Impacts of Large Satellite Constellations on Astronomy

[DigitalMan]

That is a good presentation. I am curious about the OneWeb satellites. It seemed like they implied OneWeb will de-orbit satellites that are taken out of service. Satellites that die before that would take 'centuries' to decay, if I recall the presentation correctly.

I thought Greg had indicated that satellites taken out of service would go into a graveyard orbit? I don't know if that plan also applies to the extra 48,000 that were applied for but if so is anyone concerned about that? How long does it take to decay from this graveyard orbit?

It's not clear what is OneWeb's end of life plan for the 48,000 satellites, they didn't present their deorbit plan in the FCC filing as far as I can see, they merely said they will follow regulation of UK Space Agency instead of the US space debris rules.

I was going from my recollection of Greg’s testimony to Congress, but that was long ago.

[FutureSpaceTourist]

https://twitter.com/planet4589/status/1277999718652657664

#SATCON1  Tony Tyson answers my question - why does he think the problem scales with etendue rather than just field of view?  He says because surface brightness is what counts.  (But I think it's surface brightness per pixel, so I don't really agree with him)

twitter.com/planet4589/status/1278002388545286145

If you have a 6 inch telescope with a very wide field of view and a small CCD,  I expect you will have nothing but streaks, even though its etendue is far smaller than Rubin

https://twitter.com/planet4589/status/1278012091245498370

I think Tyson's point is that the nasty CCD cross talk effects only show up in large etendue systems like Rubin. But the first-order streak effects don't scale with etendue, I believe.

[FutureSpaceTourist]

https://twitter.com/planet4589/status/1278027136721399808

#SATCON1 Jared Greene (SpaceX) we have done 4 or 5 experimental [Starlinks] with smaller component-level changes [not just Darksat and Visorsat] to understand mitigations. Collaboration with Rubin Obs folks very helpful in directing strategy

https://twitter.com/planet4589/status/1278033059850399745

#SATCON1 Jared Greene:  the issue with Darksat is the antennas absorb more Earth heat, so there's a thermal problem [hence visors are better]

https://twitter.com/planet4589/status/1278034067422941185

#SATCON1 Jared Green (SpaceX): all future sats launched (modulo maybe occasional experiment) will have visors

[eeergo]

https://twitter.com/cosmos4u/status/1279034879284559873


https://twitter.com/cosmos4u/status/1279042302812528640

[LouScheffer]

The astronomers were VERY happy when OneWeb went bankrupt.  As above, for astronomy, the higher the orbit the worse, since the satellites are illuminated longer after sunset and earlier before dawn.  1200 km orbits are MUCH worse than those lower SpaceX orbits, and SpaceX is already working with the astronomers to reduce the impact.  No such cooperation exists yet with OneWeb, and the astronomers are worried.

[RedLineTrain]

Some of these analyses are pretty jarring.  Like "LEOsats >>600km are incompatible with optical astronomy."

https://twitter.com/cosmos4u/status/1279065804898742276

[DigitalMan]

Having control of a world wide sensor layer can go a long way to shore up credibility a lot cheaper than a new carrier or sub. Additionally, there will shortly be a political need to be able to point at some economic good news, any economic good news. Being able to say the UK has become a world leader in space capabilities for cheaper than say a program to hire more teachers has much political sense. In simply political terms the 500m price tag is amazing value for money.

I'm not sure I follow your argument.  When you say "world wide sensor layer", I don't think OneWeb, I think Planet Labs.  Are you imagining the UK developing another payload to put on the OneWeb bus that would do imaging or some other remote sensing?  It's not at all clear that buying a bankrupt communications business is the best value for the money if you want a sensor network, not a communications network.

Or are you saying that they could spy on the communications that go on over the OneWeb network?  Wouldn't encryption largely negate the value there?

First remember politics and reality do not always coincide. The ability to say "Our wise investment has made us a world leader in space" does not have to reflect technological truth to be politically valuable.

The UK is already claiming they will put SatNav hardware on the birds to justify the purchase. At which point "sensor layer" is mostly a question of what gets put on the bird. Even the pure communications system has strong implications for modern military credibility particularly in the era of drone swarms.

Just as a thought experiment try this relationship between the parties: Bhati gets a constellation and help opening markets while the UK get military coms and maybe 20kg of SatNav/ secret equipment on each bird. The announced relationship would likely look exactly like the one we have seen.

Except the public doesn't know OneWeb. There was no need for the government to save it, no public outrage if it didn't.

Airbus has built the UK's military sats in the past, and it owns half of OneWeb satellites, so I'm not sure what capability the UK acquired that it didn't have access to anyway.

I'm interested to see if they will make an effort to deal with light pollution issues.

Has there been any affects on astronomical observations from the existing satellites? Because we have seen photographs clearly demonstrating this for starlink...

see below:
https://www.theverge.com/2020/3/24/21190273/spacex-starlink-satellite-internet-constellation-astronomy-coating

It seems that their satellites already have a reduced optical signature in relation to starlink. I don't know what it is. Perhaps it is the insulation applied to the outside of the body that is not flat that scatters light. You can see the outside surfaces of starlink and oneweb here:
https://spaceflightnow.com/2020/07/03/uk-government-commits-500-million-in-bid-to-rescue-bankrupt-oneweb/
https://www.starlink.com/

Regardless, OneWeb's entire gen 1 constellation is similar in count to the unmitigated starlink satellites.

From what I understand only a few of OneWebs satellites are at target orbit. Any observations at this point are not representative.

[thirtyone]

Re: OneWeb visibility, I recommend reading up on the "Impacts of Large Satellite Constellations on Astronomy." Basically, OneWeb's satellites are right in the sweet spot where they cannot be seen unaided, so they won't contribute to light pollution, but will cause major damage to the most cutting edge astronomical observatories. The detectors in such observatories can be saturated by light and cause nonlinear distortion to adjacent pixels (or even entire rows/columns) well below visible thresholds. This is largely because of their higher operating orbit, which, somewhat paradoxically, creates more damage to some astronomical experiments the satellites are visible all night in a larger swath of the sky, despite their lower brightness.

But by far the biggest practical problem is that probably because OneWeb had been on the verge of bankruptcy, they didn't even have the resources to even engage the astronomical community about ways to mitigate or even characterize the problem.

If we are talking purely about light pollution for the average person looking up, though, there's sort of no question OneWeb is less visible than Starlink, because you really can't see them without instruments, even at deployment orbit. They're just smaller birds and quite a bit higher in orbit.

At least one astronomer has commented on twitter specifically saying that OneWeb is expected to be 2-3 mag fainter than Starlink, simply based on weight and altitude:
https://twitter.com/planet4589/status/1167047947584004097

[DigitalMan]

Re: OneWeb visibility, I recommend reading up on the "Impacts of Large Satellite Constellations on Astronomy." Basically, OneWeb's satellites are right in the sweet spot where they cannot be seen unaided, so they won't contribute to light pollution, but will cause major damage to the most cutting edge astronomical observatories. The detectors in such observatories can be saturated by light and cause nonlinear distortion to adjacent pixels (or even entire rows/columns) well below visible thresholds. This is largely because of their higher operating orbit, which, somewhat paradoxically, creates more damage to some astronomical experiments the satellites are visible all night in a larger swath of the sky, despite their lower brightness.

But by far the biggest practical problem is that probably because OneWeb had been on the verge of bankruptcy, they didn't even have the resources to even engage the astronomical community about ways to mitigate or even characterize the problem.

If we are talking purely about light pollution for the average person looking up, though, there's sort of no question OneWeb is less visible than Starlink, because you really can't see them without instruments, even at deployment orbit. They're just smaller birds and quite a bit higher in orbit.

At least one astronomer has commented on twitter specifically saying that OneWeb is expected to be 2-3 mag fainter than Starlink, simply based on weight and altitude:
https://twitter.com/planet4589/status/1167047947584004097

That tweet perhaps was wishful thinking in light of one of his later tweets:
https://twitter.com/planet4589/status/1277994082896424961

[ncb1397]

From what I understand only a few of OneWebs satellites are at target orbit. Any observations at this point are not representative.

Here is my list that I compiled...
https://www.n2yo.com/satellite/?s=44062
https://www.n2yo.com/satellite/?s=44058
https://www.n2yo.com/satellite/?s=44059
https://www.n2yo.com/satellite/?s=44061
https://www.n2yo.com/satellite/?s=45131

Could have missed a couple. Seems like these should be observable and astronomers can gather basic information like magnitude? I mean magnitude is a simple thing to generate. It just seems bizarre that we don't even have that?

edit: wait, I think I found it...

• 1584 Starlinks just the start.
– SpaceX: 12,000? 42,000? At 550 km, observed V ~ 5th.
• SpaceX filed with FCC to replace 2,825 satellites @ 1,110-1,325 km with 2,824 satellites @ 540- 560 km.
– Amazon: filed for 3,236 at 590, 610, and 630 km.
– OneWeb: initially ~700, grow to 1980 (at 1200 km). At 1200 km, observed V ~ 8th

https://www.nationalacademies.org/event/04-27-2020/docs/D8142CA54A24C066B0A7BF5C749227DBA5420F633FF3

I assume this means observed visibility was 8th magnitude? Starlink is 5th magnitude.

[abaddon]

Additional to above, OneWeb has apparently done nothing at all to work with radio astronomers, whom SpaceX had been working with even before the first operational launch.  (The visibility of the satellites came as a surprise to everyone, whereas radio interference was expected).  There's a lot of astronomy done outside the visible light spectrum.

With the visor, IIRC Starlink observations dropped to about 7th magnitude, and keep in mind Starlink satellites, unlike OneWeb, are only illuminated at dawn and dusk, and only close to the horizon.

Further conversations should probably be in the dedicated thread.

[ncb1397]

Additional to above, OneWeb has apparently done nothing at all to work with radio astronomers, whom SpaceX had been working with even before the first operational launch.  (The visibility of the satellites came as a surprise to everyone, whereas radio interference was expected).  There's a lot of astronomy done outside the visible light spectrum.

With the visor, IIRC Starlink observations dropped to about 7th magnitude, and keep in mind Starlink satellites, unlike OneWeb, are only illuminated at dawn and dusk, and only close to the horizon.

Further conversations should probably be in the dedicated thread.

I think I found a counter example to this. See GMAT simulation output and script file. Plotting the indicated latitude and longitude in google earth, a generic satellite at 53 degree inclination and 550 km altitude is lit with only a ground track seperation of ~390 km (meaning the elevation angle would be 45 degrees or more). UTC time is 7:30 meaning about ~3 am solar time at La Silla Observatory (labelled as GroundStation1 in the ground plot). I guess we should more clearly different what "close to the horizon" means. Elevation of >45 degrees is closer to zenith than the horizon.

[meberbs]

Additional to above, OneWeb has apparently done nothing at all to work with radio astronomers, whom SpaceX had been working with even before the first operational launch.  (The visibility of the satellites came as a surprise to everyone, whereas radio interference was expected).  There's a lot of astronomy done outside the visible light spectrum.

With the visor, IIRC Starlink observations dropped to about 7th magnitude, and keep in mind Starlink satellites, unlike OneWeb, are only illuminated at dawn and dusk, and only close to the horizon.

Further conversations should probably be in the dedicated thread.

I think I found a counter example to this. See GMAT simulation output and script file. Plotting the indicated latitude and longitude in google earth, a generic satellite at 53 degree inclination and 550 km altitude is lit with only a ground track seperation of ~390 km (meaning the elevation angle would be 45 degrees or more). UTC time is 7:30 meaning about ~3 am solar time at La Silla Observatory (labelled as GroundStation1 in the ground plot). I guess we should more clearly different what "close to the horizon" means. Elevation of >45 degrees is closer to zenith than the horizon.

Please look up thread at the analysis that has been done already. The regions significantly affected by something at Starlink altitude are as described above. One specially chosen case does not change this, and it certainly doesn't change the fact that satellites at OneWeb altitudes cause many more problems throughout the whole sky and whole night.

Edit: fix post title

[ncb1397]

Additional to above, OneWeb has apparently done nothing at all to work with radio astronomers, whom SpaceX had been working with even before the first operational launch.  (The visibility of the satellites came as a surprise to everyone, whereas radio interference was expected).  There's a lot of astronomy done outside the visible light spectrum.

With the visor, IIRC Starlink observations dropped to about 7th magnitude, and keep in mind Starlink satellites, unlike OneWeb, are only illuminated at dawn and dusk, and only close to the horizon.

Further conversations should probably be in the dedicated thread.

I think I found a counter example to this. See GMAT simulation output and script file. Plotting the indicated latitude and longitude in google earth, a generic satellite at 53 degree inclination and 550 km altitude is lit with only a ground track seperation of ~390 km (meaning the elevation angle would be 45 degrees or more). UTC time is 7:30 meaning about ~3 am solar time at La Silla Observatory (labelled as GroundStation1 in the ground plot). I guess we should more clearly different what "close to the horizon" means. Elevation of >45 degrees is closer to zenith than the horizon.

Please look up thread at the analysis that has been done already. The regions significantly affected by something at Starlink altitude are as described above. One specially chosen case does not change this, and it certainly doesn't change the fact that satellites at OneWeb altitudes cause many more problems throughout the whole sky and whole night.

It is not a specially chosen case. The simulation was run with one randomly placed satellite for 100 orbits or ~6 days. This yielded 1 instance of a problematic transit (at least in terms of the description above about what can and cannot happen). Presumably, this suggests that with ~1500 satellites, you would get a lot more of these (like hundreds per night). The correct description of starlink is probably something more like this....

Lit satellites don't show up at high elevation in the deepest part of the night at typical observatory latitudes. They can however show up at high elevation a few hours before and after sunset (dependent on time of year and latitude).

And beyond that, the description that OneWeb causes a problem throughout the whole sky and the whole night isn't accurate either. Getting lit transits with a northern azimuth in a southern latitude (or vice versa) doesn't tend to work at 550 km or 1200 km. The geometry just doesn't work like that.

Anyways, there are pros and cons to higher and lower orbits and OneWeb's and SpaceX's approach. An orbit twice as high will tend to generate 25% of the brightness for instance. It also requires less satellites (which means less orbital planes and less satellites per orbital plane or both). Less orbital planes at polar inclinations will be easier to time observation periods around than more planes at mixed inclinations (in OneWeb's cases, the orbital planes have a 20 degree seperation with seperation between satellites being 10 degrees). OneWeb is putting up ~600 and then they are going to be done with the current deployment period. Starlink has put up about that many and they aren't stopping (and can't stop if they want to provide global services).

[ChrisWilson68]

Additional to above, OneWeb has apparently done nothing at all to work with radio astronomers, whom SpaceX had been working with even before the first operational launch.  (The visibility of the satellites came as a surprise to everyone, whereas radio interference was expected).  There's a lot of astronomy done outside the visible light spectrum.

With the visor, IIRC Starlink observations dropped to about 7th magnitude, and keep in mind Starlink satellites, unlike OneWeb, are only illuminated at dawn and dusk, and only close to the horizon.

Further conversations should probably be in the dedicated thread.

I think I found a counter example to this. See GMAT simulation output and script file. Plotting the indicated latitude and longitude in google earth, a generic satellite at 53 degree inclination and 550 km altitude is lit with only a ground track seperation of ~390 km (meaning the elevation angle would be 45 degrees or more). UTC time is 7:30 meaning about ~3 am solar time at La Silla Observatory (labelled as GroundStation1 in the ground plot). I guess we should more clearly different what "close to the horizon" means. Elevation of >45 degrees is closer to zenith than the horizon.

Please look up thread at the analysis that has been done already. The regions significantly affected by something at Starlink altitude are as described above. One specially chosen case does not change this, and it certainly doesn't change the fact that satellites at OneWeb altitudes cause many more problems throughout the whole sky and whole night.

It is not a specially chosen case. The simulation was run with one randomly placed satellite for 100 orbits or ~6 days. This yielded 1 instance of a problematic transit (at least in terms of the description above about what can and cannot happen). Presumably, this suggests that with ~1500 satellites, you would get a lot more of these (like hundreds per night). The correct description of starlink is probably something more like this....

Lit satellites don't show up at high elevation in the deepest part of the night at typical observatory latitudes. They can however show up at high elevation a few hours before and after sunset (dependent on time of year and latitude).

Whether you chose it specially or not, it is an unusual, extreme case.  The date you used was close to the middle of summer for a location in the Southern Hemisphere.  And the time of day was between 2 and 2.5 hours before sunrise.

In mid summer the sun will be less far below the horizon closer to sunrise and sunset.

You found, by accident or not, pretty much the worst case for that particular location.  It is not typical.

And you characterized it as "a few hours" when in fact it's limited to less than 2.5 hours even in this extreme case.

[envy887]

OneWeb is putting up ~600 and then they are going to be done with the current deployment period. Starlink has put up about that many and they aren't stopping (and can't stop if they want to provide global services).

This is a moot point, since minimum coverage scales exactly with minimum visibility. The higher constellation can only have fewer satellites because a larger percentage of the total number of satellites is visible at any given time.

[thirtyone]

I recommend reviewing earlier posts, especially detailed analyses by astronomers in the AAS actively looking at the issue. The analysis is quite a bit more complex and the current expectation is that OneWeb is practically going to be more problematic, particularly if they continue to not engage with astronomers. The amount of damage done to a imager from a bright satellite at different altitudes is not as obvious as it seems and the working group seems to be settling on preferring megaconstellations at lower altitudes rather than higher ones. Most of the twitter posts and analyses above are from professional astronomers.

[ncb1397]

OneWeb is putting up ~600 and then they are going to be done with the current deployment period. Starlink has put up about that many and they aren't stopping (and can't stop if they want to provide global services).

This is a moot point, since minimum coverage scales exactly with minimum visibility. The higher constellation can only have fewer satellites because a larger percentage of the total number of satellites is visible at any given time.

The number of satellites may not be determinative. The question is how easy it is to dodge the planes as their arc moves through the sky. If it was just one plane with a bajillion satellites, that is very different than a lot of planes with less satellites (and potentially much worse even if the total count is lower). We already essentially have a megaconstellation within a singular plane - there are 400-500 active satellites in geosynchronous orbit but the combination of their distance and the fact their is very few planes (1 in this case), makes their effect easier to mitigate against than what we are dealing with now (same goes with Iridium with 6 orbital planes).

[ncb1397]

It is not a specially chosen case.

As pointed out above, the observer location and time of year are in fact special, and running a simulation for days and then pulling out a single instance is a method of cherry picking.

It is entirely consistent with other simulations...like this one which are all at or lower than 550 km.



The London simulation has lit near zenith transits in summer throughout the night (granted this is at 51 degrees north...farther from the equator than major observatories like La Silla, Paranal, Mauna Kea, etc.).

[ncb1397]

People were complaining that this was cherry picking, So, I did a full year simulation. The methodology is the following. The simulation starts on january 1st, 2020 and progresses at 1 minute intervals for practically the entire year (525600 steps). Each interval, the simulation stops and the longitude and latitude geofenced to only record data for spacecraft between 24 degrees south and 34 degrees south and 66 degrees west and 76 degrees west longitude. The ground station is positioned roughly in the center of this at 29.25 degrees south, 70 degrees west which isn't exactly centered on LSO(it is about 70 kilometers off but the exact ground station location is just an example of a place an observatory could be). In addition, the software only outputs data for when the solar panels are receiving power (and therefore it can be deduced that the spacecraft is lit). This then generates a text file consisting of a myriad of data including the positions of the sun, ground station and satellite in a common coordinate system centered on the earth which is then imported into a spreadsheet(in this case google sheets and is linked below) and the spacecraft elevation from the ground station is calculated and the sun angle is calculated for recorded data point. Any negative sun angle in the spreadsheet indicates a sun under the horizon (night time or twilight) and we can see the calculated elevation angle as well. There are numerous hits for high spacecraft elevation angles. The first hit found is 19th january 1:34 UTC with an elevation angle of 37 degrees and a sun angle of -21 degrees(line 26 in the spreadsheet). A minute later, the elevation has risen to about 66 degrees elevation, etc. On January 19th, sunrise is ~9:58 UTC and sunset is ~23:43 UTC meaning this is a few hours after sunset (also indicated by the -20 degree sun elevation angle). The next hit is the next night on January 20th, etc.

spreadsheet:
https://docs.google.com/spreadsheets/d/1226QfCLWKJHdfP1CV80WaUc7HnAGwRBYxmN2hTVtYfk/edit#gid=1848490928

Linked below is the GMAT configuration text file and the a screenshot of the first hit on January 19th paused in the software. Again, this is just with one satellite. With many satellites, you will get a lot more hits over a certain amount of time and we can think of this as a floor as the combination of the geofencing and the 1 minute simulation steps could miss some stuff (an object in low earth orbit would travel about 420 km in 1 minute). I welcome any errors found in the methodology or math. Any way to speed of the simulation and still get the same data would be helpful as well as the ~half million steps took hours (this could potentially increase the fidelity as well with shorter intervals).