Seasonality of leaf area index and photosynthetic capacity for better estimation of carbon and water fluxes in evergreen conifer forests

Abstract Leaf area index (LAI), defined as one half the total leaf area per unit ground area, and Vcmax, representing the maximal carboxylation rate of leaves, are two most significant parameters used in most Terrestrial Biosphere Models (TBMs). The ability of TBMs to simulate gross primary productivity (GPP) and evapotranspiration (ET) for evergreen needle-leave forests (ENF) can be significantly hampered by uncertainties in LAI and Vcmax. Remotely sensed (RS) LAI for ENF is generally underestimated in winter, early spring and late autumn. Although constant Vcmax throughout the growing season is often used in TBMs for GPP and ET modeling, it could vary significantly under leaf aging and stressed conditions. There were recent studies that apply seasonal leaf chlorophyll constraints on GPP modeling for croplands and deciduous forests, but little attention is given to the influence of the seasonality of either LAI or Vcmax on GPP or ET estimations for the ENF biome. In this study, we pay special attention to this biome, with the purpose of investigating if the representations of seasonal LAI and Vmax variations are essential in TBMs. To serve this purpose, the University of Toronto LAI product Version 2 was corrected for its seasonal variation using leaf lifespan and in-situ measurements at eight ENF sites in Canada. Seasonal Vcmax variation was derived from the MERIS Terrestrial Chlorophyll Index (MTCI) through downscaling it to the leaf level using a scheme with a general vertical nitrogen distribution within the canopy. Leaf chlorophyll content (LCC) is thus derived from MTCI and converted to Vcmax using empirical equations. Four model cases with and without considerations of the seasonal LAI and Vcmax variations were tested and compared. Validation against eddy covariance measurements indicates that the case with both LAI and Vcmax variations produced the highest R2, lowest root mean square error (RMSE) and lowest mean absolute error (MAE) for both GPP and ET simulations, and thus outperforms all other cases without considering the variations or with consideration of one of the variations only. In this best case, the simulated daily GPP yields R2 of 0.91, RMSE of 0.91 g C m−2 and MAE of 0.65 g C m−2, while the simulated daily ET yields R2 of 0.8, RMSE of 0.52 mm and MAE of 0.34 mm. Most improvements were found in spring and autumn. Not only the correlations between the seasonal trajectories of model simulation and observation were improved, but also the annual total GPP and ET were more accurately estimated. The smallest mean absolute relative bias to eddy covariance measurements is 9% for GPP and 15% for ET, both were found in the best case. Moreover, improvements in GPP were more pronounced than in ET. Our results highlight the significance of considering both seasonal structural and physiological characteristics of leaves in TBMs. Considering the important role that evergreen coniferous forests play in global terrestrial ecosystems, global simulations of GPP and ET in space and time can benefit from the proper representation of seasonal variations in canopy structure and leaf physiology as represented by LAI and Vcmax, respectively.