Year-to-year variations in summer drawdown of Northern Hemisphere atmospheric carbon dioxide (CO2) are compared with corresponding year-to-year variations in sea-level pressure (SLP), surface air temperature, and the productivity of land vegetation as inferred from the satellite-derived normalized difference vegetation index (NDVI). Annual values of CO2 drawdown for the years 1980--2000 are estimated from smoothed time series derived directly from individual flask samples at the nine Northern Hemisphere monitoring stations with the most continuous records. The leading principal component of the nine standardized drawdown time series, in which all stations exhibit positive loadings, is used to represent the hemispheric signal in the CO2 drawdown. Linear regression analysis is used to infer the spatial patterns of anomalies in sea-level pressure, surface air temperature, and the NDVI observed during various seasons of years in which the drawdown is anomalously strong. Winters preceding anomalously high drawdown seasons exhibit patterns characteristic of the high index of the Northern Annular Mode (NAM). SLP tends to be anomalously low over the Arctic and high over midlatitudes, and Eurasia tends to be anomalously warm. The pattern of the NDVI observed during the early months of the growing season in years with anomalously high drawdown is indicative of high productivity over Eurasia. These results support the notion that the wintertime NAM influences the annual drawdown of CO2 by modulating winter temperatures that, in turn, affect the productivity of the terrestrial biosphere during the subsequent growing season.