Spatial variation is well known to exist in water quality parameters of the Great Lakes nearshore; however, strong patterns for extended reaches also have been observed and found to be robust across seasonal time frames. Less is known about robustness of inter-annual variation within parameters for water quality in the nearshore. We have conducted high-resolution surveys with towed electronic instrumentation in nearshore areas of Lake Superior and have combined several seasons (2001–2005) of measurements from multiple research efforts to investigate how spatial variation compares across years. The combined survey tows ranged across approximately 1200 km of Lake Superior's south shore. In addition to the survey tracks, we also sampled fixed stations to collect calibration data and other parameters not observed by the in situ electronic sensors. The towed sensor data provided information on the spatial and temporal variability of water quality parameters along the nearshore. We found a consistent spatial pattern over time along the south shore of Lake Superior. Nearshore water quality parameters were analyzed with respect to landscape characteristics of the adjacent watersheds (US only) using multivariate stepwise regressions and found to correlate to landscape characterization. The stressor categories of landscape character that best described the nearshore parameters were agriculture-chemical usage and land-cover attributes. Peak nearshore values corresponded with landscape position that had the most altered landuse character (e.g. Duluth/Superior region). The landscape character appears to drive and maintain the spatial pattern in nearshore water quality parameters.

Introduction

In a recent paper we provided evidence for a strong connection between landscape characteristics and water quality, as well as plankton, along a 537-km transect in western Lake Superior's nearshore zone, from Grand Maris, Minnesota, to the Keweenaw Peninsula in Michigan (Yurista and Kelly, 2009). Although strong relationships between tributaries, small lakes, and wetlands and their surrounding watersheds are well described in the literature (Allan et al., 1997; Herlihy et al., 1998; Morrice et al., 2008), we know of few studies describing a broad, watershed-based relationship for the open coastal and nearshore waters of the Great Lakes. Reavie (2007) and Kireta et al. (2007) have provided evidence of a watershed-diatom linkage that includes open nearshore waters, and our own research group has described relationships for Great Lakes coastal wetlands (Trebitz et al., 2007; Morrice et al., 2008). But nearshore waters have been poorly studied for decades (Mackey and Goforth, 2005; Kelly, 2009). In part, a belief that time and space dynamics and variability within the nearshore obscure a generalized watershed linkage has discouraged attempts at broad scale studies. To an extent this has instead promoted intensive studies in only some selected watersheds across the Great Lakes.

From initial studies, we recognized the high nearshore variability in general (Yurista et al., 2005, 2006) and in Lake Superior in particular, which was especially noted when comparing nearshore and offshore spatial patterns (Yurista et al., 2009). Our previous study used continuously towed in situ sensors to provide a high resolution, spatially continuous data set that was semi-synoptic over a brief period of time (days) and across an extensive portion of the huge Lake Superior shoreline. Our approach, using in situ technologies, was indeed a directed attempt to overcome local variability in order to reveal broader patterns. Spatial patterns were evident which paralleled some watershed patterns along the track, such that pelagic plankton and water properties were strongly predicted from adjacent watershed properties to a very high degree (Yurista and Kelly, 2009). This paper expands on that topic by using results compiled from several high-resolution sampling activities over a longer period of time (summers of 2001–2005), to examine alongshore variability in space, over time.

The compiled data sets allow us to view most of Lake Superior's US shoreline (>1200 km) within about a 5-year time frame, with some shoreline segments repeatedly sampled within this period. Operationally we are considering the nearshore to be the region of the coastal boundary that is dominated by alongshore currents and extends out to approximately the 30-m depth contour (Rao and Schwab, 2007). The remarkable observation from the compilation is that watershed-based patterns are consistent and robust spatially, and we can indeed describe in situ conditions as a function of watershed attributes for the whole US coast of Lake Superior. Our analysis further uses the data sets to make some significant space- and time- variability comparisons, leading us to conclude that the landscape properties are a prime driver that maintains the observed spatial patterns across the US nearshore.

Methods

General

Multiple sites and long stretches of shoreline in the US nearshore of Lake Superior were surveyed over a 5-year time frame (2001–2005). We towed electronic instrumentation in a variety of tow patterns in the nearshore region with bottom depths that ranged between 5 m and 30 m. Some tow paths followed a grid structure (Yurista et al., 2005, 2006) within the 20-m depth contour; others followed contour lines (5-m, 10-m or 20-m isopleths) and often included cross-contour tows; and some used a comprehensive grid at tributary pour-points with a paired nearby high energy coastline. Several tow paths were 20-km stretches of nearshore at a 20-m contour that included a cross contour at the midpoint from 5 to 30 m depths. Some tows were 10-km stretches outside embayments at the 10-m contour. Lastly, we conducted a long continuous tow (537 km) at the 20 m contour (Yurista and Kelly, 2009). Effort across years was aimed at investigating various tow strategies to efficiently extend spatial coverage of the US nearshore of Lake Superior. For Lake Superior and other Great Lakes we have determined that an efficient strategy for contrasting regional land use patterns with nearshore water quality is a long continuous tow at a defined depth contour (Yurista et al., 2011; in review, 2011; Kelly and Yurista, 2011). Depth is one factor that shapes many physical processes (swash and surf zones, alongshore currents, bottom friction) and ecological community structure (fish communities: pelagic-coastal, benthic: profundal-littoral). Alongshore currents and mixing creates a nearshore region from about 10 m to 30 m in depth and 2 to 4 km in breadth that is generally well mixed with a low dilution gradient across depth for parameters when compared particularly to the initial high dilution gradient experienced across the swash and surf zones (< ∼5 m depth; Rao and Schwab, 2007). We have operationally settled on a depth based strategy of a 20-m contour that allows us to monitor the alongshore mixing zone, remain within the coastal boundary region (Rao and Schwab, 2007), and allows us to have a margin of safety for ship operations in many locations with rapidly changing bottom contours. The combined data from all of our towing does not represent a statistically based sampling strategy (Stevens and Olsen, 2004), although, within any given year the sample sites were chosen with a probability based procedure and the total coverage extends over most of the US shoreline of Lake Superior. Also, the data over this time span are not necessarily the most desirably matched in attributes, differing targeted depth contours and differing tow lengths, but they do provide a broader observational scale than any one specific study.

An array of electronic instrumentation was mounted on a tow fish that was either a V-FIN (YSI 465) or a Mini-Bat (Guildline model 8820). The array included an optical plankton counter (OPC) or a laser-OPC (Herman, 1992; Herman et al., 2004) multiplexed with a CTD (SeaBird SBE19+), fluorometer (WetLab), and transmissometer (WetLab). The array was towed at a target speed of 2.25–2.5 m s−1 and data was recorded every 0.5 seconds referenced with global positioning data. The tow fish was undulated from approximately 2 m below the surface to 2 m above the bottom. All tows were during daytime hours and conducted during the seasonal peak biological sampling window (mid-July thru mid-September; Yurista et al., 2005) from the vessels R/V Prairie Sounder (2001), R/V Lake Explorer (2002–2004), and R/V Tullibee (2004–2005). Biological response to tributary and landscape inputs will be most easily detected during stable biological conditions such as during the seasonal summer peak period. Rapidly developing or declining growth periods (spring or fall) have increased background variance across the measurement time frame. Samples separated by a few days to weeks during population development can experience a large change in standing stock. While a large proportion of the yearly runoff which fuels the production occurs during the spring, mixing is primarily alongshore and known to experience entrapment within the coastal boundary that slows mixing to offshore waters by hydrological processes that include thermal bars (Csanady, 1970; Auer and Gatzke, 2004; Rao and Schwab, 2007). The entrapment allows for a biological response to set up in the nearshore. The late summer sampling window is a best compromise for an annual monitoring effort to detect response to trends in landscape stressors, and this period may even include some remnant influence from spring runoff.

Sensor data were converted to engineering units using manufacturer software and annual factory calibration coefficients. Sensor data were also compared with water quality point-samples to corroborate and to correlate sensors with similar point measurements (e.g. fluorescence and chlorophyll a). Data was further processed with a kriging routine (SURFER, 2002). Kriging produced point estimates at regular grid intervals that are best linear unbiased estimators of irregularly spaced data (Isaaks and Srivastava, 1989). We used GIS processing to assign every grid point along the tow track to an associated watershed (greater than 2nd order). Established boundaries (segmentsheds, below) were extended into the nearshore at midpoints between adjacent watersheds. The associated grid points for each watershed were averaged within parameters to produce estimates that were used in correlations with the watershed land use characterizations.

Land use characterization was taken from Danz et al. (2005) for each watershed, including their interfluves, and referred to as segmentsheds (Hollenhorst et al., 2007). Segmentsheds of second order and larger were characterized in seven broad categories of agriculture-chemical applications, atmospheric deposition, land cover, population density, point-sources, shoreline modifications, and soils. Landscape character was ranked by principal component scores within each general category (Danz et al., 2005). We investigated whether the nearshore water properties we measured correlated with the associated segmentshed characterizations (Yurista and Kelly, 2009). The average value for each tow parameter associated with each segmentshed was regressed against the seven principle component characterizations. We used a backward step-wise multivariate regression analysis (SYSTAT, 2004).

Inter-comparison of OPC and LOPC data

As technology advances, it is important to maintain the comparability of long-term data collection programs across sampling methods and across technological generations of instrumentation. During this study we used two versions of an OPC to monitor the spatial extent of zooplankton biomass. The second generation optical plankton counter (laser-OPC; Herman et al., 2004) has an improved detection range for smaller particles and a reduced rate of coincident counts; however, field response may possibly differ from the earlier OPC model. Both an OPC and an LOPC were mounted within inches of each other on the Mini-Bat tow device in order to compare their field response in the same environment. We towed a 17-km transect on 14 October 2004 across the western arm of Lake Superior from the Lester River, Minnesota, to the Amnicon River, Wisconsin.

We restricted our direct comparison to a range of overlap between ∼300- and ∼1800-μm equivalent spherical diameter (ESD) because of the difference in detection ranges and bin sizes between the OPC and LOPC. We used 305-μm ESD for the lower limit based on our previous experience with the OPC's apparent reduced ability to detect smaller particles at our typical towing speeds (also see below). The upper limit (∼1800-μm ESD) was based on: 1) particle abundances being very low above this value, 2) involved computations of oversized particles with the LOPC (>1920-μm ESD), and 3) to preclude concerns in the largest OPC size bins with possible electronic noise from the wing angle control of the tow body. Biomass was summed into 0.1 log-unit bins across the detection ranges to construct and compare normalized biomass size spectra (NBSS) between the OPC (305–1758 μm) and LOPC (300–1755 μm) (NBSS; Kerr and Dickie, 2001).

The best overlay of OPC and LOPC spectra was determined by readjusting the OPC field calibration shape factor f (Sprules et al., 1998). The best overlap criterion was that there should be minimum least squared error between the NBSS of the OPC and the LOPC. The shape factor f was varied such that the OPC NBSS estimate most closely matched the estimate from the field calibrated LOPC (f = 2.585) (Yurista et al., 2009). The resulting shape factor f of 1.975 provided a secondary calibration of the OPC that best correlated with the LOPC data in the range of overlap (Figure 1).

Figure 1.

Simultaneously collected data from an LOPC and an OPC. Normalized biomass size spectra (Kerr and Dickie, 2001) with OPC fit to LOPC restricted to the range of 300 to 1800 ESD as a secondary calibration with f = 1.975 for OPC spectra computations and f = 2.585 for the LOPC. Area below the curves represents the biomass mis-match measured by the OPC to biomass in an LOPC net-equivalent measurement. The mis-match in OPC tow biomass (305 μm) was extrapolated to 153 μm net biomass by adjusting field calibration to f = 1.58.

Figure 1.

Simultaneously collected data from an LOPC and an OPC. Normalized biomass size spectra (Kerr and Dickie, 2001) with OPC fit to LOPC restricted to the range of 300 to 1800 ESD as a secondary calibration with f = 1.975 for OPC spectra computations and f = 2.585 for the LOPC. Area below the curves represents the biomass mis-match measured by the OPC to biomass in an LOPC net-equivalent measurement. The mis-match in OPC tow biomass (305 μm) was extrapolated to 153 μm net biomass by adjusting field calibration to f = 1.58.

OPC field calibrations, of necessity, incorporate into the shape factor both an extrapolation for biomass across sizes not measured with the OPC (Figure 1) and compensation for coincidence counting error to obtain a correlation with zooplankton net-tow biomass (Sprules et al., 1998). In Lake Superior few to no coincident events are expected due to low plankton concentrations, so the shape factor correlating OPC to zooplankton net-tow equivalent would primarily incorporate only a biomass extrapolation. We made a direct comparison to the LOPC with 150 μm as the minimum bin size as a surrogate for 153-μm mesh nets as calibrated for Lake Superior (Yurista et al., 2009). We found the total biomass measured by the OPC in the sizes above 305 μm to be 64% of the total biomass for the LOPC 153-μm mesh net equivalent. The best iteration of the OPC shape factor was f of 1.58 to account for the unmeasured biomass below 305 μm as the direct comparison to LOPC 153-μm net equivalent biomass.

The OPC and LOPC showed good agreement across their range of overlap. There was indication at the low end of data ranges for each instrument that additional considerations of optical particle data may need to be considered (Figure 1). Detection in lower OPC size bins clearly fell off compared to the LOPC and probably reflects the effect of our tow speed on OPC sensitivity to the smaller particles. At towing speeds of 2.25 to 2.5 m s-1, particles smaller than about 338 to 354-μm ESD appeared to be under reported. In a similar manner we interpret the response in the LOPC lower size bin region likely to have reduced detection sensitivity for particles smaller than about 120 to 135-μm ESD. We do not have an analogous comparison for the LOPC to a faster and more sensitive instrument in this lower detection region, but we are extrapolating from the corresponding observation at the lower detection end for the OPC.

Results

Observable trends (LOWESS fits, SYSTAT, 2004) were noted in the water quality parameters of the nearshore (5 to 30-m bottom depth) against landscape position across all years of data (Figure 2). Temperature showed a definite gradient along the nearshore of increasing value from the western and northern most position at the Pigeon River (0 km) to the eastern most point at Sault Ste. Marie (1230 km). Peak values in the other parameters occurred in the Duluth-Superior to Apostle Island area (260–400 km, Figure 2). A second but less prominent peak occurred near the Ontonagon River (∼550 km; Figure 2) that tended to be obscured by an overlapping broader spatial pattern. There was a more consistent base level with less spatial variation across the rest of the entire nearshore. The variability across the 5-year time span was most pronounced in the Duluth-Superior region in conjunction with the peak values. The other extensive stretches did not exhibit variability that was as large near Duluth-Superior (Figure 2). The variability of fluorescence was less spatially distinct, but all the measurements were low and in a relatively narrow range.

Figure 2.

Average parameter values from electronic sensors associated with tributary watersheds of second order or larger and plotted against distance starting at the US-Canada border along the Pigeon River, MN (km 0) extending to the Sault Ste. Marie (km 1234). The trend line is a LOWESS fit to the data (Systat, 2004). Spatial variability is depicted by residuals of sample site parameter values to the LOWESS trend line.

Figure 2.

Average parameter values from electronic sensors associated with tributary watersheds of second order or larger and plotted against distance starting at the US-Canada border along the Pigeon River, MN (km 0) extending to the Sault Ste. Marie (km 1234). The trend line is a LOWESS fit to the data (Systat, 2004). Spatial variability is depicted by residuals of sample site parameter values to the LOWESS trend line.

The measured water quality parameters were correlated with the characterization of the associated segmentsheds (Figure 3; Table 1). Backward stepwise regressions were generally convincing for field ecological data (r2 of 0.27 to 0.41), explaining much of the spatial pattern in variability across the 5-year time course of all the measurements (Table 1). The stressor categories of landscape character that best described the nearshore parameters were agriculture-chemical usage and land-cover attributes. The landscape position with peak nearshore values corresponded with segmentsheds having the most altered landuse character (e.g. Duluth/Superior/Apostle Island region, ∼260–400 km). The regions of the nearshore with the least variability were associated with the more pristine areas of the lake and their segmentsheds (e.g. US Pictured Rocks National Park, ∼1000–1100 km).

Figure 3.

Regression models (lines) for measured nearshore water parameters (open circles) as a function of spatially based landscape character (Table 1).

Figure 3.

Regression models (lines) for measured nearshore water parameters (open circles) as a function of spatially based landscape character (Table 1).

Table 1.

Results of nearshore water quality variables as a function of landscape characterization using backward step-wise regression. Variable = Constant + b*AC1 + c*LC1 + d*AD1 + e*PD1 + f*PS1 + g*SL1 + h*SO1, where landscape characterization values for each watershed (AC1-Ag-Chemicals, LC1-Land Cover, AD1-Atmospheric Deposition, PD1-Population Density, PS1-Point-Sources, SL1-Shore Line alteration, SO1-Soils) are from Danz et al. (2005). There were 232 data points and 7 possible variables for each parameter of interest.

VariableSpecific ConductivityFluorescenceBeam AttenuationZooplankton Biomass
r2(μS m−1)(0.41)(μg m−3)(0.27)(m−1)(0.31)(mg m−3)(0.33)
Constant  101.4 1.80 1.99 484.4 
b  0.372 0.141 42.66 
c  −0.233 −0.176 −0.059 −30.55 
d  
e  
f  
g  0.459 
h  
VariableSpecific ConductivityFluorescenceBeam AttenuationZooplankton Biomass
r2(μS m−1)(0.41)(μg m−3)(0.27)(m−1)(0.31)(mg m−3)(0.33)
Constant  101.4 1.80 1.99 484.4 
b  0.372 0.141 42.66 
c  −0.233 −0.176 −0.059 −30.55 
d  
e  
f  
g  0.459 
h  

Discussion

Spatial variability was observed over the 1200 km of US nearshore, yet there was a consistent pattern against landscape position that emerged across the years. The nearshore region of the Great Lakes has been considered to be dynamic and functionally dependent on processes that include alongshore currents, storm event mixing, and variable landscape input from rain events and winter snowmelt. With all of these contributing factors, there remained robustness to the regional water quality character that was presumably maintained by these same processes. Local conditions may be temporarily disrupted or altered by storm events, but they also appeared to be reestablished by the synergism of all the interacting processes (landscape input, nearshore currents, etc.). A significant process in structuring the nearshore is that the water mass tends to remain entrapped in the coastal boundary (Csanady, 1970). Also, as water mass moves back and forth with the alongshore current, it is continually amended by tributary input that acts to adjust concentrations in receiving areas in proportion to the input. Tributary discharge volumes and concentrations vary from spring to fall and across years, which increases the variability in the nearshore receiving and subsequent down-current areas, yet the input signals from disturbed segmentsheds can be detected as influencing broad scale regions of the receiving waters with higher parameter values that also have greater variability. Localized episodic events will be expected to increase local variability about the alongshore pattern, however, tributary inputs, alongshore currents, and mixing processes will have a moderating effect on disruptive events, such as storms, and act to reestablish local conditions similar to the pre-event regional condition. We have conducted repeat surveys separated by several weeks within a single season in some of the other Great Lakes that have provided support for the regional robustness in condition (Yurista et al., 2011; in review, 2011). We have not yet conducted surveys following spring runoff to fully confirm land use linkages throughout the year in structuring the nearshore regional pattern. The alongshore currents and mixing, coupled with tributary and direct landscape inputs, organize an apparently robust spatial pattern at the regional level in nearshore water quality parameters that is shaped by the landscape signal.

To further investigate the spatial coherence over time, we compared the 2 years with the most data and overlap. The surveys in 2004 and 2005 were both very extensive in coverage and the tow tracks had considerable overlap in space. The sampling styles differed with a long continuous tow in 2004 over about 3–4 days compared to a series of multiple but discrete daily ∼20-km segments (coincident with a series of random points drawn across the US shoreline) and sampled over about 5–6 weeks (Figure 4). Both efforts (2004 and 2005) focused around a 20-m depth contour, were conducted parallel to shore during the late summer period (August to mid-September), and were otherwise comparable. These two datasets allow us to analyze spatial patterns observed at the same locations but in different years. If landscape contributions are primary drivers, then to some degree, similar patterns should persist over time.

Figure 4.

Comparison of spatial coherence between 2004 (solid circles) and 2005 (open circles) data. Trend lines for 2004 and 2005 are LOWESS fits to the respective annual data (Systat, 2004). Data from less spatially extensive studies from 2001–2004 are also shown (crosses). The specific conductivity axis was restricted and eliminated one high reading for specific conductivity (110 μS cm−1) near Duluth (∼260 km) in 2001.

Figure 4.

Comparison of spatial coherence between 2004 (solid circles) and 2005 (open circles) data. Trend lines for 2004 and 2005 are LOWESS fits to the respective annual data (Systat, 2004). Data from less spatially extensive studies from 2001–2004 are also shown (crosses). The specific conductivity axis was restricted and eliminated one high reading for specific conductivity (110 μS cm−1) near Duluth (∼260 km) in 2001.

We conducted several levels of the inter-annual comparison and present here prime examples with specific conductivity and zooplankton. The patterns for the 2 years of high-resolution data throughout the water column (Figure 4) have some remarkable similarities, with common regions of higher and lower concentrations and many particular locations showing even similar vertical structure. The years 2004 and 2005 exhibit different average concentration levels, but there is an overall similarity of spatial pattern trend lines (LOWESS fit; SYSTAT, 2004) where sampling spatially overlapped (Figure 4). This pattern includes peaks in the Duluth-Superior local area (∼260 to 280 km), high but variable values eastward, and usually secondary high values at the Apostle Islands-Chequamegon Bay-Bad River region (∼370 to 425 km), and values declining more moving further eastward.

The coherence of the annual spatial patterns is quantitatively demonstrated in the paired values for each segmentshed location sampled both years (Figure 5). The paired site overlap covers a range from ∼257 to 571 km, over 300 km. Significant regressions for both specific conductivity and zooplankton biomass indicate remarkable location-by-location similarity even though the average levels by annum differed for both parameters (two-tailed t-test, n = 40, p < 0.01). Zooplankton followed a pattern we have previously noted (Yurista and Kelly, 2009; Yurista et al., 2009), with lower biomass concentrations at warmer temperatures. We infer that the difference in temperature between samplings of the 2 years is a strong contributor to the observed difference in average zooplankton levels between 2004 (colder, more zooplankton) and 2005 (warmer, less zooplankton) (Figure 6). We also tested spatial pairings for other parameters. Data for the 2 years for temperature were strongly correlated (p < 0.01), fluorescence was weakly correlated (p < 0.1), and there was not a significant correlation for beam attenuation. The data ranges for fluorescence and beam attenuation were rather small and the mean values were not clearly different between years.

Figure 5.

Pair-wise comparison at sample locations that overlapped spatially between the 2004 and 2005 data sets for specific conductivity and zooplankton biomass.

Figure 5.

Pair-wise comparison at sample locations that overlapped spatially between the 2004 and 2005 data sets for specific conductivity and zooplankton biomass.

Figure 6.

Zooplankton biomass plotted as a function of temperature for the sample locations that overlapped in 2004 and 2005.

Figure 6.

Zooplankton biomass plotted as a function of temperature for the sample locations that overlapped in 2004 and 2005.

Most relevant to this discussion is the observation that spatial pattern in biology along 300 km of the shoreline is retained in spite of inter-annual differences in other fundamental environmental drivers such as temperature. While it is hard to expect that any spatial pattern would or should repeat exactly over time, we suggest the analytical evidence is strong to support the notion of the landscape being a primary driver of spatial variability in the nearshore. General observations in spatial landuse character (Figure 7) support the multivariate regression modeling results given above that suggest landscape use and anthropogenic disturbance levels broadly influence the Lake Superior nearshore waters.

Figure 7.

Spatial distribution of landscape character (Ag-chemi- cal and Land cover PC values, Danz et al., 2005) in conjunction with the 2001–2005 specific conductivity and zooplankton biomass trend lines (LOWESS). The large stars indicate the PC values at the 5 largest US tributaries that comprise approximately 48% (interfluves included) of US drainage to Lake Superior.

Figure 7.

Spatial distribution of landscape character (Ag-chemi- cal and Land cover PC values, Danz et al., 2005) in conjunction with the 2001–2005 specific conductivity and zooplankton biomass trend lines (LOWESS). The large stars indicate the PC values at the 5 largest US tributaries that comprise approximately 48% (interfluves included) of US drainage to Lake Superior.

We analyzed the temporal variability across the years but with only limited overlap at any given site. The data generally did not have enough statistical power to attribute the magnitude of temporal variability to weather patterns, alongshore currents, inputs related to the largest tributaries, or the regional landscape character. Additionally, analyses with flow-weighted watershed contributions (areal based) could not be identified as distinct and persistent local signals in the nearshore region. Adjacent segmentsheds have similar land use character (Figure 7) and any specific tributary's overall contribution to the nearshore could not be distinguished using the present data. Our data for the 20 m contour and at the scale of our summarized observations may not be best suited to discerning contributions from localized flow variability to the magnitude of variability in the nearshore average values, but this does not lessen the observational evidence for a robust landscape-nearshore linkage at the scale our data is designed to assess. The scale of observational data presented here is the average condition across broad shoreline reaches of watersheds and the adjacent interfluves, which is at a much coarser scale than what would be included in a tributary plume. The dispersed impact of a plume is observed in elevated parameter values of surrounding waters while the detailed parameter levels within a plume thalwag may be missed in summarizing the data.

The data suggest that the landscape character appears to drive the character of the nearshore region. Nearshore water parameters were correlated with land use characterization. In regions of increased landscape stress, the nearshore waters had altered parameter values that were higher and more variable than in less developed areas. Landscape activities create stress to watershed processes and alter export to nearshore areas. In watershed regions of more pristine character and little disturbance to ecosystem processes, the nearshore conditions were relatively stable both across long spatial stretches and over longer temporal periods. Both the base level export and any entrapped or residual input (Csanady, 1970) due to storm/peak level export from watersheds appear to drive the magnitude and control the variability of conditions in nearshore waters at regional scales.

Monitoring of nearshore condition is a tractable problem when an appropriate high resolution approach to sampling is used. There is variability in nearshore waters, but we now have a qualitative understanding of how it is expressed spatially and temporally across a 5-year time span during the stable late summer period of the year. We have an expectation of what can be observed and a survey approach that yields the necessary high resolution data, continuous alongshore tows in the coastal boundary region (Yurista and Kelly, 2009). The nearshore condition is also correlated with landscape stressors that would be expected to affect water quality in the receiving areas. While the volume of the nearshore (<30-m depth) in Lake Superior is approximately 1% of the whole lake, the landscape signal tends to remain entrapped within the coastal boundary region (Csanady, 1970). Therefore any landscape signal in the nearshore would be approximately 100 times the signal that could be measured in the open offshore waters once it is fully incorporated into the lake. Given that the signal from the landscape activities can be detected reasonably well in the nearshore, the nearshore may provide early warning of anthropogenic input that could be addressed before the vast volume of the lake is affected in the long term. If Lake Superior were to be undesirably perturbed it would take a long time for remedial actions to restore whole lake condition, with a water turnover time of approximately 170 years (Schertzer and Rao, 2009). Nearshore monitoring could be a potential sentinel system for linking landscape input to impending whole lake change.

Conclusions

We have (1) combined multiple years of data in the nearshore of Lake Superior that represent 1200 km of shoreline, (2) observed an emerging pattern that relates to landscape position, (3) observed variability to be greater on spatial scales than on temporal scales, and (4) correlated water quality measures with landscape character. We described a conceptual framework that views landscape as a driver in mediating the robust spatial pattern observed in nearshore variability of Lake Superior. We suggest that high resolution monitoring of the nearshore area provides a means for early identification of stress to the whole lake.

Acknowledgements

This research was funded wholly by the US Environmental Protection Agency. The views expressed in this article are those of the authors and do not necessarily reflect the views or policies of the US Environmental Protection Agency.

This article not subject to United States copyright law.

References

Allan, J. D., Erickson, D. L. and Fay, J.
1997
.
The influence of catchment land use on stream integrity across multiple spatial scales
.
Freshwater Biol.
,
37
:
149
161
.
Auer, M. T. and Gatzke, T. L.
2004
.
The spring runoff event, thermal bar formation, and cross margin transport in Lake Superior
.
J. Great Lakes Res.
,
30
(
Supplement 1
):
64
81
.
Csanady, G. T.
1970
.
Dispersal of effluents in the Great Lakes
.
Water Res.
,
4
:
79
114
.
Danz, N. P., Regal, R. R., Niemi, G. J., Brady, V., Hollenhorst, T., Johnson, L. B., Host, G. E., Hanowski, J. M., Johnston, C. A., Brown, T., Kingston, J. and Kelly, J. R.
2005
.
Environmentally stratified sampling design for the development of Great Lakes environmental indicators
.
Envir. Monit. Assmnt.
,
102
:
41
65
.
Herlihy, A. T., Stoddard, J. L. and Johnson, C. B.
1998
.
The relationship between stream chemistry and watershed land cover data in the mid-Atlantic region, U.S
.
Water Air and Soil Pollution
,
105
:
377
386
.
Herman, A. W.
1992
.
Design and calibration of a new optical plankton counter capable of sizing small zooplankton
.
Deep Sea Res.
,
39
:
395
415
.
Herman, A. W., Beanlands, B. and Phillips, E. F.
2004
.
The next generation of optical plankton counter: the Laser-OPC
.
J. Plank. Res.
,
26
:
1135
1145
.
Hollenhorst, T. P., Brown, T. N., Johnson, L. B., Ciborowski, J. J.H. and Host, G. E.
2007
.
Methods for generating multi-scale delineations for indicator development in Great Lake coastal ecosystems
.
J. Great Lakes Res.
,
33
(
Special Issue 3
):
13
26
.
Isaaks, E. H. and Srivastava, R. M.
1989
.
Applied geostatistics
,
NY
:
Oxford University Press Inc.
.
Kelly, J. R.
2009
. “
Chapter 5.1. Nutrients and the Great Lakes nearshore, circa 2002–2007
”. In
Nearshore areas of the Great Lakes
Background paper, State of the Lakes Ecosystem Conference 2008. Environment Canada and U. S. Environmental Protection Agency, ISBN 978-1-100-13562-5, EPA 905-R-09-013, Cat. No. En164-19/2009E
Kelly, J. R., Yurista, P. M., Miller, S.E., Cotter, A.C., Corry, T.C., Scharold, J.S., Sierszen, M.E., Isaac, E.J. and Stockwell, J.D.
2011
.
Challengs to Lake Superior's condition, assessment, and management. A few observations across a generation of change
.
Aquat. Ecosyst. Health Mgmt.
,
14
(
4
):
332
344
.
Kerr, S. R. and Dickie, L. M.
2001
.
The biomass spectrum: a predator-prey theory of aquatic production
,
NY
:
Columbia University Press
.
Kireta, A. R., Reavie, E. D., Danz, N. P., Axler, R. P., Sgro, G. V., Kingston, J. C., Brown, T. N. and Hollenhorst, T.
2007
.
Coastal geomorphic and lake variability in the Laurentian Great Lakes: implications for a diatom-based monitoring tool
.
J. Great Lakes Res.
,
33
(
Special issue 3
):
136
153
.
Mackey, S. D. and Goforth, R. R.
2005
.
Foreword: Special issue on Great Lakes nearshore and coastal habitats
.
J. Great Lakes Res.
,
31
(
Supplement 1
):
1
5
.
Morrice, J. A., Danz, N. P., Regal, R. R., Kelly, J. R., Niemi, G. J., Reavie, E. D., Hollenhorst, T., Axler, R. P., Trebitz, A. S., Cotter, A. M. and Peterson, G. S.
2008
.
Human influences on water quality in Great Lakes coastal wetlands
.
Environ. Manage.
,
41
:
347
357
.
Rao, Y. R. and Schwab, D. J.
2007
.
Transport and mixing between the coastal and offshore waters in the Great Lakes: a review
.
J. Great Lakes Res.
,
33
:
202
218
.
Reavie, E. D.
2007
.
A diatom-based water quality model for Great Lakes coastlines
.
J. Great Lakes Res.
,
33
(
Special issue 3
):
86
92
.
Schertzer, W. M. and Rao, Y. R.
2009
. “
An overview of the characteristics of Lake Superior meterology, hydrology and physical limnology
”. In
State of Lake Superior
, Edited by: Munawar, M. and Munawar, I. F.
3
50
.
Burlington, ON
:
Aquatic Ecosystem Health & Management Society
.
Ecovision World Monograph Series
Sprules, W. G., Jin, E. H., Herman, A. W. and Stockwell, J. D.
1998
.
Calibration of an optical plankton counter for use in fresh water
.
Limnol. Oceanogr.
,
43
:
726
733
.
Stevens, D. L. and Olsen, A. R.
2004
.
Spatially balanced sampling of natural resources
.
J. Am. Stat. Assoc.
,
99
:
262
278
.
SURFER
.
2002
.
Surface mapping system
,
Golden Software Inc
.
SYSTAT
.
2004
.
The system for statistics
,
SPSS, Inc
.
Trebitz, A. S., Brazner, J. C., Cotter, A. M., Knuth, M. L., Morrice, J. A., Peterson, G. S., Sierszen, M. E., Thompson, J. A. and Kelly, J. R.
2007
.
Water quality in Great Lakes coastal wetlands: basin-wide patterns and responses to anthropogenic disturbance gradient
.
J. Great Lakes Res.
,
33
(
Special issue 3
):
67
85
.
Yurista, P. M. and Kelly, J. R.
2009
. “
Spatial patterns of water quality and plankton from high-resolution continuous in situ sensing along a 537-km nearshore transect of western Lake Superior, 2004
”. In
State of Lake Superior
, Edited by: Munawar, M. and Munawar, I. F.
439
471
.
Burlington, ON
:
Aquatic Ecosystem Health and Management Society
.
Ecovision World Monograph Series
Yurista, P. M., Kelly, J. R. and Miller, S. E.
2005
.
Evaluation of optically acquired zooplankton size-spectrum data as a potential tool for assessment of condition in the Great Lakes
.
Environ. Manage.
,
35
:
34
44
.
Yurista, P. M., Kelly, J. R. and Miller, S. E.
2006
.
Comparisons of zooplankton community size structure in the Great Lakes
.
J. Geophysical Res. Oceans.
,
111
C05S08, [doi:10.1029/2005JC002971]
Yurista, P. M., Kelly, J. R. and Miller, S. E.
2009
.
Lake Superior zooplankton: biomass: Alternate estimates from a probability-based net survey and spatially-extensive LOPC surveys
.
J. Great Lakes Res.
,
35
:
337
346
.
Yurista, P. M., Kelly, J. R., Miller, S. E. and Van Alstine, J.
2011
.
Lake Ontario: nearshore variability in water quality parameters.
.
J. Great Lakes Res.
,
DOI: 10.1016/j.jglr.2011.09.002