5.3 Photometric accuracy

5.3.1 Comparison with SDSS-DR8

The photometric calibration of the Deep survey fields is fully described in Section 3.7. We aim in this section at conducting an external comparison to the SDSS survey: wherever possible, magnitude offsets between common stars in the CFHTLS Deep and SDSS-DR8 have been measured. SDSS magnitudes are first converted to the MegaCam AB system using the MegaCam-SDSS color transformation equations presented in the section comparing the Wide to SDSS-DR8 (Section 4.4.3, Equation 13).

The offsets are quoted in Tables 17 and 18 and in the synoptic table. Only fields D2 and D3 have sources in common with SDSS-DR8. The offsets are found to be similar to the means Δm=u*,g,r,i∕y,z(Table 10) or δm=u*,g,r,i∕y,z(Table ??) found for the Wide fields.

In both the Deep and the Wide surveys, the CFHTLS to SDSS-DR8 comparisons are made using stars with 17 < i < 21. The signal-to-noise ratio of all common stars on the Deep fields is much higher than for the Wide, hence it is reassuring to find very similar offsets in all filters for both the Deep and the Wide surveys. As explained in the Section 4.4.3 on the comparison between the Wide and SDSS-DR8, the CFHTLS is anchored to the SDSS Supernova Survey which is established in a true AB magnitude system, while SDSS-DR8 is known to exhibit some offsets to the AB system, especially in the u* and the z bands (Equation 14), offsets that we identify indirectly here. Please refer to Section 4.4.3 for a complete discussion on the topic.

—————-


PIC PIC

Figure 43: Comparison of the (u* - g)(g - r) (left) and (g - r)(r - i) (right) stellar color-color tracks of the four Deep Dk-85 fields. The tracks are shown in different colors to make shifts between fields easier to see. Green, blue red and magenta points are D1-85, D2-85, D3-85 and D4-85. The agreement is excellent. All points overlap and the distinction between the color-color tracks is difficult. The left panel show D4-D3-D2-D1 over-plotted, in that order, so D4 (magenta) is background, and D1 (green) is foreground. The right panel presents the opposite.

The stellar color-color tracks of the four fields are remarkably similar. No color offsets can be measured between D1, D2, D3 and D4. This is illustrated on Fig. 43. It shows a superimposition of the (u* - g)(g - r) and (g - r)(r - i) color-color plots for the four D-85 stacks together.

5.3.2 Internal photometric errors

The internal errors have been measured by comparing the photometry of common sources in the D-25 and D-85 stacks. The D-25 and D-85 parent samples comprise all source pairs listed in the merged catalogues of the four Deep fields. The selected sources must have magnitudes measured in all filters (no objects with magnitude/mag_err = 99.0 in any u*, g, r, i, y and z filters, for both the D-25 and D-85 catalogues), with a SExtractor extraction flag FLAGS=0 (objects with good photometry) and a Terapix object flag=1 or 0 (i.e. stellar or extended sources, only non-saturated objects in non-masked regions). Only sources fainter than u*,g=20.5 and r,i,y,z=19.5, and brighter than the D-25 80% completeness limit of extended sources are selected. This cut ensures the homogeneity and completeness of the D-25 and D-85 populations.

The internal photometric errors are derived from the distributions of magnitude differences of source pairs between the D-25 and D-85 catalogues, D25–D85, as a function of D-25 magnitudes. The analysis are carried out in three magnitude ranges, for all filters, and after a 3-σ clipping over the distributions of magnitude differences. The statistics use the mag_auto magnitudes to derive first the median and mean systematic magnitude offsets between D-25 and D-85, then the mean scatter, based on the rms. We verified that the median systematic offset is randomly distributed around zero and never exceeds 0.013 mag., for all sub-samples.

The distributions are then fitted by a Gaussian, which gives the FWHM of the mean magnitude difference of cross-identified sources in D-25 and D-85. The mean internal photometric error is then defined as σD25-D85 =FWHM2.35. We checked that the error estimate based on the Gaussian fitting is very close to the rms errors. The results are summarized in Table 21. Overall, they look very similar to the Wide survey, when sources with same signal-to-noise ratio are compared.

We use the results quoted in Table 21 to compute the photometric errors listed in the summary Tables 17 and 18. The internal photometric errors are the mean values of the three magnitude ranges. This is a reasonable but probably optimistic estimate because it does not take into account there are many more faint than bright objects. Note that these results are valid for both stellar and extended sources.

In contrast with the Wide survey, we do not have many sources in common between CFHTLS and SDSS in D2 and D3 to estimate accurately external errors. In addition, we do not have any common sources at all for D1 and D4. However taking into account the statistics for the internal errors and the remarkable consistency of the results with the Wide survey, conclusions for the Wide survey apply equally: we consider the following errors are reasonably accurate and conservative estimates of the rms photometric errors over a rather broad magnitude range of the CFHTLS Deep survey down to the 80% completeness limit (details in Table 21):

For faint sources beyond the 80% completeness limit, photometric errors rise by a factor of ~ 2 in all bands with respect to the magnitude range quoted above and in Tables 17 and 18.


PIC PIC PIC
Figure 44: Estimation of the internal photometric accuracy in Deep stacks by comparing the the mag_auto magnitudes of sources in D4-25-i and D-85-i. The sample comprises all (stars and galaxies) common sources down to the 80% completeness limits for extended sources of the D-25 stack (sharp cuts on the left panels). The catalogue is then split into three magnitude bins (from top to bottom, the red dots: 19.5 < i < 21.21, 21.21 < i < 22.71, 22.71 < i < 24.21), and for each bin we analyze the i-band magnitude differences D25–D85 as a function of magnitude (left panels). The right panels show the distribution of magnitude differences. The histograms are first sigma-clipped and then fit by a Gaussian that returns the fwhm of the distribution. The mean internal photometric error is then defined as σD25-D85 =fwhm2.35.











Filter
CFHTLS T0007 Deep data
D1
D2
D3
D4
mag. range σ D25-D85 mag. range σD25-D85 mag. range σD25-D85 mag. range σD25-D85
[AB MegaCam] [mag.] [AB MegaCam] [mag.] [AB MegaCam] [mag.] [AB MegaCam] [mag.]









u* 20.50 - 22.42 0.011 20.50 - 22.54 0.011 20.50 - 22.35 0.013 20.50 - 22.27 0.014
22.42 - 23.92 0.023 22.54 - 24.04 0.021 22.35 - 23.85 0.025 22.27 - 23.77 0.026
23.92 - 25.42 0.062 24.04 - 25.54 0.053 23.85 - 25.35 0.060 23.77 - 25.27 0.068
g 20.50 - 22.19 0.008 20.50 - 22.20 0.006 20.50 - 22.19 0.007 20.50 - 22.16 0.007
22.19 - 23.69 0.014 22.20 - 23.70 0.012 22.19 - 23.69 0.015 22.16 - 23.66 0.014
23.69 - 25.19 0.035 23.70 - 25.20 0.030 23.69 - 25.19 0.037 23.66 - 25.16 0.037
r 19.50 - 21.58 0.007 19.50 - 21.58 0.006 19.50 - 21.59 0.006 19.50 - 21.55 0.007
21.58 - 23.08 0.012 21.58 - 23.08 0.010 21.59 - 23.09 0.011 21.55 - 23.05 0.012
23.08 - 24.58 0.030 23.08 - 24.58 0.026 23.09 - 24.59 0.029 23.05 - 24.55 0.031
i 19.50 - 21.22 0.008 19.50 - 21.28 0.007 19.50 - 21.29 0.007 19.50 - 21.21 0.008
21.22 - 22.72 0.016 21.28 - 22.78 0.011 21.29 - 22.79 0.014 21.21 - 22.71 0.014
22.72 - 24.22 0.040 22.78 - 24.28 0.027 22.79 - 24.29 0.034 22.71 - 24.21 0.034
y 19.50 - 21.18 0.012 19.50 - 21.29 0.010 19.50 - 21.26 0.007 19.50 - 21.12 0.007
21.18 - 22.68 0.019 21.29 - 22.79 0.012 21.26 - 22.76 0.015 21.12 - 22.62 0.016
22.68 - 24.18 0.040 22.79 - 24.29 0.028 22.76 - 24.26 0.037 22.62 - 24.12 0.042
z 19.50 - 20.73 0.010 19.50 - 20.80 0.008 19.50 - 20.61 0.011 19.50 - 20.66 0.009
20.73 - 22.23 0.023 20.80 - 22.30 0.020 20.61 - 22.11 0.026 20.66 - 22.16 0.020
22.23 - 23.73 0.061 22.30 - 23.80 0.050 22.11 - 23.61 0.071 22.16 - 23.66 0.059










Table 21: Analysis of the internal photometric errors in the CFHTLS T0007 Deep data from the comparison of source pairs in D-25 and D-85. For each Deep field, the internal error is estimated from the mean mag_auto magnitude differences of common sources in D-25 and D-85. The statistics are based on a 3-σ clipping and a Gaussian fitting of the distribution of magnitude differences. The best fit returns the fwhm of the distribution. The mean internal photometric error is then defined as σD25-D85 =fwhm2.35. Each sample is split into three magnitude ranges, but does not make any morphological selection, so the internal errors quoted in this table are valid for any sources. The bright cuts are AB=20.5 for u* and g and or AB=19.5 for r,i,y and z, while for the faint sources, the cut is set by the 80% completeness limits of extended sources in the D-25 stacks, so both D-85 and D-25 are complete up to the depth of the analysis.