Abstract
Amazonian forest plays a crucial role in regulating the carbon and water cycles in the global climate system. However, the representation of biogeochemical fluxes and forest structure in dynamic global vegetation models (DGVMs) remains challenging. This situation has considerable implications to simulate the state and dynamics of Amazonian forest. This study aims at simulating the dynamic of the evapotranspiration (ET), productivity (GPP), biomass (AGB) and forest structure of wet tropical forests in the Amazon basin using the updated ORCHIDEE land surface model. The latter is improved for two processes: stand structure and demography, and plant water uptake by roots. Stand structure is simulated by adapting the CAN version of ORCHIDEE, originally developed for temperate forests. Here, we account for the permanent recruitment of young individual trees, the distribution of stand level growth into 20 different cohorts of variable diameter classes, and mortality due to asymmetric competition for light. Plant water uptake is simulated by including soil-to-root hydraulic resistance (RS). To evaluate the effect of the soil resistance alone, we performed factorial simulations with demography only (CAN) and both demography and resistance (CAN-RS). AGB, ET and GPP outputs of CAN-RS are also compared with the standard version of ORCHIDEE (TRUNK) for which eco-hydrological parameters were tuned globally to fit GPP and evapotranspiration at flux tower sites. All the model versions are benchmarked against in situ and regional datasets. We show that CAN-RS correctly reproduce stand level structural variables (as CAN) like diameter classes and tree densities when validated using in-situ data. Besides offering the key advantage to simulate forest's structure, it also correctly simulates ET and GPP and improves fluxes spatial patterns when compared to TRUNK. With the new formulation of soil water uptake, which is driven by soil water availability rather than root-biomass, the simulated trees preferentially use water in the deepest soil layers during the dry seasons. This improves the seasonality of ET and GPP compared to CAN, especially on clay soils for which the soil moisture potential drops rapidly in the dry season. Nevertheless, since demography parameters in CAN-RS are constant for all evergreen tropical forests, spatial variability of AGB and basal area across the Amazon remains too uniform compared to observations, and are very comparable to the TRUNK. Additional processes such as climate driven mortality and phosphorus limitation on growth leading to the prevalence of species with different functional traits across the Amazon need to be included in the future development of this model.