Leaf nitrogen from the perspective of optimal plant function

1. Leaf dry mass per unit area (LMA), carboxylation capacity (V (cmax)) and leaf nitrogen per unit area (N(area)) and mass (N(mass)) are key traits for plant functional ecology and ecosystem modelling. There is however no consensus about how these traits are regulated, or how they should be modelled. Here we confirm that observed leaf nitrogen across species and sites can be estimated well from observed LMA and V (cmax) at 25°C (V (cmax25)). We then test the hypothesis that global variations of both quantities depend on climate variables in specific ways that are predicted by leaf‐level optimality theory, thus allowing both N(area) to be predicted as functions of the growth environment. 2. A new global compilation of field measurements was used to quantify the empirical relationships of leaf N to V (cmax25) and LMA. Relationships of observed V (cmax25) and LMA to climate variables were estimated, and compared to independent theoretical predictions of these relationships. Soil effects were assessed by analysing biases in the theoretical predictions. 3. LMA was the most important predictor of N(area) (increasing) and N(mass) (decreasing). About 60% of global variation across species and sites in observed N(area), and 31% in N(mass), could be explained by observed LMA and V (cmax25). These traits, in turn, were quantitatively related to climate variables, with significant partial relationships similar or indistinguishable from those predicted by optimality theory. Predicted trait values explained 21% of global variation in observed site‐mean V (cmax25), 43% in LMA and 31% in N(area). Predicted V (cmax25) was biased low on clay‐rich soils but predicted LMA was biased high, with compensating effects on N(area). N(area) was overpredicted on organic soils. 4. Synthesis. Global patterns of variation in observed site‐mean N(area) can be explained by climate‐induced variations in optimal V (cmax25) and LMA. Leaf nitrogen should accordingly be modelled as a consequence (not a cause) of V (cmax25) and LMA, both being optimized to the environment. Nitrogen limitation of plant growth would then be modelled principally via whole‐plant carbon allocation, rather than via leaf‐level traits. Further research is required to better understand and model the terrestrial nitrogen and carbon cycles and their coupling.