Key points are not available for this paper at this time.
This paper examines the calibration of airborne laser scanning based tree attribute models to separate data by applying a best linear unbiased predictor. Firstly, single Scots pine ( Pinus sylvestris L.) trees were identified from dense airborne laser scanning data. Secondly, seemingly unrelated mixed-effects models for diameter at breast height, tree height, volume, dead branch height, and crown base height were constructed using airborne laser scanning based height metrics as predictors at both the area and individual tree level. Finally, these models were calibrated to validation stands using field measurements of some of the five abovementioned tree attributes. The models were calibrated by applying the best linear unbiased predictor to predict the random stand effects for the validation stand. In a system of several models, the correlation of random effects enabled the prediction of stand effects for all models, providing the response of at least one of the models was known for one or more sample trees of the validation stand. The results showed that the accuracy of tree attribute prediction improved in most cases as the number of sample trees increased. The level of improvement was highest for volume and dead branch height. The practical importance of the results of this study lies in applications where forest stands are visited in the field, for example, before making cutting decisions.
Maltamo et al. (Thu,) studied this question.