Earth system model simulations show different feedback strengths of the terrestrial carbon cycle under glacial and interglacial conditions

In simulations with the MPI Earth System Model, we study the feedback between the terrestrial carbon cycle and atmospheric CO 2 concentrations under ice age and interglacial conditions. We find different sensitivities of terrestrial carbon storage to rising CO 2 concentrations in the two settings. This result is obtained by comparing the transient response of the terrestrial carbon cycle to a fast and strong atmospheric CO 2 concentration increase (roughly 900 ppm) in Coupled Climate Carbon Cycle Model Intercomparison Project (C 4 MIP)-type simulations starting from climates representing the Last Glacial Maximum (LGM) and pre-industrial times (PI). In this set-up we disentangle terrestrial contributions to the feedback from the carbon-concentration effect, acting biogeochemically via enhanced photosynthetic productivity when CO 2 concentrations increase, and the carbon–climate effect, which affects the carbon cycle via greenhouse warming. We find that the carbon-concentration effect is larger under LGM than PI conditions because photosynthetic productivity is more sensitive when starting from the lower, glacial CO 2 concentration and CO 2 fertilization saturates later. This leads to a larger productivity increase in the LGM experiment. Concerning the carbon–climate effect, it is the PI experiment in which land carbon responds more sensitively to the warming under rising CO 2 because at the already initially higher temperatures, tropical plant productivity deteriorates more strongly and extratropical carbon is respired more effectively. Consequently, land carbon losses increase faster in the PI than in the LGM case. Separating the carbon–climate and carbon-concentration effects, we find that they are almost additive for our model set-up; i.e. their synergy is small in the global sum of carbon changes. Together, the two effects result in an overall strength of the terrestrial carbon cycle feedback that is almost twice as large in the LGM experiment as in the PI experiment. For PI, ocean and land contributions to the total feedback are of similar size, while in the LGM case the terrestrial feedback is dominant.