We compared 2 sampling strategies designed to estimate abundance of double-crested cormorants (Phalacrocorax auritus, hereafter cormorants) on aquaculture ponds in western Mississippi. Cormorants are a major predator of cultured channel catfish (Ictalurus punctatus) in this region; thus, estimating cormorant abundance is needed to better determine their economic impact. We independently designed a species-specific survey (i.e., cluster sampling) and a general survey (i.e., transect sampling) based on robust probability sampling theory to estimate abundance of this target population. During winters 2002–2003 and 2003–2004, we conducted 8 pairs of surveys and compared estimates of cormorant abundance and associated precision using conventional paired t-tests and complimentary equivalency tests. Abundance estimates from sampling methods did not differ given a minimum important effect size of 1,420 individuals. Precision of estimates for both survey protocols was poor (the coefficient of variation [CV] was 39.5% for cluster samples and 45% for transect samples), and we were unable to definitively conclude if precision was similar between sampling methods (due to low sample size and high variability). We found sample sizes must increase 222% for cluster sampling and 538% for transect sampling to detect a 15% change in abundance on average. Thus, neither method met our goals of detecting a given effect size at a desired level of precision. We recommend investigating additional sampling designs that may provide precise estimates of abundance more efficiently than the methods compared in this study.
Available at: http://works.bepress.com/stephen_dinsmore/24/