Predictability analysis based on the maximum Lyapunov exponent considers infinitesimal perturbations, which are associated with errors in the smallest fastest-evolving scales of motion. However, these errors become irrelevant for the predictability of larger scale motions. In this study we employ the newly developed Finite Size Lyapunov Exponent (FSLE) analysis to assess predictability of atmospheric boundary layer flows as a function of scale. We demonstrate the expected enhanced predictability at large scales and quantify the dependence of predictability on the stability of the atmospheric environment.