The Quantification of Stream Drift

Abstract

Although the purpose of many drift studies is to describe quantitatively the abundance of drifting invertebrates and make comparisons between seasons or sites, almost no investigations have employed replicate sampling. We analyzed drift collections from a Rocky Mountain stream in order to investigate the variability of drift sampling. The data were normalized and the variances stabilized for each taxon examined by data transformation. The fourth root transformation was favored for five taxa and the logarithmic transformation for three. Using the 95% confidence limits on 24-h drift density for an abundant mayfly (Baetis bicaudatus), we found that six to seven replicates are required to obtain 95% CL ± 50% of the mean. Drift sampling appears to require fewer replicates than benthic sampling for comparable precision. Investigators may fail to replicate drift samples because they elect to sample frequently over 24 h in order to quantify the diel periodicity of drift. However, when comparison between sites or dates is the principal goal, we recommend that the effort normally put into frequent sampling over 24 h be invested instead in replicated sampling just after dark, when drift normally is greatest. When we regressed drift from the first night sample against total drift from the remainder of the 24-h period, 60–90% of the variation in the latter was predicted from the single nighttime sample. Thus, little information appears to be lost by this recommended procedure.