Breakout session for End-to-End GalSim simulations for LSST

Several people ( @danielsf, @jmeyers314 etc) have been talking or working with me @rmjarvis, @rmandelb, @connolly, @roodman and others on LSST related modules for GalSim.

We could use a breakout session to further discuss progress on this front, and continue work towards getting an end-to-end focal plane LSST simulation (from reading input catalog files to producing output chip FITS files) functioning.

1 Like

FYI: “GalSim is a modular, open source package for simulating images of stars and galaxies using a range of mechanisms that has been developed to support the analysis of weak lensing.” I paraphrase a suggestion from Knut Olsen that this topic might usefully be broadened to encompass what Phil Marshall (I think) calls First Byte - alluding to the use of simulated data. Our thought is to increase the use of end-to-end simulation products by the community. In the galactic and stellar areas, I’ve not heard much about this since the program to incorporate many variable types seemed to go silent. I’d like to hear from the community what inputs and outputs from simulation would be useful now, what are essential before the observing cadence is firmed up, and what before the survey data begins to flow.

Hey @cwalter, thanks for this suggestion! Could you draft a brief abstract that describe in just a bit more detail what you have in mind for such a breakout (considering @sridgway’s suggestion as well), including your expected audience? Can post here or send to me.

Dear Beth and all,

Sorry for the delay, I have been traveling.

The purpose of this proposed session is to continue work and organize future effort towards making GalSim (https://github.com/GalSim-developers/GalSim) a end-to-end solution for doing science studies for LSST. Currently, GalSim is often used to produce and study postage stamp images. In this context, “end-to-end” envisions a set of LSST relevant modules or codes in GalSim that allows us to:

  • Read instance catalogs or directly query the CatSim databases based on OpSim or manual pointings.
  • Propagate sources through a realistically modeled turbulent atmosphere.
  • Further perturb the wavefront as it passes through aberrated LSST optics.
  • Propagate individual photons (with wavelength information) into a physics based LSST specific sensor model.
  • Write out resulting full focal plane images with the proper LSST camera and chip geometry including the readout chain.

By building on GalSim, we can have a very open toolkit that can be contributed to by all and used for a wide variety of science studies. The resulting version of GalSim will be complementary to PhoSim. Some effects which are emergent in PhoSim will be controllable by command, which can be particularly useful.

Work for most of the pieces listed above have already begun, and some even have relatively complete versions. In some cases, fast interpolations for normal running based on the output of the full calculation will need to be developed. In this session, we will take stock of the current state of the work and plan out further development. Making all of the pieces work together in an “end-to-end” manner should be a particular focus. It will also be a good opportunity for new people who wish to contribute to the effort to become involved.

Developers and people who are currently working on the above topics and should attend the meeting include @danielsf, @jmeyers314, @rmjarvis, @rmandelb, @connolly, @roodman. All others are also welcome.

Re: @sridgway’s suggestion: “I’d like to hear from the community what inputs and outputs from simulation would be useful now, what are essential before the observing cadence is firmed up, and what before the survey data begins to flow.”. I think this would be a very interesting to hear about. Personally, I am more attuned to the DESC simulation needs and know less what the other science collaborations need. However, I think this would be best served in another breakout, as this proposed session is focused on the development of tools themselves.