As we develop increasingly complex “bottom-up” models of multi-sector dynamics, sources of uncertainty multiply and, with them, the need to represent them effectively, efficiently and in a way that preserves interpretability. In this talk, using the experience developed while trying to represent climate impacts in emission and land use change scenarios of the Global Change Analysis Model (GCAM), I will give an overview of the challenges and solutions we have started to explore, in our quest to improve our understanding and quantification of the interacting influences, possible feedbacks, and uncertainties and variability affecting the outcomes of the model. Challenges are related in large part to time- and space-scale mismatches, gaps in the representation of processes, what is treated as endogenous or exogenous in a model like GCAM, computational constraints. Solutions are being proposed not only through model development, but through the development and use of emulators, used in an auxiliary role to GCAM. We are developing emulators not only – and more traditionally -- to bypass the computationally demanding earth system models supplying climate information to GCAM, but also – in a novel approach -- to emulate GCAM itself. This helps to gage model sensitivities and to better explore input and parameter spaces complementing more established scenario discovery approaches.