Fish-benthos correlations and effects on benthos that reflect significant effects on fish communities in southern Ontario streams

Abstract

This thesis attempts to determine the types of effects on benthic community composition that can be considered to have ecological consequence. The approach was based on the philosophy that changes in benthos are of little significant unless they coincide with unacceptable changes in fishery resources The derivation of critical benthic community effect sizes required four components. First, a definition of unacceptable change in a fishery resource. Second, selection of some aspect of fishery resources that could be considered important to prevent damage to. Third, a demonstration that benthic community composition was inherently related to relevant descriptors of the condition of a fishery. And finally, specific relationships between benthic community composition and the fishery descriptor, from which critical benthic effect sizes could be determined.

To define an unacceptable change in a fishery resource, I adopted the use of normal ranges which are inherent in most goal-setting forums. Historically, two problems with the use of normal ranges has been in the application of normal ranges to multivariate descriptors, and in the application of appropriate statistical tests for comparing "impact" locations against the normal range. I therefore provide an operational definition for the normal range that simplifies the use of normal ranges in hypothesis testing. I also demonstrate three statistical tests that can be used to determine when a point observation falls outside of the normal range. The consequences of using conventional two-sample contrasts in relation to one-sample non-central equivalence and interval tests are determined. For sites that are truly outside of the normal range of variation for reference locations, equivalence tests will lead to erroneous conclusions of no impact at most 5% of the time. Two-sample contrasts will lead to erroneous conclusions of no impact about 50% of the time with low sample sizes (i.e., 10-20 reference locations). In contrast to both the two-sample contrast and the equivalence test, interval tests fail to recognize sites as being impacted unless impacts are in excess of about 3 o from the reference population average (with a reference sample size of 20). Finally, the penalty for using the equivalence test is that it will fail up to 26% of sites that are truly members of the reference population. Practitioners contrasting non-random impacted locations against a set of reference locations should consider these characteristics of the various potential tests when deriving conclusions of impact.

Fish community data from 37 streams, variously affected by agriculture, urbanization and impoundments, were used to determine the statistical power of various descriptors of fish community composition. In general, single-pass estimates of biomass of all species, and using a multivariate approach to describing the community, provided more modelling and statistical power. In contrast, measuring only the abundances or biomasses of individual (preferred) species like brook trout, was a less powerful method (statistically) for characterizing the fishery of a stream, and was less useful for modelling purposes.

To demonstrate consistent associations between fish and benthos, I used three independent data sets in which fish and benthos were collected across stream size and temperature gradients. These data showed consistent and strong associations between stream fish and benthos community composition. The data also suggested that more detailed benthic taxonomy resulted in stronger associations between fish and benthos, and provided evidence that the fish-benthos association is strongly driven by coincident association of fish and benthos with environmental conditions.

Finally, critical benthic effect sizes were determined for 11 combinations of data sets to explore sources of variation in derived critical values. In general, benthic critical values that coincided with ecologically relevant effects on fish were less than the normal range of variation for benthic communities, but varied with factors such as the type of benthic sampling apparatus, the microhabitat benthos were collected from, and the nature of the study design (i.e., whether it incorporated regional-reference or site-specific reference locations). Although the fish-benthos correlations were significant regardless of the level of benthic identification, identifying benthos to species resulted in more accurate prediction of the condition of fish communities. In addition, the reliability of benthos was higher when the reference fish community was a brook trout community, or when the assessment was site specific. In both situations, changes in fish communities from reference to impacted conditions were abrupt.

The results from this study have obvious application to aquatic environmental assessments, particularly stream assessments in southern Ontario. Benthic criteria, such as those derived here, should be applied in association with relevant statistical procedures, primarily non-central interval or equivalence tests. These tests will provide exact probabilities that benthic community composition in impacted locations exceeds the benthic critical value.