Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-30490

Add delta-magnitude check to AstrometryTask

    XMLWordPrintable

    Details

    • Type: Story
    • Status: Done
    • Resolution: Done
    • Fix Version/s: None
    • Component/s: meas_astrom
    • Labels:
      None

      Description

      Recent investigations in DM-30466 shows that one of the (more common?) failure modes in the astrometry fitting is erroneous matching between source and reference objects of incompatible magnitudes.

      As discussed on that ticket, one of the challenges of a magnitude-based selection is that we don't have a zeropoint yet. However, we do have a match between source and reference objects. Therefore, it should be possible to compute an effective zeropoint. My suggestion is to compute -2.5*log10(src_flux) + 2.5*log10(ref_flux) and perform a configurable n-sigma cut about the mean/median when doing outlier rejection.

      This will obviously require some significant testing on RC2 and DC2 data to ensure that nothing breaks.

        Attachments

          Issue Links

            Activity

            Hide
            erykoff Eli Rykoff added a comment -

            I have added the option to doMagnitudeOutlierRejection to the AstrometryTask in the part of the iteration that happens after the matcher does the pattern matching, and prior to the fitter computing the wcs. Therefore, this is applicable to all matchers, not just PessimisticB.

            I have tested a random subset of Johann Cohen-Tanugi's list of failures from https://jira.lsstcorp.org/secure/attachment/49907/tract_patch_with_failed_astrometry.txt . In each case, I ran first with the latest weekly (w_2021_23) default settings, and then with doMagnitudeOutlierRejection=True. This way I could check that detectors that were failures previously in the DC2 runs are still failing. I ran 10 previously failed detectors. One of them worked fine with w_2021_23 (and still works fine with my new branch). One of them, Lauren MacArthur's "hot mess" from DM-30466, (1131081/150) which was chosen as the worst of the bunch, is not fixable with this branch (without other configuration changes that Lauren MacArthur has proposed) because the initial pattern matching is so far gone. The other 8 are still failures with the latest weekly but succeed with doMagnitudeOutlierRejection=True.

            I have also asked that James Chiang look at some prior failures (probably after it goes into the weekly) for determining if this should be turned on for DP0.2 processing.

            The summary of the final reported scatter values, as well as the number of stars matched in the photoCal task are as follows:

            Visit/Detector w_2021_23 scatter (arcsec) w_2021_23 nstar doMagnitudeOutlierRejection scatter (arcsec) doMagnitudeOutlierRejection nstar
            654437/73 0.008+/-0.005 135 0.008+/-0.005 131
            648548/105 7.517 +/- 5.048 0 0.005 +/- 0.003 117
            517994/40 0.288 +/- 0.179 24 0.005 +/- 0.003 114
            692688/65 5.550 +/- 4.842 3 0.004 +/- 0.002 114
            262006/155 0.165 +/- 0.089 93 0.003 +/- 0.002 111
            202442/58 5.924 +/- 3.496 1 0.006 +/- 0.003 107
            479591/49 0.537 +/- 0.206 8 0.009 +/- 0.005 94
            1197874/176 0.431 +/- 0.243 23 0.004 +/- 0.002 100
            939927/37 0.373 +/- 0.225 23 0.003 +/- 0.003 72
            1131081/150 8.534 +/- 4.786 2 8.639 +/- 4.752 2
            Show
            erykoff Eli Rykoff added a comment - I have added the option to doMagnitudeOutlierRejection to the AstrometryTask in the part of the iteration that happens after the matcher does the pattern matching, and prior to the fitter computing the wcs. Therefore, this is applicable to all matchers, not just PessimisticB. I have tested a random subset of Johann Cohen-Tanugi 's list of failures from https://jira.lsstcorp.org/secure/attachment/49907/tract_patch_with_failed_astrometry.txt . In each case, I ran first with the latest weekly ( w_2021_23 ) default settings, and then with doMagnitudeOutlierRejection=True . This way I could check that detectors that were failures previously in the DC2 runs are still failing. I ran 10 previously failed detectors. One of them worked fine with w_2021_23 (and still works fine with my new branch). One of them, Lauren MacArthur 's "hot mess" from DM-30466 , (1131081/150) which was chosen as the worst of the bunch, is not fixable with this branch (without other configuration changes that Lauren MacArthur has proposed) because the initial pattern matching is so far gone. The other 8 are still failures with the latest weekly but succeed with doMagnitudeOutlierRejection=True . I have also asked that James Chiang look at some prior failures (probably after it goes into the weekly) for determining if this should be turned on for DP0.2 processing. The summary of the final reported scatter values, as well as the number of stars matched in the photoCal task are as follows: Visit/Detector w_2021_23 scatter (arcsec) w_2021_23 nstar doMagnitudeOutlierRejection scatter (arcsec) doMagnitudeOutlierRejection nstar 654437/73 0.008+/-0.005 135 0.008+/-0.005 131 648548/105 7.517 +/- 5.048 0 0.005 +/- 0.003 117 517994/40 0.288 +/- 0.179 24 0.005 +/- 0.003 114 692688/65 5.550 +/- 4.842 3 0.004 +/- 0.002 114 262006/155 0.165 +/- 0.089 93 0.003 +/- 0.002 111 202442/58 5.924 +/- 3.496 1 0.006 +/- 0.003 107 479591/49 0.537 +/- 0.206 8 0.009 +/- 0.005 94 1197874/176 0.431 +/- 0.243 23 0.004 +/- 0.002 100 939927/37 0.373 +/- 0.225 23 0.003 +/- 0.003 72 1131081/150 8.534 +/- 4.786 2 8.639 +/- 4.752 2
            Hide
            Parejkoj John Parejko added a comment -

            Thank you for the updated test. I've left a few more comments, but I think we can easily sort them out.

            This is a useful simple refcat match check, and I'm glad we now have it.

            Show
            Parejkoj John Parejko added a comment - Thank you for the updated test. I've left a few more comments, but I think we can easily sort them out. This is a useful simple refcat match check, and I'm glad we now have it.
            Hide
            erykoff Eli Rykoff added a comment -

            Jenkins is green: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/34463/pipeline/ and I re-ran a test on one of the DC2 failures listed above and it is working properly.

            Show
            erykoff Eli Rykoff added a comment - Jenkins is green: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/34463/pipeline/ and I re-ran a test on one of the DC2 failures listed above and it is working properly.

              People

              Assignee:
              erykoff Eli Rykoff
              Reporter:
              erykoff Eli Rykoff
              Reviewers:
              John Parejko
              Watchers:
              Chris Walter, Eli Rykoff, James Chiang, Jim Bosch, John Parejko, Lauren MacArthur
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved:

                  Jenkins Builds

                  No builds found.