SV-A1 Limiting Magnitudes: Updated 9/16/13¶
If we take the SVA1 photometric errors at face value, what is the 10 sigma galaxy detection limit over the full area? And how much area is deeper than any given limiting magnitude? Keep in mind that this is separate question than what the completeness limit is...there are also caveats that the errors may not be correct (though with non-homogenized coadds they should be close).
I've started by only taking SVA1 objects that are observed in g, r, i, and z. I'm not using Y at the moment. Note that due to a bug in sextractor, some objects are listed as detected (via model/detmodel mags, though flux_auto == 0) even though they're not. The way to filter these out is to require NITER_MODEL_[GRIZ] > 0.
I then pixelize the sky using the mangle simplepix scheme (resolution = 8, corresponding to ~0.5 degree pixels). In each pixel I plot mag_autoerr vs mag_auto for each band (corrected for Galactic extinction), and fit a simple model with an effective exposure time and 10 sigma limiting magnitude.
First, for non i-band magnitudes I am not plotting DETMODEL magnitudes, appropriate for colors, which is more relevant for most of our science. (E.g., we are not using g band to detect objects...)
Second, regarding Huan's e-mail from 9/13/2013 to des-astro et al: Huan is correct that the way I was calculating limiting magnitude was not quite appropriate for the detection threshold. What we need to do is the following (as detailed in other of my limiting magnitude wiki pages Depth Measurements): calculate the 10sigma limiting magnitude in a 1.5" aperture (MAG_APER_3_I). Then calculate the aperture correction for these galaxies from MAG_APER_3_I to MAG_AUTO_3_I. After applying the aperture correction, this gives us the appropriate "detection" limiting magnitude appropriate to apply to the DES specifications. This typically yields an extra ~0.3 magnitude depth (I find a slightly smaller difference than Huan's 0.4 mag, but the improvement depends on the field and the seeing.)
The Results (updated!)¶
The plots with "dm" (for detmodel) in the name at the bottom of the page show the updated depth maps. I have the detmodel depth maps for g,r,i,z. In addition, "i_tot" is the total magnitude map as before. (This is probably most appropriate for simulating mock catalogs.) And "i_aper+" is the corrected aperture magnitude method.
Figuring out which regions are truly "full depth" is a bit tricky without mangle masks. However, from eyeballing, it looks like the corrected aperture depth in the close-to-full-depth regions is ~23.4-23.5. This is consistent with Huan's estimation. We will gain a bit by observing in darker skies for i-band imaging.
Finally, I should emphasize that this (a) all depends on the errors being correct, and (b) makes no representation for the completeness of the catalog at these magnitudes.
The Results (original)¶
The plots at the bottom of the page show a histogram of depth values for each band. The average depth in g is 23.8, in r is 23.6, in i is 22.8, and in z is 22.1. Of course, much of this area isn't full depth (10 tilings). The supernova fields are the deepest in all bands (due to being > full depth).
I've also created maps of the depth in the four bands, with another set zoomed in on the SPT-E region. You can see here that the full depth parts of SPT-E reach i=23.0-23.2. This is (as before) considerably shallower than our target of 23.7+ mag. Some of this is due to less than ideal image quality during the SV season, some is due to sky brightness, and some may be due to...something else. Though tests on individual images previously show that the depth does scale as expected with number of images, provided the sky is sufficiently dark.
Finally, I've created plot of the total area brighter than a given cut in any band. There are ~400 deg^2 total. About 100 deg^2 are deeper than 23 in i band.
And Now Using Tiles!¶
I've repeated the tests using the coadd tiles rather than the simple pixelization scheme, files attached. I have also plotted the nominal EXPTIME header value for each of these tiles. Finally, I have correlated the limiting magnitude with the EXPTIME values. Note that the way to do this correctly would be to use the mangle masks.
In fact, I'm not entirely sure what the EXPTIME value is: half the images have EXPTIME > 1000, which is nominally full depth....
In any event, the depth is correlated with EXPTIME, but there's a lot of scatter (~0.5 mag rms).