Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: jointcal
-
Labels:None
-
Story Points:8
-
Epic Link:
-
Sprint:AP S20-5 (April), AP S20-6 (May), AP F20-1 (June), AP F20-2 (July)
-
Team:Alert Production
-
Urgent?:No
Description
There's existing code in the stack to apply proper motions and parallaxes from the reference catalog. Find it, dust it off, and get it going in Jointcal. It may require some modification. If it requires substantial overhauls (e.g. rewriting completely because it's too slow), let's make a new ticket to capture that.
For now, we will correct the reference catalog to the mean epoch of the loaded data.
Attachments
Issue Links
- is blocked by
-
- Done
- is triggering
-
- Done
-
- Done
-
- Won't Fix
- relates to
-
- Done
-
DM-25316 Apply proper motion in AstrometryTask when loading refcats
- Invalid
Activity
Just as a heads up: since these features weren't ready when we were designing DC2, they were turned off there. They were actually on in DC1 which is how we found out they weren't supported 
.
So, probably good to have application be a config option turned on by default but possible to turn off.
Chris Walter
added a comment - Just as a heads up: since these features weren't ready when we were designing DC2, they were turned off there. They were actually on in DC1 which is how we found out they weren't supported .
So, probably good to have application be a config option turned on by default but possible to turn off. Paul Price: are you able to do this medium sized review?
As implemented, the code uses the mean of the dates in the data, which might not be ideal but I think the bigger step will be DM-24278 which will shift all of the internal positions by known PM values. That should make the question of mean/median/mode/whatever mostly irrelevant.
Also note that we do not have code to compute the parallax corrections in the stack right now: the refcat loader only does proper motion corrections. My meas_algorithms branch cleans up some docstrings related to that. I don't know if implementing refcat parallax corrections in meas_algorithms should be part of this ticket, or on a separate ticket.
Code looks good.
Can you confirm that proper motion correction can be disabled via config parameters, and that we fail over smoothly if there are no proper motion columns in the catalog?
Paul Price
added a comment - Code looks good.
Can you confirm that proper motion correction can be disabled via config parameters, and that we fail over smoothly if there are no proper motion columns in the catalog? Question answered on github: proper motions are now always applied if available and the default is to fail with an error if they are not. That can be changed by setting astrometryRefObjLoader.requireProperMotion=False, which will allow astrometry to run without PM corrections.
I ran a tract of HSC RC2 successfully with this new code, so it looks like it's fine for HSC. However, doing so led me to discover that the logic for requireProperMotion in meas_algorithms is not robust: if the refcat has a pm_ra field but it is not an Angle, then you'll just receive a warning instead of a failure if that config is True. I've filed DM-25807 about this issue.
Question arises: is the mean actually the best date to use? Is there another date which will minimize the parallax-contributed error?
On this ticket, regard that as out of scope (but feel free to think about it); but make it possible to parameterize what date is used.