Tag Archives: volcano

Auroras, Volcanoes and Bears, Oh My!

It’s amazing what you can see in a single image from the VIIRS Day/Night Band.

OK, so you can’t actually see any bears with VIIRS (even bears that have fattened up for the winter are less than 375 m across – and certainly less than the 742 m resolution of the Day/Night Band), but you can see auroras and volcanoes. And a lot more! Take a look at this image from the VIIRS Day/Night Band (DNB) taken at 12:19 UTC 7 October 2013:

VIIRS Day/Night Band image, taken 12:19 UTC 7 October 2013

VIIRS Day/Night Band image, taken 12:19 UTC 7 October 2013

What do you see? Make sure you click on it to see the full resolution image.

Well, there’s a bright arc that stretches from Siberia, over the Brooks Range in northern Alaska and into the Yukon and Northwest Territories of Canada, and there are even brighter arcs north of that (between the Brooks Range and Barrow, even extending over the Arctic Ocean). Those are examples of the aurora borealis (a.k.a. Northern Lights).

As an aside, did you know that VIIRS can provide information about the speed of the aurora? I’ll wait if you want to read that. The aurora in this case is moving more in the along-track direction than the across-track direction, so the method for calculating the speed of the aurora won’t work so well with this aurora. Although, there does seem to be some apparent motion of the auroral element immediately off the coast from Prudhoe Bay. See if you can spot the scan-to-scan “shifts”.

Speaking of Prudhoe Bay, you can also see the tremendous amount of light given off by the oil and gas operations there, making the area look like the largest city in Alaska. (Compare the size of the lit-up area with that of Anchorage, which actually is the largest city in Alaska.) If you know your Alaska and Yukon geography, you should also be able to pick out Barrow, Fairbanks, Delta Junction, Whitehorse, Wasilla, Kodiak and Juneau.

Now, what is that big, circular city over the Alaska Peninsula? (It also looks bigger than Anchorage.) That is the volcano Veniaminof, which was erupting last time we looked at it (at the end of August), and appears to still be going strong as of 7 October 2013 (or it calmed down in the meantime before acting up again, as this article suggests). It’s not that the volcano is actually larger than the city of Anchorage, as it appears. What you are seeing is the light emitted by the liquid-phase rock erupting from the volcano. Rock that is hot enough to be liquid is hot enough to emit radiation in the visible and near-infrared wavelengths the Day/Night Band is sensitive to. This light is illuminating a circular area of the clouds surrounding Veniaminof, and enough of it is reflected or scattered to outer space that VIIRS is able to detect it.

Another question you might ask is, “What is that bright band across the image that parallels the scan lines and passes over both Anchorage and Juneau?” It almost looks like an aurora, but it is in too straight a line. It is another example of “stray light”. Now, if you read the previous post (and remember what it said) you might be confused, because I said that stray light was fixed and shouldn’t be a problem anymore. Well, the stray light correction is not perfect, especially in near-new moon images such as this. For example, say that the stray light correction introduces an error in the radiance on the order of 1 x 10-10 W cm-2. During a full moon, the radiance value observed for a clear background surface pixel would be on the order of 1 x 10-7 W cm-2, so this error is too small to notice (0.1 % error). During a new moon, the radiance values of a clear background surface are more like 5 x 10-10 W cm-2 (20% error, which is noticeable).

Another thing to consider is that the stray light correction requires post-processing, and is dependent on the moon’s phase and time of year. Since this image was near new moon in October (13% of full on the 7th), it uses data from the previous new moon (in September) to do the stray light correction, since we didn’t have a similar data from October to use. (Of course, we have that data now, but it wasn’t available at the time this image came off the satellite. It will be used on next year’s October data.) Slight differences in the solar-satellite geometry between September and October are the largest source of error in the stray light correction. Since the stray light correction began in late August 2013, once we get to September 2014 the stray light correction ought to perform much better (but it will still be more likely to show errors during a new moon than a full moon).

Speaking of this being near a new moon, how is it possible that you can see clouds and sea ice and snow? It’s true that the aurora does a good job illuminating the surface underneath, but the aurora doesn’t cover everywhere. What about over the ocean? There are no city lights, no erupting volcanoes, and not enough ships in the sea to light up the sky (like there are off the coast of Korea). Where is the light coming from?

It’s actually a phenomenon known as “airglow” (or “nightglow”, since it is easiest to detect at night). The shortened version is that molecules in the upper atmosphere interact with ultraviolet radiation and, as a result, emit photons. This happens around the clock. Airglow is enough to detect a Super Cyclone at night with no moon.  That means, even on the darkest nights, the Day/Night Band is capable of viewing clouds and ice and snow. You don’t even need the moon at all!

Beginning to See the Light: an Introduction to VIIRS DNB and NCC

If you found this webpage, you are either A) a spam-bot searching for new websites to inundate with spam messages, B) interested in learning about VIIRS and it’s revolutionary Day/Night Band or C) very upset at Google right now for steering you to the wrong place. This website is for those of you in group B, but hopefully a few of you in group C will stick around and become interested to learn about the kinds of things you can do with weather satellites – particularly with a sensor as powerful as the VIIRS Day/Night Band.

First, a little background on VIIRS. Actually, just read this “Beginner’s Guide” (PDF file) that I wrote if you need background. (Hey, I wrote it. I might as well promote it.) It’s designed for people who are interested in using the data but, even if you don’t deal with VIIRS data directly, that PDF has a lot of good information in it you may find useful. Our topic today is basically an expansion on pages 23 and 24 of that document.

Typically, imaging sensors on weather satellites operate in the visible and infrared portion of the electromagnetic spectrum. This is the case with VIIRS, which has 22 channels (also called “bands”) ranging in wavelength from 0.412 µm to 12.01 µm. Perhaps the most revolutionary channel on VIIRS is the Day/Night Band (or DNB). The DNB is a broadband channel sensitive to radiation in the wavelength range from about 0.5 – 0.9 µm, which covers much of the visible and into the near-infrared (near-IR) wavelengths.

What makes the Day/Night Band unique is its ability to detect the low levels of visible light that occur at night. Most visible-wavelength sensors don’t work at night because the signal is well below the noise of the instrument. (Only the DMSP OLS was able to capture visible imagery at night prior to the launch of VIIRS and the DNB has the OLS beat in spatial resolution, radiometric resolution and quantitative applications.)

The DNB observes band-integrated radiance values at ~750 m spatial resolution over a ~3000 km-wide swath that covers the entire Earth twice a day. Since it is on a polar-orbiting satellite, Suomi NPP, it observes high latitude areas (like the Arctic) every orbit (every 101 minutes). The data produced by the DNB is very useful for Arctic applications (as we will show in future topics), but it can be difficult to work with.

The radiance values observed at night are roughly 7-8 orders of magnitude less than during the day, and they vary by several orders of magnitude between a new moon and a full moon. (Here’s a quick and dirty resource for information on the moon’s phase.) This large contrast between day and night creates a lot of problems when trying to display images near the day/night terminator since a lot of computer displays only allow 256 colors. For high-latitude places like Alaska, the terminator is present all night long in the summer months.

Here is a scene containing five VIIRS DNB granules over Alaska near 12:50 UTC on 13 August 2013. This image was created by linearly scaling the radiance values (which range from 1.4×10-3 to 7.3×10-10 W cm-2 sr-1) as a number from 0 to 255:

DNB with linear scaling

VIIRS DNB image taken 12:48 UTC 13 August 2013. This image uses linear scaling of the radiance values.

The top edge of the image is on the day side of the Earth, while the bottom is on the night side. The linear scaling only shows detail on the day side, even though the DNB can detect what’s on the night side.

Taking the base-10 logarithm of the radiance values (now dealing with a scale from -2.3 to -9.1) brings out the detail in the twilight areas, but causes saturation on the day side of the image and the night side still looks dark:

DNB with logarithmic scaling

VIIRS DNB image taken 12:48 UTC 13 August 2013. This image uses logarithmic scaling of the radiance values.

By the way, if you follow my other blog, you might be surprised to find out you only need to click on these images once to get to the full resolution version. I should probably mention that these images show the full width of the VIIRS swath, but have been reduced in resolution by a factor of two.

Here is a more sophisticated attempt at scaling, which uses information about the solar zenith angle to divide the region into strips, and each strip has it’s own scaling designed to be continuous from strip to strip:

DNB example using solar zenith angle-dependent scaling.

VIIRS DNB example from 12:48 UTC 13 August 2013. This image uses solar zenith angle-dependent scaling. Image courtesy of GINA.

This scaling allows you to see features on the day side, night side, and everywhere in between. But, see all the wave-like, broad ripples in the middle of the image? Those aren’t actually in the data – it’s a consequence of using this scaling method. The shorter wavelength ripples near the bottom of the image are caused by “striping” and “stray light”.

Striping occurs because the 16 detectors that make up the DNB may not all have the exact same sensitivity to light. In each scan (16 rows of pixels in the full resolution data), some rows of pixels appear brighter, because that particular detector is more sensitive to light than its neighbors. This was a problem particularly at low light levels, but a stray-light fix has been implemented and was put into operation on 20 August 2013 (a week after this image was taken) that should fix (or at least reduce) this.

Stray light is light that hits the detectors that isn’t coming from the Earth – it comes directly from the sun. This happens shortly after the satellite passes into the night side and shortly before it passes back into the day side of the Earth. Don’t worry, though. This was fixed along with the striping on 20 August 2013.

Now, what about those stripes that go from northeast to southwest (primarily over the Yukon Territory)? Are those some sort of artifact of the scaling method? Nope. Those are shadows cast by deep convective clouds near sunset – just like in this photo.

So, we’ve highlighted this problem with the DNB: how do you best display a 7-orders-of-magnitude change in value using only 256 colors? VIIRS already has the solution covered. It’s called the Near Constant Contrast product (often shortened to NCC).

The Near Constant Contrast product takes the radiance values observed by the DNB and coverts them into reflectance (also called albedo). Now, think about what it means to do this conversion. For most visible wavelength imagery, reflectance is relatively straight-forward to calculate. The satellite observes the reflected radiation, and we assume the incident radiation is all coming from the sun. (This is a good assumption during the daytime.) Calculating the incident solar radiation on each point on the Earth at a given time (and day of year, etc.) is considered a solved problem. But, what do we do at night?

At night, you have to take the moon into account. The NCC imagery uses a model of the sun and moon to calculate the incident radiation for all points on the Earth at all times of day, all days of the year, for all phases of the moon. Accounting for this variation in the incident radation reduces the range of values we need to display for scenes that cross the terminator. The problem that arises is that the DNB senses light from more than just the sun and moon. It can detect fires, city lights and auroras (among other things), which are sources of emitted light, not reflected light. These light sources can be 2-3 orders of magnitude brighter than the reflected component (particularly during a new moon). Nevertheless, the NCC product reduces the range of values we have to display from 7-8 orders of magnitude down to 2-3 orders of magnitude, and it produces images like this:

NCC image using logarithmic scaling

VIIRS NCC image taken 12:48 UTC 13 August 2013. This image uses logarithmic scaling.

This image was scaled using the base-10 logarithm of reflectance on a scale from -1.3 to 1.3 (roughly from 0.05 to 20 in the original reflectance units). The only loss of contrast occurs on the bottom of the image where stray light is contaminating the reflectance signal, and this has since been corrected for.

See that dot of light over the Alaska Peninsula? I’ll zoom in at full satellite resolution so you can get a better look:

VIIRS NCC image taken 12:52 UTC 13 August 2013. This is cropped and zoomed in from the previous image.

VIIRS NCC image taken 12:52 UTC 13 August 2013. This is cropped and zoomed in from the previous image.

That’s light emitted by the molten-hot magma erupting from the Veniaminof volcano. The DNB (and its sister product, NCC) are sensitive enough to see glowing-hot lava all the way from outer space!

Want to know what it looks like with the stray light removed? Here’s an image from 30 August 2013 (about a week after the stray light correction became operational):

VIIRS NCC image, taken 12:30 UTC 30 August 2013

VIIRS NCC image taken 12:30 UTC 30 August 2013. This image uses logarithmic scaling.

The Veniaminof volcano is still erupting at this time (two weeks later!), and is really easy to see. Here it is at full resolution:

VIIRS NCC image taken 12:30 UTC 30 August 2013

VIIRS NCC image taken 12:30 UTC 30 August 2013. This is cropped and zoomed in from the previous image.

Depending on how the scaling is performed, NCC and DNB imagery is very similar for daytime and nighttime scenes, but it is the twilight and near-terminator scenes where the NCC product really shines.