Hacker News Reader

Comment by hak8or | Hacker News Reader

I was curious what instruments this use, looks like a special form of radar? Does this mean it effectively gives us very accurate height maps regardless of cloud coverage, and is able to differentiate between what surface material it's seeing?

> Radar instruments can image Earth’s surface through clouds, precipitation, regardless of sunlight, making them particularly well suited for monitoring polar regions. The Sentinel-1C and -1D satellites also carry an Automatic Identification System (AIS) instrument – improving the mission capacity to detect ships and sea pollution. The Sentinel-1D AIS was also activated as the satellite passed over Antarctica capturing the presence of ships in these extreme areas.

View on HN

Synthetic aperture radar is basically building a bitmap of radar reflectivity. So what you get looks a lot like a photo. You can end up with very non-photo artifacts though - blown out pixels caused by corner reflectors, bright things can result in ghost copies in multiple places and if there’s other radar operating in the same frequency bands it can end up on the picture.

The core idea is that you send out pulses as you pass over the ground and then record the echoes. You can create an image by - for each pixel in the image - working out the response you would expect to receive back and correlating that with the actual responses you saw. That gives you a reflectivity value. You can do it in multiple polarisation to better distinguish things.

Ideally you want to have a large collecting area (aperture) for radar to get good resolution. But it isn't practical to put a big radar dish in space. So they use a small aperture and simulate a larger one by sweeping out an area over time and using some clever maths. Hence 'synthetic aperture radar'.

What you can get in a single image are 5.5cm wavelength microwave backscatter - this means surface materials can be differentiated by looking at texture differences at that scale. So - tarmac vs a ploughed field, for example. There's 2 polarizations as well, so you can identify e.g. vegetated areas also, which scatter the signal in a different way.

A single image from Sentinel-1 won't give a height map directly, but a pair can using interferometry (InSAR), as the phase of the backscattered signal is also measured. With that you can derive something about the terrain. It's not super accurate though for absolute height maps.

And yes the signals pass through cloud and it works at night.

If my understanding is correct (and I'd love to be corrected if not!), it can be used to generate super accurate differential heightmaps. It won't tell you exactly how high a peak is, for example, but it can tell you that it's dropped a few millimeters since the last time you measured.

https://en.wikipedia.org/wiki/Interferometric_synthetic-aper...

Typically you will get an image pair for an area every 6-12 days. The phase used in interferometry is massively affected by atmospheric conditions, which can vary a lot in this time, and are difficult to correct for. So, one pair is often not enough for this. But if you look at a bunch of pairs for that area over a longer time period, you might be able to correct for the atmospheric effects and get your differential height map. You can get more accurate elevation models 'out of the box' with different systems, e.g., the SRTM (one of the most well known publicly available global elevation maps) [1] was made with insar but 2 antennas on one craft, and Germany's TanDEM-X [2] is a pair of satellites flying in formation a few hundred meters apart, capturing the same area at the same time.

[1]: https://www.earthdata.nasa.gov/data/instruments/srtm [2]: https://en.wikipedia.org/wiki/TanDEM-X

Reply on news.ycombinator.com