Issue HLA-1590 was created on JIRA by Jennifer Mack:
Add logic to improve alignment logic in the MVMs?
Depends on whether the goal is sky coverage vs. alignment quality.
Goal: Run a regression test to look at the impact of requiring a common WCS solution.
{}-----{}Discussion from slack on Feb 23, 2026 ---------
Current: The MVM combines all SVMs in the same skycell/detector/filter regardless of WCS. Some are aligned to catalogs and some have only ‘a priori’ solutions. Drizzling these together can result in a smeared MVM layer.
Proposed: MVM Layers should be derived from SVMs with a common WCSNAME
__
Option 1: Prioritize Absolute Astrometry: Layers must be eDR3 or GSC242; Drawback: Excluding visits may introduce gaps across the FoV in some layers... not ideal if the goal if sky coverage. Workaround: Add an alignment step to bootstrap the WCS across the skycell/layer, if possible.
__
Option 2: Prioritize Relative Astrometry: Layers can have their own unique WCS, as long as it is consistent. Drawback: Resetting to ‘a priori’ solution for some filters may mean that the narrowband and broadband layers in a skycell are shifted. Workaround: Add an alignment step across layers.
______________
mburger:
How significant a problem has this been? It might be worth looking into how much of an improvement this would be.
______________
mack:
I imagine the results are mixed: good for star clusters and less so for deep fields, good for wide/medium and less so for UV/narrow filters . We would need to run a small regression test to check the impact. Sounds like Steve may have one already for his GS-failure checks. We could add some targets, specifically in fields like M83 with many SVMs being combined into layers.
We could even consider cross-instrument alignment at that point, e.g. FIT-MVM-catalog so that there is a common absolute WCS used for all layers in the sky-cell?
______________
sgoldman:
I'm working a little bit on trying to do a better job of only pulling in "good" images (#2101), but in theory either of these should improve the quality of the MVMs.
See HLA-960.
Issue HLA-1590 was created on JIRA by Jennifer Mack:
Add logic to improve alignment logic in the MVMs?
Depends on whether the goal is sky coverage vs. alignment quality.
Goal: Run a regression test to look at the impact of requiring a common WCS solution.
{}-----{}Discussion from slack on Feb 23, 2026---------Current: The MVM combines all SVMs in the same skycell/detector/filter regardless of WCS. Some are aligned to catalogs and some have only ‘a priori’ solutions. Drizzling these together can result in a smeared MVM layer.
Proposed: MVM Layers should be derived from SVMs with a common WCSNAME
__
Option 1: Prioritize Absolute Astrometry: Layers must be eDR3 or GSC242; Drawback: Excluding visits may introduce gaps across the FoV in some layers... not ideal if the goal if sky coverage. Workaround: Add an alignment step to bootstrap the WCS across the skycell/layer, if possible.
__
Option 2: Prioritize Relative Astrometry: Layers can have their own unique WCS, as long as it is consistent. Drawback: Resetting to ‘a priori’ solution for some filters may mean that the narrowband and broadband layers in a skycell are shifted. Workaround: Add an alignment step across layers.
______________
mburger:
How significant a problem has this been? It might be worth looking into how much of an improvement this would be.
______________
mack:
I imagine the results are mixed: good for star clusters and less so for deep fields, good for wide/medium and less so for UV/narrow filters . We would need to run a small regression test to check the impact. Sounds like Steve may have one already for his GS-failure checks. We could add some targets, specifically in fields like M83 with many SVMs being combined into layers.
We could even consider cross-instrument alignment at that point, e.g. FIT-MVM-catalog so that there is a common absolute WCS used for all layers in the sky-cell?
______________
sgoldman:
I'm working a little bit on trying to do a better job of only pulling in "good" images (#2101), but in theory either of these should improve the quality of the MVMs.
See HLA-960.