Effect of sampling protocol and volunteer bias when sampling for macroinvertebrates

Abstract

We evaluated the efficacy of different field sampling approaches for volunteers sampling macroinvertebrates in low-gradient streams.We used a series of analytical metrics to compare results using the Environmental Protection Agency (EPA) multihabitat, the Minnesota Pollution Control Agency multihabitat, and EPA single-habitat sampling protocols. We also investigated the effect of 2 scenarios in which volunteers fail to follow (and potentially bias) the widely used EPA multihabitat protocol by including either more snag and vegetated banks or more run and riffle habitat than prescribed by the protocol. We collected jab samples from cobble, snags, vegetated banks, submerged macrophytes, and sand in 4 contiguous 125-m reaches in an Anoka sand-plain stream in Minnesota. We identified up to 100 macroinvertebrates in each jab sample to family. We subjected a parent population of 40 jab samples/reach to a bootstrap analysis to sample and create metric or index scores 100 times without replacement for each of the 3 volunteer sampling methods and 2 biased scenarios. The EPA multihabitat protocol and the biased scenario in which woody debris and bank vegetation were oversampled yielded the highest diversity of organisms, whereas the biased scenario in which riffle and run habitats were oversampled yielded the lowest diversity. The EPA multihabitat protocol used correctly was more likely to indicate ‘‘good’’ water quality (on the basis of the EPA muddy-bottom narrative assessment tool designed for volunteers) than either biased sampling scenario. This result illustrates that poor field methods could result in underestimation of water quality.