Adaptive Sampling Design

In adaptive sampling, information gained during the sampling process is used to modify, or adapt, how the subsequent sample units are selected. Traditionally, the selection procedure is defined prior to sampling. For example, a sample scheme may be to select n units at random from the population, as in simple random sampling. In adaptive sampling, the select procedure may change during the survey.

In biology, plant or animal populations are often spatially aggregated, and when the population is rare, sampling can be challenging. Adaptive sampling can be used in this situation. An initial sample of plants may be undertaken within quadrats placed randomly in the area. Additional sampling is undertaken around the quadrats where plants were found. The final sample combines information from the initial sample and from the additional sampling. In this adaptive design, survey effort was concentrated in the localities where plants were found. The information from the initial sample was used to “adapt” the sample because additional sampling was directed to the localities where plants were known to be. In this regard, adaptive sampling is considered more informative than traditional sampling.

There are many forms of adaptive sampling; the one described above could be called adaptive cluster sampling. Another adaptive design is to adaptively allocate extra survey effort in stratified sampling. In traditional stratified sampling, the population is divided into homogeneous groups or regions, called strata. Survey effort is allocated among strata according to some criterion, usually estimates of the within stratum variance or mean. If there is no information on these strata statistics or if the information is poor, adaptive stratified sampling can be used. After an initial survey of the strata, estimates of stratum variance or mean are used to decide on allocation of additional survey effort. Usually this additional survey effort is allocated to the strata with the highest variances. Information on the strata gained during the survey is used to adapt the final allocation of survey effort.