Capturing the Earth, satellite imagery storage requirements at planetary scale

Coming across the capacities of the very large data centres now being built, in the exabyte range, it’s interesting to consider how much space is needed to store an image of the entire earth captured at the highest possible resolutions, with current technology, from a realistic orbit.

There currently exists databases of partial imagery of the earth in near real time, through composite stitching of various satellite outputs, providers such as Zoom.Earth have already achieved quite a bit in this field.

What if one day something akin to that is available with the most advanced imagery? The current state of the art for the top secret satellites is likely under 10 cm, though of course the exact number is classified. I firmly believe in the near future 5 cm resolution will be routinely available with advancements in optics.

Rough numbers:

1 square km at 5 cm resolution = 20000 x 2000 pixel image = 400 megapixels

In 30 bit RGB (10 bit color) that is 1.5 GB losslessly compressed (2 to 1 compression) ideally.

Earth has a surface area of 510 000 000 square km

That’s a 204 quadrillion pixel image!

Some simple math gives 765 petabytes (PB) of storage is needed for one image. Where 1 PB = 1000 TB = 1 million GB = 1 billion MB

Of course in reality the earth is not perfectly spherical, additional overhead data has to be stored, 765 PB would require redundant storage or you would lose that to bit rot quite quickly, the water images could probably be compressed more, etc.

Given that roughly 71% of the surface area is water a few hundred PB could probably be cut through lossy compression without sacrificing any perceptible image quality.

Nonetheless let’s assume minimal data overhead and complete accuracy, so we’ll need triple redundancy, at least!, with some buffer as well.

A ballpark number could 2400 PB, or 2.4 exabytes, of actual disk space needed for one image.

If you could accept the odd bit flip or compression artifact this could easily be reduced to 240 PB given the advanced state of compression algorithms nowadays.

So what about video?

At 30 fps at lossless quality that gives 72 exabytes (EB) per second of video!

At a more realistic compressed standard, perhaps as little as 1.44 EB per second, assuming 100 to 1 lossy compression efficiency.

This calculation is a bit silly as the only we have currently of capturing whole earth shots is with satellites parked at geostationary orbit, that could not reach a 5cm resolution without some truly massive optics. I really don’t expect to see this sort of capability this century.

A more realistic way is to look at what the reasonable capabilities are of geostationary orbit (GEO) earth observation satellites within the foreseeable future, with say Hubble sized optics. Although the current state of the art is at 500 meters with Japan’s Himawari 8, I believe one day we can achieve 10 meter resolution imagery from GEO.

At the 10 meter per pixel scale, that gives 1.8 PB per second of video at 2:1 lossless quality, and just 36 TB per second at 100 to 1 lossy compression. Actually feasible with current storage technologies.

Although 10 meters per pixel will barely resolve buildings, this is still quite useful for at least studying cloud formations, weather patterns, ship movements, and perhaps large plane movements at a massively improved resolution from current weather satellites.

The trickiest part would be imaging the poles since from geostationary orbit the images will be highly skewed, one day something akin to a pole sitting satellite (see ESA) may be used to provide coverage.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s