Most similar neighbour-based stand variable estimation for use in inventory by compartments in Finland

Open Access
Publisher Website
Google Scholar
Abstract
Non‐parametric regression was used to predict basal‐area diameter distribution and the stand volume of Scots pine ( Pinus sylvestris L.) in Finland. The regression is based on weighted averages of most similar neighbours (MSN) of a stand. This is a special case of the k ‐nearest‐neighbour method, in which the similarity is measured using canonical correlations. The results were compared with those obtained with percentile‐based basal‐area diameter distribution models which are currently used in Finland. When constructing the MSN models, stand mean characteristics were used as independent variables, and variables describing the shape of the diameter distribution and stand density as dependent variables. Since the relationships are mainly non‐linear, and the weights are based on linear correlations, the use of second powers of independent variables improved the results. The basic MSN model was constructed for the whole study area, but also regional and local models were tested. Regional models were tested for the vegetation zones in Finland. Local models were tested by applying two modified MSN regression methods for similarity‐based neighbourhoods. The results obtained indicated that the accuracy of the MSN method is comparable to percentile‐based basal‐area diameter distribution models in the case of stand volume. However, the description of stand structure, i.e. the number of stems could be improved by using MSN regression. The vegetation zones used seemed to be too small for reference areas. However, it was possible to get more accurate local results if the neighbourhood used was described and selected effectively.

This publication has 0 references indexed in Scilit: