A performance comparison of metric scoring methods for a multimetric index for Mid-Atlantic highlands streams

When biological metrics are combined into a multimetric index for bioassessment purposes, individual metrics must be scored as unitless numbers to be combined into a single index value. Among different multimetric indices, methods of scoring metrics may vary widely in the type of scaling used and the way in which metric expectations are established. These differences among scoring methods may influence the performance characteristics of the final index that is created by summing individual metric scores. The Macroinvertebrate Biotic Integrity Index (MBII), a multimetric index, was developed previously for first through third order streams in the Mid-Atlantic highlands of the United States. In this study, six metric scoring methods were evaluated for the MBII using measures related to site condition and index variability, including the degree of overlap between impaired and reference distributions, relationships to a stressor gradient, within-sample index variability, temporal variability, and the minimum detectable difference. Measures of index variability were affected to a greater degree than those of index responsiveness by both the type of scaling (discrete or continuous) and the method of setting expectations. A scoring method using continuous scaling and setting metric expectations using the 95th percentile of the entire distribution of sites performed the best overall for the MBII. These results showed that the method of scoring metrics affects the properties of the final index, particularly variability, and should be examined in developing a multimetric index because these properties can affect the number of condition classes (e.g., unimpaired, impaired) an index can distinguish.