| Summary: | Abstract: Gross primary production (GPP) is the primary source of all carbon fluxes in the ecosystem. Understanding variation in this flux is vital to understanding variation in the carbon sink of forest ecosystems, and this would serve as input to forest production models. Using GPP derived from eddy-covariance (EC) measurements, it is now possible to determine the most important factor to scale GPP across sites. We use long-term EC measurements for six coniferous forest stands in Europe, for a total of 25 site-years, located on a gradient between southern France and northern Finland. Eddy-derived GPP varied threefold across the six sites, peak ecosystem leaf area index (LAI) (all-sided) varied from 4 to 22 m2 m2 and mean annual temperature varied from 1 to 13 °C. A process-based model operating at a half-hourly time-step was parameterized with available information for each site, and explained 7196% in variation between daily totals of GPP within site-years and 62% of annual total GPP across site-years. Using the parameterized model, we performed two simulation experiments: weather datasets were interchanged between sites, so that the model was used to predict GPP at some site using data from either a different year or a different site. The resulting bias in GPP prediction was related to several aggregated weather variables and was found to be closely related to the change in the effective temperature sum or mean annual temperature. High R2s resulted even when using weather datasets from unrelated sites, providing a cautionary note on the interpretation of R2 in model comparisons. A second experiment interchanged stand-structure information between sites, and the resulting bias was strongly related to the difference in LAI, or the difference in integrated absorbed light. Across the six sites, variation in mean annual temperature had more effect on simulated GPP than the variation in LAI, but both were important determinants of GPP. A sensitivity analysis of leaf physiology parameters showed that the quantum ...
|
|---|