The practical use of astrometry to detect exoplanets via the reflex motion of the parent star depends critically on the elimination of systematic noise floors in imaging systems. In the diffractive pupil technique proposed for space-based detection of exo-earths, extended diffraction spikes generated by a dotted primary mirror are referenced against a widefield grid of background stars to calibrate changing optical distortion and achieve microarcsecond astrometric precision on bright targets (Guyon et al. 2010). We describe applications of this concept to ground-based uncrowded astrometry using a diffractive, monopupil telescope and a wide-field camera to image as many as ∼4000 background reference stars. Final relative astrometric precision is limited by differential tip/tilt jitter caused by high altitude layers of turbulence. A diffractive 3-meter telescope is capable of reaching ∼35 μas relative astrometric error per coordinate perpendicular to the zenith vector in three hours on a bright target star (I < 10) in fields of moderate stellar density (∼40 stars arcmin-2 with I < 23). Smaller diffractive apertures (D < 1 m) can achieve 100-200 μas performance with the same stellar density and exposure time and a large telescope (6.5-10 m) could achieve as low as 10 μas, nearly an order of magnitude better than current space-based facilities. The diffractive pupil enables the use of larger fields of view through calibration of changing optical distortion as well as brighter target stars (V < 6) by preventing star saturation. Permitting the sky to naturally roll to average signals over many thousands of pixels can mitigate the effects of detector imperfections.