Measurements on blended objects as a topic suggestion for 2016 Project & Community Workshop

Archive LSST2016
#1

Last September (2015), @jbosch started a discussion in the LSST Community DM forum on Baseline Plan for Blended Fitting, which led to a very useful draft document, Jim Bosch’s Blended Measurement Paper, and followup DM discussion in GitHub.

Since almost(?) every LSST science collaboration will need measurements made on blended objects, it would be useful to the community if some plenary time at the 2016 Project and Community Workshop could be devoted to a project update on the current baseline plan for measurements on blended objects and potential alternative strategies, the major open questions regarding strategy, and any related “research questions” that may be beyond the scope of the project’s current requirements, that science collaboration members could potentially help address. It would also be useful to get a very rough estimate of the time scale on which various algorithms in DM might be ready for testing. Even if the current state of the DM code is not particularly relevant for tackling open “research questions” related to measurements on blended objects, there may be specific questions science collaboration members could address outside the DM stack. Also, are there missing requirements that need to be quantified for blended objects — for example, for a “bake off” between different algorithms — that specific science collaborations should be addressing?

Perhaps some of these questions could be refined or addressed in breakouts that focus on different aspects of “blended measurements” — photometry, shapes, astrometry, … — and/or different types of fields (high Galactic latitude vs. bulge), for which different algorithms may be optimal.

3 Likes
#2

I’d be happy to contribute to such a session. DM’s timescale for planning should let us provide a better estimate by then of when code will be available for testing, and I’m pretty confident we can brainstorm some tests that would be useful to us that don’t require that code to be available.

I don’t expect that we’ll add any qualitatively new algorithmic options beyond what’s present in my blended measurement paper, but I do think we’ll likely have a bit more internal agreement about how we rank those options (or at least an agreement on what order we’ll try them out to determine a ranking).

#3

This sounds like a good topic and I would also be happy to contribute. Javier Sanchez and I should have a draft paper quantifying the expected levels of galaxy-galaxy and galaxy-star blending in LSST circulated before the meeting.

#4

I agree this is a great topic and would also be glad to contribute some results from forward model fitting of blends.

#5

I agree this is an interesting topic and would certainly want to attend. I have done a bit of work on storing truth information in simulations for deblending studies.

#6

@burchat thank for this! Would you be willing to lead a breakout session on blended objects? If so, could you tweak the nice description that you wrote above into a brief abstract that describes what the session would look like, who the audience would be, and who the “must attend” folks from the Project might be?

#7

@bethwillman I’d be happy to help put together a breakout session on blended objects. I’ll draft a brief abstract with the details you requested and also seek input from a few others who have responded to this breakout suggestion.

#8

Great, Pat. Thanks! Looking forward to it.

#9

@bethwillman and @jbosch –
Here’s a draft abstract, target audience, and “must attend list” for the proposed blending breakout. Feel free to edit. Please schedule this breakout after the Community Workshop Plenary called “DM in-depth report for the Community”. (It would be great if that DM plenary were scheduled on Wednesday rather than Thursday, since I suspect it will provide a baseline for a number of Community breakouts.)
…………………………….

Abstract:
Most LSST science collaborations will use measurements made on blended objects. This breakout will build on a Data Management update on the current baseline plan and algorithmic options, and a rough estimate of when DM code will be available for testing. In the breakout, we will review and/or identify

  • the major open questions regarding algorithmic options for measurements on blended objects;
  • any related “research questions” that may be beyond the scope of the project’s current requirements, that science collaboration members could potentially help address — with or without DM code;
  • any missing requirements that need to be quantified for measurements made on blended objects — for example, for a “bake off” between different algorithms — that specific science collaborations should be addressing.
    Contributions that describe on-going studies of the impact of blending on particular measurements (photometry, shapes, astrometry) and/or for different types of fields (high Galactic latitude vs. bulge) are welcome.

Resources:

Audience:

  • All science collaboration members interested in measurements made on blended objects.
  • Project folks working on the DM science pipelines.

“Must attend” folks from the Project:

  • Jim Bosch
  • Robert Lupton

“Very nice to have” folks from the Project:

  • Mario Juric
  • KT Lim
  • David Nidever
#10

I doubt the DM plenary will go into sufficient detail to provide much of a baseline for this; deblending isn’t something we’ve made much progress on in the past year or so, and it’s only one of a gazillion things we’ve been working on plans for (and those plans haven’t firmed up much since the preliminary document I circulated a few meetings ago).

#11

A group of people from the Stars, Milky Way and Local Volume discussed this at the San Diego meeting and we are concerned about Crowded Field Photometry especially at low galactic latitude. This is in many ways distinct from blending of 2 sources (see the baseline plan discussion) and I suggest it should be an agenda item.

Issues:

  1. What are the project’s plans for crowded field photometry? We want this to be on the radar.

  2. Is photometry of crowded fields an LSST Project output or does the community take it on [Level 3]? If the latter, what resources will be available?

  3. How can collaboration members with DECam experience at low galactic latitudes be involved?

#12

Pending approval from my institute, I’m planning to attend the community workshop half (Weds-Fri?) of the meeting. @hcferguson and I have been working on identifying a training set of blended galaxy images from CANDELS data with an eye to evaluating different deblending strategies. We’d both be interested in attending a session on this topic.