Jump to Main Content
An algal model for predicting attainment of tiered biological criteria of Maine's streams and rivers
- Danielson, Thomas J., Loftin, Cynthia S., Tsomides, Leonidas, DiFranco, Jeanne L., Connors, Beth, Courtemanch, David L., Drummond, Francis, Davies, Susan P.
- Freshwater science 2012 v.31 no.2 pp. 318-340
- Algae, biological assessment, biologists, discriminant analysis, fish, habitats, macroinvertebrates, model validation, models, objectives, prediction, professionals, rivers, sampling, streams, surface water, water quality, water quality standards, Maine
- State water-quality professionals developing new biological assessment methods often have difficulty relating assessment results to narrative criteria in water-quality standards. An alternative to selecting index thresholds arbitrarily is to include the Biological Condition Gradient (BCG) in the development of the assessment method. The BCG describes tiers of biological community condition to help identify and communicate the position of a water body along a gradient of water quality ranging from natural to degraded. Although originally developed for fish and macroinvertebrate communities of streams and rivers, the BCG is easily adapted to other habitats and taxonomic groups. We developed a discriminant analysis model with stream algal data to predict attainment of tiered aquatic-life uses in Maine's water-quality standards. We modified the BCG framework for Maine stream algae, related the BCG tiers to Maine's tiered aquatic-life uses, and identified appropriate algal metrics for describing BCG tiers. Using a modified Delphi method, 5 aquatic biologists independently evaluated algal community metrics for 230 samples from streams and rivers across the state and assigned a BCG tier (1–6) and Maine water quality class (AA/A, B, C, nonattainment of any class) to each sample. We used minimally disturbed reference sites to approximate natural conditions (Tier 1). Biologist class assignments were unanimous for 53% of samples, and 42% of samples differed by 1 class. The biologists debated and developed consensus class assignments. A linear discriminant model built to replicate a priori class assignments correctly classified 95% of 150 samples in the model training set and 91% of 80 samples in the model validation set. Locally derived metrics based on BCG taxon tolerance groupings (e.g., sensitive, intermediate, tolerant) were more effective than were metrics developed in other regions. Adding the algal discriminant model to Maine's existing macroinvertebrate discriminant model will broaden detection of biological impairment and further diagnose sources of impairment. The algal discriminant model is specific to Maine, but our approach of explicitly tying an assessment tool to tiered aquatic-life goals is widely transferrable to other regions, taxonomic groups, and waterbody types.