The diversity and environmental sensitivity of lotic invertebrates make them useful indicators of stream conditions; consequently macroinvertebrates have become increasingly popular in regional comparisons of streams (Hilsehoff 1988; Kerans and Kerr 1994; Rosenberg and Resh 1993). However, spatial and temporal variability in abundance and diversity among individual collections (at the reach scale) can make detection of differences among streams difficult (Norris and Georges 1993) necessitating some level of replication or physical compositing of samples. A dilemma we often face is how to determine the adequate number of samples. One approach has been to adjust the number of replicates to yield the desired level of precision (Elliott, 1977; Resh 1979); a criterion of ñ40% mean total abundances has been suggested as a practical goal (Resh and McElravy 1993; Pinder el al 1987). Another approach has been to evaluate the stability of the mean with increasing sampling effort (Elliott, 1977). These methods are limited by relying heavily on numerical abundance, with little regard for other assemblage attributes.
Our approach for assessing sampling adequacy was to consider a number of broadly representative metrics, and to identify appropriate levels of sampling effort to achieve stable values of these assemblage measures, somewhat analogous to one method suggested in Elliott (1977). Biomonitoring programs often use descriptors (often called metrics) designed to represent three basic assemblage characteristics: numerical abundance (total numbers of individuals), measures of diversity (taxa richness, diversity indices, and measures of evenness or dominance), and proportional abundances (proportions of total numbers including groups or taxa of special interest such as taxonomic order, dominant taxa, functional feeding groups or pollution-tolerant taxa). Some of these measures overlap, particularly those based on proportions that are all dependent on the same denominator, e.g. total abundance. We chose five metrics that represent different aspects of assemblage components: total abundance, taxa richness, Shannon diversity, proportional relative abundance of Ephemeroptera, Trichoptera and Plecoptera (EPT), and percent dominance of the most dominant taxon.
In the article associated with this study, our focus was on the adequacy of field macroinvertebrate sampling to discern differences among stream reaches within ecoregions. We greatly oversampled each stream reach, examining the incremental changes in metric values resulting from increases in sampling effort within each stream reach. This intense, hierarchical sampling design allowed us to quantitatively compare the relative magnitudes of macroinvertebrate assemblage metric variances across the range of spatial scales from within-habitat (eg. within a riffle) up to within-region (ie, comparing ecoregions).
Site Selection : Two hand-picked stream sites (Lookout and Mack Creeks) in the H.J. Andrews Experimental Forest/Cascade Mt. Long-term Ecological Research area were also sampled to allow comparison to the large available database at these sites. Mack Creek is a small first order stream, Lookout Creek is a fairly large third order stream. The collection at these streams were added to 16 streams (9 Cascade and 7 Valley) that were sampled for macroinvertebrate assemblages as part of a larger study.
Field Sampling and Counting: Macroinvertebrate field sampling took place in late September, 1992. From the random start point on each stream, 14 study transects (cross-sections) were marked off upstream at 10 m intervals in small streams, and 25 m intervals in large streams. Three, one ft2 Surber samples, one on the right side, center and left were taken within each transect (Fig. 2). At least 7 transects of fast water habitat (riffles, rapids or cascades) and 7 transects of slow-water habitat (pools or glides) were sampled on each stream. If 7 of each habitat were not encountered after 14 transects, the crew continued upstream sampling at the same transect intervals in only the under-represented habitat until 7 were sampled. As a result, total number of transects varied from 14 to 20 per stream. Each macroinvertebrate sample was filtered through a 500 micro-meter soil sieve at stream side. Samples were placed in whirlbags and preserved in 70% ethanol. Stream habitat at each Surber sample location was qualitatively classified as either a pool, glide, riffle, rapid, or cascade.
