The Galactic Plane and w-band Astrometry

(Up to the IPP for PS1)

MOPS has been reporting astrometry problems in the plane for some time. Richard and Larry pointed me as some problem exposures which exemplified the situation: o6783g0438o and o6783g0456o. I found that several chips failed astrometry and were reported as failing. A number of other chips had poor solutions but were not detected as poor. Looking at the astrometry analysis in detail, I discovered that the reference star selection was poorly matched to the w-band filter in the plane. Specifically, there is a density limitation to avoid excess cross correlation effort. Since the cross correlation is a problem of order N3, this is a reasonable concern. However, in the plane, the greater depth of w-band meant that the selected stars would frequently not overlap the non-saturated gpc1 detections (in magnitude range). Thus, chips could easily fail because there were not enough good stars for a solution, and poor astrometry could result from the limited numbers as well.

I modified the density limits to allow good overlap between reference catalog stars and w-band gpc2 stars. I was able to choose values which did not increase the processing time too much. I ran a large number of tests for 1100 3pi exposures scattered across the survey to prove that the new parameters did not degrade astrometry generally. In fact, the new results were nearly universally better, so I am happy to install them.

This improvement will also help diff operations in the plane. For the bad pair of exposures from Larry, the old parameters resulted in many skycells having 30-60k diff detections (due to dipoles). the new parameters generally keeps the numbers close to 10k.

The following figures illustrate the astrometry failure and the diff detections. I added the exposures to a DVO database, calibrated first with the original configuration. I then calculate the astrometric residuals between the pairs of detections. I also generated a DVO database with the images calibrated using the new psastro parameters. The left image shows (as a density histogram) the astrometric residuals as a function of the magnitude. The right image show the same measurement for the new calibration.

Aside from the overall tighter distribution, the left hand plot has a small excess of detections with w-band magnitude < ~21 and astrometry residual > ~0.2 arcsec. Selecting these detections and plotting them on the sky over the chip outlines gives the left-hand plot below. Adding as red dots all of the diff detections from this pair of exposures gives the right-hand plot. Clearly, where we have the astrometric outliers, we also have

The following table gives the number of sources detected per skycell in 3 diff analyses: the first column is the raw nightly science analysis. The second uses the 1D convolution and improved covariance math, which is supposed to result in better subtractions, but with deeper effective S/N limits. The third uses the improved astrometry as well as the 1D convolution and covariance improvements.

+------------------+---------+---------+---------+
| skycell_id       | sources | sources | sources |
+------------------+---------+---------+---------+
| skycell.1031.032 |    2283 |    3305 |    2069 | 
| skycell.1031.033 |    9406 |   13619 |    7054 | 
| skycell.1031.041 |    2767 |    4905 |    4051 | 
| skycell.1031.042 |    7256 |   14069 |   12032 | 
| skycell.1031.043 |   51606 |   53788 |   12746 | 
| skycell.1031.044 |   32441 |   42455 |   12897 | 
| skycell.1031.045 |   43884 |   55201 |   10456 | 
| skycell.1031.046 |   34249 |   23209 |    5261 | 
| skycell.1031.051 |    6969 |   13470 |   13276 | 
| skycell.1031.052 |    7178 |   13749 |   17357 | 
| skycell.1031.053 |   47787 |   50816 |   16656 | 
| skycell.1031.054 |   28970 |   31802 |   12897 | 
| skycell.1031.055 |   66951 |   56117 |   12299 | 
| skycell.1031.056 |   70969 |   75275 |   13106 | 
| skycell.1031.057 |    9127 |   11141 |    8648 | 
| skycell.1031.060 |    1954 |    2813 |    1958 | 
| skycell.1031.061 |    6909 |   14486 |   12612 | 
| skycell.1031.062 |   17338 |   23238 |   18343 | 
| skycell.1031.063 |   46593 |   53416 |   17135 | 
| skycell.1031.064 |   64355 |   70932 |   15028 | 
| skycell.1031.065 |   85954 |   79187 |   14821 | 
| skycell.1031.066 |   63364 |   67406 |   14358 | 
| skycell.1031.067 |    9634 |   15617 |   19795 | 
| skycell.1031.070 |    4379 |    9043 |    7775 | 
| skycell.1031.071 |   14078 |   25376 |   13941 | 
| skycell.1031.072 |   16118 |   27672 |   17070 | 
| skycell.1031.073 |   13908 |   23794 |   16044 | 
| skycell.1031.074 |   37034 |   46646 |   15478 | 
| skycell.1031.075 |   28113 |   36175 |   14784 | 
| skycell.1031.076 |   18122 |   29506 |   14599 | 
| skycell.1031.077 |    4841 |    8234 |   15442 | 
| skycell.1031.080 |    2862 |    5671 |    4778 | 
| skycell.1031.081 |    7296 |   13274 |    9304 | 
| skycell.1031.082 |    7725 |   13450 |   11396 | 
| skycell.1031.083 |    7485 |   16680 |   14826 | 
| skycell.1031.084 |    7485 |   17699 |   13744 | 
| skycell.1031.085 |    6051 |   15980 |   14279 | 
| skycell.1031.086 |    6422 |   16929 |   14765 | 
| skycell.1031.087 |    3947 |    9906 |   10657 | 
| skycell.1031.090 |    1028 |    2915 |    2568 | 
| skycell.1031.091 |    5044 |   10463 |    9222 | 
| skycell.1031.092 |    4814 |   12820 |   12314 | 
| skycell.1031.093 |    4264 |   11255 |   10921 | 
| skycell.1031.094 |    4219 |   12645 |   11750 | 
| skycell.1031.095 |    3971 |   12508 |   11744 | 
| skycell.1031.096 |    4446 |   12240 |   11508 | 
| skycell.1031.097 |    1556 |    3773 |    3549 | 
| skycell.1120.004 |    1107 |    3295 |    3314 | 
| skycell.1120.005 |    2693 |    8899 |    8935 | 
| skycell.1120.006 |    4127 |   14171 |   13545 | 
| skycell.1120.007 |    3705 |   13819 |   13135 | 
| skycell.1120.008 |    3628 |   10247 |    9817 | 
| skycell.1120.009 |    1422 |    3047 |    2762 | 
| skycell.1120.017 |     601 |    2004 |    1928 | 
+------------------+---------+---------+---------+

Attachments