Regional and Mesoscale Meteorology Branch

How fast does an aurora move? I “googled” it, and got answers ranging from “fast” to “very fast”. Not very scientific. It also doesn’t help that the majority of aurora videos on the Internet are time-lapse footage, and there’s no way to know how fast the footage has been sped up. Although, I did find this video that claims to be real-time footage:

When the camera is still, you could try to calculate the speed of some of the aurora elements if you knew where the cameraman was, what stars were in the view (and how far apart they are), and how high up (or how far away) the aurora was at that time. All information that I don’t have.

What if I said we could estimate the speed of the aurora by examining VIIRS Day/Night Band (DNB) images?

Here’s a DNB image of the aurora australis (a.k.a. Southern Lights) over Antarctica, taken on 1 October 2012:

VIIRS DNB image of the aurora australis, taken 00:22 UTC 1 October 2012

Compare this image with the images of the aurora borealis shown back in March 2012. Something doesn’t look right. Far from looking like smooth curtains of light, the aurora (particularly the brightest one) has a jagged appearance, like a set of steps. (This is easier to notice if you click on the image to see it in higher resolution.) This is because the aurora wouldn’t stay still, and we can use this information to estimate the speed it was moving.

The stripes that you see in the image are a caused by the 16 detectors that comprise the DNB which, for various reasons, don’t have exactly the same sensitivity to light. (This condition is given a super-scientific name: “striping”.) The DNB senses light from the Earth by having a constantly rotating mirror reflect light onto these detectors. One rotation of the mirror (particularly the part that occurs within the field of view of the sensor) comprises one scan. Each detector comprises one row of pixels in each scan, each with 742 m x 742 m resolution at nadir. There are 48 scans in one “granule” (the amount of data transmitted in one data file), and it takes ~84 seconds to collect the data that make up one granule. That means it takes ~1.75 seconds per scan.

If you watch that video again, you’ll notice that the aurora can move quite a bit in 2 seconds. Now, let’s zoom in much more closely on one of the aurora elements:

Zoomed-in VIIRS DNB image of an aurora, taken 00:22 UTC 1 October 2012

This image has been rotated relative to the original image, in case you were wondering why it doesn’t seem to match up with the first image. The brightest pixels are where the brightest aurora elements were located. The “steps” (or “shifts” as they are typically called) occur every 16 pixels, which mark out the end of one scan and the beginning of the next.  If you count the number of pixels that the brightest aurora elements shifted from one scan to the next, it varies from about 6 to 10 pixels. Assuming a constant resolution of 742 m per pixel along the scan (which isn’t exactly true, the resolution degrades a little bit as you get closer to the edge of the scan but not by much), that means this particular aurora element moved somewhere between ~4.5 and ~7.5 km in ~1.75 seconds from one scan to the next. Doing the math (don’t forget to carry the 1), that comes out to somewhere between 9000 and 15,000 km h^-1 (rounded to account for possible sources of error), which I guess counts as “very fast”. But, it’s not as fast as the coronal mass ejections that create auroras. They have an average speed of 489 km s^-1 (1,760,000 km h^-1)!

So, what looks like an oddity in the VIIRS image, actually contains some interesting scientific information about the speed of an “active aurora“.

But, we’re not done yet. Let’s get back to the striping. Along with “stray light”, it’s one of the few remaining issues in VIIRS imagery. Stray light, which you can see evidence of in the lower right corner of first aurora image, is a particular problem in the DNB. It occurs when sunlight is reflected onto the detectors when the satellite is on the nighttime side of the Earth, but close to the edge of the day/night “terminator“. Our colleagues at Northrup Grumman have been working on a correction to stray light that also reduces the striping. This correction allows for much better viewing of auroras, which have a tendency to occur right where stray light is an issue.

Here is an image of another aurora over Antarctica, taken on 15 September 2012, corrected for stray light and striping:

VIIRS DNB image of the aurora australis over Antarctica, taken 18:56 UTC 15 September 2012. The data used in this image was corrected for stray light and striping by Stephanie Weiss (Northrup Grumman).

This was the night of a new moon, so the only light in the scene (once the stray light is taken out) is the aurora. (OK, there may be some “air glow” and starlight. But, it doesn’t show up on this brightness scale.)

This aurora was a lot less “active” so it looks more like smooth curtains of light. Although, when you zoom in on the brightest swirl in the upper right corner, you can see it did move 3-5 pixels between scans:

VIIRS DNB image of the aurora australis over Antarctica, taken 18:56 UTC 15 September 2012. This image has been zoomed in and rotated relative to the previous image of the same aurora. The data used in this image was corrected for stray light and striping by Stephanie Weiss (Northrup Grumman).

This translates to 4000 to 8000 km h^-1, which still counts as “fast” even if it doesn’t count as “very fast”. See, Google was right! Auroras do move anywhere from “fast” to “very fast”. But, now we at least have an estimate to quantify that speed.

And, in case you were wondering, these estimates of the speed of auroras are consistent with earlier observations. According to the book Aurora and Airglow by B. McCormac (1967), the typical speed of auroras is between 0 and 3 km s^-1  (up to 10,800 km h^-1). So, it appears that VIIRS does give a reasonable estimate about the speed of an aurora. We just happened to catch one “typical” aurora and one “faster than typical” aurora.