Impacts of Large Satellite Constellations on Astronomy

[envy887]

If confirmed in the next few hours/days, finally good news even if it's just a small step towards an actual working mitigation. From the Reddit post:

So, if confirmed, a significant decrease in brightness and probably enough to knock most of the Starlinks (with that coating) below naked-eye visibility when at 550km. It wouldn't make much difference to the effect on astronomical observations.

The surface treatment (presumably to some part of the Earth-facing base of the Starlink) does not change the Starlink brightness when in the low-drag configuration.

The reduction in magnitude I saw suggests that only a portion of the earth-facing base of Darksat has been coated in anything vaguely black. Changing the coating of the Ku and Ka-band antenna panels on that face is probably a bigger deal and will take longer to re-qualify. If the whole of that panel could be brought down to a few percent reflectivity another 2 optical magnitudes might be knocked off.

That last part would be a bigger deal.

5.7 if confirmed, would be a very good start - they would only be naked-eye visible under very good seeing conditions and when well above the horizon. And even one more magnitude of dimming would put the 550 km shell well below the visible limit for pretty much any elevation.

The overall impact on astronomy and astrophotography would be lessened, but not to the extent as the impact on visual observing.

[Asteroza]

The optics of this issue (no pun intended) means DarkSat being below human visibility shifts the conversation from the average person thinking their night sky is ruined to "I can't see it, so it isn't a problem, why are you complaining?", which from a PR/social pressure standpoint may be sufficient to take the heat off Starlink.

[thirtyone]

Anyone find any new posts on Darksat observations yet from astronomers? Haven't caught any reposts on twitter for the usual sources. Think the threshold that's important at this point is quite a bit beyond visual observability, so I'd imagine proper equipment is necessary to determine how far it is from necessary brightness.

[envy887]

Anyone find any new posts on Darksat observations yet from astronomers? Haven't caught any reposts on twitter for the usual sources. Think the threshold that's important at this point is quite a bit beyond visual observability, so I'd imagine proper equipment is necessary to determine how far it is from necessary brightness.

The Reddit post eergo quoted below has been updated:

[Edit: Observed Darksat again on morning of 1st March. Estimate this time for Darksat brightness was 5.9mag at an elevation of 70degrees. It was 1.2mags fainter than a number of other L1.2 Starlinks observed just before, so a reduction in luminosity of a factor 3.]

This is promising for naked-eye viewers, although not enough to resolve all of the interference with observations.

https://www.reddit.com/r/Starlink/comments/f9704t/measurement_of_the_brightness_of_darksat_in_its/

[su27k]

Recent article on Science: Satellite megaconstellations menace giant survey telescope

It's behind paywall, some tidbits:
1. Study shows 1/3 of the images from Rubin Observatory (formerly LSST) during parts of the night would be ruined. (Although it's not clear what is the constellation size they're assuming)
2. Most telescopes can avoid Starlink, but if it has two out of three characteristics it would be in trouble: long exposure, big mirror, wide field of view
3. Current plan is for Starlink to be darkened by a factor of 15, which would mitigate the issue for Rubin Observatory. SpaceX is working towards a design that can do this, there will be more darkened satellites with revised design in the coming launches.
4. OneWeb will need to be darkened too.

[thirtyone]

Recent article on Science: Satellite megaconstellations menace giant survey telescope

It's behind paywall, some tidbits:
1. Study shows 1/3 of the images from Rubin Observatory (formerly LSST) during parts of the night would be ruined. (Although it's not clear what is the constellation size they're assuming)
2. Most telescopes can avoid Starlink, but if it has two out of three characteristics it would be in trouble: long exposure, big mirror, wide field of view
3. Current plan is for Starlink to be darkened by a factor of 15, which would mitigate the issue for Rubin Observatory. SpaceX is working towards a design that can do this, there will be more darkened satellites with revised design in the coming launches.
4. OneWeb will need to be darkened too.

One more really interesting statement from the article - SpaceX plans to launch "several satellites" with an "updated dark design" in the coming weeks - so probably next or next next launch (article published on the 28th).

[Dizzy_RHESSI]

ESO have released a study today on the impacts to their visible and infrared facilities.

https://www.eso.org/public/news/eso2004/

The paper:

https://arxiv.org/abs/2003.01992v1

[su27k]

Layman's version in the form of a BBC news article: Europe's major telescopes 'can meet satellite challenge'

Up to about 100 satellites could potentially be bright enough to be visible with the naked eye during twilight hours, about 10 of which would be higher than 30 degrees of elevation.

...

"Once you crunch the numbers, the numbers are not as bad as what some people feared," Dr Hainaut told BBC News. "People said 'oh no, there'll be 40,000 bright satellites in the sky? No, there won't be, simply because most of these satellites will be below the horizon, and then most of the others will be in the shadow of the Earth. That's the first bit of good news.

I'm glad an astronomer finally put this incorrect notion to rest.

[eeergo]

Layman's version in the form of a BBC news article: Europe's major telescopes 'can meet satellite challenge'

Up to about 100 satellites could potentially be bright enough to be visible with the naked eye during twilight hours, about 10 of which would be higher than 30 degrees of elevation.
"Once you crunch the numbers, the numbers are not as bad as what some people feared," Dr Hainaut told BBC News. "People said 'oh no, there'll be 40,000 bright satellites in the sky? No, there won't be, simply because most of these satellites will be below the horizon, and then most of the others will be in the shadow of the Earth. That's the first bit of good news.

I'm glad an astronomer finally put this incorrect notion to rest.

That notion was quickly and clearly dismissed in a few posts here. It's by the way pretty obvious even for the most mathematically impaired that you can't have 40k satellites in the visible sky at once if you're considering 26k to start with

The study is nice and indeed alleviates many fears. It also shows the availability to dialogue shown by the major players is not faltering. Still, in my view:

- on the one hand, it misses some of the most critical aspects of these endeavors: it ignores from the get-go the effect of satellites in deployment/decommissioning/maintenance low orbits.

This is fine if you want to address the steady-state ideal system, but it will never be like that in reality. It's like modeling traffic in a city with cars never stopping anywhere - it's useful for some basic understanding, but you'll never approach a realistic representation like that. In fact, it is reasonable to believe at least 5-10% of satellites will be in that low-orbit, low-drag situation where they are extremely visible and illuminated by barely less time than in the operational config, at any given time. Crudely extrapolating total/visible satellites for the steady case, this would mean between 1000-3000 satellites in total, of which 50-150 would be *very* visible in the sky at once, in the best case of homogeneous dispersion (obviously many more if they're still close together after deployment). Multiply that by 2-4 for the case of full-scale constellations.

- it also assumes a conservative case of 26k satellites from 14 constellations (12k Starlinks, 11k for 3 other major providers together, 4k for the remaining 10 providers). A single provider is already planning to more than *triple* that share!

- on the other hand, even with the 26k moderately optimistic assumption that ignores high-visibility units, while the title of the BBC article is true for VIS/NIR narrow and medium-field short/medium-exposure observations, is NOT the take-away message from the study, which mentions much more noticeable effects for long exposures and wide/very wide FOV telescopes, not addressable with manageable fixes.

[pochimax]

Layman's version in the form of a BBC news article: Europe's major telescopes 'can meet satellite challenge'

Up to about 100 satellites could potentially be bright enough to be visible with the naked eye during twilight hours, about 10 of which would be higher than 30 degrees of elevation.

...

"Once you crunch the numbers, the numbers are not as bad as what some people feared," Dr Hainaut told BBC News. "People said 'oh no, there'll be 40,000 bright satellites in the sky? No, there won't be, simply because most of these satellites will be below the horizon, and then most of the others will be in the shadow of the Earth. That's the first bit of good news.

I'm glad an astronomer finally put this incorrect notion to rest.

This "incorrect" notion is only completely true in the visible part of the spectrum. In the infrared domain they are very bright sources. The nearer to the Earth, the brighter.

Interesting that satellites eclipsed by the earth (so, not visible) are sources of occultation (they shadow the sky as they travel the sky sphere)

Fortunately, not important excepto for wide field telescopes.

Pending another paper on radio astronomy. I' m worried about sub millimeter astronomy. Starlink could be very bright infrarred objects, as high as 100 Jansky at 10 microns.

[edzieba]

Layman's version in the form of a BBC news article: Europe's major telescopes 'can meet satellite challenge'

Up to about 100 satellites could potentially be bright enough to be visible with the naked eye during twilight hours, about 10 of which would be higher than 30 degrees of elevation.

...

"Once you crunch the numbers, the numbers are not as bad as what some people feared," Dr Hainaut told BBC News. "People said 'oh no, there'll be 40,000 bright satellites in the sky? No, there won't be, simply because most of these satellites will be below the horizon, and then most of the others will be in the shadow of the Earth. That's the first bit of good news.

I'm glad an astronomer finally put this incorrect notion to rest.

This "incorrect" notion is only completely true in the visible part of the spectrum. In the infrared domain they are very bright sources. The nearer to the Earth, the brighter.

Unless you're observing intense Neutrino emissions, you're not going to be seeing any satellites that are below the local horizon. X satellites in orbit does not mean X satellites visible from any given point on the surface, no matter what wavelength you are observing in.

Radio astronomy has been mentioned upthread: they are more concerned about active SAR satellites that can permanently damage radio telescope hardware.

[envy887]

This is fine if you want to address the steady-state ideal system, but it will never be like that in reality. It's like modeling traffic in a city with cars never stopping anywhere - it's useful for some basic understanding, but you'll never approach a realistic representation like that. In fact, it is reasonable to believe at least 5-10% of satellites will be in that low-orbit, low-drag situation where they are extremely visible and illuminated by barely less time than in the operational config, at any given time. Crudely extrapolating total/visible satellites for the steady case, this would mean between 1000-3000 satellites in total, of which 50-150 would be *very* visible in the sky at once, in the best case of homogeneous dispersion (obviously many more if they're still close together after deployment). Multiply that by 2-4 for the case of full-scale constellations.

Homogenous dispersion isn't "best case" IMO. If the sats are closely grouped as in the trains, it much easier to avoid them.

And lower sats do go dark much earlier in the night, and light up later in the morning. Typically they are dark before astronomical twilight. See Pat Seitzer's presentation linked upthread.

And it may be entirely feasible to darken the orbit-raising configuration. AFAIK, SpaceX hasn't tried, because the benefit is much lower due to the satellites spending less than 2% of their lifetime in that configuration.

Also, the orbit-raising configuration is "low drag". In the deorbit/decommissioning phase, they will want "high drag", which may be much darker than "low drag" and possibly even darker than operational orientation.

[eeergo]

This is fine if you want to address the steady-state ideal system, but it will never be like that in reality. It's like modeling traffic in a city with cars never stopping anywhere - it's useful for some basic understanding, but you'll never approach a realistic representation like that. In fact, it is reasonable to believe at least 5-10% of satellites will be in that low-orbit, low-drag situation where they are extremely visible and illuminated by barely less time than in the operational config, at any given time. Crudely extrapolating total/visible satellites for the steady case, this would mean between 1000-3000 satellites in total, of which 50-150 would be *very* visible in the sky at once, in the best case of homogeneous dispersion (obviously many more if they're still close together after deployment). Multiply that by 2-4 for the case of full-scale constellations.

Homogenous dispersion isn't "best case" IMO. If the sats are closely grouped as in the trains, it much easier to avoid them.

And lower sats do go dark much earlier in the night, and light up later in the morning. Typically they are dark before astronomical twilight. See Pat Seitzer's presentation linked upthread.

And it may be entirely feasible to darken the orbit-raising configuration. AFAIK, SpaceX hasn't tried, because the benefit is much lower due to the satellites spending less than 2% of their lifetime in that configuration.

Also, the orbit-raising configuration is "low drag". In the deorbit/decommissioning phase, they will want "high drag", which may be much darker than "low drag" and possibly even darker than operational orientation.

Please reread the sentence you quote: homogeneous spacing is *best*-case for the situation I was explaining, where I estimate the number of satellites visible *contemporarily*. It means the satellites are at their maximum relative separation, so it will be the hardest to find large numbers together. The closer they are, the easier they are to avoid them all *on average*, but the most difficult during a certain finite period.

Of course they fall out of illumination earlier the lower they are. My point is the difference in visibility counters or even outweighs that advantage for large areas of the sky, which is something the paper doesn't study but is hugely relevant.

What percentage of their lifetime a satellite spends in low drag and/or low orbits is irrelevant, the important point is how many of them will be doing it at once. I haven't seen visibility data for "high drag"/decommissioning satellites, although I understand there are 12 of them undergoing the process at this time. Do you know of any observation suggesting lowered visibility for those? Most likely though, future decommissioned satellites will not hold attitude anyway - high drag means stronger torques on the spacecraft, which if they fulfilled their mission and are at end-of-life, they will not have enough RCS to counter for long. Unless deorbit periods are extremely quick of course, but so far they aren't following that route.

[envy887]

What percentage of their lifetime a satellite spends in low drag and/or low orbits is irrelevant, the important point is how many of them will be doing it at once.

Since the lifetime of the satellite and the amount of time SpaceX has and likely needs to build out the 12k constellation as both ~6 years, these are essentially the same thing. By the time they have but the entire constellation, they will need to start replacing the first satellites launched, at nearly the same rate. So percentage of lifetime is a good approximation for percentage in orbit-raising.

I haven't seen visibility data for "high drag"/decommissioning satellites, although I understand there are 12 of them undergoing the process at this time. Do you know of any observation suggesting lowered visibility for those? Most likely though, future decommissioned satellites will not hold attitude anyway - high drag means stronger torques on the spacecraft, which if they fulfilled their mission and are at end-of-life, they will not have enough RCS to counter for long. Unless deorbit periods are extremely quick of course, but so far they aren't following that route.

The one satellite that has deorbited so far appears to have been actively changing altitude until very shortly before reentry, which would likely means it held attitude during the entire decommissioning phase (the ion thruster isn't going to be able to do net altitude changes in a random tumble). Attitude control is done by moment wheels and mag detorquers since Starlinks have no RCS, so attitude should be held up until the point where drag imbalance overwhelms the system, which is probably low enough that reentry happens within a couple days.

This stuff should all be studied, but IMO these are all edge cases and unlikely to have nearly the same impact on the totality of observations as the operational part of the constellation will.

[eeergo]

What percentage of their lifetime a satellite spends in low drag and/or low orbits is irrelevant, the important point is how many of them will be doing it at once.

Since the lifetime of the satellite and the amount of time SpaceX has and likely needs to build out the 12k constellation as both ~6 years, these are essentially the same thing. By the time they have but the entire constellation, they will need to start replacing the first satellites launched, at nearly the same rate. So percentage of lifetime is a good approximation for percentage in orbit-raising.

Thanks for those numbers, I wasn't aware of them in such a clear way.

So 12000/6 = 2000 a year makes 2-3 launches a month. At the current rate of orbit-raising where it takes them 2-3 months to get to low-drag attitude/altitude, it means 240-540 satellites in upward transit at any given time. Multiply that by at least 2 because you'll get down-going decommissioned units in this steady-state replenish-decommission scenario, and you get 500-1200 units out of low drag.

The study takes into account 14 different constellations which even in a perfect world wouldn't line up timings so harmoniously. But to build on its assumptions, let's just scale the result: Starlink's 12k are around 40% of the 27k units considered there - so you'd end up with around a bit more than twice as many in upward/downward transit at any given time: 1000-2400 units.

I estimated 1000-3000 (5-10%), which as you can see is actually pretty much spot on when applying the projected launch/deorbit rates.

I haven't seen visibility data for "high drag"/decommissioning satellites[...]

The one satellite that has deorbited so far appears to have been actively changing altitude until very shortly before reentry, which would likely means it held attitude during the entire decommissioning phase (the ion thruster isn't going to be able to do net altitude changes in a random tumble). Attitude control is done by moment wheels and mag detorquers since Starlinks have no RCS, so attitude should be held up until the point where drag imbalance overwhelms the system, which is probably low enough that reentry happens within a couple days.

A single, fresh-out-of-Falcon v0.9 experimental satellite, first to be deorbited, is probably not a good approximation of an actual nominally-decommissioned unit. A non-negligible number of these units undergoing deorbit will likely fail on transit. We don't know the control authority of those wheels/rods, if they plan on beefing them up or lower their specs to operational needs, or even if they plan to add RCS. Visibility data for early decommissioned units should be available soon though, with 11 more v0.9s and a v1.0 undergoing the same process right now.

This stuff should all be studied, but IMO these are all edge cases and unlikely to have nearly the same impact on the totality of observations as the operational part of the constellation will.

As per my previous posts, I am showing this is not a second-order correction.

[envy887]

The Starlink v0.9 sat that deorbited showed a sustained multi-day descent rate at least 4 different times that would bring it all the way down from 550 km in ~35 days or faster. This implies that from 1100 km will take ~2 months, but from 300 km it takes less than a week. That means that in a 12k constellation with 6 year lifespan, only ~200 will ever need to be deorbiting worldwide at any given time. The suggestion that they might revert to a slower, less controlled deorbit method is highly implausible at best. The whole point of deorbiting is to get down in a quick and controlled manner, and the one satellite that deorbited already proved they can do that.

Dead at injection or worse, dead in operation satellites are probably a bigger issue. v0.9 looks like it had one DOA and 1 failure after ~60 days. The first 2 operational launches appear to have had 1 failure each, after ~30 and ~40 days. The third and fourth launches appear to have had no failures so far, though data is limited. So early mortality looks like about 4/300 or ~1% so far. This will probably improve as each batch incorporates development advances. None of the satellites have failed after reaching operational altitude, totaling 46 satellite-years in operation so far.

That implies a failure rate under 1.5%. Assume that the ~1% initial plus >1.5% operational failures (which is an upper bound from observations so far) remain in steady state, then this implies that over 6 years of operational life, a 12k constellation would have  1500 failures at 330 km, 140 failures at 550 km, and 240 failures at 1100 km. The VLEO failures would deorbit in ~1 month, so only 15 would be visible at any time. The 550 km failures would deorbit in ~5 years, so ~120 would be visible at any given time.  These 130 or so dead satellites from the lower shells are not a significant issue for observations. This is a steady state number, because atmospheric decay removes them faster than than new ones fail.

The 1100 km failures would stay up for many years, adding ~400 dead objects per decade. This isn't an immediate major issue, but I'd like to see SpaceX either commit to retrieving these actively, or move all the shells under 600 km.

So I'm getting ~200 from deorbiting, plus 130 from low failures, plus 240 from high failures, all visible at once. That is 570 or so altogether.

Graphs modified from: https://twitter.com/StarlinkUpdates/status/1235218498579820546

[envy887]

For the purposes of the study, assessing 26k was wise, but I'm HIGHLY dubious that the vast majority of those will be launched within a decade. The vast majority have no funding, no spectrum rights, and/or no available launch capacity, and no business case. Maybe those things will change e.g. with Starship coming online, but pretty soon here Starlink is going to provide the first of real-world data on 1000+ bird constellations. And that data will be assessed long before the other very large constellations get anywhere near a lauchpad.

In light of the issues Starlink has raised to date, and the observations of it in operation which will be coming, and studies like this recent one, I'm expecting the FCC and others to require consideration of the impact on astronomy, and working with optical astronomers in a manner similar to the requirement to work with radio astronomers, before approving spectrum or landing rights.

[eeergo]

The Starlink v0.9 sat that deorbited showed a sustained multi-day descent rate at least 4 different times that would bring it all the way down from 550 km in ~35 days or faster.
[...]
So I'm getting ~200 from deorbiting, plus 130 from low failures, plus 240 from high failures, all visible at once. That is 570 or so altogether.

I don't necessarily agree with those assumptions based on large-number projections of data from single-(or zero-) case statistics, but it's interesting we keep getting similar numbers even after adding up constraints and using different assumptions (330-570 downward-going yourself, between 240-540 myself, plus as many upward going ones). So again not a second-order issue.

As for the deployment/regulatory problems they mighy encounter of your second post, I agree with your assessment (minus SS ), but that's finance and politics - the concern is about the plans and their consequences.

[su27k]

- on the one hand, it misses some of the most critical aspects of these endeavors: it ignores from the get-go the effect of satellites in deployment/decommissioning/maintenance low orbits.

This is fine if you want to address the steady-state ideal system, but it will never be like that in reality. It's like modeling traffic in a city with cars never stopping anywhere - it's useful for some basic understanding, but you'll never approach a realistic representation like that. In fact, it is reasonable to believe at least 5-10% of satellites will be in that low-orbit, low-drag situation where they are extremely visible and illuminated by barely less time than in the operational config, at any given time. Crudely extrapolating total/visible satellites for the steady case, this would mean between 1000-3000 satellites in total, of which 50-150 would be *very* visible in the sky at once, in the best case of homogeneous dispersion (obviously many more if they're still close together after deployment). Multiply that by 2-4 for the case of full-scale constellations.

I'm not sure why satellites in deployment/decommissioning phase matters if it's just 5-10% of the total, that's a rounding error in this kind of ballpark estimates.

[NaN]

It's nice to see a considered analysis with all of the assumptions spelled out. The study wisely makes a number of pessimistic assumptions; notably it assumes a brightness similar to the current v1.0 Starlinks - adjusted for altitude - for all satellites. If SpaceX is able to roll darksat-like improvements into deployments this year and next year, as appears likely, then this becomes a pessimistic assumption (which is good news). Studies such as this can form the foundation of recommendations which are considered as a baseline for future constellations.

What the study does underscore is that LSST will be impacted significantly worse than other cases - severely so if tens of thousands more non-darkened LEO sats are launched - and treating it as the "high water mark" remains valid overall.