Nitrogen is limiting in many growth environments and its acquisition is energetically costly even under non-limiting conditions. In addition leaf photosynthesis and leaf N content are strongly correlated. A high rate of whole-plant carbon gain per unit nitrogen acquired could therefore be considered adaptive. Using stand-based canopy models we determined the optimal distribution of leaf nitrogen contents and leaf area indices (LAI) at which whole-stand canopy photosynthesis per unit nitrogen is maximized (simple optimization), and compared these to measured data. While certain observed trends were well predicted, there were consistent deviations between predicted and measured values. Measured N distributions were less steep than predicted ones while plants produced greater than optimal amounts of leaf area.

Why was there a difference? Simple optimization ignores that since plants interact with each other (they compete for light and resources) the optimal characteristics of a plant will depend on the density and characteristics of its neighbors. In these cases game theory (competitive optimization) rather than simple optimization should be applied. Using game theory in combination with individual-based canopy models we showed that vegetation stands with an optimal LAI can be invaded by plants that produce more than optimal leaf area. The predicted evolutionarily stable LAI, the LAI at which no mutant can increase its photosynthesis by changing its leaf area, corresponded well with measured LAIs. Similar analyses were performed for other traits and it was predicted that populations of plants of optimal height or with optimal leaf angles or root masses could be invaded by taller plants with more horizontal leaves, producing more roots. This suggests that in dense vegetation, plants should not simply maximize their own performance irrespective of neighbors, but they should do better than their neighbors.

The above proposition however still lacked an experimental validation of the supposed link between nitrogen distribution and LAI, and plant performance (fitness). We therefore used transgenic tobacco (SAG) with delayed leaf senescence, resulting in more uniform N distributions and larger LAIs than in wild type tobacco (WT). WT plants used the additional nitrogen reallocated from senescing leaves to produce more leaf area towards the top of the canopy, and we predicted that this would make them competitively stronger than SAG plants. Indeed while growth and seed production of SAG plants was similar or higher than that of WT plants in open stands, they declined more strongly with increasing density of neighbor plants. This supports the assumption underlying our game theoretical canopy models, that N reallocation and adjustment of LAI contributes to competitiveness rather than individual-based maximization of whole-plant carbon gain.