Cluster plot optimization for a large area forest resource inventory in Korea

Abstract

Cluster plots are a common plot design and have a very long history in large area forest monitoring. The aim of this study is to assist in determining a suitable cluster plot design for forest resources assessment in South Korea, to analyse both statistical and cost criteria. Twenty-five pilot cluster plots were measured, which consisted of 10 subplots arranged in a complex pattern cluster allowing the simulation of various cluster configurations. In our approaches to design optimization of the cluster plots, we took three statistical characteristics into account: the intra-cluster correlation; the geometric arrangements of the subplots; and standard error as an indicator of precision. The inventory costs were also investigated in terms of time consumption which are the basis of our analysis of economic efficiency. In the pilot cluster plot dataset, the covariance functions for the target attributes (growing stock, basal area, and number of trees) decreased with increasing distance up to about a distance of 100 m: a subplot distance of 87 m was found to be an adequate minimum distance to keep intra-cluster correlation at a low level. In our study, the intra-cluster correlation was influenced more by the spatial arrangement than by cluster size. For clusters of four subplots, a modified triangular cluster gave the best results for all target attributes. The cluster size affects the precision, as well as the total working time. From the results of both statistical and cost analysis, a modified triangular cluster of four subplots was found to be the most efficient cluster plot design for the forest conditions in the test area.