Novel Representation of Leaf Phenology Improves Simulation of Amazonian Evergreen Forest Photosynthesis in ORCHIDEE Model

Abstract

Leaf phenology in the humid tropics largely regulates the seasonality of forest carbon and water exchange. However, it is inadequately represented in most global land surface models due to limited understanding of its controls. Based on intensive field studies at four Amazonian evergreen forests, we propose a novel, quantitative representation of tropical forest leaf phenology, which links multiple environmental variables with the seasonality of new leaf production and old leaf litterfall. The new phenology simulates higher rates of leaf turnover (new leaves replacing old leaves) in dry seasons with more sunlight, which is then implemented in ORCHIDEE, together with recent findings of ontogeny‐associated photosynthetic capacity, and is evaluated against ground‐based measurements of leaf phenology (canopy leaf area index and litterfall), eddy covariance fluxes (photosynthesis and latent heat), and carbon allocations from field observations. Results show the periodical cycles of solar radiation and vapor pressure deficit are the two most important environmental variables that are empirically related to new leaf production and old leaf abscission in tropical evergreen forests. The model with new representation of leaf phenology captures the seasonality of canopy photosynthesis at three out of four sites, as well as the seasonality of litterfall, latent heat, and light use efficiency of photosynthesis at all tested sites, and improves the seasonality of carbon allocations to leaves, roots, and sapwoods. For regional simulations, the new model version not only captures better than the former version in simulating GPP, compared with gridded data-driven FLUXCOM-GPP products and four GPP proxies derived from satellites: solar-induced fluorescence (SIF), Enhanced Vegetation Index (EVI) and Near Infra-Red reflectance of vegetation (NIRv). The new version models a dry-season increase of GPP across Amazonia for grid-cells with mean annual precipitation (MAP) higher than a threshold value of 2000 mm.yr-1, consistent with observational GPP proxies and confirming previous results by Guan et al. (2015). We tested two potential climatic triggers for leaf shedding before the dry seasons: vapor pressure deficit (VPD) and short-wave incoming radiation (SW). Both SW and VPD were found be good precursors of the seasonal litter fall peak by comparing against in situ data from 14 sites. This study advances understanding of the environmental controls on tropical leaf phenology and offers an improved modeling tool for gridded simulations of interannual CO 2 and water fluxes in the tropics.