Monitoring Hurricane Harvey with the Geostationary Lightning Mapper (GLM)

Figure 1, below, shows a single image from the attached movie showing the Geostationary Lightning Mapper observations for Hurricane Harvey.

cover_image

Figure 1:  Example image from the attached movie from 0700 UTC on August 25, 2017 while Harvey is a Category 2 hurricane.  The image shows the 11.2 micron infrared imagery in grey scale (background) and the GLM group density (shaded) accumulated over 15 minutes.

The link to the animation below monitors Harvey from 1400 UTC on August 23, 2017 while it was still a remnant system coming off the Yucatan Peninsula through the initial landfall and heavy precipitation across Texas at 2345 UTC on August 27, 2017.  The animation shows the ABI 11.2 micron infrared imagery (using a greyscale color curve to emphasize GLM) as well as the GLM 8 km group density (using the SPoRT color curve tested at the Hazardous Weather Testbed and being prepared as the default, operational curve).  Given the length of time covered by the animation, the data are shown at 15 minute intervals.  Specifics on Harvey (i.e., maximum winds and minimum pressure) are from the National Hurricane Center’s product archive for this storm.  The link is for an mp4 movie and is approximately 62 MB in size.

[62 MB]  Hurricane Harvey mp4 link.

Several still images are shown below highlighting interesting features.

One feature is the distribution of the total lightning observations throughout the tropical cyclone and the magnitude of the lightning density.  Generally, the total lightning is not distributed throughout the entire storm, but concentrated in bands and sometimes in the eye wall, as seen in Figure 1.  Figure 1 can be compared to the early stages when Harvey was upgraded to a Tropical Storm (Figure 2), but also later where many times there is no lightning in the eye wall (Figure 3). Also, please note that the accumulations are for 15 minutes versus 1 or 2 minutes shown in severe weather cases.

figure2

Figure 2:  This is the same as Figure 1, but for Tropical Storm Harvey at 0415 UTC on August 24, 2017.  

figure2

Figure 3:  This is the same as Figure 1, but at 0815 UTC on August 25, 2017.  This is highlighting the distribution of total lightning is mainly in the convective bands and note in the eye wall.

Figure 4 shows Harvey as it makes landfall as a Category 4 hurricane.  Here, GLM group density values are on par with the outer convective bands as the eye wall makes landfall.

figure3

Figure 4:  Same as Figure 1, but at 0215 UTC on August 26, 2017.  This image highlights the GLM group density observations in Hurricane Harvey’s eye wall as it makes landfall.

Lastly, after the initial landfall, Figure 5 shows a large increase in the magnitude and spatial area of the GLM group densities in the outer convective band as some of the catastrophic rain impacts the Houston, Texas region.

figure4

Figure 5:  Same as Figure 1, but at 0245 UTC on August 27, 2017.  This image highlights the increased GLM group density magnitude and spatial extent during part of the catastrophic rains that impacted Houston, Texas.

NOTE:  NOAA’s GOES-16 satellite has not been declared operational and its data are preliminary and undergoing testing. Users receiving these data through any dissemination means  (including, but not limited to, PDA and GRB) assume all risk related to their use of GOES-16 data and NOAA disclaims any and all warranties, whether express or implied, including (without limitation) any implied warranties of merchantability or fitness for a particular purpose.

GLM is here! First beta-release imagery

The Geostationary Lightning Mapper (GLM) has completed a product validation review and has been cleared for distribution through the GOES-R Re-broadcast system.  The GLM data are currently in a “beta-status”.  This means that additional updates will occur with the data processing before GOES-R (now GOES-16) moves to the east position in November.  However, this is a great opportunity to get an initial look at the GLM data in real-time.  The two examples below show the first data to be received at 1454 UTC today (5 July 2017) over the eastern United States and for the GLM field of view.  The data have been manually ingested into the National Weather Service’s AWIPS display for demonstration purposes.  Stay tuned for more examples as convection becomes more active!

glm_eastConus_05jul17_1454

Figure 1:  The first GLM-beta status observations (in this example 1 minute GLM group density) from the GOES rebroadcast zoomed in over the eastern United States at 1454 UTC on 5 July 2017.

glm_fov_05jul17_1454

Figure 2:  This is from the same time as Figure 1, but is now showing the 1-minute beta-GLM group density over the entire field of view.

NOTE:  NOAA’s GOES-16 satellite has not been declared operational and its data are preliminary and undergoing testing. Users receiving these data through any dissemination means  (including, but not limited to, PDA and GRB) assume all risk related to their use of GOES-16 data and NOAA disclaims any and all warranties, whether express or implied, including (without limitation) any implied warranties of merchantability or fitness for a particular purpose. 

GLM is coming: The instrument

The first beta-release data of the Geostationary Lightning Mapper (GLM) instrument will be out this week. (Update as of 12 June 2017:  GLM beta release has been delayed until July.)  As we get closer to having real-time GLM observations, here is a quick post about the GLM instrument itself.

glm_instrument

Figure 1:  An artist’s image of the GOES-16 satellite with the Geostationary Lightning Mapper (GLM) shown as the zoom out in the upper right.

In the post describing the origin of the GLM (here), it was discussed how the GLM is not the first space-based instrument to observe lightning.  However, it is the first lightning sensor available in geostationary orbit.  Conceptually, the GLM can be thought of as a very large digital camera.  Each pixel of the camera is identifying optical brightness difference from cloud top.  Each pixel is monitoring if any light is observed and if the light observed exceeds a background threshold.  This check is occurring every 2 ms and these observations become the basic GLM “event” observations.  The background field and threshold criteria are designed to reduce false alarms.  The placement of the charge couple device, or CCD pixels, on the instrument designed to help with the instrument’s spatial resolution.  The instrument’s CCD pixels vary in size to help account for the increasing parallax the closer to the edge of the field of view the observations get.  This allows the resolution of the GLM to go from 8 km directly beneath the satellite to only 14 km at the edge of the field of view.

The actual field of view for GLM is shown in Figure 2 for both the GOES-East (eventual location of GOES-16) and -West (future position of GOES-17) positions.  The underlying, normalized annual lightning flash rate comes from observations made by the Optical Transient Detector and Lightning Imaging Sensor from 1995-2005.  Currently, the GLM is in the GOES-16 check-out location (Figure 3).  The total field of view will range from 52 degrees north and south.  Additionally, the GLM does observe total lightning, or the combination of intra-cloud and cloud-to-ground observations.  However, the GLM will not distinguish between the two.  Still, observing total lightning, particularly over such a large domain will aid in warning decision support, lightning safety, as well as situational awareness in data sparse regions.  This will be helpful for detecting flash flooding (noting where is convection) in the inter-mountain west, convection monitoring for aviation, as well as opening up new avenues of research for tropical cyclone forecasting.  Lastly, the GLM was designed to be able to detect 70% of total flashes over the entire field of view over 24 hours.  The false alarm rate was designed to be less than 5%.  Recently, a calibration and validation field campaign had been underway to investigate the GOES-16 instruments.  Early indications are that the GLM will likely exceed the design specifications.  Exact values will be provided later after the field data has been analyzed.

glm_fov_E_and_W

Figure 2:  The field of view for GLM in the GOES-East and -West position.  The normalized, annual lightning flash rate shown is derived from 10 years of Optical Transient Detector and Lightning Imaging Sensor, low-Earth orbiting instrument observations.

goes16_central

Figure 3:  Same as Figure 2, but showing the current GLM field of view through November 2017.

Subsequent posts will start to focus on actual GLM observations once they are made available.

GLM is coming: Preparations

Creating demonstration data and products to train forecasters for GLM has presented unique challenges.  Aside from the Optical Transient Detector (OTD) and the Lightning Imaging Sensor (LIS), no other space-based platform had similar capabilities to the GLM.  Furthermore, the OTD and LIS were low-Earth orbiting instruments and would only view a small portion of the Earth for no more than a couple minutes at a time.  This prevented their use as a demonstration data set as forecasters would need to see how total lightning (the combination of intra-cloud and cloud-to-ground) evolved with time.  That put the focus on ground networks, which could observe the entire life cycle of a storm.  The ground networks, unfortunately, lacked the ability to observe total lightning (or the capability was not yet available).  The exception was the ground-based lightning mapping array.

The lightning mapping array (LMA) was developed by New Mexico Tech and evolved out of earlier systems, some of which were tested at Kennedy Space Center.  By operating in the very high frequency end of the electromagnetic spectrum (~80 GHz), the LMAs could observe the entire lightning channel within a cloud.  Primarily designed for lightning research, these would become instrumental in the training activities for the GOES-R Proving Ground and the GLM.  This is because the LMAs were capable of observing total lightning.  Their accuracy was extremely good and they have been used for ground verification for OTD and LIS and will do so again for GLM.  Their primary disadvantage is a very short range; generally no more than 200 km from the center of the network.  Figure 1 shows the physical relationship of total lightning to a storm updraft as well as the lightning jump concept.

total_lightning_physical_reasoning

Figure 1:  The top two panels show the total lightning density (left) and radar reflectivity at ~20 thousand feet (right) and 1442 UTC.  The radar elevation corresponds to the mixed phase region.  Total lightning is produced by a strong, voluminous updraft extending into the mixed phase region.  This is a non-linear relationship so the strongest updrafts will produce the most lightning.  This physical connection can be harnessed by forecasters as shown in the lower left (total lightning density at 1450 UTC) and right (radar reflectivity at 20 thousand feet at 1452 UTC).  The total lightning shows a “bull’s eye” feature indicating rapid intensification, or lightning jump.  This preceded the radar update at 1452 UTC showing the updraft now extending into the mixed phase region.  This allowed the total lightning to provide additional information on the intensification of this storm and the rapid increase indicates that severe weather is imminent.  Also, the total lightning information shows the spatial extent of the lightning that can be used for safety applications.

The Melbourne, Florida forecast office was the first office to use total lightning data from a local lightning detection and ranging network (very similar to an LMA) in the late 1990s.  This was a combined effort by the forecast office, Massachusetts Institute of Technology (MIT), MIT Lincoln Lab, and NASA Marshall Space Flight Center.  The data became extremely popular with the office and the Lightning Imaging Sensor Demonstration and Display (LISDAD) system was instrumental in investigating the uses of total lightning in real-time.

In 2002, NASA’s Marshall Space Flight Center had installed the research oriented North Alabama Lightning Mapping Array (NALMA).  By March 2003, the NASA SPoRT team, in collaboration with the Huntsville weather forecast office, had made NALMA data available in the National Weather Service display system; AWIPS.  The Huntsville forecast office would then go on to issue its first warning using total lightning data that May.  NASA SPoRT would extend collaborations with a handful of other forecast offices using NALMA as well as the NASA owned Washington D.C. LMA in the late 2000s.

By 2008, the GOES-R Proving Ground was accelerating its efforts with training and hands-on activities, such as the Hazardous Weather Testbed in Norman, Oklahoma.  This required a demonstration product for GLM that could be used in real-time.  The Marshall Space Flight Center had developed the GLM proxy that was derived from NALMA to test data processing algorithms for the GLM.  The drawback was that it could not be run in real-time.  However, in 2009 NASA SPoRT produced the pseudo-GLM (PGLM) product (Figure 2).  It was not an exact replica of future GLM observations, but it represented a reasonable facsimile that allowed for hands-on training that let forecasters better learn about total lightning and its relation to storms intensity and severe weather.

pglm_example

Figure 2:  An example of the pseudo-GLM flash extent density product derived from the Washington D.C. lightning mapping array during the derecho event of 2012.  The radar reflectivity (right) shows a strong line of storms approaching Washington D.C.  The greatest reflectivities (and likely strongest storms) are towards the northeast.  The pseudo-GLM (left) shows that it has update sooner than the radar, but also emphasizes the northeastern end of the line.  In fact, over 110 flashes are observed in two minutes at one location, highlighting the strongest overall storm.  The convection to the southwest is weaker as evidenced by the lack of pseudo-GLM observations.

NASA SPoRT, thanks to funding via the GOES-R visiting scientist program, was able to reach out to each of the other LMAs that were in operation across the country (and one in Canada!).  By 2014, almost a dozen LMAs were collaborating.  This allowed for the PGLM to be produced for numerous locations as well as expand partnerships to over a dozen forecast offices, three center weather service units, and the Aviation Weather Center.  Figure 3 shows the approximate domain and collaborating organization for each available LMA.  Combined, the collaborations between the forecasters, LMA owners, and the testbeds allowed for a wide variety of feedback discussing operational uses and visualization concepts.  Much of this has directly supported my own efforts as the GLM satellite liaison for the ongoing work in the Satellite Foundational Course for GOES-R, preparing for operational applications training, and the 2017 summer assessment.

lma_collaborators

Figure 3:  The approximate domain of the 11 collaborating lightning mapping arrays that have been used as part of the GLM preparations for the GOES-R Proving Ground.  The numbers correspond to the list at top showing the owners of the collaborative network.

Next up in the “GLM is coming” series is a post describing the GLM instrument itself as we await in initial release of GLM data.

A New Total Lightning Web Display

The SPoRT Center regularly works to display unique data in products, such as total lightning from ground-based lightning mapping arrays (LMAs), in the Weather Service’s display system; AWIPS II.  However, there is occasionally an opportunity to try a different method for specific operational applications.  One of those opportunities came with the Morristown, Tennessee forecast office.  Here, the collaboration was looking for a web-based visualization in order to better collaborate with emergency managers.  Feedback to SPoRT requested the need for a real-time display that could animate the data, auto-update, and allow zooming to a feature that would not reset with an update.  Additionally, there was a need to make this functional on mobile devices.

This has resulted in the test display shown here of the North Alabama Lightning Mapping Array flash extent density from July 1, 2015 from 1:30-4:00 PM (Central).  Like the more traditional display in AWIPS II, this flash extent density highlights the main storm cores where the updraft is intensifying, shows the spatial extent of total lightning, and even highlights several long flashes into the stratiform region behind the main convection, as shown in the still images below.  While the display is just in a development state now, it is demonstrating the potential for how to bring these data to emergency managers and Weather Service forecasters who may be in the field and not in the office, such as for special outdoor events.

Demonstration total lightning web display

The North Alabama Lightning Mapping Array flash extent density animation from 1:30-4:00 PM (Central) on July 1, 2015 in a new demonstration web display.  State and county boundaries are in black, while interstates are blue and major U.S. highways are in red.  (Click for the full resolution image.)

The two images below show a still image from 2:14 PM (Central) of the total lightning flash extent density and the corresponding radar reflectivity.

A still taken from the animation above at 2:14 PM (Central).  The main storm core and stratiform region lightning are highlighted.

A still taken from the animation above at 2:14 PM (Central). The main storm core and stratiform region lightning are highlighted. (Click for the full resolution image.)

 

The corresponding radar reflectivity at 2:14 PM (Central) for the still image above highlighting the locations of the total lightning features.

The corresponding radar reflectivity at 2:14 PM (Central) for the still image above highlighting the locations of the total lightning features. (Click for the full resolution image.)

A National Center Perspective of the pseudo-GLM product

Last year, SPoRT had the opportunity to expand collaborations with the Aviation Weather Center (AWC) and Storm Prediction Center (SPC) through a GOES-R Visiting Scientist Proposal.  The effort was focused total lightning activities with the National Centers, who have different operational perspectives from a local forecast office.  In addition to providing training on total lightning, SPoRT learned a great deal about the day-to-day operations at each of these locations.  SPoRT works to put products into the end user’s decision support system.  At the National Centers, this is N-AWIPS, which presented an interesting challenge for the GOES-R Geostationary Lightning Mapper (GLM) demonstration data set, the pseudo-GLM.  Unlike a local forecast office, the National Centers needed to be able to see each network at the same time and this need resulted in the PGLM mosaic.  First transitioned in June 2012, the PGLM mosaic has been evaluated informally at AWC and SPC.

Next week, the PGLM will be a participant with the AWC’s Summer Experiment in support of the GOES-R Proving Ground.  In preparation, SPoRT has coordinated with the AWC and SPC GOES-R Satellite Champions to update the display with improved color curves, lightning mapping array range rings, and network status bars as well as producing a training module on the PGLM that is more geared towards the National Centers.  Below is a screen capture of the newest display in N-AWIPS.  SPoRT is currently coordinating with the GOES-R Satellite Champion to transition this product to the Weather Prediction Center / Ocean Prediction Center.  Lastly, generating the mosaic is possible through collaborations that provide the real-time data to SPoRT from several organizations.  There are currently seven collaborating networks.  These are the Colorado Lightning Mapping Array (Colorado State / New Mexico Tech), Houston Lightning Mapping Array (Texas A&M / New Mexico Tech), Langmuir Laboratory Lightning Mapping Array (New Mexico Tech), North Alabama Lightning Mapping Array (NASA), Oklahoma Lightning Mapping Array (University of Oklahoma), Washington D.C. Lightning Mapping Array (NASA), and West Texas Lightning Mapping Array (Texas Tech University).

new_pglm_mosaic2

A screen capture of the pseudo-GLM mosaic product in N-AWIPS for use at the National Centers with the newest color curves, range rings, and network status bars.

Total Lightning Perspective of the Moore, Oklahoma Supercell

This particular blog post focuses on total lightning observations from the Moore, Oklahoma tornado.  SPoRT is participating in the annual NOAA Hazardous Weather Testbed Spring Experiment in Norman, Oklahoma.  The Spring Experiment is demonstrating new NOAA and NASA experimental capabilities as part of the annual Experimental Warning Program.  One NASA capability being demonstrated is total lightning associated with severe / tornado weather events.  The data used were NOT from NASA, but from the Oklahoma Lightning Mapping Array operated by the University of Oklahoma.  NASA SPoRT has access to these data through a collaboration to support the Hazardous Weather Testbed and demonstrates SPoRT’s software plug-in to display these data in the National Weather Service’s AWIPS II system.  Also, this collaboration is demonstrating the SPoRT / MDL total lightning tracking tool.  This particular post discusses the connection of total lightning and tornado occurrence consistent with the “lightning jump” concept developed by Christopher Schultz (NASA Coop) and the lightning team here at the Earth Science Office.  These experimental data were not available to the Norman, Oklahoma forecast office and this post is intended as a discussion of how these data may have been used.

Figure 1 takes place at 1910 UTC and shows a 4-panel display from AWIPS II.  The lower two panels show radar observations of storm relative velocity (left) and reflectivity (right).  The top panels show two total lightning products.  The first is the source density product (left), which is used by several SPoRT partners in operations.  The pseudo-geostationary lightning mapper (PGLM – right) is the demonstration product SPoRT is providing to the Hazardous Weather Testbed this year to demonstrate what the future Geostationary Lightning Mapper observations may look like.  The PGLM data are derived from the ground-based lightning mapping array data.  In this case it is from the Oklahoma network provided to SPoRT with this collaboration.  Lastly, please note the two pop-up windows.  These display the output from the SPoRT / MDL total lightning tracking tool, which is a time series of the source densities (left) and PGLM (right) observations, respectively.  Newcastle and Moore, Oklahoma are circled for reference.

Figure 1: AWIPS II four panel display from 1910 UTC that shows the total lightning source density (upper left), and pseudo geostationary lightning mapper flash extent density (PGLM - upper right), along with the radar storm relative velocity (lower left), and radar reflectivity (lower right).  The pop-up windows show the total lightning tracking tool's time series plot for the source densities (left) and PGLM flash extent density (right), respectively.

Figure 1: AWIPS II four panel display from 1910 UTC that shows the total lightning source density (upper left), and pseudo geostationary lightning mapper flash extent density (PGLM – upper right), along with the radar storm relative velocity (lower left), and radar reflectivity (lower right). The pop-up windows show the total lightning tracking tool’s time series plot for the source densities (left) and PGLM flash extent density (right), respectively.

Both the source density and PGLM demonstrate a lightning jump around 1910 UTC, as shown by the spike in observations in the time series (~800 sources and 46 flashes, respectively).  Christopher Schultz’s official lighting jump algorithm supports this visual inspection as it too indicated a lightning jump.  Interestingly, the first severe thunderstorm warning was issued at 1912 UTC and based on radar observations at 1908 UTC.  Normally, we train that lightning jumps will precede severe weather, so why is the jump coincident with the initial severe thunderstorm warning?  The answer is that the environment in central Oklahoma was extremely favorable for tornadic supercells.  As such, as storms showed any signs of growth a warning was issued.  This is similar to how the Huntsville forecast office operated during the April 27, 2011 outbreak as there were so many violent storms across the region.  Given the environment, the total lightning would play a reinforcing role as the lightning jump at 1910 UTC indicates that this storm is rapidly strengthening and becomes rooted in the boundary layer.  One feature that the total lightning observations provide is a very rapid update cycle.  The total lightning data update every minute, versus the radar updating every 4-6 minutes.  This means that the total lightning observations are providing continuous updates into how the storm is evolving, allowing the forecaster to evaluate the storm’s growth in between radar volume scans.

We will next step forward to 1928 UTC, shown in Figure 2.

This is the same as Figure 1, but at 1928 UTC.

Figure 2: This is the same as Figure 1, but at 1928 UTC.

The total lightning observations begin to undergo a second, reinforcing lightning jump at 1928 UTC.  The time series plot is less obvious than from 1910 UTC, particularly with the source densities, but the lightning jump algorithm did flag a reinforcing jump at this time.  At this point, this is 12 minutes before the official tornado warning at 1940 UTC and 28 minutes prior to the reported touchdown time of 1956 UTC, near Newcastle, Oklahoma.  This reinforcing jump emphasizes to the forecaster that something is occurring and that the storm continues to intensify.  Given that a severe thunderstorm warning is already active, this reinforcing jump alerts the forecaster that this storm is unlikely to weaken soon.  The radar reflectivity emphasizes this as well, as it begins to take on a supercell structure and a faint hook echo may be forming (circled in reflectivity frame).

Figure 3 comes at 1940 UTC, shown in Figure 3, when the tornado warning was issued.

This is the same as Figure 1, but at 1940 UTC.

Figure 3: This is the same as Figure 1, but at 1940 UTC.

At this stage, the lightning activity has decreased somewhat after the initial jump at 1910 UTC and the reinforcing jump at 1928 UTC.  Radar continues to show intensification, particularly with the radar velocity couplet clearly evident to the west-southwest of Newcastle, Oklahoma.

We will next step ahead to 1950 UTC, just prior to the touchdown of the tornado at 1956 UTC in Figure 4.

Figure 4: This is the same as Figure 1, but at 1950 UTC.

Figure 4: This is the same as Figure 1, but at 1950 UTC.

At this stage, the tornado warning has been active for 10 minutes and the radar observations show the classic hook echo and velocity couplet signatures.  Both total lightning products show one final increase in activity, but given the high values for the past few minutes, this is not a third lightning jump.  The tornado would touchdown 6 minutes later just outside of Newcastle, Oklahoma before further intensifying and moving through Moore, Oklahoma.

Christopher Schultz provided an additional radar analysis that is a cross section of the radar azimuthal shear (a measure of the storm’s rotation) in time in Figure 5.  Red vertical bars show the occurrence of the original and reinforcing lightning jump at 1910 and 1928 UTC, respectively.  Of note is the large increase in azimuthal shear after each lightning jump prior to the tornado’s touchdown.

Figure 5: A radar azimuthal shear cross section plot from 1900-2300 UTC.  The red bars indicate the times of lightning jumps from the lightning jump algorithm.

Figure 5: A radar azimuthal shear cross section plot from 1900-2300 UTC. The red bars indicate the times of lightning jumps from the lightning jump algorithm.

Once again, I would like to re-iterate that these experimental data were not available to the forecasters in real-time and that this is a post-event analysis.  Overall, the total lightning data behaved as expected, with two lightning jumps preceding the severe weather and the tornado that would later impact Moore, Oklahoma.  Based on the extremely favorable environment for tornadic supercells, the total lightning data would play a supporting role providing insight into the storm’s development for a forecaster, particularly with its one minute update times.  Total lightning typically has more utility in marginal events, but the post-analysis here shows that the underlying concepts of what drives a lightning jump are just as valid here.