Impacts of Large Satellite Constellations on Astronomy

[eeergo]

Since GEO is ~100x further away, and meter-sized objects are ~100x larger in area than 10 cm objects, this suggests that 10-cm class debris in lower LEOs will produce a streak, especially with very large telescopes, sufficient to negate the usefulness of the directly effected pixels on that exposure.

Of course that streak will be very much dimmer, which is probably better for the surrounding pixels and especially better for reducing scattering in the optics that affects the whole frame.

The bolded part is key. Anyway, you cannot take visibility of objects in GEO and extrapolate that to visibility in LEO because the more distant the orbit, the less apparent movement the object has... with the limit being GEO, where the object doesn't move with respect to the ground. There, integration time is your friend and you can resolve much dimmer objects. In LEO you'd have to track them.

Edit: 10-cm debris is NOT "exponentially the vast majority" of the catalog. According to CelesTrak, 7636 of the 20036 on-orbit tracked objects (38%) are either payloads or rocket bodies. While some of the payloads are cubesats in that range, most of those, and all the rocket bodies, are quite a bit larger.

Check out slide 19-20 of the link I published. Debris scales approximately exponentially with size for collision-generated fragments. The vast majority (~70%) of tracked objects are very small to register much on observations, and therefore not meaningfully comparable to tens of thousands of 15-meter satellites. I don't see an issue with my observation, which anyway was a comment on top of a much more important idea.

[envy887]

Concerning back-of-the-envelope estimates - here's an actual accurate simulation of visible satellites if the full 12k constellation was up in their operational orbits. The absurd sky dominance is considering worst-case illumination conditions in the summer - of course that's also when best weather is available for observations. And this is for *just* the baseline Starlink, in visible wavelengths and not accounting for under-maintenance sats (lower orbits). It is left as an exercise to the reader to add a few competing such constellations, such as the 5 already in the works (totaling approximately as many birds as a 12k Starlink).
[...]

Showing all the existing satellites/debris with oversized/overbrightness icons like that, instead of as point sources of realistic brightness, would produce a rather similar terrifying simulation which has equally little basis in reality.

Actually, it is quite grounded in reality. Of course you're not gonna see such objects with your naked eye, but observatories mostly WILL, as long as they're somewhat illuminated - that's the issue. And this is neglecting other proposed constellations or Starlink expansions (up to 4x as many sats proposed, remember), satellites under maintenance in lower or graveyard orbits, "flashers", and impacts beyond VIS.

That's the thing with other objects in LEO, though... you can't logically say that "any illuminated Starlink has an impact" and at the same time say "the other 20k+ illuminated objects in LEO have no impact". Anything large enough to show up in the catalog is going to be visible to a large telescope, if it's illuminated.

So the relevant question is: "how bright are all of the objects, and at what magnitudes do secondary effects like bleed-over into other pixels and scattering to the whole frame become manageable?"

So then, the simulation with constant brightness icons tells us nothing useful, unless we know what the brightness means in terms of impact...

As mentioned, if a satellite is illuminated it will generally be an impact to astronomy / astrophotography, so the simulation showing every sat at the same magnitude should be a good approximation, especially since it makes some conservative assumptions on the other hand. Looks like the realistic brightness simulation could be easily modified to show only satellites over a certain magnitude. I haven't seen any figure for passes/sq2/min/mag yet.

LSST is going down to what, 24th magnitude? At that sensitivity, literally everything in LEO is going to show up as at least a streak. So you would have to adjust the simulation to show basically everything, at which point it's going to be dominated by stuff other than Starlink. Or else adjust it to the level where the impact goes beyond just a simple streak.

[eeergo]

And I acknowledged that I made a mistake with the tracking of space debries. So you can disregard that part.

Ok, but that was the whole point of your baseless claim.

It is either blatant disinformation or wild disregard for your lack of knowledge about this topic (starting by the fact you didn't know most debris is tracked by radar, not by optical telescopes, which you can find in Wiki or NASA's public outreach space debris site, the two top results if you search for "space debris") - you choose.

Not sure why you make this personal. Please stop.

My excuses if your skin is so thin as to think this is personal, it isn't. You said "disinformation isn't the right term". If you aren't intentionally downplaying this problem using wildly unresearched arguments, you're unintentionally ignorant about the matter. It's not pejorative.

It doesnt matter if a signal is instantaneous on the detector or applied over a longer period of time. What counts is the signal at the time of readout. If you take a series of exposures, a satellite would only be present in one of this stack of images. Just like cosmic rays.

:\

Do you understand how astronomical signal integration works, and the timescales associated vs the transit time of one of these satellites or the discharge time of a CR on a detector? From this comment, I don't think so.

I want to find a paper that tells me all the relevant effects. A paper that outlines all the effects that are considered and the impact on astronomical observations. Many people, me included, are not qualified to do this properly. I would bet with all the attantion, at least LSST would publish something soon.

I hope so. That's actually another side of the problem: the appalling lack of studies of this pollution by anyone, and the fait-accompli policy telecoms are being allowed to implement.

[envy887]

Since GEO is ~100x further away, and meter-sized objects are ~100x larger in area than 10 cm objects, this suggests that 10-cm class debris in lower LEOs will produce a streak, especially with very large telescopes, sufficient to negate the usefulness of the directly effected pixels on that exposure.

Of course that streak will be very much dimmer, which is probably better for the surrounding pixels and especially better for reducing scattering in the optics that affects the whole frame.

The bolded part is key. Anyway, you cannot take visibility of objects in GEO and extrapolate that to visibility in LEO because the more distant the orbit, the less apparent movement the object has... with the limit being GEO, where the object doesn't move with respect to the ground. There, integration time is your friend and you can resolve much dimmer objects. In LEO you'd have to track them.

Edit: 10-cm debris is NOT "exponentially the vast majority" of the catalog. According to CelesTrak, 7636 of the 20036 on-orbit tracked objects (38%) are either payloads or rocket bodies. While some of the payloads are cubesats in that range, most of those, and all the rocket bodies, are quite a bit larger.

Check out slide 19-20 of the link I published. Debris scales approximately exponentially with size for collision-generated fragments. The vast majority (~70%) of tracked objects are very small to register much on observations, and therefore not meaningfully comparable to tens of thousands of 15-meter satellites. I don't see an issue with my observation, which anyway was a comment on top of a much more important idea.

Optical tracking of objects in GEO is done with sub-meter scopes (typically ~25 cm as far as I can tell), and yes, they probably need some time to detect meter-class objects. LSST at 8.4 meters is going to be just a smidge more sensitive.

Even if most LEO objects are debris, that still does leave roughly 10,000 objects in Earth orbit that are not so small and are definitely going to show up on observations even with much smaller scopes than LSST.

[su27k]

You DO think so while neglecting to look at basic information with which to back your claims. You're not gonna find a paper about the effect of mosquitoes flying in front of the telescope or something, just because that isn't an issue. Similarly, papers aren't needed to disprove baseless claims.

I have been providing simulations by professional astronomers, and there are plenty of concerned such people out there speaking out, with increasingly accurate qualitative and quantitative analysis, whose message is pretty coherent. That is proper context, not orbital debris numerology.

No, your so called "simulations" (http://www.deepskywatch.com/Articles/Starlink-sky-simulation.html) is created by Michael Vlasov, an amateur astronomer, not a professional astronomer. Further more, he's just using Stellarium planetarium software to create an animation, this is can be done by anybody, it doesn't prove anything.

There is no "coherent message" from "increasingly accurate qualitative and quantitative analysis", there is only two preliminary results from LSST team, one says 12k Starlink is not an issue and impact is less than 0.01%, the other says "full constellation" would have a rather significant impact without qualifying what the full constellation is. A paper is clearly needed in order to explain the difference here.

Besides LSST results, the only quantitative result I have seen is from Olivier Hainaut, a professional astronomer at ESO, his result is that:

Most telescopes can deal with that, says Olivier Hainaut, an astronomer at the European Southern Observatory (ESO) in Garching, Germany. Even if more companies launch megaconstellations, many astronomers might still be okay, he says. Hainaut has calculated that if 27,000 new satellites are launched, then ESO’s telescopes in Chile would lose about 0.8% of their long-exposure observing time near dusk and dawn. “Normally, we don’t do long exposures during twilight,” he says. “We are pretty sure it won’t be a problem for us.”

So any claim that Starlink has big impact on things besides sky survey like LSST is baseless.

[eeergo]

That's the thing with other objects in LEO, though... you can't logically say that "any illuminated Starlink has an impact" and at the same time say "the other 20k+ illuminated objects in LEO have no impact". Anything large enough to show up in the catalog is going to be visible to a large telescope, if it's illuminated.

No, you're fudging it.

As I said before, the majority of objects in the catalog are tracked through radar, not because they're optically or shortwave-visible.

Starlinks: 12k identical 15-m satellites with reflective surfaces -> consistently and largely visible.
Debris and other satellites: 20k objects in *many* orbits, mostly of a small size and not necessarily reflective.

So yes, I can say mostly any Starlink (or other such satellite from such LEO constellations) that's illuminated will be visible, and most of the sub-meter LEO debris won't, or will have a much lesser impact.

So the relevant question is: "how bright are all of the objects, and at what magnitudes do secondary effects like bleed-over into other pixels and scattering to the whole frame become manageable?"

So then, the simulation with constant brightness icons tells us nothing useful, unless we know what the brightness means in terms of impact...

The question you pose is appropriate. The conclusion about the simulation is not. My metaphor of climate change becomes more applicable ("further research is needed" when claiming lack of meaningful impact is untenable).

LSST is going down to what, 24th magnitude? At that sensitivity, literally everything in LEO is going to show up as at least a streak. So you would have to adjust the simulation to show basically everything, at which point it's going to be dominated by stuff other than Starlink. Or else adjust it to the level where the impact goes beyond just a simple streak.

Again, you're fudging it by going straight to the limit cases.

It's not going to be dominated by stuff other than Starlink/LEO megaconstellations, because (i) there are less comparably-sized objects orbiting Earth anywhere right now, in absolute terms; (ii) smaller objects in comparable orbits are not even close in visibility, or even practically do not integrate a signal over the integration time needed to resolve 24 mag deep sky objects, and (iii) megaconstellations would dominate other trackable debris in LEO, even accounting for sub-meter debris.

[eeergo]

Optical tracking of objects in GEO is done with sub-meter scopes (typically ~25 cm as far as I can tell), and yes, they probably need some time to detect meter-class objects. LSST at 8.4 meters is going to be just a smidge more sensitive.

If the object doesn't move much, or at all, it can be blacked out. Obviously.

Even if most LEO objects are debris, that still does leave roughly 10,000 objects in Earth orbit that are not so small and are definitely going to show up on observations even with much smaller scopes than LSST.

That number is overestimated based on the numbers we've been quoting, but whatever. Is your argument that since there's a problem we shouldn't worry about making it much worse? Should we just dump another Pacific Garbage Patch since there's one or several already?

[eeergo]

No, your so called "simulations" (http://www.deepskywatch.com/Articles/Starlink-sky-simulation.html) is created by Michael Vlasov, an amateur astronomer, not a professional astronomer. Further more, he's just using Stellarium planetarium software to create an animation, this is can be done by anybody, it doesn't prove anything.

There is no "coherent message" from "increasingly accurate qualitative and quantitative analysis", there is only two preliminary results from LSST team, one says 12k Starlink is not an issue and impact is less than 0.01%, the other says "full constellation" would have a rather significant impact without qualifying what the full constellation is. A paper is clearly needed in order to explain the difference here.

Besides LSST results, the only quantitative result I have seen is from Olivier Hainaut, a professional astronomer at ESO, his result is that:

Most telescopes can deal with that, says Olivier Hainaut, an astronomer at the European Southern Observatory (ESO) in Garching, Germany. Even if more companies launch megaconstellations, many astronomers might still be okay, he says. Hainaut has calculated that if 27,000 new satellites are launched, then ESO’s telescopes in Chile would lose about 0.8% of their long-exposure observing time near dusk and dawn. “Normally, we don’t do long exposures during twilight,” he says. “We are pretty sure it won’t be a problem for us.”

So any claim that Starlink has big impact on things besides sky survey like LSST is baseless.

You do realize being an amateur astronomer (are you?) doesn't invalidate his result given it's based on pretty reasonable and accurate parameters, right? Much less make it "baseless" as your decontextualized claims.

You don't need a team of ophthalmologists and a supercomputer to tell you 30 medical lasers pointing to your eye for a week won't be good for your retina, but a laser pointer for a second will probably be ok.

I guess Cees Bassa, James Lowenthal, Anthony Tyson or Daniel Marín are also baseless amateurs, as are the many professional astronomers (disorganizedly) speaking out every day. I also guess the IAU (https://www.iau.org/news/announcements/detail/ann19035/), AURA (https://www.aura-astronomy.org/news/aura-statement-on-the-starlink-constellation-of-satellites/), LSST (40%, not 0.01%, please update your quotes), AAS (https://aas.org/press/aas-issues-position-statement-satellite-constellations), Fuji telescope... are baseless, since for sure they aren't amateurs.

[su27k]

Nope, the number of catalogued objects in orbit is >19000, and several thousands more are tracked but not catalogued: https://orbitaldebris.jsc.nasa.gov/quarterly-news/pdfs/odqnv23i1.pdf

Nevertheless, they are tracked not because they are "visible" (in visible wavelengths), but overwhelmingly observed through radar. About a third of them are attributable to ASAT tests or the Iridium-Kosmos collision (i.e. small fragments).

Are you seriously apples-to-apples comparing the visibility of a O(10 cm)-sized piece of debris (exponentially -see 1998 debris model slideplayer.com/slide/12934272/- the vast majority of the tracked debris you mention) with tens of thousands of actual 15-meter-long functioning satellites?

Are you seriously comparing the CR stochastic noise to satellites passing in front of your FOV?

No paper-quoting is needed when you are twisting basic concepts. Please spare us such crude misinformation.

Wrong, groundbased telescope can be used to track LEO debris less than 10cm.



From Limiting Future Collision Risk to Spacecraft: An Assessment of NASA's Meteoroid and Orbital Debris Programs (2011) Chapter 2:

For debris of larger sizes (between 2 mm and 10 cm in LEO), an exposed surface larger than that of a typical satellite is required to obtain a meaningful sample of “impacts,” so ground telescopes and short-wavelength radars are used.

...

An advantage of detecting uncataloged debris with both radars and telescopes is finding debris that may not be seen by one or the other alone. In 1995, NASA began operations of the NASA-built and NASA-designed 3-meter Liquid Mirror Telescope (LMT). However, as a result of budget cuts, the LMT was shut down in 2001.

...

Some of the limitations with respect to observational inclination will be resolved with remote operations by NASA of the Meter-Class Autonomous Telescope (MCAT) at Kwajalein Atoll in the Pacific Ocean beginning in 2012. This low-latitude location will permit detection of uncataloged debris at low inclinations, although not to sizes as small as can be detected by current radar capabilities. MCAT will also detect GEO debris as small as 10 cm.

Details about NASA's Liquid Mirror Telescope can be found in Orbital Debris Quarterly News, Volume 11, Issue 2:

From the outset it was desirable that the new
orbital debris telescope be able to detect objects
smaller than the ostensible 10-15 cm SATCAT
diameter size limit. For a minimum detection
diameter, 1 cm was selected because research
performed on shielding for the International
Space Station (ISS) indicated that layered
(Whipple) bumpers offered protection from
objects smaller than this. Since the ISS would
be constructed in low Earth orbit (LEO) at
approximately 500 km altitude, the NASA-LMT
would focus on providing information on the
LEO debris population in the critical 1 to 15 cm
diameter size regime. The LMT would also help
quantify the statistical accuracy of the SATCAT
by comparing the observed and predicted fl ux
of objects larger than 10 cm diameter at LEO
and middle Earth orbit (MEO) altitudes. By
necessity, direct observations below 1 cm were
relegated to Radar such as Haystack operated by
Massachusetts Institute of Technology’s (MIT)
Lincoln Labs at Millstone Hill, Massachusetts

[su27k]

No, your so called "simulations" (http://www.deepskywatch.com/Articles/Starlink-sky-simulation.html) is created by Michael Vlasov, an amateur astronomer, not a professional astronomer. Further more, he's just using Stellarium planetarium software to create an animation, this is can be done by anybody, it doesn't prove anything.

There is no "coherent message" from "increasingly accurate qualitative and quantitative analysis", there is only two preliminary results from LSST team, one says 12k Starlink is not an issue and impact is less than 0.01%, the other says "full constellation" would have a rather significant impact without qualifying what the full constellation is. A paper is clearly needed in order to explain the difference here.

Besides LSST results, the only quantitative result I have seen is from Olivier Hainaut, a professional astronomer at ESO, his result is that:

Most telescopes can deal with that, says Olivier Hainaut, an astronomer at the European Southern Observatory (ESO) in Garching, Germany. Even if more companies launch megaconstellations, many astronomers might still be okay, he says. Hainaut has calculated that if 27,000 new satellites are launched, then ESO’s telescopes in Chile would lose about 0.8% of their long-exposure observing time near dusk and dawn. “Normally, we don’t do long exposures during twilight,” he says. “We are pretty sure it won’t be a problem for us.”

So any claim that Starlink has big impact on things besides sky survey like LSST is baseless.

You do realize being an amateur astronomer (are you?) doesn't invalidate his result given it's based on pretty reasonable and accurate parameters, right? Much less make it "baseless" as your decontextualized claims.

What result? He showed some animation of Starlink going around the sky, this proves nothing.

You don't need a team of ophthalmologists and a supercomputer to tell you 30 medical lasers pointing to your eye for a week won't be good for your retina, but a laser pointer for a second will probably be ok.

Useless analog, you're intentionally ignoring the only other professional astronomer that did the calculation and showed a result of minimal impact.

I guess Cees Bassa, James Lowenthal, Anthony Tyson or Daniel Marín are also baseless amateurs, as are the many professional astronomers (disorganizedly) speaking out every day.

Show me the numbers. Without numbers, it doesn't matter what they speak, this is science, not astrology.

I also guess the IAU (https://www.iau.org/news/announcements/detail/ann19035/), AURA (https://www.aura-astronomy.org/news/aura-statement-on-the-starlink-constellation-of-satellites/), LSST (40%, not 0.01%, please update your quotes), AAS (https://aas.org/press/aas-issues-position-statement-satellite-constellations), Fuji telescope... are baseless, since for sure they aren't amateurs.

IAU/AURA/LSST/AAS statements are all focused on one thing: LSST, but LSST does not equal the whole astronomy field. And no, I won't update the quote until LSST team showed how an estimate can jump from 0.01% to 40%, this clearly requires an explanation.

Also you're ignoring the latest AAS statement: https://aas.org/posts/advocacy/2019/12/aas-works-mitigate-impact-satellite-constellations-ground-based-observing, where they stated "Until we have collected the survey data, we and SpaceX will proceed as if satisfying the Large Synoptic Survey Telescope's needs is the high bar to aim for." and "The goal of Starlink is to provide worldwide internet service, an aspiration we do not want to impede, but this requires one to two orders of magnitude more low Earth orbiting satellites (LEOs) than currently exist. We do not want to give up access to optical observations from the ground. Our group’s task is to find a path forward that accommodates both uses of the sky.", so:

1. SpaceX is cooperating with AAS to satisfy the need of LSST, this basically mitigate all the concerns in the statements you listed.
2. AAS is clearly following a live and let live path where both constellation and astronomy can co-exist, as AAS is the organization representing all US astronomers, I take this to be the sentiment of astronomers at large.

[envy887]

My metaphor of climate change becomes more applicable ("further research is needed" when claiming lack of meaningful impact is untenable).

That metaphor is hilarious considering that the operators of LSST said only 6 months ago, and I quote, that "for LSST, Starlink satellites will be a nuisance rather than a real problem", and that referring explicitly to the 12k deployment.

Is your argument that since there's a problem we shouldn't worry about making it much worse? Should we just dump another Pacific Garbage Patch since there's one or several already?

Indeed not. And this isn't a valid analog, because Starlink is inherently valuable while a garbage patch is not.

My argument is that Starlink will (if SpaceX continues to work to mitigate the brightness as they claim they will) make the existing set of problems facing optical observatories only a little worse, or at most somewhat worse, and not to an extent that negates Starlink's inherent value to space development and the rest of the world. And also that astronomers should focus on evaluating and mitigating the effects of Starlink passes on their end, in addition to working with SpaceX to reduce the brightness.

[envy887]

I won't update the quote until LSST team showed how an estimate can jump from 0.01% to 40%, this clearly requires an explanation.

They did provide an explanation in the statement: the satellites are bright enough that scattered light in the optics impacts the entire frame, and not just the pixels directly under the streak.

Why they missed that effect the first time isn't explained, nor how much of a brightness reduction is necessary to mitigate this effect, if it was due to the satellites being brighter than expected. They had to have expected at least 5th magnitude IMO, because LEO satellites are quite commonly that bright.

[Lar]

We need an increase of collegiality, STAT.

Calling an analogy "hilarious"? Not helpful. Calling research "baseless"? Not helpful. Calling people ignorant? Not helpful either.

Look, I also moderate the SpaceX group on FB and we have had one dumpster fire after another there whenever Starlink is brought up. I came over here to read more reasoned, calm, facts based analysis. The last few pages are not meeting the "be excellent to each other" standard we hold ourselves to.

Let's do better please.

[Semmel]

It doesnt matter if a signal is instantaneous on the detector or applied over a longer period of time. What counts is the signal at the time of readout. If you take a series of exposures, a satellite would only be present in one of this stack of images. Just like cosmic rays.

:\

Do you understand how astronomical signal integration works, and the timescales associated vs the transit time of one of these satellites or the discharge time of a CR on a detector? From this comment, I don't think so.

I dont think my knowledge of the topic has anything to do with the argument I make. I am always happy to learn. Can you please explain what is wrong with my statement?

[Semmel]

Thx Lar. Though, if this conversation has proved anything, its the need for some paper from professionals. There are so many variables involved, its hard to keep everything in the head. Especially when it comes to impact. I try to accumulate a list of things to consider for such a publication.

Observatory site:
Latitude matters mostly here. If the observatory is located far from the equator, the density of constellation satellites is higher than on the equator. Also, twilight during winter is longer far from the equator.

Size of the telescope:
Larger telescopes are susceptible to fainter objects. But usually also collect more sky background.

Instrument field of view:
Large field of view is obviously worse because it sees more sky. Luckily most instruments are very narrow, in the arc seconds range some see arc minutes. Very rarely is the field of view degrees in diameter. Spectrographs usually have a very narrow field of view.

Instrument exposure time:
This varies a lot from instrument to instrument. From seconds to hours, everything is included. Though, hours are usually split in multiple stacked exposures and not done in one go because the sky background noise is far brighter than the read out noise and the number of cosmics is prohibitive. Its rare that exposure times exceed 10 minutes.

Scinece operation:
There are so many different ways of using a telescope, its hard to get a good handle on this. Some scan the sky, some target specific objects, some do both. Some observe the same patch of sky occasionally over a long period of time.. its hard to make a dedicated impact analysis without also considering the science case.

Time of observation:
Obviously the sats are only visible during twilight. In this time, science operations are almost never done. Some calibration but not science. The sky is just too bright for that. Time of year also matters because it influences the length the satellites are visible and the ratio of astronomical night time vs. twilight.


Target on sky:
The target matters in relation to the location of the observatory.  Most of the time, observations are limited to close to zenith. But the Magellan clouds are high interest targets for many instruments and they are very far south and the telescope has to observe closer to the horizon than otherwise usually done. So the clouds would be kind of a worst case for most telescopes.


Brightness of the satellites:
The satellites are very bright short after launch but dim considerably. The first set of starlink satellites has currently a brightness of about 3 magnitudes (based on heavens above). This is not different to other objects. So time after launch matters. But with the current launch cadence, there seems to be a just launched train all the time. But its just one train not covering the entire sky. Rather like the ISS.


Mitigation:
I am unsure if SpaceX has the technology to stealth their satellites such that they become invisible when illuminated by the sun. Its quite a trick to do that.
Other mitigation methods are also discussworthy. Like software detection and closing shutter as suggested above. Maybe even scheduling observations to avoid satellites but that would be a major impact on observatories.


In many of the above, the LSST is the worst case. Its a large telescope, wide field of view, wide wavelength band and scanning mode, which means only 2 observations per field. The only thing that it has going for it is relatively short exposure times. In many cases, considering the LSST is enough if you want to prove that the constellation has no impact on astronomy. Because if it doesnt impact LSST (or not much), other observatories and instruments are very likely even less effected. On the other hand, if the analysis finds that the LSST is heavily impacted, more complex analysis has to be done. So the LSST is a good starting point.

[Semmel]

I won't update the quote until LSST team showed how an estimate can jump from 0.01% to 40%, this clearly requires an explanation.

They did provide an explanation in the statement: the satellites are bright enough that scattered light in the optics impacts the entire frame, and not just the pixels directly under the streak.

Why they missed that effect the first time isn't explained, nor how much of a brightness reduction is necessary to mitigate this effect, if it was due to the satellites being brighter than expected. They had to have expected at least 5th magnitude IMO, because LEO satellites are quite commonly that bright.

Their exact statement:

The LSST Project Science Team has been simulating the potential impacts to LSST observations. Their latest update of preliminary results from November 2019 indicates that (assuming the full deployment of planned satellites) nearly every exposure within two hours of sunset or sunrise would have a satellite streak. During summer months there could be a 40% impact on twilight observing time (less in winter) and saturation of sensors by the satellites can continue well past astronomical twilight. Because of scattered light in the optics by the bright satellites, the scientific usefulness of an entire exposure can sometimes be negated.

Ok, that sounds totally different already. 40% of the frames during twilight have a streak. During twilight is the key because twilight has a bright sky background. Thats what it makes twilight. Which means scattered light from the satellite is barely visible if at all. Twilight ends about 1.5 hours after sunset and starts 1.5 hours before sunrise (at least in Paranal). LSST is a bit more south (LSST: 30 deg south, Paranal 24deg south), so twilight might be longer there, the 2 hours makes sense. In that time period, the very dim objects are not visible due to the brightness of the sky. My hunch is, that this statement is not as impactful on science operation as it sounds.

Besides, Starlink satellites are about magnitude 3 (bases on heavens above). Thats normal for satellites. There are 170 stars of that brightness or brighter. You can see them with your naked eye when walking out at night. If LSST has a problem with Mag 3 objects that effects their entire frame during twilight, they will have a hell of a problem during dark time.

[envy887]

I am unsure if SpaceX has the technology to stealth their satellites such that they become invisible when illuminated by the sun. Its quite a trick to do that.
Other mitigation methods are also discussworthy. Like software detection and closing shutter as suggested above. Maybe even scheduling observations to avoid satellites but that would be a major impact on observatories.

It should be possible in many cases to close a shutter when a pass is predicted. With SpaceX publishing TLEs frequently, passes can be predicted with very high temporal and spatial accuracy.

For all but the very widest FOVs, the shutter would only need to be closed for a few seconds at most, because the satellites are moving very quickly.

[Semmel]

I am unsure if SpaceX has the technology to stealth their satellites such that they become invisible when illuminated by the sun. Its quite a trick to do that.
Other mitigation methods are also discussworthy. Like software detection and closing shutter as suggested above. Maybe even scheduling observations to avoid satellites but that would be a major impact on observatories.

It should be possible in many cases to close a shutter when a pass is predicted. With SpaceX publishing TLS frequently, passes can be predicted with very high temporal and spatial accuracy.

For all but the very widest FOVs, the shutter would only need to be closed for a few seconds at most, because the satellites are moving very quickly.

I agree, it would be possible. I know a bit about the ESO control infrastructure. The instrument control software allows the pause of an exposure without killing the observation. The control infrastructure for that exists. What doesnt exist is an automated system to command the pause of an exposure when a sattellite enters the field of view.

They would need 2 components, a global warning system that correlates current pointing and location of satellites to predict when a pause is required. This is probably the hardest part. Something similar exists to avoid the laser guide stars of other telescopes. On Paranal are 6 telescopes, 1 has a laser guide system (UT4). Whenever the laser guide system is active, any other instrument that would cross beam with the laser is prevented from starting an exposure. A similar system could be implemented to detect satellites I suppose. That would take some significant time to get done.

The second part would be an addition to the VLT instrument control framework to pause an exposure when the software receives the satellite warning. This is certainly easier but would need some convincing.

I dont know if all instruments have shutters. Also, its not always desired. For instance, if you want to observe exoplanet transits, a pause in the exposure might ruin your time series. But the blip of the satellite might not, especially if it doesnt cross exactly over the target. So it is situational and not a general solution.

[eeergo]

It doesnt matter if a signal is instantaneous on the detector or applied over a longer period of time. What counts is the signal at the time of readout. If you take a series of exposures, a satellite would only be present in one of this stack of images. Just like cosmic rays.

Do you understand how astronomical signal integration works, and the timescales associated vs the transit time of one of these satellites or the discharge time of a CR on a detector? From this comment, I don't think so.

I dont think my knowledge of the topic has anything to do with the argument I make. I am always happy to learn. Can you please explain what is wrong with my statement?

Sorry about signing off last night, but I only have limited time and will for engaging in these sort of conversations during free moments left out by *actually* doing astrophysics research (albeit not optical, so sorry for not being able to provide definitive numbers off the top of my head, but I do have more than a passing knowledge of the delicate balances at play), and reacting to the announcement the major project I work on just got full funding. It also gets tiring to get stormposted with -euphemistically- poorly meditated retorts for being "anti-SpaceX", the new "un-American", especially when I'm being careful to refer to *all* such megaconstellations.

Anyway: your knowledge of the topic *is* important when you make such definitive statements, and have continued in your latest posts even after I showed you why. In fact, from the tone of your posts here I was dubious whether you were a fellow astrophysicist or related scientist, until further ones pretty much cleared up the issue.

You're talking about data pipelines, laser guides, control software infrastructures, cosmic rays, exposures, signal at readout etc. and you don't know how an optical (or whatever) signal is integrated to optimize S/N while minimizing spurious (i.e. short-timescale) artifacts or noise? I would think that's a much more basic point you need to understand before going on to discuss fixes:

CRs can be efficiently filtered out because their discharge time on the detector is very short, while astronomical targets are fixed and/or tracked. It's not that the images without them (none) are selected out. Same concept for dim quick objects (on the timescale of the long exposures needed to image large-magnitude targets) that don't saturate the detector, BUT obviously they won't ever be as quick or random as CRs, so their impact will be greater, although in many cases they *can* be just picked out of the image stack.

Obviously, the brighter and slower with respect to the target they are, the more they'll lower the S/N and the more they'll complicate downstream work as calibration or analysis. One or two satellites crossing your image can be avoided or corrected for, generally. When you get as many large, stable and reflective satellites as all current pieces of LEO stuff (most of which are not/barely a nuisance in optical, as explained upthread) combing out the "bad" images on the stack without greatly reducing S/N, introducing unwanted biases or making your analysis shakey goes down, non-linearly. When good foreseeable business means 4x as many, with competition, and higher-impact dead/under-maintenance sats, well...

[jebbo]

CRs can be efficiently filtered out because their discharge time on the detector is very short, while astronomical targets are fixed and/or tracked.

I admire your persistence. I largely gave up on this thread a while ago.

Tangentially on CRs, depends what you mean by "filtered out". For some sensors (e.g. the Kepler CCDs), it is easy enough to flag the CRs and filter out the appropriate pixel samples, but we learned that they reduced sensitivity for quite a long time afterwards, and this was the source of a number of false-positive transit detections.

I imagine for future ground-based transit searches, depending on the instrument PSF etc, we could see false-positives when dim objects cross the optimal aperture (we can probably detect them).

Another issue is that when you consider the follow-up observation of some transiting planets, where you want to look at in/out of transit spectroscopy, some of these targets are only going to be visible, low in the sky in the sky and near dawn / dusk, at some times of year and for a very short window (particularly if you want to look at ingress/egress). If such observations get polluted, then the next observing opportunity could be a long time in the future.

It is quite hard to work out stats for such things because it is target dependent (not just on coordinates, but things like period)

More broadly, though I think mitigation in optical bands can be done (at least partially, but there are issues with solar panels where the active surface can't be coated; I worry about reflections from these in some geometries) using low albedo coatings, it is more difficult in the IR, as the satellites will be warm emitters regardless of coatings. So wide-field IR surveys could suffer.

--- Tony