Gerald V. Sgro

Diatom-based Weighted-averaging Transfer Functions for Great Lakes Coastal Water Quality: Relationships to Watershed Characteristics

ABSTRACT In an effort to develop indicators for Great Lakes near-shore conditions, diatom-based transfer functions to infer water quality variables were developed from 155 samples collected from coastal Great Lakes wetlands, embayments and high-energy shoreline sites. Over 2,000 diatom taxa were identified, and 352 taxa were sufficiently abundant to include in transfer function development. Multivariate data exploration revealed strong responses of the diatom assemblages to stressor variables, including total phosphorus (TP). Spatial variables such as lake, latitude and longitude also had notable relationships with assemblage characteristics. A diatom inference transfer function for TP provided a robust reconstructive relationship (r2 = 0.67; RMSE = 0.28 log(μg/L); r2jackknife = 0.55; RMSEP = 0.33 log (μg/L)) that improved following the removal of 13 samples that had poor observed-inferred TP relationships (r2 = 0.75; RMSE = 0.22 log(μg/L); r2jackknife = 0.65; RMSEP = 0.26 log (μg/L)). Diatom-based transfer functions for other water quality variables, such as total nitrogen, chloride, and chlorophyll α also performed well. Measured and diatom-inferred water quality data were regressed against watershed characteristics (including gradients of agriculture, atmospheric deposition, and industrial facilities) to determine the relative strength of measured and diatom-inferred data to identify watershed stressor influences. With the exception of pH, diatom-inferred water quality variables were better predicted by watershed characteristics than were measured water quality variables. Because diatom communities are subject to the prevailing water quality in the Great Lakes coastal environment, it appears they can better integrate water quality information than snapshot measurements. These results strongly support the use of diatoms in Great Lakes coastal monitoring programs.

A Diatom Quality Index from a Diatom-Based Total Phosphorus Inference Model

A diatom quality index to assess diatom community impairment in the nearshore wetlands of the Laurentian Great Lakes was developed from a diatom-based total phosphorus (TP) weighted average inference model. The index is calculated with a weighted average equation using species optima standardized to a 1–10 scale and species tolerance standardized to a 1–3 scale. Multiple regression analysis revealed a moderate fit (R 2 = 0.63) between site scores of the selected index and GIS derived watershed characteristics (agriculture, soils, and industrial facilities). These index scores more closely fit watershed characteristics than the diatom inferred TP (R 2 = 0.59). In a regression tree analysis, soil permeability separated higher index scores from lower scores identifying this variable as an important interaction factor in the analysis. The diatom quality index can be a powerful tool for analyzing habitat quality in the Great Lakes and can communicate the link between quantifiable diatom assemblage response with watershed-level disturbance.

COMPARISON OF SIMPLE AND MULTIMETRIC DIATOM-BASED INDICES FOR GREAT LAKES COASTLINE DISTURBANCE(1).

Because diatom communities are subject to the prevailing water quality in the Great Lakes coastal environment, diatom‐based indices can be used to support coastal‐monitoring programs and paleoecological studies. Diatom samples were collected from Great Lakes coastal wetlands, embayments, and high‐energy sites (155 sites), and assemblages were characterized to the species level. We defined 42 metrics on the basis of autecological and functional properties of species assemblages, including species diversity, motile species, planktonic species, proportion dominant taxon, taxonomic metrics (e.g., proportion Stephanodiscoid taxa), and diatom‐inferred (DI) water quality (e.g., DI chloride [Cl]). Redundant metrics were eliminated, and a diatom‐based multimetric index (MMDI) to infer coastline disturbance was developed. Anthropogenic stresses in adjacent coastal watersheds were characterized using geographic information system (GIS) data related to agricultural and urban land cover and atmospheric deposition. Fourteen independent diatom metrics had significant regressions with watershed stressor data; these metrics were selected for inclusion in the MMDI. The final MMDI was developed as the weighted sum of the selected metric scores with weights based on a metric’s ability to reflect anthropogenic stressors in the adjacent watersheds. Despite careful development of the multimetric approach, verification using a test set of sites indicated that the MMDI was not able to predict watershed stressors better than some of the component metrics. From this investigation, it was determined that simpler, more traditional diatom‐based metrics (e.g., DI Cl, proportion Cl‐tolerant species, and DI total phosphorus [TP]) provide superior prediction of overall stressor influence at coastal locales.

Coastal Geomorphic and Lake Variability in the Laurentian Great Lakes: Implications for a Diatom-based Monitoring Tool

ABSTRACT In an evaluation of diatoms as indicators of human disturbance in coastal ecosystems of the Laurentian Great Lakes, we characterized assemblage specificity to lake and habitat type to identify non-anthropogenic factors influencing indicator models. Surface sediment assemblages and environmental variables were collected along the U.S. coastline at 191 sample sites, which were classified by lake and geomorphic type: high-energy (HE), embayment (EB), coastal wetland (CW), riverine wetland (RW), protected wetland (PW), and open water (OP). Diatom inferred (DI) total phosphorus (TP) transfer functions (models) were developed for each lake and geomorphic type. Robust models included: the overall model (RMSEP; r2jack = 0.65; RMSEP = 0.005), Lake Superior (r2jack = 0.73; RMSEP = 0.003), Lake Ontario (r2jack = 0.73; RMSEP = 0.007), PW (r2jack = 0.64; RMSEP = 0.003), and EB (r2jack = 0.64;RMSEP = 0.007). Weaker models, indicating poorer diatom-TP relationships, included: RW (r2jack =0.03; RMSEP = 0.005), OP (r2jack = 0.15; RMSEP = 0.059), and Lake Michigan (r2jack = 0.38; RMSEP =0.006). DI TP data were regressed against landscape characteristics to quantify the relationships to adjacent watershed stressors. RW data were further scrutinized as a case study to investigate the suitability of diatom-based approaches in systems with poor diatom-TP relationships. Despite poor performance of the RW model, DI phosphorus data for riverine wetlands, derived from the overall model, were strongly related to watershed characteristics (r2 = 0.61), indicating the overall models ability to integrate stressors from the surrounding watershed in areas where measured phosphorus did not adequately characterize prevailing conditions. This study confirms that physical properties (e.g., lake or habitat type) can influence indicator models; however, weaknesses may be overcome by robust calibration techniques.

DIATOM SPECIES COMPOSITION AND ECOLOGY OF THE ANIMAS RIVER WATERSHED, COLORADO, USA

Abstract The diatom flora of selected sites in the Animas River Watershed, San Juan County, Colorado, was studied. Eighty diatom taxa were identified from 10 sites: 8 sites on the Animas River and 1 site each on the Cement and Cascade tributaries. The sample diatom abundance was dominated by Achnanthidium minutissimum, Encyonema silesiacum, Aulacoseira distans, Hannaea arcus, and Diatoma mesodon. The presence of teratologic specimens of Fragilaria and Achnanthidium in the samples indicated the possibility of metals contamination. Diatom diversity was low and Lange-Bertalot pollution index scores indicated little organic pollution evidenced from diatom composition. There was evidence that diatom composition at the sites was differentially affected by pH and possibly by the concentrations of Zn alone or in combination with Cd, Cu, and Fe.

Assessing the condition of the Missouri, Ohio, and Upper Mississippi rivers (USA) using diatom-based indicators

Diatom-based indicators were developed to assess environmental conditions in the Missouri, Ohio, and Upper Mississippi rivers. Disturbance gradients, comprising the first two principal components derived from a suite of stressor variables, included a trophic gradient (Trophic) and a gradient reflecting agriculture and other development activities (Ag/Dev). Diatom-based indicators were developed by creating models using weighted average calibration and regression-based transfer functions to relate planktonic and periphytic diatom species assemblages to each disturbance gradient. The most predictive disturbance models combined phytoplankton and periphyton assemblages into a single bioindicator model (observed versus inferred: Trophic

Evaluation of geographic, geomorphic and human influences on Great Lakes wetland indicators: A multi-assemblage approach

r_{\text{boot}}^{2} = 0. 5 6

Climate warming and changes in Cyclotella sensu lato in the Laurentian Great Lakes

; Ag/Dev