5 The photometric sequences

The experience acquired with the reduction and analysis of thousands of CCD frames from different observatories and telescopes led us to the conclusion that a thorough assessment of the photometric quality of each CCD object must be complemented by visual inspection of the frame data. On the other hand, the use of an automated pipeline has been essential to the project, by making feasible the re-reduction of a large part of the data with minimum effort, after some critical input values were established. We also verified that both PSF and aperture photometry carried out in a semi-automated way with the standard IRAF packages can have systematic errors, and a simple, yet effective rule to detect such errors is to check the two magnitudes against each other. Basically, we have adopted a set of tools for Quality Assurance (QA) of the sequences mainly serving the purpose of finding gross errors. Additionally, we have checked a number of common fields re-imaged at most observing runs in order to monitor the stability of the different equipment used and to ascertain all-sky data homogeneity.

Figure 3: The 622 GSPC-II fields, binned according to the limiting magnitude of the sequences

Figure 4: Average number of GSPC-II objects per sequence as function of galactic latitude (upper and lower panel refer to Northern and Southern hemispheres respectively)

5.1 Selection criteria and caveats

The criteria used for the selection of the sequences released with this paper are the following: a) sigma of unit weight of the fit to the Landolt stars to better than 0.05 mag; b) at least 6 stars per CCD sequence; c) single object instrumental error in each filter to better than 0.15 mag; d) agreement between PSF and aperture photometry.

The details of the implementation of criterion d) are explained in the following. For each frame a linear transformation between the PSF and aperture magnitudes is carried out according to the model $M_{\rm apt}-M_{\rm psf} = z_0 + k \cdot M_{\rm apt}$. Then, a 3-sigma rejection criterion is used to estimate the parameters of the fit, and their corresponding errors. Only sequences with |k| < 0.01, and rms (sigma of unit weight) < 0.3 in all filters are retained. These thresholds have been empirically determined, and have proven to be reliable for removing frames characterized by a poor choice of the PSF profile function. Once the frames have been selected, multiple observations (i.e., those coming from different IFCs) are averaged using equal weights to produce the best current estimate of the stellar magnitudes. As already mentioned, these sequences are primarily intended for the calibration of Schmidt plate surveys. Therefore, they should be used as a whole to obtain a calibration curve. Accurate single object photometry can only be ensured by dedicating more time (in term of human resources) to the analysis of each single frame.

Figure 5: Average differences between catalog and fitted magnitude and colors of the Landolt standard stars imaged at every GSPC-II observing run from the Kitt Peak, Cerro Tololo and ESO telescopes

Figure 6: Mean yearly extinction values in the B, V and R passbands for the Kitt Peak, Cerro Tololo and ESO-La Silla observatories

Figure 7: Cumulative distributions ($\%$) of the zero-point error of the transformations to the standard system for the B, V, and R passbands, from left to right respectively

5.2 Variability

A related issue is the handling of candidate variable stars. As expected, our database will contain a percentage of variable stars with more than one observation over the years, which can be used to detect such variability. While the ultimate goal in such cases would be to distinguish between genuine variability and photometric errors of various nature, variability is not taken into account in the current catalog release, and magnitudes coming from different frames are simply averaged to obtain the final photometry.

5.3 Global statistics

The results reported here refer to the photometric sequences selected with the criteria described in the previous subsections. Data are from 7 different telescopes (see Table 1; data from Mt. Hopkins, Lowell, and McDonald Observatories were not included in this release), for a total of 153 observing nights, 1285 IFCs and 622 different GSPC-II fields. The open circles of Fig. 1 show the equatorial position of the centers of the selected fields. The total number of catalog objects, after averaging multiple observations (see Sect. 5.1), is 219942. The histogram of Fig. 3 represents the number of sequences as function of limiting magnitude in the V passband. The distribution is well peaked around $V_{\rm lim} \simeq 20$. The 3 bright sequences ( $V_{\rm lim} < 18$) correspond to southern fields S171, S226 and S311 for which only short exposures are available in the present catalog version. The average number of objects per sequence is displayed in Fig. 4 as function of galactic latitude. It can be noted that the northern sequences are on average more populated than the southern ones, due to the larger CCD formats used.

In Fig. 5 the mean differences between the Landolt catalog values and the ones estimated by the nightly fits of the observations are plotted as function of color for the complete set of Landolt stars used to reduce our data. For each star, an empirical estimation of the error on the fitted Landolt photometry is obtained by summing the differences in quadrature, and the resulting error bars are plotted. As the figures show, the general agreement of the Landolt catalog stars with their fitted values is within few percent, and there is no evidence of residual systematic color or magnitude effects at this level of accuracy.

As the observations have been carried out over several years, we looked for possible long-term trends in the value of the extinction coefficients. Average estimates of these values as a function of time are reported in Fig. 6 for the Kitt Peak, Cerro Tololo, and ESO-La Silla telescopes, for which the data statistics are more significant. It is interesting to note that the extinction (in all filters) show the effects of the 1991 eruption from the volcano Pinatubo in the Philippines. As explained in Sect. 4.3, an important contribution to the final photometric error comes from the zero-point error of the transformation between instrumental and standard photometry. In Fig. 7 the cumulative distribution of the zero-point errors, as estimated by the IRAF task FITPARAM, is shown for all the selected nights. The percentage of nights with zero-point errors smaller than 0.06 is $\sim$95%.

Figure 8: GSPC-I/GSPC-II B magnitude (top panel) and V magnitude (bottom panel) residual plot

Figure 9: Distribution of GSPC-I/GSPC-II magnitude differences (V mag: solid line; B mag: dashed line)

5.4 Photometric accuracy

The first independent check of the photometry quality of our data has been obtained by comparison with the faint GSPC-I stars, corresponding to the bright end of the GSPC-II sequences. As GSPC-I objects appear saturated in our long CCD exposures, we have pre-selected 5 and 4 min exposures (or shorter) in the B and V band respectively before matching GSPC-II objects with GSPC-I. The match resulted in 597 objects in the V passband and 380 objects in B. From this comparison we detected a few outliers, some of which could be explained by poor observation history of the GSPC-I stars; globally, the agreement between GSPC-II and GSPC-I photometry is at the level of few percent, as the residual plots and histograms of Figs. 8 and 9 indicate. The statistics computed on these residuals, without eliminating any outlier, give an rms of 0.07 mag and a mean of 1 millimag in the V passband, while the analogue values for the B filter are 0.06 mag and 3 millimag.

Figure 10: Comparison of 46 common fields between CTIO and ESO. Each point in left panels represents the difference between two standard magnitudes of the same star as measured independently at CTIO and ESO, in the B, V, and R passbands. The graphs on the right show the corresponding $1-\sigma$ photometric error per magnitude bin, as measured from the dispersion of the points in the corresponding the left panel

Figure 11: Comparison of 10 common GSPC-II fields between CTIO and Kitt Peak observatories in the V and R filters. Only one common field (N507) was available in the B passband. See explanatory caption of Fig. 10

To test the average photometric quality of the standard sequences down to the magnitude limit, we have intercompared observations of the same GSPC-II fields coming from different runs at the main telescopes used for this program. After the selection criteria delineated in Sect. 5.1 had been applied, we found 46 GSPC-II fields in common between the CTIO and ESO-La Silla observatories, for a total of 150 different IFCs (see Sect. 4), and 10 GSPC-II fields - only one common field in the B filter - shared between KPNO and CTIO, corresponding to 33 IFCs. The magnitude differences of all objects in common between different IFCs are plotted in Figs. 10 and 11 for each passband and for the two inter-observatory comparisons. Then, by binning the stars into appropriate magnitude ranges we estimated the single-object photometric error as a function of magnitude, which is reported in the right panels.

The graphs show fluctuations of the estimated error vs. magnitude which reflect the presence of some outliers, although the general behaviour of the curves is fairly stable, with an error at 19 mag of approximately 0.07 mag. Possible causes of these outliers, which are not identifiable with any particular field or CCD frame, are: object misidentification, crowded-field effects, stellar variability. It is conceivable that, by devoting more time to the quality control of the data, a large part of such outliers can be detected and removed from the sequences. As some of them will go undetected when we have only one GSPC-II frame, it is important to note that GSC-II reductions have a considerable robustness to them because there typically are many faint stars in each sequence - and in particular in the crowded ones - so that the outliers will have large residuals in the photographic photometry and will therefore be rejected. As a last comment, photometry which suffered from light contamination by a close object(s) on the CCD frame is usually excluded from the calibration process, since those stars would appear as blends in the Schmidt photographic plate. Such cases will be detected and removed by comparison of aperture and PSF photometry in future catalog releases.

5.5 The export catalog

As outlined in Sect. 5.1, multiple observations of the same GSPC-II stars have been averaged, and the mean value is taken to be the best estimation of the object's magnitude. Each catalog record contains an object identifier, according to the GSPC-II nomenclature given in Sect. 4.1, its equatorial coordinates at equinox J2000, plus its magnitude and magnitude error in the B, V, and R bands respectively. When the B magnitude information is not present, the corresponding fields are left blank. The reported magnitude error (in each passband) is obtained by the formal error propagation of individual standard errors coming from the reduction of different sets of exposures pertaining to that object, as computed by IRAF. Since, as pointed out in Sect. 3.3, the standard error computed by IRAF can be an underestimate of the true one, we also computed an empirical $\sigma_{\rm mag}$ when more than one IFC was available. Objects for which the empirical photometric error exceed 0.25 mag, and which represented only $1\%$ of the data set selected by the criteria exposed in Sect. 5.1, did not enter the export catalog.