Discovering plausible energy and economic futures under global change using multidimensional scenario discovery


The use of scenarios has proven valuable for global change analysis as a means for organizing and communicating information about uncertain future socioeconomic conditions. A small group of well-defined scenarios can provide a set of standard reference cases for assessing the performance of candidate policies under alternative futures. However, traditional methods of defining scenarios may yield storylines that do not align well with the capabilities of downstream models. It can also be difficult to assess whether the number and scope of constructed scenarios most effectively cover the space of possible outcomes. Model-based methods of 'scenario discovery' have recently been proposed that apply statistical data-mining algorithms to a large number of model simulations to identify regions of the stochastic parameter space that lead to unacceptable policy performance. These regions then delineate practically relevant and internally consistent 'discovered scenarios'. To distinguish 'acceptable' from 'unacceptable' policy outcomes, existing methods require pre-specification of a threshold value on a single performance metric. We believe this requirement may present a barrier when decision-makers hold differing views on the relative importance of multiple policy objectives. Therefore, we describe a scenario discovery method that is multidimensional in the outcome space, thus precluding the need for users to agree on a single performance threshold or set of tradeoff weights. We demonstrate application of our approach to the results of ENGAGE, an agent-based model (ABM) of economic growth, energy technology, and carbon emissions. We believe that scenario discovery can add particular value to agent-based modeling, as ABMs typically generate a wider array of possible futures than aggregate-scale models. For this reason, a systematic method for sorting through the many stochastic model simulations to identify policy vulnerabilities and opportunities is essential. We conclude by discussing how our methodology might be applicable to the development of socioeconomic scenarios under the 'representative concentration pathways' (RCP) framework. © 2012 Elsevier Ltd.