The current methods available to estimate gravitational shear from astronomical images of galaxies introduce systematic errors which can affect the accuracy of weak lensing cosmological constraints. We study the impact of KSB shape measurement bias o
n the cosmological interpretation of tomographic two-point weak lensing shear statistics. We use a set of realistic image simulations produced by the STEP collaboration to derive shape measurement bias as a function of redshift. We define biased two-point weak lensing statistics and perform a likelihood analysis for two fiducial surveys. We present a derivation of the covariance matrix for tomography in real space and a fitting formula to calibrate it for non-Gaussianity. We find the biased aperture mass dispersion is reduced by ~20% at redshift ~1, and has a shallower scaling with redshift. This effect, if ignored in data analyses, biases sigma_8 and w_0 estimates by a few percent. The power of tomography is significantly reduced when marginalising over a range of realistic shape measurement biases. For a CFHTLS-Wide-like survey, [Omega_m, sigma_8] confidence regions are degraded by a factor of 2, whereas for a KIDS-like survey the factor is 3.5. Our results are strictly valid only for KSB methods but they demonstrate the need to marginalise over a redshift-dependent shape measurement bias in all future cosmological analyses.
