Tropical Cyclone Idai: Before, During and After

As of the time of this writing, there is currently a humanitarian crisis in Mozambique caused by what was Tropical Cyclone Idai. Here’s the situation as of 25 March 2019.

Wikipedia actually has a pretty detailed history of Idai. Long story short, one of the worst (“worst” meaning large negative impact on humans) tropical cyclones in recorded history for the Southern Hemisphere formed just off the coast of Mozambique on 4 March 2019. It quickly headed inland as a tropical storm, where it dropped heavy rains on northern Mozambique and Malawi. Then, it turned back into the Mozambique Channel, headed for Madagascar, stopped, turned around, rapidly intensified, and then hit Mozambique a second time as a Category 2 cyclone. After making it on land a second time, it stalled out and dissipated, dropping more heavy rain in the process on central Mozambique and eastern Zimbabwe. Here is a long loop from Meteosat-8 showing much of the life cycle of Cyclone Idai as it appeared in the longwave infrared (IR).

Here’s a visible (True Color) loop from VIIRS that covers most of the month of March:

Animation of VIIRS True Color images from both S-NPP and NOAA-20 (1-25 March 2019)

Animation of VIIRS True Color images from both S-NPP and NOAA-20 (1-25 March 2019)

This loop has been reduced in resolution to half of its original size to save on file size. Even with only 2-3 images per day (since we combined both S-NPP and NOAA-20 images), you can still clearly see the cyclone over Mozambique early in the loop head out to sea and then turn around and hit Mozambique again, where it dumped heavy rain for several days.

But, I want to draw your attention to several of the images in that loop: the beginning, the middle, and the end. On 1 March 2019, NOAA-20 got a pretty clear view of central Mozambique:

NOAA-20 VIIRS True Color composite image (11:32 UTC, 1 March 2019)

NOAA-20 VIIRS True Color composite image (11:32 UTC, 1 March 2019)

We’ll call this the “Before” image – and this one is full resolution (750 m). (NOTE: You have to click on it show it at full resolution.) We can also look at the Natural Color RGB (also known as the Day Land Cloud RGB and about a dozen other names), which we can make with the high resolution imagery bands I-1, I-2 and I-3:

NOAA-20 VIIRS Natural Color RGB composite image (11:32 UTC, 1 March 2019)

NOAA-20 VIIRS Natural Color RGB composite image (11:32 UTC, 1 March 2019)

This is also at full resolution (375 m). (Again, only if you click on it.)

The worst of the flooding occurred with Idai’s second landfall on 14 March 2019, and both VIIRS got great views of Idai prior to landfall:

NOAA-20 Natural Color RGB composite image (10:47 UTC, 14 March 2019)

NOAA-20 Natural Color RGB composite image (10:47 UTC, 14 March 2019)

S-NPP Natural Color RGB composite image (11:38 UTC, 14 March 2019)

S-NPP Natural Color RGB composite image (11:38 UTC, 14 March 2019)

These images were taken ~50 min. apart. And, if you couldn’t already tell, they’re the high resolution Natural Color images. This is for two reasons: 1) who doesn’t want to see tropical cyclones at the highest resolution possible? and 2) the Natural Color RGB brings out details in the cloud structure you can’t see in True Color. As we’ve discussed before, Natural Color highlights ice clouds in a cyan color, while liquid clouds are nearly white. But, if you look closely in the above images, you will see lighter and darker cyan regions in the clouds above (or at the top of) the eyewall. This is due to differences in particle size. Larger ice particles appear more cyan, while smaller ice particles appear more white. (Of course, there is also some shadowing going on, which accounts for the darkest regions.)

Another thing to note is the first image comes from NOAA-20, which was to the east of Idai. This provides a great view of the sloped structure of the west side of the eyewall. (And, not much information on the east side of the eyewall.) The second image comes from Suomi-NPP, which was to the west of Idai, looking at the east side of the eyewall. The two satellites in tandem provide an almost 3D view of the clouds in the eyewall (separated by 50 minutes, of course).

Also, see that peninsula that is just to the west of the eyewall in the last two images? (Hint: you won’t see it unless you bring up the full resolution versions.) That’s where the city of Beira is (or was). Beira was home to half a million people, and was one of the major ports in Mozambique. It took a direct hit from the eyewall of Idai, which destroyed approximately 90% of the buildings there. Beira was also ground zero for the resulting flooding, and the pictures coming out are not pretty.

This is a good segue to talk about the images from the end of the loop. NOAA-20 captured a relatively cloud-free view of Mozambique on 25 March 2019:

NOAA-20 VIIRS True Color composite image (10:42 UTC, 25 March 2019)

NOAA-20 VIIRS True Color composite image (10:42 UTC, 25 March 2019)

NOAA-20 VIIRS Natural Color RGB composite image (10:47 UTC, 25 March 2019)

NOAA-20 VIIRS Natural Color RGB composite image (10:47 UTC, 25 March 2019)

These images were collected 10 days after landfall, and the flooding is still evident. Don’t believe me? Compare these “After” images with the “Before” images shown earlier (zoomed in on Beira):

Animation comparing NOAA-20 True Color RGB composite images from 1 March 2019 and 25 March 2019

Animation comparing NOAA-20 True Color RGB composite images from 1 March 2019 and 25 March 2019

Notice the fertile, green agricultural land surrounding Beira in the “before” image that is covered by brown floodwater in the “after” image. Just like what we saw in the pictures from Beira.

But, there’s a lot flooding that is not so easy to see in the True Color that shows up better in the Natural Color RGB:

Animation comparing NOAA-20 Natural Color RGB images from 1 March 2019 and 25 March 2019

Animation comparing NOAA-20 Natural Color RGB images from 1 March 2019 and 25 March 2019

Since this VIIRS Natural Color imagery has twice the resolution of True Color, this animation is too large for WordPress to play it automatically. You have to click on it to see the animation play.

We’ve talked before about differences between True Color and Natural Color when it comes to flooding, and this example shows it quite well. You see, True Color can miss flooding, because water is pretty transparent at visible wavelengths. If the water is clear, you can see through it and, from the perspective of VIIRS, you see the ground underneath the water (as long as the water is relatively shallow). If the water is muddy, like most of this flooding, it’s easier to see (since radiation reflects off the particles in the water), but it can look the same as the mud (or bare ground) that isn’t covered by water.

Natural Color uses longer wavelengths, where water is much more absorbing, so water appears nearly black. That’s why it is typically easier to see flooding against a background of non-flooded land in Natural Color than True Color. But, the flooding around Beira is so muddy, the high reflectivity in the visible channel (which is the blue component of the RGB) starts to win out, and the floodwater appears more blue than black.

We can prove it by looking at the individual bands that make up these RGB composites. Remember to click to play the animations for the I-bands:

Comparison of NOAA-20 channel I-1 images from 1 March and 25 March 2019

Comparison of NOAA-20 VIIRS channel I-1 (0.64 µm) images from 1 March and 25 March 2019

Comparison of NOAA-20 channel I-2 images from 1 March and 25 March 2019

Comparison of NOAA-20 VIIRS channel I-2 (0.87 µm) images from 1 March and 25 March 2019

Comparison of NOAA-20 channel I-3 images from 1 March and 25 March 2019

Comparison of NOAA-20 VIIRS channel I-3 (1.61 µm) images from 1 March and 25 March 2019

Note that the flooded areas look brighter in I-1 (thanks to the dirty water) and look darker in I-2 and I-3 (because they are less sensitive to the dirt in the water and more sensitive to the water itself).

The individual M-bands that comprise the True Color RGB, shown below, have been corrected for Rayleigh scattering and scaled the same as in the True Color images above:

Comparison of NOAA-20 channel M-3 images from 1 March and 25 March 2019

Comparison of NOAA-20 VIIRS channel M-3 (0.48 µm) images from 1 March and 25 March 2019

Comparison of NOAA-20 channel M-4 images from 1 March and 25 March 2019

Comparison of NOAA-20 VIIRS channel M-4 (0.55 µm) images from 1 March and 25 March 2019

Comparison of NOAA-20 channel M-5 images from 1 March and 25 March 2019

Comparison of NOAA-20 VIIRS channel M-5 (0.67 µm) images from 1 March and 25 March 2019

It is quite difficult to detect the flooding using the visible channels (M-3, M-4, M-5 and I-1) alone. But, the flooded areas are generally brighter in the “after” images. However, the water is easy to see in the shortwave IR channels (I-2, and I-3 along with M-7 and M-10, which were not shown).

Of course, this was a very long-winded way of looking at the flooding. We could have just used the JPSS Program’s official Flood Product made with VIIRS, created by researchers at George Mason University. Here is a three day composite image (composited to reduce the impact of clouds), covering 19-22 March 2019:

NOAA-20 VIIRS Flood Detection Product using a 3-day cloud-free composite (19-22 March 2019)

NOAA-20 VIIRS Flood Detection Product using a 3-day cloud-free composite (19-22 March 2019). Image courtesy S. Li (GMU).

Red and yellow areas show where flooding is detected. Gray areas are areas that were cloudy all three days. As an interesting side note, this product is validated against the Natural Color RGB. For more on this product, click here. If you want to know how much precipitation actually fell, here is a loop provided by NASA made with observations from GPM (Global Precipitation Measurement Mission):

You get bonus points if you can read the scale below the images. But, even without a magnifying glass, you can probably guess: it’s a lot of rain!

Ice, Ice, Baby

A winter storm moved through the Northeast U.S. over the weekend of 19-20 January 2019. This Nor’easter was a tricky one to forecast. Temperatures near the coast were expected to be near (or above) freezing. Temperatures inland were expected to be much colder. Liquid-equivalent precipitation, at least according to the GFS, was predicted to be in the 1-3 inch (25-75 mm) range the day before. This could easily convert to 1-2 feet (30-60 cm) of snow. The question on everyone’s mind: who gets the rain, who gets the snow, and who gets the “wintry mix”? The fates of ~40 million people hang in the balance. This is one of the situations that meteorologists live for!

The difference between 71°F and 74°F is virtually meaningless. The difference between 31°F and 34°F (with heavy precipitation, at least) is the difference between closing schools or staying open. It’s the difference between bringing out the plows or keeping them in the garage; paying overtime for power crews to keep the electricity flowing or just another work day; shutting down public transportation or life as usual.

Of course, the obvious follow-up question is: what is the “wintry mix” going to be? Rain mixed with snow? Sleet? Freezing rain? It doesn’t take much to change from one to the other, but there can be a big difference on the resulting impacts based on what ultimately falls from the sky.

So, what happened? Here’s an article that does a good job of explaining it. And, here are PDF files of the storm reports from National Weather Service Forecast Offices in Albany, Boston (actually in Norton, MA) and New York City (actually in Upton, NY). The synopsis: some places received ~1.5 inches (~38 mm) of rain, some places received ~11 inches (~30 cm) of snow and some places were coated in up to 0.6 inches (15 mm) of ice.

Of particular relevance here are the locations that received the ice. If you took the locations listed in the storm reports that had more than 0.1 inches (2.5 mm) of ice (at least the ones in Connecticut) and plotted them on a map, they match up quite well with this map of power outages that came from the article I linked to:

Map of power outages in Connecticut as a result of an ice storm (19-20 January 2019)

Map of power outages in Connecticut as a result of an ice storm (19-20 January 2019). Image courtesy Eversource/NBC Connecticut.

Now, compare that map with this VIIRS image from 22 January 2019 (after the clouds cleared out):

VIIRS channel I-3 image from NOAA-20, 17:09 UTC 22 January 2019

VIIRS channel I-3 image from NOAA-20, 17:09 UTC 22 January 2019

As always, you can click on the image to bring up the full resolution version. This is the high-resolution imagery band, I-3, centered at 1.6 µm from NOAA-20. Notice that very dark band stretching from northern New Jersey into northern Rhode Island? That’s where the greatest accumulation of ice was. Notice how well it matches up with the known power outages across Connecticut!

The ice-covered region appears dark at 1.6 µm because ice is very absorbing at this wavelength and, hence, it’s not very reflective. And, since it is cold, it doesn’t emit radiation at this wavelength either (at least, not in any significant amount). This is especially true for pure ice, as was observed here (particularly the second image), since there aren’t any impurities in the ice to reflect radiation back to the satellite. The absorbing nature of snow and ice compared with the reflective nature of liquid clouds is what earned this channel the nickname “Snow/Ice Band” (PDF).

At shorter wavelengths (less than ~ 1 µm), ice and snow are reflective. (Note how a coating of ice makes everything sparkle in the sunlight.) This makes it nearly impossible to tell where the ice accumulation was in True Color images:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 from NOAA-20, 17:09 UTC 22 January 2019

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 from NOAA-20, 17:09 UTC 22 January 2019

The Natural Color RGB (which the National Weather Service forecasters know as the Day Land Cloud RGB (PDF file)) includes the 1.6 µm band, which is what makes it useful for discriminating clouds from snow and ice. And, as expected, the region of ice accumulation does show up (although it is tempered by the highly reflective nature of snow and ice in the visible and “veggie” bands that make up the other components of the RGB):

VIIRS Natural Color RGB composite of channels, I-1, I-2 and I-3 from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS Natural Color RGB composite of channels, I-1, I-2 and I-3 from NOAA-20 (17:09 UTC, 22 January 2019)

Another RGB composite popular with forecasters is the Day Snow/Fog RGB (PDF file), where blue is related to the brightness temperature difference between 10.7 µm and 3.9 µm, green is the 1.6 µm reflectance, and red is the reflectance at 0.86 µm (the “veggie” band). This shows the region of ice even more clearly than the Natural Color RGB:

VIIRS Day Snow/Fog RGB composite of channels (I-5 - I-4), I-3 and I-2 from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS Day Snow/Fog RGB composite of channels (I-5 minus I-4), I-3 and I-2 from NOAA-20 (17:09 UTC, 22 January 2019)

Breaking things up into the individual components, we can see how the ice transitions from being reflective in the visible and near-infrared (near-IR) to absorbing in the shortwave-IR:

VIIRS high-resolution visible channel, I-1, from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS high-resolution visible channel, I-1 (0.64 µm), from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS high-resolution "veggie" channel, I-2, from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS high-resolution “veggie” channel, I-2 (0.86 µm), from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS channel M-8 from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS channel M-8 (1.24 µm) from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS channel M-11  from NOAA-20 (17:09 UTC, 22 January 2019)

VIIRS channel M-11 (2.25 µm) from NOAA-20 (17:09 UTC, 22 January 2019)

Of course, the 1.6 µm image was already shown, so I didn’t bother to repeat it. If you squint, you can even see a hint of the ice signature at 1.38 µm, the “Cirrus Band“, where most of the surface signal is blocked by water vapor absorption in the atmosphere:

VIIRS "cirrus" channel, M-9, from NOAA-20 (17:09 UTC 22 January 2019)

VIIRS “cirrus” channel, M-9 (1.38 µm), from NOAA-20 (17:09 UTC 22 January 2019)

If the ice had accumulated in southern New Jersey or Pennsylvania, though, it would not have shown up in this channel, since the air was too moist at this time to see all the way down to the surface. But, you can compare this image with the previous images to see why they call it the “cirrus band”, since the cirrus does show up much more clearly here.

So, mark this down as another use for VIIRS: detecting areas impacted by ice storms. And remember, even though ice storms may have a certain beauty, they are also dangerous. And, not just for the obvious reasons. This storm in particular came complete with ice missiles. So, for the love of everyone else on the road, scrape your car clean of ice before risking your life out there!

Rivers of Ice

Oh, Yakutsk! It has been a long time – 2012, to be exact – since we last spoke about you. It was a different time back then, with me still referring to the Natural Color RGB as “pseudo-true color”. (Now, most National Weather Service forecasters know it as the “Day Land Cloud RGB”). VIIRS was a only a baby with less than one year on the job. Back then, the area surrounding the “Coldest City on Earth” was on fire. This time, we return to talk about ice.

You see, rivers near the Coldest City on Earth freeze during the winter, as do most rivers at high latitudes. Places like the Northwest Territories, the Yukon, Alaska and Siberia use this to their advantage. Rivers that are frozen solid can make good roads, a fact that has often been overly dramatized for TV. Transporting heavy equipment may be better done on solid ice in the winter than on squishy, swampy tundra in the summer. But, that comes with a cost: ice roads only work during the winter.

In remote places like these, with few roads, rivers are the lifeblood of transportation – acting as roads during the winter and waterways for boats during the summer. But, what about the transition period that happens each spring and fall? Every year there is a period of time where it is too icy for boats and not icy enough for trucks. Monitoring for the autumn ice-up is an important task. And, perhaps it is more important to monitor for the spring break-up of the ice, since the break up period is often associated with ice jams and flooding.

We’ve covered the autumn ice up before (on our sister blog), but VIIRS recently captured a great view of the spring break up near Yakutsk, that will be our focus today.

We will start with the astonishing video captured by VIIRS’ geostationary cousin, the Advanced Himawari Imager (AHI) on Himawari-8 from 18 May 2018:

The big river flowing south to north in the center of the frame is the Lena River. (Yakutsk is on that river just south of the easternmost bend.) The second big river along the right side of the frame is the Aldan River, which turns to the west and flows into the Lena in the center of the frame.

Now that you are oriented, take a look at that video again in full screen mode. If you look closely, you will see a snake-like section of ice flowing from the Aldan into the Lena. This is exactly the kind of thing river forecasters are supposed to be watching for during the spring!

Of course, this is a geostationary satellite, which provides good temporal resolution, but not as good spatial resolution. The video is made from 1-km resolution imagery, but we are looking at high latitudes on an oblique angle, so the resolution is more like 3-4 km here. So, how does this look from the vantage point of VIIRS, which provides similar imagery at 375 m resolution? See for yourself:

(You will have to click on the image to get the animation to play.)

Animation of VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (18 May 2018)

Animation of VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (18 May 2018)

This animation includes both Suomi NPP and NOAA-20 VIIRS. That gives us ~50 min. temporal resolution to go with the sub-kilometer spatial resolution. Eagle-eyed viewers can see how the resolution changes over the course of the animation, as the rivers start out near the left edge of the VIIRS swath (~750 m resolution), then on subsequent orbits, the rivers are near nadir (~375 m resolution) and then on the right edge of the swath (~750 m resolution again). In any case, this is better spatial resolution than AHI can provide at this latitude.

One thing you can do with this animation is calculate how fast the ice was moving. I estimated the leading edge of the big “ice snake” moved about 59 pixels (22.3 km at 375 m resolution) during the 3 hour, 21 minute duration of the animation. That works out to an average speed of 6.7 km/hr (3.6 knots), which doesn’t seem unreasonable. Counting up pixels also indicates our big “ice snake” is at least 65 km long, and the Aldan River is nearly 3 km wide in its lower reaches when it meets the Lena River. That is in the neighborhood of 200 km2 of ice!

That much ice moving at 3 knots can do a lot of damage. Just look at what the ice on this much smaller river did to this bridge:

(Make sure you watch it all the way to the end!)

The Mystery Channel

I wrote the first post on this blog more than 5.5 years ago. Since then, I have covered a multitude of instances where VIIRS imagery has helped us learn about the world we live on. But, during that time there has been one channel on VIIRS that has never been mentioned. Not once. And, what may be even more surprising is that this channel is not featured on any of the next generation geostationary satellites. It’s not on the GOES-R Program’s ABI, not on Himawari’s AHI, not on the upcoming Meteosat Third Generation FCI. Those with photographic memories will know exactly which channel I’m talking about. The rest of you will just have to guess, or go back through the archives and use the process of elimination to figure it out.

So, is this channel useless? Why is it on VIIRS, but not ABI? Which one is it? The suspense is killing me! I can’t answer that second question, but I can definitely answer the third and give some insights to #1. (The short answer to #1 is “No” – otherwise we wouldn’t be here.) But, to do this, we have to remember why Lake Mille Lacs disappeared earlier this year. It might also be good to remember our earlier posts on Greenland, because that is the location of our most recent mystery.

We begin with the view of Greenland from GOES-16 back at the end of July 2017:

This video covers the period of time from 0700 UTC 27 July to 2345 UTC 28 July. If you follow this blog, you already know that this the “Natural Color” RGB composite, which in GOES-16 ABI terms is made of bands 2 (0.64 µm), 3 (0.86 µm) and 5 (1.61 µm). Notice the whitish coloration over the central portion of Greenland. This is the feature of interest.

We know from experience (and earlier blog posts) that snow and ice are not very reflective at 1.6 µm, which is why it takes on that cyan appearance in Natural Color imagery. Whitish colors are indicative of liquid clouds. But, the feature of interest doesn’t appear to move over this two day period. (If you look closely, it does appear to shrink a little, though.) It’s hard to believe a cloud could be that stationary over a two day period.

Let’s isolate the 1.6 µm band by itself to see if we can tell what’s going on:

Shortly after the first sunrise, you can see a patch of liquid clouds over the ice that quickly dissipate, leaving our feature of interest exposed. Clouds appear again near the first sunset, and late in the second day (28 July). The feature of interest isn’t as bright as those clouds, but is brighter than the rest of the ice and snow on Greenland.

At shorter wavelengths, nearly all of Greenland is bright, so our feature of interest isn’t as noticeable. Here’s the 0.86 µm band from ABI:


 
But, it shows up at the two longer shortwave IR bands. Here’s the 2.25 µm band:


 
The same is true for 3.9 µm, but I won’t waste time showing it.

So, what is going on? What is our feature of interest?

Well, the problem is, Greenland is way off on the limb from the perspective of GOES-16’s current location. Perhaps we need a better view from something that passes directly overhead of Greenland. Hmmm. What could that be?

This is a VIIRS blog after all, so I think you know the answer to my rhetorical question. Let’s start with our good old friend, Natural Color, which we should all be familiar with:

S-NPP VIIRS Natural Color RGB composite of bands M-5, M-7 and M-10 (14:40 UTC 27 July 2017)

S-NPP VIIRS Natural Color RGB composite of bands M-5, M-7 and M-10 (14:40 UTC 27 July 2017)

You can tell by the shadows cast where the clouds are, even if they are a similar color to the background of snow and ice on Greenland. But, the feature of interest isn’t very obvious. There appears to be an area of lighter cyan over the central portions of the ice sheet, but it definitely doesn’t look like a cloud. Let’s break it up into single channels, like we did with ABI, starting with M-7 (0.86 µm):

S-NPP VIIRS channel M-7 (14:40 UTC 27 July 2017)

S-NPP VIIRS channel M-7 (14:40 UTC 27 July 2017)

Again, it’s all bright. How about M-10 (1.61 µm)?

S-NPP VIIRS channel M-10 (14:40 UTC 27 July 2017)

S-NPP VIIRS channel M-10 (14:40 UTC 27 July 2017)

Now, Greenland appears all dark. For completeness, let’s look at M-11 (2.25 µm):

S-NPP VIIRS channel M-11 (14:40 UTC 27 July 2017)

S-NPP VIIRS channel M-11 (14:40 UTC 27 July 2017)

It’s subtle, but you can see a hint of brightening over the south-central portion of the ice sheet. (In case you’re wondering why it looks so much darker in VIIRS than ABI, it’s because our visible and near-IR GOES-16 imagery uses “square root scaling” by default. In image processing terms, it’s the same as a gamma correction of 2. The VIIRS images don’t have that.) Now, for the ace up my sleeve – the one channel that has never appeared before on this blog:

S-NPP VIIRS channel M-8 (14:40 UTC 27 July 2017)

S-NPP VIIRS channel M-8 (14:40 UTC 27 July 2017)

This is M-8, centered at 1.24 µm. Its primary use is listed in the JPSS Program literature as “cloud particle size.” Based on reading the documentation for the cloud products, it appears M-8 is used operationally only as a backup for M-5 (0.67 µm) in the cloud optical thickness and effective particle size retrievals under certain conditions, or when M-5 fails to converge on solution. One of those conditions is the retrieval of cloud properties over snow and ice. As we shall see, however, M-8 is very good at determining the properties of the snow and ice itself.

M-8 shows quite clearly the bright central portion of Greenland (our feature of interest) surrounded by dark at the edges of the ice sheet (even without any gamma correction). Snow-free areas appear brighter than the edge of the ice sheet because, much like M-7/0.86 µm, vegetation is also highly reflective at 1.24 µm.

This example shows what we’ve long known. Snow and ice are highly reflective in the visible (and very near IR) portions of the electromagnetic spectrum. In the short- and mid-wave IR, snow and ice are absorbing and cold. This means they don’t emit or reflect much radiation at these wavelengths. That’s why they appear dark at 1.61 and 2.25 µm. M-8 straddles the boundary of these regions as exemplified by this graph:

Reflectance spectra of snow

Reflectance spectra of snow. The highlighted portion shows the bandwidth of VIIRS channel M-8.

The information in this graph comes from the ASTER Spectral Library created by NASA. Note that the reflectance of snow in M-8 is highly variable and a function of the snow grain size. This may explain why the central portion of Greenland’s ice sheet appears so bright, while the edges are so dark in M-8. Another explanation is that, much like in Minnesota, snow melt causes a drop in reflectance. Slush just isn’t as reflective as fresh snow, and M-8 is highly sensitive to this.

The last week in July was a very warm one for Greenland. The capitol, Nuuk, recorded highs in the 60s (°F), or upper-teens (°C), peaking at 71°F (22°C) on 29 July 2017. Normal for that time of year is 52°F (11°C).

Since Greenland is pretty far north, we can take advantage of the multiple VIIRS overpasses per day and really capture this snowmelt:

Animation of daytime VIIRS M-8 images (27-29 July 2017)

Animation of daytime VIIRS M-8 images (27-29 July 2017)

This animation, which you may have to click on to get it to play, covers the three day period 27-29 July 2017. Here’s it is obvious what impact the heat wave is having on Greenland’s ice and snow. Our “feature of interest” really shrinks over this period of time.

In early August, the snow and ice start to recover and become more reflective again. Here’s an extended animation that includes the relatively clear days of 17 July, 20 July and the entire period from 30 July to 3 August 2017:

Animation of VIIRS M-8 (17 July - 3 August 2017)*

Animation of VIIRS M-8 (17 July – 3 August 2017)*

Our “feature of interest” is unmelted snow/ice on Greenland’s ice sheet.

Now, this is the VIIRS Imagery Team Blog. We can do a better job of highlighting this snowmelt by combining it with other channels in an RGB composite. One way is to replace M-7 with M-8 in the Natural Color RGB:

Animation of VIIRS Natural Color imagery composites of channels, M-5, M-8 and M-10 (17 July - 3 August 2017)*

Animation of VIIRS Natural Color imagery composites of channels, M-5, M-8 and M-10 (17 July – 3 August 2017)*

Fresh, fine snow has the cyan color we’re all familiar with, but now coarse snow and melting snow are a deeper, more vivid blue color.

Another option takes a page out of the EUMETSAT Snow playbook. Here’s one with M-8 as the blue component, M-7 as the green component and M-5 as the red component:

Animation of VIIRS RGB composite using channels, M-8, M-7 and M-5 (17 July - 3 August 2017)*

Animation of VIIRS RGB composite using channels, M-8, M-7 and M-5 (17 July – 3 August 2017)*

Now the fresh, fine snow is pale yellow, while the coarse snow and snowmelt are a darker yellow-orange. The question is: which one do you like better?

So, I have now talked about every band on VIIRS. And, I learned that the last time I looked at melting on Greenland, I should have been looking at M-8 from the very beginning.

On the Disappearance of Lake Mille Lacs

Two weeks ago, one of Minnesota’s 10,000 lakes disappeared, leaving them with only 9,999. And, it wasn’t a small one, either. It was the state’s second largest inland lake. But, this is not like Goose Lake, which actually did dry up. The lake in question simply became temporarily invisible. So, no need to panic, fishing and boating enthusiasts. But, as you’ll see, the term “invisible” can be just as ambiguous as the term “lake”.

Let’s start with the fact that Minnesota doesn’t have 10,000 lakes. Their slogan is a lie! Depending on how you define a lake, Minnesota has 21,871, or 15,291, or 11,842. But, Wisconsin might have more (or less) and likes to argue with Minnesota about that fact. Michigan might have way more (62,798) or way less (6,537). And, they all pale in comparison to the number of lakes in Alaska. Here is an article that explains the situation nicely.

With that out of the way, today’s story comes from “current GOES” and what one colleague noticed during a cursory examination of GOES Imager images. Here’s the GOES-13 visible image from 19:30 UTC 27 January 2017:

GOES-13 visible image, taken 19:30 UTC 27 January 2017

GOES-13 visible image, taken 19:30 UTC 27 January 2017

Compare that with the visible image from 19:15 UTC 2 February 2017:

GOES-13 visible image, taken 19:15 UTC 2 February 2017

GOES-13 visible image, taken 19:15 UTC 2 February 2017

Notice anything different between the two images over Minnesota? No? Then let’s flip back-and-forth between the two, with a giant, red arrow pointing to the area in question:

Animation of the above images

Animation of the above images. The red arrow points to Lake Mille Lacs.

The red arrow is pointing to the location of Lake Mille Lacs. You might know it as Mille Lacs Lake. (Either way, it’s name is redundant; “Mille Lacs” is French for “Thousand Lakes,” making it Thousand Lakes Lake.) As the above images show, on 27 January 2017 Lake Mille Lacs was not visible in the GOES image. On 2 February 2017, it was. They both look like clear days, so what happened? Why did Lake Mille Lacs disappear?

As I said before, the lake didn’t dry up. It simply became temporarily invisible. But, this requires a discussion about what it means to be “visible”. Lake Mille Lacs shows up in the image from 2 February 2017 because it appears brighter than the surrounding land. That’s because the lake is covered with snow. Aren’t the surrounding land areas also covered with snow? Yes. However, the surrounding lands also have trees which obscure the snow and shade the background surface, which is why forested areas appear darker even when there is snow.

That leads to this question: why does the lake appear darker on 27 January? Because it rained the week before. Want proof? Look at the almanac for Brainerd (NW of Lake Mille Lacs) for the period of 18-22 January 2017. Every day made it above freezing along with several days of rain. Much of the snow melted (including the snow on the lake). Want more proof? Here’s a video taken on the lake from 20 January 2017. See how Minnesotans drive around on frozen lakes – even in the rain? And, see how wet and slushy the surface of the ice is? This makes it appear darker than when there is fresh snow on top. If you’ve ever seen a pile of slush, you know it’s not bright white, but a dull gray color. The less reflective slush on the lake reduced the apparent brightness down to the level of the surrounding woodlands. That’s why the lake appeared to disappear.

Now, this is “current GOES” imagery. We can do better with VIIRS, since we have more channels to play with. And, as we all know, GOES-R successfully launched back in November 2016 and is now in orbit as GOES-16. This satellite has the first Advanced Baseline Imager (ABI) in space. The ABI has many of the same channels as VIIRS, so the following discussion applies to both instruments. “New” GOES will have up to 500 m resolution in the visible, which is much closer to VIIRS (375 m) than “current” GOES (1 km). That’s another thing to think about when we talk about what is visible and what isn’t.

Here are the VIIRS high-resolution visible (I-1) images that correspond to the GOES images above:

VIIRS high-resolution visible (I-1) image, taken 19:35 UTC 27 January 2017

VIIRS high-resolution visible (I-1) image, taken 19:35 UTC 27 January 2017

VIIRS high-resolution visible (I-1) image, taken 19:22 UTC 2 February 2017

VIIRS high-resolution visible (I-1) image, taken 19:22 UTC 2 February 2017

Although, we should probably focus on Minnesota. Here are the cropped images side-by-side:

Comparison between VIIRS high-resolution visible (I-1) images

Comparison between VIIRS high-resolution visible (I-1) images

Remember: you can click on any image to bring up the full resolution version.

Although Lake Mille Lacs is just barely visible in the image from 27 January, it’s much easier to see on 2 February. So, we get the same story from VIIRS that we got with GOES, which is good. That means we don’t have a major fault of a multi-million dollar satellite. It’s a “fault” of the radiative properties of slush, combined with the low resolution of the GOES images above.

Keep your eyes also on the largest inland lake in Minnesota: Red Lake. The Siamese twins of Upper and Lower Red Lake didn’t get as much rain as Lake Mille Lacs and its snow never fully melted, so its appearance doesn’t change much between the two images.

The GOES Imager also has a longwave infrared (IR) channel, and a mid-wave IR channel similar to VIIRS. Since the goal of this is not to compare GOES to VIIRS, but to show how these lakes appear at different wavelengths, we’ll stick to the VIIRS images. Here are the high-resolution VIIRS longwave IR images from the same times:

Comparison of VIIRS high-resolution longwave IR (I-5) images

Comparison of VIIRS high-resolution longwave IR (I-5) images

In both images, the lakes are nearly invisible! This is because the longwave IR is primarily sensitive to temperature changes, and the slush is nearly the same temperature as the background land surface. With no temperature contrast to key on, the lake looks just like the surrounding land. Although, if you zoom in and squint, you might say that Lake Mille Lacs is actually more visible in the image from 27 January. 27 January was a warmer day (click back on that Brainerd almanac), and the surrounding land warmed up more than the slushy ice on the lake. 2 February was much colder on the lake and the land. But, let this be a lesson that, just because the lake doesn’t show up, it doesn’t mean the lake doesn’t exist!

Something interesting happens when you look at the mid-wave IR. All the lakes are visible, and take on a similar brightness, no matter how slushy they are:

Comparison of VIIRS high-resolution mid-wave IR (I-4) images

Comparison of VIIRS high-resolution mid-wave IR (I-4) images

In this wavelength range, both reflection of solar energy and thermal emission are important. Snow, ice and slush are not reflective and they are cold, making the lakes appear darker than the surrounding land. The fact that the land surrounding Lake Mille Lacs and Red Lake is darker on 2 February than it is on 27 January is further proof that it was a colder day with more snow on the ground.

Here’s where we get to the advantage of VIIRS (and, soon, GOES-16): it has more channels in the shortwave and near-IR. The 1.6 µm “snow and ice” band has a lot of uses, and I expect it will be a popular channel on the ABI. Here’s what the high-resolution channel looks like from VIIRS:

Comparison of VIIRS high-resolution near-IR (I-3) images

Comparison of VIIRS high-resolution near-IR (I-3) images

Compare these with the visible images above. Now, the reverse is true: Lake Mille Lacs is easier to see in the first image than the second! You can’t call it invisible at all on 27 January! The presence of liquid water makes the slush very absorbing – more than even ice and snow – so it appears nearly black. In fact, it’s hard to tell the difference between the slushy ice-covered Lake Mille Lacs, and the open waters of Lake Superior, which has no ice or slush on it. On 2 February, we see the fresh layer of snow on Lake Mille Lacs has increased the lake’s reflectivity, but it’s still slightly darker than the surrounding snow covered land. This is for two reasons: snow and ice are absorbing at 1.6 µm and the surrounding woodlands are more reflective.

Here’s a better comparison between the “visible” and the “snow and ice” bands:

Comparison of VIIRS I-1 and I-3 images (animation)

Comparison of VIIRS I-1 and I-3 images (animation)

You’ll have to click on the image to see it animate between the two.

Here’s an animation showing all five high-resolution bands on VIIRS for the two days:

Comparison of VIIRS high-resolution imagery channels (animation)

Comparison of VIIRS high-resolution imagery channels (animation)

Again, you have to click on it to see it animate.

Now, we can combine channels into RGB composites that highlight the snow and ice. We’ve discussed several RGB composites for snow detection before. And, we have been looking at the Natural Color RGB for a long time. This composite combines the high-resolution bands I-1 (0.64 µm), I-2 (0.86 µm) and I-3 (1.6 µm) as the blue, green and red components of the image, respectively. Here’s what it looks like for these two days:

Comparison of VIIRS Natural Color RGB composites

Comparison of VIIRS Natural Color RGB composites using high-resolution imagery bands

Lake Mille Lacs is visible on both days – first because it’s darker than the surroundings, then because it’s brighter. This composite demonstrates how vegetation can obscure the surface snow – it appears more brown in deciduous forests (and bare fields with no snow) and green in coniferous areas. But, the important point is that the wetter the snow and slush, the darker it appears. The fresher the snow, the brighter cyan color it has.

This is exaggerated in the “Snow RGB” that combines moderate resolution bands M-11 (2.25 µm), M-10 (1.6 µm) and M-7 (0.86 µm):

Comparison of VIIRS "Snow RGB" composites of channels M-11, M-10 and M-7

Comparison of VIIRS “Snow RGB” composites of channels M-11, M-10 and M-7

M-11 (2.25 µm) is sold as a “cloud particle size” band, but it also helps with snow and ice detection (and fires). The presence of water in melting snow enhances the darkening at 2.25 µm. In this RGB, that means melting snow appears more red, while fresh snow appears more pink. The slush on Lake Mille Lacs appears very dark – almost as dark as Lake Superior – so a Minnesotan might be forgiven if they see the image from 27 January and decide not to drive out on the lake to go ice fishing because they think the ice isn’t there.

Of course, VIIRS also gives us the True Color RGB – the most intuitive RGB composite – that combines the blue-, green- and red-wavelength visible bands: M-3 (0.48 µm), M-4 (0.55 µm) and M-5 (0.67 µm). If you’re curious, here’s what that looks like:

Comparison of VIIRS True Color RGB composite images

Comparison of VIIRS True Color RGB composite images

The slush on Lake Mille Lacs looks just like dirty slush and the fresh snow looks just like snow. (As it should!)

So, the second biggest lake in Minnesota never disappeared – it just changed its surface properties. And, it will always be “visible” to VIIRS in one channel or another – unless it’s cloudy (or it completely dries up).