Estimation of relative efficiency of adaptive cluster vs traditional sampling designs applied to arrival of sharks

TL;DR

This paper compares adaptive cluster sampling (ACS) to simple random sampling (SRS) for estimating shark arrivals, deriving conditions under which ACS is more efficient, especially in clustered and sparse populations, with simulation insights.

Contribution

It introduces a variance ratio inequality to decide when ACS outperforms SRS, providing a practical decision tool for sampling design in ecological studies.

Findings

ACS reduces variance in clustered populations.

Rare clusters favor ACS over SRS.

Simulation confirms conditions for ACS efficiency.

Abstract

Adaptive Cluster Sampling (ACS) is introduced as a technique to use when "natural" groupings are evident in a spatially distributed population, especially sparsely distributed populations. An ACS sampling design will allow efficient allocation of survey manpower with more effective decision rules for where/when to allocate those resources. In particular, given a clustered distribution, ACS would result in a lower variance of the mean estimator than Simple Random Sampling (SRS). This paper derives a linear inequality based on a ratio of variances and the SRS sample size to determine the conditions under which ACS would be a more appropriate sampling strategy. This inequality could be used to make preliminary decisions on the potential benefits of pursuing an ACS design over an SRS design. The relationship between relative efficiency of ACS over SRS is discussed for (1) variable conditions for neighborhood expansion (Conditions to Adpat) (2) the degree of clustering and (3) the percent of hits in a particular neighborhood. Simulation results demonstrate how 'rare' clusters should be for ACS to be a better design than SRS. The relationships that emerge from the simulation provide insights that were not apparent in the analysis. The results are important in the estimation of thresher shark landings along the California coast.