stressors - GLEI Pub

Diatom-based Integrative Water Quality Model

Training sets have become the mainstay of modeling methods to calibrate diatom indicators to water quality variables of interest. Typically the present-day distributions of diatoms (or any of several indicator organisms) are calibrated across the gradient of a selected environmental parameter. These models use a species-based weighted average (WA) approach, and have consistently provided a valuable tool to infer past conditions from sedimentary records (see associated GLEI-diatom indicator). If we are to continue refining these models we need to focus on better characterization of species-environmental relationships. Previous applications have typically been limited to single water quality variables, such as phosphorus, nitrogen, pH, or salinity, because these variables are often of interest to water quality managers. Inevitably however, weaknesses in these models are associated with a loss of ecological information, resulting from the inability of a single water quality variable to account for a large amount of variation in the diatom assemblages. This model weakness is a result of the fact that a specific variable of interest captures only a fraction of the variation in the diatom data that may be explained by other measured variables. For instance, it is well known that the measured phosphorus gradient from a suite of sites is typically correlated with nitrogen, water clarity, organic compounds and suspended solids, all of which can have water quality implications. A diatom model based on an integrated water quality gradient provides a means to reconstruct general environmental quality at a locale.

A diatom-based model to infer phosphorus and other water quality parameters in Great Lakes coastlines was developed as part of the GLEI project. This evaluation used the same set of chemical and diatom data to instead derive an integrated water quality (WQ) model. PCA of all chemical variables identified the major environmental gradient that would be considered ranging from “high” (i.e., low-nutrient, clear-water sites) to “low” (i.e., high-nutrient, high Cl, turbid sites) water quality (Fig. 1). Characterizing this dominant gradient allowed the suite of WQ data to be summed up in a comprehensive WQ variable, which was then related to diatom assemblages.

Watershed stressor data (including agricultural intensity and urban development) were regressed against (A) measured and (B) diatom-inferred WQ data using multiple linear regression. For both models, DI parameters were better correlated to watershed characteristics than measured parameters (Fig. 2). This finding endorses the use of diatoms in monitoring programs as well as their frequent applications to paleoecology; the diatoms better integrate aquatic conditions, which are subject to watershed influences, and so provide more reliable inferences of the prevailing condition than spot measurements of water chemistry. The combined WQ variable is better correlated to watershed characteristics than measured TP, a probable result of the fact that WQ is derived from several chemical parameters, and so should better reflect overall chemical condition than a single nutrient. Combination of these variables into a single WQ variable offers a notable advantage to characterizing watershed-measured water quality relationships.

Summarizing water chemistry variables into a comprehensive index of water quality appears to proffer some advantages over using specific environmental variables in diatom models. A disadvantage of using a WQ index this way would be some loss of information for water quality managers who may be interested in specific variables, such as phosphorus or water clarity. However, an integrated measure of water quality is a useful means to provide a more holistic view of limnological conditions at a site.