Hilsenhoff's family-level index (FBI) combines information about the relative abundances of taxa and their tolerances to pollution. Versions of this index are used extensively in North America to assess water quality. When faced with constraints on time, money, or expertise, bioassessment practitioners have been tempted to calculate a version of the FBI with very coarse (e.g., order-level) taxonomy. Such an approach requires a degree of within-taxon averaging of tolerance values and raises questions about the degree to which accuracy is compromised and bias is introduced. Data from thousands of streams in Ontario (Canada) demonstrated that such tolerance-value averaging produces index scores and associated water-quality classifications that are not equivalent to those calculated with the standard family-level taxonomic precision. Two methods were used in an attempt to correct the order-level FBI scores to equivalence with the family-level index: (1) tolerance scores for the orders included in the calculation were calculated as abundance-weighted means of the scores of their component families, and (2) order-level FBI scores were estimated as predicted values from a polynomial regression of the two versions of the index. The use of abundance-weighted mean tolerance scores greatly improved the accuracy of the order-level index, and the regression-based correction reduced bias by equalizing the distribution of errors across the range of observed FBI values. Nonetheless, equivalence of scores was not demonstrated, and water quality was misclassified in 12 to 80% of cases. Practitioners are discouraged from the practice of tolerance-value averaging and are advised to adhere to the standard family-level FBI.
Keywords: Biotic index; Classification success; Equivalence test; Ontario; Streams; Taxonomic precision.