The study area covers the coastline of the Sept-Îles region, between 66°56′ N and 65°54′ W (Fig. 1). Situated in the northwestern Gulf of St. Lawrence, this region features a mix of saltwater brought by the Atlantic Ocean and freshwater from the St. Lawrence River and its tributaries (Carrière et al., 2018b). The region is under a semi-diurnal tidal system where tides range between 0.39 to 3.6 m in Port-Cartier and 0.43 to 3.36 m in Sept-Îles (Fisheries and Ocean Canada, 2024). The area has a subarctic climate and has historically been covered by sea ice between November and April (Aubut Demers et al., 2018). However, this period has been shrinking rapidly, and nearly ice-free winters could become more common due to climate warming (Galbraith et al., 2024a). Since 2010, the Gulf of St. Lawrence has experienced 4 winters with nearly ice-free conditions around the end of the season (Galbraith et al., 2024a).

According to a study on the hydrodynamics of the BSI (Shaw et al., 2022), the area exhibits an estuarian circulation pattern. Currents are seaward at the surface (2–10 m) and shoreward in deeper waters (10–30 m). Surface drifter measurements indicate average and maximum speeds of 17.4 and 86.6 cms-1, respectively. While locally dominated by tides, currents in the bay result from complex interactions between tides, estuarine circulation, winds, and the Earth's rotation. Water residence time is estimated to be 5.6 ± 3.5 d.

Diverse types of land-use along the transect were observed: natural coastlines with less developed areas, residential areas (urban and suburban) and Industrial-Port (IP) Zones. Sites 5 to 8 in Port-Cartier are located in an IP zone, and the Port of Sept-Îles (PSI) is the second largest Canadian port for annual activity (PSI, 2025) where several heavy industries contribute to the economic importance of the region. The Matamec sector presents a less developed and more pristine environment, featuring an ecological reserve and the Matamec River, which is of great conservation value (Government of Quebec, 2011). The study area is affected by freshwater input from the Ste-Marguerite and Moisie rivers, damned and wild, respectively. Smaller rivers also flow in the area, including the rivière aux Rochers in Port-Cartier and Brochu River, West of the Ste-Marguerite. The Hall, des Rapides, aux Foins and du Poste rivers flow into the BSI, with around 22 m3s-1 of freshwater input into this area annually (Shaw, 2019).

The vertical profiles were obtained with a Conductivity-Temperature-Depth (CTD) probe equipped with sensors that measure temperature and salinity (Sea-Bird Electronics SBE19plusV2), photosynthetically active radiation (PAR; Biospherical QSP-2300L with a spherical quantum sensor), turbidity (Seapoint) and in vivo chlorophyll fluorescence (Seapoint). The sensors were last calibrated in 2017, so the data should be interpreted with caution and considered as relative rather than absolute values. Fluorescence was used as an indicator of phytoplankton chlorophyll a biomass (Chl a). Chl a is an indicator of phytoplankton biomass and reflects the level of primary production in coastal waters (Araújo et al., 2022). The CTD was equipped with an integrated software system that automatically calculated practical salinity, density, specific volume anomaly, Brunt-Väisälä frequency, Sigma-T, potential temperature, Sigma-Theta, freezing temperature, sound velocity, and absolute salinity. At each site, the CTD was lowered to approximately 2–3 m above the seafloor to prevent potential damage to the instrument and had an average descent speed of 1 ms-1 and a sampling frequency of 4 Hz Measurements along the transect were made monthly at all sites between May and October 2022 (24–26 May, 14–15 June, 19–21 July, 16–17 August, 13–14 September and 24 October), except at site 34 in May and site 25 in June due to technical problems. The full dataset is available at 10.5683/SP3/ALRWON (Arseneault and Saulnier-Talbot, 2025a). Meteorological data for the week prior to sampling was also documented using data from the Government of Canada (2025) (Table A1). To facilitate visualizations, the 35 sites were divided into three Zones (Table 1). Sites 1 to 15 between Port-Cartier and the west of the BSI were defined as Zone 1, sites inside the BSI (16 to 27) were defined as Zone 2, and Zone 3 encompasses the sites from the east of the BSI to Matamec (28 to 35). Monthly vertical profiles of temperature, salinity, turbidity and fluorescence at each site were generated using “ggplot” (R core Team, 2022). With the same statistical package, relationships between Chl a and environmental variables were explored. Scatter plots with LOESS curves to observe general trends were generated and Spearman correlation was used to assess the significance of these relationships. A summary of the variables measured in the surface water of the three site zones for each month of sampling is available in the annex (Table A2).

Thermocline layers were identified, if possible, monthly at every site. Determining the depths of the thermocline is not always straightforward, as several methods can be used (de Boyer Montégut et al., 2004). These methods are however rarely applied in areas as shallow as in this study. Janecki et al. (2022) found no studies of this sort except in the Baltic Sea, where depths still reached 100 m (Leppäranta and Myrberg, 2009). These authors developed a new algorithm for shallow seas to find the top of the thermocline and halocline, for sites reaching >100 m. We decided for this study to use the gradient threshold method of 0.2 °C (de Boyer Montégut et al., 2004; Leppäranta and Myrberg, 2009). We assumed that the top layer of the thermocline was located at the depth of the strongest gradient of the slope (ΔT>0.2 °C), and the lower layer was located at the depth of the deepest gradient, to capture the typical plateau, if there was one. Since the sites in our study area were shallow, between 4 and 52 m with an 18 m median, we decided to start the calculations at 3 m depth instead of the advised near-surface 10 m depth (de Boyer Montégut et al., 2004). This approach also avoided the possible effects of the instrument, the anchor and the boat, which can mix the surface waters (de Boyer Montégut et al., 2004; Maske et al., 2014). The depths of the thermocline layer were obtained with the package “rLakeAnalyzer” (Winslow et al., 2019), with some modifications of the “therm.depth” function. A cutoff of 1 °C was also applied, a default criterion in the R function. If the change between surface and maximum depth temperatures was less than this threshold, thermocline depth was not calculated. Thermocline depths were not calculated for sites with a maximum depth <5 m. The same gradient threshold method was applied to salinity and density profiles to identify the halocline and pycnocline. We chose to use temperature in the present study as a primary indicator of vertical stratification, consistent with recent studies that have focused on the thermocline in similar environments with dynamic coastal waters influenced by water masses and anthropogenic activities (Lim et al., 2025, and references therein). The euphotic depth was also calculated using the depth at which PAR declined to 1 % of the surface value (Kirk, 1994), and the stratification index as the density difference (kgm-3) between the maximum sampled depth and the surface (3 m) sampled depth levels (Blais et al., 2019). The diffuse attenuation coefficient Kd (PAR) (m-1) was calculated for each CTD profile using linear regression of the natural log of PAR against depth, following Lund-Hansen (2004). Profiles with a maximum depth <5 m, or r2<0.8 were excluded. Simple and multiple linear regressions were then performed between Kd and depth-averaged in vivo fluorescence and turbidity to examine the drivers of light attenuation. One site out of the 35 was chosen to illustrate seasonal changes in the vertical profiles of temperature, salinity, Chl a and PAR during the sampling season.

In May, the water temperature was the coldest and was mostly even across depth at all sites (Fig. 2). Surface temperature averaged 5.56 °C along the transect. At maximum depth, temperatures ranged between 0.85 °C (Site 31) and 3.83 °C (Site 11), excluding site 25 which was only 3 m deep. During this period, a near-surface thermocline was beginning to form (Figs. 2 and 3), usually between 3 and 4 m depth (Table A3). In May, 22 sites showed the onset of stratification. In June, surface temperatures began to rise, ranging from 8.69 °C (site 32) to 11.99 °C (site 19). The water column remained largely mixed, but stratification became noticeable at deeper sites, such as sites 9, 27 and 31, where temperatures were between 1.72 °C (Site 31) and 4.3 °C (Site 9) at maximum depth. Sixteen sites exhibited thin thermocline layers (1–2 m thick), mostly near the bottom of the water column. Waters in June were generally well-mixed.

Between July and September, clear stratification developed with increasing temperatures, though a few shallow sites, <6 m depth, lacked distinct layers. At most sites in July, a thermocline was observed between 4 and 5 m depth. Surface temperatures ranged from 5.52 °C (Site 28) to 14.80 °C (Site 29). While surface warming was observed at some sites, 14 sites recorded a colder surface temperature than in June. For instance, recorded temperature at site 34 was 10.15 °C in June versus 5.63 °C in July. Temperature became colder at shallower depths in July, contrasting with June's well mixed conditions where colder waters only appeared at the deepest sites. In August, surface temperatures ranged between 5.18 °C (Site 29) and 17.92 °C (Site 9). Stratification was still persistent but started to deepen. August showed a deepening of the thermocline: up to 4 m at some sites (5 and 6) in the west, while sites in the BSI, like sites 19 to 24, showed the same depth layers as in June. At the eastern sites, the thermocline deepened between 4 and 12 m. September marked the warmest month throughout the water column, although deeper sites like sites 31, 30 and 9 retained colder bottom temperatures, around 4.2 °C. At several western sites, the thermoclines continued to deepen, while they began to disappear at the BSI sites and became shallower toward the east, near the Moisie River (sites 30 and 31). By October, the entire transect showed uniform temperatures with no significant differences between surface and maximum depths, and the absence of a thermocline. In early fall, the overall mean surface temperature was 6.83 °C, with 5.55 °C at maximum depth, resulting in an average difference of only 1.28 °C throughout the water column. All sites were mixed, except for site 31, the deepest one, that showed a weak persistent thermocline at around 10 m depth. The stratification index confirmed this trend, with mean values peaking in August and September (2.9 kmm3) and dropping in October (0.7 kmm3) across all zones (Table A3).

In summary, surface temperatures increased from May to September, with stratification becoming prominent in July. By October, conditions in the transect had reversed back to cold, uniform temperatures across depths. The largest changes in surface temperatures occurred between May and June (increase of 5.02 °C), and September and October (a decrease of 9.30 °C). Site 9 was chosen as an example (Fig. 3) to show how the thermocline layer, using the threshold method, matched with the vertical profiles of temperature and other parameters. This site was approximately 25 m deep and had no missing values, which made visualization easier. At this specific site, the thermocline, as observed in general, started to appear in May at a shallow depth, deepened during summer and disappeared in October.

Signals of freshwater input at sites near river mouths were apparent in May (Fig. 4), particularly at sites 29 (17.8 PSU), 31 (19.7 PSU) and 30 (20.1 PSU), located near Moisie River. These sites recorded the lowest surface salinity, likely due to increased volume of river discharge following spring snowmelt. The overall average salinity of the water column in May was 30.0 PSU. Despite this localized freshwater influence, thermocline, halocline and pycnocline depths were closely aligned in May, within approximately 1 m at most sites (Table A3). June recorded an average of 27.8 PSU, the lowest salinity of the season. Thermocline and halocline depths differed among zones. In Zone 2, the halocline (5.3–10.9 m) occurred above the thermocline (14.9–16.1 m), suggesting that stratification was mostly driven by surface salinity associated with freshwater inputs. In contrast, in Zone 3, the halocline (20.5 m) was deeper than the thermocline (13.8–15.2 m). Halocline and pycnocline were aligned in all 3 zones in June. July showed the highest salinity with an average across sites around 30.6 PSU. Stratification became more visible at some sites in July where salinity was lower above the thermocline, such as sites 13 and 15, that recorded 28 PSU versus 30 PSU below. In August, salinity above the thermocline decreased to around 27 PSU, showing the deepening of stratification. A strong signal of freshwater input by rivers was again observed in September, this time near the Brochu and Ste-Marguerite rivers, at sites 11 (17.6 PSU) and 12 (19.1 PSU). Between July and September, the onset depths of thermocline, halocline, and pycnocline were generally aligned, but the thermocline extended deeper, resulting in a thicker thermally-stratified layer compared to the halocline and pycnocline. In October, salinity was more uniform below the surface, while surface salinity remained lower, especially at sites located west of the transect, where salinity at Site 1 reached 20 PSU at the surface. Below surface, salinity increased to over 30 PSU along the water column. Although the thermocline was absent across the transect by October, the halocline and pycnocline remained detectable at some sites. Throughout the study period, halocline and pycnocline depths were mostly aligned (Table A3), indicating that salinity had more influence on density stratification in the water column, especially in June.

In these data, salinity inversely followed temperature trends in the water column, where it increased with depth while temperature decreased. This relationship was visible in vertical profiles such as Site 9 (Fig. 3). A Spearman correlation analysis confirmed that temperature and salinity were inversely correlated across the dataset (r=-0.84, pvalue<0.01). Overall, salinity at maximum depth showed little seasonal variation, averaging over 30 PSU, except in June (28.5 PSU). The most pronounced seasonal differences in salinity were observed at surface. May and June had the lowest surface salinities whereas July had the highest salinity, and August only showed a slight decrease, in general. A drop in surface salinity was observed in September, especially at sites near the Brochu River (Site 11). Salinity exhibited variability at the surface (first 2 m), usually at sites located near rivers, and became uniform at depth.

Because turbidity mostly varied at surface, only data down to a depth of 5 m was shown for better clarity (Fig. 5). Turbidity across the transect varied between 0 and 8.4 Formazin Turbidity Units (FTU) with limited variability, except at certain sites mostly located inside the BSI (sites 16 to 21). These sites showed a turbidity range of 4.8 to 8.4 FTU, with the highest value measured at site 20.

Higher values of turbidity, superior to 2 FTU, were also recorded in May, in the BSI and at sites near river mouths, such as sites 11 (4.6 FTU), 30 and 31. A similar pattern was observed in October, with higher turbidity values in the BSI and at sites close to rivers, where Site 20 exhibited a surface turbidity of 3.2 FTU and Site 11 a turbidity of 2.6 FTU. In July and September, turbidity remained below 2 FTU throughout the transect, with sites within the BSI being slightly more turbid than the others.

A seasonal variation of fluorescence (Fig. 6), reflecting concentration of Chl a, was observed along the transect.

Concentrations were the lowest in May and began to rise in June, reaching a peak in July, and then decreased throughout August to October. In September, a localized resurgence was observed at the surface of some sites, before fluorescence levels declined again in October at most sites.

In May, surface fluorescence throughout the transect averaged 0.37 mgm-3 and surface PAR 943.8 µmolphotonsm-2s-1. All sites had a concentration below 1 mgm-3 at all depths. In June, surface fluorescence increased to a mean of 1.07 mgm-3, with surface PAR averaging 1517.5 µmolphotonsm-2s-1. Sites 17, 18 and 19 recorded the highest Chl a concentration. These were the only sites across the entire transect where fluorescence exceeded 2 mgm-3 in June, all between 1 and 3 m. In July, a subsurface fluorescence maximum was observed at several sites, with the highest concentrations found at sites 30, 31, 16, and 35, followed by Site 23, where concentrations were all superior to 4 mgm-3, at depths between 15 to 24 m. All these sites were located near rivers, except for Site 16 which was positioned in front of an aluminum smelter. At this site, the highest concentration of 4.28 mgm-3 was measured at 24 m, which was the maximum depth at this site. All fluorescence peaks were below the thermocline layer, except at Site 35 where fluorescence concentrations were higher at the thermocline. Surface fluorescence averaged 0.53 mgm-3 in July, while surface PAR averaged 746.1 µmolphotonsm-2s-1, lower than in June. August showed the highest surface PAR of the season, averaging 1707 µmolphotonsm-2s-1, while surface fluorescence averaged 0.57 mgm-3. Sites 3, 4, 5, and 6, all located in Port-Cartier near the aux Rochers River estuary, and Site 31 at the Moisie River, had a concentration superior to 2 mgm-3. Maximum fluorescence (2.73 mgm-3) was observed at Site 4 at a depth of 5 m. Highest fluorescence in August occurred at shallower depths than in July, between 5 and 26 m. In contrast to July, peaks of Chl a in August were mostly found at shallower depths, either above or within the thermocline layer.

In September, surface fluorescence reached its seasonal peak, averaging 1.59 mgm-3, with surface PAR averaging 326.5 µmolphotonsm-2s-1. Sites 7 (3.49 mgm-3), 12 (3.84 mgm-3) and 13 (4.06 mgm-3) recorded higher fluorescence concentrations than in August, all at the surface. Concentrations higher than 2 mgm-3 were observed at 8 sites, all located in the western part of the transect. Fluorescence declined again in October, with a mean surface of 1.11 mgm-3. The maximum fluorescence for this month (1.79 mgm-3) was measured at the surface of Site 32, followed by the surface of Site 9 (1.71 mgm-3).

The diffuse attenuation coefficient, Kd (PAR) ranged from 0.12 to 1.05 m-1, with an overall mean of 0.32 m-1, showing both seasonal and spatial variability along the transect (Fig. 7). June had the highest mean Kd (0.43 m-1), where sites 18, 20, 21 and 22 reached values between 0.58 and 1.05 m-1, the highest value recorded along the transect. October showed the second highest mean (0.37 m-1), particularly at sites influenced by rivers such as Site 11 (0.78 m-1) and Site 28 (0.68 m-1), where higher turbidity was observed that month. July had the lowest mean Kd (0.24 m-1), indicating clearer water conditions during the summer stratified period. Throughout all months, the eastern sites showed the lowest Kd values (0.24 m-1), while the BSI had the highest (0.35 m-1).

A significant positive correlation was found between Kd and depth-averaged turbidity (r2=0.59, p<0.001), while the relationship between Kd and depth-averaged fluorescence was weak (r2=0.02, p<0.05). Fluorescence was not a significant driver when included alongside turbidity in the multiple regression (p>0.05), suggesting that turbidity, rather than phytoplankton biomass, was the dominant driver of light attenuation, consistent with the elevated turbidity values observed at BSI and river-influenced sites.

In this data set, temperature, salinity, turbidity and PAR all had significant effects on Chl a, and the scatterplots and LOESS smoothers revealed nonlinear relationships (Figs. 8 and A1–A4). The highest Chl a concentrations were mostly observed at colder temperatures, particularly between 1 and 5 °C, and at higher salinities between 30 and 32 PSU. However, the monthly breakdown revealed that the direction of these relationships changed across the season. In May and June, Chl a increased with temperature (r=0.89 and 0.81), while in July the relationship reversed (r=-0.61), reflecting the subsurface fluorescence maximum where the highest concentrations occurred at deeper cold temperatures (Fig. A1). Concentrations then decreased with increasing temperature, with a secondary rise at warmer values, as suggested by the trend. This could reflect the fewer peaks of Chl a near the surface observed in autumn. A negative relationship between Chl a and salinity was observed in most months, suggesting that higher concentrations tended to occur at lower salinities, possibly linked to localized riverine inflow and associated nutrient inputs. July was an exception, where a positive relationship was observed, likely reflecting the subsurface fluorescence maximum (Fig. A2). Peaks of fluorescence were most abundant under low to moderate light conditions. When focusing on Chl a peaks only (>2 mgm-3), the relationship with PAR showed a weak negative correlation (r=-0.46, p<0.01) as opposed to a weak positive one with all the data (r=0.42, Fig. 8), consistent with the subsurface fluorescence maximum observed in July (Fig. A3). This was also visible at Site 9 (Fig. 3), where in July, peaks (>3 mgm-3) were below the thermocline at low PAR between 19.8 and 29.2 µmolphotonsm-2s-1. Turbidity measurements were mostly between 0 and 2 FTU during the sampling period, which coincided with most of the Chl a peaks and was positively associated with Chl a in most months (Fig. A4) Overall, the strength and direction of the relationships between Chl a and the environmental variables varied across months, highlighting the seasonal dynamics along the transect.