Avoiding common pitfalls

Stars should be excluded

As a general rule, any stars within a field should be masked out when using the photometric image-fitting method of the imphot module. This can be done using a region file, as described in the Regions section). The reasons for excluding stars are described below.

The image fitting procedure assumes that all differences between a MUSE image and an HST image, are due to their different PSFs, their different calibration errors and different pointing errors. Stars can break this assumption in two ways:

  1. Stars that have significant proper motions may appear to have moved between the HST and MUSE observations of a field. For example there were 11 years between the HST observations of the UDF and the start of the MUSE observations of the UDF. In that time, all of the stars in the UDF moved significantly. Estimated proper motions of these stars were reported in [1] by Pirkal et al (2005), and the ones seen in the MUSE UDF fields are listed below.

    RA (deg)

    Dec (deg)

    HST ID

    Field

    Flux [erg/Å/s/cm²]

    Type

    PM (mas/year)

    53.1630

    -27.7672

    9397

    UDF01

    12.3e-18

    QSO

    0.70 ± 1.47

    53.1580

    -27.7692

    9230

    UDF01

    21.0e-18

    Star

    10.26 ± 0.41

    53.1485

    -27.7702

    9212

    UDF04

    1.1e-18

    Star

    12.35 ± 0.54

    53.1583

    -27.7949

    3166

    UDF05

    49.0e-18

    Star

    8.84 ± 0.66

    53.1766

    -27.7997

    2150

    UDF06

    17.3e-18

    Star

    25.16 ± 0.45

    53.1323

    -27.7829

    5921

    UDF07

    56.3e-18

    Star

    3.27 ± 1.26

    The proper motions of these stars range from 3.3 to 25.2 mas/year. Over the 11 years that elapsed between the HST observations and the MUSE observations, this corresponded to an accumulated motion of 0.04 arcsec to 0.3 arcsec. These are significant distances compared to the MUSE pixel size of 0.2 arcsec, so these stars significantly degrade the image fitting procedure if they aren’t excluded.

  2. Another problem is that the HST WFC suffered from a problem called CTE charge loss. Although this affected all sources in an image, it is was much worse for faint unresolved point sources against a dark background. In [2], Reiss (2003) determined that for faint stars against a background, as much of 10% of the flux of the source could be lost. This was before the deeper UDF observations, and for the UDF, with its darker backgrounds and sensitivity to fainter stars, the effect is worse. The photometric image-fitting technique reveals that the flux of stars in the HST UDF images typically need to be scaled by 10% to 20% more than other sources in the same images, to match the MUSE image. If stars are left in the fit, then this results in a poor fit, unless the fit is restricted to a small area centered on a single star.

QSOs should also be excluded

Bright QSOs (Quasi-Stellar-Objects) are point-sources that look similar to stars. Unlike most stars, they don’t have measurable proper motions, but their fluxes are often significantly variable on both short and long timescales. This, along with the CTE effect which was described above for stars, results in poor fits when any bright QSOs are included during the photometric image-fitting procedure.

If the photometric image fitting procedure is limited to a small region centered on a QSO, then this should yield a good fit, which can be used to estimate the pointing errors in the image. However the flux scale that it reports should not be used to scale the rest of the image. Beware that an estimate of the pointing error that is obtained from a single QSO, will be affected by any localized geometric distortions at the position of the QSO in the image. For this reason, it may be better to use a region file to mask out the QSO and perform a global image fit on the other sources in the image. This should yield an estimate of the pointing error that averages out localized distortions.

Exclude variable sources

For the reasons already discussed in the above sections, any sources with varying fluxes or positions should be excluded using ds9 region files, both when aligning images before summing, and when fitting for the PSF of an image. Weak variable sources may not have much affect on the fits, but it is easy to mask them out, so this is worth trying. The quickest way to find out if there are any problematic sources of this form, is to perform a photometry image-fit without specifying a region file, and request that the residual image be displayed, using the –display option of the fit_photometry script. Any problematic sources will then be visible in the residual image of the fit.