While significant effort has been invested in understanding the centennial-scale behavior of the carbon cycle, carbon management and mitigation strategies require better quantification of annual- to decadal-scale carbon cycle processes and feedbacks. To advance the predictive ability of global carbon cycle models over shorter time scales, a new perspective and deeper understanding of intrinsic properties of terrestrial, marine, and human systems is required. In particular, effects of land use change, vegetation disturbance and recovery, and seasonal climate variability all impact the short-term trajectory of the carbon cycle, and many of those processes are not under human control. Similarly, the ocean carbon cycle depends strongly on upwelling, seasonal or storm-induced mixing, and wind-driven ventilation of the Southern Ocean. Carbon cycle feedbacks from a wide variety of human activities, which depend on economic activity, agricultural demand, and governmental policy, are poorly quantified or understood. This session is designed to identify processes and mechanisms important for improving the predictability of the carbon cycle across the Earth system. Discussed will be challenges in coordinated observation and modeling activities, opportunities for applying "Big Data" approaches for improved model-data integration, and model development needs for future Earth system models. With better predictive ability such models could better account for natural and anthropogenic effects on the atmospheric CO₂ growth rate and offer reduced uncertainty regarding the potential success of near-term mitigation strategies.