Diatom-based models to infer nutrient concentrations are proven robust indicators, but evidence suggests that in the future these models will be little improved by using larger training sets. I present a simple means to summarize the water quality (WQ) data from a suite of coastal Great Lakes locations and develop a diatom-based WQ model using standard weighted-averaging methods. A one-dimensional WQ index was derived by summarizing measured environmental data (nutrients, pigments, solids) using dimension-reducing ordination and calculating the primary WQ gradient of interest. Evaluations of weighted-averaging diatom model predictions (WQ index model: r²_jackknife = 0.62, RMSEP = 1.32) indicate that the model has reconstructive power similar to a comparative model for total phosphorus concentrations (TP model: r²_jackknife = 0.65, RMSEP = 0.26 log[μg/L 1]), but that predictive bias was lower for the WQ model. Also, inferred WQ index data had a higher correlation to adjacent watershed characteristics than inferred TP data. We attribute this to the ability of an integrated WQ index to better characterize the overall quality of a site than a single nutrient variable such as phosphorus. The diatom-based WQ model may be advantageous for management where it is necessary to provide a summary inference of water quality condition at a coastal locale.