Improved Gold 1.0 Depth Maps

Contact: Eli Rykoff

See Translating Mangle Masks Into Data Units for details on the problem and the original runs.

See SVA1 Gold Catalog v1.0 for download links.

These are improved and run with a Random Forest classifier to correlate the systematics with the measured depth.


A draft of a short paper on the methodology as applied to SDSS DR8 is here .


In addition to the files linked to from the SVA1 Gold Catalog v1.0 page, there are additional small files describing the relationship between t_eff and m_lim (see paper linked above).

Gold 1.0.2 t_eff parametrization for MAG_APER_3
Gold 1.0.2 t_eff parametrization for MAG_APER_4
Gold 1.0.2 t_eff parametrization for MAG_AUTO
Gold 1.0.2 t_eff parametrization for MAG_MODEL
Gold 1.0.2 t_eff parametrization for MAG_DETMODEL

Each of these files is a small structure with 5 elements (grizY). Each element has a two-element FIT array. The header has the pivot point (23.0 in all cases). To compute t_eff, the effective exposure time, from the 10sigma limiting magnitude m_lim:

t_eff = exp(fit0+fit1*(m_lim-23.0))

Translating 10sigma Limits into 5sigma Limits

With m_lim and t_eff, we can solve eqns 1,5,6,7 in the depth draft to translate the 10sigma limits into 5sigma limits:

flim10 = 10.^((m_lim - zp)/(-2.5))
fn = ((flim10^2.*t_eff)/(10.0^2.) - flim10)
flim5 = (n2 + sqrt(n2^2. + 4*t_eff*n2*fn))/(2.*t_eff)

m_lim_5 = zp - 2.5*log10(flim5)

Some Pictures

The reconstructed depth map for all of SVA1, auto, i-band, along with the residuals:

And a zoom in on SPTE:

And SPTE for r (auto):

And i (detmodel):

Note that the residuals are definitely worse in the corner with RA~85, Dec ~-57. However, as shown in the paper above some of this may be caused by rapid depth changes and the relatively coarse pixelization used to measure and plot the depth. I have not yet had a chance to run the full analysis using fake data which will be able to quantify these effects.