Item

Abstract

The authors assess the ability of 18 Earth system models to simulate the land and ocean carbon cycle for the
present climate. These models will be used in the next Intergovernmental Panel on Climate Change (IPCC) Fifth
AssessmentReport (AR5) for climate projections, and such evaluation allows identification of the strengths
and weaknesses of individual coupled carbon–climate models as well as identification of systematic biases of
themodels. Results show thatmodels correctly reproduce the main climatic variables controlling the spatial
and temporal characteristics of the carbon cycle. The seasonal evolution of the variables under examination
is well captured. However, weaknesses appear when reproducing specific fields: in particular, considering
the land carbon cycle, a general overestimation of photosynthesis and leaf area index is found for most of
the models, while the ocean evaluation shows that quite a few models underestimate the primary production.
The authors also propose climate and carbon cycle performance metrics in order to assess whether
there is a set of consistently better models for reproducing the carbon cycle. Averaged seasonal cycles and
probability density functions (PDFs) calculated from model simulations are compared with the corresponding
seasonal cycles and PDFs from different observed datasets. Although the metrics used in this study allow
identification of somemodels as better or worse than the average, the ranking of this study is partially subjective
because of the choice of the variables under examination and also can be sensitive to the choice of reference
data. In addition, it was found that the model performances show significant regional variations.