Day 1129

Thursday.

1607.08679
Accurate covariance estimation of galaxy-galaxy weak lensing: limitations of jackknife covariance
Shirasaki, Takada, MIyatake, Takahashi, Hamana, Nishimichi, Murata

Develop a method to simulate gg WL by utilizing all-sky, light-cone simulations.  Populate a real catalog of source galaxies into a light cone simulation realization, simulate the lensing effect on each galaxy, and then identify lensing halos that are considered to host galaxies or clusters of interest.  Use the mock catalog to study the error covariance matrix of galaxy-galaxy weak lensing and find that the super-sample covariance (SSC), which arises from density fluctuations with length scales of comparable with or greater than a size of survey area, gives a dominant source of the sample variance.  Then compare the full covariance with the JK covariance, the method that estimates the covariance from the resamples of the data itself.  Show that, although the JK method gives an unbiased estimator of the covariance in the shot noise or Gaussian regime, it always over-esimtates the true covariance in the sample variance regime, because the JK covariance turns out to be affected by SSC of JK subregion area that has a greater power than that of the total survey area.  Apply the method to SDSS data; the source galaxy catalog and the lensing samples of the redMaPPer clusters and the LRGs.  First show that the 48 mock SDSS catalogs nicely reproduce the signals and the JK covariance measured from the real data, and then argue that the use of the accurate covariance, compared to the JK covariance, can yield an improvement in the cumulative signal-to-noise ratio of the 2-halo term measurement by a factor of 2.4--3.2, which is equivalent to the survey area of about 20,000-30,000 sq. degrees.  Discuss that such large-separation WL signals with the accurate covariance allows improvements in constraints on cosmo parameters such as neutrino mass.