varsim.md

Simulations of Variables and Transients

Abstract

Introduction

• Large sky surveys in Time domain astronomy are becoming common

• Large amounts of incoming data:

  • require new methods to detect objects
  • more accurate methods to study objects, as precision increases

• Brings to fore survey science and data science methods.

• How can preparation be done without the data? Test analysis methods using:

  • reprocess old datasets: important as it includes real characterstics, but misses the appropriate size necessary to study effects.
  • Simulated datasets: Get the appropriate size for testing precision effects, but the challenge is including appropriate correlations and subtle effects.
  • Important for analysis methods, survey design: eg. survey strategy requirements.

• Simulating Data :

  • The astrophysical sources:
    • Transients/Variables are among some of the rich and complex astrophysical systems, and therefore of interest to frontier research in astrophysics : typically generated by computationally intensive codes involving solutions of partial differential equations describing transport phenomena, dynamics and nucleosynthesis.
    • Simulating transients from data driven models: Fast, low computational requirements, with sufficient data, often better tuned at describing realistic populations of astrophysical transients.
  • Equally important is the survey science aspect:
    • Representing a survey appropriately, including the proper cadence, seeing, depths. Emphasized here as work is supposed to be general.
    • Misses effects present in the image processing pipeline: emulators necessary.
  • Which data product to emulate ?
    • Forced photometry results
    • Forced photometry of detected objects, or all objects? helps in understanding the impact of selection. Over what time scales ? Data Size issues.
    • What information should be recorded?

• Code architecture choices:

  • Parallelizability
  • Reproducibility
  • Library vs script
  • Choice of primitives

• Inputs and Outputs in this work and relationship to other work:

  • Intended to work with LSST, must interact with LSST data products, and software stack
  • Output of survey strategy in OpSim outputs
  • SNCosmo
  • SNANA
  • MAF

Description of calculational setup: Code primitives

Realizer :

• a model that knows the time evolution of the transient. The model is expected to provide the correct time evolution in different bands.
• Such a model may depend on a set of parameters. For a particular set of parameters, it describes a particular class of transients as observed from the top of the earth's atmosphere.
• In order to produce light curves as observed in a survey, this model requires an input of the survey strategy, as well as observational parameters.

Population :

• a model, and a sequence of parameters. The sequence of parameters may be drawn from a stochastic distribution or a deterministic list.
• It should include coordinate positions (ra, dec) for calculation of MW extinction. However, we have a Population class which does not include this.
• There is further support of primitives to build up a population object, in terms of rate distributions, cosmological volume calculations, etc.
• While such a sequence may include all objects of interest in the universe, or the survey, reproducibility demands a definition of a quantum which will never be split in between nodes during simulations.
• A convenient identificiatin of such quanta for variables of roughly uniform densities in different sky locations is over quasi equal spatial areas.
• For objects with different spatial densities depending on the location in the sky (eg. variable stars), a similar method may be chosen, but it may help to identify quanta based on positions.
• Both cases may handled easily through hierarchical tesselations of the sky.

Functioning

• In order to simulate a population, the code

Discussion

• Adding chips functionality might be desirable. Check on Mangle help for this.

Notes :

• Should include discussion of parallelization and reproducibility on parallelizing SN vs tiles.
• Keep all 3 codes together
• APJ Supplement