Axis ratio and PA from ls_dr7.galaxy (more)

There are fields (columns) like fracdev, shapeexp_e1, and shapedev_e2.

First question: in which band(s)?

Second: is the axis ratio simply {x}_e1 divided by {x}_e2 (or maybe vice versa)?

Third: is there any way to determine the PA?

Minor note: typo in shapedev_e2 Description (should be Ellipticity component 2 (not 1).


Re: bands: what we report is a best-fit value across all bands. If the galaxy has different shapes in different bands, what we’ll end up reporting is basically a depth-weighted average.

For the rest, have a look at the “Ellipticities” section here:


Thanks! :slight_smile:

The Ellipticity section is very helpful, and answers my questions.

I’m a bit surprised at the approach you take (best-fit value across all bands), not so much for shapes but profile and size. From my experience with Galaxy Zoo and SDSS, size can vary quite a bit from g, say, to z (for example, a galaxy dominated by blue star-forming regions will look much bigger in g - at low redshift - than in z).

My bigger surprise is with fracdev (fracDeV in SDSS) … you can turn an unambiguous spiral (fracdev = 0) into an elliptical (fracdev = 1) simply by going from the g band to the z band (say) (assuming a redshift below ~0.38). In fact, a pet project I’ve never really done much with is to show that the band-specific trends in fracDeV correlate well with Hubble type (at least the S0/Sa to Sd/m grade, perhaps more for no-bars spirals, SAs). Do you know if anyone in DECaLS has looked at this?

Well, we mostly care about making accurate color measurements of small, faint galaxies. Using the same profile for each band ensures that we are weighting different parts of the galaxy the same, and using all bands to do the fitting uses all the available signal-to-noise. The goal isn’t truly to produce accurate models of galaxies – though if you look at the residual images, I think you might be surprised how well they work for our small target galaxies.

I don’t think anyone on the DECaLS / Legacy Surveys team has looked at morphological questions (except for faint-end star/galaxy separation, where there is intense focus by the quasar targeting team).


Ah yes, of course! For small faint galaxies that makes a great deal of sense.

Morphologies are more important for bigger brighter galaxies, out to z ~0.35 say. And for many spirals, a bulge+disk 2-component model is necessary (strangely, I see little in the literature on modeling bars, even though they are often the dominant morphological feature).

One thing that did strike me on the residuals, of obviously ellipticals, is how many ‘donuts’ there are. I suspect - but have not done the work to confirm - that’s because the Sérsic index needs to be something other than 4 (deVaucouleurs), not because there is a distinct bulge. I should dig up the Tal&vanDokkum paper on “red galaxies at z ~0.34” (my gloss); from memory, the best fit Sérsic index varied as they went from (SDSS) u to g to …

If, one day, the team is interested in morphological questions, I’d be happy to share! :slight_smile:

Yes, it’s on my list to run general Sersic models. There’s some technical work required in the modeling code to make this happen.

An example of some donuts (and a misalignment, the red/green thingie, maybe movement?):

The blue guy in the middle is a star. The “beach ball” residuals are probably because of small misalignments in the images. We do take the Gaia measured proper motions (small, in this case) into account when measuring the individual images.

Thanks. :slight_smile:

The “blue guy” is a zsp=0.364 QSO. I was kinda hoping (well, a bit; it’s too bright and distant to hope for much) that DECaLS would resolve the host galaxy (no sign of any such in SDSS).

I think the donuts/beach balls may be more due to poor modeling than small misalignments: wrong Sérsic index, as I mentioned above; and perhaps for some of those with a bright spot at the centre an AGN (unless from different tiles/bricks, one would expect misalignment to produce a consistent pattern across the whole image, no?)

:slight_smile: Touche’ regarding the blue guy!

Small misalignments would come in the form of offsets, changes of scale/rotation/shear (4 matrix components). There are also higher-order distortion terms (that we don’t re-fit). So the exact pattern for anything but a simple shift with a consistent set of images would be hard to predict.

No doubt some of the galaxies would be better fit with a varying Sersic index, and indeed those give you donuts, or variations of bright-faint-bright ripples.

Thanks! :smiley:

In hindsight, yes of course misalignments may produce a number of apparent features, and only simple shifts/translations would produce a consistent pattern. My bad :frowning: (but hey, if I didn’t make mistakes, how could I truly learn? :wink: )