Blizzards, dust storms, fires, floods, thunderstorms and tornadoes were the result of very strong back-to-back negatively tilted shortwave troughs through the eastern half of the U.S. From Texas to Wisconsin, New York to Florida, severe wind gusts, large hail, and deadly tornadoes were recorded. The hardest hit states during the three-day severe weather outbreak (as far as thunderstorm hazards are concerned) were Missouri, Arkansas, Mississippi, and Alabama.

The GOES-16 ABI water vapor RGB animation below (courtesy of College of Dupage) shows the development and occlusion of the low-pressure system on 3/14-15, evidenced by the beautiful spiral vortex in the middle of the country. On its heels, another shortwave trough further south spawned several rounds of intense thunderstorms, tapping into the rich Gulf moisture on 3/15. This latter system would then menace the eastern U.S. on 3/16, particularly Ohio, West Virginia, Pennsylvania, and New York.

The NOAA/CIMSS ProbSevere system of models are used to provide forecasters automated guidance to convective weather threats in the short term, generally 0-60 minutes. While there are plenty of examples to look at, I will focus on some strong tornadic supercells in southern Mississippi.

LightningCast, a trained AI/ML model, uses solely GOES-R ABI data to predict next-hour lightning. It is intended to aid forecasters in providing impact-based decision support (IDSS) to partners, but also to provide general convective initiation guidance. In the line of quickly developing storms from far southwest Mississippi, zipping down into eastern Louisiana, LightningCast was able to give anywhere from 15 to 30 minutes of lead time to lightning initiation (measured from the yellow 30% contour in the animation below). Considering the rapid evolution of these cells, and the overall busy situation of the day, this could really help provide forecasters with some advanced notice of sustained convection in a dangerous environment.

Above: The background is the GOES-16 day-cloud-phase-distinction RGB. The foreground blue-to-red pixels is the GOES-16 GLM flash-extent density. The labeled contours are predictions made by LightningCast.

Once these supercells in Louisiana and southern Mississippi matured, the ProbSevere IntenseStormNet provides an estimate of “intensity” using an image-based AI/ML model. Specifically, it uses images of the 0.64 µm reflectance, 10.3 µm brightness temperature, and GLM flash-extent density to make predictions. While the IntenseStormNet could be used as a satellite-only model for severe weather guidance, it has been incorporated into ProbSevere v3, providing better satellite information at the mature phase of storm development.

Below, in south-central Mississippi, the poor village of Tylertown was hit twice by tornadic supercells in the span of 40 minutes, causing devastating damage. We can see > 90% probabilities from the IntenseStormNet and a string of tornado reports following the line of recent tornado reports. The robust overshooting tops, cold-U/above-anvil cirrus plume signatures, and strong GLM flash cores all contributed to the high probabilities.

Storm chaser Tanner Charles shows the velocity couplets from the Jackson, MS radar passing over very nearly the same location.

Not a loop! Two tornadoes going over the same area twice with 40 minutes of each other. Geez #tornado #mswx pic.twitter.com/1KtnNxV8rA

— Tanner Charles ? (@TannerCharlesMN) March 15, 2025

ProbSevere version 3 is an upgrade to the operational version 2, and uses more radar, satellite, lightning, and short-term NWP data to better predict severe weather hazards such as wind, hail, and tornadoes.

In the first supercell that hit Tylertown, the probability of tornado was 68% at the time of the first NWS tornado warning. This is an extreme value for ProbTor v3. Unsurprisingly, the low-level MRMS AzShear (i.e., low-level rotation) was the top contributor, with the environmental shear, low-level mean wind (closely correlated with 0-1 km shear), and mid-level rotation the next top-contributing predictors. The storm’s flash rate, IntenseStormNet probability, and MLCAPE from the HRRR model provided additional boosts to the probability.

After this first tornadic supercell hit Tylertown, it would produce major tornadoes in Taylorsville, MS, and would go on to produce tornadoes in western Alabama, nearly 200 miles away from Tylertown! The secondary maximum in the ProbTor v3 probability (top center panel, red line) occurred before and during the tornado in western Alabama. Note also how the PTv3 increased before PTv2 as the storm was developing (compare red lines in top left and top center panels), a trend we’ve observed in a number of tornadic storms. ProbSevere v3 is slated to be operational in NOAA this summer (see paper here for more details on the models).

View only this post Read Less

Next-Generation Fire System (NGFS) data within a RealEarth instance (direct link), above, shows the rapid development and growth of multiple fires during the day on 14 March 2025. The RealEarth instance might be an easier way to track fires on an Extreme Fire Weather day like 14 March; the NGFS Alerts Dashboard (https://cimss.ssec.wisc.edu/ngfs) below (showing only detections in Texas and Oklahoma), in a screenshot from just after 0000 UTC on 15 March), shows many many detections and unless you are keenly aware of Oklahoma geography and county names, it’s hard to know at a glance which fires are which. When plotted on a map as shown above, they are easier to track.

The RealEarth instance has imagery that is probe-able. The probe shown below of the Hickory Hill Road fire lists out different satellite-derived properties at the point probed.

---

Multiple Fire Warnings have been issued over Oklahoma. At 2336 UTC, the Norman OK WFO website (above) showed the location of active warnings. The NGFS RealEarth display for that time is shown below. Not all of the fires indicated at that time below have Fire Warning associated with them. A Fire Warning occurs as a request to the NWS Forecast Office by their partners. For example, if you were to click on the map above within the Fire Warning polygon southwest of Stillwater, you would have access to the Fire Warning text shown at bottom, issued at the request of the Oklahoma Forestry Service.

The graphic below, from WFO OUN, testifies to the historic nature of this wildfire outbreak. The extraordinary winds — Stillwater OK airport saw gusts exceeding 70 knots — are helping to drive the fires. The widespread dust is apparent in the RealEarth imagery above

There was a very timely and informative Satellite Book Club presentation on Wildfire operations on 13 March 2025. Click here to listen.

View only this post Read Less

Aided by strong southwesterly winds gusting as high as 72 knots (83 mph) behind a cold front, more than 130 wildfires rapidly developed and spread across 44 counties in Oklahoma on 14 March 2025. 1-minute Mesoscale Domain Sector GOES-16 (GOES-East) Visible images (above) included overlays of 1-minute GOES-16 Fire Mask derived product (a component of the Fire Detection and Characterization Algorithm, FDCA) and 30-minute Peak Wind Gusts —  which showed a marked increase in wildfire thermal signatures after about 1900 UTC. Notable wildfires affected the Oklahoma City (KOKC) area (and nearby Norman), in addition to Guthrie (KGOK) and Stillwater (KSWO) — prompting the issuance of Fire Warnings and evacuation notices for those (and many other) locations.

Incidentally, the wildfire just southwest of Stillwater (KSWO) burned through an Oklahoma Mesonet site from 2041-2045 UTC (2043 UTC GOES-16 image):

A fire burned through our Lake Carl Blackwell Mesonet site during 3:41-3:45pm. One of our thermometers measured a 1-minute average temperature of 173 F at 3:42pm. The 2 meter anemometer is now reporting 0, which indicates that it likely was completely melted. #okwx #okmesonet

— Oklahoma Mesonet (@okmesonet) March 14, 2025

The strong winds responsible for the wildfires were also lofting large amounts of blowing dust across the Southern Plains. 5-minute CONUS Sector GOES-16 True Color RGB images from the CSPP GeoSphere site (below) displayed the broad swath of dense blowing dust (shades of tan) originating from New Mexico/Texas — as well as another plume of blowing dust moving south-southeastward from eastern Colorado to southwestern Kansas and northwestern Oklahoma (on the back side of a deep low pressure center). In addition, 2 large smoke plumes (pale shades of white) originating in Oklahoma were rising above the blowing dust as they eventually moved eastward and northeastward.

1-minute GOES-16 Visible images that included plots of Ceiling/Visibility (below) showed how the blowing dust and smoke restricted surface visibility at many locations across Oklahoma, down to values as low as 1/4 mile at times.

===== 15 March Update =====

A 24-hour animation of GOES-16 daytime True Color RGB + Nighttime Microphysics RGB images (above) showed the long-range transport of airborne dust from New Mexico/Texas to the western Great Lakes. Dust exhibited brighter shades of magenta in the Nighttime Microphysics RGB imagery.

A longer 36-hour animation of GOES-19 (Preliminary/Non-operational) Dust RGB images created using Geo2Grid (below) also highlighted the blowing dust as brighter shades of magenta. However, toward the end of the day on 15 March the magenta signature of the airborne dust became more muted as it moved northward from Wisconsin and Michigan toward far southern Ontario.

View only this post Read Less

Imagery from CSPP Geosphere (direct link), above, shows the condensation trail produced by the launch of the SpaceX rocket. Only routine CONUS 5-minute scanning was available to view the launch; very active weather over the southern Plains caused the GOES-16 mesoscale sectors to be positioned north and west of Florida. The 2303 UTC launch (news link), shortly before sunset, created a crescent-shaped condensation trail at 2306 UTC as denoted by the arrow in the animation above. After sunset, Night Microphysics RGB imagery highlights the plume (also highlighted by the arrow) because of its different temperature compared to its surroundings. The rocket is destined for the Space Station and a crew exchange (News article).

__________

A special Mesoscale Domain Sector from GOES-19 (Preliminary/Non-operational data) provided imagery at 30-second intervals — and Rocket Plume RGB images created using Geo2Grid (below) revealed thermal signatures of the SpaceX Falcon 9 rocket’s Stage 1 booster engines during the first 4 minutes after launch. The “boost-back burn” (to initiate the Stage 1 booster’s return to the launch site) signature was particularly notable at 23:06:55 UTC.

A closer view using 30-second GOES-19 Near-Infrared (1.38 µm / Band 4, 1.61 µm / Band 5 and 2.24 µm / Band 6) and Shortwave Infrared (3.9 µm / Band 7) images (below) displayed thermal signatures of the Stage 1 booster engines as the Falcon 9 rocket rapidly ascended northeast away from Cape Canaveral during the initial 2 minutes post-launch. In its wake, a more subtle signature of the rocket condensation cloud (that was highlighted above in GOES-16 RGB imagery) could be seen in the GOES-19 Band 4/5/6 images.

View only this post Read Less