IAn evaluation of an inshore bottom trawl survey design for American lobster (Homarus americanus) using computer simulations

Abstract

This paper evaluates the performance of six possible sampling designs to estimate the population abundance index for American lobster using computer simulations. These designs include simple random sampling (SRS), systematic sampling (SYS) and stratified random sampling with four stratification schemes (i.e., based on region, depth, sediment and region × depth). For the stratified random design with region and depth being used for stratification, we evaluated the performances of different strategies for allocating sampling efforts. Simulations were implemented on the “true” populations which were estimated annually from 2002 to 2008 for both spring and fall based on a general additive model model developed in a separate study. Relative Estimation Error (REE), Relative Bias (RB) and design effect were used to measure the precision, accuracy and efficiency of mean estimation for different designs. On average, SYS tended to yield the most precise and efficient estimate of mean with specified sample size. However, its estimates tended to be biased and its performance varied with sample sizes and realizations of “true” population, thus changed with lobster distribution. Appropriate stratification, such as using depth to determine strata, significantly improved the precision and efficiency over SRS. Sediment, which is related to lobster distribution, was found to have little contribution to the improvement of the performance over SRS when it is used to determine strata. Also, allocating samples to each stratum based on variance or mean of previous year improved precision and efficiency. This study suggests that current design ( i.e., region-depth stratified design) used in the survey had stable performance across years and seasons.