LSSTDESC / Twinkles

10 years. 6 filters. 1 tiny patch of sky. Thousands of time-variable cosmological distance probes.
MIT License
13 stars 12 forks source link

Make DIAObjects and DIA[Forced]Sources #399

Open drphilmarshall opened 7 years ago

drphilmarshall commented 7 years ago

As explained in the cookbook recipe written to address #394, we now have the ability to do image differencing and measure DIASources. To make light curves, we need to make DIAObjects and then do forced photometry on them. These "reprocessed Level 1 DIASources," each associated with a "Level 1 DIAObject". For more details, see Section 4.2 of LSE-163, the LSST DPDD.

This is an Epic: DIAForcedSource Light Curves

@SimonKrughoff Let's start by writing the cookbook recipe that we want to see, and use it to layout our plan. This is:

Visit by visit:

  1. Associate the DIASources (which we currently have) into DIAObjects, making new DIAObjects as required;
  2. Updating the DIAObject positions to be their current 12-month averages;

We will skip this next step:

  1. Do ForcedPhotometry at the new DIAObject positions to obtain "precovery" psfFlux measurements for the previous 30 days, and add these "forced source" measurements to the DIASource table, "appropriately flagged." That 30 day window seems like a lot of book-keeping, and we're not going to use the Level 1 light curves for cosmology anyway.

Then, we emulate "periodic reprocessing" to make a DIAObject and DIASource table in a "reprocessed Level 1 database" (because "periodic" does not necessarily mean "annual"...)

  1. Copy the Level 1 DIAObjects to a "reprocessed Level 1" DIAObjects table, preserving the positions and average measurements. This step could be skipped, to keep things simple?
  2. Do forced photometry at these DIAObjects' positions, to make "reprocessed Level 1" DIASources.

To reiterate: we expect the reprocessed Level 1 DIASources to provide our cosmological light curves.

What should we assume for the cadence of Level 1 reprocessing? We can do it annually, but then choose to query any of the DR1, DR2, DR3 etc "Reprocessed Level 1 DIASources" for light curves.

We can write this recipe with code snippets that we want to see exist - and then realize that vaporware.

drphilmarshall commented 7 years ago

Hmm - @SimonKrughoff we got a few bits of terminology wrong in our first attempt at the above description! We should use "reprocessed Level 1" instead of "Level 2" . There are no promised Level 2 (annual data release) DIAForcedSource light curves...

drphilmarshall commented 7 years ago

OK, after a bit more discussion, we're going to punt on the Level 1 forced photometry, and just make Level 1 DIAObjects - and then emulate the reprocessing of Level 1 to make Level 1 DIASources by doing forced photometry at the Level 1 DIAObject positions. Updating the top comment now...

drphilmarshall commented 7 years ago

BTW there is some community discussion of the Level 1 reprocessing frequency here.

drphilmarshall commented 7 years ago

To ease the other book-keeping worry, how about we approximate this 12-month window as extending all teh way back to the beginnig of the survey? This will make our object positions a bit over-precise, but it will allow us to use a far simpler online update to the centroids. As for the DIAObject shape, proper motion etc, I suggest we ignore those for now: in Twinkles 1 we are only really interested in light curves. What do you think? This way we can write the simplest possible tasks, and then they can be extended to include more options later.

SimonKrughoff commented 7 years ago

@drphilmarshall Sounds good to me. I'll get started on the cookbook now.