Urbanization gradients in streams of Anchorage, Alaska: A comparison of multivariate and multimetric approaches to classification

The classification of stream sites into groups of differing habitat quality is central to the evaluation of water quality. In bioassessment approaches using macroinvertebrates, debate exists as to the merit of multivariate approaches compared to multimetric techniques for site classification. We analyze similarity coefficients between pairs of sites within classification groupings, compared to pairs of sites between groupings, as a test of the strength of different approaches to site classification. Streams within the Municipality of Anchorage in Alaska were used to investigate the effects of marked urbanization gradients on benthic communities without the influence of other disturbances in their watersheds (e.g. agriculture or timber harvest). In 1989, macroinvertebrates were collected at 27 stream sites in 13 different streams and then sites were classified into groups of apparent habitat quality in one of four ways; (1) an a-priori classification of the sites based on land use characteristics and known inputs of contaminants, (2) a TWINSPAN classification using abundances of taxa, (3) a multimetric approach using 5 metrics combined into a single index which was then trisected into three groups, and (4) a TWINSPAN classification using the metric data. Analyses of faunal similarities between and within habitat quality groups showed that the TWINSPAN classification using numerical data produced the classification with the highest overall strength, while the multimetric approaches resulted in the lowest strength of classification. However, all classification approaches identified essentially the same suite of sites in the 'poor' group, suggesting considerable robustness among bioassessment approaches in identifying sites most in need of monitoring and remediation.