Archive by Author | hhsieh00

More on the Hyper-SuprimeCam Survey Main-Belt Comet Search

By now, hopefully many of you have had the chance to try out the newest iteration of our Comet Hunters main-belt comet search that features data from the Hyper-SuprimeCam (HSC) Subaru Strategic Program, or the HSC Survey for short.  This survey is a ~300-night, 5-year observing program on the 8-meter Subaru telescope (which is the telescope used to obtain all of the data you’ve been previously classifying for the original Comet Hunters archival data search, though using a different camera) that aims to address a wide variety of scientific areas from cosmology to galaxy evolution to searches for distant objects in our own solar system.  While HSC survey observations are not specifically optimized for searching for main-belt comets (see below), with a few tweaks to our original classification interface, we still hope to use the data to search for previously unknown comets, with the added advantage of the data being relatively new compared to the data being used in our archival search, and also set the stage for an even bigger future expansion of Comet Hunters.

The main difference between HSC survey observations and the archival observations that we have asked you to review until now is that each HSC survey image has an exposure time of several minutes or more.  By comparison, for our archival search, we have tried to only use images with exposure times of less than 1 minute each.  Exposure time is how long the camera’s shutter remains open collecting light for a particular image.  For regular hand-held cameras that you might be used to, longer exposure times can be used to take photos in low-light conditions.  For astronomy, longer exposure times are used to study very faint objects.  The usual analogy used by astronomers is that of a bucket collecting rain drops.  The longer you leave the bucket out, the more rain drops it will collect (or for astronomy, the more photons from an object in the sky it will detect).  For a very faint and/or distant object for which only a very small number of photons reach us on Earth, we want to leave the camera shutter/bucket open for a longer time to collect more photons/rain drops.  As a side note, the size of a telescope’s primary mirror corresponds to the diameter of our hypothetical bucket, so by using a large telescope like Subaru and using longer exposure times, we can study very faint objects indeed!

The issue for us though is that long exposures work great for studying objects that don’t move in the sky (other than the steady, predictable motion caused by the Earth’s rotation), but not so much for nearby things in our solar system, like asteroids, which typically appear to move in the sky relative to background stars.  You can think of asteroids as fence posts along a road while you’re driving, while more distant objects like stars and galaxies are mountains far in the distance.  From your perspective, the nearby fence posts appear to move (whizzing past you as you drive) while the distant mountains appear to be essentially stationary.  Using the short exposure times for our archival search, the total apparent motion of asteroids compared to background stars was relatively minimal.  However, with the longer exposure times used by the HSC survey, the motion of the asteroids we are trying to study causes their images to be noticeably elongated (e.g., somewhat sausage-shaped), just as a photo of a fence post (or other nearby objects) might appear smeared out if you tried to take a picture while driving past.


What this means for Comet Hunters is that identifying comets just became a bit harder.  For the archival search, we asked you to compare the appearance of each asteroid with reference stars chosen from the same image and make a note of any differences you might see.  With the HSC survey data, every asteroid image will look different from the background stars because they will be elongated.  Furthermore, any cometary activity that might be present will also be “smeared out”, possibly making it harder to see.  Nonetheless, given the large size of the Subaru telescope, we expect to find some comets with activity bright enough to detect even when smeared out.  Then the trick is just to try to train your eyes to spot what elongated comets might look like, since they will not really look like most people normally think of when they think of what a comet looks like.  To help you out, we have generated some images of what elongated comets could look like and included them in the HSC Survey comet search tutorial (if you need a refresher, click on the “Show the project tutorial button” at the bottom of each classification page and go to the third panel).


The other major change that experienced Comet Hunters will notice is that instead of asking you to compare the appearance of a single asteroid to two comparison stars, we now ask you to compare the appearances of a single asteroid that has been imaged (at least) twice in the same night.  In part, this is due to the fact mentioned above that comparisons to “stationary” background stars are less useful when most of the asteroids we will show you appear elongated.  The other big reason we’ve made this change is to see if this helps to distinguish overlaps (or blends) from real comets.  This subject has been discussed in previous blog posts here and here, so we will not discuss them much here, except to say that background objects (mostly stars and galaxies) that might mimic cometary activity typically will not appear to move in the same way as our target asteroids.  So, by only focusing on asteroids that consistently show activity from one image to the next, we hope that you will be able to immediately identify cases where “activity” is only present in one image of an asteroid, rather than having to do a time-consuming manual check (as described here).

While it may initially take some getting used to, we hope that you will get the hang of searching for comets in HSC data before too long.  As mentioned in previous blog posts, the exciting thing here is that this data is much more recent than the archival data we have previously been having you review, meaning that if we discover any comet candidates, we may be able to trigger immediate follow-up observations to confirm the activity and study it further.  Another exciting aspect of this sub-project is that there is a lot more archival data available than we are currently using for our original SuprimeCam archival search.  As mentioned above, for this initial search, we intentionally restricted ourselves to short-exposure images, but if our attempt to use elongated asteroid images to search for comets is successful, we will be able to draw on the even larger pool of long-exposure SuprimeCam data, and then even include long-exposure data from other medium and large telescopes.  As I also mentioned above, astronomers like to use long exposures in tandem with large telescopes to allow them to study very faint objects, so being able to take advantage of such archival observations to do comet searches will greatly increase the amount of data available to us.  Since we expect that main-belt comets will be relatively rare compared to inactive asteroids, screening a large number of asteroids is key to finding the rare active ones, so the more data we can use, the better!

If you have any other questions about our new search, please comment below.  Otherwise, we thank you in advance for helping out with the new search, and wish you happy hunting!

New Comet Hunters Data Available!

Thanks to everyone who has helped out with Comet Hunters so far.  As noted in a previous blog post, with your help, we’ve completed the first batch of images from launch and have compiled a preliminary list of potential comet candidates based on your classifications, and are currently in the process of vetting those candidates.

We are pleased to announce that new data is now available!  We’ve fixed some issues with our data processing software (in particular, a bug that led to a large number of off-center asteroids that many of you noticed), and so this new set should be easier to analyze and classify.  With the newly uploaded batch of brand-new images as well as some re-processed images that we’ve shown before, we hope to identify many more main-belt comet candidates with your help.  Good luck and happy hunting!

Advanced comet hunting (Part 1)

So, for those of you who have been itching to do deeper analysis of Comet Hunters images, this blog post is for you!  With the information provided here, you will be able to do many additional types of analyses if you are so inclined, but the two tasks we will focus on here are (1) checking for cases of overlapping objects, and (2) checking for consistency in an object’s appearance.

Subject information: The key to being able to perform follow-up analyses of Comet Hunters images is this table (*UPDATE* see link to updated table below) that lists the corresponding meta data for all asteroid subjects currently being shown on Comet Hunters (excluding images of known comets).  The format of the file looks something like this:

SubjID  ObjNum ObjName  Date  Time     ExpT  ObjRA   ObjDec  Filename    ObsTarget
1288778 172 Baucis 2007-06-15 12:53:37 60.0  0.35268 0.08548 SUPA00547638.fits F04
1288832 172 Baucis 2007-06-15 13:01:48 60.0  0.35446 0.08665 SUPA00547678.fits F04
1288891 172 Baucis 2007-06-15 13:09:56 60.0  0.35623 0.08782 SUPA00547718.fits F04

The listed columns are subject ID (SubjID), asteroid number (ObjNum), asteroid name (ObjName), date of the original observation (Date), time of the original observation (Time), exposure time (ExpT; in seconds), Right Ascension (ObjRA; in decimal degree format), Declination (ObjDec; in decimal degree format), name of the original data file (Filename), and the original observation target (ObsTarget).

We should note that there are some duplicate images in this list where the same asteroid appears as two different subjects, one generated for the beta test (which used red crosshairs for the asteroid) and another for the official launch of the project (which uses purple crosshairs), so just be aware of this.

Overlapping objects:  So, as explained in this previous blog post, determining whether something that looks like a comet tail is actually just a background star or galaxy behind our target asteroid involves looking for images of the same part of the sky, and even taken on the same night if possible, and checking to see if the activity “moves” along with the asteroid (in which case, it may be a real tail), or if it instead appears to remain stationary in the sky (in which case, it is a background object).  The science team refers back to the original data files corresponding to each subject to make this determination, but it’s actually possible to do the same type of check simply using the Comet Hunters website.

This technique was actually demonstrated in our previous blog post about detecting overlaps, except that now with the individual subject data we provide you, you can search for other images of an asteroid of interest on  your own.  So to follow the previous example, if you found that Subject 1295819 looked like a possible comet:

You could then refer to the data table to search for the subject number “1295819” and find out that it refers to an image taken of asteroid (27473) 2000 GV78 on 2008-11-29 at 5:45:06.727 (Universal Time).  The file is currently sorted by asteroid number, so you will then note that there are several other images of asteroid (27473) also taken on 2008-11-29 that have been turned into Comet Hunters subjects, namely Subjects 1295829, 1295842, 1295854, 1295866, and 1295876.  To retrieve these images, simply replace the subject number at the end of the following URL:

and then by looking at the images side by side, you may be able to see the asteroid moving across the background object, revealing that the original suspected cometary activity was a case of the asteroid overlapping a background star or galaxy.

overlap_subject1295819 overlap_subject1295842 overlap_subject1295854 overlap_subject1295866

Checking for overlaps will not always be easy though.  Consecutive images of other asteroids may not have been obtained close in time, so background stars may change significantly from image to image.  We use software to automatically set the contrast levels of the images we display and so images of the same asteroid may sometimes be set to different levels, making the brightness of both the asteroid and background stars/galaxies appear different in different images.  Image quality may also change from one image to the next, or some images may have tracking problems but not others.  We are also still fixing the problem that we’ve found where objects were improperly centered by our software, so asteroids may be centered in some images but not others.  In many cases though, this simple analysis should allow you to identify subjects that are clear overlap cases, which will help the science team as we work to weed those out in order to get to genuine comet candidates.

Checking for consistent appearances:  Besides checking for overlaps, looking at other images of the same asteroid taken on the same night allows us to check to see if suspected cometary activity, especially faint activity, remains consistent from one image to the next. If an object shows activity in one image but not others, it is probably just caused by noise and is not what we would consider a good candidate for further follow-up.

For example, Subject 1295853 looks like it may have a faint tail at the 11 o’clock position:

However, looking at Subjects 1295818, 1295839, and 1295865, which are also images of asteroid (336184) 2008 RN107 from 2008-11-29, the same feature does not appear in any of the other images.

consistent_subject1295853  consistent_subject1295865consistent_subject1295818  consistent_subject1295839

In this case, we would then conclude that the original suspected cometary feature is not real.

Other Analyses:  In (a) future blog post(s), we will describe other types of analyses you can do with the data table we now provide, including using the raw FITS files, searching for other instances of observations of an asteroid of interest from different telescopes, retrieving and interpreting information about the orbit and viewing geometry of an asteroid of interest at the time of observations, and more.

*Update (2016-06-02): Here is the updated data table that includes new data uploaded to the Comet Hunters project since this blog post was first posted.  Happy hunting!

*Update (2017-05-16): Due to a change in Dropbox’s file settings, the previously posted link to the data table no longer works. Go here instead for the updated data table posted on 2016-06-02.

What is an overlap and how can we tell?

Many of the comet candidates that have been flagged on Comet Hunters so far have turned out to be cases where an asteroid is “overlapping” a background star or galaxy, and the question has come up as to how do we identify such cases.  First, just to explain, an “overlap” is what we call a situation where an asteroid passes very close to, or even over, a background star or galaxy, as viewed in the sky (they are of course very far away from each other in physical distance!).  If the asteroid and background object are positioned just right relative to one another, it can make the asteroid look like it has cometary activity, when in fact it does not.

In terms of identifying overlap situation, we first note that in most cases, it is actually often not possible to say anything definitive from the displayed image alone, such as the one shown here, which at first glance, certainly looks like an excellent comet candidate:

overlap1Subject 1295819

What we do in these cases is go back to the original image and search for other images of the same field, ideally from the same night and same telescope (since this means the images will look similar, having been taken under similar observing conditions and probably with similar exposure times), to compare them.  Astronomers often take multiple images of the same field within the same night because they are looking to see how a target changes over short periods of time, or because they want to be able to reduce the effects of image artifacts (adding together multiple images can aid the removal of cosmic ray hits, or by moving the telescope slightly between exposures, they can remove the effects of bad pixels on the detector), so in many cases, we can take advantage of this fact to allow us to check for overlaps.

If we can identify observations from the same night, we can look for a consistent appearance of the comet candidate from one exposure to another.  If a “tail” appears to switch sides between two different exposures, this is a good sign that it is actually due to the asteroid moving across a background object (real comet tails do occasionally appear to switch directions as seen from Earth, but not in the time span of a single night), as in the example below where we have identified another image of the same asteroid:

Subject 1295819 and Subject 1295854

or if the tail appears to simply disappear from one image to the next:

Subject 1295819 and Subject 1295842

Looking for a consistent appearance between exposures also helps us distinguish real comets from asteroids that only look like they have tails because they are faint, meaning that background noise can randomly appear to mimic activity.

Finally, if we can find many images of the same field/asteroid from the same night, we can sometimes actually see the asteroid moving across a background object and perhaps even see the background object emerge completely from behind the asteroid in question:

overlapSubjects 129581912958421295854, and 1295866

Of course, if multiple images are not available of the same field/asteroid from the same night, this task becomes somewhat more complicated, where we might need to retrieve data taken on different nights or even by different telescopes besides Subaru to see if we can identify a background source at the location of the suspected activity.  This is more difficult and may be less conclusive than the method described above because smaller telescopes or shorter exposures or images taken under worse observing conditions may not be sensitive enough to detect the suspected background object, making the check inconclusive.

Another check that can be done is to search for other images of the same asteroid taken at a similar time (e.g., using the SSOIS web tool, which we will discuss in more detail in a future blog post), or if that is not possible, then perhaps at least at a similar point in its orbit.  If similar activity is seen, then there is a good chance that the activity is real.  If not though, then this raises the chances that the original suspected activity was due to an overlapping source, or perhaps just noise or poor observing conditions.

It is of course possible that none of these checks will reveal anything conclusive.  This is one of the limitations of using archival data, that we don’t have control over the way the data is/was obtained, and so sometimes we will simply lack enough information to directly confirm or reject overlap cases.  In these cases, we have to rely on our intuition.  Is the star field crowded or not?  If it is, it suggests that the likelihood of an overlap is higher since there are more potential background objects to overlap.  Is the image quality good or not?  If not, it means that it is easier for a background star to be fuzzy and indistinct enough that it could reasonably be mistaken for a tail.

Finally, we might also ask if the object’s orbit and orbit position at the time of the observations consistent with those of other known main-belt comets.  We want to try very hard to avoid resorting to this particular line of reasoning because it creates a bias against discovering main-belt comets in unusual orbits or at unusual orbit positions, but in the event that no other information is available, we may use this to make a judgement call…i.e., if the orbit and orbit position are consistent with previously known main-belt comets, the object might be considered a stronger candidate than if those are not consistent.  We will still not be able to absolutely confirm that the activity is real, but we can put the object higher on our list of candidates to observe in the future when they again reach a point in their orbits at which activity might be expected.

What do main-belt comets look like?

Main-belt comets have a wide variety of appearances, or “morphologies”, depending on the strength of their activity, nucleus size, angle at which they are viewed from the Earth, just to name a few of the factors involved. As you look through objects in Comet Hunters, you may wonder what kind of features you should be looking out for.  This can be a hard question to answer given the diversity of possible morphologies of main-belt comets though, so in this case, it is perhaps easier to show, rather than tell.

So, first, here is a gallery of representative images of nine of the main-belt comets known to date.

As you can see, main-belt comets can have long, thin tails like 133P, or broad ones like P/2013 R3 or P/2012 T1, or curved ones like 324P or 313P.  Some are strongly active, like 238P and P/2013 R3, while others are only weakly active, like 133P, 176P, and P/2012 T1.  Some may even have two dust tails, like 288P in the above images, or may be in the process of breaking apart, like P/2013 R3.  As you can see, it is not really possible to describe the “typical” appearance of a main-belt comet.

Complicating matters more, the appearance of an individual main-belt comet can vary over time, as its activity strength changes as it approaches and then passes through perihelion (its closest approach to the Sun in its orbit), and/or due to changes in observing conditions between different observations (i.e., on different nights or even different times in the same night, or at different observatories at different locations in the world under different weather conditions).  Below are several series of observations of various main-belt comets taken over periods of several weeks, months, or years using a variety of telescopes (ranging in sizes in terms of primary mirror diameter from 1.5m to 10m) to help illustrate this point:

images_133p133P/Elst-Pizarro in 2002 and 2007 (from Hsieh et al. 2010, Monthly Notices of the Royal Astronomical Society, Vol. 403, p. 363-377)

images_176p176P/LINEAR in 2005 (from Hsieh et al. 2011, Astronomical Journal, Vol. 142, article 29)

images_238p_2005238P/Read in 2005 (a-c) and 2007 (d) (from Hsieh et al. 2009, Astronomical Journal, Vol. 137, p. 157-168)

images_238p_2010238P/Read in 2010 (from Hsieh et al. 2011, Astrophysical Journal Letters, Vol. 736, article L18)

images_324p324P/La Sagra in 2010-2011 (from Hsieh et al. 2012, Astronomical Journal, Vol. 143, article 104)

288P/(300163) 2006 VW139 in 2011 (from Hsieh et al. 2012, Astrophysical Journal Letters, Vol. 748, article L15)

images_p2012t1P/2012 T1 (PANSTARRS) in 2012 (from Hsieh et al. 2012, Astrophysical Journal Letters, Vol. 771, article L1)

313P/Gibbs in 2003 (a-c) and 2014 (d-h) (from Hsieh et al. 2015, Astrophysical Journal Letters, Vol. 800, article L16)

Note: Many of these images are created by adding together several individual images to make a composite image equivalent to leaving the telescope shutter open for up to several hours in some cases.  During this time, the main-belt comet appears to move relative to the background stars.  From the comet’s perspective though, the stars appear to move, and so the series of dotted streaks you see in many of the above images are background stars or galaxies that have been imaged several times (while “moving” between exposures) and then combined together into a single image, keeping the comet at the center of the image at all times.

This rich variety in main-belt comet morphologies is a big reason why we started the Comet Hunters project, given the difficulty of creating computer algorithms capable of identifying several different types of activity.  There are still some things that computers can do better than the human eye (such as measure small differences in the profile “widths” of candidate objects as compared to nearby stars), but we hope that the combination of citizen science and modern computing, we will be able to discover many more new main-belt comets.

Comets Lurking in the Asteroid Belt


Asteroids and comets are small bodies (several hundred miles across, or less) leftover after the construction of our Solar System’s planets. Asteroids are commonly assumed to be mostly rocky or metallic with the majority orbiting between Mars and Jupiter.  Comets on the other hand are a mixture of rock, dust, and ice with orbits more very elongated or hyperbolic orbits that typically originating from the outer Solar System and Oort Cloud (reservoir for long period comets)  As a result of their icy composition, when comets come close to the Sun and get hotter, they become “active” as their ice sublimates (changes from solid to gas), releasing gas and dust, creating the distinctive fuzzy halo (known as a coma) and tails that we associate with comets.

The assumed separation between comets and asteroids was seriously challenged in 2006 with the discovery of main-belt comets (MBCs), which have orbits in the main asteroid belt but have been observed showing cometary activity. Most of the MBCs known today were discovered by telescopes dedicated to surveying the night sky, which so far has been the most efficient way to discover these extremely rare objects. Unfortunately, these dedicated telescopes are rather small (less than 2 meters, or 6 feet, across). Much larger telescopes are available, but these are typically used by individual astronomers to conduct specific observations, and are not used for general surveys. However, the images taken by these telescopes is archived and later made publicly available after a certain amount of time (typically one year), and it turns out that we can re-purpose these observations to search for active asteroids.

Left: Images of the first three main-belt comets (MBCs) to be discovered, each showing the fuzzy appearance and tails characteristic of cometary activity - Image credit: H. Hsieh

Left: Images of the first three main-belt comets (MBCs) to be discovered, each showing the fuzzy appearance and tails characteristic of cometary activity – Image credit: H. Hsieh


Asteroids frequently appear by chance in astronomical observations of other targets (such as stars or galaxies). By compiling accidental, or “serendipitous”, observations of asteroids in archival observations of other targets, we can effectively conduct a “survey” of public data archives using much larger telescopes than are currently available for dedicated all-sky surveys. This is important because we believe that we should be able to discover many more active asteroids if we can detect fainter activity, and the way to detect fainter activity is to use larger telescopes. Detecting activity itself is not easy, however. Comets can have a wide variety of appearances, and while computer algorithms can be designed to detect some comets, it is difficult to design an automated system to detect all the different types of comets that might exist. In contrast, the human eye is much more flexible in terms of spotting different combinations and levels of “fuzziness” and tails that could indicate cometary activity.

For Comet Hunters, we have extracted images of known main-belt asteroids from the archives of the Subaru Telescope, one of the largest telescopes in the world (at 8.2 meters, or 27 feet, across), located on Mauna Kea in Hawaii, one of the best astronomical observing sites in the world. None of the objects we are targeting have been previously known to show activity, but most have so far only previously been studied using small telescopes that may have missed faint activity. By going through and classifying the images on this website, you will help us identify candidates for new comets that we will then analyze in detail and possibly re-observe with follow-up observations to confirm the activity. If we can confirm a comet candidate is real, you will have helped discover a new comet! You will also have helped us along the way to our goal of greatly increasing the number of known active asteroids, and in doing so, contributing to the progress of this new and exciting field of astronomy.

Fancy giving it a try?

%d bloggers like this: