I wonder what it would take to do this for in-flight aircraft? I.e. simultaneously surveying the entire sky and tracking all aircraft in flight? This could allow finding downed aircraft much more quickly. Would a megaconstellation like starlink with cameras suffice?
Their current best satellite in the constellation is the SuperDove and it takes images ~637 sq km at a time so, as a naive calculation, to cover the whole earth at any single instant you'd need over 800 thousand of Planet's SuperDove ignoring the difficulty of making a constellation that efficient and ignoring that you can also save a lot of birds by confining the area of interest to a narrower band of latitude without missing much activity. Your raw efficiency per satellite does take a small hit as you shrink the latitude coverage because you get a concentration of satellites along the edges but that's just orbital mechanics for you. (you can see this on the starlink constellation trackers, they're moderately denser right along the edge of the latitude band of their orbit: https://satellitemap.space/?norad=59877)
Now if we reduce the time resolution somewhat and accept we may not know at the exact moment where any particular plane is but that it will get captured on a satellite every X minutes you can really start cutting into those numbers.
Most planes broadcast their position via ADS-B while they are in the air. On land those are picked up by ground stations. For a long time we had bad coverage of planes at sea, but in part in response to Malaysia Airlines Flight 370 the Iridium satellite network now also picks those signals up, giving us global coverage. The satellite coverage is limited to the frequencies used by airliners and business jets, but those are the ones we are most concerned with finding if they go down.
As a point of curiousity, air geophysical survey craft (planes, helicopters) don't broadcast their position via ADS-B .. except perhaps at the start and end of a survey grid or at the turn point of a survey line.
If they "disapear" it's no great mystery as to where they might be .. they have thoroughly planned to within a metre flight paths and are on contract to drape specific areas.
However, they are loaded with instrumentaton: magnetic sensors, radiometric sensors, EM sensors for EM surveys, multiple GPS receivers and calibrate to neutral for data collection runs (like zero'ing a scale with a empty tub on when cooking to just measure the tub contents and not the tub).
There's an aversion to unshielded equipment, extranuous broadcasts, and pilots wearing glow underwater divers watches with radium dials.
Sure, we kept it off during aquisition on all 14 airframes for hundreds of millions of line kilometres gathered across multiple countries; Mali, Fiji, PNG, Central and Northern Australia, the DRCs, various 'stans, India | Pakistan and elsewhere.
Perhaps others had different practices or flew different instruments such as optical photogrammetry.
If you're seeing gridding at 80m ground clearence perhaps it's crop spraying or perhaps your software is infilling lines from turnpoint position blipping with nothing for the next 50km.
EDIT: I note you submitted this article and have an interest in aerial surveillance programs, I'm simply recounting the practices of one rather busy Australian geophysics company thta was subsumed into Fugro some time past. We considered aircraft and crew safety and had a 20+ year no crash no lives lost record and regularly considered aircraft position signals .. that said we avoided "polluting" the spectrum with broadcasts when aquiring signal and largely relied on off grid checkins (even just three radio blips) ever few completed lines.
I dare say some operators contracted for non EM work (say, radiometric survey) might broadcast during aquisition lines but I find it hard to believe that a good operator would broadcast during EM runs.
I was thinking of airborne electromagnetic (AEM) surveys (e.g., https://www.ramboll.com/projects/water/aem-mapping-californi...) when I wrote that, but after checking my notes, I found that I had misremembered: I actually hadn't been able to find any ADS-B tracks that were clearly survey flights!
That tracks .. geophys surveys, whether active (sending out signal to ground and listening to the return) or passive (bat airs across the spectrums), are predicated on having as 'neutral' a recording post as possible .. blatting out radio spectrum data that would have to be removed from the record in post just isn't "the way".
Aeromag and radiometrics tend to be low level and tight; 80m ground clearance, 200m line spacing (unless for regional levelling of multiple small infill surveys).
Photogrammatic sureys are higher with much wider spacing, the frames are large and there's really only need for a 50m or less overlap to get a good broad area mosaic.
I'm guessing you have routines to analyse and classify flight activity by path characteristics and coarsely bucket activity?
It would be much less costly to implement even wider spread use of ADS-B than already exists (down to very small size of aircraft), and increase the number of ADS-B receiving satellites in low earth orbit (which already exist, and intelligence agencies also have their own), than to attempt to do this in the strictly visual spectrum, based on frequent repeat visits of a particular few square km of sky imagery analysis (multiplied by the vast area of square km of sky that is the planet).
Radar would probably work better, assuming that works through the atmosphere, and assuming it should act as a backup to ADS. That could work on ships as well, which also need to be tracked due to recent "accidental dragging of anchor" incidents.
Radar has been used on satellites for ocean surveillance before but it takes a lot of power. That means carrying a nuclear reactor, or enormous solar panels plus batteries. And a huge constellation would be needed to achieve anything like continuous coverage.
It probably wouldn't be feasible to have a single constellation tracking both aircraft and ships. Different types of radar operating on different frequencies would be needed for those applications.
