Comparison of ground sampling methods for estimating canopy cover

Abstract

Knowledge of the canopy structure is essential to improving our understanding of forest structure. While numerous sampling techniques have been developed to estimate attributes of the forest canopy, these require either additional measurements or a sampling design and measurement techniques that differ substantially from the ones that are used to estimate more traditional forest attributes, such as basal area, number of stems, or volume. The root of the problem is that the sample element for a design that estimates canopy attributes is the tree crown, whereas the sample element is the bole for a design that estimates an attribute such as basal area. For example, if a fixed-area plot is used to estimate basal area, canopy cover cannot be estimated using the same design because a portion of the plot invariably is covered by the crowns of trees whose boles lie outside the plot boundary and would not be included in the sample under the standard sampling design. In this study, a technique called "morphing" is used to model the trees outside the plot boundary. For the purpose of comparison, the morphing technique is used to estimate canopy cover using data from a circular fixed-area plot, and this technique is compared with both dot count and line intersect sampling using a simulation study and two small forest populations. For the study, the populations were sampled using circular fixed-area plots with radii ranging from p = 3.05-6.10 m (10-20 ft) and line lengths ranging from L = 3.05-22.9 m (10-75 ft). For both populations, the bias of the canopy cover estimator derived from the morphing technique was negligible. The estimator based on line intersect sampling is design-unbiased, but it generally had a much larger variance than the one based on the morphing technique. The dot count method consistently had the highest variance.