Testing the performance of a forest ecosystem model (FORECAST) against 29 years of field data in a Pseudotsuga menziesii plantation

Abstract

The ability of the forest ecosystem management model FORECAST to project a 29-year record of stand response to factorial thinning and fertilization treatments in a Douglas-fir ( Pseudotsuga menziesii (Mirb.) Franco) plantation at Shawnigan Lake (Vancouver Island, British Columbia, Canada) was assessed. Model performance was evaluated firstly using for calibration a regional data set and secondly with site-specific data from control plots. Model output was compared against field measurements of height, diameter, stem density, component biomass (aboveground), and litterfall rates and estimates of nutrient uptake, foliar N efficiency, and understory vegetation biomass. When calibrated with regional data, results from graphical comparisons, three measures of goodness-of-fit, and equivalence testing demonstrated that FORECAST can produce predictions of good to moderate accuracy depending on the variable of interest. Model performance was generally better when compared with field measurements (e.g., top height, diameter at breast height, and stem density) as opposed to outputs derived from allometric and volume equations. Use of site-specific data to calibrate the model always improved performance, although improvements were modest for most variables, with the exception of branch and foliage biomass. The benefits of site-specific calibration, however, should be weighed against the costs of obtaining such data. The intended use of the model will likely determine the level of effort expended in its calibration.