We use twenty-two monthly GRACE (Gravity Recovery and Climate Experiment) gravity fields to recover nonsecular mass change in Greenland. The results show large seasonal variability. We compare with modeled precipitation, evaporation, and runoff derived from ERA40 (the 40-year ECMWF Re-Analysis of the global atmosphere). The model's seasonal amplitude is controlled by runoff and agrees reasonably well with GRACE. Both GRACE and the model show an April/May maximum. But the GRACE results show a delayed minimum relative to the model. This difference is probably associated with omissions in the runoff model, ice discharge, subglacial hydrology, mass loss by blowing-snow, and hydrology in ice-free regions. The discrepancy is smaller, but still significant, for south Greenland alone. When we include a proxy for ice discharge the agreement is improved.