The dominant weather story across much of the central United States remains the continuing impact of the Minnesota and Ontario wildfires on the air quality in the upper Midwest. You can see just how impactful the smoke was by looking at this slider of two images from the Blog’s home office on the campus of the University of Wisconsin-Madison. These two iamges were taken 24 hours apart, with the one on the left showing Madison’s isthmus and downtown on the hot and humid morning of Wednesday 15 July, while the one on the right shows the smoke-filled skies of 16 July. Watch as the prominent State Capitol, just 1.1 miles (1.9 km) away, disappears under the cloak of smoke.

Here’s a map of PM2.5 from Airnow.gov valid for 1900 UTC (2 PM CDT) on 16 July 2026. Hazardous air levels stretch all the way from north central Minnesota all the way to central Pennsylvania. The challenge of monitoring this outbreak of hazardous air was complicated by the fact that the day before, GOES-19 experienced an anomaly and was unable to transit observations until the early afternoon of the next day. In this post, we’ll explore a few of the alternatives for keeping track of what the air was doing until the workhorse geostationary platform could be returned to service.

Here in the Blog’s Headquarters City of Madison, the air quality degraded in the hours after sunset. While it’s easy to follow smoke during the day using satellites, tracking smoke at night can be challenging because the smoke particles are generally not significant sources of infrared radiation. However, there are other tools at our disposal that can be used to monitor this situation. Among the most important tools are ground-based air quality networks, such as those operated by the EPA or PurpleAir. Here’s a time series of a PurpleAir sensor in downtown Madison showing an initial increase in particulate matter around 10:00 PM CDT (0300 UTC) where the Air Quality Index (AQI) plateaued close to 200, then another increase around 4:30 AM (0930 UTC) to truly astonishingly high levels.

We can actually see the atmospheric current that brought the flow in. Here’s a loop from the Milwaukee/Sullivan NEXRAD radar showing the inland penetration of a lake breeze well into central Wisconsin. That lake breeze brought some relief from the humid conditions that had been dominating Madison’s weather for the previous few days, but it also cleared out air that originated from the south and instead replaced it with smokey air from the north. It’s also interesting to see how the lake breeze impacted the propagation of the radar beams. Lake breezes create shallow inversions as the cool air undercuts and lifts warm air. In this case, we see that shortly after the lake breeze front passed over the radar, the ground clutter increased significantly. This is because the lake breeze passage fostered an inversion over southeastern Wisconsin, refracting the radar beams in unexpected ways and fostering anomalous propagation.

The lake breeze can also be tracked with surface weather sensors. The Reliable Automated Instrumentation Network (RAIN) suite of sensors on the rooftop of Blog HQ provide a multiyear archive of weather conditions on the scale of a minute. Here’s a meteogram of 24 hours of conditions on 15-16 July. The arrival of the lake breeze can be seen right at 0300 UTC (10:00 PM CDT) with the rapid drop in temperature, increase in wind speed, and sharp change in wind direction from northwest to east.

The same cameras that we looked at above can also be used in the overnight hours. Here’s a movie between 8:00 PM and midnight CDT that shows the arrival of the smoke. See how the clouds disappear and the lights on the far shore of the lake vanish as the smoke arrives. While not appearing as dramatic as it would if the

The Space Science and Engineering Center (SSEC) is, along with NOAA, a parent of CIMSS. Researchers at SSEC are renowned experts in instrument design and deployment, and one of the highlights of SSEC’s work is the High Spectral Resolution Lidar, a laser-based instrument designed specifically to look at how aerosols like smoke are distributed through the atmosphere. This is a time-height cross section of aerosol backscatter running from 0000 to 0600 UTC (7:00 PM to 1:00 AM). We can see the smoke as the bright layer that first appears right after 0300 UTC (10:00 PM) and is about 1 km thick.

All of these things: the radar-indicated lake breeze front, the surface weather conditions, the HSRL cross sections, and the visible camera movie; all of these show significant events that are coincident with the sharp increase in AQI measured by the surface station.

The polar orbiting satellites operated by NOAA’s Joint Polar Satellite System (JPSS) also provide some unique perspectives on the event that we can’t get from geostationary view. This VIIRS true color image from NOAA-20 at 1756 UTC (12:56 PM CDT) perfectly encapsulates how widespread the smoke is. Nearly the entire states of Wisconsin and Michigan are severely affected, with additional impacts stretching from the Red River Valley to the Acela Corridor.

The VIIRS true color product has a spatial resolution of 750 m, which is slightly better than the 1 km resolution of its GOES counterpart. However, the practical resolution is a bit better because the GOES resolution is defined at the equator and it degrades the further poleward you go. By contrast, the VIIRS resolution doesn’t have a latitudinal dependence. That enables us to zoom in a little more than we might for a geostationary product. When we do, we see yet another interesting phenomenon.

Do you see how the cumulus clouds are almost exclusively outside the smoke, while the smokey regions are otherwise generally cloud free? This is likely because the smoke is attenuating enough of the incoming solar radiation to prevent surface-based convection from starting. When we talk about “fair weather cumulus,” we should make sure that the definition of “fair weather” includes no smoke.

View only this post Read Less

1-minute Mesoscale Domain Sector GOES-18 (GOES-West) Infrared Window images (above) included plots of Flood Watch/Warning/Advisory polygons (Flash Flood Warnings appeared as red polygons, while Flash Flood Emergencies appeared as bold red polygons) across the Hill Country of Texas on 16 July 2026. After the previous few days of heavy rainfall — which produced amounts in excess of 25-28 inches in Kerr County and Uvalde County — the high rainfall rates of these latest thunderstorms caused considerable flash flooding (with at least 2 fatalities being reported). Note: a GOES-18 Mesoscale Sector was positioned over Texas at that time, since GOES-19 was still in “safehold” mode after experiencing an anomaly on the previous day.

A less-cluttered view of 1-minute GOES-18 Infrared Window images (below) made it easier to see the pulses of cold overshooting tops, and the 1-hour precipitation amounts that occasionally exceeded 1-2 inches at a couple of METAR sites. GOES-18 Rainfall Rate derived product values were as high as 1.31 inches per hour just north of Uvalde at 0804 UTC (the Uvalde METAR site had just received 1.20 inches of rainfall during the previous hour).

The coldest GOES-18 cloud-top infrared brightness temperature associated with the convection was -77.83 C; according to a plot of rawinsonde data from Del Rio, Texas (below) this temperature was not far below the Most Unstable (MU) air parcel’s Maximum Parcel Level (MPL), which represented a significant overshoot of its Equilibrium Level (EL).

1-minute GOES-18 Infrared Window images with plots of 1-minute GLM Flash Points (below) revealed considerable lightning activity with these thunderstorms that produced heavy rainfall.

View only this post Read Less

Numerous wildfires began to increase in size and intensity from northern Minnesota to central Canada in mid-July 2026, as previously discussed on this blog (13 July | 14 July | 15 July). A large scale view of the GOES-19 (GOES-East) Aerosol Optical Depth (AOD) derived product on 15 July (above) displayed the long-range transport of this wildfire smoke across parts of the central/eastern Lower 48 states and the western Atlantic Ocean. The southernmost plume of high AOD was being recirculated inland from the Atlantic. Surface weather symbols of smoke or haze indicated that some of this smoke was reaching the ground (and adversely affecting visibility and air quality).

Even though the Aerosol Optical Depth derived product has a default AWIPS color table range from 0 to 1, the actual sampled AOD values associated with particularly dense smoke were as high as 3.22 over central New York (below).

Gaps in the AOD were due to the Cloud Mask product preventing AOD creation where thick clouds were present — so simply looking at GOES-19 Visible imagery (below) perhaps offered a smoother depiction of the hazy smoke transport.

GOES-19 True Color RGB images (below) provided the best qualitative view of the wildfire smoke transport.

The GOES-19 imagery shown in this blog post ended at 2021 UTC — shortly thereafter, the satellite suffered an anomaly, and went into “safehold” mode while NOAA/NESDIS engineers worked to diagnose the problem.

View only this post Read Less

Friend of the Blog Patrick Ayd, Science Operations Officer at NWS Duluth, passed on this interesting observation from the morning of 15 July 2026. The ongoing wildfires in Minnesota and Ontario, which we discussed in a blog post a few days ago, show up amazingly well when looking at the VIIRS Day/Night Band (DNB). The view below shows a string of brilliant blotches stretching across the middle of the image. Compare these bright spots to the city lights of metro Winnipeg (center of the left edge), Fargo (lower-left corner) or Duluth (western tip of Lake Superior). These are clearly some very large areas aflame.

To further convince ourselves that these are fires, let’s do a quick comparison to the 3.9 micron channel from the GOES-19 (GOES East) Advanced Baseline Imager. We use that latter channel for fire detection due to its extreme sensitivity to hot spots. Below is a slider that allows you to toggle back and forth between the low-earth orbiting VIIRS and the GOES ABI images. The dark spots indicate where active fires are, while the most intense fires have colors associated with them. You can see how the 3.9 micron fire spots line up perfectly with the bright spots in the middle of the DNB image, but the bright spots associated with the cities have no counterpart in the 3.9 micron imagery. In fact, you can even see an advantage of the VIIRS view: there are some locations right on the US/Canada border where smoke or clouds are obscuring the thermal signal of the fire but the visible brightness still shines through. The moon phase was new at this time and therefore was not illuminating the clouds.

As a bonus, here’s an animation for the GOES Fire Temperature RGB showing just how widespread (and hot!) these fires are. This product is available from SSEC’s Real Earth.

Starting now and continuing over the next few days, the smoke from these fires is going to impact large swaths of the continental United States. Stay tuned to the CIMSS Satellite Blog for more on that topic. Thanks, Patrick, for bringing this interesting satellite view to our attention!

View only this post Read Less