A Two-Phase Sampling Strategy for Forest Inventories

Abstract

A two-phase sampling strategy is investigated in order to perform forest inventories on large areas for which aerial photos are available. In the first phase the study area is partitioned into N quadrats of equal size and a point is randomly selected within each quadrat. Subsequently, the N points are classified as wood or non-wood on the basis of the aerial photos. Then in the second phase a sample of n points out of N is selected according to probabilistic sampling. The selected points are visited in order to determine their actual classifications and to quantify the amount for an interest variable X in the circular plots of adequate radius centered at the n points. On the basis of the collected data, unbiased estimators of the wood coverage and the total of X over the whole study area are obtained. Moreover, the sampling variances of these estimators are derived together with their corresponding conservative estimators.