A nonparametric technique for taper function estimation

Abstract

The most common method for determining tree profile models is by fitting an analytical function to a set of sample trees. In situations where the forester has relatively little a priori knowledge about the nature of profiles, choosing the analytical form of the parametric model is a critical problem without a satisfactory solution. An alternative is to use a nonparametric approach in which the model of the profile is explicitly specified by a tabulation of diameters at discrete heights. In this paper, it is shown how the model can be determined from the sample trees by means of a classical nonparametric probability density function estimation technique. Field tests and a measure of goodness of fit are used to express how well the models match the actual stem profiles. Using the Moroccan cedar (Cedrusatlantica Manetti) tree as an example, it is shown how the new nonparametric model can be compared with any classical parametric model. The results achieved demonstrate the advantages in using a nonparametric representation of stem profiles, which, moreover, is well suited to computer calculation constraints.