Sampling variability and the design of bacterial abundance and production studies in aquatic environments

Abstract

Published data for aquatic bacterial abundance and production in benthic and planktonic environments were collected from the literature to describe relationships between sample means and variances, to explore the factors that affect these relationships, and to estimate the number of samples needed to detect specified differences among means with adequate power. Between 75 and 94% of sample log10(variance) was explained by log10(mean) for both bacterial abundance and production. Differences in mean-variance relationships of bacterial abundance and production due to habitat (river, lake, marine), quantification method, and experimental manipulation (planktonic bacteria) or substrate type (benthic bacteria) were negligible (less than 11% of residual variance from regressions explained). Between 12 and 69 replicates are necessary to detect a 20% difference in means for bacterial abundance and production with a power of 80%. Given the median rate of replication of 3 to 4, the majority of published studies reviewed here are, at best, able to detect differences in means of 50% (planktonic bacterial abundance) or 100% (planktonic production and benthic abundance and production) with 80% power. If effect sizes less than these values are deemed biologically meaningful, then future studies will have to increase sampling effort to enable detection of such differences.