To assess the effect of enhanced nitrogen deposition and CO2 on European carbon uptake we used a terrestrial biogeochemical model BIOME-BGC (version 4.1.1 with carbon and nitrogen allocation routine from 4.1), which calculates water, carbon, and nitrogen pools dynamics as well as their fluxes on a daily basis[16,30]. The model is driven by maximum and minimum air temperatures, precipitation, air humidity, and solar radiation data. Carbon dynamics include calculations of the plant growth onset and senescence periods, allocation of assimilates to the different plant organs, mortality as well as litter production, and soil organic matter decomposition. Nitrogen dynamics include calculations of plant and soil microbial demands based on carbon to nitrogen ratios of plant organs, litter, and soil microbial community. The amount of nitrogen available to satisfy these demands is determined by nitrogen deposited from atmosphere, biological nitrogen fixation and nitrogen mineralized during soil organic matter decomposition. Nitrogen loss from ecosystem is determined by the amount of soluble mineral nitrogen available, water outflow, and soil water content.
Possible forest decline caused by high nitrogen inputs were not included in the model, because it was not relevant on the coarse grid scale considered in this study.
The Biome-BGC model was successfully evaluated for a number of hydrological and carbon cycle components [31-33]. In recent years, the model has been corroborated with eddy-covariance data at the sites with high and low deposition of atmospheric nitrogen [17,18,34,35].
The model was parameterized for seven vegetation types: deciduous broadleaf forest, evergreen needleaf forest, evergreen broadleaf forest, evergreen deciduous forest, scrubland, C4 and C3 grasslands. Ecophysiological parameters (e.g. carbon to nitrogen ratios of forest ecosystem's pools) for evergreen needleaf and deciduous broadleaf forests were optimized from field measurements of net carbon fluxes . Parameterizations for evergreen broadleaf forest, deciduous needleaf forest, grasslands, and shrubland were as in White et al. .
Model input data and simulations
All input data were transformed to 1° × 1° spatial resolution and subsequent model simulations were performed at this spatial resolution as well. Input land surface characteristics included digital elevation map, soil texture map, and land cover classification. Since in this study we did not consider land use changes, for all spatially explicit simulations of carbon, nitrogen, and water fluxes, a vegetation map  was held constant.
The model was first run with constant annual atmospheric nitrogen deposition (2 kgN/ha yr, Holland, 1999 #728) and CO2 concentrations (283 ppm) as well as daily climate data from NCEP Reanalysis  for 1948–1957 at a spatial resolution of 1° × 1° until an ecological equilibrium was reached.
Simulations with mature and re-growing forests
Previous studies  suggested age-related response of forest growth to changing abiotic conditions (increased temperature, CO2 and nitrogen deposition). To capture these effects we estimated carbon uptake of three different groups of forests: mature forests planted long before nitrogen deposition started to increase; middle age forests planted when the nitrogen deposition was still relatively low (1950), but exponentially increasing thereafter; and young forests planted when nitrogen deposition was already high (1970) and exponentially increasing during their lifetime (Figure 2).
We isolated the effects of elevated nitrogen deposition on forests at different succession stages by simulations with CO2 increase, but no nitrogen deposition increase and with both CO2 and nitrogen increases. In each simulation we assumed that forests had the same age distribution.
Simulation with increasing CO2
Between 1860 and 1999, ambient CO2 increased from 283 ppm to 368 ppm. To isolate the effect of increased CO2 on carbon uptake we performed model simulations with nitrogen deposition kept constant at 1860 level and CO2 changing as described above.
Simulation with increasing CO2 and nitrogen deposition
To isolate the additional effect of increased nitrogen deposition on carbon uptake we performed model simulations with changing both atmospheric CO2 and nitrogen deposition.
The BIOME-BGC model simulations were driven by the state-of-the-art atmospheric nitrogen deposition for 1860–1999 (Figure 2). Spatial distribution of atmospheric nitrogen deposition was estimated with three dimensional atmospheric chemical transport model TM3 . The estimates included deposition of both NOy and NHx, which were added to get the total atmospheric nitrogen deposition.
These data were produced using the TM3 global chemistry transport model that has a horizontal resolution of 5° longitude and 3.75° latitude, and has 19 vertical levels to 10 hPa. It is driven by six-hourly meteorological data obtained from the European Centre for Medium-Range Weather Forecasting. Mid-1990s deposition rates are based on Rodhe et al.  combining output for reduced and oxidized nitrogen deposition (both wet and dry). Comparisons between modelled and measured deposition rates show model accuracy to be within 50% (and often substantially better). The main sources of error arise from emissions inventories, atmospheric transport, removal, and chemical transformations. The original annual model outputs at decadal time step were transformed to annual outputs at one year time step using linear interpolation for each grid pixel.