Abstract
Biological nitrogen fixation (BNF) is a key contributor to sustaining the terrestrial carbon cycle, providing nitrogen input that plants require. This is particularly salient for projections of carbon uptake under increased atmospheric carbon dioxide concentrations, which may allow for so-called ‘carbon dioxide fertilisation’ if other plant requirements, such as nitrogen, do not prevent increases in productivity. The amount, processes, and global distribution of BNF is highly disputed and consequently land surface models represent it in varying ways. Looking at the latest generation of CMIP6 earth system models with terrestrial nitrogen cycles, we consider their performance with regard to BNF. We assess models against a new comprehensive meta-analysis of BNF field measurements that gives a global range and site-specific values. We find that compared to the wide range of upscaled observations, the models still have a larger range, with under and overestimates.