I think there may be a problem with the way we're doing the astrometric uncertainty propagation.
@dlakaplan pointed out a while ago that there's some sources that are clearly the same as one another but are not correctly associated, e.g. https://dev.pipeline.vast-survey.org/sources/54484320/ and https://dev.pipeline.vast-survey.org/sources/54484847/. Previously there was a 1" fudge-factor inserted into all pipeline runs, but that's been removed for the latest pipeline runs because we correct the astrometry. However, now it seems that the astrometric uncertainties are too small and sources aren't being correctly associated.
To take that source as an example, in SB54901 the RA and Dec error reported by the pipeline is 0.26" and 0.28" respectively. However if you look at the corrections database (/data/VAST/vast-data/TILES/corr_db/) the uncertainty in both corrections is 0.35", so the uncertainty in the pipeline numbers should be at least that large.
Looking at the actual corrected catalogues used for that run (/data/vast-survey/VAST/pipeline-symlinks/galactic/EPOCH47/TILES/STOKESI_SELAVY_PROCESSED/selavy-image.i.VAST_1753-18.SB54901.cont.taylor.0.restored.conv.components.processed.v2.xml), the uncertainty seems to be sensible (0.44" and 0.41") so the problem must be caused by the pipeline.
I think there may be a problem with the way we're doing the astrometric uncertainty propagation.
@dlakaplan pointed out a while ago that there's some sources that are clearly the same as one another but are not correctly associated, e.g. https://dev.pipeline.vast-survey.org/sources/54484320/ and https://dev.pipeline.vast-survey.org/sources/54484847/. Previously there was a 1" fudge-factor inserted into all pipeline runs, but that's been removed for the latest pipeline runs because we correct the astrometry. However, now it seems that the astrometric uncertainties are too small and sources aren't being correctly associated.
To take that source as an example, in SB54901 the RA and Dec error reported by the pipeline is 0.26" and 0.28" respectively. However if you look at the corrections database (/data/VAST/vast-data/TILES/corr_db/) the uncertainty in both corrections is 0.35", so the uncertainty in the pipeline numbers should be at least that large.
Looking at the actual corrected catalogues used for that run (
/data/vast-survey/VAST/pipeline-symlinks/galactic/EPOCH47/TILES/STOKESI_SELAVY_PROCESSED/selavy-image.i.VAST_1753-18.SB54901.cont.taylor.0.restored.conv.components.processed.v2.xml), the uncertainty seems to be sensible (0.44" and 0.41") so the problem must be caused by the pipeline.