The Utilization of GLM during the January 11th Tornado Event

December and January have been very busy for tornadoes across the NWS Huntsville County Warning Area. Eight tornadoes, including two EF-2 tornadoes, impacted the area on December 16. One additional tornado occurred on December 29. Forecasts and outlooks leading up to January 11 suggested that it, too, would be another busy day for the region.

Flash Extent Density data from GLM were crucial while issuing warnings during the December 16 event. So one of the first things I did on January 11th was load a procedure that combined the mesoscale sector legacy IR band (10.3 µm), one-minute ENTLN cloud to ground and cloud flash lightning data, and the one-minute GLM flash extent density. This was done as a quick way to identify strengthening updrafts, as it is easy to see cloud tops cool and lightning activity increase using one-minute data. This proved to be especially useful on December 16th as lightning jumps were noted before most tornadoes occurred, and therefore became a key component of increasing lead times with warnings. On January 11th as the QLCS was starting to surge into west central Alabama, we noticed a strong lightning jump that preceded the deadly Pickens County tornado.  So, from that point we knew GLM was going to be useful once again in warning operations. 

Tornado warnings had been issued based on the Three Ingredients Method earlier as the line crossed Cullman County, as can be observed in the image loop below. However, it became increasingly challenging to determine whether to extend tornado warnings downstream due to data limitations:

  • few if any damage reports
  • limited radar data since KHTX was down at the time
  • KGWX and KBMX only capturing the mid-to-upper level portions of the storms
  • ARMOR data coming in sporadically


Image 1.  Upper Left: Loop of RAP13 km bulk shear vectors (kts) and KGWX 0.5 Refl (dBZ), Upper Right: KGWX Storm Rel Velocity (kts), Lower Left: GLM Flash Extent Density, Lower Right: SVR (yellow ploygons) and TOR (red polygons) issued by the HUN WFO.  Radar, GLM and warning polygon data between 1859-1939 UTC 11 Jan 2020, RAP13 0-3 km bulk shear vectors at 19 and 20 UTC.

KGWX and KBMX indicated rotation aloft over Cullman and Marshall counties, but we did not have a good feel for what was happening closer to the surface. A modest increase in lightning over Cullman County provided the confidence to issue a downstream tornado warning at 1907 UTC.  At 1910 UTC, we noticed a relatively sharp uptick in lightning activity on GLM, the sharpest we had seen in our area all day, which further increased our confidence. The strongest uptick was noted at 1919 UTC, which preceded the damage at Union Grove by approximately 5 minutes. This, combined with the line segment becoming oriented more perpendicular to the 0-3km bulk shear vectors (upper left box), created enough confidence to go ahead and issue another warning downstream into portions of Jackson and Madison counties.

The first warning including Union Grove was issued 17 minutes before an EF-2 tornado hit Brindlee Mountain Primary School (indicated by Home marker in image below), and the second was issued as the tornado was impacting the school. We also noticed that the modest increase in lightning over Cullman County occurred close in time to both the Holly Pond and Joppa, AL tornadoes.

Image 2.  Upper Left: 1900 UTC 11 Jan 2020 RAP13 km bulk shear vectors (kts) and 1924 UTC 11 Jan 2020 KGWX 0.5 Refl (dBZ), Upper Right: 1924 UTC 11 Jan 2020 KGWX Storm Rel Velocity (kts), Lower Left: 1924 UTC 11 Jan 2020 GLM Flash Extent Density, Lower Right: 1924 UTC 11 Jan 2020 SVR (yellow ploygons) and TOR (red polygons) issued by the HUN WFO.

GLM was critical to our warning decision-making process at a time when radar data were limited. Furthermore, it helped us communicate the urgency of the situation to Marshall County first responders as the Union Grove event unfolded.


— Ashley and Brian (WFO Huntsville)

Observing the First Major Thundersnow Outbreak of the 2019-2020 Winter Season

Written by Sebastian Harkema and Emily Berndt

The first major heavy-banded snowfall event of the 2019-2020 winter season occurred from Oct. 9-12 and produced over two feet of snowfall in North Dakota. Throughout the event, the NESDIS merged snowfall rate (mSFR; Meng et al. 2017) product tracked the heaviest snowfall rates, including bands with snowfall rates greater than 2 in/hr. With a temporal resolution of 10 minutes, this product can be used in real-time to forecast the location and evolution of snowbands producing heavy snowfall, and even anticipate cloud-seeding. SPoRT has collaborated closely with NESDIS to experimentally transition and assess the passive microwave and merged snowfall rate products with NWS forecast offices (Ralph et al. 2018).  Therefore, this product is available in AWIPS and forecasters can select different snow-to-liquid ratio values to best fit the situation.

Figure 1: NESDIS mSFR product and GOES-EAST ABI (Ch. 13) on October 9, 2019.

Figure 1 demonstrates the mSFR product overlapping GOES-East ABI (channel 13) for October 9th as the snowband traversed across Montana. While the mSFR product provides a unique way to monitor snowfall, the phenomenon known as thundersnow captivated the attention of some operational forecasters as well as the general public, in part by the availability of Geostationary Lightning Mapper (GLM) observations. Recent work from NASA SPoRT has shown that the overlap of GLM and mSFR data can be used to objectively identify and characterize electrified snowfall (i.e., thundersnow; Harkema et al. 2019a). In fact, Harkema et al. 2019a demonstrated that thundersnow flashes identified by GLM contain on average more total optical energy per flash area than other flashes in the GLM field-of-view. Harkema et al. 2019a also demonstrate that thundersnow flashes observed by GLM are spatially larger compared to non-thundersnow flashes and is likely a result of weaker mesoscale updrafts and slower charging rates compared to severe summertime convection.

Figure 2: NESDIS mSFR product, GOES-EAST ABI (Ch. 13), and GLM flash extent density observations on October 10, 2019.

Figure 2 demonstrates the objective identification of thundersnow based on the overlap of mSFR and GLM flash extent density observations on October 10th around the Colorado/Nebraska/Wyoming border region. From the loop, this region experiences an enhancement of snowfall rates approximately 30-40 minutes after the first occurrence of thundersnow. Even though it appears as though thundersnow can be used as a precursor for enhancement of snowfall rates in the near future, thundersnow has a spatial offset of 131±65 km from the heaviest snowfall rates (Harkema et al. 2019b, In Review). This spatial offset is evident when examining the thundersnow that occurred along the Minnesota/Manitoba border between 12-15 UTC on October 11th (Fig. 3).

Figure 3: NESDIS mSFR product, GOES-EAST ABI (Ch. 13), and GLM flash extent density observations on October 11, 2019.

The thundersnow observed by GLM occurs on the northern extent of the heaviest snowfall rates (purples/whites). The separation of thundersnow and the heaviest snowfall rates is likely caused by hydrometeor lofting of the snowfall as it descends to the surface because of the low terminal fall speed of the ice crystals.

Winter is fast approaching and the NESDIS mSFR product and GLM can be used in tangent with each other to improve situation awareness. NASA SPoRT is at the forefront of understanding the operational implications of electrified snowfall and continues to investigate the thermodynamic and microphysical properties that are associated with it. See the official JPSS Quick Guide and a past JPSS Science Seminar for more product information.


Harkema, S. S., C. J. Schultz, E. B. Berndt, and P. M. Bitzer, 2019a: Geostationary Lightning Mapper Flash Characteristics of Electrified Snowfall Events. Wea. Forecasting, 43(5), 1571–1585,

Harkema, S. S., E. B. Berndt, and C. J. Schultz, 2019b: Characterization of Snowfall Rates, Totals, and Snow-to-Liquid Ratios in Electrified Snowfall Events from a Geostationary Lightning Mapper Perspective. Wea. Forecasting. In Review.

Meng, H., Dong, J., Ferraro, R., Yan, B., Zhao, L., Kongoli, C., Wang, N.‐Y., and Zavodsky, B. ( 2017), A 1DVAR‐based snowfall rate retrieval algorithm for passive microwave radiometers, J. Geophys. Res. Atmos., 122, 6520– 6540, doi:10.1002/2016JD026325.

Reconstructing a Rare Bolt from the Blue Event Using Multiple Lightning Datasets

Reconstructing a Rare Bolt from the Blue Event Using Multiple Lightning Datasets

Written by Chris Schultz

On August 20, 2019, much of the Midwest was impacted by several rounds of severe thunderstorms.  These electrically active thunderstorms produced wind damage across Iowa, Illinois, Indiana, Ohio, Kentucky, and Missouri. However, it wasn’t the large flash rates that got the attention of those of us in SPoRT, but a rare bolt from the blue event that occurred nearly 50 miles (76 km) outside any surface precipitation.

During the 40 minutes leading up to the lightning event, the closest thunderstorm activity was located approximately 50 miles south of Dittmer, MO, across parts of Phelps, Dent, Washington, St. Francois, and Ste. Genevieve Counties (Fig. 1A).   Between 400 pm and 440 pm CDT zero lightning flashes occurred in Franklin, Jefferson, Warren, or St. Charles Co., MO (Fig. 1B).

Figure 1 – A- Radar reflectivity at 0.4 degrees elevation at 2140 UTC from KLSX in Weldon Spring MO, and  B- NLDN lightning detections between 21:00:00 and 21:40:16 UTC (4:00:00-4:40:00 pm CDT).

Then at 4:40:15 pm CDT, a positive lightning flash was observed by Vaisala’s National Lightning Detection Network well outside of any precipitation (Fig. 2).  This flash was positive polarity, was approximately 136 kiloamps, and located in an area that had not observed any lightning in the previous 40 minutes. This +CG flash was accompanied by 5 additional incloud flash detections, and one negative cloud to ground flash detection by the NLDN.  All 7 detections occurred within 1 second of each other, indicating that they were part of the same lightning event.  However, the question remained, where did this flash originate? Radar and previous lightning data from the NLDN indicate that there are 2-3 areas of thunderstorm activity to the south of this location which could be a possible origination point. But there wasn’t a definitive prospect because the NLDN point locations are spatially separated by several miles. 

Figure 2 – Radar reflectivity at 0.4 degrees elevation at 2140 UTC from KLSX in Weldon Spring MO (A) and NLDN lightning detections at 21:40:15 UTC (4:40:15 pm CDT).

Bringing in Geostationary Lightning Mapper Flash Extent Density data product for the same point in time (Fig. 3), there is a better idea of which thunderstorm this flash originated from.  There is a distinct lightning path from the thunderstorms over Dent and Phelps Counties in up to the NLDN flash locations in Jefferson and Franklin Counties. This single flash travelled nearly 57 miles (~ 92 km) from its original start location to the ground location, and actually propagated further north into Warren and St. Charles Counties.  

Figure 3 – GOES GLM Flash Extent Density overlaid on 0.64 µm ABI data at 2141 UTC (441 pm CDT).

Taking a vertical slice of the radar data between the parent thunderstorm and the location where the flash came to ground, there is a distinct path of precipitation aloft between 20,000 and 30,000 ft (Fig. 4).  Thus the lightning traveled through an anvil region before coming to ground approximately 41 miles (76 km) outside of the main precipitation near the surface.  Large bolt from the blue events have been reported in the literature previously (e.g., Kuhlman et al. 2009, Weiss et al. 2012, Lang et al. 2016). This flash was also a unique event because any lightning safety protocols would not have been in place for the location due to the absence of lightning within 6 miles during the previous 40 minutes.

Figure 4 – A vertical cross section of reflectivity from KLSX at 2140 UTC (440 pm CDT)

When GLM data are combined with ground based lightning networks like the NLDN or Earth Networks Total Lightning Network, the GLM Flash Extent Density can be used to connect point locations and determine where additional electrification may be present aloft that is not readily apparent at the surface.

Geostationary Lightning Mapper (GLM) Data Used to Aid in Warning Decision…

The NWS office in Huntsville, AL (HUN) has had a long history with the use of total lightning data in operations, which stretches back to the office’s inception (after NWS modernization) in 2003.  Back then, and until its removal to South America for GOES validation testing, the HUN office largely used data from the North Alabama Lightning Mapping Array (NALMA). Lightning data sources from NLDN and ENTLN have also been used to varying degrees, but the advent of the GLM aboard GOES-16 brought a new era of lightning observations.  Because of the office’s participation in early operational testing of the GLM, its use and familiarity have gradually increased over the past year.  This was probably made easier due to our familiarity with total lightning data from the NALMA network.  Generally, GLM data have been used in much the same way as those from the NALMA network, especially with regards to situational awareness purposes (i.e., airport weather warnings, real-time weather watches for EM partners, initial cell electrification, etc.).  The use of the data to aid in severe weather warning decisions has been a bit slower to evolve, as might have been expected.  After all, there are differences in the way the NALMA observes lightning as compared to the GLM.  Values from the GLM have typically been “muted” compared to those from NALMA, so forecasters have had to make internal adjustments and recalibrate, if you will, what is considered significant.  However, the physical mechanisms that generate increases in total lightning, that is, increases in mixed-phase updraft volume, are essentially observed either way.  Thus, GLM data can still be useful to relate important information about storm/cell evolution, and can help to “tip the scales” in the balance of evidence about whether or not a warning may be needed.

This particular application of the GLM data occurred this morning with operational meteorologists at the HUN office.  The short image loop below shows thunderstorms moving across northwestern portions of Alabama between 1226 and 1300 UTC.  The top panel of the image contains data from the KGWX radar (0.5 degree reflectivity), while the bottom panel contains GLM 1-minute Flash Extent Density (FED) data.  Notice that lightning activity is relatively limited initially as the storm moves across western Franklin County, AL (near center of image), with 1-min FED values ranging between 2 and 6 flashes per minute.  Then, at the 12:37 UTC time mark, flashes begin an increase that manifests in a statistical lightning “jump” (GLM sigma > 2).  The warning meteorologist at the time was watching this cell for potential severe weather, and observed the sudden increase in FED values.  This, together with other radar and satellite observations (not shown here), suggested that a severe weather warning was necessary as wind signatures aloft gradually increased.  A warning was subsequently issued at 1245 UTC.  Incidentally, this thunderstorm did end up producing some wind damage, with trees reported down in south-central portions of Franklin County.  Notice also that a number of strong cell signals were detected by radar as indicated by the higher dBZ values across the domain.  Another use of the GLM is allowing meteorologists to focus on the cells with the strongest updrafts, making the overall radar interrogation and warning process more efficient.  This case can help to demonstrate that the GLM can be used as an important indicator of storm evolution and as a useful operational tool for the evaluation of severe weather potential.

-Kris W.


[Top] KGWX 0.5 deg Refl with NWS Severe Thunderstorm Warning (yellow box), [Bottom] GLM 1-min Flash Extent Density, 1226-1300 UTC 17 July 2019.


GLM Aids Convective Situational Awareness–at 37 degrees

There has been a great deal of focus on the potential for heavy rain and flooding across the southeastern United States.  However, today has been marked with an interesting mix of winter and spring.  The day began with reports of sleet, snow, and rain as the first wave of precipitation spread across the region.  Temperatures rose into the upper 30s to lower 40s later in the morning, and the focus shifted–to thunderstorms.

Multi-Radar/Multi-Sensor Reflectivity, valid 1846 UTC and GLM Flash Extent Density , valid 1847 UTC 19 February 2019

Flash Extent Density (FED) data from the Geostationary Lightning Mapper has been lighting up (literally and figuratively) over the Mid-South and Mississippi Valley as precipitation lifts northward.  While many of the flashes have been focused within the convective elements along the southern edge of the precipitation shield, there have been numerous “long” flashes (greater than ~30 km) advancing north well beyond the convective cells (such as the one indicated above at 1847 UTC).

Use of the GLM aided forecasters in issuing an airport weather warning for the Northwest Alabama Regional Airport (Muscle Shoals) at 11:52 AM.  While this was not much earlier than the first report of thunder at KMSL, it was much farther north of the convective cells than originally anticipated.  GLM aided the forecasters’ situational awareness of an unusual situation.

Multi-Radar/Multi-Sensor Reflectivity, valid 1750 UTC and GLM Flash Extent Density, valid 1751 UTC 19 February 2019. White circles denote the KMSL (Muscle Shoals) and KHSV (Huntsville) airports.

GLM data have also been used to assess convective potential for sub-severe thunderstorms as the more convective cells have moved into the Huntsville county warning area.  A persistent GLM centroid of 4-8 flashes per minute corresponded to a report of dime-size hail in Cullman, Alabama around 1945 UTC.

Multi-Radar/Multi-Sensor Reflectivity, valid 1924-1954 UTC and GLM Flash Extent Density, valid 1925-1954 UTC 19 February 2019

GLM “sees” apparent meteor flash in Western Cuba…

So, I was seeing some news reports on Twitter this afternoon about an apparent meteor that struck Western Cuba.  Pulling up data/imagery from the GLM in AWIPS, I was able to see some relatively high Flash Extent Density (FED) values from that area at the same time of the meteor report.  The first image below shows FED values (1818 UTC) overlaying GOES-16 Visible (0.64 µm) imagery at 1817 UTC.


Image 1. GLM data shows an apparent meteor flash over western portions of Cuba at ~1818 UTC 1 Feb 2019. The GLM Flash Extent Density overlay GOES-16 visible (0.64 um) imagery from ~1817 UTC.

Also, notice the large amount of lightning observed by the GLM in central portions of the Gulf of Mexico.  Here’s a short 30-min image loop around this time period (the suspected meteor flash shows up about midway through the loop).  Importantly, before the GLM sensor, the amount and extent of lightning activity over open ocean areas, away from ground networks, was generally not known, especially at such high spacial/temporal resolution.


Image 2. GLM (Flash Extent Density) and GOES-16 visible imagery (0.64 µm) loop from 1802-1830 UTC, 01 Feb 2019. An apparent meteor shows up in western Cuba at 1818 UTC in the loop. Also, notice the active deep convection and lightning over the Gulf of Mexico during the period.



GLM & Public Safety: An Important Caveat

As great as is to use data from the Geostationary Lightning Mapper, there is an important caveat forecasters have to consider when using the data.

During the afternoon of July 11, typical “air mass” showers and thunderstorms were developing across northern Alabama, including several south of the Huntsville International Airport.  At 12:18 PM CDT, the GLM Flash Extent Density data started to light up with these cells, including one larger flash at 12:21 PM.  (Huntsville airport is marked by the eastern concentric yellow circles.)

KHTX radar valid 1722 UTC 11 July 2018 and GLM FED valid 1721 UTC

As I’ve noted before, we issue airport weather warnings for Huntsville if lightning is within 5-10 miles, heading towards the terminal.  So forecasters were justifiably alarmed that GLM flashes were starting to show up within the 10-mile range ring, and just barely edging towards the 5-mile ring.

But here is where that caveat comes into play: the parallax effect.  Radar showed the actual echoes associated with these flashes to be well to the south of the GLM flashes.  Earth Networks Total Lightning data from the same time period showed lightning confined to these cells well outside the 10-mile range ring.  Furthermore, the cells were moving away from the field.

KHTX radar valid 1722 UTC 11 July 2018 and Earth Networks total lightning valid 1721 UTC

KHTX radar valid 1722 UTC 11 July 2018 and Earth Networks total lightning valid 1721 UTC

It’s worth noting that GLM showed a greater spatial extent during some of these flashes, but Earth Networks was much closer in location to the radar.

So while a cursory glance at the GLM data might lead to an airport weather warning, it was important to double-check GLM against the radar–and recognize that an AWW was not necessary in this case.  The same caution will need to apply as we begin applying GLM to other public safety situations.

GLM Detection of Anvil Lightning

During the evening hours of June 27th, we noticed that GLM average flash area data was indicating a substantial flash from the main core of storms over northeast AL/northwest GA towards middle TN.  This occurred around June 28th, 0130Z-0140Z, and I have included both a screen shot and animation for this flash.  Though no DSS events were scheduled over this particular area, it would be of great benefit to utilize this for potential lightning impacts at outdoor events–where lightning is occurring well away from the storm core/updraft. Also of note was the GLM flash extent density showing a noticeable jump throughout the evening with the stronger updrafts. This aided in the decision to keep a significant weather advisory in place for these storms.

For purposes of calibration and given my experience with utilizing GLM data at a recent HWT (May), I prefer having a four panel display in D2D. ENTLN data overlaid on Flash Extent Density while the other available parameters (Average Flash Area and Total Optical Energy) are on two other panels.

GLM_062818_0130Z_LightningFlashAnvil-AnnotatedFigure 1: Four panel of GLM data with MRMS Composite reflectivity in bottom left, annotated Average Flash area in bottom right, flash extent density with overlaid ENTLN data in top left, and Total Optical Energy in top right.


GLM_062818_0130Z_LightningFlashAnvil_AnimationFigure 2: Same as in Figure 1, just animated.

The nice thing is that there were generally no latency issues noted other than intermittent 2-3 minute delays to real time.

-Stephen Latimer (NWS Huntsville)


Using GLM for Airport Weather Warnings

As part of our impact weather responsibilities, NWS Huntsville issues Airport Weather Warnings (AWWs) for Huntsville International Airport (KHSV) and Northwest Alabama Regional Airport (Muscle Shoals; KMSL).  AWWs are issued for the threat to personnel working outside at the terminal and neighboring operations, not for the threat to aviation as a whole.  One of our criteria for AWW issuance is the threat of cloud-to-ground lightning within 5 miles.

On May 30, a relatively small line of thunderstorms developed over northeast Mississippi and tracked to the east, producing a few pockets of straight-line wind damage along the way.  The line approached the 10-nm range of KHSV at 2320 UTC (seen below).  At this point, GLM Flash Extent Density (FED) data wasn’t the most helpful data set for deciding to issue an AWW; the National Lightning Detection Network (NLDN) was already detecting a great deal of cloud-to-ground lightning with the storms.


Multi-Radar/Multi-Sensor (MRMS) Reflectivity At Lowest Altitude (RALA) valid 2320 UTC 30 May; 5-minute NLDN flashes valid 2321 UTC; and GLM Flash Extent Density valid 2321 UTC.  The 5 and 10-nautical mile range rings around KHSV are illustrated in coral near the center of the image.

The GLM FED data were far more useful later on in the event, after the line passed well east of KHSV and the Huntsville metro area.  With nightfall, the convective updrafts weakened and lightning along the line generally decreased–but flashes within the trailing stratiform region of the quasi-linear convective system (QLCS) did not.  One of the more spectacular examples occurred around 0059 UTC (seen below).  The FED product really illustrated the spatial threat simply and effectively, especially when combined with the NLDN data.  (This example uses a slightly different color curve from the GLM baseline to enhance lower flash rates.)


MRMS RALA valid 0058 UTC 31 May; 5-minute NLDN valid 0059 UTC; and GLM FED valid 0059 UTC.  5 and 10-nm range rings around KHSV are also noted.

GLM helped forecasters acquire and retain situational awareness of these trailing stratiform “long flashes”, which helped with AWW extension/reissuance.  As a result, the airport weather warning for Huntsville was re-issued until the trailing stratiform region cleared the airport and the threat subsided.

Geostationary Lightning Mapper detects lightning in the Volcanic Plume from the Fuego Volcano in Guatemala (3 June 2018)

During the eruption of the Fuego Volcano on the afternoon of Sunday June 3, 2018, the Geostationary Lightning Mapper observed a rare, but important, phenomenon: volcanic lightning.  A total of five lightning flashes were observed between 1814 and 1834 UTC with the rising plume.  The first three flashes between 1814 and 1818 UTC were 8 to 10 km north-northeast of the volcano. Then as the plume continued to advect to the northeast of the volcano, the position of the lightning followed. A flash at 1822 UTC was 13 km from the cone of the volcano, and the flash at 1833 UTC was farthest away from the cone of the volcano at 15 km.

Animation of GLM group densities from 1810 to 1835 UTC on 3 June 2018. The location of the volcano is circled.

One of the unique features of GLM is the ability to measure the size of the flash directly as each flash is observed.  In thunderstorms, flash size is a good indicator of vertical motion, which is often hard to directly measure within thunderstorms, or in this case, a volcanic plume.  Between 1814 and 1822 UTC, flashes were occurring approximately every 2 to 4 minutes, and their size was on the order of 500-1000 km2.  Then between 1822 and 1834 UTC there was a lull in GLM-detected lightning activity, followed by the largest flash detected by GLM during the event at 1500 km2 as the plume continued to expand over central Guatemala.

There was a 3 hour lull in GLM-detected activity, until 2141 UTC, when a second set of 21 flashes was detected by GLM between 2141 and 2203 UTC. These flashes were located between 1 and 8 km south-southeast from the highest point of the volcano and correspond in time and reported location of the deadly lahar and pyroclastic flow that came down the south side of the volcano. GLM flash sizes ranged from 64 km2 up to 1500 km2 and there was not as clear of an increase in size as observed with the volcanic plume.

Animation of GLM group densities from 2139 to 2205 UTC on 3 June 2018. The location of the volcano is circled.