This thesis' ultimate goal is to improve the risk and uncertainty analysis of global environmental issues. It starts with a review of the various uncertainty analysis techniques used in integrated assessment models. Then it presents results of the Dynamics of Inertia and Adaptability Model (DIAM) about the optimal timing of global change policy. That model follows the classical approach of precaution as sequential decision-making, mathematically formalized using stochastic dynamic programming. But that classical approach requires precise subjective probabilities. The central claim of this thesis is that uncertainty about long-run issues such as climate change are better represented using (objective or subjective) imprecise probabilities. This claim is supported both theoretically and with applications. Chapters 5 to 7 are an introduction to the mathematical theory of imprecise probabilities. Applications form the last section of this work. Applications in chapters 8 to 10 deal with expert elicitation, information fusion and scenario-making. They use special kinds of imprecise probabilities, namely those of the transferable belief model for chapter 8 and 9, along with distributions of possibility in chapters 9 and 10.