A negatively tilted shortwave trough and associated dynamical response created a rapidly deepening surface low in the Dakotas on May 15. These features worked together to bring high theta-E air from the southern U.S. up into Minnesota and Wisconsin, creating ample instability in the region. A strong mid-level jet streak provided deep-layer shear, and a surface dryline and large-scale dynamical lift provided the forcing mechanisms.

In the animation below, one can see the spiraling low pressure system in the Dakotas heading towards Minnesota. Deep convection lights up on and ahead of a surface dryline, which separates the dry air to its west from the warm, moist air to its east.

The ProbSevere LightningCast model uses AI and images of GOES-R data to predict the probability of next-hour lightning. It is expected to be operational in NOAA this summer, but NWS forecasters currently use an experimental feed from CIMSS to help monitor convection and provide messaging for decision support on lightning potential to stakeholders (e.g., event managers).

In the animation below, LightningCast probabilities were computed and displayed every minute, using a GOES-East mesoscale sector. The LightningCast output follows the deep convection heading into the Twin Cities region of Minnesota. Further south, LightningCast “sees” bubbling cumuliform clouds and anticipates electrification down the dryline by predicting higher probabilities before the observed lightning. Ahead of the dryline, LightningCast predicts increased probabilities on more-agitated cumulus clouds in southwest Wisconsin, despite some obscuring cirrus clouds (bluish-gray). Even with some explosive cloud growth, LightningCast provided 10-20 minutes of lead time to the first observed GLM flashes (GLM=Geostationary Lightning Mapper) in many new convective cores. This tool can provide automated, quantitative, calibrated guidance to forecasters for short-term lightning potential.

As many storms matured into supercells, all hazards were possible, including tornadoes. ProbSevere v3 uses satellite, radar, lightning, and NWP analysis data to predict next-hour severe-weather probabilities in a storm-object-based framework. Just about every storm had a severe-weather warning on it, issued by the NWS. ProbSevere can help forecasters perform triage in a busy situation (as this day certainly was), prioritizing threats. Many of these storms produced large hail, including 4″ hail in Eau Claire, Wisconsin and numerous 2″ hail reports across the state.

Only ProbTor and tornado warnings are plotted below (and MRMS MergedReflectivity background). From this, we can see that ProbTor does a good job highlighting the storms with enhanced tornado probabilities (blue-to-white-to-pink contours). Most of these storms had tornado warnings, and several tornadoes were reported in Wisconsin.

Along with the color-coded probabilities, forecasters can “hover” over storms with their cursor to see storm attributes. Let’s take a look at a couple of tornadic storms.

This storm east of the Twin Cities in Minnesota exhibited excellent radar rotational parameters (“AzShear”), strong deep layer shear, and very strong low-level kinematic parameters. Curiously, the ground-based total lightning flash rate was 34 flashes/min, whereas the space-based GLM lightning showed only 2 flashes in 5 minutes (FED=flash-extent density). This might be a case where the large cloud mass is dissipating the optical signal of lots of small flashes in the storm (“optical extinction”). In any event, ProbTor v3 had a 50% probability of tornado on this storm when a tornado was first reported. Users of ProbTor understand that 50% is a very high prediction for this model. The time series below (Figure 6) shows a large “jump” in tornado probability from 18% to 40%, about 4 minutes prior to the first tornado warning, due to an increase in low-level rotation. This storm went on to congeal with neighboring storms and produced numerous damaging wind reports in northwest Wisconsin.

Another storm congealed from hail-producing storms in Dane County, into a large supercell that ravaged Dodge County with hail, wind, and tornadoes. In the image below, we again see that the environment was very favorable for tornadoes (high shear and storm-relative helicity), the expected hail size was large (MESH=2.21 in), the lightning flash rate was high, and the storm rotational parameters were exceptionally high. This all resulted in a probability of tornado of 65%. In the time series below, a tornado warning was issued after a steady climb in probability from 10% to 65%. While the reports in the time series are preliminary, tornadoes likely occurred in the 22:30-23:00 UTC time frame. The town of Mayville declared a state of emergency due to the extensive damage. Hopefully residents were able to take proper shelter due to the timely NWS tornado warning.

One satellite predictor of ProbSevere v3 is the “intense convection probability” from IntenseStormNet, which is computed via a trained AI model and GOES imagery. In the animation below, we can see that high values of “intensity” (> 90%) follow the storm cores from south-central to east-central Wisconsin, highlighting strong overshooting tops, high storm-top texture, and high GLM flash rates (not shown). This product is particularly valuable where radar data is sparse, but it still skillfully contributes to the ProbSevere models even in good radar coverage, by providing an assessment of storm intensity from a geostationary satellite perspective.

View only this post Read Less

Tens of thousands of baseball fans took to Chicago’s North Side on the evening of 14 May 2025 to watch a Major League Baseball game between the Miami Marlins and the Chicago Cubs. Fans of the Cubs might not remember much about the game itself (the Marlins won 3–1 with all runs on both teams scored via solo home runs) but it might be a while before thy forget the weather. A fog began to envelop the friendly confines of Wrigley Field early in the game, and it continued until well after the last out was recorded.

GOES-19 was well situated to capture the development and evolution of this fog event. The fog formation and propagation straddled the transition from day to night, and thus different products are necessary to best assess what happened since GOES ABI channels like Band 7 (3.9 microns) have drastically different interpretations depending on the time of day. During the day, Band 7 is largely capturing shortwave reflection from the sun which reduces its quantitative value for daytime cloud products. However, at night the same band represents thermal emission. Not only that, but the emission at that wavelength is a function of cloud phase and particle size. The embedded movie from CSPP Geosphere shows the transition from day to night and the switch from a true color RGB to the Night Microphysics RGB. (You can adjust the playback speed by clicking on the three dots in the lower right corner).

At the start of this loop, in the late afternoon, strong evidence of a lake breeze can be found. There is substantial clearing in eastern Wisconsin, northeastern Illinois, and northwestern Indiana with the boundary between the cloudy and clear regions largely paralleling the shore of Lake Michigan. The effect of the lake breeze is even more readily apparent if one looks at the evolution of the cloud field throughout the day via the GOES-19 True Color RGB.

As the wind shifts to easterly flow, it brings a shallow layer of cold air ashore. This cold air is much more stable than the air it is replacing, and is often less moist too even though it originated over a vast lake. This is because the lake air is so much cooler than the land air that the absolute humidity levels are lower (even though the relative humidity levels are higher). The combination of cold, stable, and comparatively dry air kills off any convection that was taking place, so as the lake breeze propagates outward, it can easily be followed via satellite. Eagle-eyed readers might also notice that other, much smaller lakes in the area were also inhibiting convective development, including Lake Winnebago in the far upper part of the animation as well as the four lakes that characterize Madison, Wisconsin, in the middle left.

A quick visual inspection shows that the air quality of the existing airmass was much worse than the lake air that replaced it, too. A milky white pervades the regions away from Lake Michigan due to the enhanced shortwave scattering brought on by the high aerosol content; region closer to the lake are much more clear. The GOES Aerosol Optical Depth product confirms this, with notably higher aerosol optical depths further away from the shore; note the deep blues in eastern Wisconsin and northwestern Illinois with lighter blues further inland.

This persistent easterly flow was critical for what happened next. Air temperatures over the lake continued to fall, and thus reached the point of saturation. GOES-19 captured this perfectly as a filling in of the clouds over the lake, and only over the lake. Note that this was contemporaneous with sunset, so the animation segues from the daytime true color based on visible and near-infrared channels to the nighttime product which depends on longer wavelength channels.

The easterly flow persisted, however, and pushed the low-level foggy air ashore. The Night Microphyscis product clearly shows the extent of the fogginess. While it only benetrated slightly into Illinois, much of western Wisconsin experienced notable fog.

The shallowness of the fog layer is readily apparent from aircraft soundings. Over 100 commercial aircraft in the continental United States are equipped with temperature and water vapor sensors. Southwest Airlines is a major participant in this program, and with Chicago Midway being one of their most significant airports, there are several of these soundings every day. The Skew-T from 0208 UTC (9:08 local time) shows that the saturated later is only a few tens of millibars thick; above that, the air is too dry to support broad cloud formation in the absence of daytime convection. This image was obtained from Casey Webster’s wxster.com. Profiles from all over the continental United States using this publicly-available data source can also be found at that site and can be quite useful to help fill in the spatiotemporal gaps of the operational radiosonde network.

To see just how foggy it was, check out the highlights of the bottom of the 9th inning for the Cubs/Marlins game. You can see Major League Baseball’s video here.

View only this post Read Less

1-minute Mesoscale Domain Sector GOES-19 (GOES-East) Visible (0.64 µm) with an overlay of the FDCA Fire Mask derived product (above) showed the large thermal anomaly associated with a wildfire in the Nopiming Provincial Park (located just west of the Manitoba/Ontario border) as it made a rapid northeastward run of ~20 miles on 13 May 2025 — and also displayed the periodic bursts of pyrocumulus clouds that began shortly after 1800 UTC.

The corresponding 1-minute GOES-19 Infrared images (below) indicated that many of these pyrocumulus clouds eventually exhibited cloud-top 10.3 µm brightness temperatures of -40ºC (dark blue enhancement) or colder — a necessary condition to be classified as pyrocumulonimbus (pyroCb). This large wildfire generated a series of at least 8 pyroCb clouds from 1843 UTC on 13 May to 0007 UTC on 14 May.

Toggles between VIIRS True Color RGB and False Color RGB images from Suomi-NPP and NOAA-20 (as visualized using RealEarth) are shown below. Brighter-white pyrocumulus and pyroCb clouds were seen rising above the  wildfire smoke in both the True Color RGB and False Color RGB images; active fires appeared as brighter shades of pink in the False Color RGB images.

 

 

A larger-scale view using 10-minute Full Disk scan GOES-19 Infrared images (below) showed several of the pyroCb clouds as they were transported eastward across Ontario — with two of the larger pyroCbs exhibiting cloud-top 10.3 µm brightness temperatures as cold as the -60s C (shades of green).

———-

A cursor sample of the coldest cloud-top pixel of the large pyroCb located north of Pickle Lake, Ontario (CYPL) at 0050 UTC on 14 May (above) displayed a 10.3 µm infrared brightness temperature of -62.1ºC, along with a corresponding Cloud Top Temperature of -63.21ºC and Cloud Top Height of 41362.71 ft (12.6 km) at that pixel location.

Those values of cloud-top temperature and height appeared to be comparable to those of the tropopause on a plot of rawinsonde data from Pickle Lake at 0000 UTC on 14 May (below).

GOES-19 Infrared images with an overlay of GLM Flash Extent Density (below) revealed that there was a brief 10-minute period of lightning activity with one of the larger pyroCbs over Ontario.

The Nopiming Fire (along with the Ken Fire, just across the Ontario border) continued to burn aggressively throughout the subsequent nighttime hours — and the elongated nocturnal glow of those 2 wildfire footprints were very apparent in VIIRS Day/Night Band imagery from NOAA-20 and NOAA-21 (below), northeast and east of Winnipeg (CYWG).

The Nopiming Fire in Manitoba ran 250,000 acres yesterday (100,000 hectares) and the Lac Du Bonnet Fire ran over 10,000 acres. “Gone.” 1,000 people have been evacuated.

Reports that the fire displayed multiple column collapses and was making its own weather including lightning… pic.twitter.com/4ebrJyYARg

— The Hotshot Wake Up (@HotshotWake) May 14, 2025

View only this post Read Less

1-minute Mesoscale Domain Sector GOES-19 (GOES-East) GeoColor RGB images with an overlay of Next Generation Fire System fire detection polygons (above) displayed the thermal anomalies and dense smoke plumes associated with the Camp House Fire (near Brimson) and Jenkins Creek Fire (near Fairbanks) in northeastern Minnesota on 12 May 2025. These 2 wildfires prompted the issuance of a few evacuation orders.

Thick smoke drifted northeastward across the Boundary Waters Canoe Area Wilderness and crossed the Canadian border — GOES-19 Aerosol Optical Depth values with these smoke plumes were as high as 1.0 (below).

A NGFS probe of the Jenkins Creek Fire at 2120 UTC on 12 May (below showed various parameters of the thermal anomaly at that time.

The Camp House Fire began on the previous day, and had continued burning through the subsequent nighttime hours — its nocturnal glow was evident in a NOAA-21 VIIRS Day/Night Band image at 0831 UTC (3:31 AM local time), just northeast of Brimson (below).

1-minute GOES-19 Visible images with an overlay of the Fire Detection and Characterization Algorithm (FDCA) Fire Mask derived product (below) also showed the thermal signatures and smoke plumes of the 2 larger wildfires (along with a 3rd, smaller wildfire that began to the southwest at 1953 UTC near Shaw). South-southwest winds were gusting to 20-25 knots (23-29 mph) across the region, contributing to the extreme fire behavior.

An overpass of Landsat-8 provided a 30-meter resolution image of the Jenkins Creek and Camp House wildfires at 1657 UTC (11:57 AM local time) on 12 May, as viewed using RealEarth (below). The active fires appeared as brighter shades of pink, with their hazy smoke plumes extending northward.

View only this post Read Less