[A-CG35] Global Carbon Cycle Observation and Analysis

[ACG35-10] Climate and carbon cycle feedback under idealized CO₂ and non-CO₂ overshoot pathways

*Irina Melnikova^1,2, Philippe Ciais², Katsumasa Tanaka^1,2, Hideo Shiogama¹, Tokuta Yokohata¹, Olivier Boucher³ (1.National Institute for Environmental Studies (NIES), 2.Laboratoire des Sciences du Climat et de l’Environnement (LSCE), 3.Institut Pierre-Simon Laplace (IPSL))

Keywords:non-CO2, carbon cycle, climate, overshoot, ESM, mitigation

Reducing emissions of non-CO₂ greenhouse gases complements CO₂ mitigation in limiting global warming. However, estimates of the CO₂ and non-CO₂ carbon-climate feedback remain fraught with uncertainties, especially under overshoot scenarios. Here we carry out a series of idealised ramp-up and ramp-down experiments using the advanced Earth System Model IPSL-CM6A-LR and compare the effects of CO₂ and non-CO₂ gases with nearly equal magnitudes of the effective radiative forcing on the climate and the carbon cycle. As the model’s effective radiative forcing of methane and nitrous oxide do not compare well with more accurate published estimates, we first present a method to recalibrate these radiative forcings while keeping the original radiative scheme. We then confirm the model’s performance and show that on a global scale, the impacts of increasing and decreasing CO₂ and non-CO₂ gases on the surface climate are nearly equivalent but there are regional differences. We further investigate the carbon cycle feedback and show that as CO₂ concentration decreases, even in the absence of carbon-climate feedback, land and ocean become carbon sources on a global scale, but with large regional variability. The CO₂ and non-CO₂ radiative forcing and climate-carbon feedback are not equal and not additive. These features should be accounted for in the climate and carbon cycle emulators and considered in the greenhouse gas emission metrics. Our experimental framework for disentangling the compound feedback pathways involving CO₂ and non-CO₂ forcers could be applied to other models.