Quantifying photosynthetic capacity and its relationship to leaf nitrogen content for global‐scale terrestrial biosphere models

Abstract

Photosynthetic capacity and its relationship to leaf nitrogen content are two of the most sensitive parameters of terrestrial biosphere models (TBM) whose representation in global‐scale simulations has been severely hampered by a lack of systematic analyses using a sufficiently broad database. Here, we use data of qualitative traits, climate and soil to subdivide the terrestrial vegetation into functional types (PFT), and then assimilate observations of carboxylation capacity, V max (723 data points), and maximum photosynthesis rates, A max (776 data points), into the C 3 photosynthesis model proposed by Farquhar et al. to constrain the relationship of ( V max normalised to 25 °C) to leaf nitrogen content per unit leaf area for each PFT. In a second step, the resulting functions are used to predict per PFT from easily measurable values of leaf nitrogen content in natural vegetation (1966 data points). Mean values of thus obtained are implemented into a TBM (BETHY within the coupled climate–vegetation model ECHAM5/JSBACH) and modelled gross primary production (GPP) is compared with independent observations on stand scale. Apart from providing parameter ranges per PFT constrained from much more comprehensive data, the results of this analysis enable several major improvements on previous parameterisations. (1) The range of mean between PFTs is dominated by differences of photosynthetic nitrogen use efficiency (NUE, defined as divided by leaf nitrogen content), while within each PFT, the scatter of values is dominated by the high variability of leaf nitrogen content. (2) We find a systematic depression of NUE on certain tropical soils that are known to be deficient in phosphorous. (3) of tropical trees derived by this study is substantially lower than earlier estimates currently used in TBMs, with an obvious effect on modelled GPP and surface temperature. (4) The root‐mean‐squared difference between modelled and observed GPP is substantially reduced.