# Which are stars and which are noise in this comet photo?

The Forbes news article NASA's Hubble Telescope Snaps Best Images Yet Of Our First Interstellar Comet shows the image below (which I have since annotated) of comet 2I/Borisov (C/2019 Q4) and links to this series of Hubble images: https://archive.stsci.edu/proposal_search.php?mission=hst&id=16009

Looking through the previews of those images, they all seem to have streaks of similar length and direction, so I am assuming that the telescope is tracking the comet's motion in each exposure and the streaks are stars, suggesting that all the small dots are noise and not stars.

1. Is that right?
2. Is the fuzzy one an extended object?
3. What causes so many isolated pixels to be so much brighter than the background? Is this just the tail of a statistical distribution of shot noise, or are there other mechanisms that can produce single pixel noise many standard deviations above the statistical shot noise distribution?

• To what you refer by the fuzzy extended object? I do only see the comet (and of course strikes, stars and the noise). – Alchimista Oct 16 '19 at 10:13
• @Alchimista there are five annotated tracks, the bottom one is fuzzy. There's another fuzzy one not annotated at the top of the image (above the comet). I'm calling these two fuzzy tracks "extended". – uhoh Oct 17 '19 at 0:36
• Ah ok now I see them – Alchimista Oct 17 '19 at 9:23
• – uhoh Oct 17 '19 at 14:30

1. Is that right?

Yes.

1. Is the fuzzy one an extended object?

That would certainly be my guess (probably a distant galaxy).

1. What causes so many isolated pixels to be so much brighter than the background? Is this just the tail of a statistical distribution of shot noise, or are there other mechanisms that can produce single pixel noise many standard deviations above the statistical shot noise distribution?

Some of those might be "hot pixels", which are manufacturing defects that cause individual pixels to overproduce electrons. Most of them are due to so-called "cosmic rays", a generic term used in astronomical imaging to refer to the effects of energetic charged particles hitting the detector. In ground-based images, these are often not actual cosmic rays, but instead are particles from the decay of radioactive isotopes in or near the detector. In the case of space-based images like these, they more often are proper cosmic rays, though many of them are really charged particles trapped in the Earth's magnetosphere. (There is a part of Hubble's orbit which takes it within the South Atlantic Anomaly, where the Van Allen belt dips down toward the Earth's surface; observations are not scheduled during those periods, because of all the extra cosmic-ray hits the detectors receive.)

For example, here is a preview image from one of the other exposures in the sequence of observations. If you look closely, you can see that many of the cosmic-ray hits are slightly elongated, because the particle was traveling at an intermediate angle and passed through two or more adjacent pixels. The bright linear streak right next to the comet is an extreme example of this, where the particle was traveling almost parallel to the detector and so passed through an extended sequence of pixels.

The usual approach to dealing with this is to take several short exposures, one right after another, and combine them afterwards by rejecting extreme outlying pixel values from a single exposure (stars and other genuine objects will have similar pixel values from one exposure to the next, while cosmic rays are randomly located and do not repeat). This is why the HST images you see usually don't have cosmic-ray artifacts: they are the result of combining multiple exposures. (See here for a discussion and an example.) There are also statistical analysis tricks one can use to try to clean up individual exposures, though this necessarily involves interpolating data.

• I see, so there's probably some simple formula for the length distribution of randomly oriented minimum ionizing particles for a given pixel sensitive depth/width ratio (not that these have to be minimum ionizing) such that most are one or a few adjacent pixels, and only a few can cover tens or hundreds of pixels. Thanks for the great answer! – uhoh Oct 15 '19 at 7:59

This GIF is made (via giphy.com) from the new NAASA Goddard video Hubble's New Image of Interstellar Object. It shows the comet moving at quite a clip!

This shouldn't be a surprise.

From the link in the question https://archive.stsci.edu/proposal_search.php?mission=hst&id=16009 the coordinates for the first and last exposure are:

    RA             Dec            Time
09 47 45.181   +18 07 30.70    2019-10-12  13:44:39
09 48 17.077   +17 59 20.37    2019-10-12  20:42:23


In 7 hours the comet moved roughly 0.2 degrees!

From JPL's Horizons, the state vectors for the comet and Earth around the middle of the image sequence are:

      JDTDB        Calendar Date (TDB)        X     (km)      Y               Z               VX   (km/s)     VY              VZ
comet 2458769.250, 2019-Oct-12 18:00:00, -1.9092457E+08,  2.9804744E+08,  3.2507629E+07, -1.3003446E+01, -2.9440986E+01, -2.7477326E+01
Earth 2458769.250, 2019-Oct-12 18:00:00,  1.4082817E+08,  4.9358841E+07, -1.4600024E+03, -1.0121471E+01,  2.8065445E+01, -7.7695706E-04


The comet is about 416 million km from Earth, moving at roughly 64 km/s relative to Earth and 42 km/s relative to the solar system barycenter.

Is that right?

No.

The streaks are fast-moving objects in the view of the telescope -- Other satellites. Hubble is in a fairly low orbit (it can be visited) and there are plenty of satellites around and above it, including in its field of view. (For obvious reasons, the streaks are not caused by aeroplanes, a usual contender for streaks). What gives it away are the differing lengths of the streaks, indicating different orbital heights, and the fact that there are streaks going in other directions, meaning other orbits.

The larger fuzzy object towards the bottom that you marked might be larger satellite that is closer by and is thus imagined as a fuzzy blob streaking across the image.

Further to this point, the image time here is 5 minutes, as indicated by the start and stop dates given on the proposal. 5 Minutes is much too short for a comet to move significantly against the background, causing streaks when tracked. For the five minutes of imaging, all objects are essentially fixed on the viewing frame.

Lastly, to your point of noise: The imaging qualities of Hubble are superb [citation needed]. That amount of noise, even in raw data, would indicate an awful pixel sensor - on par to a cheap camera.

• @DanHanson -- Your DSL is sitting at the bottom of the Earth's atmosphere, which protects it from the cosmic rays that afflict Hubble. – Peter Erwin Oct 15 '19 at 19:35
• I have been working with HST images for over twenty years. That is the standard kind of noise you get in HST images. – Peter Erwin Oct 15 '19 at 19:36
• "5 Minutes is much too short for a comet to move significantly against the background" -- This is wrong. Per the linked article, the comet's apparent motion was 75 arcsec/hour, which means 5.4 arcsec in a 260 sec exposure. That's over 130 pixels on the detector they were using. They absolute had to move the telescope in order to track the comet. – Peter Erwin Oct 15 '19 at 19:51
• @PeterErwin You're right, I'm wrong. I'm an amateur astronomer and use stacking all the time to improve S/N ratio in my images, but I had no idea there were that many cosmic rays in a hubble frame. Your link was very educational. Thanks! – Dan Hanson Oct 15 '19 at 20:30
• For those who are curious as to what an actual satellite trail might look like, there's one in another of the HST images of this comet, taken about an hour and a half earlier, here. (It's the diagonal streak from upper left to lower right, from an object moving so fast that it crossed the camera's field of view before the telescope had moved far enough to produce visible curvature.) – Peter Erwin Oct 16 '19 at 13:53

Is that right?

I guess not.

It is easy to correct for biases in pixel sensitivity, these objects are stars and the telescope tracks them such that their light is collected into the same photon bin over time. Correction for pixel sensitivity is generally done (own experience) by taking images completely out of focus in faint light conditions. In this way one assures that the incoming number of photons is the same for each bin (in a statistical sense) and one can calculate the relative sensitivity of each pixel. I do not know if this is the way this is done for the Hubble telescope as well.