Diagnosing Variability of Forest Productivity Using Eddy Covariance Data and a Light-use Efficiency Model

Nine years of eddy covariance measurements in a mature coastal Douglas-fir (Pseudotsuga menziesii) forest were used to calibrate a light-use efficiency (LUE) model. Daily total gross primary production (PG) was simulated as a function of absorbed photosynthetically active radiation (Qa), temperature (T), vapour pressure deficit (D) and relative extractable water (REW). Light absorption was modelled using Beer's law, assuming constant leaf area index (LAI). The objectives were to calibrate the model, assess the individual importance of environmental stresses in controlling PG, and quantify the overall fraction of inter-annual variability of PG attributed to climate variability. Using constrained nonlinear regression, the model calibration was able to explain 84 % of the variation in daily PG, however, model errors were not randomly distributed and the model failed to match observed inter-annual variability. Model residuals indicated strong seasonality, corresponding closely with indirect estimates of stand architecture. This suggests that use of Beer's law and the assumption of static LAI in the simulation of Qa may limit the ability of simple LUE models to simulate productivity of temperate evergreen forest ecosystems. After accounting for light saturation, the model indicated that T, D, and REW all exerted significant influence over daily PG. Residual analysis suggested that REW substantially reduced annual PG in 6 of 9 years. The individual effects of T and D on PG were more difficult to assert, owing to interdependence between driving variables, acclimation of photosynthesis to winter and summer temperature regimes, and seasonal variation in anisohydric behaviour.