To: 2MASS Calibration Working Group and other interested parties
From: Roc Cutri
Re: ProtoCam Data Photometric Calibration
Photometric calibrations have been evaluated for data from the April/May ProtoCam run. Average extinction corrections for the run were calculated using all stars in calibration fields with repeated observations on each night, and the extinction coefficients agree well with the canonical values for Kitt Peak. Average photometric zero points were determined for each night using the airmass corrected standard star observations. RMS scatter in the nightly zero point measurements ranges typically from 2-5%. The trends in zero point are used to evaluate the photometric quality for each night.
These calibrations have been applied to the ProtoCam data on-line at IPAC.
Standard star observations for calibration purposes were made periodically during each night of the April/May ProtoCam run at the KPNO 50-inch. A standard observation consisted of six one-degree scans roughly centered on the standard star (or stars) repeated in quick succession with approximately 4" cross-scan stepping between each scan. Usually, observations were made at J, H and Ks for each standard pointing. Most standard fields observed contain a single known photometric standard, with the exception of FS16/M67 which contains three (FS15, 16 and 17). The standards observed on each of the photometric nights are summarized in the following table.
Table 1 Standard Star Observation Summary Date / Number of times observed Name 04-22 04-23 04-24 04-25 04-26 05-03 05-04 05-08 05-09 05-10 FS16(15,16,17) 2 2 2 2 1 2 0 0 1 1 S860D 0 0 0 0 0 0 0 0 0 2 P221C 3 3 3 2 0 1 1 1 2 0 FS23 0 0 0 0 0 1 1 1 0 0 FS24 0 0 0 0 2 0 0 0 0 0 FS26 0 0 0 0 0 0 1 0 0 0 Oph_n9(a,b) 1 1(J) 1(HK) 0 0 0 0 0 1 0 P330E 0 0 0 0 0 1 0 0 0 0 FS28 0 1 1 1 1 0 1 0 1 1All "FS" standards are from the UKIRT catalog of faint infrared standards, S860D, P221C and P330E are from the NICMOS standard catalog, and Oph_n9 is a personal communication from Jay Elias. The catalog magnitudes of the UKIRT standards have been converted to the CIT system using the transformations supplied in the catalog. Note that for these stars, we have only catalog K magnitudes and no correction to Ks has been made. The photometry of S860D by Eric Persson (personal comm.) is used, and he does provide Ks magnitudes. P221C and P330E have not yet been calibrated by the NICMOS team. As discussed below, we derive calibrated brightnesses for these objects. No attempt has been made to correct all standards to a common photometric system.
Review of the observer's logs indicate that the nights of April 22, 23, 24, 25, May 3, 9 and 10 appeared to be completely clear - that is there were not obvious signs of clouds at any time during the nights. Clouds were reported at the very end of April 26, the first third of May 4, and for about three quarters of the night of May 8. In this analysis, we examine the calibration data for the nominally photometric time.
We assume that the atmospheric extinction terms were approximately constant in each band for the duration of the run. Photometric zero points are expected to have varied night-to-night.
The global extinction and night zero point corrections were evaluated using the point source data extracted from the cleaned, R1_merged source lists (*.p.ps_s files) from standard star scans. As described in my Protocamera Run Data Processing Update memo of August 21, 1995, these lists contain psf-fit photometry that has been normalized to aperture-corrected aperture photometry. These corrections assume that a single aperture correction and psf describe point sources within the entire scans.
There is evidence for occasional, significant seeing variations within scans in the 1995 ProtoCam data. Such variations lead to photometric errors caused by mismatches between the psf's and some fraction of the point sources in the scan, and because a single aperture correction no longer applies to all of the point source data. Intra-scan seeing variations usually manifest themselves in the aperture curves-of-growth that were evaluated for every scan. When the seeing varies in a scan, the curves exhibit considerable scatter and often fail to converge. We have reviewed the curves-of-growth for all of the calibration scans and have attempted to use only those scans that appear to have stable seeing in the following calibration analysis. Undoubtedly, though some scans with variable snuck in and likely contribute to the scatter in the calibration measurements.
Atmospheric extinction (airmass) corrections were evaluated using repeated
standard star field observations made on the same night at two or more
airmasses. All stars within these scans in the brightness ranges
2MASS data have the virtue of containing many stars in each scan in
addition to the standard star that can be utilized to estimate the extinction
term. Although we do not know a priori the "catalog" magnitude
of every star in the field, we can replace Mcat with a consistent fiducial
magnitude equal to the average observed magnitude for each star over all
of the observations. This approach implies that only repeated
observations of the same field can be used to estimate the extinction
If a particular standard star field was scanned i=1,2...N times, we can
use the average brightness of each star in all observations as the fiducial
magnitude. The first order extinction correction, A, and pseudo zero-point,
B', are then given by a least squares fit to the relation:
M - Mcat = A(X) + B
In this expression, M and Mcat are the observed and catalog magnitudes
for each standard, and X is the airmass at which each is observed.
AVGi(Mi-Mavg) = A(Xi) + B'
evaluated for all stars in all scans of that field. In this expression, Mi is
the magnitude of a star in the ith scan, Mavg is the average magnitude of
that star in all N scans (Mavg = [M1+M2+...+MN]/N), AVGi(...) is the average
value of the magnitude difference (Mi-Mavg) for all stars in the ith
observation, and Xi is the airmass of the ith scan. Note that B' is not the
true photometric zero point in this case.
2MASS data have the virtue of containing many stars in each scan in addition to the standard star that can be utilized to estimate the extinction term. Although we do not know a priori the "catalog" magnitude of every star in the field, we can replace Mcat with a consistent fiducial magnitude equal to the average observed magnitude for each star over all of the observations. This approach implies that only repeated observations of the same field can be used to estimate the extinction correction.
If a particular standard star field was scanned i=1,2...N times, we can use the average brightness of each star in all observations as the fiducial magnitude. The first order extinction correction, A, and pseudo zero-point, B', are then given by a least squares fit to the relation:
Table 2 Median Extinction Corrections A(J) = 0.089 +/- 0.030 mag/airmass A(H) = 0.087 +/- 0.044 mag/airmass A(Ks) = 0.076 +/- 0.035 mag/airmassThe large rms uncertainties on these values most likely reflect measurement uncertainties rather than true nightly variations in the extinction. The scatter in the extinction estimates is probably dominated by inadequate seeing corrections, since most of the high airmass observations during the early part of the run suffered from very poor seeing. Despite the large uncertainties, these values are consistent with "historical" extinction coefficient for Kitt Peak, and we have adopted them for the all of the 1995 ProtoCam data.
The photometry for each standard star in scans that had nominally stable seeing was corrected to unit airmass using the coefficients in Table 2. The first order photometric zero point correction in each band on each night was then taken to be the average value of (Mcat - Mcorr) for all standards, where Mcat is the catalog standard magnitude and Mcorr is the extinction corrected instrumental magnitude (Mcorr=Mobs + A(lambda)*(X-1.0)). Note that approximate photometric zero points that were determined over the last several years of ProtoCam observations were applied to the instrumental photometry in the pipeline processing, so the net zero-point corrections we derive here will be small.
The summaries of the zero point calculations for each night are available for any interested below. The summaries give for each standard observation in each band on a given night: the catalog magnitudes and uncertainties, the observed (aperture-corrected, normalized) magnitudes and measurement errors, the airmass, the difference between the catalog and airmass corrected magnitude (i.e. zero point offset for each star) and the "calibrated" magnitude. The net zero point for the night is given at the end of each band list. The "calibrated" magnitude for each star was determined by applying the net zero point correction to the airmass corrected observed magnitude. I have included entries in these tables for standards that do not yet have catalog magnitudes for informational purposes (i.e. P221C and P330E). Obviously, they do not figure into the net zero point calculations, but their "calibrated" magnitudes, which are enclosed in parentheses, are of useful for observing photometric trends.
Nightly Zero Point Calibration Files...
The table below lists the mean photometric zero point correction and rms deviation from the mean (in magnitudes) in each band for each night, and the number of measurements going into each evaluation. Note that this rms deviation is equivalent to applying the net zero point to each standard measurement on the night, and evaluating the rms difference between the "calibrated" and catalog magnitudes, so it is a good measure of the overall accuracy of the constant zero point fit.
Table 3 Mean Nightly Zero Point Corrections Date J rms N(J) H rms N(H) Ks rms N(Ks) ---------------------------------------------------------------------------- 95-04-22 0.019 +/- 0.019 21 -0.037 +/- 0.021 18 -0.026 +/- 0.023 19 95-04-23 0.134 +/- 0.028 37 0.022 +/- 0.020 40 -0.009 +/- 0.027 38 95-04-24 0.096 +/- 0.039 12 -0.022 +/- 0.040 14 -0.055 +/- 0.047 18 95-04-25 0.132 +/- 0.023 15 0.007 +/- 0.023 18 -0.165 +/- 0.095* 16 95-04-26 0.113 +/- 0.023 28 0.010 +/- 0.047 28 -0.023 +/- 0.046 30 95-05-03 0.054 +/- 0.016 22 -0.019 +/- 0.028 24 -0.040 +/- 0.023 20 95-05-04 0.040 +/- 0.038 20 -0.068 +/- 0.121 24 -0.058 +/- 0.041 23 95-05-08 -0.077 +/- 0.009 4 -0.137 +/- 0.009 6 -0.071 +/- 0.021 5 95-05-09 -0.042 +/- 0.048 18 -0.075 +/- 0.025 24 -0.053 +/- 0.018 25 95-05-10 -0.070 +/- 0.046 36 -0.111 +/- 0.048 37 -0.057 +/- 0.033 40 * - 95-04-25 Ks w/o FS15,16,17 -0.006 +/- 0.011 N=4In Figures 1-9 below are shown the difference between Mcat and Mcal for each standard observation on each night as a function of UT time. Note that in these plots, we include the points for P221-C using the calibrated magnitudes discussed below. However, P221-C was not used in the zero point determinations.
We know from repeatability tests that the photometric precision for individual measurements of high SNR stars is 2-3% under conditions of good, stable seeing. Thus, we expect zero point dispersions of similar accuracy on classically "photometric" nights. It is clear from Table 3 that there are nights where in one or more bands the accuracy of the zero point determination is much worse than the levels expected from measurement error of individual stars alone. The challenge is to determine whether these deviations are caused by real variations in the photometric conditions, observing technique, differences in photometric systems, or some aspect of the data processing. Below, we discuss the quality of each night in the context of these possibilities.
Ultimately, we are at the mercy of the accuracy and consistency of the standard star catalog magnitudes to estimate our zero point accuracy. We know for certain that at least one UKIRT standard, FS15, had listed magnitudes that were inconsistent with those of other UKIRT standards. Since it was covered in the same scans of FS16 and FS17, we examined the relative brightness of FS15 to FS16 and FS17, and found that averaged over the entire run FS15 is 0.039 and 0.027 brighter at J and K, respectively, than reported in the UKIRT catalog. We used the corrected "2MASS" magnitudes for FS15 throughout this analysis. Unfortunately, we do not have the luxury of multiple standards in any other scans, or of having observed many different standards on multiple nights, so detecting systematic catalog magnitude offsets is difficult. I welcome any suggestions for good tests to search for other systematic offsets.
S860D is the only non-UKIRT standard used in the zero point calculation, and the source of its calibrated photometry is different than for the UKIRT data. Both of its measurements appear systematically too bright relative to the UKIRT standards on the one night it was observed, May 10. At least some of these discrepancies may be due to differences in the UKIRT, NICMOS and 2MASS photometric systems. How much, though is not currently known, but this does serve to emphasize the need for a uniform standard star network for 2MASS.
We derive calibrated photometry for the NICMOS standard P221-C by applying the extinction corrections and zero point corrections for each observation taken under good, stable seeing. The average calibrated magnitudes and net uncertainty in the "2MASS" system are as follows:
J: 10.870 +/- 0.005 N=64 H: 10.648 +/- 0.005 N=59 Ks: 10.627 +/- 0.007 N=51where N is the number of independent measurements used in each average. The net uncertainty is the rms scatter in the average divided by the square root of the number of measurements.
In general, for scans taken in reasonably good (<2.5") and stable seeing, and when the observer set filters correctly, the calibrations from the ProtoCam observations can be determined to 2-3%, levels consistent with the photometric precision for bright star measurements. However, seeing variations, real zero point variations and catalog magnitude discrepancies quickly degrade the calibration accuracy to below what is necessary for the Level 1 Requirements.
The plots of zero point correction as a function of time for each night suggest that more than the typical 4-5 standard observations obtained during the ProtoCam run will be necessary during the survey to track better the many possible systematic trends in the photometric quality of the night. The baseline plan to observe 2 standard fields every two hours during the night, making sure to follow at least one field across the sky, appears to be a reasonable minimum. This would provide us with 10 standard fields each with 20-400 sufficiently bright stars (e.g. 2 degree scans) for calibration purposes. It is most probable that we will be able to assemble a reasonable calibration network from the existing faint near infrared catalogs, such as the NICMOS and UKIRT catalogs. However, we must make certain that we understand how our photometric system transforms into those used by others, particularly how the standard magnitudes translate into our system. It is fine to quote magnitudes on the "2MASS system". However, we must be able to calibrate to stars with brightnesses we know to be consistent in the 2MASS system.
The extinction and zero point correction described in Tables 2 and 3 have been applied to all 1995 ProtoCam data obtained on the fully and partially photometric nights. The calibrated data is contained in files having a ".cal" suffix (i.e. j010.p.ps_s.cal). No attempt has been made to subdivide nights into photometric and non-photometric sections so it is advisable to refer to the observer's logs for each night when examining data from partially photometric nights. The calibrated source lists have been combined and band-merged and access to them will soon be available via the 2MASS XCATSCAN service.