AP Pipeline QA Meeting, 2022-12-05

2pm Pacific Time

Zoom: https://washington.zoom.us/j/98846655020

attending:

• Eric Bellm 
• Ian Sullivan 
• Leanne Guy 
• Andrew Connolly 
• John Parejko 
• Brianna Smart 
• Krzysztof Suberlak 
• Erin Howard 
• Colin Slater 
• Kenneth Herner 
• Yusra AlSayyad
• Mario Juric 
• Zeljko Ivezic 
• Bryce Kalmbach 
• Nima Sedaghat 
• Ari Heinze 
• Jeffrey Carlin 
• Krzysztof Suberlak 

regrets:

Topics for discussion:

• Announcements
  • none
• Precursor processing (AP team):
  • sprint dataset update (Kenneth Herner ):
    • (used to tease out issues of template vs. science seeing, etc.)
    • best and middle-third seeing templates, science w/ seeing between those.  Testing forced convolution in both directions, as well as "auto" mode
    • showing results with w46: several bugfixes, including fix of patch boundary variance plane underestimate (credit to Ian and Yusra)
    • auto-mode appears to be working with both best and middle-third seeing templates
    • bright (SNR>20) sources showing agreement to ~5-10%
  • takeaway is that we seem to have fairly functional difference imaging that's independent of seeing–doing a larger run, but hope to claim DM-AP-12 once that's in
  • new hire Bruno Sanchez will explore tradeoffs of alternative approaches (preconvolution, ZOGY)
• SITCOM team updates (Andrew Connolly, Mario Juric   et al.):
  • not much for updates, but a question:
    • they have a Jenkins CI process (#aos_build) but sometimes weeklies break it–how can they stay updated?
    • Ian: #dm-science-pipelines has a bookmark to the DM changelog–has weeklies + updates since last
    • runs daily, but it sounds like T&S may also be using the DM weekly
  • continuing to integrate John Franklin's ML donut model, and also working on masking overlaps in traditional model fitting to compare
• DIA Transients & Variables Science Unit update (Eric Bellm ):
  • made contacts with the contributing group leads.  Will use the DESC DIA biweekly meeting initially at least.  Will probably invite interested SU members to this meeting in ~February
• DIA Solar System Science Unit update (Mario Juric ):
  • SS group is smaller: just two external contributors (Matt Holman & Siegfried Eggl), plus lots of local folks. 
  • Fornightly meetings Mondays at 10:30 Pacific.  Notes at https://dirac.us/ngc for now
  • Last meeting discussed impact of schedule changes, implications for commissioning SSP pipelines.  Implies heavier reliance on simulations, looking for ways to exercise MPC connection.  Ari suggested that PS1 has never been analyzed with a MOPS-like algorithm for a long time, will investigate whether that's interesting
  • 2023 data exchange test with MPC: will run a "realistic but not automated" test: 2022 test tested scale but sent perfect data.  2023 will include more warts–missed linkages, etc. 
• PreOps V&V (Colin Slater ): 
  • ramping up V&V on some AuxTel processing --so far lots of sensor effects, flat fielding effects, etc.  Hope to get those resolved soon.
  • AuxTel data being run through processCcd in batch mode after each run, excited for prompt processing soon
    • Ian: how much work to create templates?  Colin: no one has tried coadds recently, still an experiment in the USDF and /repo/embargo era.  Yusra: no plans yet.  PSF is really bad right now, waiting for better seeing
• Review CI (https://chronograf-demo.lsst.codes/):
  • new DC2 dataset!  Will let us assess runtime and initial real/bogus scoring
  • template variance bugfix, nothing else notable.

• Next meetings (all at 2pm Pacific):
  
  • (skip January: holiday/meeting-free week, AAS, MLK day...)
  • February 6
• AOB

Action Items