Link to article.
Models estimated for non-wood forest products (NWFPs) are used to support the multi-objective use of forest and related decision making in various ways. In Finland, bilberries (Vaccinium myrtillus) and cowberries (Vaccinium vitis-idaea) are among the most important NWFPs, and yield models for them have been estimated and linked into forest simulators. So far, berry yield models have not been evaluated using independent data that were not used in model estimation. In this study, independent data on berry yields, as well as stand and site characteristics used as predictors in yield models, were measured from a total of 230 sample plots in North Karelia, Finland. The predictions of four bilberry and four cowberry models published previously were evaluated on the basis of the models’ prediction bias, precision and applicability in locating the best berry stands for picking. The bias and precision of the models varied highly. Due to bias, the models should be calibrated before applying them in forest decision support systems. The models for bilberry located the best berry stands more reliably than did those for cowberry. Visits to twice as many stands, as guided by the models, were needed to harvest a given amount of bilberries, but, for cowberry, the number of stands to be visited was four-fold in comparison to the situation where the measured yields were known. Besides the forest inventory data measured in the field, we also used the freely available multi-source national forest inventory (MS-NFI) data covering the whole of Finland as input data for the berry models. Both data sets were equally suitable to be used as the input data for the berry models.