I. Introduction

Scans of 4 southern tiles were taken at high airmasses from Mt. Hopkins on the night of 001016 UT for the purpose of comparing with observations of the same tiles made near the zenith earlier at CTIO. The table below summarizes the tiles and northern and southern observations, the starting airmass (X) at which they were obtained, and the quality score assigned each scan (Q). As a reminder, the survey tile numbering strategy places tiles in the southern equatorial sky observed from Mt. Hopkins in the range 100000-199999. Southern sky tiles observed from CTIO are numbered 300000-399999. For example, tiles 112000 and 312000 refer to the same patch of sky, observed from the northern and southern facilities, respectively.

II. Data Reduction and Analysis

All of these scans were placed through nominal pipeline processing and the standard calibrated point source output files (sNNN.cal) were used for the following analyses.

a. Mean Photometric and Position Offsets

The mean photometric and positional offsets between the high and low airmass scans of each tile were evaluated. For this comparison, only 3-band detected sources having photometric SNR>10 (mag-sigma < 0.1086) not flagged as being or influenced by and artifact were used. The table below lists the trimmed average J, H and Ks magnitude offsets and RMS and the trimmed average RA and DEC offsets and RMS. A 2-sigma clipping algorithm was used in evaluating the means to minimize the impact of outliers. Also listed is the number of sources going into each average.

b. Photometric and Position Offsets as a Function of Magnitude and Scan Position

The plots below show the distribution of magnitude and positional offsets for each star in the tile pairs plotted as function of source brightness in each band and absolute position. For each tile pair listed, the first plot is the photometric offset versus magnitude and DEC, and the second plot is the position offset version RA and DEC. Each panel in the plots contains a line at zero offset and a line at the mean magnitude and position offset value indicated in the table above. Click on the "X" to see the plots.

3. Discussion

a. Photometric Offsets

All of the tile pairs show small systemmatic biases in brightness between the new, high airmass and earlier low airmass scans in the sense that sources in the high airmass scans are slightly fainter than the earlier measurements. The magnitude offsets are statistically significant in the J-band, marginally significant at H in some of the tiles, and arguably insignificant in K_s. The plots show that there is little apparent dependence on the photometric offset with either source brightness or in-scan position (DEC). Although the plots are not shown here, there is also no dependence in magnitude offset on RA.

The caveat to this discussion is that the last two high airmass scans made on 001016n received poor quality assessment during nightly QA. They would have received scores of "1", but were artifically raised to "3" in order to put them into the Working Database for comparison purposes. Some care will have to be taken to not propagate these data.

The photometric offsets' inverse dependence on wavelength suggest relationship to either the nightly zero points or the atmospheric extinction coefficients. The extinction coefficients are largest in the J-band, and the nightly and intra-night variation in atmospheric transparency (zero point) are much larger in J than in the other bands. For reference, the table below lists the nightly zero point uncertainties (RMS) for each scan, along with links to the nightly zero point plots.

The root-sum-square of nightly zero point uncertainties can account for about half of the observed J-band offsets between the high and low airmass scans. It is of interest that the largest J-band offset occurs for the 114700/314700 tile pair, and the CTIO data had the largest nightly J-band zero point uncertainty.

Can the resulting J-offsets arise from using inappropriate atmospheric extinction coefficients? To account for the remaining J-offset between tiles 114700 and 314700, the largest bias in the group, would require the northern extinction coefficient to be approximately 30% too large (~0.03 mag/airmass). Seasonal drifts of that amplitude are seen in the plots in John Gizis' 2MASS Extinction Corrections page, but it is not obvious if the October 2000 data were taken during a period when the extinction is significantly different than the costant values used for the northern pipeline. For reference, the nightly J-band extinction coefficient measured for 001016n is 0.096 mag/airmass, is very close to the mean pipeline value of 0.109 mag/airmass, though.

We can examine the photometric offsets as a function of in-scan position in the high airmass scans to look for evidence of residual extinction terms not removed by the standard pipeline calibration. The two plots below show the J-band photometric differences as a function of airmass in the high airmass scans calculated for each source scans (small points), along with the trimmed-average offset in six equal length airmass intervals within each scan (large solid points). The errorbars on the trimmed average offsets represent the RMS (sigma-population) in each airmass bin. For the purpose of these calculations, we assumed that each scan was taken at a constant hour angle equal to the midpoint of the scan.

112000/312000 and 112001/312001 J-band Offsets versus airmass
114700/314700 and 114701/314701 J-band Offsets versus airmass

The following table gives the trimmed average J-band magnitude offsets, rms of the offsets, and the number of stars used in the trimmed averages in each of the six airmass intervals for the scan pairs. The first sets of columns refer to tiles 112000/312000 and 112001/312001, respectively, and the second sets refer to 114700/314700 and 114701/314701, respectively.

Linear regression fits to the trimmed average points in the four scan pairs give slopes of -0.046, -0.052, +0.018 and -0.062 mag/airmass, respectively. However, the in-scan structure seen in the offset plots indicates that the formal slopes measurements are low confidence, at best. Taken at face value, though, they suggest that the high airmass scans may be over-corrected by about 40-60% from the nominal slope of 0.109 mag/airmass.

b. Positional Offsets

The positional comparisons between low and high airmass scans show no consistent biases. In most cases, the average offsets are insignificant. However, the position offset plots show that there is differential structure dependent on cross-scan (RA) and in-scan (DEC) position.

All of the scan pairs exhibit a similar bowed cross-scan dependence in RA offset, and a much smaller bowing in DEC offset that is a function of RA. Howard McCallon indicates that this structure is due to the difference in the distortions of the northern and southern cameras/telescopes. Recall that position reconstruction in the preliminary pipeline does not include distortion corrections, but the final processing will include them.

The positional offsets in all of the scan pairs exhibit high-frequency structure of varying amplitude as a function of in-scan position (DEC). The most dramatic is the drift seen in the northern end of the 001016n/s101-990726s/s150 comparison. This deviation occurred because 001016n/s101 just missed covering an astrometric reference star at the north end that 990726s/s150 did cover, so the astromeric solution was able to random walk off in the 2000 scan. This highlights the need for using the overlap information in the final pipeline processing.