A validation and evaluation of the Prognosis individual-tree basal area increment model

We subjected the individual-tree, aspatial basal area increment model developed for the Inland Empire Variant of the Forest Vegetation Simulator to validation and evaluation tests. We used a large set of independent data from the Forest Inventory and Analysis program that covers the geographic extent to which the model is usually applied. Equivalence tests did not validate the model as a predictive tool using nominated criteria, though they usually did validate the model structure as a theory. Design-unbiased estimates of prediction error suggest that the model overpredicts diameter and volume increment by 14% and 2%, respectively. Relationships between species, bias, and predictor variables suggest the model may overpredict most on productive sites. We spatially interpolated the model performance across the study area using thin-plate splines. The observed regional patterns are examined using a selection of cross-sectional transects, and reveal a complex relationship between bias and the way climate effects are incorporated in the model structure that involve differences between the fitting and testing data. The model structure is surprisingly robust, but the representation of climate effects should be a priority in future revisions.