Improving the performance of ecological indices by balancing reference site quality and representativeness.

Reference site networks should consist of minimally disturbed sites that collectively characterize the ranges of natural settings within a region. Compromise between reference-quality and representativeness is required. We evaluated how tradeoffs between reference-quality and regional representativeness affected applicability, performance, and interpretation of multi-metric (MMI) and Observed/Expected (O/E) indices developed for streams in eastern China. We emphasized reference-quality by applying the most-stringent objective criteria and expert-judgment to select reference-group1 (G1). We emphasized representativeness by applying the least-stringent criteria to select reference-group2 (G2) sites from different strata based on watershed size. We balanced reference-quality and representativeness in G3 by applying intermediate stringent criteria from each watershed size stratum used previously. Increasing representativeness using G2 improved index applicability to almost more than twice the number of test sites than when reference-quality maximized using G1. Bias in O/E index was almost eliminated only when reference-quality and representativeness balanced using G3. MMIs developed when reference-quality maximized using G1 eliminated all bias and had the highest precision. High-quality reference with limited representativeness affected the metrics selected for inclusion in MMIs and restricted the sites to which both types of indices could be applied. A balanced approach worked best in this instance and similar approaches should be tested in other regions. [ABSTRACT FROM AUTHOR]