Meteorological satellites don’t just observe the current weather. They can also keep track of the weather’s aftermath. A great example of this is with flooding. Floods are significant economic disasters, causing upwards of $180 billion in damage (almost 1% of the US GDP) annually to structures, agriculture, and transportation infrastructure. Unlike many weather events that are over in a matter of hours to days, a flood can linger for weeks or even months.
It’s critical for emergency managers, governmental leaders, and others to be able to monitor the extent and magnitude of floods in order to plan for responses like sandbagging operations or evacuations. Of course, this is a task that can be managed by satellite. Detecting floods by satellite is conceptually quite simple: different satellite bands can be used to detect surface water. We know what pixels are supposed to be covered by surface water when water levels are normal. If we start to see other pixels start to show water coverage, then we know we have a flood.
Let’s take a look at a currently ongoing flooding event on the Illinois River in, well, Illinois. The Illinois River is one of the United States’s most important inland waterways. With over 12 billion ton-miles of shipping annually, it ranks as the third most-heavily used river behind only the Mississippi and Ohio Rivers. It’s a vital link in the global agricultural supply chain as it provides an easy way for midwest farmers to export their corn and soybeans to either the Great Lakes or down the Mississippi to the Gulf.
A map of the Illinois’s drainage basin shows that it envelops much of central Illinois with some tributaries extending into southeast Wisconsin and northwest Illinois. Frequent storms over the past month have caused significantly higher than normal rainfall totals. For the city of Peoria, right on the banks of the Illinois in the middle of the state, the observed rainfall was 134% higher than normal for the month of June, with over 5 inches extra rain than is typically seen in that month. Here’s a time series of the cumulative precipitation over the last month (4 June – 3 July 2026) at Peoria. The brown line represents the normal accumulation over that time while the green line shows the observations. It’s clear from this plot that over the last month, Peoria has experienced more than twice its typical amount of rain. Fortunately, there’s been very little additional rain over the last week.
All of that water has to go somewhere, and we can see it in the river gauges operated by the US Geological Survey. Here’s a plot of the last seven days of river levels at Beardstown, Illinois. The thick blue line shows the observed river stage, which peaked on 27th and 28th of June 2026 before starting a slow decline. Still, as late as July 3 2026, the river level is still several feet above flood stage and at least ten feet above normal.
For some time the VIIRS Flood Mapping package has been available from the Community Satellite Processing Package (CSPP) group at CIMSS, allowing detection of flooding using data from low-earth orbiting (LEO) satellite instruments. The software is based on a NOAA flood detection algorithm that was developed by Sanmei Li. Polar orbiting satellites, of course, have fine-scaled spatial resolution allowing for the detection of relatively small flooding events. However, the coarse temporal resolution of LEO satellites can prove to be a problem, especially when clouds are present. If a cloud is in the way the one or two times a day that VIIRS passes over a point, then the infrared bands used for flood detection are blocked by the cloud and it’s impossible to see the state of the surface.
Recently the ABI Flood Mapping package was released by the CSPP Geo group, allowing detection of flooding using data from geostationary satellite instruments. That package is also based on the NOAA algorithm developed by Sanmei Li. This enables flood detection at a far better temporal resolution than is possible with their existing LEO software by taking advantage of the rapidly refreshing views made possible by a geostationary platform. CSPP has users around the globe, and they can run the flood detection on arriving files and generate a composite view every hour. The higher cadence of geostationary data (every 10 minutes in the normal instrument mode) increases the probability that at least one scan will have a clear sky view, thus increasing the overall robustness of the product. Both the LEO and GEO algorithms rely on both infrared and visibly imagery, and thus are processed during daytime hours only.
Let’s take a look at how the CSPP implementation of geostationary flood detection is capturing this ongoing event. This animation shows five hours of CSPP flood detection, from 1200-1600 UTC on 3 July 2026. Blue represents pixels that are typically covered by water. The yellow, orange, and red pixels, however, represent an increasing fraction of normally dry pixels being covered by water. The gray pixels are where clouds obscured the surface preventing its characteristics from being measured. While this was a clear morning over the Illinois, the Mississippi (along the western edge of the animation) experienced more cloud coverage and thus surface water was harder to detect. This illustrates a clear advantage of the hourly product, in that you’re more likely to have clear skies at some point and can thus get more frequent updates than you would with a LEO satellite.
Let’s take a moment to compare the CSPP GEO output to the VIIRS-based CSPP LEO output. Here’s an animation of that latter product over the last several days. Due to the infrequency of the VIIRS overpasses, the flooding product is calculated on a daily cadence. It’s clear that the LEO product has an advantage in spatial resolution: the pixels are much smaller and thus more detail is readily identifiable. However, clouds cover parts of the river on multiple days because they happened to be in the way just at the time that the NOAA LEO satellites were passing overhead.
By contrast, the daily CSPP GEO flooding analysis has many more frames to select from and can therefore produce a composite view of a particular day that is far less likely to be impacted by cloud cover. Here’s the same period as the LEO product above, but from the GEO view. Note there’s much less interference from clouds. Some times persistent stratus may cover a location for an entire 24 hour period and flood values aren’t available. However, many more of the gaps from the LEO product get filled in with the GEO one.
The VIIRS and ABI flooding products are complementary. The polar orbiting detection from VIIRS not only has the higher spatial resolution when compared to ABI, it is also far more useful in Alaska and other northerly regions where geostationary resolution is degraded and polar orbiting overpasses are much more frequent; in fact, the VIIRS flooding product is available globally. However, the geostationary product has far more frequent updates in the midlatitudes. End users can benefit from consulting both products when identifying and monitoring flooding situations. Future CSPP releases are expected to include a joint VIIRS/ABI product.
The VIIRS Flood Mapping package is available to download from the CSPP LEO website, and the ABI Flood Mapping package is available to download from the CSPP Geo website. In addition, flooding products from CSPP, both LEO and GEO, can be plotted using the SSEC RealEarth online visualization platform. For GEO, look to the “Products and Layers” menu on the left-hand side. Scroll to “Flood Detection – GEO”, then select “River Flood ABI-Daily” or “River Flood ABI-Hourly” depending on your needs. For LEO, in the same “Products and Layers” menu, scroll just a bit further to “Flood Detection – Global” and select “River Flood 1 Day VIIRS Composite.”
Thanks to Graeme Martin, David Hoese, and Kathy Strabala for their insights in the operation and use of the flood detection packagers.
View only this post Read Less
