Abstract
Adaptive cluster sampling has been suggested as an efficient sampling technique for rare and patchy populations. In this thesis the question asked is: how useful is the design for ecologists? Simulated and real ecological populations were used to compare the variance of adaptive cluster sampling to more traditional sampling techniques such as simple random sampling. It was found that the adaptive technique is efficient for very patchy populations, but for a population that is not patchy, it can be highly inefficient compared to simple random sampling. The way the sample is designed is critical. The relative variance of the sample design is dependent on the size of the initial sample, the quadrat size and the adaptive selection procedure - the value of the condition used to decide when to adaptively select adjacent quadrats for sampling, and the definition of the neighbourhood to sample adjacent quadrats. Adaptive cluster sampling is sufficiently "adaptive" to allocate sampling effort to areas of high density and there may be little gain in stratifying a study area (other than for logistical reasons).
In adaptive cluster sampling, the size of the initial sample is set prior to sampling but the size of the final sample is unknown until sampling is complete and can be quite large for some populations. Restricted adaptive cluster sampling is suggested where a limit is set prior to sampling which restricts the size of the final sample. In this modified design the initial sample size is unknown.