Abstract Ice core data are combined with Regional Atmospheric Climate Model version 2 (RACMO2) output (1958–2010) to develop a reconstruction of Greenland ice sheet net snow accumulation rate, Ât(G), spanning the years 1600–2009. Regression parameters from regional climate model (RCM) output regressed on 86 ice cores are used with available cores in a given year resulting in the reconstructed values. Each core site’s residual variance is used to inversely weight the cores’ respective contributions. The interannual amplitude of the reconstructed accumulation rate is damped by the regressions and is thus calibrated to match that of the RCM data. Uncertainty and significance of changes is measured using statistical models. A 12% or 86 Gt yr−1 increase in ice sheet accumulation rate is found from the end of the Little Ice Age in ~1840 to the last decade of the reconstruction. This 1840–1996 trend is 30% higher than that of 1600–2009, suggesting an accelerating accumulation rate. The correlation of Ât(G) with the average surface air temperature in the Northern Hemisphere (SATNHt) remains positive through time, while the correlation of Ât(G) with local near-surface air temperatures or North Atlantic sea surface temperatures is inconsistent, suggesting a hemispheric-scale climate connection. An annual sensitivity of Ât(G) to SATNHt of 6.8% K−1 or 51 Gt K−1 is found. The reconstuction, Ât(G), correlates consistently highly with the North Atlantic Oscillation index. However, at the 11-yr time scale, the sign of this correlation flips four times in the 1870–2005 period.