Probabilistic climate outcomes from prediction aggregation

Currently, much of the literature around the Paris Agreement, Paris Compliance and manging the transition to net zero requires heavy use of integrated assessment models (IAMs). IAMs provide economic projections of future emissions, conditional on idealised scenarios. However, for most adaptation and cost-benefit analysis, policymakers require predictions, which IAMs do not even attempt to provide. How can we use aggregated estimates of emissions and resulting climate change to give probability distributions of climate impacts? We outline why human computation likely out-performs other prediction methods and present a flexible method to collect intended predictions from a variety of people to effectively estimate future emissions, temperatures and climate impacts via prediction aggregation platforms. These can subsequently be used to inform estimates of climate impacts. It can also highlight deficiencies in the IAM scenarios literature and indicate relative probabilities of scenarios. We estimate all-uncertainty temperatures in 2050 and outline extensions of the work.