Water level data from 9 permanent tide gauges operated by CHS were downloaded from the Canadian Tides and Water Levels Data Archive (https://www.meds-sdmm.dfo-mpo.gc.ca/isdm-gdsi/twl-mne/index-eng.htm, last access: 23 September 2026). For each regular station, three independent water level sensors were co-located with a common data logger. The instrumentation consisted primarily of vented differential pressure transducers with built-in atmospheric pressure compensation, namely OTT PLS-L or OTT PLS 500 sensors deployed in freshwater, and OTT PLS-C sensors deployed in saline environments. Only Banc du Cap Brûlé (3075) and Baie-Sainte-Catherine (3424) were equipped with a combination of one FTS64 radar and two OTT PLS-C sensors. These data underwent routine data quality control by CHS (cf. Sect. ) and were retrieved once their status had changed from preliminary to validated. The resulting minute-resolution water level corresponds to the mean of the three sensor measurements over each sampling interval, except when one sensor is unavailable or flagged as problematic, while no averaging is performed at radar stations. Overall, measured water levels at the CHS permanent stations have a total propagated uncertainty below 3 cm (2σ), except at Baie-Sainte-Catherine (3424), where it reaches 5 cm (2σ).

Additionally, 4 temporary tide gauges were installed by CHS between May/June and September 2023 to support this study. Each station was equipped with a Solinst Levelogger that measured absolute pressure, temperature, and conductivity, either mounted inside a pipe attached to a vertical structure (Fig. a), or fixed to a concrete block placed at the bottom of the river (Fig. b). At the L'Anse-Saint-Jean site (3440), 3 additional collocated OTT PLS-C vented pressure probes with conductivity cells were attached to the same concrete block; data from this triplet was later averaged into a single time series. Owing to the temporary nature of the installations and the use of sensors different from those at the permanent stations, the water level uncertainty at these sites is expected to be below 5 cm (2σ).

Five additional stations were installed nearshore after the loss of ice, between May and October 2023. Four of these stations were equipped with two sensors: one Solinst Levelogger measuring absolute pressure and temperature, and one Seametrics CT2X sensor measuring conductivity, temperature, salinity, total dissolved solids, and absolute pressure. The fifth station, at Isle-Verte (ilvn), was equiped with a single Solinst Levelogger. The sensors were placed inside perforated PVC pipes, which were then attached to metal rods hammered into the riverbed (Fig. c). For the five stations, the uncertainty is estimated to be less than 7 cm (1σ), with 7 cm representing the upper bound of the individual uncertainty estimates. The survey height error for each sensor is provided in Table ; see Sect.  for details on the datum corrections and associated uncertainties.

Absolute pressure readings were barometrically compensated using collocated barometers from independent Solinst Barologgers installed above water, with an accuracy of ±0.05 kPa. The data from each sensor was converted to water levels using the equation for hydrostatic pressure, which reads 1h=P/ρg, where h is the water column height above the sensor (m), P is the compensated pressure (kPa), ρ is water density (kg m⁻³) and g is the gravitational acceleration (m s⁻²). The derived water heights were then referenced to a common datum following a procedure described in Sect. .

In the estuarine transition zone, downstream of Orleans Island (O.I.; I.O. in French), water density oscillates with tides between freshwater and brackish water. Water density in Eq. () was computed using salinity and temperature data where available ; a constant value of 1000 kg m⁻³ was used otherwise. Both density-corrected and uncorrected data are provided in the dataset, along with temperature and salinity data, when available.

Ground-based GNSS-IR antenna arrays were used to monitor water levels in the St. Lawrence Estuary. These sensors were chosen for their ease of installation (especially in wintertime) and for their potential to provide additional information about rough (icy or wavy) surface conditions . A total of 13 low-cost GNSS-IR sensors from Precipice Sensors Inc. were attached on poles or lampposts facing the water a few meters away from the shoreline (Fig. d). Each sensor contains four u-blox SAM-M10Q antenna modules that are mounted on ground planes and connected to a data logger. Most sensors were powered with batteries and connected to a solar panel, except for two sites with access to continuous power. Three sensors were equipped with a cell connection for real-time data transmission.

The GNSS-IR processing technique was based on and consisted of two key stages. First, Signal-to-Noise ratio (SNR) measurements were analyzed using a Lomb-Scargle Periodogram (LSP) to obtain an unevenly spaced time series of water level estimates (these estimates are referred to henceforth as arcs). Second, a moving-window spline-fitting technique was applied on the arcs to obtain a regularly spaced and smoothly varying time series of water levels. Tropospheric delay and height rate corrections were applied during the spline-fitting . Both the arcs and spline-fitted water levels are provided in this dataset. All GNSS-IR water levels were referenced to a common vertical datum, as described in Sect. .

GNSS-IR parameters that were previously shown to be useful for monitoring ice and wind speed were calculated along with water level estimates as part of arcs processing stage. These parameters were calculated after applying an LSP and also a wavelet transform on SNR measurements. Certain parameters, including Confidence Level Retrieval (CLR) and Area Factor (AF) are taken from other literature . The full list of parameters and their definition is provided in Table .

A time-varying estimate of the error in the spline-fitted GNSS-IR water levels was calculated. This error estimate refers to a one standard deviation error of the hourly water level nodes that were estimated during the spline fitting process. Refer to for more information about the water level nodes. The water level nodes and errors were estimated using the least squares algorithm called leastsq in the scipy Python library . On average, spline-fit errors ranged from 0.051 m (grbr) to 0.151 m (ptio) for the data presented here (see Table ). More details about other sources of uncertainty in GNSS-IR data are discussed in .

Water levels were referenced to a common vertical datum, the static Canadian Geodetic Vertical Datum of 1928 (CGVD28, hybrid geoid model HTv2.0). CHS data are typically referenced to chart datum based on GPS-leveled water levels during installation and demobilization. They were converted from the chart datum to CGVD28 using factors listed in Table , provided by the solution from Canadian Spatial Reference System Precise Point Positioning (CSRS-PPP) at reference benchmarks taken by CHS. The remaining stations were referenced using real-time kinematic (RTK) GNSS corrections based on a Virtual Reference Station (VRS) network (either Can-Net or, when not available, Quebec’s Ministry of Natural Resources and Forests geodetic network), capable of centimeter-level accuracy. RTK-measured water levels taken near the sensors at different stages of the tide were used to reference pressure transducer data. A weighted average was applied on each group of RTK-point measurements based on the GNSS vertical accuracy (Table ), with weights equal to the inverse of the vertical errors. The regularly-sampled sensor data was interpolated to the exact RTK timestamps before computing the transformation. In contrast, RTK-measured heights of the GNSS-IR sensors were used to convert reflector heights data into absolute water levels, instead of using direct water level measurements. In fact, direct RTK-measured and GNSS-IR derived water level comparisons were not always possible due to the difficulty of access to a representative open water point. Overall, the accuracy in absolute water levels is estimated to range between 0.01 and 0.07 m. Survey height errors are provided in Table  for all stations. Further validation details are provided in Sect. .

Different coordinate systems and tidal conventions are used in SWOT products and in the Canadian geodetic reference system . SWOT water surface elevations (i.e. water levels) over continental waters are expressed relative to the mean-tide EGM08 geoid, while the associated position variables are provided in the International Terrestrial Reference Frame 2014 (ITRF2014) at the measurement epoch. To facilitate direct comparison between the in situ observations and SWOT water surface elevations, the following conversion procedure can be applied:

First, for observations referenced to CGVD28 using HTv2.0, the orthometric height is converted back to an ellipsoidal height using the HTv2.0 realization and reference epoch 1997.0 adopted for the dataset. The ellipsoidal height is obtained as h=H+N, where H is the CGVD28 orthometric height and N is the HTv2.0 geoid undulation. This conversion can be performed with the Natural Resources Canada (NRCan) GPS⋅H tool (https://webapp.csrs-scrs.nrcan-rncan.gc.ca/geod/tools-outils/gpsh.php, last access: 23 September 2026).

The complete conversion from the Canadian vertical reference system to the vertical reference used by SWOT can be expressed as: 3HSWOT=HCGVD28+NHTv2.0+ΔhTRX+Δhpd-NSWOT, where HCGVD28 is the water surface elevation relative to CGVD28, NHTv2.0 is the geoid undulation from the HTv2.0 model used to convert the CGVD28 orthometric height to an ellipsoidal height, ΔhTRX is the ellipsoidal-height change resulting from the transformation from NAD83(CSRS) at epoch 1997.0 to ITRF2014 at epoch 2023.5233, Δhpd is the permanent deformation correction given by Eq. (11.7) of the SWOT User Handbook , and NSWOT is the mean-tide geoid height provided by the SWOT geoid variable at the corresponding location. The resulting HSWOT is the water surface elevation relative to the mean-tide EGM08 geoid used by SWOT. The workflow is summarized in Table , and the corresponding conversion factors are provided in Table  for each station.

Quality control performed by CHS at their permanent stations consisted in (1) comparing measured water levels against minimum and maximum thresholds defined at each station, (2) assessing the differences between water levels measured by collocated sensors when possible, (3) comparing the measured water levels against tide forecasts when available, and (4) manual verification by an expert on a daily basis. The four temporary stations deployed by CHS underwent manual quality control only, as neither collocated sensors nor forecasts were available, with the exception of L'Anse-Saint-Jean (3440) where both Solinst and OTT PLS-C sensors were installed.

Quality control measures for the remaining pressure transducers were mainly based on the temperature and/or salinity data measured by the sensors. Sudden changes or erroneous values in temperature or salinity, often related to the sensor being above water, were flagged and the corresponding water levels disregarded. The procedure for detecting erroneous values of temperature or salinity is as follows. First, temperature values lower than 0 or higher than 100 °C, as well as salinity values lower than 0.5 or higher than 100 PSU, were disregarded. These thresholds are chosen considering that the pressure and conductivity sensors were installed from spring to fall in the brackish waters of the estuarine transition zone. Additional outliers were detected by calculating the scaled Median Absolute Deviation (MAD) in a moving window of 14 d, corresponding approximately to the neap-spring tidal period. If the absolute difference between a point and the median value in the window was greater than five, then the point was considered to be an outlier and removed. The results from applying this algorithm on salinity data are presented in Fig. .

Quality control measures were applied during both stages of the GNSS-IR processing. To produce the arcs, threshold-based criteria were applied so that any water level estimate obtained from an LSP with a peak-to-noise ratio or CLR smaller than 3 was disregarded. During the spline-fitting stage, large errors often occurred when there was a large gap in time between water level estimates. Therefore, any water level node with a standard deviation error greater than one meter was disregarded. The spline-fitted water level data also contains a “ground flag” that is raised whenever the measured levels are near or below the local bed elevation. This flag was only raised at sites where the shoreline recedes sufficiently during low tide to expose extensive intertidal flats, and where RTK-GNSS measurements of the river bed in front of the sensor were taken. Under these conditions, the sensor may measure the exposed bed surface rather than the water level, resulting in a truncated or nearly constant water level signal until the rising tide inundates the area again. At other sites where the river bed was not surveyed with RTK-GNSS, similar truncation may occur during exceptionally low tides but cannot be identified with the same confidence.

The IML-4 is a multidisciplinary platform that forms part of the AZMP by Fisheries and Oceans Canada (DFO), a network of seven similar platforms gathering extensive physical and biogeochemical data at a high rate in the St. Lawrence Estuary and Gulf (cf. Sect. ). These tailor-made buoys conceived and developped by DFO and MTE Instruments operate exclusively during the ice-free season, typically from May to October. With their diameter of 2.24 m and height of 4.51 m, they are significantly larger than some compact-sized (40–70 cm, 7–30 kg) and medium-duty (70–150 cm, 50–250 kg) commercial standards such as the Sofar Spotter, Datawell Waverider series, NextSens CB series, and Triaxys products.

The AZMP buoys carry a wave-monitoring module positioned 20 cm above the still-water level, which tracks the wave field using a 4.0 Hz SCC1300-D02 3-axis accelerometer component. The module follows the standard AZMP sampling protocol, in which surface motions are sampled for 10 min bursts every 15 min, resulting in a 5 min gap between successive records. This sampling strategy is used across all sites to manage shared computing resources among onboard sensors, thereby limiting the data volume allocated for wave monitoring. Wind data were collected by an onboard Vaisala Ultrasonic WMT700 anemometer. Buoy heading and gyroscopic measurements were obtained from an external GPS and Rion-Tech CDM260B compas mounted externally to the wave module.

All additional data collected in conjunction with the wave module was sampled only once every half hour, including GPS and compass readings. Hence, it is important to note that acceleration data are likely to be biased towards trends or sporadic fluctuations caused by rolling, pitching, or tilting, due to the inability to account for gyroscopic measurements at the acceleration sampling rate. Such trends or fluctuations are perceived as steady or slowly evolving imbalances caused, for example, by biofouling, tidal flows, or slowly varying surface currents dragging the buoy out of its position of equilibrium. During the validation of the wave data, we noticed that the absolute mean of the compass readings for rolling and pitching was less than 1°, except for a few instances where it reached or exceeded 3°, under strong wind conditions. Although these unrectified biases are generally small, we acknowledge that this absence of high-rate gyroscopic readings may degrade the accuracy of acceleration data. To address this limitation, we flag time series showing significant mean shifts during the quality control procedure (see Table , test 4.12). Despite the same limitation when accounting for the buoy's heading, we noticed during wave data validation that correcting for its heading is crucial for accurate wave directional moment estimations. By considering the lower rate 30 min GPS heading as a mean value for correction, we obtained interpretable values for wave provenance and spreading, as shown later in Sect. .

IML-4 post-acquisition processing was performed in order to provide the users with a faithful description of ambient wave conditions in the SLS. We recall that IML-4 supplies a rich variety of data alongside wave monitoring, which can be accessed in the datasets that we promote. This includes environmental conditions and auxiliary physical quantities that are available every 30 min: acoustic Doppler current profiler (ADCP) measurements of near-surface water velocity (bin #1), conductivity-temperature-depth (CTD) observations of the water body, and atmospheric relative humidity, pressure and temperature.

We carried out a comprehensive spectral analysis of the linear acceleration to describe dominant scales and the provenance of surface motions evolving under wind forcing. We used a fifth-order high-pass Butterworth filter to remove low-frequency motions that unlikely corresponds to the orbital motion of waves, such as horizontal displacements caused by tidal currents, wind blowing against the buoy, or Stokes drift. For filtering, a cutoff period pc≈11.3 s equivalent to a wavelength of 200 m for deep-water linear waves is chosen. Such scale embeds the most significant part of wave spectrum related to surface gravity waves. It also filters out slow horizontal oscillations that are allowed by the mooring cable, which is the most important source of signal contamination. Subsequently, bulk wave parameters were calculated from the average cross-spectral density of paired, consecutive 10 min data bursts recorded every 15 min, yielding an effective temporal resolution of 30 min. Lastly, the quality control of ST and LT time series was performed (see Sect. ).

Additional Spotter buoys deployed during SWOT's Cal/Val fast-sampling and early Science phases were operational from late May to mid August 2023, with the aim to increase and densify wave measurements under the SWOT Cal/Val orbit. Surface displacements of the water body were measured through a GPS-based approach , leading to a continuous acquisition at 2.5 Hz, and resolving the 0.029 to 0.8 Hz frequency bands. Similar to IML-4, spectral analysis was used to quantify wave field characteristics such as height, period, provenance (angle relative to its sources) and spreading, derived from high-pass filtered wave spectra. Validation results are presented in Sect. .

For each ST time series of surface motions and the corresponding LT time series of wave parameters, a quality flag is assigned, consisting of a primary code and a secondary code. The primary code is determined based on the nature of quality control, while the secondary code indicates the reliability in the event of hardware failure and anomalous or severe gradients in environmental conditions that could affect the buoy’s measurements of wave conditions.

The quality control is assessed following standard testing procedures described in , categorizing data as “1” (good), “3” (suspect), “4” (fail) or “9” (missing). The assigned secondary code corresponds to the test number that has served for classifying the data as faulty or suspect. For AZMP buoys, we provide an additional quality-control flag indicating whether horizontal acceleration could not be corrected for buoy heading due to the absence of readings. In such cases, we assign primary and secondary quality codes of “4.NH” for horizontal acceleration. We continue to provide horizontal acceleration, but users should note that the wave processing cannot proceed further beyond this point, and therefore directional wave parameters cannot be calculated. Otherwise, data are considered physically reliable if all tests are satisfied. Finally, if a LT time series of wave parameters was calculated using a suspect or flawed ST time series with quality codes of “4.9”, “4.12” or “4.NH”, then the derived bulk wave parameters may not accurately reflect real-world wave conditions due to insufficient, scarce data or the lack of rectification for tilting, rolling, pitching and heading.

As erroneous or erratic measures can be detected by multiple tests, we first tested for faulty data, then for suspect ones. The secondary code is established by first checking for data lying outside the operator-defined range, then we flag missing or flattened data, which may indicate measurement error. Next, we looked for data with prominent mean shifts, indicating that the buoy has shifted relative to its equilibrium position which we could not account for during post-processing. Finally, we analyzed whether there are acute rates of change in the data, or if they fall outside the statistically expected range based on historical wave data. Table  presents the sequence of tests used to establish the primary and secondary quality codes for ST and LT time series. Statistical ranges were calculated from data acquired by the IML-4 station since 2014, available from St. Lawrence Global Observatory (SLGO) at https://www.ogsl.ca/en/home-slgo/ (last access: 21 September 2026). The resulting set of statistical computations is summarized in Table , which include minimum, maximum, mean and standard deviation values for each wave parameters, with three considerations to note. Firstly, the minimum significant wave height is set to a small non-null value, as a zero-value could imply either a perfectly still surface or sensor failure. Thus, to avoid ambiguity it was set arbitrarily to 0.02 m. Secondly, the minimum wave period was determined from the inverse of the Nyquist frequency, using the lowest sampling frequency fs of the two buoys, yielding minperiod=1.25 s. Third, it is unlikely that the maximum value for wave spreading is 90°, as this would suggest an equal distribution of the wavefield in all directions; therefore, a value of 80° was established.

This section describes each water level data type and format contained in this dataset. The data folders are structured by station name, sensor type or provider, and processing level. File naming follows the convention “[4-character ID]_[sensor type or provider]_[level].csv”, where the station ID is provided in Table , and the sensor type or provider is one of “baro”, “sol”, “ct2x”, “gnssir”, or “chs”, which stand for Solinst Barometer, Solinst Levelogger, Seametrics CT2X, GNSS-IR, and CHS (Solinst or OTT PLS-C), respectively. Although most of the CHS data was acquired using Solinst sensors, we differentiate them from the Solinst data that was acquired by our team due to differences in the level of processing and quality control applied.

For each sensor, CSV files are provided following a definition of the data levels as follows:

Solinst Barometer (“baro”):

Level 0: corresponds to atmospheric pressure and temperature readings.

We organize the data for each wave buoy into distinct levels, analogous to water levels. For each buoy, year, and data level, we produced NetCDF files that facilitate simple and efficient access to all the variables and their attributes for a particular season and locations. For each variable, attributes comprise a concise description, units, and the related quality flag and code time series. One should be aware that quality control is not offered for ancillary variables, as the primary focus is on wave data.

The data folders are structured by buoy type (either “azmp” or “spot”), station name, and processing level. The file names follow a straightforward convention that combines the buoy type and ID (as listed in Table ), the processing level, the data type, and the year the data was collected. The format is as follows: “[buoy_ID]_[level]_[type]_[year].nc”. The wave buoy datasets are organized into three levels that increase upon the degree of processing:

AZMP platform (“azmp”):

Level 0, Type “accelerations”: Contains all 10 min acceleration time series, sampled at 4.0 Hz and acquired every 15 min, along with the corresponding ST quality control flag.

A comparison of water level time series obtained from three different sensors at two nearby stations in the Saguenay Fjord (L'Anse-Saint-Jean and Petit-Saguenay) is presented in Fig. . As a reference, time series from downstream and upstream stations are also included to illustrate how tides are evolving between Baie-Sainte-Catherine (3424), located at the fjord entrance, and Port-Alfred (3460), about 95 km upstream. Midway between these stations, at L'Anse-Saint-Jean (3440), both CHS Solinst Levelogger and OTT PLS-C sensor data were highly consistent with each other, showing only very slight differences, with root-mean-squared and maximum absolute differences of 0.008 and 0.08 m, respectively, and a mean bias of 0.004 m for the entire record length. Similarly, comparisons between CHS data at L'Anse-Saint-Jean (3440) and Solinst levelogger data at Petit-Saguenay (ptsg), located 5 km apart, show root-mean-squared differences (RMSD) of 0.045 m and a bias of 0.005 m. These differences are mainly associated with the Petit-Saguenay sensor drying out at low tide (see Fig. , bottom panel). When put in relation with downstream and upstream stations at Baie-Sainte-Catherine and Port-Alfred, the standing-wave character of the tides in the Saguenay fjord clearly stands out. While tidal signals are almost perfectly in phase throughout the entire fjord length, they are amplified landward from 5.1 to 6.3 m in range during spring tide (grey-shaded area in Fig. ).

The Seametrics CT2X sensor at Petit-Saguenay was intentionally left out of the analysis as it was showing a systematic bias compared to other gauges in the Saguenay Fjord. Even more so, the observed water levels were often falling outside the limits naturally imposed by the two boundary gauges at Baie-Sainte-Catherine (at the fjord entrance) and Port-Alfred (upstream). Assuming that the tide is being amplified from downstream to upstream and that mid-fjord stations were all in concordance except the CT2X one, it was chosen to disregard data from this sensor in Fig. . However, CT2X data were not removed from the database as they include relevant salinity and temperature information, as discussed in the next subsection.

In the Saguenay Fjord and in the St. Lawrence estuarine transition zone (downstream of Orleans Island), water density variations can be observed during the tidal cycle. As a result, differences in the estimated water levels can exceed one decimeter if not adjusted for density, due to the large tidal ranges in this area, reaching up to 7 m. In fact, density directly influence the calculated height of water above the sensor, mostly around high tide, when saline waters enter the estuary and when the water column is the deepest. While density corrections assume that the salinity and temperature measured at the sensor depth are representative of the water column, this may be only valid in shallow waters and less so as water depth increases. Nevertheless, the estimated effect of density on water levels is captured in Fig.  and further analyzed in Table .

Given that conditions oscillate between saltwater and freshwater as a function of the incoming tide, errors mostly occurred around high tide when a peak in salinity was reached, creating an apparent bias in the data, as shown in Table . While on average the error is comparable to the survey height accuracy (cf. Table ), maximum differences (around high tide) can amount to up to 0.155 m (ptsg) based on comparisons of the CT2X L1 non-corrected heights and density-corrected heights. Figure  illustrates an example of these differences at Saint-Roch-des-Aulnaies (srda) located in the St. Lawrence estuarine transition zone. Density corrections associated with salinity and temperature variations show variability at both the semi-diurnal and neap-spring tidal scales. Further monthly variability is observed, likely linked to the seasonal cycle in river flow.

In Table , a comparison between collocated Solinst Leveloggers and Seametrics CT2X sensors is provided, using both non-corrected and density-corrected water levels from the CT2X sensors. “No density correction” refers to L1 data that were datum-adjusted using coefficients provided in Table  but with minimal quality control (leading to more data points). “Density-corrected” entries refer to L2 water levels that have undergone quality control and are based on the density-corrected CT2X data. Note that in both cases, no density corrections were applied to the Solinst data, as salinity was only measured by the CT2X sensors and started behaving erratically after a few weeks or months after deployment, leaving the salinity records incomplete (see Fig. ). This table should be looked at in conjunction with Table  where the number of survey points used for each sensor is presented. In general, less survey points were used for CT2X sensors because of sensor malfunctioning after a certain period of recording. As a result, CT2X records are generally shorter, they were referenced using less surveying points (since some of them were taken at the end of the deployments), and they show larger differences with nearby official CHS tide gauges. Overall, we consider these sensors less reliable; nevertheless, they provide information on the error made by neglecting density variations. In fact, the various error metrics shown in Table  tend to degrade when density-corrections are applied to the CT2X data, as they depart from the non-density-corrected Solinst data. A few centimeters difference can be observed on average beween non-corrected and density-corrected RMSD, with increasing biases and maximum absolute differences between the sensors.

Water levels from collocated pressure gauges and GNSS-IR sensors closely match at the sites Saint-Laurent I.O. (3110) and Vieux-Québec (3248) (Table ). At Saint-Laurent I.O., the RMSD between the (level 2) GNSS-IR and pressure-gauge-derived water levels is 0.0696 m from the period from 17 March 2023 to 23 May 2024 and the mean offset between the two water level time series (i.e. GNSS-IR minus tide gauge heights) is -0.0015 m. Furthermore, comparing GNSS-IR and pressure gauge water levels at Vieux-Québec for the time period from 20 March 2023 to 12 May 2024 results in an RMSD of 0.1017 m and a mean offset of -0.0339 m. GNSS-IR sensors and antennas differ between sites, which partially account for performance variability with site. A comparison between GNSS-IR water levels – both arcs (level 1) and spline-fitted (level 2) – and a collocated pressure transducer is shown in Fig. . The largest differences between CHS and GNSS-IR level 2 data were observed around high and low tides and likely arise from imprecision of the spline-fitting procedure around tidal extremas .

For consistency, all of the level 2 GNSS-IR water level time series were levelled to the CGVD28 datum using RTK measurements of the sensors themselves. At the Vieux Québec and Saint-Laurent I.O. sites, the water levels could also be levelled by using the pressure gauges. In this case, the RMSE between the GNSS-IR and pressure gauge water levels remains nearly unchanged at Saint-Laurent I.O. (0.06955 m) and reduces to 0.0959 m at Vieux-Québec. It would also be possible to level the GNSS-IR water levels using RTK measurements of water levels, however this is only possible at sites where overlapping measurements exist. Furthermore, direct RTK measurements of water levels taken in front of the GNSS-IR sensors are affected by wave conditions and can depart from water levels retrieved using GNSS-IR data across the instrument’s footprint. Therefore, no further adjustments were made.

In some areas along the St. Lawrence Estuary, extensive floodplains can be revealed at low tide. The associated intertidal topography is captured by some of the GNSS-IR sensors and flagged accordingly (only where RTK-GNSS measurements of the river bed are available), while low tides at other pressure sensors appear truncated. In the latter case, the recorded barometrically-compensated (L1) water levels under dry conditions are close to zero or become negative, with correspondingly abnormal (air) temperatures flagged during quality control.

Figure illustrates water levels at Isle-Verte taken from two sensors: a Solinst Levelogger located on the North shore (ilvn_sol) and a GNSS-IR sensor positioned on the South shore (ilvs_gnssir). While the pressure sensor dried out during the lowest low tides and those segments were removed during quality control (as confirmed by comparing L1 and L2 Solinst data, not shown), the GNSS-IR sensor consistently showed truncated low-water levels, reflecting the extensive intertidal flats that border the narrow South shore channel at Isle-Verte. In this case, the ground flag was activated. However, these observations of the intertidal topography still suffer from some inaccuracies, as these points take part in the spline-interpolation procedure used to produce the water level time series, distorting the lower water levels before hitting the ground. Furthermore, given the footprint of the GNSS-IR (Fig. a), as the water recedes, one can suppose that only a portion of the observation points happen to be over dry topography, while the rest correspond to the water surface. This explains the observed variability in heights of the ground points in Fig. b. Nevertheless, comparisons of the ground-flagged elevations with SWOT observations during low tide could be conducted in a future study to assess the potential of SWOT in mapping intertidal topography, as demonstrated by .

Further insights into tidal dynamics is revealed in Fig. , relating to tidal distortion in shallow water. While both sensors on the North and South shore of Isle-Verte present highly synchronized high waters, the low waters on the South shore (shallower) occur later than on the North shore (deeper). This leads to more abrupt rising tides and longer falling tides, a tidal asymmetry that typically translates into larger amplitudes of the shallow water tides (e.g. M4).

In tidal rivers and estuaries, water levels and hydraulic gradients exhibit spatial and temporal variability, ranging from gradual changes driven by the long-wave nature of tides to abrupt shifts induced by local topography or wind effects. These variations occur both in the cross- and along-channel directions . While such gradients are often linked to river discharge, they are also influenced by processes including reversing tidal flow, Stokes drift compensation flow, neap-spring storage effects, secondary circulation, and differential water level setups within channel networks . Despite these complicating factors, understanding how the spatiotemporal dynamics of water surface slopes vary in tidal systems is critical for advancing dynamical analyses. Field measurements of water surface slopes remain inherently challenging, traditionally relying on costly and logistically demanding deployments of dense station networks or boat-based surveys. High-resolution remotely sensed data, however, emerges as a promising alternative, offering the necessary spatial detail without the associated constraints.

The SWOT mission requires slope accuracy of 1.7 cm km⁻¹ (1σ) when averaging over a water area >1 km², which poses challenges for observing low-slope environments that fall below its expected detection threshold. In this context, the present dataset serves as an ideal testbed for assessing SWOT's slope accuracy. The Saguenay Fjord, characterized by standing-wave tides (Fig. ), exhibits very low instantaneous longitudinal slopes in water levels during most of the tidal cycle (Fig. , 1st column), except near high and low tides, where slopes slightly increase or reverse. During neap tides, the reduced tidal range results in even smaller longitudinal gradients compared to spring tides, when current reversals and associated salinity intrusions are most pronounced (Fig. ). In contrast, the St. Lawrence Estuary, where tides propagate upstream, generates significantly larger instantaneous slopes that are likely detectable by SWOT throughout most of the tidal cycle, except during transitional slack-water periods. As illustrated in Fig.  (2nd column), slopes during rising and falling tides in the North and South channels of Orleans Island highlight asynchronous tidal propagation due to differences in channel depths. These disparities are evident in both water levels and slopes between the channels as well as longitudinally, with faster rising tides and slower falling tides upstream due to the increased tidal asymmetry. Such surface gradients, combined with riverbed topography, govern discharge distribution and tidal propagation in this diffluence/confluence region. Despite limited water level station coverage in these channels, SWOT is expected to provide unprecedented spatial detail on these interactions, including cross-sectional gradients in wider estuary sections (not shown).

The St. Lawrence deep-water wave system can be synthesized as a wind-driven sea limited to the estuarine stretch. In specific circumstances, swells may form locally or penetrate westward from the estuary's outlet, where it widens into the Gulf. Such wave regimes are likely to evolve on scales larger than, or comparable to, the selected cutoff wavelength of 200 m (corresponding to a period of 11.3 s). Henceforth, the choice of a cutoff criterion intends to mitigate the contamination from slow-frequency motions unrelated to wave motions, although its selection has not been fully tested against all possible wave regimes. The objective is to ensure that the buoy's measurements remain physically meaningful and interpretable for wind seas. Considering this, we delved a bit further using the wave age criterion as defined by , which classifies the filtered wave system as either wind sea or swell, based on whether 4cp≤1.5U10cos⁡(δ). where cp is the wave phase velocity, 0≤δ≤π/2 represents the angle between the wind and wave directions, while U10 is the wind speed measured at 10 m, which follows a logarithmic profile, U10=(10/z)0.11Uazmp(z), as described in . This equation uses the wind speed Uazmp(z) measured at an approximate height of z=4.3 m atop the buoy's mast.

Figure illustrates the relationship between the wind and wave system over the course of the 2023 and 2024 seasons. The wave's direction of propagation adjusts to the wind provenance (panel a), emphasizing that two states of which wind and waves propagate coincidentally along St. Lawrence are more likely observed, that is, approximately 50 and 230° relative to true north. Figure b presents a comparison between the wind component U10 parallel to waves and the dominant wave's phase velocity. Partitions that satisfy Eq. () are above the gray line cp=1.5U10, and they are highlighted in red in both panels, representing approximately 80 % of all the wind conditions measured in 2023 and 2024. The vast majority of the wave dynamics captured by IML-4 are likely representative of wind-driven wave regimes in the selected frequency range.

We continued our exploration of the St. Lawrence wave system by comparing wave spectra and bulk wave parameters between the IML-4 and Spotter buoy in Les Escousmins (spot-1082). Figure a, b illustrate the spectrograms of the variance wave spectra for both buoys. Strikingly, the observations are very similar, despite the buoys being more than 60 km apart. Dominant waves propagate with an average phase velocity of 1.5–2.5 m s⁻¹. The waves are slightly faster and weaker nearshore, as measured by Spotter-1082. Panels (c), (d), (e) present the significant wave height hm0, mean zero-crossing period Tm02, and mean wave provenance θ¯ measured by each station, with colored stars indicating wave parameters that have been raised for failing quality control, regardless of the test. The significant wave height can reach up to 2 m, but it may increase further in response to strong localized wind events (Fig. c). In a similar fashion, the mean zero-crossing period and provenance are provided for SWOT validation (Fig. d, e). Differences between the two buoys are summarized in Table .

While the two buoys experience similar variability, mean differences are visible and likely linked to their respective location in the estuary (offshore vs. nearshore). Moreover, Spotter and IML-4 platforms differ in several ways. First, the GPS signal's intermittency can cause temporary disruptions in the Spotter data stream. Second, the Spotter hub, weighing around 8 kg and measuring 42 cm × 31 cm, is more sensitive to high-frequency wave motions than the larger IML-4 platform. Third, by measuring the second derivative of surface displacements, IML-4 has a responsiveness to high-frequency waves reduced by a factor of ω-4, where ω is the wave frequency, rendering it insensitive to oscillations above approximately 0.7 Hz. A detailed performance comparison of the two instruments, however, lies beyond the scope of this work.