Interannual variation of carbon fluxes can be attributed to a number of biotic and abiotic controls that operate at different spatial and temporal scales. Type and frequency of disturbance, forest dynamics, and climate regimes are important sources of variability. Assessing the variability of carbon fluxes from these specific sources can enhance the interpretation of past and current observations. Being able to separate the variability caused by forest dynamics from that induced by climate will also give us the ability to determine if the current observed carbon fluxes are within an expected range or whether the ecosystem is undergoing unexpected change. Sources of interannual variation in ecosystem carbon fluxes from three evergreen ecosystems, a tropical, a temperate coniferous, and a boreal forest, were explored using the simulation model STANDCARB. We identified key processes that introduced variation in annual fluxes, but their relative importance differed among the ecosystems studied. In the tropical site, intrinsic forest dynamics contributed approximately 30% of the total variation in annual carbon fluxes. In the temperate and boreal sites, where many forest processes occur over longer temporal scales than those at the tropical site, climate controlled more of the variation among annual fluxes. These results suggest that climate-related variability affects the rates of carbon exchange differently among sites. Simulations in which temperature, precipitation, and radiation varied from year to year (based on historical records of climate variation) had less net carbon stores than simulations in which these variables were held constant (based on historical records of monthly average climate), a result caused by the functional relationship between temperature and respiration. This suggests that, under a more variable temperature regime, large respiratory pulses may become more frequent and high enough to cause a reduction in ecosystem carbon stores. Our results also show that the variation of annual carbon fluxes poses an important challenge in our ability to determine whether an ecosystem is a source, a sink, or is neutral in regard to CO2 at longer timescales. In simulations where climate change negatively affected ecosystem carbon stores, there was a 20% chance of committing Type II error, even with 20 years of sequential data.