Ecologists have long recognized that the composition and structure of vegetation varies consistently at a landscape-scale due to variation in soils, hydrology, and topography. We expect that this variation to interact with climate change in complex ways, potentially allowing some species to persist in refugia while shifting other species to locations that may be edaphically unfavorable. Despite the recognized importance of this variation, it has not been incorporated into regional-scale models because this heterogeneity occurs at a finer spatial scale than can be captured explicitly by refining model resolution. Rather than represent landscape-scale variability explicitly, we develop a spatially implicit approach to capture variation in soils, lateral hydrologic flow, and the effects of topography on microclimate and radiation interception. This scheme is incorporated in the Ecosystem Demography model. We tested this approach by first calibrating the model to inventory and eddy-covariance data from the Bartlett Experimental Forest in central NH and then validating it against 40+ years of vegetation and hydrology data from the Hubbard Brook Ecosystem Study, located 40km in forests of similar composition. The model will then be applied at a regional scale to forecast forest change over the next 100 years
Results/Conclusions
Using a Bayesian data assimilation scheme we show that the ED2.1 model is able to capture the variability in carbon, water, and energy fluxes at the Bartlett flux tower as well as observed tree growth rates. Model calibration suggests that rates of assimilation and fine root turnover among late successional hardwoods and conifers were lower than suggested by a previous calibration at Harvard Forest. When applied to Hubbard Brook the model is able to capture watershed streamflow at monthly to interannual scales but underestimates peak discharge following storm events. The model captures the variation in growth rates with topography, soils, and hydrology, and reproduces observed NPP during the forest growth phase. Growth rates were overestimated during the latter portion of the record, likely due to the cumulative impacts of acid rain which are not yet accounted for in the model. By sequentially switching off each source of edaphic variation, we find that the effect of elevation on microclimate has the greatest impact on the within-watershed distribution of NPP. The effects of slope and aspect on radiation are strongest at mid-elevation while lateral hydrology is most important on ridges and in valley-bottoms.
