Galaxy Zoo Starburst Talk

Fuzzy Blob blues :-/

  • Notes by Notes

    All these "Quench" project objects seem to be the smallest, fuzziest most indistinct blobs we see from GZ. Presumably in the next stage someone will start to tell us exactly what we are looking for?

    Posted

  • lamperti by lamperti

    I totally agree! What is there to discuss about featureless blobs? Would also be curious to see the data that does get discussed for a paper. Seems to be a fruitless endeavor so far with waning enthusiasm. Glad that we are over 50% done.

    I'll go back to Space Warps.

    Posted

  • JeanTate by JeanTate in response to lamperti's comment.

    Hmm, I disagree.

    I've found that many of the post-Q candidates are white, which is unusual. Also, rather a lot seem to be small, or have a bright, point-like core (nucleus), which is also unusual. Perhaps even more surprising is that, before looking at the spectrum, I'd've said many have AGNs (active nuclei), only to find that they don't! There seem to be far more with signs of interaction (tidal tails, distorted, close-by galaxies) than among the general population of SDSS galaxies (like what we found in the original Galaxy Zoo, for example).

    And so on ...

    Posted

  • james_involute by james_involute in response to JeanTate's comment.

    I'm very interested to read your comments Jean. I'm know I may be a little enthusiastic to see AGN but, having come from Galaxy Zoo, this seemed to be appearing very frequently in the quench data set, even taking to care to look at the bigger picture for any wide spread imaging artefact or other cause.

    I also feel caught out by the imaging quality of many star's. Perhaps this is simply due to more of the subjects coming from the deep sky with a consequently lower resolution/greater levels of noise. Ok - I'm very much the amateur, but during the course of 4k Galaxy Zoo classifications I felt I'd gained a fairly accurate intuitive feel for identifying stars, even in borderline cases.

    Having been caught out recently I've referred back to SDSS more frequently and been very surprised to find that, to my eye are exemplar point sources classified as galaxies and, equally, objects with (granted with a AGN/QSO core) distinctively imaged area often quite plausible tidal debris, classified as stars. For example:

    SDSS Star Classifications:
    http://quenchtalk.galaxyzoo.org/#/subjects/AGS00001oq (7 o'clock object)
    http://quench.galaxyzoo.org/#/examine/AGS00003w8 (1 o'clock object)

    SDSS Galaxy Classifications:
    http://quench.galaxyzoo.org/#/examine/AGS00003ii (central object)
    http://quench.galaxyzoo.org/#/examine/AGS00000by (central object)

    My apologies if this is not very relevant to the quench science.

    Posted

  • lamperti by lamperti

    Perhaps what really prompted my comments about featureless blobs is the question "Do you want to discuss". There is really not much to say other than being creative about each one. That is too time consuming. I'd rather see all of the data points and try to make something of those, maybe see a pattern, as opposed to taking the time to talk about each one.

    Posted

  • JeanTate by JeanTate in response to james_involute's comment.

    Hi james,

    I'll go through those four in particular later, but just a general comment now: I find the DR8 images much fuzzier than the DR7 ones, but the most recent GZ4 images the best of all (of course, there are lots from poor quality fields, and you can't do anything about them!).

    As I understand it, the DR8 photometric analysis is completely new, done by a pipeline developed independently from the DR7 one. Ditto for the GZ4 images. In both DR7 and DR8, there is a trade-off between classifying too many extended sources ('galaxies') and as point sources ('stars'), and vice versa; there'll always be mistaken IDs, of both kinds. For isolated sources, which are not too faint and are in parts of the sky that is not too 'noisy', the distinction should be pretty straight-forward. Where it often breaks down is where there's something that might be a point source embedded in an extended source, especially if the extended source is clumpy/blotchy. Our eyes are often - but not always - better than the computer algorithms; however, we must also keep in mind that there are, for every object, at least five images (one for each band), and what we get to see is a single image built from just three of the five. More than five? Where fields overlap, there'll be ten, or even 15; in Stripe 82, there could be dozens and dozens of images.

    Posted

  • JeanTate by JeanTate in response to lamperti's comment.

    I swing, sometimes wildly, between wanting to find something to comment on just about every object I get to classify, and asking myself "Comment, really? Stop wasting time! Get on with the valuable work (i.e. classifying)!"

    Posted

  • Mgorman by Mgorman

    I am joining in here because I also recently noted the fuzzball quality of these objects. I toyed with the idea of a fuzzball hashtag, but thought the better of is since it would apply to nearly all the Quench images. But, even allowing for the image issues discussed above, the very fuzziness of these galaxies may be telling a story about their lack of spiral structure and their transition into something else. What do you think?

    Posted

  • lamperti by lamperti

    Is the lack of spiral structure observed due to galactic evolution or low resolution from being so distant and small? Something for discussion in paper.

    Posted

  • JeanTate by JeanTate

    They may be small (physically, when measured in kpc); my impression is that, for the redshifts of the post-Q candidates I've seen, I'd expect more obvious disk-galaxy features (e.g. arms, bars), and bigger objects. To me, the most obvious feature is the near point-like nature of the core (nucleus? baby bulge?) and its relative luminosity (especially for spirals with these redshifts) - bright! Oh, and the color (often white, or delicate pastel shades of white).

    Maybe 'fuzzball with bright pip' is an appropriate description?

    Posted

  • JeanTate by JeanTate in response to james_involute's comment.

    Returning to the four examples:

    Having been caught out recently I've referred back to SDSS more frequently and been very surprised to find that, to my eye are exemplar point sources classified as galaxies and, equally, objects with (granted with a AGN/QSO core) distinctively imaged area often quite plausible tidal debris, classified as stars. For example:

    SDSS Star Classifications: http://quenchtalk.galaxyzoo.org/#/subjects/AGS00001oq (7 o'clock object) http://quench.galaxyzoo.org/#/examine/AGS00003w8 (1 o'clock object)

    SDSS Galaxy Classifications: http://quench.galaxyzoo.org/#/examine/AGS00003ii (central object) http://quench.galaxyzoo.org/#/examine/AGS00000by (central object)

    The corresponding SDSS DR8 ObjIds are, respectively, 1237664668981461196 ("1196"), 1237658304889487382 ("7382"), 1237668589188022437 ("2437"), and 1237654030326104193 ("4193").

    There's something called "petroRad_r" in the panel to the right of the SDSS Explore tool image, bottom row; the unit is arcsecs. This is a quantitative, automated estimate of the object's "size"; the "Petrosian radius in the r band" (DR7 details here). If this size estimate is close to the estimated PSF at this location - the FWHM of the seeing, approximately - then the object is more likely to be a star (or quasar) than a galaxy (or other 'extended source'). Take the seeing as 1.4" (a good general approximation), the first two objects have petroRad_r values of 1.310 and 1.456, respectively; the second two 2.453 and 2.659.

    Pretty open-and-shut, eh? Where this goes off the rails is when the pipeline tries to 'decompose' multiple, or overlapping, objects; a foreground star on a galaxy perhaps, or overlapping galaxies. The algorithms for doing this are pretty darn good, but human eyes are (IMHO) much better. Sadly, there aren't enough of the latter 😉

    The automated pipeline's determination of 'star' vs 'galaxy' actually involves a set of computed values, for parameters with names like "lnLStar_r" and "lnLExp_r"; these can be found by clicking on "PhotoObj" in the left-hand panel (and scrolling waaay down). Interested to learn more about this?

    Posted

  • james_involute by james_involute in response to JeanTate's comment.

    Gosh Jean, many thanks for taking so much time and trouble to illustrate the scientific and algorithmic mechanisms employed to generate the SDSS classifications. Thank you for your gentle introduction to the next tier of object parameterisation. I find the quantity of data available somewhat daunting - but I've not helped myself by jumping straight to the data inspection pages bypassing the substantial documentation. These should provide me with plenty of reading matter (even if a little knowledge continues to be 'a dangerous thing'!).

    Kind regards, James

    Posted

  • Mgorman by Mgorman

    I have also noticed the very bright nuclei. The classifications should give us stats on this, though I am wondering what the comparator arm will be. I recall that the set we are classifying has an equal number of control images. But with regard to your observation about the bright nuclei, I am still wondering "compared to what?". The GZ classifications do collect relevant information but we will need to define the comparator carefully. For example, you have suggested disks. My impression from the Quench project images is that its control images closely resemble the post-quench galaxies. The fuzzieness we have discussed seems a common feature. So what will the comparators be?

    Posted

  • Notes by Notes

    Well all I can say is 'wow', and Thank You all for your helpful and intelligent responses. I do apologise if my OP appeared flippant, I had no idea it would spark such a debate. After GZ I was a little disappointed by the GZQ images, having done a few sessions of classifying. They mostly seem rather at the limits of observable detail (are there no known examples at closer range?) and I wanted to gauge the opinion of fellow quenchites - 'all a bit fuzzy here, how's it going at your end chaps', so to speak.

    Clearly with a little understanding of the available tools (and my knowledge is currently basic), one can derive considerable data, as JeanTate has indicated. I'm not sure how 'tangible' this data is to the average zooite, but no doubt the experts will be able to find what they need. As I have continued to classify the Quench database images I have certainly noticed a preponderance of 'bright' nuclei, I've even hashtagged a couple, though how significant they are I don't know. I need to do some more reading on the subject matter, when time permits.

    Posted