We used a process-based ecosystem model (Marine Biological Laboratory General Ecosystem Model (MBL-GEM III)) to predict and analyze biogeochemical responses of Arctic tundra ecosystems to past (1921--2000) and future (2001--2100) changes in climate and atmospheric CO2 in the Kuparuk River Basin, Alaska. We first calibrated the model by deriving a single parameter set that closely simulated the response of moist tussock tundra to decade-long experimental manipulations of nutrients, temperature, light, and atmospheric CO2 at Toolik Lake on the North Slope of Alaska. We then applied the parameterized model to the entire Kuparuk River Basin over 180 years. The model predicted that warming and drying resulted in a short-term source of CO2 on annual timescales but resulted in a CO2 sink on decadal timescales. These predictions are consistent with recent measurements. A time series analysis has identified that while the immediate response to warming is to release C, the response a year later is to store C. This 1-year lag is consistent with other work that has shown a similar lag in C storage and normalized difference vegetation index (NDVI) on a global scale. Our simulation results indicated that by 2100 high CO2 and warming will increase C sequestration, mostly as a result of (1) an increase in vegetation C:N ratio, which occurs across the Kuparuk Basin, and (2) a redistribution of N from soils (with low C:N ratios) to vegetation (with high C:N ratios), which occurs mainly in ecosystems in the basin that are initially productive, dry, and warm. These results are consistent with the observation of increased shrubiness in Alaskan tundra over the past few decades. Our application of the model has been hindered by the lack of climate data for the region, especially precipitation. A number of other general issues have been identified for making progress in modeling spatial and temporal C dynamics of Arctic tundra.