Kenneth Clark, Yude Pan, Richard Birdsey, John Hom, Steve Van Tuyl, and Nicholas Skowronski. USDA Forest Service, Northern Research Station, Newtown Square, PA 19073
Scaling eddy flux data collected from towers with ca. 1 km2 footprints up to landscape and regional scales is an important challenge for the scientific community. Integration across temporal and spatial scales involves a number of assumptions concerning carbon sink strengths, land surface characteristics, stand ages, and disturbance history. Here we present results from a network of flux towers, nested forest census plots, LIDAR remote sensing products and PnET CN modeling to estimate forest structure and productivity in the Pine Barrens of New Jersey. Eddy flux measurements indicated that annual productivity of mature upland forests types were similar, but seasonal C fluxes differed substantially; Oak-dominated stands forests were the largest sinks for CO2 in the summer, while Pine-dominated stands were larger sinks in early spring and in fall, following patterns of leaf display. Annual forest productivity predicted using PnET CN driven with meteorological data collected at the flux towers was similar to integrated net CO2 exchange measurements, but notable seasonal differences resulted from the timing of leaf expansion in the spring, and down-regulation of physiological activity and senescence in the fall. However, the largest deviation between measured and modeled fluxes resulted from defoliation by Gypsy moths at one of the flux sites in early summer 2006. Prescribed fire treatments released the equivalent of 1-4 years of accumulated C, which also resulted in large deviations between field estimates and model predictions. Both census plots and LIDAR data indicate that Oak-dominated stands have greater aboveground biomass than Pine-dominated stands across this landscape, with patterns strongly affected by fire history, herbivory, wind throw, and other disturbances. Our results indicate that this framework is useful for scaling up tower-based measurements, but also highlight the complexity that disturbances introduce in estimating patterns of C dynamics in forest ecosystems.