lindzey

This is a pulsed system - we send out 1us wide chirped signals, then listen for their returns before sending out the next signal. So - we know how far away a return is, but the tricky bit is knowing which direction it came from, since our beam pattern is so wide.

We do use a chirped signal in order to improve vertical resolution - convolving the outgoing pulse with the returned trace gives us the equivalent resolution of a ~80ns pulse, but with more power.

We record on two channels, "high gain" and "low gain", separated by ~50dB. I showed the high gain products in this post, and the surface absolutely does saturate the detectors. The system was designed so that near-surface returns don't saturate the low gain channel.

We transmit 8kW, and the air/ice surface reflection coefficient is ~0.08 (~-11dB). Flying at ~600m above the surface, spreading loss actually contributes more to signal attenuation (1/(2*h)^2 ~= -62dB).

Our instrument is optimized for seeing through the entire ice sheet, mapping deep layers and the bed. Other (also airborne) instruments operate at higher frequencies, trading higher resolution for less penetration. I'm not super familiar with groups using ground-based ice-penetrating radar, but one big tradeoff is $$$. The airplane is hugely expensive to operate, whereas ground-based just needs a snowmobile.

Yes - this idea has definitely taken over my imagination. A different research group (https://www.cresis.ku.edu/) has been working on this, but the processing techniques are still evolving.

Do you have papers describing y'all's processing? I'd love to learn more!

Absolutely! I only included our lowest-level data products in this post. We don't sweep the radar side-to-side, but since it's on a moving platform, we see the same thing from multiple angles. In the data I showed, there has been some very limited stacking applied, mostly just to improve the signal-to-noise ratio. The next level of processing would be focusing/migration (depending on whether you're in the radar or seismic world) - it reasons about possible incident angle and collapses the hyperbolas to a single point.

Unfortunately, this only improves the resolution along the flight track; features parallel to the flight path and offset to the side are the hardest to filter out.

Using our sensor geometry, there's an ambiguity from where off to the side the energy was returned from. So, the solution would wind up being non-unique whatever you do.

There has been some fun work along the lines you describe that uses maps of the surface shape (generated from camera imagery or scanning laser data) to discriminate which apparently-subsurface echoes are most likely due to the surface topography. So far as I know, this hasn't been automated - it's more an aid to human interpretation (we have an army of undergrads that "picks" the most likely bed location).

I'm a student in the current batch, and can confirm that their ability to make money off recruiting fees is not a factor in the admissions decisions - I made it very clear in the admissions process that I would not be looking for a job after the batch, because it only seemed fair to let them filter on that if they wanted to. It was and has continued to be a non-issue. In fact, the founders/facilitators I've talked to are enthusiastic about my plans, and offered me great advice.