WFPC2 Associations QA Test Results

WFPC2 Associations

QA Test Results

Results of Photometry Tests on Combined WFPC2 Images

We have tested a subset of combined WFPC2 images produced by the WFPC2 Associations Products Pipeline (WASPP) for photometric accuracy, as part of a general quality assessment of these products. Our tests consisted of comparing the instrumental magnitudes of objects in these images to those of the "leader" image in the association of images from which they were formed, checking for systematic deviations from linear correlations, and comparing the respective errors. These tests were performed using version 2.2.2 of the SExtractor package of Bertin & Arnouts. The source detection and measurement parameters were set to fairly "standard" values, e.g., 3-sigma thresholds for detection. Varying these parameters, as well as the method of background subtraction (local versus global), was not found to change the basic results of the tests.

We tested images containing a large number of objects with a wide range in brightness, such as images of crowded stellar fields. We also used images of the Chandra Deep Field, in which most of the sources are galaxies. We also chose images obtained in a wide range of filters, including F300W, F555W, F813W, and F656N, and, to the extent possible, covering a range in the number of images in the input associations and in the exposure times of those images.

The bottom-line result of these tests is that we find no systematic offsets from correlation between the combined and leader images that would point to a flaw in the image combination algorithm. The errors returned by SExtractor for the magnitudes measured from the combined images are also lower than those of the leader image by approximately 1/sqrt(n), where n is the number of images in the association, which is also expected if the combination algorithm performed correctly. In this respect, we see no reason not to consider the combined images photometrically accurate and suitable for scientific analysis. We did notice deviations from correlations in the case of saturated stars, but believe we understand the origin of these differences and that they do not signify a flaw in the combination algorithm. Errors in combined images not tested may of course be present and may signify problems in the image combination software; users discovering any such errors are urged to contact one of the CADC, ST-ECF, or MAST scientists involved in the maintenance of the WASPP system.

Below we provide three examples of our test results, which are representative of our findings. No significant differences were noticed between images taken in different filters, or with different numbers of images in the input associations. Similarly, no significant differences were found in the results for the different chips of the detector for a given image (including the PC), or between objects classified as galaxies and those classified as stars. This indicates that the combination algorithm is basically robust.

Example 1: Image u68j0205b (Chandra Deep Field, F814W, 8 input images, Chip 4)

This plot shows the basic agreement between the instrumental magnitudes of the combined image and the leader image, where m is defined as -2.5log(counts/s) for each image (brighter objects thus have more negative values). The relative brightness of some of the fainter objects in the leader image is almost certainly due to cosmic ray contamination, since the combination process removes cosmic rays. None of the galaxies in the images are saturated, thus (in contrast to the next two examples) there is no deviation from correlation above the saturation level.

Example 2: Image u2he0701b (Stellar Cluster, F814W, 9 input images, Chip 3)

This plot shows the good correlation between the instrumental magnitudes of the combined and leader images (modulo cosmic ray contamination) up to the point of object saturation. Comparison of the images, including forming difference and division images, shows that the bleeding columns and diffraction spikes in saturated stars in the combined image are reduced relative to the leader image, leading to a higher count rate in the latter. This is presumably a benefit of the "artificial skepticism" algorithm. Note that all the nine images in the association in this case have equal integration times (1000 s).

Example 3: Image u4woc00cb (Stellar Cluster, F300W, 6 input images, Chip 4)

This plot shows a relatively higher count rate for saturated stars in the combined image. We believe this to be a result of the fact that the combined image has a lower mean count rate than the leader image, since some of the images in the association have lower exposure times than the leader image. Since count rate is the ratio of counts to integration time, once saturation is reached, this ratio is larger for lower integration times. We confirmed this by comparing the instrumental magnitudes for individual images in the associations having different integration times, as the same behavior for saturated sources is observed.

In any event, in all of the above cases agreement is found for non-saturated sources. For saturated sources the above results indicate that the combined images will also provide better magnitude limits due to the apparent rejection of some saturated pixels.

Results of Astrometry Tests on Combined WFPC2 Images
For combined images containing stars listed in the USNO2 catalog, WASPP registers the images to the reference frame of this catalog by computing a mean shift from the image header coordinates to the coordinates of the USNO2 stars, and updating the header coordinates accordingly. We have found this to produce a typical positional accuracy of approximately 0.3 - 0.4 arcseconds relative to the USNO2 reference frame. In images where no USNO2 stars are present, WASPP defaults to the image header coordinate values. We have however identified some images with larger systematic shifts (up to approximately 10 arcseconds) of the image coordinate values relative to the USNO2 reference frame. These frames typically contain large diffuse sources such as galaxies, and the shifts are likely caused by confusion of the USNO2 stars in the images with regions in these objects in the source identification software (SExtractor) applied to the combined images. Users concerned with astrometric accuracy are thus advised to compare the world coordinate values of combined images containing such sources with the positions of the USNO2 stars within them (using image display tools such as SkyCat and SAOImage-DS9), and correct the WCS values in the combined images if necessary. Users discovering any gross astrometric errors in the combined images are asked to contact any of the scientists in the WFPC2 Associations project.