Estimating top height with variable plot sizes

Abstract

Conventional top height estimates are biased if the area of the sample plot differs from that on which the definition is based. Sources of bias include a sampling selection effect and spatial autocorrelation. The problem was studied in relation to the use of data sets with varying spatial detail for modelling Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) plantation growth. Improved top height estimators, developed taking into account the selection effect, eliminated the bias. Bias was reduced, but not eliminated completely, when the estimators were tested using more highly autocorrelated eucalypt data.