COMPARISON OF THE PERFORMANCE OF LINE SAMPLING TO OTHER FORMS OF CLUSTER SAMPLING

Line sampling is a widely applied method in field surveys to assess spatially distributed variables. The values of the variable of interest are recorded along line-shaped survey elements. Line sampling can be viewed as a special case of cluster sampling with fixed cluster form where the number of elements in the cluster is large and the distance between the elements small. In this study the precision of estimators for the following survey elements is examined: line segment, L-shape, cross, circle, square and equilateral triangle, i.e. we estimate lines of limited length c (in contrast to transects which cross the entire survey region). The precision of a line sample depends on the probability density function (PDF) of interpoint distances on the line. The PDFs are given for the six survey elements listed above. In an example with a simple covariance function cov(d)=e(-d) the variance for given line length c increases in the order: line segment, L-shape, circle, square, cross, equilateral triangle. A comparison of line sampling and cluster sampling with the same form and size of survey element shows that, except for the circle, clusters of only a few cluster points give more precise results than line sampling when small survey elements are used, assuming the given covariance function is correct.