Carbonate chemistry data have been sampled in the Atlantic Zone since the fall of 2014 (except for the high-frequency Rimouski station where the sampling started earlier; see below). Depending on the region, between two and three surveys have usually occurred every year since. An overview of all missions, their dates and the parameters sampled is provided in Table .

The NL region, due to its extensive shelf and slope spatial extent, has a survey schedule that surveys the NL region three times a year, generally from the Southwest St. Pierre Bank (SWSPB) section up to the northern sea ice extent: usually the Bonavista Bay (BB) section during spring (April) surveys, from the Flemish Cap (FC) section to Beachy Island (BI) during summer (July) surveys and from the SWSPB section to Seal Island (SI) during fall (November) surveys (Fig. a). The surveys are most often incomplete due to operational constraints caused by sea ice, weather or research vessel downtime. The NL region survey schedule aims to sample at the surface (∼ 5 m), 10, 50, 75, 100, 150, 250, 500, 1000 m and the bottom. Sampling down the shelf slope to depths greater than 1200 m was, however, only feasible during fall surveys (maximum 3677 m) because of limited vessel capacity during other seasons.

The Maritimes region has respective spring (April) and fall (September/October, except November/December in 2017) surveys, which have been consistently carried out between 2014 and 2022 (Fig. a). The Scotian Shelf and Slope were sampled at the surface (∼ 2 m), 10, 50, 80, 100, 150, 250, 500, 1000, 1500 m and the bottom.

The GSL is sampled during spring (June) and fall (October/November) surveys along the standard sections at the surface (∼ 2 m), 15, 50, 100, 150, 200, 300, 400 m and the bottom (Fig. b). Samples have also been acquired during groundfish surveys in the GSL (Fig. b, Table ) during summer (August/September) 2017–2022 (Trip_Name: “IML2017027”, “PER2017018”, “PER2017024”, “PER2017016”, “PER2017026”, “MLB2017001b”, “IML201841”, “PER2018041”, “PER2018042”, “PER2018043”, “TEL2018196”, “IML2019036”, “PER2019140”, “PER2019008”, “TEL2019201”, “PER2020019”, “PER2021156”, “PER2021151”, “CJC2021222”, “IML2021030”, “PER2021021”, “IML2022039”, “PER2022022”, “CAR2022025” and “PER202223330”). Although these surveys are not within the regular AZMP program, they are staffed by the AZMP for oceanographic sampling and provide extensive spatial and added temporal (summer) coverage of carbonate parameters for the GSL.

The fall 2014 and spring 2017 AZMP surveys had the most complete spatial coverage of the Atlantic Zone among the entire time series. The other surveys are partial due to operational limitations specific to each region (see Table  for the timing of the different surveys per region).

In addition, two coastal stations are sampled at a greater temporal frequency. Rimouski station (labelled “Rimouski” in Figs.  and ), located in the LSLE, is sampled by the Québec region on a near-weekly basis. Station 27 (labelled “STN27” in Figs.  and ), located on the path of the coastal Labrador Current just outside St. John's Harbour is sampled by the NL region on a near-monthly basis. To illustrate their respective temporal coverage, the time series of pH at these stations (measured at Rimouski station and derived from TA–DIC at Station 27) are presented in Fig. . Note that Rimouski station is the only time series starting in the spring of 2014 in this dataset (all other sampling starts in the fall of 2014).

The collection of discrete water samples for carbonate analyses was achieved using a 12 or 24 Niskin bottle rosette equipped with a SeaBird conductivity–temperature–depth (CTD) unit coupled with an oxygen sensor to continuously measure depth, temperature, salinity and the dissolved oxygen (DO) concentration. Aliquots of water were subsampled from each Niskin bottle for nutrients (PO4, SiO3 and NO3+NO2) and DO, and they were then processed and analyzed according to established standard protocols . Although the DO concentration (in mL L-1) is available in this dataset, it was sampled less frequently than the other oceanographic parameters because its purpose was to calibrate the DO sensor measurements. In the GSL region, DO was sampled at the surface and bottom, whereas DO was sampled at the surface, the bottom and middle of the water column in the MAR region. In the NL region, DO was sampled at all depths but at a subsampled number of stations (roughly every two to four stations along a section). In the GSL and NL regions, the subset of bottle DO values was used to calibrate the DO sensor on the CTD unit, providing a DO value for each sample bottle (even when no Winkler titration was realized). For MAR data, only DO data obtained from Winkler titration are provided. Oxygen saturation (O2sat, in %) was calculated as the ratio between the measured DO concentration and its solubility referenced to the surface using the TEOS-10 toolbox and corresponding T–S observations.

Aliquots of water for the analysis of pH, total alkalinity (TA) and total dissolved inorganic carbon (DIC) (or a combination thereof) were subsampled following the protocols established by . Each aliquot was carefully drawn from the Niskin bottles, transferred into 500 mL gas-tight borosilicate reagent bottles (Corning, USA) and allowed to overflow by one volume while precluding air bubbles. A 5 mL volume was then removed to create a headspace for thermal expansion. The samples were preserved with 100 µL of saturated HgCl2 to prevent further biological activity. Each bottle was sealed with a ground-glass stopper, high-vacuum (Apiezon M) grease and a rubber band, and they were then stored in the dark between 4 and 18 ∘C for up to 6 months prior to analysis.

In 2014, the first year of the program, all three carbonate parameters (pH, TA and DIC) were analyzed for the Maritimes region at the Bedford Institute of Oceanography (BIO, Dartmouth, NS). In subsequent years, TA and pH were analyzed for the GSL region at the Maurice Lamontagne Institute (MLI, Mont-Joli, QC); DIC was added to the parameters analyzed in fall 2019, and pH was only measured in the spring of 2015 (Trip_Name “IML201516”). After 2014, TA and DIC were analyzed for the Maritimes (BIO) and NL (Northwest Atlantic Fisheries Centre – NAFC, St. John’s, NL) regions by their respective facilities. An overview of the carbonate parameters measured during each mission is provided in Table .

DIC (in µmol kg-1 of seawater) is extracted as CO2 by purging an acidified (1 M, 8.5 % phosphoric acid in excess) aliquot of water (warmed to 25 ∘C) with ultrahigh-purity (UHP) nitrogen gas using an automated sampling and gas extraction system. The dried gas, including the CO2, is transferred and absorbed into a coulometric cell and analyzed by titration and photometric detection . Various makes and models of auto-samplers and analyzers were used by the facilities in the different regions. Duplicate or triplicate sample measurements provided an analytical precision of 0.03 % at BIO and <0.1 % at MLI and NAFC.

TA (in µmol kg-1 of seawater) is measured by open-cell potentiometric titration with an automated sampling system. Each 50 mL (NAFC and BIO) or 104 mL (IML) sample, warmed to 25 ∘C, is titrated with 0.1 M hydrochloric acid to the Gran equivalence point (nonlinear curve-fitting method) using a computer-controlled Dosimat (Metrohm AG, USA) dispenser and combination glass electrode. In the NL region, aliquots of 0.1 M hydrochloric acid are added in a 25 ∘C thermostated jacket open cell while gently mixing the sample, and pH measurements are allowed to stabilize between successive readings. The analytical precision of TA at BIO was calculated using repeat analyses of bulk seawater and reported as ±0.05 %, while duplicate sample analyses revealed a precision of <0.1 % at IML and NAFC.

Seawater pH, expressed using the total hydrogen ion scale (pHT), was determined by spectrophotometry and calculated following the dye perturbation method . Purified m-cresol purple solution (University of South Florida) was added to seawater held at 25±0.05 ∘C in a 10 cm path-length quartz cell and mixed thoroughly. The ratio of blank-corrected absorbances measured at 434, 578 and 730 nm with a spectrophotometer (Agilent Technologies, USA) was used to determine pHT. pH measurements are not corrected for dye perturbation, but the m-cresol dye is visually inspected each day and measurements are made to ensure its stability over time. Precision and accuracy, evaluated daily by repeat measurements of a tris(hydroxymethyl)aminomethane (TRIS) buffer solution (Andrew Dickson, Scripps Institution of Oceanography, San Diego, CA) prepared with a salinity of 30 , were typically ±0.002 pH units at BIO and ±0.003 pH units or better at IML.

All analytical methods were calibrated with a series of seawater certified reference materials (CRMs; Andrew Dickson, Scripps Institution of Oceanography, San Diego, CA), which allowed for performance evaluations of the various instruments and normalization of the TIC, TA and pH measurements. As the analyses were conducted in three different regions, we will consider the largest analytical error provided for all measurements: TA and DIC values will be ±0.1 % or ±2.1 µmol kg-1 and pH will be ±0.003 units. These uncertainties are comparable to those suggested by for modern analytical techniques using reference materials. All data flagged as suspect because of analytical error by each laboratory have been removed from the dataset.

The aragonite saturation state (Ωarg), calcite saturation state (Ωcal, which is included in the dataset but not discussed in this report), pHT,is (total scale, in situ) and the partial pressure of CO2 (pCO2, in µatm) were calculated using the CO2SYS program modified for Python (https://github.com/mvdh7/PyCO2SYS/tree/v1.2.1, last access: 8 January 2023; ) as recommended for “best practices” by . The dissociation constants (K1 and K2) of , as refit by (Mdm); the total boron constant from ; and the KHSO4 constant from were also used as recommended for best practices . Although it has been suggested that dissociation constants formulated for estuarine waters should be used to avoid differences in the calculation of carbonate parameters at low salinity , there is some evidence that formulations other than those for best practices may produce discrepancies . The dissociation constants by Mdm were formulated for salinities >20. Samples with salinities <20 constitute only 0.15 % of this dataset (1 % of the data from the GSL), and the differences in calculated carbonate parameters using the constants by and Mdm are approximately double for samples with salinities ≥ 20 compared with those with salinities <20 (Table a). Therefore, the dissociation constants by Mdm were used in order to avoid discrepancies in the other 99.85 % of the dataset. The chosen combination of constants listed above was also suggested by because they generated the best internal consistency in their Labrador Sea (AR7W) dataset. However, their choice of constants provided less uncertainty than their choice of input parameters. The carbonate parameters in this dataset were calculated using the TA (µmol kg-1 SW, where SW denotes seawater) provided by all regions and either DIC (µmol kg-1 SW) or pHT, which is region dependent. According to the results by , the TA–DIC-based calculations of pH (e.g., NL and MAR regions) produce low accuracy but low uncertainty, and the TA–pH-based calculations of DIC (e.g., GSL dataset) provide the best consistency. The TA–pH pair will thus be used for calculating pCO2 when both pH and DIC were collected in addition to TA (see Table ). As the GSL region consistently measures the three carbonate equilibria parameters since the fall of 2019 (TA, pH and DIC; see Table ), we estimated the variability in the other derived parameters (in situ pH, Ω and pCO2) using the different combinations of pairs for this region between 2019 and 2020 as well as for 2022 (in 2021, only TA and pH were measured). Results are presented in Table d and discussed later in the text.

The spring 2015 GSL dataset (Trip_Name “IML201516”) consists solely of pH data (in vitro at 25 ∘C). In order to obtain in situ values and other carbonate parameters, TA was estimated using the TA–salinity linear relationship for S>20. Using spring data at all depths and remaining years in the GSL (excluding the high-resolution Rimouski station data) led to the following fit: TA = 43.09×S+808.31 (r2=0.98, p<0.001; see Fig. b). The estimated freshwater TA end-member (S=0) is 808.31 µmol kg-1. This value is lower than the measured values of 1204±99 µmol kg-1 or 1081±30.2 µmol kg-1 by near Québec City or the 1000 µmol kg-1 measured by west of Saguenay Fjord. It is, however, higher than the 186 µmol kg-1 calculated by as a measure of the Saguenay Fjord and North Shore rivers and the 80 µmol kg-1 measured by in the Saguenay River. Using a localized dataset for this calculation accounted for the low-salinity water exiting the St. Lawrence River. Considering the measurement error of up to 0.1 %, there was no significant difference in the predicted TA between the spring dataset and all seasons in the GSL. To estimate the degree to which the predicted TA had an effect on the calculation of the other carbonate parameters, the parameters for the spring 2015 GSL dataset were calculated in CO2SYS (TA–pH) with the maximum and minimum values of TA±2 SD (1.28 µmol kg-1), the TA–salinity regression uncertainty. This exercise shows the low sensitivity of this method, with an error of less than 0.03 % (Table c).

The calculated carbonate parameters were calibrated to in situ conditions using temperature (∘C), salinity (psu) and pressure (dbar). The nutrient alkalinity parameters, represented by total soluble reactive phosphorus (PO4, in µmol kg-1 SW) and total soluble reactive silicate (SiO3, in µmol kg-1 SW), were used as additional information when available, as they contribute to the calculation of the total carbonate alkalinity . There can be a slight offset in calculated carbonate parameters when the nutrient data are absent, especially in high-concentration regions. This offset has been calculated for the Atlantic Zone by obtaining the difference in carbonate parameters calculated with and without nutrients (nutrients set to zero in CO2SYS; ). The calculation was performed on two datasets: the TA–DIC input pair from the NL and MAR regions and the TA–pH input pair from the GSL region. The results (Table b) indicate that all calculated carbonate parameters except for pCO2 will be slightly higher when nutrient data are lacking, with the TA–DIC input pair having the largest offset. The differences in the calculated pH (0.002 pH units) and DIC (1.02 µmol kg-1) values are below the analytical uncertainty (±0.005 pH units and ±2.1 µmol kg-1, respectively). The mean saturation states and pCO2 have changed by <0.6 % for the TA–DIC pair and <0.1 % for the TA–pH pair. The larger nutrient–no nutrient difference with the TA–DIC pair may be due to the higher nutrient concentrations in the GSL region (>17.3 µM PO4 and >49.1 µM SiO3) or may reflect the input-parameter-propagated error as suggested by and , the latter of whom estimated the uncertainty in calculating the saturation state to be 3.7 % using the TA–pH and 1.7 % using TA–DIC.

However, the entire dataset used here is provided for follow-up studies (see the “Data availability” section). This dataset includes the following physical, biogeochemical and carbonate parameters: temperature (T); salinity (S); dissolved oxygen (DO) concentration; oxygen saturation (O2sat); nutrient data (PO4, SiO3 and NO3+NO2); measured and calculated (same as measured when available) TA and DIC; measured pH in the laboratory (when available); and calculated in situ pH, pCO2, Ωarg and Ωcal.

A series of temperature and salinity maps (Figs. a–d, , ) illustrate the physical oceanographic properties of the Atlantic Zone. The Arctic waters of the coastal Labrador Current on the northern NL Shelf are characterized by low salinities and temperatures near freezing, whereas the Scotian Slope surface is characterized by warmer (>18 ∘C) and more saline (>33) Gulf Stream waters (Fig. a). The lowest salinities (<27) are found in the surface waters of the LSLE (ignoring the Saguenay Fjord and the USLE; see Fig. ).

The moderately cold and saline surface waters along the southeastern NL Shelf and within the GSL consist of Arctic waters that have been modified by the Gulf Stream, St. Lawrence River or warmed in summer. The freshwater exiting the LSLE can be traced at the surface along the coast of the Gaspé Peninsula into the southern GSL and Northumberland Strait (mean S∼29), where it warms significantly in summer (mean summer T∼16 ∘C) (Figs. , ). The GSL encompasses the greatest sea surface temperature differences between spring and summer; however, the largest differences are from spring to fall and are found on the Scotian Shelf (>15 ∘C). As the AZMP does not include carbonate winter sampling, the coldest temperatures in this dataset, down to -1.7 ∘C, are located within the CIL, in the subsurface waters of the NL Shelf (e.g., hydrographic section SI; Fig. a) and in the GSL (e.g., hydrographic section LC; Fig. a). Warm, saline bottom waters influenced by the Gulf Stream can be traced along the southwest Scotian Shelf and Laurentian Channel into the GSL. Shallow bottom waters of the GSL can warm in summer and exceed 17 ∘C (not shown).

The oxygen saturation (O2sat; see Sect. 3) is greatest at the surface, with mean regional values >90 % and supersaturation occurring during spring plankton blooms (Fig. e). Along the southern NL and Scotian outer shelf and slope, an oxygen minimum zone with O2sat down to 45 % is centred around 250 m (e.g., hydrographic section SESPB; Fig. ). Mean bottom values on the NL Shelf range from nearly 100 % off Labrador to ∼ 55 % off southern NL. The O2sat reaches supersaturation in the shallow bottom waters of the southern GSL and undersaturation in deeper areas associated with the Laurentian Channel. Hypoxia, which occurs at the bottom of the LSLE and deeper areas of the GSL, such as the Laurentian, Anticosti and Esquiman channels (Fig. ; see also ), is shown as dark circles in Fig. f.

Changes in the physical environment partially drive the variability in the carbonate system. In recent years, the proportion of Labrador Current waters entering the deepest layers of the GSL in the Laurentian Channel has decreased at the expense of an increase in the North Atlantic Central Water . This translated into a significant increase in the temperature of the bottom layers of the GSL and a concurrent decrease in dissolved oxygen and pH . On the NL Shelf, the variability in the cold and fresh Arctic-origin waters flowing southward along the coast influences the carbonate system, with more acidic conditions expected during the colder and fresher years . On the Scotian Shelf, warmer and saltier conditions usually imply that the system is less prone to the undersaturation of the carbonate saturation states, but the changes in water masses' compositions may influence the interannual variability in carbonate parameters. In the next subsection, we describe the carbonate system and make links, when possible, with the oceanography of the region.

This section comprises a brief description of the spatial and temporal variations in the carbonate parameters of the Atlantic Zone. Except for 2015, the values and seasonal cycles among annual surveys are relatively comparable. For this reason, we chose the year 2017, one of the most complete year to date, both seasonally and spatially. Thus, in the following, descriptions made of the Atlantic Zone refer to maps and sections from 2017 (Figs. –, A1–A9). Comparing seasonal changes (spring, summer and fall) in the Atlantic Zone or changes by region is difficult due to the surveying protocols established in each region. The hydrographic sections FC and BB on the NL Shelf (Fig. a) are the only ones planned for the three seasons every year; however, the station separation is greater along BB, and the resolution is consequently poorer. Although the standard AZMP GSL sections are only surveyed in spring and fall, the summer groundfish surveys between 2017 and 2022 provide enough data to evaluate some seasonal trends in the GSL (see below).

An overview of the Atlantic Zone oceanographic setting and a description of the different carbonate parameters are provided in Sect.  and , respectively. A large range of values and seasonal variations were found for the different parameters presented. These elements are now discussed together in a broader regional and seasonal context.

The surface of the Atlantic Zone is characterized by cold (Arctic) and warm (Atlantic) waters that undergo a seasonal warming and freshening. This seasonal freshening generally lowers TA and, in turn, the capability of the system to buffer acidification. On the other hand, seasonal dynamics of phytoplankton blooms generally decrease inorganic carbon (e.g., DIC) in surface waters during the spring/summer as a result of photosynthesis. At depth, these values generally increase from spring to fall as plankton from the spring bloom are degraded and inorganic material is remineralized near the bottom.

The lowest pCO2 values are located in the cold coastal Labrador Current, whereas the highest values are found in the fresh waters of the LSLE. The lowest pH values are found in the GSL and the LSLE. This is especially true for the southern portion of the gulf and in the deep (>300 m) layers of the Laurentian Channel. The most acidic waters of the Atlantic Zone are found in the Lower St. Lawrence Estuary, where the pH decreased by 0.2–0.3 between the 1930s and the early 2000s; this region is also affected by severe hypoxia . These low and undersaturated conditions, which cover a large part of the Atlantic Zone, are concerning and unfavourable for calcifying organisms.

Ωarg increases at the surface and in shallow areas of the shelf from spring to fall at most MAR and NL stations as a consequence of the seasonal temperature increase and the decrease in S and TA coupled with the high DIC of respired organic matter (see also , for a description of the CO2 system seasonal cycle on the Scotian Shelf). Most of the bottom GSL is also undersaturated with respect to aragonite (Ωarg<1), except for a few saturated samples in shallow areas and along the Cabot Strait section. Consistent with , low Ωarg values are found in the cold and fresh (thus lower TA) Arctic-origin waters of the coastal Labrador Current. These cold, acidic waters are a prominent feature of the Atlantic Zone, as they contribute to the LSLW formation southwest of the Grand Banks. The LSLW, which has a Ωarg<2, partly forms the deep waters of the Laurentian Channel in the GSL and then continues southward on the Scotian Slope and along the northeastern US coast. These low Ωarg values can be traced to the deep Gulf of Maine. As the water temperature increases during the journey towards the southern US, Ωarg increases, reaching ∼ 4.5 in the Gulf of Mexico .

This study demonstrates clear links between the physical environment and the carbonate system in the Atlantic Zone. Waters around Atlantic Canada have warmed over the last century as a consequence of anthropogenic climate change, and this trend is expected to continue in the future . In this context, it is crucial to continue to monitor the physical and biogeochemical environment in the Atlantic Zone. While the 9-year time series presented here is relatively short, the data presented generally support other studies suggesting that the rate of change in ocean acidification and deoxygenation in Atlantic Canada may be greater than the global average . The situation is especially dramatic in the GSL where the warm waters entering the deepest layers of the GSL have significantly decreased the dissolved oxygen and pH since about 2019 (Fig. ; see also ).