A synoptic assessment of water quality in two Great Lakes connecting channels

Abstract

We conducted a probabilistic water quality assessment of two Great Lakes connecting channels, the St. Marys River, and the Lake Huron-Lake Erie Corridor (HEC) in 2014–2015. We compared the condition of the channels to each other and to the up- and down-river Great Lakes with data from an assessment of the Great Lakes nearshore conducted in 2015. We assessed the condition of each channel as good, fair, or poor by applying the most protective water quality thresholds for the down-channel lake. Condition in the St. Marys River rated mostly fair for total phosphorus (TP) and mostly good for chlorophyll a, and area-weighted mean concentrations were intermediate to nearshore Lake Superior and Lake Huron. A large proportion of the area of the St. Marys River was in poor condition for water clarity based on Secchi depth; while nearshore Lakes Superior and Huron were mostly in good condition for water clarity. Area-weighted mean concentrations of TP and chlorophyll a in the HEC were more like nearshore Lake Huron than Lake Erie. For those indicators, most of the area of the HEC was rated good. The HEC appears more degraded when Lake Huron thresholds are applied rather than Lake Erie thresholds. Appropriate thresholds for the connecting channels should align with assessment objectives and be at least as protective as thresholds for the down-channel lake. Future iterations of this assessment will allow evaluation of water quality trends in the connecting channels.

Introduction

Under the Clean Water Act (33 U.S. Code §1251 et seq., 1972), the U.S. Environmental Protection Agency (USEPA) is obligated to report on the condition of the nation’s surface water resources to Congress and the public. USEPA collaborates with state and tribal partners to conduct resource-specific quinquennial National Aquatic Resource Surveys (NARS) to address this responsibility. The NARS are environmental assessments designed to allow spatially-explicit estimation of aquatic ecosystem condition or “health” based on biophysical indicators. NARS use statistical survey designs and standardized field and laboratory protocols to generate estimates of the proportion of the nation’s waters, by area, that are in good, fair, or poor condition. The National Coastal Condition Assessment (NCCA), one of four NARS¹, is an assessment of marine and Laurentian Great Lakes coastal waters (USEPA, 2016a). The waterbodies connecting the Laurentian Great Lakes (hereafter, the connecting channels) have not been included in any NARS or other comprehensive ecological assessments (Edwards et al., 1989; GLEC, 2006; UGLCCMC, 1988).

Here we report results of the first probabilistic assessment of water quality in two connecting channels, the Huron-Erie Corridor (HEC) between Lake Huron and Lake Erie, and the St. Marys River between Lake Superior and Lake Huron. Applying standardized protocols, we evaluated water quality indicators of trophic condition for these waterbodies from 2014 to 2016 and report on them alongside results of the 2015 NCCA assessment of the United States part of the Great Lakes nearshore. The connecting channels include both riverine and lake-like reaches (Fig. 1) that could potentially be assessed as part of NCCA, NLA, or NRSA¹. They are hydrologically connected, culturally affiliated, and economically vital to the coastal communities of the Great Lakes, so we assessed them using NCCA protocols (e.g., Manny et al., 2015; Moerke and Werner, 2011; Ripley et al., 2011).

Figure 1.

Map of St. Marys River (a) and HEC (b). The assessed resource area for each system is shown in darker blue.

The NCCA assessment of trophic conditions in the Great Lakes is based on chlorophyll a, total phosphorus (TP), water clarity as Secchi depth, near-bottom dissolved oxygen (DO), and a multi-metric water quality index that combines these indicators. NCCA assessments apply condition class thresholds or cutoff values to define trophic condition classes (good, fair, and poor, USEPA, 2016a). Assessments based on condition classes rather than statistical estimates (e.g., mean nutrient concentrations or index scores) facilitate communication of environmental information among scientists, policymakers, and the public. NCCA condition class thresholds are based on established criteria, guidelines, or thresholds (Table 1; USEPA, 2016a). For the Great Lakes, NCCA adopted TP, chlorophyll a, and Secchi depth thresholds based on definitions of trophic condition classes (Gregor and Rast, 1979; PMSTF, 1980) and adopted DO thresholds based on a definition of hypoxic conditions (Krieger and Bur, 2009; Brandt et al., 2011; USEPA, 2016a).

Table 1.

Water quality thresholds for the Great Lakes used for NCCA. Lakes Superior, Michigan, and Huron thresholds are associated with oligotrophic conditions, Eastern and Central Lake Erie and Lake Ontario thresholds are associated with oligo-mesotrophic conditions, and Western Erie and Saginaw Bay thresholds are associated with mesotrophic conditions.

Lake/Basin	Chlorophyll a (μg/L)		Total phosphorus (μg/L)		Dissolved oxygen (mg/L)		Secchi depth (m)
	Poor/Fair	Fair/Good	Poor/Fair	Fair/Good	Poor/Fair	Fair/Good	Poor/Fair	Fair/Good
Superior	2.6	1.3	10	5	2	5	5.3	8.0
Michigan	2.6	1.8	10	7	2	5	5.3	6.7
Huron	2.6	1.3	10	5	2	5	5.3	8.0
Saginaw Bay	6.0	3.6	32	15	2	5	2.1	3.9
Erie
Western	6.0	3.6	32	15	2	5	2.1	3.9
Central/Eastern	3.6	2.6	15	10	2	5	3.9	5.3
Ontario	3.6	2.6	15	10	2	5	3.9	5.3

Identifying appropriate condition-class thresholds for the connecting channels is complicated by the channels’ status as both receiving waters of and tributaries to Great Lakes. Ideally, thresholds would be based on spatially-explicit reference conditions that reflect the expected trophic status, ecological integrity (e.g., healthy native biota) and/or the delivery of ecological benefits (e.g., recreational opportunities) to human communities. A pristine reference condition could be defined as pre-European settlement trophic conditions, but these are difficult to quantify without historic data (e.g., Stoddard et al., 2006); and may no longer be attainable. NRSA and NLA (see footnote ¹) identify least-disturbed reference sites representing the “best available chemical, physical and biological habitat conditions given the current state of the landscape” (Stoddard et al., 2006, p. 1271), and with natural attributes representative of the assessed resource (e.g., similar geology, hydrology, climate), which are used to define thresholds (Hughes et al., 1986). However, the history of industrial and commercial use and abuse of the connecting channels make it likely that even least-impaired areas are significantly impacted (USEPA, 2017).

Because we could not define channel-specific thresholds based on a reference expectation, we applied the down-channel Great Lakes thresholds to assess the connecting channels. Because it is protective of the down-channel lake, this approach is consistent with antidegradation provisions of the Clean Water Act (33 U.S. Code §1251 et seq., 1972). In this paper, we summarize the connecting channel water quality assessment in the context of condition of the up- and down-channel lakes’ nearshore. We also briefly discuss an alternative approach to assessment based on the thresholds of tributary inputs to the connecting channels.

Methods

Study Area

The HEC extends for 170 km between Lake Huron and Lake Erie and includes three waterbodies: the St. Clair River, Lake St. Clair (surface area: 1,114 km²), and the Detroit River (Fig. 1). Adjacent to the HEC are the metropolitan areas of Detroit, Michigan - Windsor, Ontario, Port Huron, Michigan, and Sarnia, Ontario. Outside of these urban areas, land use is mostly suburbs and agriculture. The HEC includes or receives tributary inputs from four Great Lakes Areas of Concern (AOC), a designation under the Great Lakes Water Quality Agreement for areas with ongoing impairments of beneficial uses primarily due to historic water pollution and sediment contamination (USEPA, 2017). These include the St. Clair River AOC, the Clinton River AOC, the Detroit River AOC, and the Rouge River AOC. The HEC receives about 4.4% of its flow from tributaries, based on differences in flow between U.S. Geological Survey (USGS) gages at Port Huron, MI on the St. Clair River and at Fort Wayne, MI on the Detroit River (USGS, 2017).

The St. Marys River, which extends for 120 km between Lake Superior and Lake Huron, flows over rapids near its head at Sault Ste. Marie, MI (Fig. 1). Downriver, the river splits into several channels and shallow lakes before debouching into the North Channel and main basin of Lake Huron. Outside of Sault Ste. Marie, land cover in the region is mostly forest. The entire St. Marys River is designated as an AOC. The St. Marys River receives 1.7 – 2.5% of its flow from tributaries. The lower value is based on an average annual flow-drainage area regression from available regional gages, extrapolated for the tributary area draining directly to the St. Marys River. The higher value is based on flow estimates from the Extended Unit Runoff Method (USEPA and USGS, 2012; McKay et al., 2012).

Study Design

The NCCA uses a probabilistic survey design to estimate, with known error, the ecological condition of a spatially-explicit aquatic resource area. The sample design for this study was based on an NCCA Great Lakes nearshore sample design described elsewhere (Kelly et al., 2015; USEPA, 2016a). The Great Lakes nearshore was defined as the area with depths less than 30 m within 5 km of shoreline in United States’ waters. The design included a separate nearshore stratum for each lake. The Great Lake connecting channel resource area was defined separately from the Great Lake nearshore resource area and included the entire area of the HEC and St. Marys River between the up- and down-channel lakes (Fig. 1). The HEC included three assessment strata: the St. Clair River, Lake St. Clair, and the Detroit River. The St. Marys River was assessed as a single stratum.

Sampling sites representative of the Great Lakes nearshore and connecting channels were selected using a generalized random tessellation stratified survey design (GRTS; Fig. 2; Stevens and Olsen, 1999; 2004). For assessment estimates, each sampled site was assigned a weight reflecting the area of the assessed resource it represented. Weights for each stratum sum to the total assessed resource area, which can be slightly less than the “design” resource area if sites are deemed unsampleable in the field due to rapids, ship traffic, or other access issues. Assessed resource areas and number of probability sites sampled are given in Table 2.

Figure 2.

Map of sites sampled in the St. Marys River (a) and HEC (b) in each year.

Table 2.

Assessed resource areas and probability sampling sites collected on the Great Lakes, St. Marys River, and HEC in 2014–2016. Two sites in the HEC were sampled twice in 2014; seven sites sampled in 2014 were revisited in 2015 (1 site in St. Clair River, 4 sites in Lake St. Clair, and 2 sites in Detroit River). Totals for GLCC include the total sites counted in the two-year combined study. For sites that were revisited in the same year or the second year, only the first visit is counted in total and used to estimate condition. Lake Huron sites included 24 sites in Saginaw Bay. Strata are listed from up- to down-system, with connecting channels in italics.

Stratum	Assessed resource area (km2)	Number of probability sites sampled
		2014	2015	2016	Total
Lake Superior	3048	-	78	-	78
St. Marys River	1219	-	50	44	94
Lake Michigan	7321	-	100	-	100
Lake Huron	3272	-	67	-	67
HEC	1287	55	47	-	95
St. Clair River	57	11	8	-	18
Lake St. Clair	1126	28	24	-	48
Detroit River	104	16	15	-	29
Lake Erie	2656	-	57	-	57
Lake Ontario	1375	-	59	-	59

Data Collection and Analysis

The HEC was sampled in September of 2014 and 2015; the St. Marys River was sampled in July and August of 2015 and 2016. Ninety-five and 94 probability sites were sampled in the HEC and the St. Marys River, respectively (Table 2). In the Great Lakes, 361 nearshore sites were sampled from June – September 2015. Due to logistic constraints, the connecting channels were sampled during a shorter interval than the lakes. Given the possibility of transient events (e.g., algal blooms), this complicates direct comparison between channels and lakes. Seasonal variation cannot be addressed in GRTS designs (Messer et al., 1991); the sample or “index” period is assumed to be representative for annual assessment purposes.

The NCCA field and laboratory protocols used to collect TP, total nitrogen (TN), chlorophyll a, DO, and Secchi depth are described in detail elsewhere (USEPA, 2015a; 2015b). Briefly, water samples were collected using a Niskin sampler deployed 0.5 m below the surface. Water samples were filtered with a Whatman GF/F 0.7-μm filter (GE Life Sciences, Marlborough, MA) for chlorophyll a (μg/L), and the filtrate was used to measure dissolved nutrients (μg/L). A Hydrolab DS5 or HL4 multiparameter probe (Sutron, Sterling, Virginia) was used to measure DO (mg/L) 0.5 m above the bottom.

Water clarity was measured with a weighted 20-cm diameter black and white Secchi disk. Condition estimates based on Secchi depth used the mean of three Secchi depth measurements where available. If Secchi depth was deeper than station depth (disk visible resting on bottom), but station depth was greater than the applicable good/fair thresholds described below (38 sites: 9 sites in Lake Superior, 8 sites in Lake Michigan, 9 sites in Lake Huron, 8 sites in HEC, and 4 sites in Lake Ontario), the site was classified as good by default. If not, then the site was classified based on an estimated Secchi depth from on light availability profiles. Profiles were measured with a LI-COR LI-1500 photosynthetically active radiation (PAR) meter (LI-COR, Lincoln, Nebraska). We fit a linear regression to the combined downcast and upcast PAR profiles to estimate the coefficient of extinction (K_d) for each site (Tyler, 1968). We then fit a regression between mean Secchi depth and extinction coefficients, which we used to estimate Secchi depth at sites where Secchi depth was unknown. Sites in the Great Lakes were estimated using a model of all Great Lakes sites sampled in 2015 (Secchi depth = 1.389 K_d^−0.98, r² = 0.90), while sites in the connecting channels were estimated using a model of all Great Lakes sites sampled in 2015 plus all connecting channels sites sampled 2014–2016 (Secchi depth = 1.304 K_d^−0.98, r² = 0.86).

Sample preservation, laboratory and field duplicate samples, and data quality assurance procedures followed NCCA protocols (USEPA, 2015a; 2015b). Data were combined across years for each GLCC assessment. Area-weighted means, cumulative distribution functions, and area estimates of condition were calculated in R using the package “spsurvey” (Kincaid and Olsen, 2012). Half the detection limit was used if measured TP concentration was below the detection limit.

Water Quality Thresholds and Assessment

Great Lakes NCCA thresholds for fair/good condition for chlorophyll a, TP, and Secchi depth represent ambient trophic status expectations for each lake (Table 1). Fair/poor thresholds are based on indicator values presumed to be associated with a shift to a more nutrient-enriched trophic status (Gregor and Rast, 1979; PMSTF, 1980). Distinct thresholds have been defined for Saginaw Bay of Lake Huron and for each basin of Lake Erie. The lake thresholds were developed primarily for open water, but their derivation included analysis of coastal data and they are considered protective of nearshore waters (USEPA, 2016b). NCCA DO thresholds were based on literature values for hypoxic conditions (e.g. Brandt et al., 2011; USEPA, 2000a). For GLCC condition estimates, we applied the most protective threshold from the down-channel lake (Table 1, USEPA, 2016a). For the St. Marys River and HEC, the most protective thresholds were for Lake Huron and Central Lake Erie, respectfully. For the HEC, we also compared condition estimates applying down-channel Central Lake Erie thresholds to results applying up-channel Lake Huron thresholds. Condition estimates are for the assessed resource area. For some sites, some indicator results were missing (See Appendix A). We provide limited results for Lake Michigan because it does not directly receive water from or discharge water into a GLCC, the primary focus of this study.

Water quality was assessed as good, fair, or poor for TP, chlorophyll a, and bottom DO, using the condition class thresholds. The assessed status of each parameter was integrated into the NCCA water quality index (WQI), determined as follows: if no component metrics (TP, chlorophyll a, Secchi depth and DO) were classed as poor, and not more than one was classed as fair, then the WQI trophic condition was assessed as good; if one component metric was classed as poor and/or two or more were classed as fair, then the WQI was assessed as fair; if two or more metrics were classed as poor, then the WQI was assessed as poor (USEPA, 2016a). We do not report categorical condition estimates for TN or include TN in the WQI because thresholds for TN in the Great Lakes are not currently available (USEPA, 2016a).

Results

Water quality estimates

Weighted mean Secchi depth in the nearshore of Lake Erie was lower than the other lakes but similar to the St. Marys River and the Detroit River stratum of the HEC (Fig. 3). After Secchi depth, by far the greatest variation in water quality among the lakes and connecting channels was between the Lake Erie nearshore and the other Great Lakes and GLCCs. Weighted mean TP, TN, and chlorophyll a in Lake Erie’s nearshore were at least 3 times greater than elsewhere (Fig. 3). Furthermore, weighted mean DO concentration approached hypoxic conditions only in Lake Erie. Nearshore Lake Michigan, which is not connected by a GLCC, was intermediate in water quality measures to nearshore Lake Superior and Lake Huron but generally had better water quality than the St. Marys River.

Figure 3.

Weighted means and 95% confidence intervals for each indicator for each stratum assessed, from up- to down-system, with Lake Michigan, which does not receive direct inputs from a connecting channel, shown between St. Marys River and Lake Huron. Circles denote Great Lakes and triangles denote GLCC.

Weighted mean TP in the St. Marys River was intermediate to that of nearshore Lake Superior and nearshore Lake Huron (Fig. 3); median TP in the St. Marys River was higher than both Lakes Superior and Huron (Fig. 4a, Appendix A). This is caused by a skewed distribution of TP in the St. Marys River where 15 samples (16 ± 6% of the assessed area) were at or below the method detection limit of 5 μg/L. These were all samples from 2016, when detection limits were atypically high compared to other years, which had detection limits as low as 1.5 μg/L. Though detection limits were lower, no sites had TP concentrations at or below 5 μg/L in the St. Marys River in 2015. The good/fair threshold for TP is 5 μg/L for both Lakes Superior and Huron, so sites at or below the detection limit were classified as good. Weighted mean and median TP in the HEC were similar to nearshore Lake Huron and much lower than nearshore Lake Erie (Figs. 3, 5a, Appendix A). Weighted mean TP was higher in nearshore Lake Huron than in the St. Clair River, probably because the Lake Huron estimate of weighted mean TP includes sites in nutrient-enriched Saginaw Bay which is far from the Lake Huron outlet to the St. Clair River. Within the HEC, weighted mean (Fig. 3) and median TP were higher in the Detroit River than in Lake St. Clair or the St. Clair River (Fig. 6a).

Figure 4.

Cumulative distributions by area of water quality measures for the St. Marys River, with Lake Superior and Lake Huron for comparison. Horizontal axes for TP, TN, and chlorophyll are logarithmic. Concentration at 50 percent area is the median concentration. In plot a, the CDF starts at 16% because 16% of St. Marys River area was estimated to have TP concentrations at or below the detection limit 5 μg/L in 2016 (see text for more details). Secchi depths shown in plot d were measured Secchi depth or estimated using PAR attenuation profiles, even if the categorical estimates were inferred from station depth being deeper than the applicable Secchi threshold for good condition.

Figure 5.

Cumulative distributions by area of water quality measures for the HEC, with Lake Huron and Lake Erie for comparison. Horizontal axes for TP, TN, and chlorophyll are logarithmic. Concentration at 50 percent area is the median concentration. Secchi depths shown in d. were measured Secchi depth or estimated using PAR attenuation profiles, even if the categorical estimates were inferred from station depth being deeper than the applicable Secchi threshold for good condition.

Figure 6.

Cumulative distributions by area of water quality measures for each stratum in the HEC. Horizontal axes for TP, TN, and chlorophyll are logarithmic. Concentration at 50 percent area is the median concentration. Area in the HEC is dominated by Lake St. Clair, so the entire HEC plots along approximately the same line as Lake St. Clair. Secchi depths shown in plot d were measured directly or estimated using PAR attenuation profiles, even if the categorical estimate was inferred from station depth being deeper than the applicable Secchi threshold for good condition.

The weighted mean TN in the St. Marys River was higher than in the nearshore of both Lake Superior and Lake Huron (Fig. 3). Median TN in the St. Marys River and nearshore Lake Superior were similar and were both higher than Lake Huron (Fig. 4b). The St. Marys River had a higher percentage of area with high (>400 μg/L) TN than the nearshore of Lakes Superior or Huron (Fig. 4b). Mean and median TN in the HEC were similar to nearshore water of Lake Huron but lower than the nearshore of Lake Erie. Lake Erie had a much higher percent of nearshore area with high TN (>1000 μg/L) than either the HEC or nearshore Lake Huron (Fig. 5b). Within the HEC, weighted mean and median TN (Fig. 6b) were similar across strata.

Chlorophyll a concentration in the St. Marys River was intermediate to that of the nearshore of Lake Superior and Lake Huron (Figs. 3, 4c). Chlorophyll a concentration in the HEC was slightly lower than nearshore Lake Huron, and much lower than nearshore Lake Erie (Figs. 3, 5c). Within the HEC, weighted mean and median chlorophyll a were similar across strata. Most of the area in HEC strata had low chlorophyll a, but the Detroit River had some areas with elevated chlorophyll a compared to the rest of the HEC (Fig. 6c). For example, 100% of the assessed area of the St. Clair River had a concentration less than 3 μg/L, while only 79 ± 13% of Detroit River had a concentration less than 3 μg /L.

Weighted mean and median Secchi depth in the St. Marys River were shallower than the nearshore of Lakes Superior and Huron (Figs. 3, 4d). Weighted mean and median Secchi depth in the HEC were much shallower than nearshore Lake Huron, and slightly deeper than nearshore Lake Erie (Fig. 5d). Within the HEC, Lake St. Clair had a deeper weighted mean and median Secchi depth than the St. Clair River or Detroit River (Fig. 6d, Appendix A).

Weighted mean and median DO in the St. Marys River and nearshore Lake Huron were similar and both lower than that of nearshore Lake Superior (Figs. 3, 4e). Weighted mean and median DO in the HEC were similar to nearshore Lake Huron, and higher than nearshore Lake Erie (Fig. 5e). Within the HEC, weighted mean and median DO were similar across strata (Fig. 6e).

Condition estimates

For the St. Marys River, less of the area was in good condition based on TP, than either Lake Superior or Lake Huron (Fig. 7a, Appendix A). Based on chlorophyll a, most of the area in the St. Marys River was in good condition, as were Lake Superior and Lake Huron (Fig. 7b). As a result of applying Central Lake Erie thresholds to the HEC, more of the area was in good condition based on both TP and chlorophyll a, than Lake Huron or Lake Erie. (Fig. 7a, b).

Figure 7.

Condition estimates (±95% confidence intervals) for the Great Lakes and connecting channels. Blue represents good condition, yellow represents fair condition, and red represents poor condition. Lake Michigan and Lake Ontario condition estimates can be found in Appendix A.

Condition estimates based on dissolved oxygen were good in nearly all cases (Fig. 7c). Based on Secchi depth, 97 ± 3% of the area of the St. Marys River was in poor condition. Most of the nearshore of Lakes Superior and Huron were in good condition based on Secchi depth (Fig. 7d, Appendix A). Most area of the HEC and Lake Erie’s nearshore were in poor condition based on Secchi depth.

Based on the WQI, more of the St. Marys River was assessed as poor condition than any other GLCC or Great Lake, except Lake Erie (Fig. 7e, Appendix A). Nearly all the St. Marys River (97 ± 3%) and Lake Erie (73 ± 9%) were classified as being in poor condition based on Secchi depth, so if either TP or chlorophyll a were classified as poor, the WQI was classified as poor for that site. Applying Central Lake Erie thresholds, conditions in the HEC were similar to Lake Huron and better than Lake Erie. Within the HEC, the proportion of poor conditions for WQI increased in each stratum in the down-channel direction, but confidence intervals were large due to low sample size (Fig. 8g).

Figure 8.

Condition estimates (±95% confidence intervals) for the HEC and each stratum for two threshold options. Blue represents good condition, yellow represents fair condition, and red represents poor condition. Dissolved oxygen is not included because thresholds are the same in all lakes.

More of the HEC was assessed as poor using Lake Huron thresholds than when using Central Lake Erie thresholds, because the Lake Huron thresholds are more protective. If Central Lake Erie thresholds are applied to the HEC, 47% more area was rated good for TP, and 17% more area rated good for chlorophyll a compared to condition based on Lake Huron thresholds (Fig. 8, Appendix A). Based on Secchi depth, 24% of the HEC was rated good with Central Erie thresholds and no area rated good with Lake Huron thresholds. Conditions in the HEC based on TP and chlorophyll a with Lake Huron thresholds were similar to (within 95% confidence intervals) conditions in nearshore Lake Huron and better than the Lake Erie nearshore (Figs. 7 and 8, Appendix A). Conditions in the HEC based on Secchi depth with Lake Huron thresholds were worse than both Lake Huron and Lake Erie.

Discussion

Condition estimates for both TP and chlorophyll a suggest that the HEC is in better condition than the nearshore of both Lake Huron and Lake Erie (Fig. 7). However, comparing assessments of the HEC to Lake Huron and Lake Erie is complicated by multiple sets of thresholds reflecting different trophic expectations for each waterbody (Table 1). The HEC was assessed with thresholds protective of the oligo-mesotrophic conditions of Central Lake Erie receiving water. The nearshore of Lake Huron was assessed with thresholds reflective of oligotrophic conditions except for Saginaw Bay, which was assessed with thresholds protective of mesotrophic conditions. The nearshore of Lake Erie was assessed with thresholds protective of mesotrophic and meso-oligotrophic conditions. Unlike condition, weighted mean and median TP and chlorophyll a concentrations and cumulative distributions can be compared directly to evaluate variation in water quality across the Great Lakes system. Concentration was similar for the HEC and nearshore of Lake Huron, and both were lower than the nearshore of Lake Erie suggesting that water quality in the HEC is more like Lake Huron than Lake Erie, and that the HEC is probably a relatively minor contributor to eutrophication of western Lake Erie, compared to Western Lake Erie’s tributaries (e.g., the Maumee River). This is consistent with recent studies of nutrient loading and eutrophication of Western Lake Erie (e.g., Scavia et al., 2014; Michalak, 2013).

Based on TP, condition estimates for Lake Huron were better than both St. Marys River and Lake Superior, despite higher concentrations of TP in Lake Huron. Most of Lake Huron nearshore sites were assessed using the same thresholds as Lake Superior, but the nutrient-enriched sites within Saginaw Bay were assessed using more relaxed thresholds protective of mesotrophic conditions (Table 1). Because each lake and connecting channel have a distinct trophic state expectation, differences in condition estimates among waterbodies must be interpreted with caution. Comparison of weighted means or cumulative distributions of indicators is a straightforward way to assess the relative water quality of waterbodies in the Great Lakes, while categorical condition estimates facilitate communication of results to non-scientists.

The AOC program provides another example how thresholds may complicate inferences and the integration of findings. The “eutrophication or undesirable algae” beneficial use impairment for the St. Mary’s River Area of Concern was removed in 2016 based on long-term monitoring at two sites: near Brush Point and near Point Aux Frenes, Michigan (Fig. 1; MDEQ, 2016). Most long-term monitoring samples for TP were within the mesotrophic/oligotrophic threshold of 25 μg/L and the AOC beneficial use impairment target of 12 μg/L, which was based on USEPA ecoregion criteria (USEPA, 2001a; MDEQ, 2016). Our closest sites to these longterm sites also had TP concentrations within these limits (7 μg/L near Point Aux Frenes and 5 μg/L near Brush Point). If we apply the 12 μg/L TP threshold directly to our dataset, 87 ± 6% of the entire St. Marys River area meets the threshold, compared to only 16 ± 6% with the NCCA threshold of 5 μg/L (Table 1). In this case, the choice of threshold affects the assessment outcomes and limits integration of assessment results across programs.

Down-channel worsening of conditions within the HEC (Figs. 6, 8) corroborates previous findings that TP increases downriver in both the HEC and St. Marys River (Edwards et al., 1989; Liston and McNabb, 1986; MDEQ, 2013; MDEQ, 2014; UGLCCMC, 1988). Previous studies were based on long-term monitoring at just a few widely-spaced locations, which are not necessarily representative of conditions across the entire system. Because of the probabilistic design, these data provide reliable estimates of current background conditions across entire GLCC systems. Individual site data can also be used to identify spatial patterns and suggest areas for further sampling. For example, our data revealed poor water quality conditions in Munuscong Lake in the St. Marys River; and in eastern Lake St. Clair in the HEC (Appendix B). These may be a result of nutrient loading from the Munuscong River and the Thames River, respectively, which both drain predominantly agricultural areas.

TN was higher in the St. Marys River than in either Lake Superior or Lake Huron (Fig. 4). Oligotrophic systems like Lake Superior tend to have high TN:TP ratios (Sterner, 2011). Elevated TN in St. Marys River may be due to high-TN inputs from Lake Superior and/or tributary inputs along the channel. No thresholds have been established for TN for the Great Lakes, although it is used in the NCCA to assess marine estuarine condition (USEPA, 2016a). Establishment of TN thresholds for the Great Lakes would help refine the NARS assessment especially for Lake Erie, for which nitrogen limitation has been documented (Dove and Chapra, 2015; Chaffin et al., 2013).

Based on Secchi depth, the St. Marys River and the HEC were degraded compared to the Great Lakes nearshore except Lake Erie nearshore, to which they were similar (Fig. 3, and Appendix C). However, shallow Secchi depths were not coupled with similarly degraded conditions for TP and chlorophyll a, suggesting suspended sediment may account for low water clarity in the GLCCs. Lake St. Clair had a slightly deeper weighted mean Secchi depth (3.5 ± 0.3 m) than St. Clair River or Detroit River (3.2 ± 0.5 m, 2.6 ± 0.3 m, respectively). Previous studies have documented high turbidity in some areas of the St. Marys River associated with wind-induced sediment resuspension, especially in Lake George (Liston et al., 1983). Turbidity in the Detroit River has increased since the 1990s (MDEQ, 2013). Reduced clarity in the connecting channels may be a result of higher shoreline development, greater impacts of boating/ship traffic on shorelines, or higher proportions of shallow, wave influenced areas in GLCCs compared to the Great Lake nearshore.

Most of the St. Marys River and the HEC were assessed as good based on DO. Dissolved oxygen concentrations in the St. Marys River were similar to Lake Huron and consistently lower than Lake Superior, but still well above thresholds for hypoxic conditions (e.g. Brandt et al., 2011). These results agree with previous water quality studies which did not identify concerns with DO in the GLCCs (Edwards et al., 1989; Liston et al., 1983; MDEQ, 2013).

The WQI was designed to be an indicator of a past, current, or imminent algal bloom in marine systems (USEPA, 2016a) and was adapted for the Great Lakes based on expected trophic status (minus TN, for which thresholds were not available). The ecological meaning of the WQI as applied in the Great Lakes has not been evaluated. Weighted means and medians of constituent indicators (TP, chlorophyll a, water clarity, DO) may provide more interpretable information about conditions in the GLCCs relative to the larger Great Lakes system than the WQI.

We applied downriver thresholds for assessment of GLCCs to be consistent with the protection of receiving waters under the Clean Water Act (33 U.S. Code §1251 et seq., 1972). An alternate approach to setting water quality thresholds is the application of input-water thresholds. If GLCC “tributary” inputs (i.e., streams and upriver Great Lakes inputs) are all meeting their respective water quality thresholds, then conditions within the GLCC should be maintained at an ecological condition defined by the least protective tributary threshold. Tributaries to the Great Lakes and connecting channels generally have higher ambient concentrations of nutrients and lower water clarity than the waterbody into which they flow, as well as higher water quality thresholds than the downstream Great Lake (e.g., Hem, 1989; USEPA 2000a, b, c). Tributary input-weighted thresholds for the GLCC could be derived based on thresholds for input waters (Great Lakes and tributaries), weighted by the fraction of flow coming from each input source. However, most published thresholds for streams are based on achieving specific use criteria rather than ecological assessment and therefore may not be directly applicable. Also, a tributary-input-weighted threshold is only valid for the outlet of the GLCC after all tributaries have entered the GLCC. To avoid this problem more assessment strata defined by the location of significant tributaries would be required.

Tributary input to the HEC is generally very small, with approximately 95% of total flow at the HEC outlet coming from Lake Huron. Using Lake Huron TP and chlorophyll-a thresholds, the HEC had a smaller percentage of area rated as good than when calculated using Central Erie thresholds. The HEC appears more degraded when Lake Huron’s oligotrophic thresholds are applied than when Central Erie’s oligo-mesotrophic thresholds are applied. Based on weighted means and the cumulative distribution of the data, water quality in the HEC is similar to Lake Huron. However, Lake Huron thresholds may not be realistic or achievable thresholds for the HEC because although most of the flow in the HEC comes from Lake Huron, tributary inputs, even those meeting their respective water quality thresholds, increase nutrient concentrations in the HEC relative to Lake Huron.

Conclusion

The Great Lakes and their connecting channels are an immense, biophysically diverse resources of international significance. Here we present the first probabilistic assessment of water quality in the St. Marys River and the HEC connecting channels. Weighted means and medians of indicators allow comparison of these GLCCs to the larger Great Lakes system, while categorical estimates provide a foundation for tracking water quality relative to societal expectations for each Great Lake and connecting channel into the future. Synoptic assessment of the Great Lakes system as called for by the binational Great Lakes Nearshore Framework (ECCC and USEPA, 2016) requires the identification of defensible and protective water quality thresholds for all constituent waterbodies, including connecting channels. Trophic expectations change through the Great Lakes system, so outcomes of the assessments depend on the thresholds applied. In the absence of channel-specific thresholds, we applied the thresholds for the down-channel Great Lake for the NCCA assessment. Additional thresholds relevant to current and anticipated connecting channel management needs should be considered.

APPENDICES

Appendix A.

Tabular weighted means and medians (A1); categorical population estimate results (A2), and summary of number of sites included in each indicator assessment (A3) can be found online at [supplementary file].

See attached tabular results.

Appendix B.

Maps showing site-specific assessment results for the Huron-Erie Corridor and St. Marys River for TP (B1), TN (B2), chlorophyll a (B3), Secchi depth (B4), and DO can be found online at [supplementary file].

Spatial variation in total phosphorus, total nitrogen, chlorophyll a, Secchi depth, and dissolved oxygen for the St. Marys River and HEC.

Figure B1.

Spatial variation in total phosphorus for St. Marys River (a) and the Huron-Erie Corridor (b). Darker colors and larger symbols indicate higher concentrations, with the lower limit of each bin representing the 0, 20^th, 40^th 60^th and 80^th percentile of the concentrations measured each connecting channel. ND = not detected.

Figure B2.

Spatial variation in total nitrogen for St. Marys River (a) and the Huron-Erie Corridor (b). Darker colors and larger symbols indicate higher concentrations with the lower limit of each bin representing the 0, 20^th, 40^th 60^th and 80^th percentile of the concentrations measured each connecting channel.

Figure B3.

Spatial variation in chlorophyll a for St. Marys River (a) and the Huron-Erie Corridor (b). Darker colors and larger symbols indicate higher concentrations, with the lower limit of each bin representing the 0, 20^th, 40^th 60^th and 80^th percentile of the concentrations measured each connecting channel.

Figure B4.

Spatial variation in Secchi depth for St. Marys River (a) and the Huron-Erie Corridor (b). Lighter colors and larger symbols indicate higher Secchi depths, with the lower limit of each bin representing the 0, 20^th, 40^th 60^th and 80^th percentile of the concentrations measured each connecting channel.

Figure B5.

Spatial variation in bottom dissolved oxygen for St. Marys River (a) and the Huron-Erie Corridor (b). Lighter colors and larger symbols indicate higher DO concentrations, with the lower limit of each bin representing the 0, 20^th, 40^th 60^th and 80^th percentile of the concentrations measured each connecting channel.

Appendix C.

Cumulative distribution function (CDF) plots for water quality indicators for each of the Great Lakes can be found online at [supplementary file].

Cumulative distributions of water quality measures for each of the Great Lakes and GLCC. Concentration at 50 percent area is the median concentration. Horizontal axes for TP, TN, and chlorophyll a are logarithmic. In plot a, the CDF for St. Marys River starts at 16% because 16% of St. Marys River area was estimated to have TP concentrations at or below the detection limit 5 μg/L in 2016 (see text for more detail). Secchi depths shown in plot d were measured Secchi depth or estimated using PAR attenuation profiles, even if the categorical estimates were inferred from station depth being deeper than the applicable Secchi threshold for good condition.