Abstract　 　　
　Sampling the cosmic shear field through observation is inevitably biased, since cosmic shear can only be measured where there are galaxies and the galaxy distribution is correlated with the lensing signal. This source-lens clustering (SLC) effect has two sources, intrinsic source clustering and cosmic magnification (magnification/size bias). The former decreases with decreasing width of the source galaxy redshift distribution. However, this reduction is limited. Hence it remains unclear whether the residual can be controlled below the statistical error of stage IV lensing surveys. Furthermore, SLC induced by cosmic magnification can not be reduced by lensing tomography. Through N-body simulations, we quantify the impact of SLC on lensing E/B-mode power spectrum in the context of lensing tomography. We consider two favorite estimators in measuring the lensing power spectrum, standard estimator and pixel-based estimator. We find that none of the two estimators can satisfactorily handle both sources of SLC. For standard estimator, SLC induced by both the two sources can bias the lensing power spectrum by O(1%)-O(10%) and the exact value depends on the galaxy bias and the flux/size of source galaxies. SLC induced by intrinsic source clustering also increases statistical uncertainties in the measured lensing power spectrum. However, standard estimator suppresses intrinsic source clustering induced SLC in cross power spectrum. In contrast, pixel-based estimator is efficient to suppress SLC induced by cosmic magnification. However, it fails to suppress SLC induced by intrinsic source clustering and the measured lensing power spectrum can be biased low by O(1%)-O(10%). We conclude that SLC effect is a severe systematic error for weak lensing cosmology, even with the aid of lensing tomography. We present useful scaling relations to self-calibrate SLC effect.