Lens position matching via some suitable LRG catalog

[mbaumer]

We use CFHTLS red galaxies to decide where to plop down our simulated lenses on sky, by matching nearest neighbors (in redshift and velocity dispersion--although the latter is not for long) between catalogs.

Apparent cuts on the CFHTLS LRG magnitudes in our current reference catalog ($OM10_DIR/data/CFHTLS_LRGs.txt) limit our matching ability on an OM10-PS1 sample to 73%

@drphilmarshall , can you investigate loosening these cuts to expand our CFHTLS reference catalog? You can use the following plot as a guide:

[mbaumer]

In the future, we might also consider drawing from PS1 itself, or other surveys. See #2

[drphilmarshall]

OK, I made an initial foray, trying to select sensible galaxies from CFHTLS (basically the brightest galaxies in each of 6 broadish redshift bins - I checked that we cannot query by color directly). Here's my notebook, in a CFHTLS branch (that I pulled your notes into as well, Mike).

From this I think I conclude that its best to first paint the OM10 lenses with reasonable colors, extrapolating from SDSS (like we do with the quasars), and then go looking for a suitably faint and correctly colored object to position the system. I worry that we won't have enough red objects unless we go looking for them specifically - I might need to download a lot of CFHTLS galaxies at some point soon.

[mbaumer]

Hey Phil,
The painting script now assigns the colors from the SDSS reference, rather than the magnitudes...sorry for messing up something so simple!
This means you're now good to go to paint a PS1-like sample of lenses using db.paint(), and then find CFHTLS galaxies that match those colors. The painting method now operates on db.sample, so feel free to paint before or after down-sampling!

[drphilmarshall]

Sweet! I'll give it a try :-)

On Mon, Jul 20, 2015 at 12:02 PM, Mike Baumer notifications@github.com
wrote:

Hey Phil,
The painting script now assigns the colors from the SDSS reference,
rather than the magnitudes...sorry for messing up something so simple!
This means you're now good to go to paint a PS1-like sample of lenses
using db.paint(), and then find CFHTLS galaxies that match those colors.
The painting method now operates on db.sample, so feel free to paint before
or after down-sampling!

—
Reply to this email directly or view it on GitHub
#22 (comment).

[drphilmarshall]

Downloading CFHTLS galaxies is getting painful. Let's switch to using SDSS LRGs for the sky positioning - and start using the painted-on lens colors to find a good LRG to use when simulating images.

[mbaumer]

Try matching on one color; check that others are sensible?

[drphilmarshall]

Hmm. I downloaded the SDSS LRGs and checked them in. They're very bright! Mostly brighter than i=19 or so. This is too bright for PS1QLS. So, back to CFHTLS. I think the next thing to do is to use the SDSS-painted OM10 lens colors to define the CFHTLS wget statement, in a download_CFHTLS_LRGs() function. I can work on this while you upgrade the nearest neighbor matching, @mbaumer.

[mbaumer]

That’s unfortunate, but I can work on the matching algorithm in parallel while you sort out the CFHTLS selection. Thanks for doing all that wget legwork!

On Jul 27, 2015, at 8:33 PM, Phil Marshall notifications@github.com wrote:

Hmm. I downloaded the SDSS LRGs and checked them in. They're very bright! Mostly brighter than i=19 or so. This is too bright for PS1QLS. So, back to CFHTLS. I think the next thing to do is to use the SDSS-painted OM10 lens colors to define the CFHTLS wget statement, in a download_CFHTLS_LRGs() function. I can work on this while you upgrade the nearest neighbor matching, @mbaumer https://github.com/mbaumer.

—
Reply to this email directly or view it on GitHub #22 (comment).

[drphilmarshall]

OK, I upgraded the download script and it is running. Can you try it out as well, @mbaumer, and see whether the new CFHTLS_LRG.v2.txt catalog works better in the NN position match? We might still need to go to downloading the data in redshift bins as well as magnitude bins... Thanks!

[drphilmarshall]

BTW You'll see from the issue22 notebook that download errors are common - it still might be worth looking for a better way to get all the CFHTLS galaxies we need...