Improving terrestrial carbon cycle simulations with eco-evolutionary optimality: Including the P model in LPJ-GUESS

Terrestrial gross primary productivity (GPP) is a linchpin flux in the terrestrial carbon cycle and its simulation is a central component of dynamic global vegetation models (DGVMs). When calculating GPP, DGVMs typically rely on a light use efficiency (LUE) model which relates the amount of absorbed solar radiation to the amount of carbon fixed by photosynthesis. Recent theoretical advances utilising eco-evolutionary optimality (EEO) theory have led to the development of the P model, a parameter-sparse LUE model which has been well-validated at both local and global scales.

Here we implemented the P model into LPJ-GUESS, an established, community-developed DGVM. We compared LPJ-GUESS’s performance with and without the P model to remotely-sensed GPP estimates. The inclusion of the P model reduced the error in the simulated spatial pattern of annual GPP by 17% and markedly improved of the timing of the northern hemisphere spring green up. In order to disentangle the causes of data-model mismatch, we also investigated the GPP errors as a function of the environmental variables such as elevation, and in the case of elevation we found a strong model bias which was similar both with and without the P model.

In addition to the improved model skill, the P model version of LPJ-GUESS uses far fewer parameters (none of which are PFT specific), encapsulates a coherent body of theory reflecting more recent understanding of photosynthetic responses to changing environmental conditions, and has a reduced model run time. Based on this, we conclude that the P model has the potential to improve LPJ-GUESS and other DGVMs.