Agenda And Minutes for 3/31 10am CDT

Note that we're now in Daylight Saving Time in the USA!

Call-in Info

Call-in instructions:

1. Dial Toll-Free Number: 866-740-1260 (U.S. & Canada)
2. International participants dial:
Toll Number: 303-248-0285
Or International Toll-Free Number:
3. Enter your 7-digit access code, followed by “#”

Your access code: 7691719


  1. Photo-z Updates
  2. Multi-epoch, multi-band detection:
  3. Mask Updates
  4. Completeness...
  5. Plans?

Attending: Eli, Diego, Aurelien, Huan, William, Yuanyuan, Alex D-W, Don, Felipe, Doug T., and more!

1. A quick update of the photo-zs from Huan. More tests have been done with additional spectroscopic sets and things look "mostly ok" for biases, with some sub-populations of galaxies underrepresented in the training having some problems. Tests continue, but there are no dealbreakers so far...

2. We discussed Dustin Lang's multi-epoch and multi-band detection talk from the Stripe 82 meeting. (Talks are all available if anybody is interested at ). Most of this, especially the multi-epoch detection, is food for thought for the future. However, the question of how to weight the bands in the detection image is more relevant in the near term. Right now, by doing a coadd of r+i+z for the detection image, we are emphasizing the detection of objects that have flat (blue) colors: r-i = 0; i-z = 0. Detection S/N can be increased by 50% by weighting the bands appropriately for a red object. This is particularly relevant for high redshift cluster finding. However, implementing these ideas may require multiple detection runs rather than a "one-size-fits-all" catalog. In the meantime, perhaps we should be weighting the images differently so that we can enhance objects of typical colors (none of our target objects, clusters or high z sources, are as blue as r-i=0; i-z=0).

3. Anne briefly discussed the new masks (that Eli will be releasing soon after a bit more tidying). The residuals in the mask model have correlations for auto mags that are low enough that they are below the statistical power of our 5 year survey. (yay!) However, it looks like the detmodel magnitudes are not modeled as accurately. This will have to be figured out.

4. More discussion of the completeness question. We will poke Eric Huff to see what timescale Balrog tests can be run. We may have to bite the bullet and run SN stacks ourselves to see what comes out. There was some concern from the photo-z group via Huan that processing products that are not coming from DESDM may not be as trustworthy (or have the "stamp" of approval). On the other hand, DESDM have their hands full. In keeping with the original mandate of the SVA1 group, we will push forward outside of DESDM for these completeness tests in the short term for SVA1, but ensure that what we develop gets rolled into the standard processing going forward.