Strange blobby source distribution in some HAP point catalogs

[stscijgbot-hstdp]

Issue HLA-1240 was created on JIRA by Rick White:

There is a fairly rare failure mode where the HAP point catalog source detection goes crazy and generates big blobs with thousands of spurious sources scattered around the image. I don't know what triggers this, but my guess is that something has gone wrong with the background estimation. That could lead to this kind of structure.

Not a high priority to fix since it is rare, but if the problem were identified and fixed, it would probably lead to quality improvements in other catalogs that are not so obviously bad.

A sample image is appended below in the attachments and comments. An additional comment lists 16 datasets that are known to have the problem. The problem has only been found in ACS/WFC catalogs. It is likely that at least 125 ACS/WFC visits have the problem (about 0.8% of the ACS/WFC HAP-SVM visits).

[stscijgbot-hstdp]

Comment by Rick White on JIRA:

An example image hst_13057_12_as_wfc_jc3f12 is shown. The image itself looks fine: it is a crowded field with a lot of saturated stars, which may be a clue, but is not bad. The segment catalog looks good. The point catalog is crazy. The image below shows the HAP-SVM image with the point sources overplotted in red. Sources flagged as bad have been removed. The big patches of sources are clearly wrong.

!hst_13057_12_as_wfc_jc3f12-point.png|width=70%!

[stscijgbot-hstdp]

Comment by Rick White on JIRA:

I have identified multiple additional datasets with this problem. So far I have only seen this issue in ACS/WFC datasets. We do not use the other ACS detectors in the Hubble Source Catalog, but it is likely that if the problem occurred for WFC3/UVIS or WFC3/IR, it would have been discovered. So it is probable that this is an ACS-only problem. (No WFPC2 catalogs have been examined.)

I estimate there are at least 125 such datasets in the sample we are using for the HSC. There are a total of about 16,400 ACS/WFC visits in the sample, so approximately 0.8% of the HAP ACS/WFC catalogs probably have the issue.

The table lists 16 ACS visits that are known to have bad, blobby point catalogs. The table includes the number of good sources (according to the flags) in the HAP source lists and a link to the HLA interactive display that overlays the point catalog.

||datasetname||segmentCatNobj||pointCatNobj||display||
| hst_10775_19_acs_wfc_total_j9l919 | 4164 | 20361 | display |
| hst_10775_47_acs_wfc_total_j9l947 | 3453 | 7951 | display |
| hst_12586_17_acs_wfc_total_jbts17 | 2017 | 7378 | display |
| hst_13057_10_acs_wfc_total_jc3f10 | 2406 | 11789 | display |
| hst_13057_12_acs_wfc_total_jc3f12 | 3145 | 15025 | display |
| hst_13057_21_acs_wfc_total_jc3f21 | 2399 | 7257 | display |
| hst_13057_39_acs_wfc_total_jc3f39 | 2495 | 13203 | display |
| hst_13057_a5_acs_wfc_total_jc3fa5 | 2588 | 14094 | display |
| hst_13463_09_acs_wfc_total_jc8x09 | 2093 | 12420 | display |
| hst_13463_11_acs_wfc_total_jc8x11 | 3791 | 26904 | display |
| hst_13463_30_acs_wfc_total_jc8x30 | 1943 | 15342 | display |
| hst_13463_40_acs_wfc_total_jc8x40 | 2687 | 21429 | display |
| hst_13463_42_acs_wfc_total_jc8x42 | 3333 | 15304 | display |
| hst_13463_59_acs_wfc_total_jc8x59 | 2321 | 10742 | display |
| hst_9700_01_acs_wfc_total_j8kr01 | 164 | 321 | display |
| hst_9750_19_acs_wfc_total_j8q619 | 3688 | 13247 | display |