7/10/03

A Plan for Weak Lensing Simulation Software


In the next ~1 year SNAP's need for weak lensing simulations will be to analyze the effect of hardware and survey design changes on the cosmological parameter estimation accuracy from WL measurements. Since the flow of information in studies like this is fairly linear and repeatable, it will probably be more rapid and efficient for us to construct a series of stand-alone programs, which exchange information via ASCII files, then it would be to create a suite of routines that work within the SNAP simulation framework that is currently being planned. Following is a proposal for what each element of this stand-alone chain should do, what the input and outputs would be, and who might be willing to write them. Much of this has been done in nearly the desired form, and for each task I have tried to select the name of the person(s) who has already done the most related work.

"UM" means Tim McKay, Kevin Janka, and Dave Gerdes at University of Michigan.

A good goal would be to have this just about working before the November SNAP collaboration meeting. I don't see any individual pieces that are so far from completion that this is impossible. The next step would be for all involved to comment on this breakdown of algorithms and personnel, and finalize the nature of the data to be exchanged between each unit. Then we can proceed to produce the programs for each STEP (in language of your choice for now). Next year we can worry about integrating this into SNAPSim framework - translating the algorithms into Java should be much less work than figuring out how to do it the first time.

-------------------------------------------------
STEP 1: Generate galaxy distribution.

Input: Redshift and/or flux limit of planned survey, number of square degrees to simulate.

Output: Parent galaxy catalog. Author: Huan Lin

Description: The job here is to come up with some distribution N(z,M*,r,SED) for the galaxies that will be visible in the wide-field survey. This distribution is of course not yet known, but it should be possible to construct one that reproduces the known projections of this distribution, such as:
Since our goal is to see how many galaxies dN/(dAdz) have measurable size and redshift, getting the size right will be important.

-------------------------------------------------
STEP 2: Calculate galaxy apparent magnitudes & sizes

Input: parent galaxy catalog
          SNAP filter descriptions

Output: Parent galaxy catalog augmented with total AB mag in each SNAP filter angular size

Author: Huan Lin and/or UM? (Who is UM?)

Description: Take each galaxies' stellar mass and SED, predict rest-frame spectrum, redshift and integrate over specified SNAP filters to get flux at telescope in each band. Take intrinsic size r and use LCDM angular diameter distance to give angular size.

------------------------------------------------
STEP 3: Determine measurement accuracy, realize measurements

Input: parent galaxy catalog
          SNAP instrument specifications (PSF, detector noise, dithering, etc.)

Output: Author: Gary Bernstein

Description: Use analytic S/N estimators to determine flux uncertainty for this galaxy in each band given the observing scenario. Then realize a measurement using the true flux and the uncertainty. Use analytic formula to estimate accuracy of shape determination for this galaxy after perfect PSF correction; take the best single filter shape measurement as the indication of how well shape is measured.

-------------------------------------------------
STEP 4: Realize an image.

Input: parent galaxy catalog
         SNAP instrument specifications (PSF, detector noise, etc.)

Output: FITS image(s)

Author: Richard Massey

Description: For each galaxy in the parent catalog, take its flux in a chosen band (likely 800 nm) and apparent size, then realize a galaxy in shapelet space using the HDF/GOODS/etc. galaxies at similar mag & size as the training set. This implicitly assumes that shape details are independent of redshift, SED, etc., and just depend upon mag/size, but that's all we can do right now.

Future Extensions:
        * More sophisticated models of how shape depends upon wavelength, z, SED for given galaxy.

-------------------------------------------------
STEP 5: Measure shapes on the image

Input: FITS image

Output: Galaxy shapes & uncertainties (?? what exactly here...)

Author: Jason Rhodes

Description: Use existing shape-measurement algorithms (e.g. RRG) to measure the shapes of the galaxies in the images. This will provide a lower limit to shape-measuring ability while the analytic STEP 3 provides an upper limit. Output of this step can be used to tweak STEP 3 to reflect limitiations of real algorithms.

Future Extensions:
         * Implement more complex shape-measurement methodologies.

--------------------------------------------------
STEP 6: Photometric redshift estimates

Input: Observed galaxy catalog

Output: Observed galaxy catalog, augmented with photo-z and uncertainty.

Author: UM t (Who is UM?)

Description: Implement photo-z methods on the mock SNAP filter-band measurements. Various methods are possible here, might be useful to first use same template SEDs that were used to generate galaxy fluxes in STEP 2, but also try a different method (e.g. Connolly-style) as well.

--------------------------------------------------
STEP 7: Shear measurement error

Input: Observed galaxy catalog

Output: Shear variance function: (inverse) variance of shear per square degree per dz, as function of z.

Author: Bernstein/Rhodes

Description: Take the analytic (STEP 3) or measured (STEP 5) shear accuracty estimation (model for shape noise might be required for former) and determine the accuracy of shear estimation per square degree for galaxies with sufficiently accurate photo-z's. This should be done as a function of galaxy z. ---------------------------------------------------
STEP 8: Cosmological constraints

Input: Shear variance function
Survey size & shape.

Output: Cosmological-parameter Fisher matrices

Author: Some combination of Jain, Stebbins, Refregier, Takada

Description: From the shear variance function vs z and the size of the survey it should be possible to calculate the cosmological parameter accuracy for any chosen method of lensing analysis. There are many possible lensing statistics to choose: