A note on the application of a distance-independent individual-tree growth model to inventory data.

摘要

An increasing use of forest sampling and a huge progress in individual-tree growth modelling offered a new field of application for individual-tree growth models: predicting forest growth by updating tree data of individual sample plots. The fact that the sampling design for collecting the parameterization data of the model, in particular plot size, has an effect on specific model coefficients is usually overlooked. Therefore, biased predictions will be obtained if a different sampling design is used for collecting data that should be used for growth predictions. Hans and Zumrawi [See Forest Science (1991) 37 1641-1655] showed that this bias plays only a subordinated role for a single-period growth prediction. However they pointed out that it is unknown how this error will propagate, but they expected an accumulation in case of long term projections. In this study, this hypothesis was tested by means of two simulation examples using the distance-independent individual-tree growth simulator PROGNAUS. Data from a 20-year-old stand of Norway spruce (Picea abies) in Germany were used. Results show that the bias of the simulated basal area increases as the length of the simulation period increases. In general, results indicate that the tree record splitting is an appropriate method to prevent a deviation from the original sampling design of the parameterization data. Therefore, it allows for a differentiation with regard to tree size (diameter at breast height), making the simulation results of an existing model more realistic..