Variability in Benthic Invertebrate Density Estimates from Stream Samples

The problem of variability in benthic density estimates from stream sampling was reassessed using data from a number of stream studies, including our own. When data were analyzed assuming a negative binomial distribution and a tolerated precision of 40% of the mean, it was apparent that earlier estimates of hundreds of samples required to estimate density were inaccurate. Based on total numbers of organisms, in most cases, six samples were sufficient for a precision of 40%. Notably, from three samples on, variability (as measured by the coefficient of variation) actually increased with increased number of samples. Examination of two long-term studies revealed that taking three samples yielded reliable estimates of benthic density with reasonable coefficients of variation. However, such sampling schemes yielded significantly lower numbers of species than when six or 12 samples were taken. We suggest that six replicates, each a composite of two samples, would be sufficient for stream studies, by providing a reliable estimate of total density and a fairly complete species list.