posted on Nov, 1 2012 @ 04:54 PM
Ok, back to this great paragraph from the website. There is another key phrase here - the data is a synthetic blend from five satellites. That's
telling you that they're tiling the data, and manipulating it. In other places here they'll tell you that they're actually morphing the image
shapes, and that always causes some distortion as well. But any tiled mosaic from multiple image sources has a propensity for dropped tiles or edge
mismatch. The software will try to compensate for it. But you're stitching together a lot of images taken from different perspectives. And this will
always provide a possibility for an entire droppped tile. I'll get back to that.
Now, to continue on with "why are there these spiral shapes". What have we learned already? It's a microwave radiometry imager. It's a DMSP SSM/I
channel. And it's a synthetic image composed of shape morphed swaths of data from five satellites. The data comes from looking at variations in the
background microwave glow from the surface, caused by falling water.
Next step, how do you make a microwave radiometry imager, in particular, how does the DMSP SSM/I channel do this? Well, a radiometer is typically made
from a very very sensitive microwave receiver. It pretty obviously can't be a camera - glass lenses don't have much effect on microwaves, so that's
out. Nope, you start with a very sensitive microwave receiver. It's not the sort of thing that makes an image, though. It only looks at one place, a
spot. One spot. It can look at one spot and tell you what the microwave level at that frequency is, at that one spot. And then only through some hard
work - the level of radiation is so low, you have to use a lot of tricks to get the data out of the noise. One thing you do is that almost all
microwave radiometers use some sort of chopping technique, constantly comparing the target to a black body calibration source. Chopping is an old way
of getting better performance out of crap signal, but that's another dissertation.
"But, Tom," you might ask, "if it only looks at one small spot, how do I get a picture out of that?". Easy! I'd reply. You get it by scanning.
Just like an old CRT display - you scan across the scene below with your radiometry antenna, looking at one spot after another in a systematic way,
then you put the image together in computer memory. Sweet, eh? Now, how do you go about that. Well, if you had a really mechanically agile antenna,
you could do this on a Cartesian scan - left to right, top to bottom, moving the receiving antenna in a rectangular grid. This causes issues in that
big motions like this will also move the satellite around its COG a lot. And it takes a lot of motion, which is power use, and it wears out your
bearings/servos in an environment that you can't maintain.
What's another way, one might ask. Well, that other way is to use the motion of the satellite. Remember, a lot of these things spin. If the spin axis
of the satellite is oriented vertically, whoopee! you just got most of your work done for free. If not, you can contrive to have your scanner spin,
which adds issues with satellite orientation due to gyroscopic force, but that's another dissertation. Let's assume the bird has a vertical spin
axis. So, your satellite is spinning around an imaginary line pointing straight down towards the ground. You can now get an image just by looking off
to one side at the outside of your image, and slowly bring your antenna down until you're pointing straight down. As the craft spins, you'll scan a
spiral image, much like a tone arm on an old LP player scanned the record, outside to in (or inside to out...samey same), with no motion other than
canting the antenna down slowly. This, too, moves the COG, but not as much, as the spin of the craft will stabilize against this in that axis. Two
more birds for one stone!
But what you get is an image with spiral artifacts in it. With a cartesian scan, you'd have gotten rectilinear artifacts. What could make this worse,
you might ask, if you're thinking about some scans looking more spirally than others. Well, that's an old issue called interlace artifacts, which
we'll address in part III.