When Canada Looks Like China

No, I’m not talking about Chinatown in Vancouver. Or Chinatown in Toronto. Or any other Chinatown in Canada. I’m talking about this. Or, more exactly, this. Poor air quality is making it difficult to breathe in Canada and elsewhere.

Unlike the situation in China, you can’t really blame the Canadians for their poor air quality. (Unless, of course, some serial arsonist is wreaking havoc unfettered.) You see, it has been an active fire season in western Canada, to put it mildly. Here’s a not-so-mild way to put it. That article, from 3 July 2014, put the number of fires in the Northwest Territories alone at 123, with most of them caused by lightning. But, after a check of the Northwest Territories’ Live Fire Map on 30 July 2014 it looks like there are more than that:

"Live Fire Map" from NWTFire, acquired 17:00 UTC 30 July 2014

"Live Fire Map" from NWTFire, acquired 17:00 UTC 30 July 2014. This is a static image, not an interative map.

I estimated 160-170 fires in that image (assuming I didn’t double count or miss any). How many fires can you count?

At one point earlier in July, it was estimated that battling the fires was costing $1 million per day! The fires have been impacting power plants, causing power outages, impacting cellular and Internet service, closing the few roads that exist that far north, and doubling the number of respiratory illnesses reported in Yellowknife, the territory’s capital.

It’s no secret that this area is sparsely populated. At last count, the territory had roughly 41,000 residents in 1.3 million km2. (Fun fact: the Northwest Territories used to make up 75% of the land area of Canada. It has since been split up among 5 provinces and into two other territories. With the formation of Nunavut in 1999, it was reduced to being only twice the size of Texas.) If so few people live there, why should we care if they have a few fires?

If you are so heartless as to ask that question, you are also short-sighted and selfish. For one, I already explained the damage that the fires are doing. For two, fires like these impact more than just the immediate area and more than just Canada. Let me explain that but, first, let me show you the fires themselves – as seen by VIIRS – over the course of the last month.

Animation of VIIRS Fire Temperature RGB images 24 June - 25 July 2014

Animation of VIIRS Fire Temperature RGB images 24 June - 25 July 2014

You will have to click on the above image, then on the “933×700” link below the banner to see the animation at full resolution. It is 15 MB, so it may take a while to load if you have limited bandwidth. What you are looking at is the Fire Temperature RGB in the area of Great Slave Lake, the area hardest hit by this fire season. There are a lot of fires visible over the course of the month!

See how the larger fires spread out? They look like the large scale version of an individual flame spreading out on a piece of paper. (Don’t try to replicate it at home. I don’t want you catching your house on fire!) Of course, the spread of the fires is dependent on the winds, humidity, moisture content in the vegetation, and the firefighters – if they’re doing their job.

Now, these weren’t the only fires in Canada during this time. Check out this Fire Temperature RGB image from 15 July 2014 and see how many (rather large) fires there are in British Columbia and Saskatchewan:

VIIRS Fire Temperature RGB composite of channels M-10, M-11 and M-12, taken 21:08 UTC 15 July 2014

VIIRS Fire Temperature RGB composite of channels M-10, M-11 and M-12, taken 21:08 UTC 15 July 2014

Make sure to click through to the full resolution version. I counted 9 large fires in British Columbia, 1 in Alberta (partially obscured by clouds) and 6 in Saskatchewan. If you look closely, you might also spot 3 small fires in Washington plus more small fires in Oregon. (“Small” here is compared to the fires in Canada.)

Now, all these fires means there must be smoke and, because VIIRS has channels in the blue and green portions of the visible spectrum, we can see the smoke clearly. This is one of the benefits of the True Color RGB (in addition to what we discussed last time). If I tried to create another animation, like I did above, showing the extent of the smoke plumes it would be so large it might crash the Internet. Instead, here are some of the highlights (or low-lights, depending on your point of view) from the last month.

On 6 July 2014, the smoke is largely confined to the area around Great Slave Lake:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:35 UTC 6 July 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:35 UTC 6 July 2014

The very next day (7 July 2014) the smoke is blown down into Alberta and Saskatchewan (almost as far south as Calgary and Saskatoon):

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:16 UTC 7 July 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:16 UTC 7 July 2014

One day later (8 July 2014) smoke is visible down into Montana, North Dakota and beyond the edge of the image in South Dakota (a distance of over 2000 km [1200 miles] from the source!):

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 19:57 UTC 8 July 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 19:57 UTC 8 July 2014

 

On the 12th of July, you could see a single smoke plume stretching from Great Slave Lake all the way into southwestern Manitoba (plus smoke over British Columbia from their fires):

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:23 UTC 12 July 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:23 UTC 12 July 2014

When the fires really get going in British Columbia a few days later, the smoke covers most of western Canada. On 15 July 2014, smoke is visible from the state of Washington to the southern reaches of Nunavut and Hudson Bay:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 19:27 UTC 15 July 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 19:27 UTC 15 July 2014

One day later (16 July 2014), and it appears that smoke covers 2/3 of Alberta, nearly all of Saskatchewan, all of western Manitoba, southern Nunavut, southeastern Northwest Territories, and most of Montana and North Dakota. There is also smoke over Washington, Oregon and northern Idaho:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:48 UTC 16 July 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 20:48 UTC 16 July 2014

A quick estimate puts the area of smoke in the above image at 2.5 million km2, which is roughly a third the size of the contiguous 48 states!

With renewed activity in the fires in the Northwest Territories last week, the smoke was still going strong over Canada, impacting Churchill, Manitoba (home of polar bears and beluga whales):

VIIRS True Color RGB composite of channels M-4, M-4 and M-5, taken 20:17 UTC 23 July 2014

VIIRS True Color RGB composite of channels M-4, M-4 and M-5, taken 20:17 UTC 23 July 2014

I guess if the melting polar ice caps don’t kill off the polar bears, they can still get cancer from all this smoke. Maybe the “world’s saddest polar bear” will want to stay in Argentina.

I should add that some of my colleagues at CIRA and I have sensitive noses and were able to smell smoke right here in town (Fort Collins, Colorado) earlier this month. Plus, there were a few smoky/hazy sunsets. (Although it should be clarified that we don’t know if it was from the fires in Canada or the fires in Washington and Oregon. There weren’t any fires in Colorado at the time.) Nevertheless, the areal coverage and extent of the smoke from fires like these is immense, and can have impacts thousands of miles away from the source. And, it’s all carbon entering our atmosphere.

 

UPDATE (8/1/2014): Colleagues at CIMSS put together this image combining two orbits of data over North America from yesterday (31 July 2014), where you can see smoke stretching from Nunavut all the way down to Indiana, Ohio and West Virginia. There may even be some smoke over Kentucky and Tennessee. Witnesses at CIMSS reported very hazy skies across southern Wisconsin as a result.

Hell Froze Over (and the Great Lakes, too)

This has been some kind of winter. The media has focused a lot of attention on the super-scary “Polar Vortex” even though it isn’t that scary or that rare. (I wonder if Hollywood will make it the subject of the next big horror movie in time for Halloween.) Many parts of Alaska have been warmer than Georgia, with Lake Clark National Park tying the all-time Alaskan record high temperature for January (62 °F) on 27 January 2014. (Atlanta’s high on that date was only 58 °F.) Sacramento, California broke their all-time January record high temperature, reaching 79 °F three days earlier. In fact, many parts of California had record warmth in January, while everyone on the East Coast was much colder than average. Reading this article made me think of an old joke about statisticians: a statistician is someone who would say: if your feet are stuck in a freezer and your head is stuck in the oven, you are, on average, quite comfortable.

One consequence of the cold air in the eastern United States is that Hell froze over. No, not the Gates of Hell in Turkmenistan. This time I’m talking about Hell, Michigan. Hell is a nice, little town whose residents never get tired of people telling that joke.

It has been so cold in the region around Hell that the Great Lakes are approaching a record for highest percentage of surface area covered by ice. This article mentions some of the benefits of having ice-covered Lakes, including: less lake-effect snow, more sunshine and less evaporation from the Lakes, which would keep lake levels from dropping. Although, that is at the cost of getting ships stuck in the ice, and reducing the temperature-moderating effects of the Lakes, which allows for colder temperatures on their leeward side.

This article (and many other articles I found) uses MODIS “True Color” images to highlight the extent of the ice. Why don’t they show any VIIRS images? Well, I’m here to rectify that.

First off, I can copy all those MODIS images and show the “True Color” RGB composite from VIIRS:

VIIRS "True Color" RGB composite of channels M-3, M-4 and M-5, taken 17:27 UTC 11 February 2014

VIIRS "True Color" RGB composite of channels M-3, M-4 and M-5, taken 17:27 UTC 11 February 2014

While it was a rare, sunny winter day for most of the Great Lakes region on 11 February 2014, it’s hard to tell that from the True Color imagery. I mean, look at this True Color MODIS image shown on NPR’s website. Can you tell what is ice and what is clouds?

There are ways of distinguishing ice from clouds, which I have talked about before but, it doesn’t hurt to look at these methods again and see how well they do here. First, let’s look at my modification of the EUMETSAT “Snow” RGB composite:

VIIRS "Snow" RGB composite of channels M-11, M-10 and M-7, taken 17:27 UTC 11 February 2014

VIIRS "Snow" RGB composite of channels M-11, M-10 and M-7, taken 17:27 UTC 11 February 2014

This “Snow” RGB composite differs by using reflectances at 2.25 µm in the place of the 3.9 µm channel that EUMETSAT uses. (Their satellite doesn’t have a 2.25 µm channel.) It’s easy to see where the clouds are now. Of course, now the snow and ice appear hot pink, which you may not find aesthetically pleasing. And it certainly isn’t reminiscent of snow and ice.

If you don’t like the “Snow” RGB, you may like the “Natural Color” RGB composite:

VIIRS "Natural Color" RGB composite of channels I-01, I-02 and I-03, taken 17:27 UTC 11 February 2014

VIIRS "Natural Color" RGB composite of channels I-01, I-02 and I-03, taken 17:27 UTC 11 February 2014

This has the benefit of making snow appear a cool cyan color, and has the added benefit that you can use the high-resolution imagery bands (I-01, I-02 and I-03) to create it. There is twice the resolution in this image than in the Snow and True Color RGB images. Here’s another benefit you may not have noticed right away: the clouds, while still white, appear to be slightly more transparent in the Natural Color RGB. This makes it a bit easier to see the edge of the ice on the east side of Lake Michigan and the center of Lake Huron, for example.

If you’re curious as to how much ice is covering the lakes, here are the numbers put out by the Great Lakes Environmental Research Laboratory (which is about a 25 minute drive from Hell) from an article dated 13 February 2014:

Lake Erie: 96%; Lake Huron: 95%; Lake Michigan: 80%; Lake Ontario: 32% and Lake Superior: 95%. This gives an overall average of 88%, up from 80% the week before. The record is 95% set in 1979, although it should be said satellite measurements of ice on the Great Lakes only date back to 1973.

Why does Lake Ontario have such a low percentage? That last article states, “Lake Ontario has a smaller surface area compared to its depth, so it loses heat more slowly. It’s like putting coffee in a tall, narrow mug instead of a short, wide one. The taller cup keeps the coffee warmer.”  Doesn’t heat escape from the sides of a mug as well as the top? And isn’t Lake Superior deeper than Lake Ontario? Another theory is that “Lake Ontario’s depth and the churning caused by Niagara Falls means that it needs long stretches of exceptionally cold weather to freeze.”  Does Niagara Falls really have that much of an impact on the whole lake?

So, what is the correct explanation? I’m sorry, VIIRS can’t answer that. It can only answer “How Much?” It can’t answer “Why?”

 

BONUS UPDATE (17 February 2014):

It has come to my attention that the very next orbit provided better images of the Great Lakes, since they were no longer right at the edge of the swath. Here, then, are the True Color, Snow and Natural Color RGB composite images from 19:07 UTC, 11 February 2014:

VIIRS "True Color" composite of channels M-3, M-4 and M-5, taken 19:07 UTC 11 February 2014

VIIRS "True Color" composite of channels M-3, M-4 and M-5, taken 19:07 UTC 11 February 2014

 

VIIRS "Snow" RGB composite of channels M-11, M-10 and M-7, taken 19:07 UTC 11 February 2014

VIIRS "Snow" RGB composite of channels M-11, M-10 and M-7, taken 19:07 UTC 11 February 2014

 

VIIRS "Natural Color" composite of channels I-01, I-02, and I-03, taken 19:07 UTC 11 February 2014

VIIRS "Natural Color" composite of channels I-01, I-02, and I-03, taken 19:07 UTC 11 February 2014

 

UPDATE #2 (18 March 2014): The Great Lakes ice cover peaked at 92.2% on 6 March 2014, just short of the all-time record in the satellite era. March 6th also happened to be a clear day over the Great Lakes, and VIIRS captured these images:

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 18:35 UTC 6 March 2014

VIIRS True Color RGB composite of channels M-3, M-4 and M-5, taken 18:35 UTC 6 March 2014

 

VIIRS Natural Color RGB composite of channels M-5, M-7 and M-10, taken 18:35 UTC 6 March 2014

VIIRS Natural Color RGB composite of channels M-5, M-7 and M-10, taken 18:35 UTC 6 March 2014

Wild Week of Wildfires, Part III

The last two posts covered flooding. Now, a month later, we are back to covering last year’s most common topic: wildfires. This time, we’ll make a game out of it. Keep in mind that, for many operational fire weather forecasters, this isn’t a game – it is information that could prove useful in saving lives or homes from destruction. If you have read the earlier posts on fire detection and haven’t forgotten what you’ve been told (here’s a good one to go back and read), this should be easy for you.

The following images are the unmapped data from three consecutive VIIRS granules over the Southwest U.S., starting at 20:36 UTC 11 June 2013. The “raw” data has been processed to produce the “True Color”, “Natural Fire Color” and “Fire Temperature” RGB composites. Plus, the brightness temperature data from channel M-13 (4.0 µm) has a color table applied to it to aid in fire detection. Satellite channels near 4 µm are the “industry standard”, so to speak, for detecting fires as they are highly sensitive to sub-pixel heat sources like fires. The “Natural Fire Color” and “Fire Temperature” composites are RGB composites developed just for VIIRS that both had their debut on this very blog.

The question is: how many fires can you see? Remember, you have to allocate resources (firefighters, helicopters, planes, etc.) based on your assessment. The media is hounding you for all the latest statistics on each blaze and they can’t wait until the 5:00 briefing. They need the scoop now to get higher ratings. Plus, the crew is loading fire retardant on the plane as you read this. Where should the pilot fly to? Everyone is counting on you! (Of course, you would never have just satellite data by itself in a real-life scenario – but, do you want to play this game, or just think of flaws?)

I’ll give you a hint: You won’t see any fires unless you view each image at full resolution. Click on the image, then on the “3200×2304” link below the banner to see the full resolution version. (You could even open each full resolution image in a new tab, and click between the tabs for easy comparison, assuming you’re not using some archaic version of Internet Explorer or another old browser that doesn’t allow tabs. When you would click on the “3200×2304” link, instead right-click and select “Open in New Tab”. Another option would be to save the images and open them in an image viewing software program that will allow you to zoom in more than 100% but, that is starting to sound like a lot of work and I’m not sure I want to play this game anymore. It’s too complicated. By the way, if that’s the way you feel, don’t become the manager of a fire incident team.)

I’ll give you another hint: Many of the hot spots that indicate fires are only 1-2 pixels in size. Be prepared to look for needles in the haystack, and make sure you have your reading glasses on, if you need them.

VIIRS "True Color" composite of channels M-03, M-04 and M-05, taken at 20:36 UTC 11 June 2013

VIIRS "True Color" composite of channels M-03, M-04 and M-05, taken at 20:36 UTC 11 June 2013

VIIRS "Natural Fire Color" composite of channels M-05, M-07 and M-11, taken 20:36 UTC 11 June 2013

VIIRS "Natural Fire Color" composite of channels M-05, M-07 and M-11, taken 20:36 UTC 11 June 2013

VIIRS "Fire Temperature" composite of channels M-10, M-11 and M-12, taken 20:36 UTC 11 June 2013

VIIRS "Fire Temperature" composite of channels M-10, M-11 and M-12, taken 20:36 UTC 11 June 2013

VIIRS channel M-13 image, taken 20:36 UTC 11 June 2013

VIIRS channel M-13 image, taken 20:36 UTC 11 June 2013

So, did you see them all? You should have identified 12 fires. Did you find less than 12? Some of them are hard (or impossible) to see in some of the images. Did you find more than 12? The color scale used on the M-13 image led to false alarms, so you can be forgiven if that’s what caused you count too many.

This example shows some of the complicating factors when trying to identify fires from satellites. It also shows why fire managers never rely on satellite data alone. Now, having said that, VIIRS can and does provide useful information on fires.

First, here’s the answer (link goes to PDF) from the National Interagency Fire Center. They identified 15 active “large incident” fires on 12 June 2013. (They update their maps once per day, so all the fires that started on 11 June make it on the 12 June map.) But, there are differences between their map and what VIIRS saw.

First, the Mail Trail fire (#5 in the PDF) is outside the domain of these three VIIRS granules, so you couldn’t have found that in these images. Fires #3, 4 and 7 (Healy, Porcupine and Ferguson) are obscured by clouds, and/or were mostly contained, transitioning from active to inactive. The Tres Lagunas Fire (#13) started back in May and is undergoing mop up activities. The hot spots from that fire (if there are any left) aren’t visible in the images, but the burn scar is. That leaves the Stockade (#1), Crowley Creek (#2), Hathaway (#6), Fourmile (#8), Silver (#9), Thompson Ridge (#10), Jaroso (#11), Big Meadows (#12), Royal Gorge (#14), and Black Forest (#15) – 10 fires which are all visible in the VIIRS images. Plus, VIIRS saw two more fires that are not included on that list: one in southern California (near the Salton Sea) that I couldn’t find any information on, plus a pellet plant fire in Show Low, Arizona. (Small fires in towns are usually outside the scope of the National Interagency Fire Center, so they don’t bother to list those.)

I would argue that the “Fire Temperature” composite worked the best at identifying each of these fires, but all 4 images have their uses. Here’s the Fire Temperature RGB image with the visible fires identified:

VIIRS "Fire Temperature" composite of channels M-10, M-11 and M-12, taken 20:36 UTC 11 June 2013

VIIRS "Fire Temperature" composite of channels M-10, M-11 and M-12, taken 20:36 UTC 11 June 2013

Answer honestly. Which fires did you see, and which fires did you miss?

The Fire Temperature RGB takes advantage of the VIIRS channels in the portion of the electromagnetic spectrum ranging from the near-infrared (NIR) to the shortwave infrared (SWIR). The blue component is M-10 (1.61 µm), the green component is M-11 (2.25 µm) and the red component is M-12 (3.7 µm). As wavelength increases over this range, the contribution of the Earth’s emission sources increases and the contribution from the sun decreases. As a result, only the hottest hot spots show up in M-10, as they have to be seen over the large signal of radiation from the sun reflecting off the Earth’s surface. In M-12 (as in M-13), hot spots from fires produce more radiation at that wavelength than the amount of reflected solar radiation. M-11 is somewhere in the middle. That means relatively cool (e.g. smoldering) or small fires only show up in M-12, which makes those pixels appear red. Pixels containing fires hot enough or large enough to show up in M-11 will take on an orange to yellow color. Pixels containing fires hot enough or large enough to show up in all three channels will appear white.

You have to be careful, though, as some pixels in the Fire Temperature RGB appear red, even though there aren’t any fires in them. A few of these pixels show up red in the M-13 image, and are labelled as “not a fire/false alarm”:

VIIRS M-13 image, taken 20:36 UTC 11 June 2013

VIIRS M-13 image, taken 20:36 UTC 11 June 2013

According to the color table used, any pixel with a brightness temperature above 340 K (67 °C) will be colored, with colors ranging from red to orange to pale yellow as temperature increases. Now, look at that area in the True Color image (or on Google Maps):

VIIRS "True Color" composite of channels M-03, M-04 and M-05, taken 20:36 UTC 11 June 2013

VIIRS "True Color" composite of channels M-03, M-04 and M-05, taken 20:36 UTC 11 June 2013

That area is very dark – almost black – volcanic rock with very little vegetation that has been baking in the sun all day. It has managed to acquire a brightness temperature that is higher than some of the active fire pixels. The Crowley Creek fire doesn’t show up as red in the M-13 image (the Stockade fire is the one with the yellow and orange pixels) and the Fourmile fire is barely visible. (It has two pixels warmer than 340 K, even though 10 pixels appear red in the Fire Temperature RGB). The color scale in the M-13 image could be applied to a different temperature range, but you’ll always have that trade-off: have the colors start at too high a temperature, and you’ll miss some fires; have the colors start at too low a temperature, and you’ll increase the false alarms.

The True Color image should have helped you identify 5 of the fires. The smoke plumes that show up are a dead giveaway. I’m talking about the Big Meadows, Royal Gorge, Jaroso, Thompson Ridge and Silver fires, of course. There may be smoke with the Hathaway fire, but it would be mixed in with the cirrus clouds and hard to see. Not all fires produce a lot of smoke, though. Having information on the ones that do aids in issuing air quality alerts, among other benefits.

Lastly, the Natural Fire Color image highlights most (but not all) of the fires. Look for the red pixels:

VIIRS "Natural Fire Color" composite of channels M-05, M-07 and M-11, taken 20:36 UTC 11 June 2013

VIIRS "Natural Fire Color" composite of channels M-05, M-07 and M-11, taken 20:36 UTC 11 June 2013

The Natural Fire Color doesn’t show active hot spots at Crowley Creek, and the Hathaway and Fourmile fires are difficult to see, because they aren’t quite hot enough. (Generally speaking, any fire that shows up red in the Fire Temperature RGB is too cold to show up as red in the Natural Fire Color.) But, this composite has the advantage of showing burn scars in addition to the active fires. Burn scars appear dark brown. The Fourmile and Crowley Creek burn scars are visible. Plus, burn scars from last year’s fires still show up: The Whitewater-Baldy, High Park and Waldo Canyon scars are identified. The Tres Lagunas was mentioned above, and it’s burn scar is visible. If you look closely, I’m sure you could find more burn scars from last year’s long fire season.

Here are all four images, zoomed in on each fire at 800%, combined into an animation to highlight how each fire appears in each image:

Animation of M-13, True Color, Natural Fire Color and Fire Temperature imagery zoomed in each fire (20:36 UTC 11 June 2013)

Animation of M-13, True Color, Natural Fire Color and Fire Temperature imagery zoomed in each fire (20:36 UTC 11 June 2013)

For some reason, you have to click to the full resolution version of the image before the animation will display.

Hopefully, this exercise is useful in demonstrating the complications that arise when trying to detect fires from satellites in space, as well as the strengths and weaknesses of some of the various methods VIIRS has at it’s disposal to aid the fire weather community.

Chinese Super-Smog

No, not a Super-Smörg, super smog. Smog that is so thick, you can taste it. The smog in many parts of eastern China has been so bad this winter, it is literally “off-the-charts“. Based on our Environmental Protection Agency‘s not-very-intuitive Air Quality Index (see pages 13-16, in particular) any value above 300 is hazardous to everyone’s health. The scale doesn’t even go above 500 because the expectation is that the air could never get that polluted. Applying this scale to the air in Beijing, the local U.S. Embassy reported an Air Quality Index value of 755 on 13 January 2013. Visibility has been reduced to 100 m at times. This video (from 31 January 2013) gives a vivid description of the problems of the smog:

If that wasn’t bad enough, here’s video from NBC News where Brian Williams reveals a factory was on fire for three hours before anyone noticed because the smog was so thick!

Did you happen to notice in the beginning of the NBC video that the “air pollution is so bad that the thick smog can now be seen from space”? Of course, the satellite image shown in that clip came from MODIS. (It must have friends in high places. That, or people get the MODIS images out on their blogs less than two weeks after the event occurred, unlike this blog.) Needless to say, VIIRS has seen the smog, too, and it is terrible.

For comparison purposes, here’s what a clean air day looks like over eastern China:

VIIRS "true color" RGB composite of channels M-03, M-04 and M-05, taken 05:21 UTC 28 September 2012

VIIRS "true color" RGB composite of channels M-03, M-04 and M-05, taken 05:21 UTC 28 September 2012

This is a “true color” composite taken 05:21 UTC 28 September 2012. (As always, click on the image, then on the “2040×1552” link below the banner to see the full resolution image.) There appears to be some air pollution in that image (look near 33° N latitude between 112° and 116° E longitude), but it’s not that noticeable.

Here’s what it looks like when Beijing is reporting record levels of air pollution (04:56 UTC 14 January 2013):

VIIRS true color RGB composite of channels M-03, M-04 and M-05, taken 04:56 UTC 14 January 2013

VIIRS true color RGB composite of channels M-03, M-04 and M-05, taken 04:56 UTC 14 January 2013

You may have heard of a “brown cloud of pollution“. Here the clouds actually appear brown thanks to all that pollution. Notice the area around Shijiazhuang – the most polluted city in China – and how brown those clouds are in comparison to the clouds on the left and right edges of the image. Then look south from Shijiazhuang to where everything south and west of the cloud bank has a dull gray color. That is all smog! It’s enough to make anyone with a respiratory condition want to cough up a lung just from seeing this.

Now, this is a complicated scene with clouds, snow, ice and smog. So, to clear things up (in a manner of speaking), here is the same image with everything labelled:

VIIRS true color RGB composite of channels M-03, M-04, and M-05, taken 04:56 UTC 14 January 2013

VIIRS true color RGB composite of channels M-03, M-04, and M-05, taken 04:56 UTC 14 January 2013

The gray smog can be seen around Beijing as well, but it pales in comparison to the rest of eastern China. Think about that! Replay the videos above and consider that might not have even been the worst smog in China at the time!

Too bad there are a lot of clouds over the area. What does it look like on a “clearer” day? (“Clearer”, of course, refers to the amount of clouds, not air pollution.) It looks worse! The image below was taken at 04:32 UTC on 26 January 2013:

VIIRS true color RGB composite of VIIRS channels M-03, M-04, and M-05, taken 04:32 UTC 26 January 2013

VIIRS true color RGB composite of VIIRS channels M-03, M-04, and M-05, taken 04:32 UTC 26 January 2013

The area covered by smog rivals the area of South Korea, which is visible on the right side of the image. (One of the reports I linked to above put the figure at 1/7th of the land area of China covered by smog around this time, which is actually a lot bigger than South Korea!) I’m just counting the smog in the image that is thick enough to completely obscure the surface. There is likely smog that isn’t as obvious (and isn’t labelled) in that image. The snow between Shijiazhuang, Tianjin and Beijing is covered by smog that isn’t quite thick enough to totally obscure it. And the large area of snow south of Tianjin is likely covered with smog. (It sure is a lot dirtier in appearance than the snow near the top of the image.)

If you don’t believe my labels, the “pseudo-true color” or “natural color” RGB composite clearly identifies the low clouds (which usually appear a dirty, off-white color even without smog), ice clouds (pale cyan) and snow (vivid cyan):

VIIRS false color RGB composite of channels M-05, M-07 and M-10 (a.k.a. "natural color"), taken 04:32 UTC 26 January 2013

VIIRS false color RGB composite of channels M-05, M-07 and M-10 (a.k.a. "natural color"), taken 04:32 UTC 26 January 2013

Notice the smog in this image. It is an unholy grayish-greenish color with a value near 70-105-93 in R-G-B color space. The “natural color” composite is made from channels M-05 (0.67 µm, blue), M-07 (0.87 µm, green) and M-10 (1.61 µm, red), which are longer wavelengths than their “true color” counterparts. Longer wavelengths mean reduced scattering by atmospheric aerosols, so the higher green value may be due to the strong surface vegetation signal in M-07 being able to penetrate through the smog. (Either that or the smog is composed of some chemical compound that has a higher reflectivity value in M-07 than in the other two channels.)

I’ve looked at the EUMETSAT Dust, Daytime Microphysics and Nighttime Microphysics/Fog RGBs, which you might think would show super-thick smog and they don’t. At least, it’s not obvious.

The EUMESAT Dust RGB applied to VIIRS, valid 04:32 UTC 26 January 2013

The EUMESAT Dust RGB applied to VIIRS, valid 04:32 UTC 26 January 2013

The Dust RGB above uses M-14 (8.55 µm), M-15 (10.7 µm) and M-16 (12.0 µm) and requires there to be a large temperature contrast between the dust (cool) and the background surface (hot). Smog almost always occurs when there is a temperature inversion (the air at the ground is colder than the air above) so the necessary temperature contrast won’t exist.

The Daytime Microphysics RGB shows the smoggy areas are a slightly different color than other cloud-free surfaces, but that color can be confused with other non-smoggy surfaces. The clouds really stand out, though:

The EUMETSAT Daytime Microphysics RGB applied to VIIRS, valid 04:32 UTC 26 January 2013

The EUMETSAT Daytime Microphysics RGB applied to VIIRS, valid 04:32 UTC 26 January 2013

Perhaps, with a different scaling, the smog might stand out more.

The Nighttime Microphysics RGB from the night before (18:50 UTC 25 January 2013) is interesting. Notice the cloud identified by the letter “B” and the non-cloud next to it, “A”:

The EUMETSAT Nighttime Microphysics/Fog RGB applied to VIIRS, valid 18:50 UTC 25 January 2013

The EUMETSAT Nighttime Microphysics/Fog RGB applied to VIIRS, valid 18:50 UTC 25 January 2013

Now compare this with the Day/Night Band image from the same time:

VIIRS Day/Night Band image of eastern China, taken 18:50 UTC 25 January 2013

VIIRS Day/Night Band image of eastern China, taken 18:50 UTC 25 January 2013

This was a day before full moon. Thanks to the moon, clouds, snow and smog are visible in addition to the city lights. Points “A” and “B” have nearly identical brightness in the Day/Night Band, but only “B” shows up as a cloud in the Nighttime Microphysics RGB. These lighter areas around “A” and “B” are partially obscuring city lights, indicating “B” is a cloud, while “A” is smog. (If either was snow, you’d be able to see the city lights more clearly. See the lighter area northwest of Beijing, which is snow.)

Nothing sees super-smog like the true color composite, but the Day/Night Band will see it as long as there is enough moonlight. Smog as optically thick as a cloud… *hacking cough* … Yuck!

End of Autumn in the Alps

Much of the United States has had a below-average amount of snow this fall (and below-average precipitation for the whole year). Look at how little snow cover there was in the month of November. Parts of Europe, however, have seen snow. It’s nice to know that it’s falling somewhere. But, can you tell where?

Here is a visible image (0.6 µm) from Meteosat-9, taken 12 December 2012 (at 12:00 UTC):

Meteosat-9 visible image of central Europe, taken 12:00 UTC 12 December 2012

Meteosat-9 visible image of central Europe, taken 12:00 UTC 12 December 2012. Image courtesy EUMETSAT.

And here’s the infrared image (10.8 µm) from the same time:

Meteosat-9 IR-window image of central Europe, taken 12:00 UTC 12 December 2012

Meteosat-9 IR-window image of central Europe, taken 12:00 UTC 12 December 2012. Image courtesy EUMETSAT.

These are images provided by EUMETSAT. Can you tell where the snow is? Or what is snow and what is cloud?

Here’s a much higher resolution image from VIIRS (zoomed in the Alps), taken only 3 minutes later:

VIIRS visible image of central Europe, taken 12:03 UTC 12 December 2012

VIIRS visible image (channel I-01) of central Europe, taken 12:03 UTC 12 December 2012

Now is it easy to differentiate clouds from snow? Just changing the resolution doesn’t help that much.

This has long been a problem for satellites operating in visible to infrared wavelengths. Visible-wavelength channels detect clouds based on the fact that they are highly reflective (just like snow). Infrared (IR) channels are sensitive to the temperature of the objects they’re looking at, and detect clouds because they are usually cold (just like snow). So, it can be difficult to distinguish between the two. If you had a time lapse loop of images, you’d most likely see the clouds move, while the snow stays put (or disappears because it is melting). But, what if you only had one image? What if the clouds were anchored to the terrain and didn’t move? How would you detect snow in these cases?

EUMETSAT has developed several RGB composites to help identify snow. The Daytime Microphysics RGB (link goes to PowerPoint file) looks like this:

Meteosat-9 "Daytime Microphysics" RGB composite of central Europe, taken 12:00 UTC 12 December 2012

Meteosat-9 "Daytime Microphysics" RGB composite of central Europe, taken 12:00 UTC 12 December 2012. Image courtesy EUMETSAT.

Snow is hot pink (magenta), which shows up pretty well. Clouds are a multitude of colors based on type, particle size, optical thickness, and phase. That whole PowerPoint file linked above is designed to help you understand all the different colors.

The Daytime Microphysics RGB uses a reflectivity calculation for the 3.9 µm channel (the green channel of the RGB). Without bothering to do that calculation, I’ve replaced the reflectivity at 3.9 µm with the reflectivity at 2.25 µm (M-11) when applying this RGB product to VIIRS, and produced a similar result:

VIIRS "Daytime Microphysics" RGB composite of the Alps, taken 12:03 UTC 12 December 2012

VIIRS "Daytime Microphysics" RGB composite of the Alps, taken 12:03 UTC 12 December 2012

Except for the wavelength difference of the green channel (and minor differences between the VIIRS channels and Meteosat channels), everything else is kept the same as the official product definition. Once again, the snow is pink, in sharp contrast to the clouds and the snow-free surfaces. We won’t bother to show the Nighttime Microphysics/Fog RGB (link goes to PowerPoint file) since this is a daytime scene.

EUMETSAT has also developed a Snow RGB (link goes to PowerPoint file):

Meteosat-9 "Snow" RGB composite of central Europe, taken 12:00 UTC 12 December 2012

Meteosat-9 "Snow" RGB composite of central Europe, taken 12:00 UTC 12 December 2012. Image courtesy EUMETSAT.

This also uses the reflectivity calculated for the 3.9 µm channel. Plus, it uses a gamma correction for the blue and green channels. Is it just me, or does snow show up better in the Daytime Microphysics RGB?

If you switch out the 3.9 µm for the 2.25 µm channel again and skip the gamma correction when creating this RGB composite for VIIRS, the snow stands out a lot more:

VIIRS "Snow" RGB (with modifications as explained in the text), taken 12:03 UTC 12 December 2012

VIIRS "Snow" RGB (with modifications as explained in the text), taken 12:03 UTC 12 December 2012

Now you have snow ranging from pink to red with gray land areas, black water and pale blue to light pink clouds. This combination of channels makes snow identification easier than the official “Snow RGB”, I think.

All of this is well and good but, for my money, nothing beats what EUMETSAT calls the “natural color” RGB. I have referred to it as the “pseudo-true color“. Here’s the low-resolution EUMETSAT image:

Meteosat-9 "Natural Color" RGB of central Europe, taken 12:00 UTC 12 December 2012. Image courtesy EUMETSAT.

And the higher resolution VIIRS image:

VIIRS "Natural Color" RGB of central Europe, taken 12:03 UTC 12 December 2012

VIIRS "Natural Color" RGB composite of channels M-5, M-7 and M-10, taken 12:03 UTC 12 December 2012

The VIIRS image above uses the moderate resolution channels M-5, M-7 and M-10, although this RGB composite can be made with the high-resolution imagery channels I-01, I-02 and I-03, which basically have the same wavelengths and twice the horizontal resolution. Below is the highest resolution offered by VIIRS (cropped down slightly to reduce memory usage when plotting the data):

VIIRS "Natural Color" RGB composite of channels I-01, I-02 and I-03, taken 12:03 UTC 12 December 2012

VIIRS "Natural Color" RGB composite of channels I-01, I-02 and I-03, taken 12:03 UTC 12 December 2012

Make sure to click on the image and then on the “2594×1955” link below the banner to see the image in full resolution.

This RGB composite is easier on the eyes and easier to understand. Snow has high reflectivity in M-5 (I-01) and M-7 (I-02) but low reflectivity in M-10 (I-03) so, when combined in the RGB image, it shows up as cyan. Liquid clouds have high reflectivity in all three channels so it shows up as white (or dirty, off-white). The only source of contention is that ice clouds, if they’re thick enough, will also show up as cyan.

Except for the cyan snow and ice, the “natural color” RGB is otherwise similar to a “true color” image. Vegetation shows up green, unlike the other RGB composites where it has been gray or purple or a very yellowish green. That makes it more intuitive for the average viewer. You don’t need to read an entire guide book to understand all the colors that you’re seeing.

Compare all of these RGB composites against the single channel images at the top of the page. They all make it easier to distinguish clouds from snow, although some work better than others. Now compare the VIIRS images with the Meteosat images. Which ones look better?

(To be fair, it’s not all Meteosat’s fault. The images provided by EUMETSAT are low-resolution JPG files [which is a lossy-compression format]. The VIIRS images shown here are loss-less PNG files, which are much larger files to have to store and they require more bandwidth to display.)

As a bonus (consider it your Christmas bonus), here are a few more high-resolution “natural color” images of snow and low clouds over the Alps. These are kept at a 4:3 width-to-height ratio and a 16:9 ratio, so they make ideal desktop wallpapers.

VIIRS "natural color" composite of channels I-01, I-02 and I-03, taken 12:29 UTC 14 November 2012

VIIRS "natural color" composite of channels I-01, I-02 and I-03, taken 12:29 UTC 14 November 2012. This is an ideal desktop wallpaper for 4:3 ratio monitors.

That was the 4:3 ratio image. Here’s the 16:9 ratio image:

VIIRS "natural color" composite of channels I-01, I-02 and I-03, taken 12:29 UTC 14 November 2012

VIIRS "natural color" composite of channels I-01, I-02 and I-03, taken 12:29 UTC 14 November 2012. This is an ideal desktop wallpaper for 16:9 ratio monitors.

Enjoy the snow (or be glad you don’t have to drive in it)!