Impacts of CO2 concentration and climate change on the terrestrial carbon flux using six global climate–carbon coupled models

•Although the simulations indicate increases of NPP and NEP owning to CO2 fertilization effect, the strength of trends significantly differs among the CMIP5 models.•Low agreement in the magnitude of precipitation and soil moisture among CMIP5 models can be an important source of uncertainty for NPP.•Nitrogen cycle component of CMIP5 is very vital with restraining the overestimated effect of CO2 fertilization.

Based on the simulations of the fifth phase of the Coupled Model Intercomparison Project (CMIP5), we estimated the response of net primary production (NPP) and net ecosystem production (NEP) to rising atmospheric CO2 concentration and climate change on global and regional scales. The modeled NPP and NEP significantly increased by about 0.4 Pg C yr−2 and 0.09 Pg C yr−2, respectively, in response to the rising atmospheric CO2 concentration. However, adverse trends of the two variables were driven by climate change on a global scale. Regarding the spatial pattern, the decreases were mainly located in tropical and temperate regions. Thus, the terrestrial carbon sink was accelerated not only by a rising atmospheric CO2 concentration, but also by global warming at high latitude and altitude regions, e.g. Tibet and Alaska. Although the simulations indicated increases of NPP and NEP owing to the CO2 fertilization effect, the strength of the trends significantly differed from the CMIP5 models. The enhanced trend in the terrestrial carbon sink simulated by MPI-ESM-LR was about 47 times larger than that simulated by CESM-BGC considering the CO2 fertilization effect. Differences in the modeled responses of NPP and NEP resulted from the various processes of the land surface component accounting for the nitrogen limitation effect and plant functional types (PFTs). We also found that the difference in the accelerating terrestrial carbon loss forced by global warming between CMIP5 models, ranged between 6.0 Tg C yr−2 in CESM-BGC and 52.7 Tg C yr−2 in MPI-ESM-LR. Such a divergence was partially responsible for the difference in the simulated climate between the CMIP5 models: the difference in increasing temperature was about 1.4 K.