Improvements in the terrestrial carbon cycle in CMIP models evaluated with satellite observations

With impacts of climate change already noticeable in every region of the world, understanding and accurately simulating the drivers of climate change is crucial. In particular, the global carbon cycle and its responses to changing carbon dioxide (CO2) emissions plays an important role. This thesis aims to identify important improvements and key processes relevant to accurately simulate the carbon cycle under climate change and to provide recommendations for further model developments. This is achieved by a comprehensive evaluation of historical simulations from earth system models (ESMs) participating in the last two phases of the Coupled Model Intercomparison Project (CMIP) with satellite observations. In a first study of this thesis, column-average CO2 mole fraction (XCO2) from CMIP5 and CMIP6 emission-driven ESM simulations are compared to satellite observations. A previously found discrepancy between a negative trend of the seasonal cycle amplitude (SCA) with increasing XCO2 in the northern midlatitudes shown by the observations with models showing an insignificant trend could be attributed to spatial sampling. Furthermore, while ESMs overestimate mean and growth rate of XCO2 while underestimating the SCA, the CMIP6 ensemble performs better than CMIP5 ensemble. In a second study, the present-day land carbon cycle is evaluated. While some long-standing biases could be resolved in CMIP6, such as the photosynthesis overestimation which was resolved through the inclusion of the interactive nitrogen cycle, other biases remain. Despite the increased process complexity in emission-driven simulations that fully account for the influence of climate-carbon feedbacks on atmospheric CO2, they perform just as well as CO2 concentration-driven simulations. Therefore, both the use of emission-driven over concentration-driven simulations, as well as the inclusion of interactive nitrogen cycles are recommended as a default setting for future CMIP phases.