After some consensus has been achieved that the Earth's global mean temperature will have increased by 1.4 to 5.8C at the end of this century with ”business as usual” greenhouse gas emissions, society has to decide if or which mitigation measures should be taken. I begin this talk with results from a new integrated assessment project on this very issue. The portfolio of mitigation options that we consider simultaneously is constituted by the transformation of the energy system to renewable sources, the increase of energy efficiency, and carbon capturing and sequestration. Novel methods provide more realistic capture of innovation dynamics than in standard Integrated Assessment approaches, and renewable sources turn out to be more competitive. As the talk progresses, we acknowledge the fact that models are uncertain and include model parameter uncertainty in our optimization procedure. We utilize the Bayesian scheme of combining subjective and objective knowledge, but go beyond the Bayesian approach to include weaker uncertainty representations such as random sets rather than standard probability distributions. The quantitative scheme is complemented by a qualitative risk assessment of mitigation options. Options are ranked according to risk categories that are to be geared to the principles of reversibility and the degree of involved systems' understanding. I discuss whether the indicator "spatial range" - as successfully operationalized in environmental chemistry for the assessment of potentially harmful substances - could play a major role in this scheme.