of mature convection moving through the Duluth CWA.

While data still remains a bit jumpy, I did notice that it seems that the average flash area near the overshooting tops (used here as a proxy for approximate updraft location) seems to be lower/smaller than in surrounding anvil areas or in areas where updrafts do not appear quite as strong (or have become weaker). The average flash area appears to increase as the convection becomes "older" (i.e. new/strengthening updrafts not apparently seen on visible imagery). Fig. 1 shows an animation through time while Fig. 2 shows a still frame at 21:40 UTC.

Fig. 1: animation of 5-min/1-min GLM Average Flash Area overlaid on GOES-East Ch. 2

Fig. 2: still of 5-min/1-min GLM Average Flash Area overlaid on GOES-East Ch. 2 at 21:40 UTC


Rosie Red

DLH - Nifty Convection Monitoring Procedure

2150Z: I decided to try looking at smoothed GLM data overlaid on visible satellite by using the "interpolate image" & "interpolate colors" options along with some transparency in AWIPS. On the same image I also overlaid ProbSevere polygons & GOES CI probabilities. It was much easier for me to interpret GLM trends while having some of the texture from the visible satellite imagery. The identification of new updraft growth in the GLM flash extent density data was much more evident with this procedure vs just using the GLM data alone. ProbSevere polygons were useful in monitoring the intensity trends of the storm at a glance. Finally having the GOES CI probability helped in a zoomed-out view to identify potential area of new convection. I'm impressed at the amount of data I'm able to view in this procedure without the whole thing looking too cluttered.

Loop of above procedure showing storm evolution in SW MN:
Peter Sunday

T-storm Warning Decision Process

The storms are heading into a more prime environment based on what I've been looking at (see earlier blogs).  

The ProbSevere increased with each scan leading up to the warning.  The lightning slighly increased.  There were so many cells that it was a tad confusing which one was strongest. The flash centriod density helped a bit...it's a nice small scale with the pixels that I overlapped with the storm.
-Penny Gardens


*Side note the t-storm warnings verified with storm reports from NWS.

000
NWUS53 KFSD 242246
LSRFSD

PRELIMINARY LOCAL STORM REPORT
NATIONAL WEATHER SERVICE SIOUX FALLS SD
546 PM CDT THU MAY 24 2018

..TIME...   ...EVENT...      ...CITY LOCATION...     ...LAT.LON...
..DATE...   ....MAG....      ..COUNTY LOCATION..ST.. ...SOURCE....
            ..REMARKS..

0522 PM     HAIL             3 W SCOTLAND            43.15N 97.78W
05/24/2018  E1.00 INCH       BON HOMME          SD   PUBLIC 
 
 000
NWUS53 KFSD 242216
LSRFSD

PRELIMINARY LOCAL STORM REPORT
NATIONAL WEATHER SERVICE SIOUX FALLS SD
516 PM CDT THU MAY 24 2018

..TIME...   ...EVENT...      ...CITY LOCATION...     ...LAT.LON...
..DATE...   ....MAG....      ..COUNTY LOCATION..ST.. ...SOURCE....
            ..REMARKS..

0502 PM     HAIL             6 S DANTE               42.95N 98.19W
05/24/2018  E1.25 INCH       CHARLES MIX        SD   PUBLIC           

            HAIL COVERED THE GROUND. 

Parallax Issues: Beware in the Northern Plains and Northern Mississippi Valley regions!

Just looked at a cluster of showers/storms in southwest MN and noticed that the IR satellite image and GLM data are offset (to the north) from MRMS "reflectivity at lowest altitude" radar data by as much as a county!

This parallax issue is something that WFOs and CWSUs across the Northern Plains and Northern Mississippi Valley regions will certainly need to be aware of when they use GOES-16 data, as it may lead them to issue warnings/advisories for the wrong storms!


- Thomas Bell