Abstract. Global biosphere models vary greatly in their projections of future changes of global soil organic carbon (SOC) stocks and aggregated global SOC masses in response to climate change. We estimated the certainty (likelihood) and quantity of increases and decreases on a half-degree grid. We assessed the effect of changes in controlling factors, including net primary productivity (NPP), litter quality, soil acidity, water saturation, depth of permafrost, land use, temperature, and aridity associated with probabilities (Bayesian network) on an embedded, temporally discrete, three-pool decomposition model. In principle, controlling factors were discretized into classes, where each class was associated with a probability and linked to an output variable. This creates a network of links that are ultimately linked to a set of equations for carbon (C) input and output to and from soil C pools. The probability-weighted results show that, globally, climate effects on NPP had the strongest impact on SOC stocks and the certainty of change after 75 years. Actual land use had the greatest effect locally because the assumed certainty of land use change per unit area was small. The probability-weighted contribution of climate to decomposition was greatest in the humid tropics because of greater absolute effects on decomposition fractions at higher temperatures. In contrast, climate effects on decomposition fractions were small in cold regions. Differences in decomposition rates between contemporary and future climate were greatest in arid subtropical regions because of projected strong increases in precipitation. Warming in boreal and arctic regions increased NPP, balancing or outweighing potential losses from thawing of permafrost. Across contrasting NPP scenarios, tropical mountain forests were identified as hotspots of future highly certain C losses. Global soil C mass will increase by 1% with a certainty of 75% if NPP increases due to carbon dioxide fertilization. At a certainty level of 75%, soil C mass will not change if CO2-induced increase of NPP is limited by nutrients.