Modelling carbon and water cycles in a beech forest

A forest ecosystem model (CASTANEA) simulating the carbon balance (canopy photosynthesis, autotrophic and heterotrophic respirations, net ecosystem exchange, wood and root growth) and the water cycle (transpiration, soil evaporation, interception, drainage and soil water status) is tested with data from a young beech forest (Fagus sylvatica L.). For this purpose, the model validity is assessed by comparison between net CO2 and H2O fluxes simulated and measured by the eddy flux technique over one year. In addition, most of the sub-models describing the processes mentioned above are tested using independent measurements from the same forest stand: tree growth, branch photosynthesis, wood and soil respirations, sap flow and soil water content. Most of the input parameters (both weather and plant characteristics) are measured in the same experimental site (i.e. Hesse forest) independently of the validation dataset (none has been fitted to match the output data, except rainfall interception parameters); some are from other beech sites or from literature. Concerning the radiative transfer, the model reproduces the measured exponential PAR extinction and provides a good estimate of the net radiative budget, except during winter. At the branch scale, simulated photosynthesis and transpiration of sun-leaves are close to the measurements. We show also, using model simulations, that the seasonal decrease of measured net photosynthesis at the branch level could be explained by a decrease in leaf nitrogen content during the leafy season. At stand scale, a good correlation was obtained between simulated and observed fluxes both on a half-hourly basis and on a daily basis. Except at the end of the leafy season, the model reproduces reasonably well the seasonal pattern of both CO2 and H2O fluxes. Finally, even if there are some discrepancies between model estimations and fluxes measured at stand scale by eddy covariance, the model simulates properly both annual carbon and water balances when compared with the sum of the measured local fluxes. The remaining differences question the scaling up process when building such a model and the spatial footprint of eddy fluxes measurements.