France has been experiencing record rainfall this winter. Last week (the final week of February 2026), the country ended a positively-biblical 40 day streak of consecutive days of rain, defined as an average of at least 1 mm of rain from all observing sites across continental France (about 80% of the size of Texas). The previous record was 32 days in 2023. Further details, for all francophone readers, are available from MétéoFrance. Some interesting highlights: since January 1st, Bordeaux received 321 mm (12.6 inches) of rain; they’d expect a total of 260 mm (10.2 inches) for the entire winter. Toulouse is in a similar situation, having received 203 mm (8.0 inches) so far in 2026 when they’d expect only 139 mm (5.5 inches) for the whole season. Across the country, this registered as the wettest February since 1959 with total accumulation more than twice the normal value.

Satellites are an excellent tool for monitoring not only the short-term weather conditions that lead to flooding, but also the longer-term extent of the floodwaters. The VIIRS Flood Mapping Product (quick guide here) provides one such look. Polar-orbiting satellites like those that host VIIRS are well-suited for flood observations because the higher resolution compared to geostationary enables a more detailed view of the extent of the flooding, while the slowly-evolving nature of floods means that the coarser temporal resolution of the polar-orbiting satellites is still adequate to capture the evolution of flooding events. The identification of floodwaters via satellite is conceptually very simple: surface water can be readily identified via satellite, and surface water in a location where water is not supposed to be implies a flood. There are some more challenging aspects to this, however, as clouds, surface snow, and terrain shadows can create regions of false positives and thus a flood detection algorithm needs to accommodate these and other issues.

Flooding products are available from SSEC’s Real Earth. Here is a link to the VIIRS 5 day composite flood product. This product is available once a day over the continents and selected island regions. The advantage of the 5 day composite is that it can help ameliorate the impact of clouds that would otherwise be in the way. The following animation shows the last two weeks of the VIIRS 5 day composite flood product over the Loire river valley in western France, a region famous for chateaux and vineyards. The colors on this product are representative of the fraction of a pixel that is covered by flooding waters: yellow is more than 40% and red is more than 80%. The rapid jump in the flood extent on the 24th is likely a result of the composite nature of this product with many of the preceding days featuring extensive cloud coverage.

Flood detection can be further enhanced with the inclusion of digital elevation models (DEMs). VIIRS observations can be used to calculate a percentage of a pixel that is covered by water. Assuming that the lowest portions of the pixel will be filled with water first, the higher resolution DEM can be used to downscale the macro-scale flooding information to more finely-detailed flood maps. SSEC is developing an experimental 30 m flood depth product that connects the areal coverage of the satellite to the DEM to produce highly-detailed observations of localized flooding. Here’s a sample image from that product, showing the Garonne and Dordogne rivers just downstream from Bordeaux, another famous winemaking region.

Different satellites can give an even more detailed view. Sentinel 2 is a European polar-orbiting satellite, analogous to the United States’ Landsat mission, designed for small-scale mapping and land classification. Among its bands include true-color red, green, and blue channels at 10 m spatial resolution. The downside of this high resolution is that the imager has a very narrow swath, and thus a given location doesn’t get an overpass every single day and clouds can further limit the number of usable views. An archive of Sentinel multispectral observations is available from the EU Copernicus Browser. Below is a slider that enables comparison between two different Sentinel 2 views of the Loire between Nantes and Angers: one from late January before significant flooding and another from late February when flooding is rampant. You can drag the bar back and forth to see how the environment changes between the two dates. It’s clear where the Loire has escaped its banks, and is especially evident in the middle of the image.

Finally, we can also look at the Normalized Difference Water Index (NDWI). Like its more famous cousin, the Normalized Difference Vegetation Index (NVDI), the NWDI takes the difference between the reflectance observed by two satellite bands and divides that by the sum of those bands. In this case, this is the difference between the green (560 nm) and the near IR (842 nm) bands. Water tends to be positive while vegetation and bare soil tends to be negative. Here is a comparison of the NDWI for the two dates, and here the impact of the flooding is obvious.

Fortunately, the amount of rainfall over France has lessened and waters appear to have started to recede.

View only this post Read Less

10-minute Full Disk scan GOES-18 (GOES-West) Infrared and Water Vapor images (above) showed the evolution of a mountain wave cloud south of the Brooks Range in northern Alaska on 01 March 2026. The coldest cloud-top infrared brightness temperatures were -61C (darker shades of red).

A toggle between GOES-18 Water Vapor and Infrared images at 1200 UTC (below) included a Topography image and 700 hPa wind barbs from the GFS model — which showed northerly winds flowing across the terrain of the Brooks Range, with the mountain wave cloud displaced to the south.

A toggle between GOES-18 Infrared and Water Vapor images at 1800 UTC (below) included an image of 500 hPa Vertical Velocity — the bulk of the mountain wave cloud was co-located with the zone of middle-tropospheric upward vertical velocity (brighter shades of green).

A Suomi-NPP VIIRS Infrared image (below) displayed the mountain wave cloud at 1402 UTC. This cloud appeared to play a role in keeping the surface air temperature at Fort Yukon (PFYU) significantly warmer than surrounding sites (by limiting radiational cooling across the Yukon Flats).

View only this post Read Less

5-minute CONUS Sector GOES-19 (GOES-East) Near-Infrared “Snow Ice” images along with Radar Reflectivity images (above) showed a band of clouds producing precipitation that was moving southward across the North Dakota/Minnesota border region on 27 February 2026. Moderate to heavy snow produced by this feature — along with blowing snow created by strong winds — prompted the issuance of a Snow Squall Warning at 2008 UTC that included a portion of Interstate 29 in the vicinity of Grand Forks (2011 UTC GOES-19 image). Peak wind gusts reached 51 kts (59 mph) at Grand Forks ND (KGFK), and the surface visibility was reduced to less than 1/4 mile with heavy snow at Crookston MN (KCKN).

Day Snow-Fog RGB images created using Geo2Grid (below) displayed the southward progression of this cloud band (brighter shades of white: 2011 UTC image) — as well as narrow northwest-to-southeast oriented horizontal convective rolls (a signature that highlights areas where significant blowing snow is occurring) in the wake of the cloud band responsible for the Snow Squall Warning. Areas of snow cover exhibited brighter shades of red.

View only this post Read Less

A narrow, but intense snow band formed over an area stretching from central Iowa to southwestern Wisconsin over the the night of 19 February into the morning of 20 February. Localized accumulations reached a foot in parts of northeast Iowa. This map from the La Crosse, WI, National Weather Service office shows the extent of this snowfall. (A lengthier discussion of this event from NWS La Crosse is available here).

A week later, this narrow corridor of snow is still around and still exerting influence on local weather conditions. Take a look at the following plot of surface temperatures from AWIPS in the early afternoon of 27 February 2026. North-central Iowa is experiencing temperatures in the lower 60s while parts of eastern Iowa are even reaching 70. However, there’s a cooler alley in northeastern iowa where temperatures are in the low-to-mid 40s.

Without any context, an analyst might opt to put some sort of synoptic-scale feature in northeast Iowa and assume that larger-scale flow is causing this pool of cooler temperatures. However, this being the CIMSS Satellite Blog, we’re going to recommend that you check out contemporaneous satellite imagery to verify what might be happening. Here’s the same figure, this time with the 0.64 micron highest resolution satellite imagery included as the base layer.

It’s clear that the surface snow cover is exerting a downward influence on temperatures. But why? There are two main reasons. The first is that snow has a high shortwave albedo: it reflects a substantial fraction of the light that shines on it. The albedo of new fallen snow can be as high as 90%, thoough for older snow such as this it’s much closer to 40 or 50%. Still, that is a substantial fraction of the incoming solar heating that is redirected away from the surface and cannot help to change the temperature. The other impact of snow is through phase changes. The solar energy that isn’t reflected from by the snow goes into changing its phase through melting or sublimation.

Here’s a pair of animations captured from AWIPS (thus the color dithering is not ideal). The first shows the true color RGB with the surface temperatures, while the second depicts the Day Cloud Phase Distinction RGB. Both are useful for identifying the location of the snow band surrounded by bare ground, while the latter animation helps discriminate between the snow (green) and ice clouds aloft (pink).

Finally, we should also take a look at the impact of the snow band on nighttime temperatures. While the earth’s surface generally has very high infrared emissivities, surface snow has a near perfect emissivity up to 0.99. This means that the snow is a very effective emitter of infrared radiation, more so than other surface types. Thus, we expect snow-covered ground to be colder at night than uncovered ground. This, of course, is on top of the fact that the snow-covered ground didn’t get as warm during the day as the other locations did, so it’s already starting from a lower temperature. This image shows the Band 13 (10.35 micron) imagery for 12:40 AM on the 27th. The color table has been adjusted to enhance contrast. Recall that for this window channel, in the absence of clouds we’re reading the brightness temperature of the surface. That surface is colder where the snow is, and the surface temperature observations back that up.

View only this post Read Less