The effect of habitat-specific sampling on biological assessment of water quality using a predictive model

1. Detection of impairment in macroinvertebrate communities using rapid biological assessment depends on the ability to compare sites, with confidence that differences obtained result from water quality. However, collections from more than one habitat type may introduce variation that can potentially mask water quality differences among sites. Data were collected from the riffle, edge, pool-rock and macrophyte habitats at reference (minimally disturbed) and test (disturbed) stream sites throughout the Australian Capital Territory. The effect of habitat-specific sampling on predictive models for detecting impairment in macroinvertebrate communities was determined. Four models were used: riffle only, edge only, each habitat as an individual object, and all habitats sampled at a site considered as a composite sample. 2. Macroinvertebrates from individual habitats generally clustered into separate groups because collections from the same habitat at different sites were more similar than collections from different habitats within a site. Thus, in the habitats as individual objects model, the taxa predicted to occur at a test site may be an indication of habitat type rather than water quality. The outputs of the composite habitats and riffle and edge models were similar. However, the variable number of habitats included at each site in the composite model may confound the detection of biological impairment because of unequal sampling effort. The riffle and edge models were the most robust because they were less confounded by inter-habitat variation and were based on comparisons made between equivalent environmental units. 3. Comparison of observed/expected taxa ratios for test sites showed that each model could detect biological impairment, indicating considerable data redundancy was introduced by sampling several habitats. In particular, the pool-rock and macrophyte habitats contributed no information with regard to macroinvertebrate taxon occurrence or detection of biological impairment that could not be obtained from either the riffle or edge habitats within the study area.