Single-Purpose Flour

Posted on November 16, 2016 by Curtis_Seaman

Think of a snowflake. What happens when that snowflake hits the ground? Now, picture other snowflakes – millions of them – all hitting the ground and piling up on top of each other, crushing our first poor snowflake. Skiers love to talk (and dream) about “fresh powder.” But, what happens when the “powder” isn’t so fresh?

Those delicate, little snow crystals we imagine (or look at directly, if we click on links included in the text) undergo a transformation as soon as they hit the ground. Compression from the weight of the snow above, plus the occasional partial thaw and re-freeze cycle (when temperatures are in the right range), breaks up the snow flakes and converts the 6-pointed crystals into more circular grains of snow. As more and more snow accumulates on top, the air in between the individual snowflakes/grains (which is what helps make it a good insulator) gets squeezed out, making the snow more dense. If enough time passes and enough snow accumulates, individual snow grains can fuse together. These bonded snow grains are called “névé.” If this extra-dense snow can survive a whole summer without melting, then a second winter of this compaction and compression will squeeze out more air and fuse more snow grains, creating the more dense “firn.” After 20 or 30 years of this, what once was a collection of fragile snowflakes becomes a nearly solid mass of ice that we call a “glacier.” Glaciers can be made up of grains that are several inches in length.

But, you don’t need to hear me say it (or read me write it), you can watch a short video where a guy in a thick Scottish accent explains it. (Did you notice his first sentence was a lie? Snow is made of frozen water, so glaciers are made of frozen water, since they are made of snow. I think what he means is that glaciers aren’t formed the same way as a hockey rink, but the way he said it is technically incorrect.) At the end of the video, the narrator hints at why we are looking at glaciers today: glaciers have the power to grind down solid rock.

When a glacier forms on a non-level surface, gravity acts on it, pulling it down the slope. This mass of ice and friction from the motion acts like sandpaper on the underlying rock, converting the rock into a fine powder known as “glacial flour” or, simply, “rock flour.” In the spring and summer months, the meltwater from the glacier collects this glacial flour and transports it downstream, where it may be deposited on the river’s banks. During dry periods, it doesn’t take much wind to loft these fine particles of rock into the air, creating a unique type of dust storm that is not uncommon in Alaska. One that can be seen by satellites.

And, wouldn’t you know it, a significant event occurred at the end of October. Take a look at this VIIRS True Color image from 23 October 2016:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 (21:24 UTC 23 October 2016)

See the big plume of dust over the Gulf of Alaska? Here’s a zoomed in version:

Zoomed in version of above image.

That plume of dust is coming from the Copper River delta. The Copper River is fed by a number of glaciers in Wrangell-St. Elias National Park, plus a few in the Chugach Mountains so it is full of glacial sediment and rock flour (as evidenced by this photo). And, it’s amazingly full of salmon. (How do they see where they’re going when they head back to spawn? And, that water can’t be easy for them to breathe.)

Notice also that we have the perfect set-up for a glacial flour dust event on the Copper River. You can see a low-pressure circulation over the Gulf of Alaska in the above picture, plus we have a cold, Arctic high over the Interior shown in this analysis from the Weather Prediction Center. For those of you familiar with Alaska, note that temperatures were some 30 °F warmer during the last week in October in Cordova (on the coast) than they were in Glennallen (along the river ~150 miles inland). That cold, dense, high-pressure air over the interior of Alaska is going to seek out the warmer, less dense, low-pressure air over the ocean – on the other side of the mountains – and the easiest route to take is the Copper River valley. The air being funneled into that single valley creates high winds, which loft the glacial flour from the river banks into the atmosphere.

Now, depending on your preferences, you might think that the dust shows up better in the Natural Color RGB composite:

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (21:24 UTC 23 October 2016).

But, everyone should agree that the dust is even easier to see the following day:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 (21:01 UTC 24 October 2016)

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (21:01 UTC 24 October 2016)

You can also see a few more plumes start to show up to the southeast, closer to Yakutat.

Since Alaska is far enough north, we get more than one daytime overpass every day. Here’s the same scene on the very next orbit, about a 100 minutes later:

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (22:42 UTC 24 October 2016)

Notice that the dust plume appears darker. What is going on? This is a consequence of the fact that glacial flour, like many aerosol particles, has a tendency to preferentially scatter sunlight in the “forward” direction. At the time of the first orbit (21:01 UTC), both the sun and the dust plume are on the left side of the satellite. The sunlight scatters off the dust in the same (2-dimensional) direction it was traveling and hits the VIIRS detectors. In the second orbit (22:42 UTC), the dust plume is now to the right of the satellite, but the sun is to the left. In this case, forward scattering takes the sunlight off to the east, away from the VIIRS detectors. With less backward scattering, the plume appears darker. This has a bigger impact on the Natural Color imagery, because the Natural Color RGB uses longer wavelength channels where forward scattering is more prevalent. Here’s the True Color image from the second orbit:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 (22:42 UTC 24 October 2016)

The plume is a little darker than the first orbit, but not by as much as in the Natural Color imagery. Here are animations to show that:

Animation of VIIRS True Color images (24 October 2016)

Animation of VIIRS Natural Color images (24 October 2016)

There are many other interesting details you can see in these animations. For one, you can see turbid waters along the coast in the True Color images that move with the tides and currents. These features are absent in the Natural Color because the ocean is not as reflective at these longer wavelengths. You can also see the shadows cast by the mountains that move with the sun. Some of the mountains seem to change their appearance because VIIRS is viewing them from a different side.

The dust plumes were even more impressive on 25 October 2016, making this a three-day event. The same discussion applies:

VIIRS True Color composite of channels M-3, M-4 and M-5 (20:43 UTC 25 October 2016)

VIIRS True Color composite of channels M-3, M-4 and M-5 (22:26 UTC 25 October 2016)

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (20:43 UTC 25 October 2016)

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (22:26 UTC 25 October 2016)

Full disclosure, yours truly drove through a glacial flour dust storm along the Delta River on the north side of the Alaska Range back in 2015. Even though it was only about a mile wide, visibility was reduced to only a few hundred yards beyond the hood of my car. It felt as dangerous as driving through any fog. The dust event shown here was not a hazard to drivers, since it was out over the ocean, but it was a hazard to fisherman. Being in a boat near one of these river deltas means dealing with high winds and high waves. To forecasters, these dust plumes provide information about the wind on clear days (when cloud-track wind algorithms are no help), which is useful in a state with very few surface observing sites to take advantage of.

The last remaining issue for the day is one of terminology. You see, “glacial flour dust storm” is a mouthful, and acronyms aren’t always the best solution. (GFDS, anyone?) “Haboob” covers desert dust. “SAL” or “bruma seca” covers Saharan dust specifically. So, what should we call these dust events? Something along the lines of “rock flour”, only more proactive! And, Dusty McDustface is right out!

Remote Islands V: St. Helena and Ascension

Posted on May 2, 2016 by Curtis_Seaman

You may have missed it in the news, but history was made last week:

A plane landed! Wow!

But, that’s not any old plane – that’s the first commercial airliner to land on St. Helena Island, which just completed the construction of their very first airport. That means there may be no more commercial sailing to this tiny island.

People prone to seasickness may be cheering the news. People afraid of flying might not. Did you notice it took three attempts to land that plane in the video above? The first pass was getting everything all lined up with no intention of landing. The landing gear wasn’t even down. The second – which looked like a roller coaster – was waived off due to the heavy crosswinds. The third time was the charm. However, it was such a shaky first landing, they’ve postponed the official opening of the airport.

So, where is St. Helena (pronounced Ha-LEEN-a), anyway? And why should I care?

Well, to answer the first question, it’s somewhere in this image:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 (12:45 UTC 26 April 2016).

Did you find it? To help you with your bearings, Africa is just outside this VIIRS swath on the right side of the image. Two hints: click on the image to bring up the full resolution version. St. Helena is just northwest of the center of the image. It’s the only island in the image not covered by clouds. Fun fact: every island within this VIIRS swath is part of the British Overseas Territory of St. Helena, Ascension and Tristan da Cunha. We already looked more closely at Tristan da Cunha, so let’s take a look at the other two.

We can get a higher resolution look if we use the I-band Natural Color RGB composite:

VIIRS Natural Color RGB composite of channels I-01, I-02 and I-03 (12:45 UTC 26 April 2016).

Notice the island appears green in the center, surrounded by a ring of brown – just the way it looks on a really high resolution satellite image. VIIRS has the resolution to pick this out!

As for why you should care, I don’t know if I can answer that. If your first thought is to ask that question, you probably don’t care. But, there are a few interesting things to note about St. Helena (besides its new airport):

– It was once an important stopping point for ships sailing from Europe to India in search of spices. At least, until the Suez Canal opened.

– It later became a prison, housing those who fought against the British government and lost, including Napoleon Bonaparte, Dinuzulu, King of the Zulu Nation, and POWs from the Boer War.

– Along with Ascension Island, St. Helena helped inspire the modern environmental movement. And it was here that the first large scale experiments in weather modification took place. (Not counting rain dances, of course.)

After witnessing the effect of deforestation on the island in the late-1700s and early-1800s, it was believed that re-foresting would help keep moisture on the island, which would lead to more clouds and more rainfall. Ascension Island, which was essentially a barren wasteland when first discovered, was also planted with trees, creating it’s Green Mountain, which is clearly visible on very high resolution satellites.

Speaking of Ascension Island – where is that located? In the first image above, showing most of the Southern Atlantic, Ascension is near the upper left corner. It’s hard to see because it is covered by clouds. Just follow the 8 °S latitude line in from the left edge of the image.

Here it is at high resolution during a clear day:

VIIRS Natural Color RGB composite of channels I-01, I-02, and I-03 (14:03 UTC 20 April 2016).

If you look closely, you’ll see that there is a small cloud or two right over Green Mountain, so maybe the efforts of the early environmentalists paid off!

For completeness, Tristan da Cunha is in the lower left of the True Color image I posted at the top. While it is covered by clouds, you can tell it’s there because it is creating its own waves. Here it is on the next orbit, where it is closer to satellite nadir:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5 (15:24 UTC 26 April 2016).

If I’ve inspired you to visit these islands, ask the government to give me a commission. But, seriously, don’t forget to say “Hi!” to Jonathan. Or see the many other plants and animals that are found nowhere else on Earth.

UPDATE (16 October 2017): Reuters has reported that the airport is now officially open to commercial flights (only a year and half after I wrote the original blog post)!

The Great Flood of 2015

Posted on January 5, 2016 by Curtis_Seaman

As we begin 2016, struggling to get back into the swing of things at work and vowing not to overeat or over-drink ever again, it’s appropriate to bid farewell to 2015 – not just for all the weird weather events that we covered on this blog over the year, but also for the weird, wacky weather that ruined many people’s holidays. I’m not sure of the exact number, but this article mentions 43 weather-related fatalities in the U.S. in the second half of December. Let’s see, between 23-30 December 2015, there were:

– 77 tornadoes (including 38 on the 23^rd and 18 on the 27^th);

– Parts of New Mexico and west Texas got over 2 ft (60 cm) of snow from a blizzard that created drifts upwards of 10 ft (3 m) on the 27^th;

– Record warmth was observed in the Northeast before and during Christmas and the site of Snowvember went until 18 December before the first measurable snow of the season;

– Chicago received almost 2″ of sleet (48 mm) on the 29^th when any accumulation of sleet is quite rare;

– And – what will be our focus here – St. Louis received over 3-months-worth of precipitation in three days (26-28 December), from a storm that flooded a large area of Missouri, Illinois and Arkansas. In fact, the St. Louis area had the wettest December on record, right after having the 7th wettest November on record, which put it over the top for wettest calendar year on record. Current estimates place 31 fatalities at the hands of this flooding, which caused the Mississippi River to reach its highest crest since the Great Flood of 1993.

What kind of satellite imager would VIIRS be if it couldn’t detect massive flooding on the largest river in North America? (Hint: not a very useful one. Or, a less useful one, if you’re not into hyperbole.) Hey, if it works in Paraguay, it works here – or it isn’t science!

I shouldn’t have to prove that the Natural Color RGB is useful for detecting flooding (since I have done it many, many, many, many, many, many times before), so we can go right to the imagery. Here’s what the Midwest looked like on 13 November 2015 – before the flooding began:

VIIRS Natural Color RGB composite of channels I-1, I-2, and I-3 (19:02 UTC 13 November 2015).

And, here’s what the same area looked like on New Year’s Day:

VIIRS Natural Color RGB composite of channels I-1, I-2, and I-3 (18:45 UTC 1 January 2016).

Notice anything different? This is actually the reverse of the last time we played “Spot the Differences” – we’re looking for where water is now that wasn’t there before, instead of searching for bare ground that used to have water on it.

Of course, the first thing to notice is the large area of snow covering Iowa, Nebraska and northwest Missouri that wasn’t there back in November. Next, we have more clouds over the southern and northern parts of the scene. Those are the easy differences to spot. Now look for the Missouri River in eastern Missouri, the Arkansas River in Arkansas, the Illinois River in Illinois, the Indiana River in Indiana… Wait! There is no Indiana River. I fooled you! (Although, there are rivers in Indiana that are flooded.)

The most significant areas of flooding are in northeast Arkansas and the “Bootheel” of Missouri (which I think looks more like a toe or a claw than a heel), and the Mississippi River along the border of Tennessee shows signs of significant flooding as well. (If only it were the Tennessee River!) Here’s a before and after comparison, zoomed in on that part of the region:

You may have to refresh the page to get this to work right.

There’s a lot more water in the image from 1 January 2016 than there was back in November 2015! Since we are looking at the high-resolution Imagery bands, our quick-and-dirty estimate of water volumes still applies like it did for California’s drought: multiply the number of water-filled pixels by the depth (in feet) of the flooding, and by 100 acres to get the floodwater volume in acre-feet. Then multiply that by 325,852 gallons per acre-foot to get the volume in gallons. Even though this estimate is not exact, you can see how the gallons of floodwater add up. And, if you live in California, you can dream of seeing that much water! If you live in Missouri and can think of an economical way to transport this water to California, you’d be rich.

Now, see how many other areas of flooding you can find when you compare the two images in animation form:

Click to view an animation of VIIRS Natural Color RGB images from 13 November 2015 and 1 January 2016.

You will have to click on the image to see the animation. You can click on the image again to see it in full resolution (with most web browsers).

One thing you might notice is that some of the floodwaters appear more blue than black. Take a look at the Arkansas River in particular. As we discussed with the Rio Paraná and Rio Paraguay, this is due to the increased sediment that increases the albedo of the water at visible wavelengths. In other places the floodwaters are shallow enough that VIIRS can see the ground underneath – again making the water appear more blue in this RGB composite.

Wouldn’t it be nice to identify areas of flooding without having to play a “Spot the Differences” game? Maybe something that would automatically detect flooded areas? Well, you’re in luck:

VIIRS-based Flood Map (18:48 UTC 1 January 2016). Image courtesy S. Li (GMU).

This image is an example of the VIIRS-based flood detection product being developed by the JPSS Program’s River Ice and Flooding Initiative. This initiative is a collaboration between university-based researchers and NOAA forecasters who use products like these to help save lives. Thanks to S. Li for developing the product for and providing the image!

If you want to know what the flooding looks like from the ground, here is a nice video. Or, you can look at some pictures here.

As a final note, the American Meteorological Society is holding its Annual Meeting in New Orleans next week. This event will be held at the Convention Center – right on the bank of the Mississippi River – right at the time the river is forecast to crest from these floodwaters. The world’s largest gathering of weather enthusiasts might be directly impacted by this flood. Let’s hope no one has to swim their way to any poster sessions or keynote speeches! (I don’t think local residents want to deal with any flooding, either.)

Goose Lake is Gone (Again)

Posted on July 24, 2015 by Curtis_Seaman

We’ve covered mysteries before on this website. Well, here’s one from 150 years ago:

The emigrants, coming west on the Applegate Trail to Oregon in the 1870s, were puzzled. The trail was, of course, a seemingly unending set of wagon-wheel ruts stretching from the jumping-off points in the Midwest over deserts and mountains and all sorts of obstacles that seemed insurmountable, but weren’t.

But this one seemed impossible. Had the wagons before them really plunged directly into the enormous lake that lay before them? The ruts led directly into the water, and there was no sign of them having come out again.

It was miles across – the other side lay almost invisible on the horizon, much too far to float a caulked wagon. And yes, it was deep – far too deep to ford.

There was nothing for it but a trip around the lake, since the western sky lay on the other side. And so, around they went – making a detour of something like 100 miles.

On the other side, they found the wagon ruts again. They emerged from the water and headed on westward toward the Cascades. Once arrived at the West Coast, none of the previous emigrants knew anything about any lake there.

Was it aliens who came down to Earth to put a lake where there was none before? Did the earlier emigrants have covered wagon submarine technology (and very short term memories)? Maybe it was a very localized, very short-term Ice Age – a glacier snuck down from the Cascades and into the valley in the middle of the night and then melted without anyone noticing. What about that?

SPOILER ALERT: None of those theories is true. Anyone who would come up with these ridiculous ideas should be ashamed of themselves. Oh, wait – I came up with them. Hmmm. What I meant to say is: those are all good theories that are worthy of scientific exploration. Unfortunately, VIIRS wasn’t around in the 1870s. Plus, this mystery has already been solved. As our source explains:

It remained a mystery until, several years later, a drought struck and the lake dried up again.

What we’re talking about is Goose Lake, which is at times the largest lake that’s at least partially in Oregon. (In terms of surface area, not volume.) It’s right on the border between Oregon and California. When Goose Lake is at its fullest, it has a surface area of 147 square miles (380 km²), but it’s only 26 ft (8 m) deep. Maybe, if the emigrants weren’t so cowardly, they could have walked across it (although they might have gotten stuck in the mud). It would have saved 100 miles of extra walking (although they might have gotten stuck in the mud).

As you are probably well aware, California and Oregon are under a long-lasting, extreme drought. So, if you live near Goose Lake, it’s probably no surprise that the lake has dried up again. And, since this is 2015, VIIRS can tell us something about it this time.

Have you ever played one of those “spot the differences” games? (Don’t play them at work, or you’ll never get anything done.) Well, here’s a “spot the differences” game you can play at work – at least if your work involves detecting evidence of drought.

Here’s what Goose Lake looked like three years ago, according to VIIRS Natural Color imagery:

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (20:40 UTC 15 July 2012)

Note that it’s not as dark in color as the other lakes because it is so shallow. Now, here’s the same scene just last week:

VIIRS Natural Color RGB composite of channels I-1, I-2 and I-3 (21:40 UTC 16 July 2015)

Notice anything different? Now, for this spot-the-differences game, we’re going to ignore clouds, because they are always going to be different between the two images, difficult to count, and irrelevant to this discussion. (Except that clouds can obscure the view of a lake and can cast shadows that look like lakes.)

Since I labelled Goose Lake on those images, you have no excuse for not spotting that difference. Besides, if you can’t see that 147 square miles of lake surface are missing from the second image, you have no hope to see any of the other differences.

I counted at least 20 lakes or reservoirs that are present in the 2012 image that have dried up and vanished in the 2015 image. Plus, there are about as many lakes or reservoirs that have noticeably shrunk since 2012. Can you spot them all? Can you see more than I did?

After you’ve declared yourself done, compare your results with mine:

Comparison of the above VIIRS Natural Color images of Goose Lake.

As always, click on it to see the full resolution image. I’ve marked with red arrows those lakes that are visible in the 2012 image that are not visible in the 2015 image. Yellow arrows indicate the lake has lost surface area (but not totally vanished) between 2012 and 2015. And, there are a few spots that look like surface water visible in the 2015 image that are not present in 2012 – I’ve marked those with green arrows. There are a couple of lakes visible in the 2012 image that are covered by clouds in the 2015 image. Those are left unmarked. I’ve also labelled a burn scar left over from a pretty big wildfire in south-central Oregon visible in 2012 that has since disappeared. That’s the main non-lake, non-cloud related difference between the two images.

Most notably, Upper Alkali Lake (southeast of Goose Lake) dried up, which you should have noticed without me pointing it out. Drews Reservoir on the northwest side of Goose Lake in Oregon appears to have dried up, as does New Year Lake right across the border from Upper Alkali Lake in Nevada. Thompson Reservoir (the northernmost red arrow) looks bone dry and Gerber Reservoir (west of Drews Reservoir) has very little water left. The eastern half of Clear Lake Reservoir is now empty and the western half is significantly reduced in size. Three big reservoirs (lakes) on the southern edge of the image have also lost quite a bit of water (Trinity Lake, Shasta Lake and Eagle Lake).

Even if you don’t care that a bunch of salty, alkaline lakes in rural Jefferson (as they might prefer you to call it) have dried up, you should care about the reservoirs. And not just for the boating and other water recreation activities, which are now hazardous. When towns run out of water, prime agricultural land lays fallow, and Tom Selleck gets in trouble with the law, you know things are serious.

The reservoirs closer to central California are down quite a bit as well, and these impact a lot of people. Use your honed-in spot-the-difference skills in these VIIRS I-2 (0.865 µm) images from the same dates and times as the above images:

VIIRS I-2 image (20:40 UTC 15 July 2012)

VIIRS I-2 image (21:40 UTC 16 July 2015)

I-2 is one of the components of the Natural Color imagery (the green component). What makes it good for this purpose is that land and, particularly, vegetation are highly reflective at this wavelength, so they appear bright. Water is absorbing, so it appears black (or nearly so if the water’s dirty or shallow). It also has 375 m resolution at nadir. If you click to the full resolution versions of the above images, you can see that most of the reservoirs have lost quite a bit of surface area between 2012 and 2015.

If you’re too lazy, or have poor eyesight, click on this image below to better compare the two images:

Comparison of VIIRS I-2 images from the same dates and times as above

One more point that needs to be made: 375 m resolution at nadir is good for weather satellites like VIIRS, but the fact that you can see the loss of water in these images is testimony to how bad this drought is!

As you may or may not know, the resolution of VIIRS in these images degrades from 375 m at nadir to 750 m at the edge of the swath. As a reasonable approximation, that’s means each pixel is a quarter mile to a half mile wide. That means each pixel of missing water represents between 40 and 160 acres. We’ll say 100 acres, given that these images were taken roughly halfway between nadir and edge of scan. If the water was only 1 foot deep in these pixels, that would be a loss of 100 acre-feet. That’s 32.5 million gallons of water. (By the way, the average household uses between 0.5 and 1 acre-foot per year in water.)

Multiply the number of pixels that have lost water by 100 to get the area in acres. Multiply that by the average depth of the water lost to get the volume in acre-feet. And then multiply that by 325,852 gallons per acre-foot and that’s a lot of gallons of missing water!

(In case you’re interested, this PDF document says the average depth of Goose Lake is 8 ft. At 147 sq. mi. of surface area, that’s 245 billion gallons of water gone, give or take.)

Germany’s Magic Sparkle

Posted on March 16, 2015 by Curtis_Seaman

You may or may not have heard that a small town in Italy received 100 inches (250 cm; 2.5 m; 8⅓ feet; 8 x 10^-17 parsecs) of snow in 18 hours just last week (5 March 2015). That’s a lot of snow! It’s more than what fell on İnebolu, Turkey back in the beginning of January. But, something else happened that week that is much more interesting.

All you skiers are asking, “What could be more interesting than 100 inches of fresh powder?” And all you weather-weenies are asking, “What could be more interesting than being buried under a monster snowstorm? I mean, that makes Buffalo look like the Atacama Desert!” The answer: well, you’ll have to read the rest of this post. Besides, VIIRS is incapable of measuring snow depth. (Visible and infrared wavelengths just don’t give you that kind of information.) So, looking at VIIRS imagery of that event isn’t that informative.

This is (or was, until I looked into it in more detail) another mystery. Not a spooky, middle-of-the-night mystery, but one out in broad daylight. (We can thus automatically rule out vampires.)

It started with a comparison between “True Color” and “Natural Color” images over Germany from 9 March 2015:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 11:54 UTC 9 March 2015.

VIIRS Natural Color RGB composite of channels M-5, M-7 and M-10, taken 11:54 UTC 9 March 2015.

The point was to show, once again, how the Natural Color RGB composite can be used to differentiate snow from low clouds. That’s when I noticed it. Bright pixels (some white, some orange, some yellow, some peach-colored) in the Natural Color image, mostly over Bavaria. (Remember, you can click on the images, then click again, to see them in full resolution.) Thinking they might be fires, I plotted up our very own Fire Temperature RGB:

VIIRS Fire Temperature RGB composite of channels M-10, M-11 and M-12 from 11:54 UTC 9 March 2015.

I’ve gone ahead and drawn a white box around the area of interest. Let’s zoom in on that area for these (and future) images.

VIIRS True Color RGB (11:54 UTC 9 March 2015). Zoomed in and cropped to highlight the area of interest.

VIIRS Natural Color RGB (11:54 UTC 9 March 2015). Zoomed in and cropped to highlight the area of interest.

VIIRS Fire Temperature RGB (11:54 UTC 9 March 2015). Zoomed in and cropped to highlight the area of interest.

Now, these spots really show up. But, they’re not fires! Fires show up red, orange, yellow or white in the Fire Temperature composite (which is one of the benefits of it). They don’t appear pink or pastel blue. What the heck is going on?

Now, wait! Go back to the True Color image and look at it at full resolution. There are white spots right where the pastel pixels show up in the Fire Temperature image. (I didn’t notice initially, because white spots could be cloud, or snow, or sunglint.) This is another piece of evidence that suggests we’re not looking at fires.

For a fire to show up in True Color images, it would have to be about as hot as the surface of the sun and cover a significant portion of a 750-m pixel. Terrestrial fires don’t typically get that big or hot on the scale needed for VIIRS to see them at visible wavelengths. Now, fires don’t have to be that hot to show up in Natural Color images, but even then they appear red. Not white or peach-colored. If a fire was big enough and hot enough to show up in a True Color image, it would certainly show up in the high-resolution infrared (IR) channel (I-05, 11.45 µm), but it doesn’t:

VIIRS high-resolution IR (I-05) image (11:54 UTC 9 March 2015).

You might be fooled, however, if you looked at the mid-wave IR (I-04, 3.7 µm) where these do look like hot spots:

VIIRS high-resolution midwave-IR (I-04) image (11:54 UTC 9 March 2015).

What’s more amazing is I was able to see these bright spots all the way down to channel M-1 (0.412 µm), the shortest wavelength channel on VIIRS:

VIIRS “deep blue” visible (M-1) image (11:54 UTC 9 March 2015).

So, what do we know? Bright spots appear in all the bands where solar reflection contributes to the total radiance (except M-6 and M-9). I checked. (They don’t show up in M-6 [0.75 µm], because that channel is designed to saturate under any solar reflection so everything over land looks bright. They don’t show up in M-9 [1.38 µm] because solar radiation in that band is absorbed by water vapor and never makes it to the surface.) Hot spots do not coincide with these bright spots in the longer wavelength IR channels (above 4 µm).

What reflects a lot of radiation in the visible and near-IR but does not emit a lot in the longwave IR? Solar panels. That’s the answer to the mystery. VIIRS was able to see solar radiation reflecting off of a whole bunch of solar panels. That is the source of Germany’s “magic sparkle”.

Don’t believe me? First off, Germany is a world leader when it comes to producing electricity from solar panels. Solar farms (or “solar parks” auf Deutsch) are common – particularly in Bavaria, which produces the most solar power per capita of any German state.

Second: I was able to link specific solar parks with the bright spots in the above images using this website. (Not all of those solar parks show up in VIIRS, though. I’ll get to that.) And these solar parks can get quite big. Let’s take a look at a couple of average-sized solar parks on Google Maps: here and here. The brightest spot in the VIIRS Fire Temperature image (near 49° N, 11° E) matches up with this solar park, which is almost perfectly aligned with the VIIRS scans and perpendicular to the satellite track.

Third: it’s not just solar parks. A lot of people and businesses have solar panels on their roofs. Zoom in on Pfeffenhausen, and try to count the number of solar panels you see on buildings.

One more thing: if you think solar panels don’t reflect a lot of sunlight, you’re wrong. Solar power plants have been known reflect so much light they instantly incinerate birds*. (*This is not exactly true. See the update below.)

Another important detail is that all of the bright spots visible in the VIIRS images are a few degrees (in terms of satellite viewing angle) to the west of nadir. Given where the sun is in the sky this time of year (early March) and this time of day (noon) at this latitude (48° to 50° N), a lot of these solar panels are in the perfect position to reflect sunlight up to the satellite. But, not all of them. Some solar panels track the sun and move throughout the day. Other panels are fixed in place and don’t move. Only the solar panels in the right orientation relative to the satellite and the sun will be visible to VIIRS.

At these latitudes during the day, the sun is always to south and slightly to the west of the satellite. For the most part, solar panels to the east of the satellite will reflect light away from the satellite, which is why you don’t see any of those. If the panel is pointed too close to the horizon, or too close to zenith (or the sun is too high or too low in the sky), the sunlight will be reflected behind or ahead of the satellite and won’t be seen. You could say that this “sparkle” is actually another form of glint, like sun glint or moon glint – only this is “solar panel glint”.

Here’s a Natural Color image from the very next day (10 March 2015), when the satellite was a little bit further east and overhead a little bit earlier in the day:

VIIRS Natural Color RGB composite of channels M-5, M-7 and M-10 from 11:35 UTC 10 March 2015.

Notice the half-dozen-or-so bright spots over the Czech Republic. These are just west of the satellite track and in the same position relative to satellite and sun. (The bright spot near the borders of Austria and Slovakia matches up with this solar farm.) The bright spots over Germany are gone because they no longer line up with the sun and satellite geometry.

As for the pastel colors in the Natural Color and Fire Temperature RGBs, those are related to the proportional surface area of the solar panels relative to the size of each pixel as well as the background reflectivity of the ground surrounding the solar panels. The bright spots do generally appear more white in the high-resolution version of the Natural Color RGB from 9 March:

VIIRS high-resolution Natural Color (I-01, I-02, I-03) RGB image (11:54 UTC 9 March 2015).

See, we learned something today. Germany sparkles with green electricity and VIIRS can see it!

UPDATES (17 March 2015): Thanks to feedback from Renate B., who grew up in Bavaria and currently owns solar panels, we have this additional information: there is a push to add solar panels onto church roofs throughout Bavaria, since they tend to be the tallest buildings in town (not shaded by anything) and are typically positioned facing east, so the south-facing roof slopes are ideal for collecting sunlight. The hurdle is that churches are protected historical buildings that people don’t want to be damaged. Also, it’s not a coincidence that many solar parks have their solar panels facing southeast (and align with the VIIRS scan direction). They are more efficient at producing electricity in the morning, when the temperatures are lower, than they are in the afternoon when the panels are warmer. They face southeast to better capture the morning sun.

Also, to clarify (as pointed out by Ed S.): the solar power plant that incinerates birds generates electricity from a different mechanism than the photovoltaic (PV) arrays seen in these images from Germany. PV arrays (aka solar parks) convert direct sunlight to electricity. The “bird incinerator” uses a large array of mirrors to focus sunlight on a tower filled with water. The focused sunlight heats the water until it boils, generating steam that powers a turbine. Solar parks and solar panels on houses and churches do not cause birds to burst into flames.

JPSS (SNPP and NOAA-20)

VIIRS Imagery and Visualization Team Blog

Tag Archives: Natural Color

Single-Purpose Flour

Remote Islands V: St. Helena and Ascension

The Great Flood of 2015

Goose Lake is Gone (Again)

Germany’s Magic Sparkle