Sensor arrays based in a hydropower dam allow synoptic observations of water quality variations in a large river

Methods to observe changes in water quality in large rivers are constrained by differing water masses due to tributaries and nearshore influences as well as logistical challenges presented by seasonal constraints such as freezing in winter. A novel method was developed to monitor seasonal water quality at high temporal resolution in the Upper St. Lawrence River (annual average discharge of 7,250 m³·s⁻¹). The method utilized hydrodynamic modeling combined with upstream surveys using vessels to characterize water masses (via transverse and longitudinal transects) influencing water quality observed by automated sensor arrays and discrete water sample collection at fixed stations located in a run-of-the-river hydropower dam. The sensor locations in the dam were based on hydrodynamic modeling used to hind-cast water sources upstream that passed through the sensors that draw water from water turbine intakes. The advantages of the fixed (Eulerian) protocol with sensors in specific locations outweigh vessel and buoy-based sensor arrays since they have lower cost to operate, secure and safe access, simpler logistics, and provide year-round observational capacity for this large river that freezes over annually. The results demonstrate that this approach of characterizing water masses using a combination of hydrodynamic modeling with high spatial resolution water quality surveys for selection of fixed sensor locations can be applied to other hydropower dams in the Great Lakes region and large rivers elsewhere.

Graphical abstract

Graphical Abstract.

1 Introduction

Great Lakes-St. Lawrence River system represents 20% of the world’s surface freshwater, and is home to over 38 million people (Fergen et al., 2022). This globally significant ecosystem has a low ratio of water area surface to catchment area (0.32; Wang et al., 2015), resulting in its outflows, the rivers, strait, and fluvial lakes collectively referred to as Great Lakes connecting waters being strongly influenced by lake properties. The Great Lakes-St. Lawrence River system has been subject to sustained water quality monitoring for over fifty years. Article 5 of the 1972 Great Lakes Water Quality Agreement between Canada and the United States requires the two nations to establish water quality objectives. The research, surveillance and monitoring programs that were formally established by Canada and the United States in response to this treaty continue today. This binational effort has resulted in the ability to detect long-term changes in water quality of the Great Lakes related to pollution controls (e.g., Chapra and Dolan, 2012), land use (e.g., Mackie et al., 2022), aquatic invasive species (e.g., Dove, 2009), and climate change (Mahdiyan et al., 2021).

Approaches used for water quality surveillance and monitoring all have limitations (Behmel et al., 2016; Glasgow et al., 2004; Rundel et al., 2009). The traditional technique regularly employed by Canada and the United States in Great Lakes water quality sampling uses ships to access fixed stations in pelagic regions in spring and summer, where thermal profiles are observed and used to collect discrete or depth integrated samples. Remote sensing technologies, such as satellite imagery, are used to observe surface temperatures (Moukomla and Blanken, 2016), ice cover (Mason et al., 2016), and surface blooms of phytoplankton (Binding et al., 2020). These observations are limited by cloud cover and require some degree of ground-truthing. Buoys have been used to collect discrete samples and to house sensors capable of real-time data streaming, although they are available only during ice-free conditions and are often removed prior to the shoulder seasons bracketing winter. Some opportunistic observations have been made by using drinking water treatment facility intakes to acquire data for nearshore waters across all seasons (Nicholls et al., 2001), although there remain issues with fouling by dreissenid mussels in long intake pipes, which can affect water quality and plankton community composition, and the source of water can be affected by seasonally variable currents and seiches. Advances in technology have provided opportunities to utilize underwater automated vehicles, some capable of operation below ice, and other innovations such as networked subsurface buoy arrays (Verhamme et al., 2024), yet the costs and logistics of operation limit widespread use in the Great Lakes.

Water quality monitoring and surveillance has primarily occurred in pelagic waters of the Great Lakes, and not within nearshore regions, including the connecting waters. Moreover, observations are almost exclusively made during the spring and summer months. Connecting waters are ecologically significant, economically and socially important water bodies in the Great Lakes-St. Lawrence River system, with numerous threats to water quality but limited monitoring and surveillance programs for observations (Twiss et al., 2025). To address the need for water quality observations in large rivers in this system that satisfies both spatial (shallow and deep water locations) and temporal challenges (making year-round observations), the authors developed a protocol for water quality sensor array installations in a hydropower dam located in the St. Lawrence River, also known locally as Kaniotarowanenneh, in Kanienʼkéha (Mohawk language). The intention of this publication is to demonstrate the approach that uses existing commercially available instrumentation and provide examples of the advantages and limitations.

In brief, water is drawn using ambient pressure from penstocks that reflects water from both nearshore and main channel locations as determined by hydrodynamic modeling to distinguish water masses, and passes through a series of enclosed electronic sensors. At periodic intervals the sensors were cleaned, re-calibrated, and the opportunity used to collect discrete water samples for additional laboratory analyses. Possible applications for this approach are in the other connecting waters in the system that contain large hydropower dams (e.g., St. Marys River, Niagara River), as well as in other rivers to assess nutrient flux from catchments that have impoundments at outflows. A similar Eulerian observational approach as developed here can be applied to water intakes (e.g., drinking water, irrigation, cooling) in rivers.

2 Study sites and methods

2.1 Study site description

Fluvial Lake St. Lawrence (Figure 1) is one of the main geomorphic provinces along the Upper St. Lawrence River (Table 1) and the focus area of this study. This section of the river was created in 1958 by the construction of the Moses-Saunders run-of-the-river hydroelectric generating dam, the Long Sault control structure, and locks of the St. Lawrence Seaway; it is characterized by entry of river water at high velocity with subsequent slowing of the current with impoundment (Twiss et al., 2025). Lake St. Lawrence freezes over during winter (early January to late March). Annual discharge of the Upper St. Lawrence River at Massena, NY and Cornwall, ON is 7,250 m³·s⁻¹ (average of annual means for 1936–2024; U.S. Geological Survey, 2025).

Figure 1

Figure 1. Map of the upper St. Lawrence river (Kaniotarowanenneh). The geomorphic regions of the river: TIA, Thousand Islands Archipelago; STR, Single Thread Reach; LSL, fluvial Lake St. Lawrence.

Table 1

Table 1. Physical properties of geomorphic provinces within the Upper St. Lawrence River; $D=\frac{S}{\sqrt{4\mathit{\pi A}}}$.

2.2 Study objectives

The objective of this study is to compare the advantages and limitations of high resolution methods (Figure 2) used to characterize water masses in fluvial Lake St. Lawrence by: (1) conducting transverse transects across the channel; (2) conducting longitudinal transects along the shoreline at the 2 m isopleth and in the main channel along the entire reach of this fluvial lake; and (3) establishing a fixed sensor arrays in the Moses-Saunders hydropower dam that detect nearshore and main channel water quality continuously over annual temporal scales at high frequency resolution. The instrumentation used are not unique; they are all commercially available and can be substituted for by other instruments with similar capacity.

Figure 2

Figure 2. Schematic diagram of three techniques for observing water quality parameters in a large river. Dashed line indicates water flow direction (left to right). (A) Transverse channel transects; (B) longitudinal channel transects; (C) fixed location (*) observations. Techniques A and B follow geographic constraints (shore line and bathymetry) and are conducted from a vessel; observations in technique C take place at a geographically static point, such as a location in a hydropower dam.

2.3 Water sampling methodologies

The following describes two mobile and one stationary sampling protocols used to observe water quality in the St. Lawrence River. The mobile sampling protocols (§2.3.1, 2.3.2) were conducted prior to the establishment of the stationary protocol (§2.3.3). Details of the instrumentation used in each are found in §2.4.

2.3.1 Transverse transects across the Lake St. Lawrence channel

Water quality along the shores and the main channel of fluvial Lake St. Lawrence was measured using continuously recording sensors onboard a vessel in order to assess differences between near-shore and off-shore water quality. The RV Lavinia (25 ft. Boston Whaler) travelled at roughly 2.6 m·s⁻¹ (five knots) along each transect in a net upstream direction: the vessel crossed from shore to shore and turned toward the far shore when the vessel crossed the 2-m isopleth.

River water was pumped continuously from an in-hull portal (depth of 0.7 m) at a rate of 0.18 L·s⁻¹ though a fluorometer (TD model 10-AU; see below) into a ferry box (an 8-L opaque polyvinyl chloride cylinder) containing a phytoplankton-specific spectrofluorometer (FluoroProbe; see below) and a specific conductivity and temperature sensor (YSI model 600XL; see below), that overflowed back into the river. All instruments recorded time stamped data (10-AU at 1 Hz, YSI at 2 Hz, FluoroProbe at 0.8 Hz, GPS at 0.09 Hz). Instruments were provided with electrical connections from 12 VDC vessel batteries converted to 120 V/60 Hz AC. Three regions of the river (each ranging in area of 5–12 km²) were observed using this method on three dates between June 19 and July 7, 2011.

Surveying took approximately 3–6 h per region, and covered areas from −75°05′ east to −75°12′ east. Data were processed using Ocean Data View (ver. 4; Schlitzer, 2011) with kriging to interpolate water quality between vessel tracks. The spatial resolution was approximately 114 m determined by the residence time of the ferry box multiplied by the velocity of the vessel.

2.3.2 Longitudinal transects along the length of Lake St. Lawrence

As in the transverse transect observations, water quality along the shores and the main channel of fluvial Lake St. Lawrence was measured using continuously recording sensors onboard a vessel moving upstream as described above (§2.3.1.). Because data were recorded at different intervals, a MATLAB (MathWorks, Natick, MA) code was created to compile the data values into 2 s intervals, with the latitude and longitude values interpolated between the recorded points. Latitude and longitude were recorded continuously using navigational software (Offshore Navigator, ver. 5.07, Maptech; Amesbury, MA).

The 2-m isopleth marked the path for the near-shore transects, while the off-shore transect followed the path of the greatest water velocity as determined by 2-D hydrodynamic model output (River 2-D; Steffler and Blackburn, 2002). Longitudinal transects were approximately 45–50 km long, and started at approximately −74°49′ east near the Moses-Saunders hydropower dam and shipping locks, and ended at the Iroquois dam, −75°18.5′ east. The transect along the north shore 2 m isopleth was conducted on June 29; a transect along the south shore 2 m isopleth was conducted on July 3; and the main channel transect was completed on July 4 (all dates in 2012). The south shore transect did not begin further east of −74.9° east due to a fine clay suspension in the water attributed to operation of the Wiley-Dondero shipping channel that caused spurious responses by the FluoroProbe.

2.3.3 Fixed sensor arrays in the Moses-Saunders hydropower dam

The Eulerian design to sampling by placing sensors in fixed locations required a priori knowledge of which river water mass was being observed by the sensors. In brief, hydrodynamic modeling was used to determine the extent that water upstream was sensed while traversing the hydropower dam sensor arrays. This characterization of “what upstream water was reaching the sensors” involved a two-step modeling process: (a) use the River2D model to determine time-varying currents, then (b) use those estimated currents to model the path of an ensemble of particles. The key inputs required to drive that first model include a bathymetric description of the water bed (topography and roughness), total water flow at the input (upstream) boundary, and water surface elevation at the discharge sections. The output of that first model is a gridded set of time-varying current over the modeled area.

Water currents in the Upper St. Lawrence River were simulated every 3 h using the Upper St. Lawrence River Forecasting System, a real-time hydrodynamic model formerly integrated into the Great Lakes Coastal Forecasting System of the US National Oceanographic and Atmospheric Administration (NOAA). Using historical data to prescribe the boundary flows and water heights for the period covered by a particular sample collection, the River2D model estimates a time varying depth-averaged representation of the velocity field, which is denoted as:

$\nu =\mathbf{u}(x,t)$

where 𝒗 describes the velocity vector at time 𝑡 at position 𝒙 in the river (where a 2-d vector 𝒙 gives the position measured in Universal Transverse Mercator coordinates). Taking a Lagrangian perspective of the flow, individual fluid parcels are considered to be governed by that velocity field, where the fluid parcel is observed as it moves in time. Taking 𝒂 to be the position of the fluid particle at time 𝑡0, the position of the particle is given by X(𝒂,t), and its evolution is related to the velocity field by the relationship.

$\begin{array}{ll}d\frac{X(a,t)}{\mathit{dt}}=\mathbf{u}\left(X\right(a,t),t)& (1)\end{array}$

such that the fluid parcel movement is governed by the velocity field. Numerical integration was used to solve Equation 1, where a stochastic term was incorporated to account for dispersion of the flow. This particle trace simulation used an integration time step of 60 s, and then validated by reducing the time step to 2 s and verifying that the original simulation values were within acceptable tolerances. As the primary goal was ensemble characterization rather than any specific trace, the 60-s time step was sufficient for our qualitative analysis.

For a particular sample 𝑖, taken at location 𝒂𝑖, an ensemble of locations was chosen from a bivariate normal distribution centered on the measured sample location. For each point in the ensemble, its position backwards in time used the time varying flow field, e.g., looking 72 h into the simulated flow. For each point in the ensemble its velocity at each of those hourly time marks was determined. In other words, for the sample location in the hydropower dam, it was determined: 1) where the water had come from over the previous 72 h; and 2) what the velocity history of each water parcel was. The ensemble of velocity measurements is 𝒂{𝑣𝑖}. This sample is interpreted as describing the distribution of velocity history for the water mass in the vicinity of the sample location. An example of this model output is provided in Figure 3. Based on these calculations, sensor installations (Figure 4) were established to observe water at turbines Unit 32 (southern nearshore, established June 2014), Unit 17 (main channel water, established June 2017), and Unit 01 (northern shore, established May 2017).

Figure 3

Figure 3. Map of fluvial Lake St. Lawrence (Upper St. Lawrence River) showing 72-h hind cast of modeled water locations that passed through three fixed locations on the Moses-Saunders hydropower dam (dam). River water flow is from the southwest to the northeast. North shore (red), main channel (green) and south shore (blue) denote water masses that pass through turbines (Units) 32, 17, and 01, respectively, in the Moses-Saunders hydropower dam.

Figure 4

Figure 4. (Left) Photograph of the Moses-Saunders hydropower dam showing fixed sensor locations (view from northeast). For scale, the dam is 980 m wide. Source: Ontario Power Generation. (Right) Cross section of the Moses-Saunders hydropower dam showing the location of fixed sensor arrays drawing water from the reservoir prior to passing through the turbine. Hydraulic head from reservoir surface to turbines is 25 m. Source: New York Power Authority Visitor Center.

River water for sensors entered through a 1.27 cm diameter stainless steel tubing from high flow stator cooling pipes (30 cm diameter) that drew water from the turbine scroll case, prior to passing through the turbine (Figure 4). Water pressure was regulated using a ¾ inch gate valve and piped at 10 L·min⁻¹ through ½ inch braided nylon tubing to the C6 Multi-sensor (Turner Designs) contained in a water-tight flow-through housing then to conductivity and temperature sensors (YSI Model 600XL) contained in a flow-through housing, and disposed of in a scupper. All tubing was insulated with water pipe insulation, and the instrument housings were insulated with ¾ inch neoprene sheets. The instruments and accompanying portable computer were secured on shelves bolted to a bulkhead and used a 120 V/60 Hz AC electrical connection.

The sensor locations were used to measure water quality from both high frequency observations (0.017 to 0.033 Hz) and discrete water sample analysis (see §2.5) from 1-L samples collected in acid-cleaned polycarbonate bottles at 2–3 week intervals. Water samples were collected in acid-cleaned polycarbonate bottles to analyze for size-fractionated chlorophyll-a (Chl-a), nutrients (total phosphorus, dissolved nitrate, dissolved silicate), major anions (sulfate, chloride), and phytoplankton community composition using a FluoroProbe. Sensor data collection in the fixed locations was coordinated to Greenwich Mean Time.

2.4 Instrumentation

The following is a description of the instruments used, their maintenance, and calibration protocols.

C6 Multisensor Platform, with Cyclops 7 sondes (Turner Designs, Sunnyvale, California) — The sensors of this instrument were housed in an opaque flow-through housing and located at Units 32 and 01. The fluorometric sensors measured in vivo chlorophyll-a, in vivo phycocyanin, and coloured dissolved organic matter (CDOM). Turbidity was measured using a nephelometer (light back-scatter), and a thermocouple measured water temperature. The instrument contained an automated wiper to clear biofouling on the sensors: the wiper made two rotations immediately prior to measurements being made. The wiper motor was replaced after 4 years of operation. The instrument was controlled with a C-Soft computer interface (Turner Designs, Sunnyvale, California). Calibration was conducted at each sampling and cleaning interval (2–4 weeks) with solid standards (chlorophyll-a [3.79 μg·L⁻¹], phycocyanin [3.74 μg·L⁻¹], CDOM [4.69 μg·L⁻¹]), and a liquid standard (10 nephelometric turbidity units [NTU] turbidity standard (diluted from a 1,000 NTU solution; AMCO Clear Turbidity Standard, GFAS Chemicals, Powell, Ohio). The solid standards were calibrated with known chlorophyll-a concentrations determined by the spectrophotometric trichromatic method, a solution of phycocyanin prepared from powdered pigment isolated and provided by G. Boyer, SUNY-ESF, Syracuse, New York) and quinine sulfate (Sigma-Alrich, St. Louis, Missouri) was used to calibrate the CDOM solid standard. The following are the minimum detection limit (DL; ppb) and linear range (ppb) for each analyte determined by the C6: CDOM, 0.1, 0–1,500; chlorophyll-a, 0.03, 0–500; phycocyanin, 2, 0–4,500; and turbidity had a minimum DL of 0.05 NTU and is linear from 0–1,500 NTU (Turner Designs, 2025). Deionized water was used to wash and blank the instrument during calibration.

YSI Multi-Parameter Water Quality Monitor, Model 600XL and YSI Model ProPlus (Yellow Springs Instruments [YSI], Yellow Springs, Ohio) — These instruments measured specific conductivity and temperature at Unit 32 and Unit 01, respectively. The instruments were housed in a flow-through housing and interfaced with a computer through Ecowin and Ecowatch Lite software (Yellow Springs Instruments, Yellow Springs, Ohio). The specific conductivity was calibrated at 2–4 week intervals using a 1.413 mS·cm⁻¹ conductivity solution (0.01 M KCl; Labchem Inc., Zelienople, Pennsylvania) following a thorough cleaning with deionized water.

YSI Model 6,920 V2-2 (Yellow Springs Instruments, Yellow Springs, Ohio) — This instrument was used to measure in vivo chlorophyll-a, specific conductivity, turbidity, and temperature, with an automated wiper blade to clear biofouling that rotated once prior to each measurement. The standard for chlorophyll-a calibration was a solution of rhodamine red; all other standards used are listed above.

TD 10-AU (Turner Designs, Sunnyvale, California) — This instrument located at Unit 17 was used in flow-through mode to measure CDOM with an internal standard calibrated to Suwannee River fulvic acid (International Humic Substances Society, St. Paul, Minnesota). Data were collected in the internal flash memory in the instrument and downloaded as text files. At 2–4 week intervals, the quartz flow-through cell was cleaned with a nylon brush and de-ionized water.

FluoroProbe (bbe Moldaenke, GmbH, Germany) — Phytoplankton community composition was assessed using a phytoplankton pigment-specific spectrofluorometer (FluoroProbe) equipped with a magnetically stirred 25-mL quartz sample cell (Work Station 25) on a dark adapted (>15 min) sample. The background fluorometric signature of each discrete sample collected from the power dam was determined using river water filtered using a syringe filter (<0.2 μm) from that specific sample. For the continuous sensor investigative surveys (see §2.3.1. and §2.3.2.), a single main channel water sample (filtered <0.2 μm) was used for the background fluorimetric signature for the main channel surveys (transverse sampling and main channel of the longitudinal survey), whereas a single nearshore water sample collected near the mouth of a tributary (Coles Creek; 44.8911°N, −75.1152°E) was used to represent the nearshore environment (e.g., along the 2 m isopleth; §2.3.2.). Glass surfaces (lenses of the FluoroProbe and the quartz cuvette used for discrete samples collected when visiting the sensor locations in the hydropower dam), were washed after use with a mild detergent, and rinsed with deionized water and a lint-free wipe.

The FluoroProbe identifies phytoplankton classes based on precalibrated excitation and emission spectra fingerprints programmed into the instrument, namely (i) Chlorophyta and Euglenophyta; (ii) phycocyanin (PC)-rich Cyanobacteria; (iii) Heterokontophyta, Pyrrophyta, and Haptophyta (note: there are no Haptophyta in the Great Lakes); and (iv) the phycoerythrin (PE)-rich Cyanobacteria and Cryptophyta. Water samples were kept in the dark for at least 15 min prior to measurement; in the field (§2.3.1 and §2.3.2), the period of darkness was reduced to the turnover time of the ferry box (ca. 45 s) since all transects were conducted during daylight.

2.5 Chemical analyses

Total phosphorus and silicate: Phosphorus content was determined for duplicate 35 mL unfiltered water samples stored in borosilicate glass tubes. A fresh solution of potassium persulfate was added to each sample and the calibration standards to achieve a final concentration of 0.7% (mass:volume), and then digested in an autoclave (10 min at 121 °C). Phosphorus was determined colorimetrically using the antimony-molybdate-tartrate method (Wetzel and Likens, 2000) and light absorption was measured (885 nm) using a 10 cm pathlength quartz cuvette. Dissolved silicate was determined colourimetrically using the molybdate method (Wetzel and Likens, 2000), at 815 nm in a 5 cm pathlength cell. River water was filtered (<0.2 μm pore size) using a syringe filter and retained in a polypropylene tube at 4 °C in the dark prior to analysis.

Dissolved anions: River water was filtered through a 0.2-μm pore-sized syringe filter into a polypropylene container and kept at 4 °C in the dark until analysis of chloride, sulfate and nitrate by ion chromatography at the Center for Air and Aquatic Resources Engineering and Sciences, at Clarkson University, Potsdam, New York.

Extracted chlorophyll-a: Size fractionated chlorophyll-a concentrations in river water were analyzed fluorimetrically following extraction of seston captured on filters of specific porosity. Parallel filtrations using 47-mm-diameter filters were conducted in duplicate within 2 h of collection using the following sequence: approximately 100 mL were collected onto a 20-μm pore-size woven nylon filter; approximately 80 mL were collected onto a 2-μm pore size polycarbonate filter; and approximately 70 mL were collected onto a 0.2-μm pore size polycarbonate filter. This approach provided details on the abundance of chlorophyll-a in the microplankton (>20 μm), nanoplankton (2–20 μm), and picoplankton (0.2–2 μm) size fractions, respectively. Filters were extracted in 10 mL of 90% acetone in the dark at 4 °C, for 8–24 h and chlorophyll-a was measured by fluorimetry (Welschemeyer, 1994) using a calibrated fluorimeter (Turner Designs model TD-700; Sunnyvale, CA).

2.6 Details of precision and accuracy and limits of detection or quantification, where applicable

There was a constant decrease in the values of data collected by the TD 10-AU used to observe CDOM for extended periods (days) at Unit 17 (none of those data are presented here). The internal quartz flow-through cuvette had no internal anti-fouling mechanism to prevent matter from adhering to the quartz surface, and this fouling was suspected to impede excitation and emission light. Thus, sensors that require light must have antifouling devices to ensure data veracity. An acceptable application is the use over 4–6 h during the vessel-based surveys; no appreciable biofouling is expected over such a brief time. Sensors such as thermocouples and electrodes (used to measure specific conductivity) are not subject to the same degree of interference.

The C7 sensors of the C6 Multiparameter instruments at Units 32 and 01 were cross-calibrated by using the C7 solid standards for chlorophyll-a, phycocyanin, and CDOM to calibrate the C6 instrument used at Unit 01. The solid standards for that instrument were then determined in units equivalent to the Unit 32 instrument and used subsequently for calibrations at Unit 01.

Observation of standard values measured by instruments (C6 and YSI at Unit 32) calibrated 5 to 19 days prior showed no significant (regression not significantly different from zero at p < 0.05) drift in the accuracy of the instruments.

3 Results from vessel-based and fixed location observational protocols

3.1 Observations from transverse transects

The river can be characterized by separate zones, some of which are localized and influenced highly by tributary inputs, while others are considerably larger, such as zones that are part of the main channel river flow. Typical of the season pattern for the time of year during this data collection, tributary inputs are characterized by areas of high water temperature, (Figure 5A) and high CDOM (Figure 5B). However, two of the tributaries surveyed here (Brandy Brook and Houasic Creek) diverge strongly with greater specific conductivity of Houasic Creek water and lower specific conductivity of Brandy Brook waters, compared to the main channel river water (Figure 5C).

Figure 5

Figure 5. Surface water quality of a 4 km stretch of fluvial Lake St. Lawrence showing distinct water mass properties. Data collected from a vessel moving at 5 knots upstream (right to left) in continuous transverse transect directions. (A) Water temperature. (B) Coloured dissolved organic matter (CDOM; Suwannee river fulvic acid equivalents). (C) Specific conductivity.

Low discharge from tributaries limited nearshore impact on downstream nearshore water quality. In addition, extensive shallow waters (< 1 m deep) several km downstream of Houasic Creek make it difficult to detect any tributary influence at the 2 m isopleth, which was 300–600 m offshore (south) of the shoreline. The CDOM present in the main channel along the south shore is attributed to inputs from the Oswegatchie River (Figure 1), which had a noticeable effect on CDOM concentrations along the south shoreline nearshore region of the St. Lawrence River up to 70 km downstream from this main tributary (see Ball et al., 2018, and §3.3).

3.2 Observations from longitudinal transects

Transects along the length of fluvial Lake St. Lawrence show a high degree of variability along the 2 m isopleths on both the north and south shore lines and relatively invariant change in parameters in the main channel (Figure 6). Shapiro-Wilkinson test for normality of all observed variables for all transects showed that data sets varied significantly (p < 0.001) from a normal distribution; however, the main channel water had smaller variance in data (Table 2). Levene’s test comparing variances of the observed variables showed that all variables had significantly different (p < 0.001) variances between observations made on the north versus south shoreline transects. These observations support the notion that water in the main channel of the river has homogenous properties while the nearshore areas, delineated here as the 2 m isopleth, are heterogeneous and influenced by their respective nearshore features, such as embayments, shallow reaches, and tributaries.

Figure 6

Figure 6. Surface water quality of a 40 km stretch of fluvial Lake St. Lawrence shows distinct water quality along longitudinal transects of the nearshore (north, red; south, blue) compared to the main channel (green). Data (smoothed to 30 s intervals; rolling average) collected from a vessel moving continuously upstream (right to left) at 5 knots. (A) Water temperature. (B) Specific conductance. (C) Coloured dissolved organic matter (CDOM; Suwannee River fulvic acid equivalents). (D) In vivo chlorophyll-a fluorescence.

Table 2

Table 2. Variability of water quality parameters measured along the main channel and nearshore transects (approx. 40 km) of fluvial Lake St. Lawrence (St. Lawrence River) measured during three separate days.

The average residence time of water in the Upper St. Lawrence River is 11.3 days (volume/average annual discharge). Since the average transit time in the main channel is estimated to be 6.2 days based on the average velocity in the main channel (0.3 m·s⁻¹; Twiss et al., 2025) it follows that appreciable amounts of water are retained in slackwater areas adjacent to the shore and within archipelagoes. The 3-day hind-cast of water origin (Figure 3) passing through the Moses-Saunders dam supports this statement, as illustrated by the greater distance of water travelled in the main channel versus the nearshore regions.

3.3 Fixed location sensors: Eulerian observation of flow fields

3.3.1 Spatial detection

Over the period of 2014 to 2018, the Oswegatchie discharge was 0.71 ± 0.68% of the Upper St. Lawrence River discharge. The sensor array at Unit 32 detected waters from the Oswegatchie River, which enters the Upper St. Lawrence upstream of the Moses-Saunders hydropower dam. Despite the river’s relatively small contribution to total flow, this entry of highly coloured, CDOM-rich water draining the Adirondack Mountains was still distinguishable in the receiving waters. The evidence for this capacity is shown in Figure 7. Whereas CDOM concentrations and the discharge of the Upper St. Lawrence River were positively correlated, r (896) = 0.019, p < 0.00001 (Figure 7A), that relationship is very weak when compared to the correlation with CDOM observed at Unit 32 that was more closely correlated with discharge of the Oswegatchie River, r (794) = 0.24, p < 0.00001 (Figure 7B). Thus, the effect of the Oswegatchie River discharge on nearshore slackwater regions is maintained for at least 68 km. This retention of water mass is characteristic of large rivers (Frenette et al., 2003), wherein the large mass of water and depth restricts immediate full mixing of tributaries.

Figure 7

Figure 7. Coloured dissolved organic matter (CDOM) observed at the Unit 32 turbine, closest to the south (New York) shoreline of the St. Lawrence River showing (A) weak correlation of observed CDOM concentrations with St. Lawrence River discharge and (B) closer correlation with discharge of the Oswegatchie River, a tributary approximately 68 km upstream of the hydropower dam on the south shore.

The two nearshore stations (Unit 01 and Unit 32) reflected heterogeneity as observed using the longitudinal transects (§3.2). The 2017 freshet created a strong influence on water quality in the south shore station Figure 8). At Unit 32, the influence of the Oswegatchie River, low ionic strength with high CDOM content draining a portion of the Adirondack Mountain watershed (Driscoll et al., 1991), is evident with a decrease in specific conductivity, concomitant with an increase in CDOM concentration detected at the hydropower dam.

Figure 8

Figure 8. Water quality of the spring freshet of 2017 observed from the south (New York) nearshore region of the Upper St. Lawrence River.

3.3.2 Temporal resolution

The choice to conduct long-term (multiple-year) observations at high resolution (0.017 to 0.0028 Hz) provided several advantages. Firstly, it showed the seasonal changes in water quality from year to year. For a large river such as the Upper St. Lawrence, the influence of solar heat flux on a large scale (e.g., the headwater at Lake Ontario; surface area 18, 960 km²) dominates seasonal changes in temperature (Figure 9), with slight year to year changes due to weather. Of interest to users of the St. Lawrence River as a drinking water source, proliferations of potentially toxigenic phycocyanin-rich Cyanobacteria (Plaas and Paerl, 2020) follow a predictable pattern with increases taking place with increasing water temperatures and a presence lasting into October, after river water temperature has decreased (Figure 9).

Figure 9

Figure 9. Long-term observation period at the south shore fixed observation station depicting the annual elevation of phycocyanin (PC)-rich cyanobacteria concentrations (using the phycocyanin fluorescence measured continuously as a proxy for PC-rich cyanobacteria) in the Upper St Lawrence River. Values are daily averages.

The fixed sampling stations provided the opportunity to collect discrete river water samples year-round during routine (2–3 week intervals). Analysis of total chlorophyll-a (a proxy for phytoplankton biomass) from the three fixed stations revealed similar seasonal trends in chlorophyll-a concentrations with lowest values during the months of November and February (Figure 10). A statistical analysis of chlorophyll-a concentrations from Units 32 and Unit 17 that were sampled on the same dates showed that despite greater total phosphorus concentrations at the nearshore station (Unit 32) there was no significant difference in total chlorophyll-a concentrations between the nearshore and main channel waters (paired t-tests, p < 0.05, N = 13).

Figure 10

Figure 10. Concentrations of total chlorophyll-a in river water over time from the three sensor locations in the Moses-Saunders hydropower dam, Upper St Lawrence River. Values are based on solvent-extracted measurements of particles >0.2 μm. (A) Water collected at Unit 32. (B) Water collected at Unit 17. (C) Water collected at Unit 01.

Analysis of anion concentrations in discrete water samples from Unit 32 (Figure 11) showed annual variability with an additional effect of the flooding in 2017, which diluted nitrate (Figure 11A), chloride, and sulfate (Figure 11B) concentrations. A latent effect was observed in the following year (2018) with increased nitrate concentrations, whereas chloride and sulfate concentrations remained relatively stable in comparison with other non-flood years. The molar N: P ratio (total phosphorus to dissolved nitrate) increased in 2018 and deviated from the relatively constant pattern seen in previous years (Figure 11C).

Figure 11

Figure 11. Nutrient and anion concentrations observed at the south nearshore fixed station in the Moses-Saunders hydropower dam, Upper St. Lawrence River over a four-year period. (A) Total phosphorus (circles) and dissolved (<0.2 μm) nitrate (triangles). (B) Silicate (squares), chloride (diamond), and sulfate (circle). (C) The molar ratio of N (measured as nitrate) to phosphorus shows a latent impact on water quality following the flood year (spring and summer 2017) on Lake Ontario and the St. Lawrence River.

4 Discussion of the assessment of the various protocols for observing water quality in large rivers

4.1 Advantages and limitations

The three observational protocols were compared with respect to spatiotemporal properties as well as logistical considerations (Table 3). The two vessel-based protocols (transects) provided the opportunity to make measurements during ice-free conditions with high spatial resolution, however, the cost to conduct these assays would be high due to labour and fuel costs, and the ability to cover large sections of the river would be constrained because of the time required to complete the transects, and the logistical constraints (movement of equipment, weather, daylight).

Table 3

Table 3. A comparison of attributes of river observation techniques that predispose them for various environmental monitoring and surveillance needs.

The time to gather the data depicted in Figure 10 is estimated to be 138 5-h efforts to gather samples (including 1.5 h travel time), process and analyze water samples, clean and re-calibrate instruments; this is equivalent to 86 8-h single person work days, over a 5 year period. Note that visits to U32 and U17 occurred simultaneously. The data acquired from a 1 day effort by two people for the transverse transect would require approximately 58 days to survey the entire Upper St. Lawrence River (696 km²; Table 1) at 12 km² per day under suitable weather conditions. Given the temporal changes evident in water quality over a 2 month period, the transverse transect sampling protocol would be comparatively expensive and difficult to interpret due to the time required to make the observations. Similarly, the longitudinal transects (also requiring a two person per day effort) would require approximately 12 days to cover the estimated 380 km of 2-m isopleths (north, and south) and 140 km of main channel river water. Thus, the cost (labour and transportation to and from the laboratory and on the water) of both vessel protocols would be much greater than that of the fixed station protocol. For example, the operational cost of the three fixed stations over a five-year period, primarily involving technician time for bi-weekly maintenance was approximately 82 8-h work days; that work effort would only be able to accomplish 8 complete transverse surveys of the entirety of Lake St. Lawrence (126 km²) and only during the ice-free state.

The fixed stations also provide protection of instrumentation and a safe and stable working environment for technicians. One uncontrollable constraint in the fixed station is the adherence to hydropower dam operations. An interruption of 4 months occurred in 2016 due to closure of the turbine for large-scale maintenance and repairs.

4.2 Possible pitfalls and artifacts: caveats

Regular interruptions (weekly during late summer and autumn) of water flow for the fixed stations occurred for up to 3 h to allow for personnel to clean out filters used to trap material, e.g., fish, senescent submerged aquatic vegetation, from entering the stator cooling apparatus. Due to the constant and warm temperature (25 to 30 °C) in the hydropower dam, the stoppage of water flow was detectable as an increase in water temperature in the flow through instrument housings. Using the R program (RStudio version 3.3.0, © 2016 The R Foundation for Statistical Computing Platform) custom scripts for automated processing of the raw data were created to produce datasets that removed water quality measured by the instruments during flow stoppages. An important component of a water quality surveillance system is the integration of observations with data management and control that involves data verification and quality assurance. Our focus here was the development of the observational protocol for making year-round observations in a large river; integration of such an observational platform into a data visualization and communication network is reviewed by Chen et al. (2025) who discuss various advantages and disadvantages of remote monitoring systems, such as those described here.

Since this sensor array relies upon in vivo fluorimetry for several parameters (viz., chlorophyll-a, coloured dissolved organic matter, phycocyanin), care must be exercised in the use of discrete values from these instruments. Using high temporal resolution provides the ability to assess both value variability and the influence of factors controlling the values. For example, a higher resolution of data collected at Unit 32 within the same period as depicted in Figure 9, show diel changes in temperature (explainable by solar flux) and fluorometry of pigments (chlorophyll-a) and fluorochromes (CDOM) (Figure 12). The diel variations in fluorescence yield are attributed to photoquenching of chlorophyll-a and CDOM and not irreversible photobleaching (Osburn and Morris, 2003; Rousso et al., 2021) as observed by decreases during daylight hours and recuperation of elevated values during night. One advantage to high temporal frequency of measurements during nighttime is that it avoids photo-quenching, and at a time that is logistically challenging for vessel-based surveillance. There exist model corrections for adjusting in situ chlorophyll-a fluorescence to extracted chlorophyll-a concentrations (Sharp et al., 2025) that require ambient light flux; such an approach would be useful if chlorophyll-a concentrations are needed at a frequency greater that the 2–6 week intervals between site visits to collect river water for chlorophyll-a extractions that were done here. Decreases in CDOM as the season progresses from spring to summer may be due to both photobleaching of the source water (Lake Ontario) and reductions in tributary loading attributed to hydrology and influences of microbial activity (Shousha et al., 2022).

Figure 12

Figure 12. High resolution of water quality recorded at the Unit 32 turbine location, which observes the nearshore water quality at the southern shore of the Upper St. Lawrence River at the Moses-Saunders hydropower dam. Data are recorded from June 18–23 at one-minute intervals: symbols at the top (red) are river water temperature, at the middle (brown) are coloured dissolved organic matter, and at the bottom (green) are chlorophyll-a. Values are data recorded at 1 min intervals (Greenwich Mean Time); chlorophyll-a and CDOM are fluorimetric determinations.

Operating water quality sensor arrays in a hydropower dam requires sensitivity to the operators of the dam and their principle objective: to safely and efficiently produce electricity. Arrays had to be secured in bulkheads that avoided food traffic and used 120 V/60 Hz power that is periodically required by hydrodam workers. In addition, interruption of water supply due to normal dam operations (see above) required a post data collection protocol to account for the interruption rather than requesting hydropower dam workers to manipulate instruments or record the interruptions.

Real-time data collection and communication from this large hydropower dam would require telecommunication that is restricted due to the concrete, steel and copper structures that interfere with radiofrequencies making cellular modems non-functional. Using existing cyber networks in a hydropower dam is a cybersecurity risk that would not be allowed by the operators. Thus, any real-time observation would require installation of a separate dedicated cyber network (cables to the dam exterior).

We acknowledge that comparing the transverse (2011) and longitudinal (2012) and observations with the fixed station observations (2014–2020) is fraught with the different time periods studied. However, notwithstanding the 2017 flood year and its latent effect on some water quality variables, interannual variability of water quality in the Upper St. Lawrence River is low. This is likely due to the fact that this is a water-level regulated river and the influence of a very large water mass as its headwater (Lake Ontario).

The flooding of 2017 likely caused some change in hydrodynamic patterns, particularly in nearshore regions. Some minor river bed modification (erosion, deposition) was observed, as evidenced by observations of large (several m²) mats of emergent aquatic vegetation (Typha spp.) that were scoured and transported downstream. Determining the extent of river bed modification would require high resolution bathymetric surveys to determine if the morphometry was altered. Given the magnitude of this straight, non-meandering river (Table 1) it is unlikely that the flood created any profound river bed modification.

4.3 Summary of the methodology

The development of fixed station water quality monitoring stations in the hydropower dam evolved from the spatial, temporal, and logistical challenges and the expense required to routinely observe large areas of a very large river. This is the first known application of sensor arrays established in hydropower dams to measure water quality as it passes through the dam. This approach may prove useful in other run-of-the-river hydropower plants, such as the three hydropower dams on the St. Marys River (known regionally as Baawaating in Anishinaabemowin language) that drains Lake Superior, the other location in the Great Lakes where dams regulate upstream water levels. Moreover, the application of this approach could be used for water quality surveillance in other large rivers elsewhere that have dams to provide the platform for sensor installations, e.g., the Rhine, Rhône, Volga, Angara, Nile, Mekong, Yangtze, and Tietê rivers.

In this study, the transverse and longitudinal vessel-based surveys were conducted first to better understand this large river system. We concluded that the fixed stations in the hydropower dam were the best option for long-term surveillance of water quality. Nevertheless, it might be advantageous to periodically use vessel-based surveys to probe potential influences on water quality or the impact of specific events, such as flooding of tributaries.

An important component of our approach is the hydrodynamic modeling that hindcasts where water originates that is detected at different locations in the dam. In rivers, water masses do not always fully mix due to differences in water velocity and depth, and retention by substrates (Elhadi et al., 1984). The vessel-based observation protocols were invaluable to demonstrate the extent of water quality heterogeneity and influencing factors in the nearshore zones and the relative homogeneity in the main channel. Thus, knowledge of differing water masses is required in order to design an appropriate water quality detection system.

The advantages of high spatial resolution of observations provided by the transverse and longitudinal protocols are lost in light of the logistical constraints and costs required for long-term measurements possible with the fixed stations. In addition, the fixed stations produce very large data sets amenable to statistical analysis of data structures to help explain the natural phenomena that are being observed, such as scale-dependent processes that increase predictive power and interpretation of ecosystem function (Mimouni et al., 2020; Mimouni et al., 2021). One direct application of the fixed station approach in the Upper St. Lawrence River is the observation of relationships between water quality variables and water levels as a new water level regulation plan (Plan 2014) is put in place in an effort to restore ecosystem damage due to overly controlled water level regulation in the past (Clamen and Macfarlane, 2018), in addition to climate change impacts on water quality. The fixed stations can monitor water quality at all times of day each day of the year, and thus provide insight into water quality at times of the year that are not routinely surveyed (Xenopoulos et al., 2025).

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

MT: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. JS: Conceptualization, Formal analysis, Methodology, Project administration, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. JR: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Writing – review & editing. SL: Formal analysis, Methodology, Writing – review & editing, Investigation. HS: Formal analysis, Investigation, Methodology, Writing – review & editing. FN: Formal analysis, Investigation, Methodology, Writing – review & editing. CL: Formal analysis, Investigation, Methodology, Software, Writing – review & editing. AR: Formal analysis, Investigation, Methodology, Writing – review & editing. EM: Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – review & editing. LS-P: Data curation, Investigation, Methodology, Writing – review & editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. The authors declare that financial support was received for the project from the following grants to MT: the St. Lawrence River Research and Education Fund (New York Power Authority); the New York State Water Research Institute; New York’s Great Lakes Basin Small Grants Program (New York Sea Grant); and the Beacon Institute for Rivers and Estuaries (Clarkson University); SL, HS, FN, and CL were funded by the US National Science Foundation -Research Experience for Undergraduate Program (award no. EEC-1359256). The Ontario Ministry of the Environment and Climate Change, the Five B Family Foundation, and the National Research Council Canada -Industrial Research Assistance Program provided support for a research fellowship for EAM.

Acknowledgments

We thank Jonathan Mayette for providing safe and secure access to the hydropower dam, and Sean Doyle and Jeffrey Farrell (New York Power Authority) for assistance with permitting and logistical support; we thank Carmen Ulrich for assistance with GPS analysis and Matthew Windle for creating figure one.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

Ball, E. E., Smith, D. E., Anderson, E. J., Skufca, J. D., and Twiss, M. R. (2018). Water velocity modeling can delineate nearshore and main channel plankton environments in a large river. Hydrobiologia 815, 125–140.

Google Scholar

Behmel, S., Damour, M., Ludwig, R., and Rodriguez, M. J. (2016). Water quality monitoring strategies—a review and future perspectives. Sci. Total Environ. 571, 1312–1329.

Google Scholar

Binding, C. E., Stumpf, R. P., Shuchman, R. A., and Sayers, M. J. (2020). “Advances in remote sensing of Great Lakes algal blooms” in Contaminants of the Great Lakes. ed. W. Lick (Cham: Springer International Publishing), 217–232.

Google Scholar

Chapra, S. C., and Dolan, D. M. (2012). Great Lakes total phosphorus revisited: 2. Mass balance modeling. J. Great Lakes Res. 38, 741–754.

Google Scholar

Chen, H., Gao, X., and Yuan, R. (2025). Advances in remote sensing and sensor Technologies for Water-Quality Monitoring: a review. Water 17:3000. doi: 10.3390/w17203000

Crossref Full Text | Google Scholar

Clamen, M., and Macfarlane, D. (2018). Plan 2014: the historical evolution of Lake Ontario–St. Lawrence River regulation. Canadian Water Resources J. 43, 416–431.

Google Scholar

Dove, A. (2009). Long-term trends in major ions and nutrients in Lake Ontario. Aquat. Ecosyst. Health Manag. 12, 281–295.

Google Scholar

Driscoll, C. T., Newton, R. M., Gubala, C. P., Baker, J. P., and Christensen, S. W. (1991). “Adirondack mountains” in Acidic deposition and aquatic ecosystems: Regional case studies. eds. S. Christie and D. F. Charles (Springer New York: New York, NY), 133–202.

Google Scholar

Elhadi, N., Harrington, A., Hill, I., Lau, Y. L., and Krishnappan, B. G. (1984). River mixing—a state-of-the-art report. Can. J. Civ. Eng. 11, 585–609.

Google Scholar

Fergen, J. T., Bergstrom, R. D., Twiss, M. R., Johnson, L., Steinman, A. D., and Gagnon, V. (2022). Updated census in the Laurentian Great Lakes watershed: a framework for determining the relationship between the population and this aquatic resource. J. Great Lakes Res. 48, 1337–1344.

Google Scholar

Frenette, J. J., Arts, M. T., and Morin, J. (2003). Spectral gradients of downwelling light in a fluvial lake (Lake Saint-Pierre, St-Lawrence River). Aquat. Ecol. 37, 77–85.

Google Scholar

Glasgow, H. B., Burkholder, J. M., Reed, R. E., Lewitus, A. J., and Kleinman, J. E. (2004). Real-time remote monitoring of water quality: a review of current applications, and advancements in sensor, telemetry, and computing technologies. J. Exp. Mar. Biol. Ecol. 300, 409–448.

Google Scholar

Mackie, C., Lackey, R., Levison, J., and Rodrigues, L. (2022). Groundwater as a source and pathway for road salt contamination of surface water in the Lake Ontario Basin: a review. J. Great Lakes Res. 48, 24–36.

Google Scholar

Mahdiyan, O., Filazzola, A., Molot, L. A., Gray, D., and Sharma, S. (2021). Drivers of water quality changes within the Laurentian Great Lakes region over the past 40 years. Limnol. Oceanogr. 66, 237–254.

Google Scholar

Mason, L. A., Riseng, C. M., Gronewold, A. D., Rutherford, E. S., Wang, J., Clites, A., et al. (2016). Fine-scale spatial variation in ice cover and surface temperature trends across the surface of the Laurentian Great Lakes. Clim. Chang. 138, 71–83.

Google Scholar

Mimouni, E. A., Ridal, J. J., Skufca, J. D., and Twiss, M. R. (2020). A multiscale approach to water quality variables in a river ecosystem. Ecosphere 11:e03014.

Google Scholar

Mimouni, E. A., Ridal, J. J., and Twiss, M. R. (2021). A multi-year assessment of phytoplankton fluorescence in a large temperate river reveals the importance of scale-dependent temporal patterns associated with temperature and other physicochemical variables. Front. Water 3:740309.

Google Scholar

Moukomla, S., and Blanken, P. D. (2016). Remote sensing of the north American Laurentian Great Lakes’ surface temperature. Remote Sens 8:286.

Google Scholar

Nicholls, K. H., Hopkins, G. J., Standke, S. J., and Nakamoto, L. (2001). Trends in total phosphorus in Canadian near–shore waters of the Laurentian Great Lakes: 1976–1999. J. Great Lakes Res. 27, 402–422.

Google Scholar

Osburn, C. L., and Morris, D. P. (2003). “Photochemistry of chromophoric dissolved. UV effects in aquatic organisms and ecosystems” in UV effect in aquatic organisms and ecosystems, comprehensive series in Photochemistry & Photobiology. eds. E. W. Hlebing and H. Zagarese (New York, NY: Royal Society of Chemistry).

Google Scholar

Plaas, H. E., and Paerl, H. W. (2020). Toxic cyanobacteria: a growing threat to water and air quality. Environ. Sci. Technol. 55, 44–64. doi: 10.1021/acs.est.0c06653,

PubMed Abstract | Crossref Full Text | Google Scholar

Rousso, B. Z., Bertone, E., Stewart, R. A., Rinke, K., and Hamilton, D. P. (2021). Light-induced fluorescence quenching leads to errors in sensor measurements of phytoplankton chlorophyll and phycocyanin. Water Res. 198:117133. doi: 10.1016/j.watres.2021.117133,

PubMed Abstract | Crossref Full Text | Google Scholar

Rundel, P. W., Graham, E. A., Allen, M. F., Fisher, J. C., and Harmon, T. C. (2009). Environmental sensor networks in ecological research. New Phytol. 182, 589–607. doi: 10.1111/j.1469-8137.2009.02811.x,

PubMed Abstract | Crossref Full Text | Google Scholar

Sharp, S. L., Forrest, A. L., Roberts, D. C., Cortés, A., and Schladow, S. G. (2025). A correction model for quenching effects on chlorophyll a fluorescence measurements in lakes. Limnol. Oceanogr. Methods 23, 543–561. doi: 10.1002/lom3.10697

Crossref Full Text | Google Scholar

Shousha, S., Maranger, R., and Lapierre, J. F. (2022). Contrasting seasons and land uses alter riverine dissolved organic matter composition. Biogeochemistry 161, 207–226. doi: 10.1007/s10533-022-00979-9

Crossref Full Text | Google Scholar

Steffler, P., and Blackburn, J. (2002). River2D: Two-dimensional depth averaged model of river hydrodynamics and fish habitat - introduction to depth averaged modeling and user’s manual. Edmonton, AB: University of Alberta.

Google Scholar

Turner Designs (2025). C3 and C6 submersible fluorometers, document S-0096 rev. AJ. Available online at: https://docs.turnerdesigns.com/t2/doc/brochures/S-0096.pdf (accessed December 20, 2025).

Google Scholar

Twiss, M. R., Ridal, J. J., Rooney, R. C., Christie, G. C., Bratton, J. F., and Wang, L. (2025). Critical intersections of flow: connecting waters among the Laurentian Great Lakes. J. Great Lakes Res. 51:p.102585. doi: 10.1016/j.jglr.2025.102585

Crossref Full Text | Google Scholar

U.S. Geological Survey (2025). Water-year summary for site USGS 0426433. Available online at: https://waterdata.usgs.gov/ny/nwis/wys_rpt/?site_no=04264331&agency_cd=USGS (accessed August 10, 2025).

Google Scholar

Verhamme, E. M., Bratton, J. F., Austin, J. A., Binding, C. E., Collingsworth, P. D., Dick, G. J., et al. (2024). The use of advanced and emerging technologies for adaptive ecosystem-based management of the Great Lakes. Aquat. Ecosyst. Health Manag. 27, 96–107.

Google Scholar

Wang, L., Riseng, C. M., Mason, L. A., Wehrly, K. E., Rutherford, E. S., McKenna, J. E. Jr., et al. (2015). A spatial classification and database for management, research, and policy making: the Great Lakes aquatic habitat framework. J. Great Lakes Res. 41, 584–596.

Google Scholar

Wetzel, R. G., and Likens, G. E. (2000). Limnological analyses. New York, NY: Springer-Verlag.

Google Scholar

Xenopoulos, M. A., Twiss, M. R., and Child, M. (2025). “Great Lakes winter science: summary report” in A report submitted to the international joint commission (IJC) Great Lakes science advisory board (Windsor, ON: IJC-Great Lakes Regional Office).

Google Scholar