Optimal Survey Design: Lessons from a Stratified Random Sample of Macrobenthos

Abstract

A stratified random sample of macrobenthos in Western Port (Victoria, Australia) provided adequate data to take an a posteriori look at the efficiency of various random survey designs in terms of their ability to provide precise estimates of the mean number of individuals per taxon, i.e. mean estimates with the smallest possible variance. Emphasis was placed on the efficiency of the stratified simple random sampling design. The analyses showed that bay-wide estimates resulting from various stratified designs would not have been substantially more precise than those from simple random sampling. This conclusion was not influenced by the allocation strategy used in the stratified design. On the other hand, substantial gains in precision could have been made, for the same total number of grab samples, by increasing the number of stations at the expense of the number of grab samples per station. Our data suggested that the optimal number of grab samples per station is only one, contrary to the common practice of multiple grab samples per station. Key words: stratified, random, survey, optimal, macrobenthos, sample, Western Port