Dynamic scenario discovery under deep uncertainty: The future of copper

摘要

Scenarios are commonly used to communicate and characterize uncertainty in many policy fields. One of the main challenges of scenario approaches is that analysts have to try and capture the full breadth of uncertainty about the future in a small set of scenarios. In the presence of deep uncertainty, this is even more challenging. Scenario discovery is a model-based technique inspired by the scenario logic school that addresses this challenge. In scenario discovery, an ensemble of model runs is created that encompasses the various uncertainties perceived by the actors involved in particular decision making situations. The ensemble is subsequently screened to identify runs of interest, and their conditions for occurring are identified through machine learning. Here, we extend scenario discovery to cope with dynamics over time. To this end, a time series clustering approach is applied to the ensemble of model runs in order to identify different types of dynamics. The types of dynamics are subsequently analyzed to identify dynamics that are of interest, and their causes for occurrence are revealed. This dynamic scenario discovery approach is illustrated with a case about copper scarcity.