Projecting future concentrations of atmospheric CO2 with global carbon cycle models: The importance of simulating historical changes

Abstract

Projections of future atmospheric CO2 concentrations using global carbon cycle models and assumed time series of future anthropogenic CO2 emissions are only useful if simulations agree reasonably well with the observed history of past changes in atmospheric CO2. In this article we compare simulations from a set of eight global carbon cycle models with observations of atmospheric CO2 from the Siple Station, Antarctica, ice core and the monitoring station at Mauna Loa Observatory, Hawaii, USA. Our comparisons reinforce previous assessments that early estimates of biospheric CO2 emissions derived by reconstruction of historical land-use change are incompatible with the understanding of atmosphere-ocean CO2 exchange codified in conventional carbon cycle models and the observed history of changes in atmospheric CO2. More recent estimates of the history of CO2 emissions associated with land-use change do not significantly resolve this incompatibility. Terrestrial biospheric emissions estimated by deconvolution of atmospheric CO2 observations provide reasonable correspondence between simulation and observation, but the deconvolution estimates differ dramatically from the estimates by land-use reconstruction. Resolution of this difference is a challenge for modelers of the global terrestrial biosphere. In the interim, caution is required in interpreting atmospheric CO2 projections from models that have not yet resolved the basic inconsistencies among emission estimates, models of oceanic uptake, and observations of atmospheric CO2.