Identification of Linear Climate Models from the CMIP3 Multimodel Ensemble

Abstract

In this paper, we investigate an ensemble of atmosphere–ocean general circulation models (AOGCMs) participating in phase 3 of the Coupled Model Intercomparison Project (CMIP3), and whose dynamic behavior is emulated by the reduced-complexity climate model MAGICC6. Using a superimposed impulse in solar radiative forcing for the purposes of system identification, we identify 12 AOGCMs in the CMIP3 ensemble for which low-order linear, time-invariant (LTI) models are able to very closely approximate MAGICC6 surface temperature projections under each of the four greenhouse gas (GHG) emission scenarios known as the Representative Concentration Pathways (RCPs) in the IPCC Fifth Assessment Report (AR5), even when extended to centennial timescales. The linear climate models identified in this paper are suitable for the analysis of feedback-based approaches to mitigation aimed at stabilizing global-mean surface temperature, and will inform future quantitative assessment of closed-loop approaches to geoengineering of the climate based on solar radiation management (SRM).