Abstract
Marine organism survey designs can take a wide range of forms, and an important decision is how to spread limited sampling effort over space. This study compared systematic, simple random, stratified random, and a sophisticated spatially balanced technique called generalized random tessellation stratified (GRTS), sampling designs. The comparison used data from an established optical survey of Atlantic sea scallop (Placopecten magellanicus) populations in the northwest Atlantic Ocean. Two field studies showed limited differences in the performance of simple random and systematic sampling, despite considerable differences in sample sizes. Simulations expanded this study considerably by analyzing six sampling techniques over a greater range of scallop populations. The six techniques were simple random sampling, three variants of stratified random sampling that differed in their methods for assigning stations to strata, systematic sampling, and GRTS sampling. The performance of each technique was assessed through the mean scallop density estimate and the variance of this mean. All techniques were accurate throughout, although all variants of the stratified random technique were inaccurate and imprecise in one small area with low sample size. The stratified random and simple random techniques were among the best at closely estimating the true precision. The GRTS and systematic techniques were highly accurate but weaker at estimating true precision. These results highlight that all techniques have tradeoffs and show that optimal statistical survey design is highly specific to the spatial nature of the species being targeted.