Abstract
Robust inferences of environmental condition come from bioindicators that have strong relationships with stressors and are minimally confounded by extraneous environmental variables. These indicator properties are generally assumed for assemblage-based indicators such as diatom transfer functions that use species abundance data to infer environmental variables. However, failure of assemblage approaches necessitates the interpretation of individual dominant taxa when making environmental inferences. To determine whether diatom species from Laurentian Great Lakes sediment cores have the potential to provide unambiguous inferences of anthropogenic stress, we evaluated fossil diatom abundance against a suite of historical environmental gradients: human population, agriculture, mining, atmospheric nutrient deposition, atmospheric temperature and ice cover. Several diatom species, such as Stephanodiscus parvus, had reliable relationships with anthropogenic stress such as human population. However, many species had little or no indicator value or had confusing relationships with multiple environmental variables, suggesting one should be careful when using those species to infer stress in the Great Lakes. Recommendations for future approaches to refining diatom indicators are discussed, including accounting for the effects of broad species geographic distributions to minimize region-specific responses that can weaken indicator power.