Weak Lensing Peak Finding: Estimators, Filters, and Biases

Large catalogs of shear-selected peaks have recently become a reality. In order to properly interpret the abundance and properties of these peaks, it is necessary to take into account the effects of the clustering of source galaxies, among themselves and with the lens. In addition, the preferred selection of lensed galaxies in a flux- and size-limited sample leads to fluctuations in the apparent source density which correlate with the lensing field (lensing bias). In this paper, we investigate these issues for two different choices of shear estimators which are commonly in use today: globally-normalized and locally-normalized estimators. While in principle equivalent, in practice these estimators respond differently to systematic effects such as lensing bias and cluster member dilution. Furthermore, we find that which estimator is statistically superior depends on the specific shape of the filter employed for peak finding; suboptimal choices of the estimator+filter combination can result in a suppression of the number of high peaks by orders of magnitude. Lensing bias generally acts to increase the signal-to-noise \nu of shear peaks; for high peaks the boost can be as large as \Delta \nu ~ 1-2. Due to the steepness of the peak abundance function, these boosts can result in a significant increase in the abundance of shear peaks. A companion paper (Rozo et al., 2010) investigates these same issues within the context of stacked weak lensing mass estimates.