Comment by themeiguoren | Hacker News Reader

Comment by themeiguoren | Hacker News Reader

Parent Full thread

From a first-principles perspective, it's baffling to me why this would ever be a serious problem. We know the ephemerides of all objects in Earth orbit, and they are traveling at a fast enough angular speed to not linger in an observational spot. We're not working with long-exposure film here - we can manipulate pixels over time. On a technical level it should be trivial to mask out the pixels where satellites are located at each moment. I think 1% is far too high a guesstimate, but say that artificial satellites end up covering 1% of the full night sky (for context the moon takes up under 0.001% of the sky's area). That still leaves you with 99% observability - hardly an effect on the amount of data you can bring in, especially compared to such problems as clouds.

Dead-simple algorithms to erase satellite trails are available for hobbyists right now: https://www.skyandtelescope.com/astronomy-blogs/imaging-foun...

And more complex ones for professional astronomers should still be super simple to implement compared to many of the algorithms astronomers have to wrangle with.

We don't want to manipulate pixels; the goal is not to produce pretty pictures of known things. The goal is to collect as much data as possible about things which are not well understood (because, in part, of a lack of data), especially those that are time-varying on short human timescales.

The objects in question are often extremely faint, and even if one knows a lot of the details of periodic noise, removing the noise does not recover the faint signal.

> baffling

Stand by, there will be lot of professional science communication on this topic in due course.

> moon, 1%, ...

Suppose right now at the very limb of the sun we detect an extremely interesting light curve that is clearly not extragalactic yet still atypical of supernovae, but know that it is in the process of moving directly behind the sun (from our perspective). How do you propose to remove the sun from the image of the suspected supernova?

Here's a list of solar-system objects that moved directly behind the sun in 2019: https://sungrazer.nrl.navy.mil/index.php?p=transits/transits... This Christmas would have been an interesting time for 2010: The Year We Make Contact's great black spot of Jupiter to appear.

The moon also occults Jupiter fairly regularly. If Jupiter was doing something strange behind the moon, think of the loss for all the Earth-based observatories. Or would it be trivial for them to faithfully reconstruct the unusual behaviour of Jupiter?

Yes, the sun and moon are slow-moving, bright, and high-solid-angle obstructions compared to any spacecraft. However, so is Jupiter and a galactic supernova, and neither of them are usually interestingly variable on the scale of a glitching millisecond pulsar (for example) or the electromagnetic counterparts to a gravitational wave detection.

In short

> algorithms to erase satellite trails

are doing cosmetics, not astronomical or astrophysical research.

And if cosmetically you erase the wrong pixel, you miss out lots of observation time of events like https://en.wikipedia.org/wiki/ASASSN-15lh#Discovery and end up regretting missing early data (as in e.g. SN 1997D).

http://www.astronomy.ohio-state.edu/asassn/index.shtml exists to avoid missing out on such things again, and if you have a thick skin, there's https://twitter.com/SuperASASSN

A recent tweet is interesting: 'This is, by the way, a discovery made in "astronomical twilight" - one way Itagaki and other amateur astronomers can discovered such events before professional astronomers is by observing fields (galaxies) quite close to the Sun.' (although that's in large part because bigger observatories deliberately aim away from the sun, because it's very noisy and its luminosity may damage sensitive detectors).

I understand the goal is not pretty pictures. But you're not using a hammer to remove random white noise from all your observations, you're using a scalpel to mask out completely predictable errors for very short periods of time. My amateur astronomer link was not meant as a scientific how-to, but as an illustration that even simple hammer-y methods work to mask out satellite trails.

There are very few non-periodic events in the universe that happen over the fraction-of-a-second timescale for which a satellite occludes an area of the sky, the linking of EM to gravitational wave events being perhaps the only notable exception. Meteor impacts on planets happen over the course of seconds. Supernovae occur over the timeframe of hours. Glitches in millisecond pulsars (the millisecond name notwithstanding) occur over the timescale of days.

This is also the issue with using occlusion by the sun and moon as analogy - their occlusions do operate on the timescale of interesting astronomical events. If the moon whipped around the Earth at the same rate as LEO satellites it might occlude Jupiter, but you would have plenty of vision between the shutters to reconstruct events on the timescales we are interested in. I'll also note that the analogy breaks down with the focus on objects in the solar system since they all roughly lie on the ecliptic - you've reduced the occlusions over a 2D sky to occlusions between objects that live on a 1D slice of it. The Sun and Moon both take up 0.3% of the angular diameter of the sky, but only 0.001% the angular area.

Are you losing real data with a sky full of 10's of thousands of LEO satellites? Yes. Is it a significant amount? Not by any reasonable metric. Is it over a significant timescale? No.

An analogy: If you have some video footage and for whatever reason have a punchhole in the corner of every 100th frame, that by no means precludes you from doing analysis on the video. I'd expect the video conservationists to be up in arms about any blemishes on their original film, but I'd argue that that is the position from aesthetics. To answer questions such as "how many times does the ball in this video bounce" or "how does the lighting in this video change over time" or "how does that person's hair react to the wind visible in the video" the hole is insignificant to obtaining the answer, even if the area of interest is directly covered every 100th frame.

I'll close with the acknowledgement that radio interference is a legitimate concern here. But satellites are hardly the only emitters which are increasing in number at an increasing rate, and that commons will be polluted whether we launch them or not.

Reply on news.ycombinator.com