Process-based models are a powerful approach to test our understanding of biogeochemical processes, to extrapolate ground survey data from limited plots to the landscape scale, and to simulate the effects of climate change, nitrogen deposition, elevated atmospheric CO2, increasing natural disturbances and land use change on ecological processes. However, in most studies, the models are calibrated using ground measurements from only a few sites, though they may be extrapolated to much larger areas. Estimation accuracy can be improved if the models are parameterized using long-term carbon (C) stock data from multiple sites representative of the simulated region. In this study, forest biomass C stocks measured in 61 forested plots located in three research sites in the Delaware River Basin (DRB) were used to modify the PnET-CN model in three ways: (1) field measured mortality rates in each forest type were used to parameterize the wood turnover rate; (2) a numerical approach was used to calibrate the relationship between foliage N and maximum photosynthesis rate; and (3) stand age was incorporated into the model as an input variable, which determines the year of the last disturbance.
The results showed that these model modifications improved model performance in capturing the spatial variation of forest C dynamics in the DRB forests. The spatial distribution of forest C pools and fluxes in the three sites were mapped using the modified model. The modified model was also used in experimental scenarios, which predicted that 39% of forest C sequestered over the past decade could be attributed to the combined effects of elevated CO2 and N deposition. This study demonstrated an effective method for using long-term biometric measurements of forest biomass C stocks to improve a process-based ecosystem model at a regional scale. Further research should target on improving the model parameters which are sensitive to the spatial variation of forest C dynamics.