Severe Weather in southeast Texas

January 6th, 2021 |

GOES-16 Day Cloud Phase Distinction RGB, 1646-2136 UTC on 6 January 2021, along with surface METARs (Click to animate)

The Storm Prediction Center in Norman issued a Slight Risk (click for map, from here) of severe weather over portions of southeast TX on 6 January 2021. The Day Cloud Phase Distinction RGB, shown above (click the image to animate) shows a developing line of convection stretching through the SLGT RSK area (The tallest convective cloud tops acquire a yellowish tint as they glaciate; lower clouds are blue/green/cyan).  The Day Cloud Phase Distinction RGB also allows for easy visualization of vertical wind shear:  the high cirriform clouds (orange and red) move in a distinctly different direction than the low cumuliform clouds (blue and green).  A Severe Thunderstorm Watch (Watch #2 on the year) was issued at 1900 UTC (Click here for Radar image that accompanied the watch issuance).  How could various satellite-based (or satellite-influenced) products be used to anticipate and to quantify the likelihood of severe weather during the day?

Polar Hyperspectral Sounding (PHS) data (from CrIS on Suomi NPP/NOAA-20 or from IASI on MetOp, for example) can augment Advanced Baseline Imager (ABI) data from GOES-16 (or GOES-17) to allow for better initialization of moisture fields in models. PHS data are linked to ABI information at the time of the polar orbiting overpass, and that relationship is carried forward in time. This data fusion process (PHSnABI) combines the excellent spectral resolution of the PHS with the superior spatial and temporal resolution of the ABI. When those data are used to initialize a model, it is frequently the case that the better moisture distribution within the PHSnABI fields leads to a more refined forecast of convection. (See this website for more information and for current model fields) Was that true on this day?

The toggles below show data from models runs initialized at 1400 and 1500 UTC, with model fields at 1800, 2000 and 2200 UTC. Lifted Index fields are shown with data from a Rapid Refresh-type simulation (that is, with no incorporation of fused PHSnABI data) identified as ‘RAP’ in the label; with data from a Single Data Assimilation (‘SDA’) system; and with data from a Continuous Data Assimilation (‘CDA’) system.

The CDA model system does appear best at simulating the timing of the convection that moves through southeast Texas (if one can use simulated Lifted Index as a proxy for the leading edge of convection).

Lifted Index at 1800 UTC from Models (RAP, SDA, and CDA) initialized at 1400 UTC (Click to enlarge)

Lifted Index at 1800 UTC from RAP, SDA and CDA models initialized at 1500 UTC (Click to enlarge)

Lifted Index at 2000 UTC from RAP, SDA and CDA models initialized at 1400 UTC (Click to enlarge)

Lifted Index at 2000 UTC from RAP, SDA and CDA models initialized at 1500 UTC (Click to enlarge)

Lifted Index at 2200 UTC from RAP, SDA and CDA models initialized at 1400 UTC (Click to enlarge)

Lifted Index at 2200 UTC from RAP, SDA and CDA models initialized at 1500 UTC (Click to enlarge)

NOAA-20 VIIRS imagery at 1823 UTC: 1.61 µm, True Color and False Color (Click to enlarge)

NOAA-20 overflew the convection at 1823 UTC, and the imagery above was processed at the Direct Broadcast site at CIMSS. (It is available for AWIPS via an LDM feed, and also as imagery for one week at this website; data for other days is here). VIIRS I3 (1.61 µm), True-Color and False-Color imagery from VIIRS all show a well-developed convective system at 1823 UTC.

As the convective event is unfolding, NUCAPS profiles derived from NOAA-20 can be used to diagnose the thermodynamic state of the atmosphere.  The toggle below shows 5 different profiles over southeastern Texas (along a line to the west of Galveston Bay) at ca. 1830 UTC.  The green points are NUCAPS profiles for which the infrared retrieval has converged to a solution.  A general decrease in stability (and increase in moisture) is apparent for profiles closer to the convection.  The red point (a profile for which the infrared and microwave retrieval both failed) is included as well.

NUCAPS profiles at select points as indicated over southeast Texas, 1830 UTC on 6 January 2021 (Click to enlarge)

A simpler, faster way to view the thermodynamic fields within NUCAPS profiles is to use gridded fields.  NUCAPS data are gridded onto constant pressure surfaces (using Polar2Grid software). The Total Totals Index field, below, shows a corridor of instability inland over southeast Texas with values exceeding 50.

Total Totals index from gridded NOAA-20 NUCAPS values, ca. 1830 UTC (Click to enlarge)

During the actual convective outbreak, NOAA/CIMSS ProbSevere (available online here) offers a data-driven way to highlight the radar echoes most likely to be producing severe weather in the next 60 minutes. The animation below shows values at 15-minute timesteps (for simplicity); ProbSevere values can change every 2 minutes, however. Use ProbSevere in combination with radar scanning to increase confidence in warning issuance.

NOAA/CIMSS ProbSevere, every 15 minutes, 1715 – 2300 UTC on 6 January 2021 (Click to enlarge)

Severe Weather reports(source) for 6 January are shown below.

SPC Storm Reports from 6 January 2021 (Click to enlarge)

Using GOES ABI and deep learning to nowcast lightning

September 2nd, 2020 |

NOAA and CIMSS are developing a product that uses a deep-learning model to recognize complex patterns in weather satellite imagery to predict the probability of lightning in the short term. Deep learning is a branch of machine learning based on artificial neural networks, which have the ability to automatically learn targeted features in the data by approximating how humans learn.

A convolutional neural network (CNN) was trained on over 23,000 images of GOES-16 Advanced Baseline Imager (ABI) data to predict the probability of lightning, within any given ABI pixel, in the next 60 minutes. The CNN was trained using 118 days of data collected between May and August of 2018. Images of GOES Geostationary Lightning Mapper (GLM) flash-extent density (created with glmtools) were used as the source of lightning observations. Note that GLM is an optical sensor that observes both in-cloud and cloud-to-ground lightning.

The CNN currently uses four ABI channels: band 2 (0.64-µm), band 5 (1.6-µm), band 13 (10.3-µm), and band 15 (12.3-µm). Bands 2 and 5 are only utilized under sunlit conditions. Utilization of additional channels and time sequences of images is under investigation. The model uses a semantic image segmentation architecture to assign the probability of lighting in the next 60 minutes to each pixel in the image. The model is very computationally efficient, only needing 30 seconds to process the ABI CONUS domain and 3 seconds to process an ABI mesoscale domain using multithreading on a 40-CPU linux server.

Currently, the model only utilizes satellite radiances. Thus, it can be applied to nearly any spatial domain covered by the ABI or an ABI-like sensor (e.g. AHI). Based on near-real-time testing, the model routinely nowcasts lightning initiation with 10-30 minutes of lead-time. We expect the skill and lead-time will increase as new predictors (e.g. more ABI fields, NWP, radar where available) are added to the model.

Below are a sampling of recent examples. The base images are 0.64-µm reflectance, with GLM-derived flash-extent density overlaid as filled semi-transparent polygons. The flash-extent density is the accumulated number of flashes within the previous 5 minutes. The CNN-derived probabilities are displayed as contours at select levels (near-real-time output is available through RealEarth).

The overall objective is to improve lightning nowcasts in support of aviation, mariners, and outdoor events/activities. Beyond improving the CNN, our work will focus on packaging the output into actionable information for forecasters and other decision makers.

A cold front in Iowa

 

Thunderstorm development on sea-breeze boundaries in Florida and the Bahamas

 

Diurnally and orographically forced storms in the Southwest U.S. and Rocky Mountains

 

Storms in central Oklahoma, on the edge of Hurricane Laura’s cloud shield

 

A couple of examples over the Northeast U.S.

 

A boundary of convection on the southern bank of Lake Ontario

 

Storms in a warm sector in IL/IN/OH, perhaps along an outflow boundary

 

Southeast U.S. offshore region

The background image in this example transitions from 10.3-µm brightness temperature to 0.64-µm reflectance, while the flash-extent density enhancement also changes, in an attempt to enhance contrast.

Outbreak of severe thunderstorms across the Deep South

April 12th, 2020 |

GOES-16 "Red" Visible (0.64 µm) images [click to play animation | MP4]

GOES-16 “Red” Visible (0.64 µm) images [click to play animation | MP4]

A major outbreak of severe thunderstorms (SPC Storm Reports) occurred across the Deep South on 12 April 2020. 1-minute Mesoscale Domain Sector GOES-16 (GOES-East) “Red” Visible (0.64 µm) images (above) showed the development and propagation of deep convection during the 1200-2359 UTC period. The corresponding GOES-16 “Clean” Infrared Window (10.35 µm) images are shown below.

GOES-16

GOES-16 “Clean” Infrared Window (10.35 µm) images [click to play animation | MP4]

Some of the strongest long-track tornadoes occurred in southern Mississippi — a closer view of GOES-16 Visible, Infrared and Visible/Infrared Sandwich Red-Green-Blue (RGB) images (below) revealed the pulsing nature of overshooting tops — which exhibited cloud-top infrared brightness temperatures as cold as -77ºC at 2038-2039 UTC, about 35 minutes prior to the destructive tornado that moved through Bassfield — and well defined “enhanced-v” signatures were apparent in the Infrared and RGB imagery, with that signature’s warm wake immediate downwind (east) of the overshooting tops indicating the likely presence of Above-Anvil Cirrus Plumes.

GOES-16 "Red" Visible (0.64 µm ), "Clean" Infrared Window (10.35 µm), and Visible/Infrared Sandwich RGB images [click to play animation | MP4]

GOES-16 “Red” Visible (0.64 µm ), “Clean” Infrared Window (10.35 µm) and Visible/Infrared Sandwich RGB images [click to play animation | MP4]

GOES-16 "Red" Visible (0.64 µm) images, with time-matched SPC Storm Reports plotted in red [click to play animation | MP4]

GOES-16 “Red” Visible (0.64 µm) images, with time-matched SPC Storm Reports plotted in red [click to play animation | MP4]

1-minute GOES-16 Visible images (above) and Infrared images (below) include plots of time-matched SPC Storm Reports.

GOES-16 "Clean" Infrared Window (10.35 µm) images, with time-matched SPC Storm Reports plotted in cyan [click to play animation | MP4]

GOES-16 “Clean” Infrared Window (10.35 µm) images, with time-matched SPC Storm Reports plotted in cyan [click to play animation | MP4]

NOAA/CIMSS ProbSevere is a tool that could have been used during this outbreak to identify which radar cells were most likely to produce severe weather.  The image below, from here, shows the reports of severe weather, the warning polygons, and ProbSevere locations (a closer view of the Mississippi tornadoes can be seen here).

Severe weather reports from 12 April 2020 (Green: Hail; Blue: Wind; Red: Tornado), NWS Warning Polygons and ProbSevere locations (plotted as boxes when ProbSevere exceeded 50% (Click to enlarge)

===== 14 April Update =====

GOES-16

GOES-16 “Red” Visible (0.64 µm) and Normalized Difference Vegetation Index images [click to enlarge]

Southwest-to-northeast oriented tornado damage paths in southern Mississippi were evident in a toggle between GOES-16 Visible and Normalized Difference Vegetation Index (NDVI) images (above). NDVI values within the damage path were generally 0.6, compared to 0.7-0.8 in adjacent areas. According the the NWS Jackson storm survey, the maximum path width of the longest-track (~67 mile) EF-4 tornado that began near Bassfield was about 2 miles — the widest ever measured in Mississippi, and one of the widest tornado damage paths ever measured in the US.

In a toggle between Aqua MODIS NDVI and Land Surface Temperature (LST) images (below), LST values were 5-10ºF warmer — low 80s F, darker shades of red —  within the tornado damage path, compared to areas adjacent to the path.

Aqua MODIS Normalized Difference Vegetation Index and Land Surface Temperature images [click to enlarge]

Aqua MODIS Normalized Difference Vegetation Index and Land Surface Temperature images [click to enlarge]

The tornado damage paths were also apparent in a comparison of before (26 March) and after (14 April) Aqua MODIS True Color RGB images (below) from the MODIS Today site. Note that 2 smoke plumes were seen on the 26 March image.

Aqua MODIS True Color RGB images from 26 March and 14 April [click to enlarge]

Aqua MODIS True Color RGB images from 26 March and 14 April [click to enlarge]

True and False-color imagery from NOAA-20 (from this (temporary) website) also show the damage path.

True- and False-Color imagery from the afternoon NOAA-20 overpass on 14 April 2020 (Click to enlarge)

NOAA-20 True Color RGB imagery of the Mississippi EF-4 tornado damage path that had a maximum with of 2 miles is shown below, using RealEarth.

NOAA-20 VIIRS True Color RGB image, including county outlines and map labels [click to enlarge]

NOAA-20 VIIRS True Color RGB image, including county outlines and map labels [click to enlarge]

Strong Tornado through Nashville Tennessee

March 3rd, 2020 |

GOES-16 Clean Window (10.3 µm) Infrared Imagery at 1-minute intervals, 0313 – 0722 UTC on 3 March 2020 (Click to play mp4 animation)

A long-track tornado moved through Nashville (north Nashville to Lockeland Springs then towards Lebanon) early in the morning on 3 March 2020. (ProbSevere for this event is also discussed in this blog post; A GLM View of the storm is here) The Band13 mp4 animation, above (Click here for an animated gif), from 0313 through 0722 UTC shows the long-lived storm forming over western Tennessee and rolling eastward into Nashville after midnight. As the storm approached Nashville, the storm top frequently showed isolated cold pixels (yellow in the enhancement) that are testament to the strong convection, and an above-anvil cirrus plume (AACP) might be there too. All of these structures are consistent with a tornadic storm. The 00 UTC Nashville Sounding shows a favorable environment as well.

1-minute GOES-16 Infrared images with plots of time-matched SPC Storm reports is shown below, covering the period 0300-0802 UTC. The color enhancement applied to those images is slightly different — and the pulsing overshooting tops are easier to see as highlighted with darker shades of orange.

GOES-16

GOES-16 “Clean” Infrared Window (10.35 µm) images, with SPC Storm reports plotted in red [click to play animation | MP4]

Shortly after the tornado passed through Nashville, NOAA-20 overflew Tennessee. The toggle below shows the GOES-16 ABI Band 13 (10.3 µm) infrared imagery, the NUCAPS Sounding Location, and the computed Total Totals Index from the NUCAPS thermodynamic information. Values exceeding 50 are south and west of the thunderstorm complex. NUCAPS soundings are showing a thermodynamically unstable environment.

GOES-16 ABI Band 13 (10.3 µm) Infrared Imagery along with NUCAPS Sounding Points and Gridded Values of Total Totals Index computed from NUCAPS Soundings (Click to enlarge)

The high-resolution VIIRS instruments on NOAA-20 gives a high-resolution view of the storm, as shown below.  The figure includes the 11.45 µm VIIRS I05 infrared imagery (from Real Earth) and also shows a 5-minute GLM Group Density accumulation with a profound maximum (values exceeding 250!) over Lebanon TN.  The morning NOAA-20 orbit on 3 March was to the east of Tennessee (source), so a parallax shift between the GLM data from GOES-16 and the VIIRS I05 data will occur.

Real Earth captures of 11.45 µm VIIRS I05 infrared imagery and GOES-16 GLM 5-minute Group Density, 0700-0705 UTC

A zoomed-out toggle between VIIRS I05 and the Day Night Band (below, courtesy William Straka, CIMSS), (Click here to view the I05 with a different enhancement) shows the storm to the east of Nashville.  The quarter New Moon (that was below the horizon) offered no clear view of the cloud tops, but lightning illumination is apparent.

NOAA-20 11.45 µm VIIRS I05 infrared imagery and Day Night Band 0.70 µm visible imagery at 0717 UTC on 3 March 2020 (Click to enlarge)

This long storm in a favorable environment was captured well by the NOAA/CIMSS ProbSevere product.  The animation below tracks the storm across much of northwest and north-central Tennessee.  Time-series plots of the radar object (#181389) associated with the Nashville tornado are here:  ProbTor (including components), ProbSevere (including components), and ProbSevere, ProbHail and ProbTor values.  The cyclic nature of the storm is apparent.  ProbSevere/ProbTor/ProbHail values were very high when the Nashville tornado was on the ground (around 0640 UTC).

NOAA/CIMSS ProbSevere display over northwest to north-central Tennessee, 0330 to 0730 UTC. (Click to enlarge)


A project spearheaded by Bill Smith Sr and Qi Zhang at Hampton University (they are also affiliated with CIMSS) takes Polar Hyperspectral (PHS) data and blends it with GOES-16 ABI data.  Those fields that exploit the high spectral resolution from sounders in polar orbit (such as IASI or CrIS) and the high spatial and temporal resolution from GOES are then input into numerical models.  Output is here.  The better representation of moisture fields in these fields can help a model better define where strong convection will or will not occur.  The animation below shows initial fields of the Significant Tornado Parameter from 0300 – 0800 UTC on 3 March 2020. Compare the animation to initial fields of STP in a model that does not include the assimilated hyperspectral data here.

Significant Tornado Parameter at initial model run times from 0300 through 0800 UTC on 3 March 2020 (Click to enlarge). The number of assimilated retrievals is indicated in red; no retrievals were assimilated at 0700 UTC, and the STP field was affected!

What did the forecast for 0600 UTC look like from this model that includes polar hyperspectral data?  That is shown below, with a series of model runs.  There is run-to-run variability, but overall the forecast simulations show a peak in the STP parameter near Nashville. (This PDF compares model runs with and without Polar Hyperspectral Data; adding the satellite data helps the model focus convection where it occurs, mostly because the moisture fields are more accurately defined).

Forecasts valid at 0600 UTC on 3 March 2020, initialized at 0000, 0100, 0200, 0300, 0400 and 0500 UTC (Click to enlarge)