Tropical forests play a major role in global carbon cycling, accounting for about half of the carbon stored by land plants and more than a third of total carbon fixation by the terrestrial biosphere. However, there is considerable uncertainty regarding the mechanisms governing spatial and temporal variation in tropical forest carbon stocks and fluxes, and a great debate concerning their future trajectories under global change. Thus, there is a recognized need to improve the representation of tropical forests in Earth System Models (ESMs) to better predict future global carbon budgets.
Here, we adapted the vegetation modules of the land model LM4-PPA to analyze tropical forest dynamics at Barro Colorado Island (BCI), Panamá. LM4-PPA is a fully mechanistic, individual based model that features the Perfect Plasticity Approximation, an approach to model height-structured dynamics for forest canopies. Taking advantage of extensive data sets on tree morphological traits and performance at BCI, we developed new parameterizations for tree allometry, mortality and growth allocation. We evaluated model performance against field-based estimates of forest biomass, productivity and structure, as well as new diagnostics based on growth rates on different canopy layers.
Results/Conclusions
The analysis of tree allometric scaling at BCI revealed a saturating relationship between tree height and trunk size, in contrast to the power function scaling traditionally used in this type of models. This new parameterization proved important to improve estimates of aboveground biomass and to avoid the simulation of very large trees. We parameterized a new allometric relationship for branch biomass scaling in tropical forests to fix a dynamic target on growth allocation. Finally, we also incorporated species-specific, background mortality rates that varied depending on the position within the canopy.
With the inclusion of these new parameterizations, LM4-PPA was able to reproduce forest patterns at a variety of temporal scales and levels of organization. The model provided realistic estimates of total biomass and production, including diurnal and seasonal cycles comparable to flux tower measurements at BCI. It also reproduced emergent growth patterns across different vegetation layers that matched observations based on long term monitoring of tree survival. Finally, LM4-PPA was able to simulate community level emergent patterns, including a realistic size structure and biomass partitioning among canopy layers. These improvements represent a major advancement towards the projection of tropical forest response to climate change.