Seasonal dynamics of carbonate chemistry, nutrients and CO₂ uptake in a sub-Arctic fjord

Coastal oceans and marine shelves are regions of enhanced uptake of atmospheric carbon dioxide (CO₂) and disproportionately large fraction of primary productivity relative to their areal coverage (Thomas et al., 2004; Borges et al., 2005; Thomas et al., 2009; Cai, 2011). This biological productivity is fuelled by oceanic, atmospheric, and terrestrial nutrient inputs that drive biogeochemical exchanges between the coastal and open ocean and enhanced burial of organic and inorganic carbon (Thomas et al., 2004; Chen and Borges, 2009). Observations in high latitude fjord and coastal regions have shown these regions to be predominantly sinks for atmospheric CO₂, despite large regional variability (Omar et al., 2005; Else et al., 2008b; Signorini et al., 2013; Fransson et al., 2014; Evans et al., 2015; Omar et al., 2016; Yasunaka et al., 2016; Ericson et al., 2018). Mid- and low-latitude coastal regions were thought to be predominantly oceanic CO₂ sources, owing to higher water temperatures and organic matter inputs (Borges et al., 2005; Cai et al., 2006; Chavez et al., 2007; Chen and Borges, 2009). However, increased observations and the development of high-resolution regional models have enabled better resolution of carbon cycling in these regions and identified areas of annual CO₂ uptake in the coastal ocean (Takahashi et al., 2009; Laruelle et al., 2014; Bourgeois et al., 2016; Fennel et al., 2019). Seasonal dynamics are more variable and typically amplified, with temporal changes occurring faster, in the coastal ocean compared to open ocean environments. The processes controlling the biogeochemical cycling in coastal systems are difficult to determine without sustained seasonal measurements of the hydrographic, biogeochemical and metrological variables (Borges et al., 2005; Bozec et al., 2006). Recent research efforts have significantly improved the understanding and quantification of carbon cycling in the coastal ocean, e,g, Fennel et al. (2019). However, seasonal baseline estimates of biogeochemical cycling and air–sea CO₂ exchange in some sub-Arctic and Arctic regions are limited. Greater spatial surveys would enable a better understanding of how climate change may affect carbon cycling at high latitudes.

The continental shelf and fjords of the Norwegian coast support a rich food web, cold-water coral reefs, large mammals (whales, seals), seabirds, valuable fish stocks and spawning grounds, and are used for recreation and aquaculture (Matthews and Sands, 1973; Erga and Heimdal, 1984; Salvanes and Noreide, 1993; Fosså et al., 2002; Asplin et al., 2014; Brattegard et al., 2011). The marine environment of the fjords is influenced by water mass circulation, tides, wind regimes, and freshwater inputs. Biogeochemical cycling and primary production exhibit strong seasonality that is controlled by variations in meteorology and hydrography (Eilertsen et al., 1984; Noji et al., 1992; Skarðhamar and Svendsen 2005; Eilertsen and Frantzen, 2007; Eilertsen and Degerlund, 2010; Wiedmann et al., 2016). Steep topography leads to orographic steering of winds, which strongly control mixing and stratification in the fjord (Cottier et al., 2010; Myksvoll et al., 2012), in addition to effects from tides and currents. Persistent down-fjord (to the fjord opening) winds induce upwelling of nutrient-rich coastal water that can stimulate phytoplankton growth in the fjord (Skarđhamar and Svendsen, 2005; Cottier et al., 2010). Up-fjord winds push surface waters coastwards and inhibit the upwelling of subsurface water and create a pressure gradient, which results in pulsed outflow of deep water (Skarđhamar and Svendsen, 2005; Cottier et al., 2010). Katabatic winds descend from the surrounding mountainous regions into the fjord and cool the surface water and enhancing mixing of the water column (Spall et al., 2017). Wind-induced turbulent mixing resuspends detrital and lithogenic material in shallow areas and influences biogeochemical cycling and export fluxes in the water column (Noji et al., 1993; Keck and Wassmann, 1996; Reigstad and Wassmann 1996).

Water column stratification is largely driven by salinity changes from freshwater input and warming/cooling in addition to currents, tides, and winds that drive vertical mixing with saltier subsurface water masses (Inall and Gillibrand, 2010; Cottier et al., 2010). These features influence particle transport, phytoplankton production and biogeochemical cycling (Klinck et al., 1981; Syvitski et al., 1987). The Norwegian Coastal Current transports fresher waters and supplies the fjords with nutrients and oxygen (Aure and Stigebrandt, 1989) to support enhanced primary production and a diversity of marine life (Erga and Heimdal, 1984; Salvanes and Noreide, 1993). From late autumn to winter, cooling, convective mixing and low light levels reduce phytoplankton activity and net respiration by heterotrophic organisms and organic matter remineralisation enriches the water column with inorganic carbon and nutrients (Noji et al., 1993; Eilertsen and Frantzen, 2007; Eilertsen and Dagerlund, 2010). From spring, light availability, stratification, temperature and day length increase with increases in phytoplankton biomass and productivity (Reigstad and Wassmann 1996; Eilertsen and Frantzen, 2007). Autotrophic activity utilises the winter stock of nutrients and drives biological carbon uptake during the growing season. Zooplankton abundance typically peaks during the spring bloom, declines throughout the summer as phytoplankton stocks diminish and often slightly increases during the smaller autumn blooms (Grønvik and Hopkins 1984; Michelsen et al. 2017).

High latitude surface waters are particularly sensitive to increases in atmospheric CO₂ and are likely to be the first areas to experience widespread ocean acidification; i.e., the lowering of pH and carbonate mineral saturation (Ω) states (Orr et al., 2005; Fabry et al., 2009; Doney et al., 2009). This sensitivity is due to the naturally low seawater carbonate ion concentrations as a result of greater freshwater inputs from melting sea ice, glacial meltwater, precipitation and river runoff (Chierici and Fransson, 2009; Fransson et al., 2013; Fransson et al., 2015). Compared to seawater, freshwater sources are low in total alkalinity (A_T), the natural buffer against acidity, and dilute the carbonate ion concentrations in seawater and decrease calcium carbonate (calcite or aragonite, CaCO₃) saturation. These processes enhance surface water acidification in coastal and seasonally ice-covered regions (Chierici and Fransson, 2009; Yamamoto-Kawai et al., 2009; Azetsu-Scott et al., 2014; Evans et al., 2014; Reisdorph and Mathis, 2014; Fransson et al., 2015). This enhancement occurs at potentially faster rates than the decreased pH and carbonate concentrations due to anthropogenic CO₂ uptake by the ocean; the impact on fjords will depend on the potential for atmospheric CO₂ uptake, as driven by biological production, and their geochemical buffering capacity to ocean acidification. Ocean acidification impacts the growth, metabolic processes and life cycles of marine organisms, especially those that precipitate CaCO₃ to form shells and skeletons (Orr et al., 2005; Fabry et al., 2008). When the calcium carbonate saturation decreases below the equilibrium threshold (Ω = 1) for carbonate precipitation and dissolution in seawater, the potential for CaCO₃ to dissolve increases. The rate of acidification in the open ocean of the Norwegian Sea has been well documented with long-term decreases in pH and aragonite saturation state (Lauvset et al., 2015; Jones et al., 2019). Few studies have been carried out to investigate the seasonal biogeochemical cycling and CO₂ uptake in northern Norwegian fjords, in contrast to the numerous studies from other northern fjord systems such as Svalbard (Omar et al., 2005; Fransson et al., 2014; Fransson et al., 2015; Ericson et al., 2019a; 2019b) and Greenland (Rysgaard et al., 2012; Meire et al., 2015).

This study presents the first time-series measurements of carbonate chemistry (dissolved inorganic carbon, C_T, and total alkalinity, A_T), macronutrients (nitrate + nitrite, phosphate, silicate) and δ¹⁸O covering a full annual cycle in a sub-Arctic fjord, Kaldfjorden. The data provide baseline hydrographic and biogeochemical measurements in the full water column during spring, summer, autumn and winter and emphasise the importance of time-series sampling to unravel the processes controlling the seasonal variability of carbon cycling in this region of the coastal ocean. The main objectives of this work were to (1) investigate the seasonal cycling in carbonate chemistry in the context of physical (freshwater inputs and water mass mixing) and biogeochemical (photosynthesis/respiration, remineralisation, calcification, air–sea fluxes) forcing; (2) estimate net community production (NCP) and annual air–sea CO₂ exchange; and (3) determine the current ocean acidification state.

Kaldfjorden (69.75°N, 18.68°E) is an ice-free fjord, 15 km long and about 2 km wide, on the western part of Kvaløya, Tromsø county, northern Norway (Figure 1). The fjord has a north–south orientation with a typical structure of U-shaped valley bounded by steep, glacially carved sides and is connected to the north Atlantic Ocean across the Norwegian shelf. Kaldfjorden has a partial sill between 75 and 135 m at the mouth and two basins (150–220 m) separated by a ridge. The seafloor shallows towards the inner part of the fjord. The Kaldfjorden marine environment hosts pelagic and benthic calcifiers, plays a role in seasonal migration of herring and supports aquaculture production of Atlantic salmon (Register of Aquaculture Permissions, 2018). The region experiences the polar night from the end of November to the end of January and 24-hour daylight from the end of May until the end of July. The growing season in northern Norway is typically between the end of March and October/November, with peaks in phytoplankton biomass during the spring bloom in mid/late April and a smaller autumn bloom can occur by the end of August or early September (Eilertsen and Frantzen, 2007).

The hydrography of Kaldfjorden is influenced by the circulation of water masses of coastal and Atlantic origin, mixed with local freshwater. The North Atlantic Current carries warm and saline Atlantic Water (S > 35; 5 < T ≤ 10°C) northwards along the Norwegian continental slope (Skarðhamar and Svendsen, 2005). The Norwegian Coastal Current carries colder and less saline Norwegian Coastal Water (S < 35; 4 < T ≤ 12°C) along the continental shelf (Saetre, 2007), which is freshened by riverine inputs (Nordby et al., 1999; Skarðhamar and Svendsen, 2005; Albretsen et al., 2012). Winter Mode Water is formed from cooling and convective mixing of the local fjord water during wintertime. Several small streams transport freshwater (terrestrial snow and ice melt, precipitation) into the fjord.

Time-series measurements of precipitation, air temperature and wind speeds were recorded at an hourly resolution by the Norwegian Meteorological Institute (www.eKlima.met.no) at the Tromsø observation site (Figure 2). Wind speeds recorded at Tromsø are a better proxy for conditions in Kaldfjorden as the orientation is very similar for both sites and orographic effects will therefore be similar; wind data at other proximal sites at Maasvik and Hekkingen fyr showed more dramatic orographic effects. Data are freely available and were retrieved on 13 September 2018.

Hydrographic measurements and water samples were obtained along three transects across the outer (T1), middle (T2) and inner (T3) parts of Kaldfjorden from small motorboats or larger research vessels (Table 1). Vertical conductivity-temperature-depth (CTD) profiles were obtained east–west along each of the transects at near-monthly resolution from November 2016 to July 2018. On small boats, a handheld CTD (SAIV SD208) was used. On research vessels, onboard Seabird Electronics SBE911+ were used. Onboard RV Johan Ruud and on the small boats, water sampling was carried out using a single Niskin bottle mounted above the CTD. The bottle was lowered several times per station to each sampling depth. On the larger research vessels, a CTD rosette with 12 Niskin bottles attached was available, and all samples were taken on the upcast of the CTD profile. Table 1 summarizes details of the sampling, vessels and CTD sensors that were used during this study. The SAIV CTD #1321 used from September 2017 until September 2018 was calibrated in summer 2017. When possible, the SAIV CTD was deployed simultaneously with the SBE CTD frame for calibration and intercomparison of the sensors. Castaway CTD sensors were used as further indicators for potential differences between the SAIV CTDs. All Seabird CTD sensors are factory-calibrated annually and the conductivity cells are further calibrated against salinity samples throughout the year. All CTD data were averaged into 1-dbar pressure bins, and the upcasts of the CTD were used for analyses due to more stable vertical speed during recovery. Salinity data are reported on the Practical Salinity Scale.

Freshwater fractions (FW) were determined using the in situ salinity (S) relative to the mean salinity (34.17 ± 0.24; S_ref) of Norwegian Coastal Water (Equation 1), as measured below 200 m at the outer fjord (T1) during the study period:

Biogeochemical water samples were taken from the central hydrographic station on each transect (T1, T2, T3) from 3–6 depths in the water column, typically at the surface (0–5 m), 25 m, 50 m, 70 m, 100 m, 150 m, and bottom, depending on the water column depth, which was 230 m (T1), 175 m (T2) and 115 m (T3). Samples for carbonate chemistry were drawn from the Niskin bottle via a silicon tube into 250-mL borosilicate bottles, preserved with saturated mercuric chloride (60 µL) and stored in the dark at 4°C. Analyses for total dissolved inorganic carbon (C_T) and total alkalinity (A_T) were carried out at the Institute of Marine Research in Tromsø, Norway, within 6 months. Following methods outlined in Dickson et al. (2007), C_T was determined using gas extraction of acidified (8.5% H₃PO₄) samples followed by coulometric titration and photometric detection (Johnson et al., 1987) using a Versatile Instrument for the Determination of Titration carbonate (VINDTA 3D, Marianda, Germany). The determination of A_T was carried out by potentiometric titration with 0.1 M hydrochloric acid in a semi-open cell using a Versatile Instrument for the Determination of Titration Alkalinity (VINDTA 3S, Marianda, Germany). The average standard deviation for C_T and A_T, determined from replicate sample analyses, was within ± 2 µmol kg^–1. Measurements were calibrated against Certified Reference Materials (CRM, provided by A. G. Dickson, Scripps Institution of Oceanography, USA).

Samples for macronutrients nitrate + nitrite (NO₃ + NO₂), nitrite (NO₂), phosphate (PO₄) and silicic acid (Si[OH₄]) were collected from the Niskin bottles into 20-mL vials, preserved with chloroform and stored at 4°C. Analysis was carried out at the Institute of Marine Research, Bergen, Norway, using a Flow Solution IV analyzer from O.I. Analytical, United States, following Grasshoff et al. (2009). The analyser was calibrated using reference seawater from Ocean Scientific International Ltd., United Kingdom. The semi-conservative tracer N* ([NO₃ + NO₂] – 16[PO₄]; Gruber and Sarmiento, 1997) was used to identify anomalies of nitrate + nitrite relative to phosphate, compared to global averages. As such, N* indicates periods of nitrogen replenishment or loss, as negative values suggest nitrogen deficits due to denitrification and positive values suggest nitrogen excess due to nitrogen fixation. However, these changes also include signals of advection of different water masses with different nutrient signatures.

Samples for determination of the ratio of the stable oxygen isotope of seawater (δ¹⁸O) were transferred into 20-mL vials and stored in the dark at 4°C until analysis using a Thermo Fisher Scientific Delta V Advantage mass spectrometer with Gasbench II. Data were standardised relative to Vienna Standard Mean Ocean Water (VSMOW) for δ¹⁸O (‰) with a reproducibility of replicate analyses of ±0.04‰.

Calcium carbonate (CaCO₃) saturation state (Ω) for the biomineral aragonite and the surface water fugacity of CO₂ (fCO₂) were determined from C_T and A_T, and in situ temperature, salinity, pressure and macronutrient concentrations using the CO₂ system program CO2SYS (Lewis and Wallace, 1998; van Heuven, 2011). The carbonic acid dissociation constants (pK₁ and pK₂) of Mehrbach et al. (1973) as refit by Dickson and Millero (1987) were selected, as they have shown good agreement between measured and calculated values in Arctic waters (Chen et al., 2015; Woosley et al., 2017) and were selected for similar studies in sub-Arctic/Arctic regions (Chierici et al., 2019; Ericson et al., 2019). Ω is used as an indicator for changes in carbonate chemistry in relation to ocean acidification. When Ω < 1, waters are undersaturated with respect to CaCO₃ and thus minerals are sensitive to dissolution.

Variability in surface water fCO₂ can be partitioned into temperature and biological signals by applying a temperature normalisation to the average surface water temperature (T_ave, 7.51 ± 3.17°C, n = 49) for all seasons (Takahashi et al., 2002):

where fCO_{2 T} (Equation 2) is normalised fCO₂ and T_obs is the temperature corresponding to the surface water fCO₂. The remaining fCO₂ variability is attributed to changes in C_T (assuming constant A_T) due to biological processes. This approach does not account for variations in C_T from other processes including A_T, riverine inputs, sediment fluxes and air–sea CO₂ exchange; these variations are incorporated into the biological signal.

The potential for air–sea CO₂ exchange is determined by the difference between CO₂ in the sea and overlying air. Fluxes of CO₂ (Equation 3) were calculated from the quasi-monthly air–sea gradient in fCO₂ (ΔfCO₂), the solubility coefficient of CO₂ (K₀) from Weiss (1974) and the gas transfer coefficient (K) which is a function of wind speed (Wanninkhof, 2014):

The uncertainty in the parameterization of the gas transfer velocity k was taken as 20% (Wanninkhof, 2014). The ΔfCO₂ is the difference between the calculated fCO₂ in surface seawater and the daily mean air fCO₂ value as determined from atmospheric CO₂ values. Atmospheric CO₂ data were obtained as mixing ratios (xCO₂) at hourly resolution at the Sammaltunturi observation site in Pallas, northern Finland (67.9736°N, 24.1158°E), operated by the Finnish Meteorological Institute (FMI). Data were accessed via the World Data Centre for Greenhouse Gases on 5 October 2019. Daily atmospheric CO₂ concentrations were averaged into monthly means and converted to partial pressures, using the mean monthly air pressures and the seawater vapor pressure as determined from seawater salinity and temperature of the surface waters (Ambrose and Lawrenson, 1972; Millero and Leung, 1976). The average fCO₂ in dry air was 397 ± 9 μatm (n = 46) during the study period (30 September 2016 to 06 September 2018). Negative values of ΔfCO₂ and CO₂ flux indicate surface water CO₂ undersaturation and uptake of atmospheric CO₂. The uncertainty in the atmospheric CO₂ concentrations is estimated as ±6 ppm as the maximum standard deviation of monthly means from daily concentrations, and the CO₂ flux uncertainty is estimated as ±0.1 mmol m^–2 day^–1. Wind speed data were averaged into monthly means and corrected to 10 m above sea level (Hartman and Hammond, 1985). The uncertainty in the wind speed data is estimated as ±10 m s^–1 as the root mean square error of one standard deviation per monthly mean from hourly observations. Air–sea CO₂ exchange per month was estimated from the daily CO₂ flux calculated for each (quasi-monthly) sampling event multiplied by the number in the respective sampling month.

The temporal evolution of C_T in the upper 50 m was determined from the quasi-monthly changes of C_T (Equation 4) during the full annual cycle in 2017–2018. The 0–50 m depth range was selected to encompass the seasonal mixed layer, and changes were determined by integrating from the surface to 50-m depth. The total change in C_T (ΔC_{T total}) is determined from the main processes that influence seawater C_T: salinity changes (ΔC_{T sal}), mixing with subsurface water (ΔC_{T mix}), photosynthesis/respiration (ΔC_{T bio}), air–sea CO₂ exchange (ΔC_{T flux}) and calcium carbonate formation/dissolution (ΔC_{T CaCO3}):

where ΔC_{T sal} is determined from the difference between the total monthly C_T change and the change in salinity-normalised C_T. Salinity normalisation using the traditional method (Equation 5) removes effects of dilution/enrichment, where variable (X) measured at in situ salinity (S) was salinity-normalised (X_sal) to the subsurface Norwegian Coastal Water salinity reference (34.17; as described above) following Friis et al. (2003):

From the C_T-salinity and C_T-δ¹⁸O relationships, values of C_T at zero salinity were estimated at -696 μmol kg^–1 and 651 μmol kg^–1, respectively (see section on Seasonality in freshwater and deep-water effects). The negative and positive estimates yield different interpretations of the freshwater endmember and the resultant C_{T sal} values, which are dependent upon the salinity normalisation method used (Friis et al., 2003). Therefore, in consideration of the significant difference in C_T endmember estimates, with respect to the sign of the value, and following the normalisation of inorganic nutrient data (with Equation 5), the C_T data were normalised using the traditional method, further discussed below (see section on Uncertainty assessment). ΔC_{T sal} integrates the signal from salinity changes due to freshwater inputs (decreased C_T and A_T) and advection of different water masses. ΔC_{T mix} was estimated from monthly changes in the mixed layer (inferred from changes in potential density) and the difference between the average C_T in the upper 50 m and the average C_T in the subsurface Norwegian Coastal Water (2137 ± 13 μmol kg^–1, n = 15). Here a deepening of the mixed layer (increased potential density) infers vertical mixing and increased C_T, adapted from Chierici et al. (2011). ΔC_{T mix} = 0 if there is no change or a shallowing of the mixed layer and integrates the signal from potential density increases (increased C_T and A_T) due to vertical mixing between sampling events.

Monthly changes in C_T due to photosynthetic fixation of C_T and production of organic carbon (ΔC_{T bio}) were determined by (1) using monthly changes in salinity-normalised nitrate and the C/N Redfield ratio of 6.6 (Redfield et al., 1963) to estimate C_T uptake (ΔC_{T bio N}) and by (2) residual difference between the total C_T change and the sum of all other contributing factors (ΔC_{T bio C}). These estimates yield the net community production (NCP), which describes the net primary production minus heterotrophic respiration. ΔC_{T bio} is negative when respiration exceeds photosynthesis, thus the reverse transformation recycles organic carbon back into its inorganic form. ΔC_{T flux} is determined from the air–sea CO₂ flux estimated at the time of each sampling event (quasi-monthly), multiplied by the number of days per respective sampling month. Changes in the ΔfCO₂ and wind speed are assumed to be linear (or the net result between positive and negative fluctuations) between each quasi-monthly sampling event. Negative fluxes indicate CO₂ undersaturation in surface water and atmospheric CO₂ uptake and thus yield a positive ΔC_{T flux}, i.e., input of C_T to the surface water. ΔC_{T CaCO3} accounts for changes in A_T that influence C_T, as outlined below.

Following the approach for C_T, monthly A_T changes were determined (Equation 6) from contributions due to salinity changes (ΔA_{T sal}), mixing (ΔA_{T mix}), a minor contribution from photosynthesis/respiration (ΔA_{T bio}) and calcium carbonate (CaCO₃) formation/dissolution (ΔA_{T CaCO3}), which likely includes terrestrial/benthic fluxes:

Changes in A_T due to salinity variations (ΔA_{T sal}) were determined using the salinity normalisation method that accounts for a non-zero freshwater endmember (Friis et al., 2003), with A_T of 337 μmol kg^–1 determined from linear regression analysis with S = 0 from the A_T-salinity relationship (presented in the section on Seasonality in freshwater and deep-water effects). ΔA_{T mix} was estimated as described for ΔC_{T mix} using an average A_T in the subsurface Norwegian Coastal Water (2282 ± 10 μmol kg^–1). Changes in A_T associated with the uptake and release of nitrate during photosynthesis/respiration (ΔA_{T bio}) can be estimated as one unit of NO₃ uptake increases A_T by one unit, therefore ΔA_{T bio} = –0.15 ΔC_{T bio} (Brewer and Goldman, 1976). Changes in A_T due to CaCO₃ formation/dissolution (ΔA_{T CaCO3}) are estimated by considering the potential alkalinity (A_T*); the sum of salinity-normalised A_T and NO₃ (Brewer and Goldman, 1976). Thus, $ΔCT CaCO3=12(ΔAT CaCO3)$⁠, which accounts for carbonate mineral precipitation and dissolution (Zeebe and Wolf-Gladrow, 2001). The ΔA_{T CaCO3} term is also likely to include any terrestrial and sediment/benthic carbonate fluxes.

Monthly ΔC_T and ΔA_T were used to determine the corresponding changes in ΔΩ for aragonite (ΔΩ aragonite). For each process (salinity changes, mixing, photosynthesis/respiration, air–sea CO₂ exchange, calcium carbonate formation/dissolution) the associated ΔC_T and ΔA_T was added to the C_T and A_T of the previous sampling event with in situ temperature, salinity and macronutrient concentrations in the surface layer. The CO2SYS program was then used to calculate the perturbed Ω to yield monthly estimates of changes in surface water Ω from the key contributing processes: ΔΩ_sal, ΔΩ_mix, ΔΩ_bio, ΔΩ_flux and ΔΩ_CaCO3.

Uncertainties in the determined effects of the physical and biogeochemical processes on ΔC_T, ΔA_T and ΔΩ were estimated as follows. Errors associated with monthly ΔC_{T total} and ΔA_{T total} were estimated to be ±0.2 mol m^–2 based on analytical precision of C_T and A_T (±2 μmol kg^–1). Uncertainties in ΔC_{T sal} and ΔA_{T sal} were estimated by consideration of the different normalisation methods that were applied (see section on Marine carbonate system). For the A_T endmember of 337 μmol kg^–1, the difference between salinity normalised and measured A_T ranged between –23 μmol kg^–1 and 131 μmol kg^–1. For the 1340 μmol kg^–1 endmember, the differences varied from –11 μmol kg^–1 to 58 μmol kg^–1. Thus, the salinity-derived (lower) endmember value yields a greater correction to the A_T values. For C_T and the traditional normalisation method, the difference between C_{T sal} and C_T ranged between –25 μmol kg^–1 and 143 μmol kg^–1. Using the 651 μmol kg^–1 endmember and the non-zero freshwater endmember normalisation, the differences were from –17 μmol kg^–1 to 95 μmol kg^–1. The upper bound of the A_{T sal} uncertainty can be considered as the maximum difference between the A_{T sal} values from each endmember, i.e., 131–58 μmol kg^–1, which is 73 μmol kg^–1. Similarly, the uncertainty in the C_{T sal} is estimated as 48 μmol kg^–1. Therefore, the upper bound of the ΔA_{T sal} and ΔC_{T sal} uncertainty (maximum difference between ΔA_{T sal} and ΔC_{T sal} for the two normalisation methods used for each) is 1.75 mol m^–2 month^–1 and 1.14 mol m^–2 month^–1, respectively. Using the traditional normalisation technique likely over-corrects C_{T sal}, as explained in Friis et al. (2003); however, the uncertainty is less compared to that estimated for the A_{T sal} methods, and therefore the standard normalisation method is considered suitable for the inorganic carbon (and nutrient) data in this study. Uncertainties in ΔC_{T mix} and ΔA_{T mix} were estimated to be ±0.4 mol m^–2 based on analytical precision of C_T and A_T (±2 μmol kg^–1). Uncertainties in ΔC_{T bio N} and ΔA_{T bio N} were estimated as ±0.1 mol m^–2 from the analytical precision of NO₃ of ±3% and the uncertainty in the C/N ratio, which was set to ±1 μmol kg^–1 to account for variations in the ratio from 6.6 (Redfield et al., 1963) to 6.7 (Frigstad et al., 2014). The uncertainties in ΔA_{T CaCO3} and ΔC_{T CaCO3} were estimated as ±0.2 and ±0.1 mol m^–2, respectively, from the analytical precision of NO₃ (±3%) and A_T (±2 μmol kg^–1).

Uncertainties in the calculated surface water fCO₂ (xCO₂) result from uncertainties in C_T, A_T, salinity, temperature, K1 and K2 that were added to each value of each property and used as inputs in CO2SYS to yield an upper bound error as ±9 μatm (±9 ppm). Based on a max value for k of 2.3, the associated error (20%; Wanninkhof, 2014) for k is ±0.46, thereby the uncertainty for ΔC_{T flux} is estimated as ±0.01 mol m^–2. Uncertainties in ΔC_{T bio C} and ΔA_{T bio C} are estimated as ±1.1 mol m^–2 from the sum of all associated uncertainties for each contributing term. Following the same approach, errors associated with ΔΩ _total were estimated to be ±0.08 based on uncertainties of Ω ± 0.04 from the input parameters run through CO2SYS. Using the associated errors for each ΔC_T and ΔA_T term in CO2SYS, uncertainties for ΔΩ were determined as ΔΩ _sal ± 0.12, ΔΩ _mix ± 0.12, ΔΩ _{bio N} ± 0.08, ΔΩ _flux ± 0.02, ΔΩ _CaCO3 ± 0.09 and ΔΩ _{bio C} ± 0.43.

Air pressure was stable during the summer and more variable for the rest of the year. Air temperatures were warmest (>20°C) in June, July and August and coldest (<–10°C) from January to March (Figure 2). Average daily precipitation was 2.7 ± 4.7 mm (n = 706), with intense events (10–32 mm day^–1) occurring sporadically throughout the year. Wind data measured every 6 hours showed that the average and the most frequent wind direction were 175 ± 90° (n = 2826) and 194°, respectively, which shows the impact of orographic effects at the Tromsø site. The prevailing southerly winds in Tromsø indicate that up-fjord (from the inner to outer part of the fjord) winds likely prevailed in Kaldfjorden. Wind speeds were on average 3.3 ± 2.1 m s^–1 (n = 2826), ranging from calm periods to strong gusts (up to 12.9 m s^–1). High pressure systems at the end of September and early October 2017 and February was accompanied by elevated temperatures and low rainfall. Winter storms (in December and January) were characterised by increased wind speeds, warmer air, and precipitation. June 2018 was a notably wet month relatively to the rest of the year. July 2018 had the warmest air temperatures (25–28°C) and very low precipitation as a result of higher and stable air pressure. Surface water temperatures were at the seasonal maximum during this period.

The water column was strongly stratified across Kaldfjorden from June to October (Figure 3). Highest potential temperatures (θ up to 12.27°C) were found in the upper 50 m from June to October 2017 and 2018 (Figure 4). Lower salinity (S ~ 32) water occupied the upper 20 m during this period (Figure 5). Subsequent cooling and convective mixing eroded the stratification in November and the water column was well mixed from December until May. During periods of weak stratification, the effects of tides are likely greater across the region (Skarðhamar and Svendsen, 2005). The fjord water resembled a modified variety of Norwegian Coastal Water (S < 34.8, 4 < θ ≤ 12°C; Nordby et al., 1999) with local effects of cooling, warming and freshening (Figure 6). The pycnocline persisted from May until November with modified Norwegian Coastal Water as the warmer and saltier watermass below 100 m. In the outer part of the fjord (T1), subsurface coastal water (below 200 m depth) had mean S of 34.17 ± 0.24 and θ of 5.35 ± 1.64°C (n = 16). Lowest salinity surface water (S of 31.85), thus highest freshwater fraction (Figure 7a,i,q), was found in summer in the inner fjord. Temperature maxima occurred in the surface layer in July 2018. The stable oxygen isotope (δ¹⁸O) varied between –0.26‰ and 0.76‰ in the water column (Figure 7b,j,r). Lower (isotopically depleted) δ¹⁸O is a signal of meteoric water input and was found in the fresher, stratified surface layer from May to November. Higher δ¹⁸O (isotopically enriched) values are indicative of coastal water (with δ¹⁸O 0.44 ± 0.21‰) in the full water column from December until April.

Water column concentrations of nitrate (NO₃), phosphate (PO₄) and silicic acid (Si(OH)₄) had ranges of 0–10.69 μmol kg^–1, 0–0.97 μmol kg^–1 and 0–12.53 μmol kg^–1, respectively (Figure 7). Lowest concentrations and episodic depletion of all macronutrients occurred in the surface layer between June and October across the fjord. Concentrations typically increased with depth but were spatially variable; enriched NO₃ and Si(OH)₄ was found at 100-m depth in the inner fjord in December (Figure 7s,u). The modified Norwegian Coastal Water (below 200-m depth at T1) had average (n = 16) concentrations of NO₃, PO₄ and Si(OH)₄ of 3.79 ± 3.55 μmol kg^–1, 0.35 ± 0.31 μmol kg^–1 and 4.49 ± 1.19 μmol kg^–1, respectively, that re-supplied the upper layers during vertical mixing.

The distribution of C_T and A_T showed lower values in the upper 50 m from June to November across the fjord. Surface water C_T was lowest (<2050 μmol kg^–1) in summer (Figure 7f,n,v) and autumn, and A_T was lowest (<2210 μmol kg^–1) from spring to autumn (Figure 7g,o,w). Reductions in sea surface C_T and A_T occurred from April, reaching lowest values of 1958 μmol kg^–1 and 2136 μmol kg^–1, respectively, in June 2018. Concentrations of C_T increased with depth to >2105 μmol kg^–1 below 100 m. The modified Norwegian Coastal Water (below 200 m depth at T1) was characterised by high A_T and C_T of 2257–2296 μmol kg^–1 and 2108–2154 μmol kg^–1, respectively, throughout the study period. Highest C_T of ~2170 μmol kg^–1 was found close to the seafloor in November in the inner fjord. Aragonite saturation (Ω) was highest (2.20–2.33) in the upper 50 m from June to November across the fjord (Figure 7h,p,x). Values decreased with depth to low Ω of 1.34–1.66 below 150 m and lowest near the seafloor in the inner fjord.

Surface waters (0–5 m) were relatively warm (9.95–12.27°C) and fresh (<33.10) with isotopically light δ¹⁸O (between –0.1‰ and –0.2‰) from July to October (Figure 8a–c). NO₃, PO₄ and Si(OH)₄ were rapidly reduced in April, with NO₃ and PO₄ nearly totally depleted from May to October (Figure 8d–f). Lowest C_{T sal} (2059–2065 μmol kg^–1), controlled by biological drawdown, and A_{T sal} (2250 μmol kg^–1), from a likely calcification signal, occurred in September and October (Figure 9b–c). Depleted Si(OH)₄ also occurred at this time. Surface water fCO₂ was lowest (270–294 μatm) and strongly undersaturated (ΔfCO_2(sea-air) of –131 μatm) in April and May (Figure 9d–e). Greatest CO₂ uptake (2.7 mmol m^–2 day^–1) occurred when ΔfCO_2(sea-air) was large (–129 μatm) and monthly wind speeds exceeded the yearly average (3.2 m s^–1) in May. Ω increased to highest saturation states (2.26–2.33) in September (Figure 9f). Seawater CO₂ is influenced by temperature (+1°C raises fCO₂ by about 10 μatm; Takahashi et al., 1993). For a seasonal surface water increase of 10.1°C, increases in fCO₂ up to 100 μatm could be expected. Surface water fCO₂ was found to increase by ~60 μatm, thus the thermodynamic effects were compensated by photosynthetic CO₂ uptake. When the effects of temperature were removed, variations in fCO_{2 T} showed similar changes compared with fCO₂ (Figure 9d) but with a larger seasonal amplitude, suggesting that biological processes dominated over thermodynamic control of fCO₂ to drive strong seasonality in surface water fCO₂ across the fjord (Figure 10).

The lowest salinity and highest freshwater fractions (5–6%) in June 2018 were accompanied by increases in Si(OH)_{4 sal} (~0.8 μmol kg^–1) and C_{T sal} (~25 μmol kg^–1), indicating terrestrial sources of (remineralised) organic matter and weathered minerals entered the fjord during intense precipitation and runoff events. Cold (2.17–3.76°C) and saline (33.58–33.87) surface waters in March and April had enriched δ¹⁸O due to vertical mixing with subsurface coastal water. Respiration, remineralisation and vertical mixing of the water column increased C_{T sal}, NO_{3 sal}, PO_{4 sal} and Si(OH)_{4 sal}, with a potential contributions from silicate dissolution, from January to March. The maximum absolute difference between measured and salinity-normalised concentrations for NO₃, PO₄ and Si(OH)_{4 sal} was 0.13 μmol kg^–1, 0.01 μmol kg^–1 and 0.15 μmol kg^–1, respectively, and relatively minor compared to 143 μmol kg^–1 for C_T and 131 μmol kg^–1 for A_T. Increases in A_{T sal} by ~40 μmol kg^–1 occurred from early autumn to November. Surface water Ω was less variable during winter/early spring and lowest values (1.59) in March. fCO₂ was highest (381–388 μatm) between September and February, and fCO_{2 T} peaked in January–March, which coincided with the period of convective mixing and colder temperatures. Surface waters were undersaturated with respect to atmospheric CO₂ with an average ΔfCO_2(sea-air) of –58 ± 33 μatm (n = 46), driving air–sea CO₂ exchange of –0.86 ± 0.63 mmol m^–2 day^–1 (n = 46; atmospheric CO₂ uptake). On an annual basis, the surface layer in Kaldfjorden is estimated to be a sink for atmospheric CO₂ of 0.32 ± 0.03 mol C m^–2 yr^–1 (n = 12).

The largest total monthly change in C_T (ΔC_{T total}) in the upper 50 m was –4.1 mol m^–2 month^–1, in the inner fjord, driven by changes in salinity ΔC_{T sal} of –3.3 mol m^–2 month^–1 (accounting for 81% of ΔC_{T total}) due to freshwater inputs in June (Figure 11). ΔC_{T mix} showed little variation in its range of 0–0.1 mol m^–2 month^–1, constituting up to 5% of ΔC_{T total}. The upper percentages of ΔC_{T mix} contributions coincided with the timing of erosion of water column stratification and increased C_T in the upper layers from mixing with subsurface Norwegian Coastal Water, notably from December to April. ΔC_{T bio} (ΔC_{T bio C}; determined from the residual difference between ΔC_{T total} and all other factors) ranged between –0.4 mol m^–2 month^–1 and –0.8 mol m^–2 month^–1 during the spring bloom and summer productive period from April to June. Biological production accounted for 20–30% of ΔC_{T total} and coincided with maximum ΔC_{T bio N} of –1.2 mol m^–1 month^–1 from the rapid NO₃ drawdown in April. A second peak in biological C_T drawdown occurred in August with ΔC_{T bio C} up to –0.5 mol m^–2 month^–1 as a signal of late summer production and a possible autumn bloom. Greatest increases in ΔC_{T total} of 1.4–1.9 mol m^–1 month^–1 were driven by ΔC_{T sal} of 1.1 mol m^–2 month^–1 (57% of ΔC_{T total}) and ΔC_{T bio C} of 0.4–0.7 mol m^–2 month^–1 (37% of ΔC_{T total}) from September to October. These increases were a result of reduction in freshwater fluxes and increased surface layer C_T from respiration and remineralisation (as indicated by higher nutrient concentrations) in post-bloom conditions. $ΔCT CaCO3(12[ΔAT CaCO3])$ was lowest at –0.3 mol m^–2 month^–1, which accounted for up to 8% of the reduction in ΔC_{T total} due to the presence of calcifying phytoplankton. Highest ΔC_{T CaCO3} of 0.3 mol m^–2 month^–1 represented up to 15% of ΔC_{T total} at the time of intense freshwater runoff, indicating terrestrial inputs of A_T, and a decaying coccolithophore bloom in September. Excess A_T (positive ΔC_{T CaCO3}) could also result from particulate inorganic carbon, e.g., CaCO₃ shells in the water column that are captured by sampling and dissolve upon analysis. ΔC_{T flux} range was 0–0.1 mol m^–2 month^–1 from the increased C_T in the upper layers due to uptake of atmospheric CO₂, which was highest in May and June (up to 4% of ΔC_{T total}).

Monthly change in Ω for aragonite (ΔΩ _total) ranged between –0.43 and 0.37 across Kaldfjorden (Figure 12). Greatest increases in ΔΩ were driven by biological production (decreased C_T, slight increase in A_T) as ΔΩ _{bio C} (and ΔΩ _{bio N}) varied between 0.25 and 0.40 from April to June. Biological production dominated the other processes and accounted for up to 99% of the monthly change in Ω. Increases in ΔΩ _{bio C} up to 0.20 in August and September coincide with biological C_T drawdown in a late summer/autumn bloom. Respiration (increased C_T, slight decrease in A_T) in post bloom and winter conditions dominated the monthly decreases in ΔΩ _total with ΔΩ _{bio C} of –0.50. Lowest (negative) ΔC_{T CaCO3} corresponded to lowest (negative) ΔΩ _CaCO3 of –0.09, and 21% of ΔΩ _total, as a result of calcification (greater decreases in A_T relative to C_T) in spring and summer. Lowest (negative) monthly changes in ΔC_{T sal} corresponded to highest (positive) ΔΩ _sal up to 0.11 to reveal the net effect of C_T dilution offset the effect of decreases in A_T from freshwater inputs. Minor changes in ΔΩ resulted from mixing with subsurface Norwegian Coastal Water, which was largest in winter and early spring with ΔΩ _mix of –0.04 (representing up to 10% of ΔΩ _total in December). ΔΩ _flux ranged between –0.03 in December and January and 0.05 in August from changes in monthly CO₂ uptake and subsequent increased C_T in the surface layer.

Higher freshwater fractions (4–7%) and a shift towards depleted δ¹⁸O (–0.26‰) from June to September coincided with warmer air temperatures and higher precipitation. The effects of increased precipitation and river runoff dominated the greatest monthly change in C_T of –4.1 mol C m^–2, representing reductions of C_T from dilution effects that accounted for 81% of monthly C_T deficits in the inner part of the fjord in June. Summertime values of C_T and A_T in the surface layer of Kaldfjorden were similar to those reported for Svalbard fjords (Fransson et al., 2015; Ericson et al., 2019b) but generally higher than those of other high-latitude coastal and fjord systems (Table 2). The differences are largely due to greater meteoric water inputs in the other high-latitude regions compared to Kaldfjorden, which include glacial meltwater and contributions from sea-ice melt that result in higher dilution effects that lower A_T (and C_T). In combination with strong biological carbon uptake, the fjord and coastal regions at higher latitudes and inland seas have nominally lower surface water A_T and C_T compared to Kaldfjorden.

The importance of freshwater in controlling the carbonate system and nutrient dynamics in Kaldfjorden is evident from the relationship between salinity and A_T and C_T during a full annual cycle (Figure 13a). The strong correlation between A_T and salinity yielded A_T = 56.9S + 337 (r² = 0.84, se = 55 μmol kg^–1, p < 0.0001, n = 234), which shows evidence of a freshwater (zero salinity) endmember for A_T of 337 ± 55 μmol kg^–1. For C_T and salinity, the relationship yielded C_T = 82.9S – 696 (r² = 0.76; se = 102 μmol kg^–1; p < 0.0001: n = 235), which indicates a deficit in C_T in freshwater, as reported for freshwater A_T endmember estimates from salinity relationships by Turk et al. (2016). A second approach in estimating the freshwater A_T endmember is using the relationship with δ¹⁸O, as a comparison. Firstly, the meteoric δ¹⁸O signature of freshwater was estimated as –10.1‰ from the relationship between δ¹⁸O and salinity (r² = 0.43, se = 0.79‰, p < 0.0001, n = 226). As no direct measurements of freshwater endmembers are currently available for Kaldfjorden, the δ¹⁸O (–10.1‰) value determined here indicates that the freshwater sources are predominantly of meteoric origin (snow melt, precipitation and river runoff). The estimated δ¹⁸O value in Kaldfjorden is similar to that of the Hudson Bay rivers (latitude ~60°N) and falls at the higher end (isotopically heavier) of previously reported ranges of δ¹⁸O in meteoric water endmembers in other high latitude fjord and coastal systems (Table 3). The spatial variability of the freshwater δ¹⁸O signature is due to the fact that precipitation becomes increasingly light isotopically at higher latitudes, in addition to localised variations such as influences of glaciers in sub-Arctic regions, as suggested by Turk et al. (2016).

When applied to the A_T-δ¹⁸O linear regression A_T = 88.7δ¹⁸O + 2236 (r² = 0.44, se = 6.7 μmol kg^–1 ‰^–1, p < 0.0001, n = 223), the A_T endmember is estimated at 1340 ± 9 μmol kg^–1. The range of the estimated freshwater A_T endmember (337–1340 μmol kg^–1) for Kaldfjorden falls within the range of Arctic rivers (Cooper et al., 2008) and for the Svalbard fjords (Fransson et al., 2015; Ericson et al., 2019a, 2019b), which are influenced by glacial runoff with a watershed containing carbonate and silicate bedrock. Values were higher than those for meteoric endmembers dominated by glacial meltwater in Greenland (Meire et al., 2015) and for Cumberland Sound (Turk et al., 2016). The associated freshwater source of A_T likely originates from terrestrial inputs, e.g., weathered minerals of surrounding rock. The regression analysis showed that freshwater in Kaldfjorden had a diluting impact on surface water A_T. The freshwater-derived A_T would act to slightly decrease dilution effects; however, the overall effect from freshwater inputs is a lowering of A_T (and C_T) in the surface layer. Variations in the A_T content of meteoric water have been previously attributed to river drainage over carbonate and silicate-rich rocks that subsequently become enriched with minerals and transported into the fjord and coastal waters (Hjalmarsson et al., 2008; Azetsu-Scott et al., 2014; Fransson et al., 2015; Ericson et al., 2019a).

Similarly to A_T, C_T decreased with increasing freshwater inputs in the upper layer of the fjord. From linear regression analysis with the local δ¹⁸O endmember (–10.1‰), A_T = 140.1δ¹⁸O + 2066 (r² = 0.47, se = 9.9 μmol kg^–1 ‰^–1, p < 0.0001, n = 224), the freshwater C_T endmember was estimated as 651 ± 13 μmol kg^–1. In contrast to the endmember estimates for A_T, the C_T estimates are similar in magnitude but opposite in sign. These findings suggest that the freshwater salinity and δ¹⁸O system could be influenced by contrasting processes of inorganic carbon removal and enrichment, respectively, depending upon which estimation method was selected. The δ¹⁸O-based endmember is similar to the estimated C_T endmember in Tempelfjorden, Svalbard (Ericson et al., 2019a). This similarity leads to the hypothesis that freshwater runoff could contain a source of C_T derived from atmospheric CO₂ uptake and terrestrial organic matter, which was subsequently remineralised upon transport to and release into the fjord. Increased surface water Si(OH)₄ and C_{T sal} of ~25 μmol kg^–1 that was linked to persistent precipitation, most notably at the inner part of the fjord, indicated a supply of dissolved silica, perhaps from weathered silicate minerals, within freshwater runoff (Figure 13b). The enhanced surface water C_{T sal} associated with freshwater fluxes enriched the C_T pool of the fjord and led to reductions in Ω. However, any additional C_T was likely assimilated during biological production during the summer and thus constitutes a more transient signal following prolonged precipitation as observed in June 2018. Enhanced A_T and silicates in freshwater delivered to Kaldfjorden could constitute a minor buffer in the surface layer against CO₂ increases and provide an additional source of silicate to siliceous plankton species, e.g., diatoms. Future warming and increased precipitation and runoff in the sub-Arctic will result in higher freshwater inputs but, in regions of calcareous and siliceous bedrock, a terrestrial supply of dissolved minerals could act to slightly counteract dilution effects on seawater A_T.

The seasonality in A_T and C_T in the modified Norwegian Coastal Water below 200 m at the mouth of the fjord varied from lower values in winter and higher values in summer and autumn as evidence of convective mixing of the surface layer and subsurface coastal water when stratification becomes eroded (Figure 7f–g). Deep vertical mixing in winter and spring, as indicated by a shift in δ¹⁸O from higher (~0.6‰) to lower (~0.1‰) isotopic values, homogenised the water column and enabled the coastal water source of macronutrients to be entrained into the upper layers and the freshwater-influenced surface to permeate deeper (Figure 7a–b). As such, the A_T and C_T signal from the productive surface layers was dispersed into the water column and lowered A_T and C_T. Following spring and summer biological production, enriched C_T and macronutrients were found at 150–250 m depth, with highest concentrations in November and December. This is due to enrichment from organic matter sinking out of productive surface waters and being remineralised in the subsurface and deep water. Seasonal variability in deep water A_T and Si(OH)₄ closely followed variations in salinity and δ¹⁸O. Higher A_T in November and December likely includes a contribution from sediment resuspension, superimposed onto the relatively elevated A_T signal from Norwegian Coastal Water (with δ¹⁸O of ~0.8). Increased concentrations of Si(OH)₄ arise from remineralisation of silicates, e.g., from diatoms and sediment resuspension, during wind-induced mixing.

The impact of these competing processes that increase and decrease C_T and A_T, largely removed net seasonality from Ω variations below 150 m, yielding a narrow range of Ω (1.5–1.7) throughout the year. A notable increase in concentrations of all macronutrients and C_T was found close to the sea floor at the inner part of the fjord in December 2017, which resulted in the lowest Ω of 1.3 during the time series. This low Ω could result from remineralisation, sediment fluxes or perhaps a manmade signal, for example water discharge in the shallower part of the fjord closest to land. This signal appeared to be localised and showed impacts to carbonate chemistry in areas exposed to greater land-water interactions, i.e., higher C_T inputs reduced Ω. Removing the effects of salinity changes on C_T and A_T (Figure 13c) reveals further seasonal dynamics driven by biological production, calcification and atmospheric CO₂ uptake (discussed in the following section).

The importance of biological processes on the carbonate chemistry in Kaldfjorden is reflected in the seasonal NCP and substantial C_T drawdown during spring and summer (Figure 11). With greater light availability, phytoplankton biomass increased in early spring and developed into a bloom of diatoms and Phaeocystis with chlorophyll a concentrations up to 10 mg m^–3 in April (Persson, 2018). Weak stratification and nutrient replenishment from subsurface waters and recycling in the upper water column fuelled biological production. Peaks in particulate matter fluxes and chlorophyll a fluxes, out of the surface layer during the spring bloom, have been observed in a neighbouring fjord, Balsfjorden in Tromsø (Eilertsen et al., 1981; Eilertsen and Degerlund, 2010). Coupled to strong biological carbon uptake and conversion to organic carbon, subsequent export of organic matter would contribute to reductions in water column C_T. Some organic matter may settle out of the water column and become buried in the sediments, driving seasonal ΔC_T depletion in the upper layer of the fjord during April. The abundance of zooplankton, e.g., copepods, was low, and export of particulate organic carbon likely continued prior to the development of strong stratification by late spring that would impede export of particulate matter to depth (Wassmann et al., 1991; Walker, 2018). By June and July, stronger stratification likely inhibited export of organic matter across the pycnocline, and thus respiration and recycling in the upper layers contributed to reduced monthly change in ΔC_{T bioN} and net respiration signals relative to April.

NCP estimates were determined from the residual difference between the total monthly change in depth-integrated C_T and the sum of all other contributing factors (ΔC_{T bio C}; NCP_C) and from the total monthly change in depth-integrated NO₃, corrected for salinity changes (ΔC_{T bio N}; NCP_N). During the productive months (growing season) of April–August, NCP_C was 14 ± 2 g C m^–2 (n = 5), which is about twice as large than the NCP_N estimate of 6 ± 2 g C m^–2 (n = 5). The differences in NCP_C and NCP_N for both seasonal and annual estimates arise from the sensitivity of the nitrate-based estimates to the C/N ratio selected. NCP is often computed using the Redfield ratio of C:N:P 106:16:1 (Redfield, 1963), which is most suitable when nitrate and phosphate are not depleted and C_T, NO₃ and PO₄ are assimilated and regenerated following the Redfield proportions (Arrigo, 2005). The relationship between C_T and NO₃ (Figure 13d) yielded an average C/N of 7.5 (r² = 0.26, se = 0.83 μmol kg^–1, p ≪ 0.001, n = 233) with a tendency towards lower C/N uptake ratios in the spring as a result of rapid NO₃ consumption as C_T remained high, followed by a shift to higher C/N uptake ratios in the summer and early autumn upon intense drawdown of C_T during high NCP. These C/N uptake ratios are higher than the Redfield C/N ratio (6.6) and consistent with previous observations of Frigstad et al. (2014) of C/N ratios that are higher relative to Redfield stoichiometry in northern high latitude regions, such as 6.7–7.0 in the Norwegian Sea region.

The strong de-coupling of C_T and NO₃ at near-total depletion of NO₃ persisted during the growing season, suggesting that rapid recycling of nutrients, production of nitrogen-poor organic matter and/or other sources of nitrogen, such as ammonia, could be important factors (Kähler and Koeve, 2001). Furthermore, the C/N uptake ratio is likely to vary due to phytoplankton species composition (Sambrotto et al., 1993) and availability of dissolved iron (Takeda, 1998). As such, carbon-based estimates often exceed nitrogen-based estimates of NCP (Bozec et al. 2006; Tremblay et al. 2008; Ericson et al., 2019b). In addition, changes in light, temperature, salinity and availability of micro-nutrients are all likely to play a role. Therefore, using the traditional Redfield utilisation/replenishment ratio would not constrain the depletion in C_T relative to NO₃ in Kaldfjorden, and N-based NCP estimates would not capture the extent of biological carbon uptake. Lower C/N in the winter and spring coincided with the smallest biologically driven monthly C_T deficits and was dominated by enrichment from respiration and remineralisation. Breakdown in water column stratification, mixing and diminishing light led to a steep decline in chlorophyll a concentrations (0.03–0.12 mg m^–3) and zooplankton abundance from October to February (Walker, 2018). The closer coupling of C and N at higher concentrations is evidence of export and remineralisation of organic matter in (sub-)surface waters, with likely contributions from sediment resuspension following episodic high winds (Walker, 2018) during winter.

The relationships between NO₃ and PO₄ (N/P; Figure 13e) and Si(OH)₄ and NO₃ (Si/N; Figure 13f) inform about the differences in supply and consumption of macronutrients. The average N/P was 11.6 (r² = 0.83, se = 0.34 μmol kg^–1, p ≪ 0.001, n = 240) and shows close coupling of inorganic nitrogen and phosphorus. The N/P values were slightly lower in winter and higher in autumn. The temporal trends in N* (Figure 14a) and lower N/P observed during winter and early spring result from the rapid removal of NO₃, and PO₄, in the spring bloom. NCP increased from April as primary production exceeded respiration and caused a rapid reduction in C_T in the surface layer. Reductions in N* from May show a shift in the system as nitrogen is depleted by ongoing biological production combined with denitrification and/or advective losses and remineralisation in subsurface water of organic matter with lower N/P ratios. The average seasonal Si/N was 0.3 (r² = 0.28, se = 0.03 μmol kg^–1, p ≪ 0.001, n = 240) and shows variability and weak coupling with an excess of Si(OH)₄ relative to NO₃ at low NO₃ concentrations. Surface water NO₃ was depleted from May to October, while Si(OH)₄and C_T were continually consumed due to biological production by a diatom community. In addition, any iron limitation in the fjord would result in phytoplankton assimilating less NO₃ relative to Si(OH)₄, thus contributing to higher Si/N (Takeda, 1998), as observed in autumn.

The time lag in Si(OH)₄ drawdown of about one month, accompanied by lower N/P, reflects the species succession and prevalence of diatoms in the phytoplankton assemblage. The lower N/P uptake ratios coinciding with near depleted NO₃ concentrations suggests that intense nitrogen recycling occurred in the mixed layer and that other sources of nitrogen, such as ammonium, could be important to sustain phytoplankton production. The variable NO₃ (and PO₄) concentrations below the mixed layer integrate the effects of uptake by phytoplankton in surface waters and regeneration from remineralisation of sinking organic matter and mixing with underlying waters. Due to the shallowing seafloor towards the inner fjord and deep convective mixing, a substantial fraction of organically fixed carbon and nutrients (and exported silicates) is likely respired and returned to the water column. This mechanism would resupply the surface layers with NO₃, and possibly iron, to maintain biological production and induce high seasonality of C_T and likely lower the potential for organic carbon burial in the sediments (Smith et al., 2015) in the inner fjord.

Surface waters were undersaturated with respect to atmospheric CO₂, but oceanic CO₂ uptake of 3.9 ± 0.3 g C m^–2 yr^–1, estimated for the full annual cycle, would resupply the upper layers with C_T. However, a net monthly decrease in the deficit in C_T is observed due to the dominating effects of freshwater inputs and NCP. The growing season NCP_C estimate for Kaldfjorden (14 ± 2 g C m^–2) is smaller compared to carbon-based NCP estimates of 49 g C m^–2 for the Amundsen Gulf region (Shadwick et al., 2011) and 65–85 g C m^–2 in Adventfjorden in Svalbard (Ericson et al., 2019a). The seasonal NCP_C is slightly less than the equivalent NCP_N,P estimates for the Nordic and Barents seas and Canadian Archipelago of 30–40 g C m^–2 (Codispoti et al., 2013). The annual NCP_C estimates for Kaldfjorden (5 ± 2 g C m^–2 yr^–1) are modest in comparison to the carbon-based NCP estimates of 34 g C m^–2 yr^–1 in Adventfjorden in Svalbard (Ericson et al., 2019b) and 108 g C m^–2 yr^–1 in the Fram Strait–Svalbard region (Vaquer-Sunyer et al., 2013). These variations show enhanced seasonal and annual NCP in the higher latitude Arctic regions, likely due to higher productivity in sea-ice-influenced areas, enhanced oceanic exchanges for nutrient resupply, and greater export of organic matter from the productive surface layer to subsurface waters.

Accounting for freshwater effects, mixing and NCP, the remaining seasonal variations in surface water C_T can be attributed to air–sea CO₂ exchange and residual changes that integrate variations in A_T, such as calcium carbonate formation/dissolution. To investigate this scenario, potential alkalinity (A_T*) was evaluated to show that the water column average was 2285 ± 12 μmol kg^–1 with largest variations across the fjord from July to October (Figure 14b). Decreases in surface water A_T* of ~50 μmol kg^–1 are indicative of CaCO₃ formation during biogenic calcification, which was most pronounced from July to September. This period encompassed the summer phytoplankton bloom, when bloom-forming coccolithophores disturb the optical properties of surface waters and can be detected by signals in light reflectance by remote sensing (Tyrrell et al., 1999). Coccolithophores are widely abundant in the global oceans and influence seawater carbonate chemistry through the synthesis of CaCO₃ shells. The CaCO₃, or particulate inorganic carbon, that is exported out of the surface can either dissolve upon transport to deeper water layers or become buried in sediments (Rost and Riebesell, 2004). MODIS-Aqua satellite-corrected reflectance available from NASA Worldview shows development of a coccolithophore bloom in Kaldfjorden during this study (Figure 15). Images taken at the time of each sampling event (8 June, 25 July, 6 September), or those closest in time that are not obscured by cloud cover, and the 1^st of each month show the bloom extended across Kaldfjorden by 26 July until 26 August. No reflectance in clear water was seen by the next sampling event on 6 September.

These observations support the proposed mechanism of removal of A_T by calcification (and concomitant reduction in monthly ΔC_{T CaCO3}), which would equate to a drawdown in C_T of ~25 μmol kg^–1. The ΔC_{T bio C} accounts for about 20% of the ΔC_{T total} and ΔC_{T alk} made a minor contribution (up to 8%) to ΔC_{T total}, which shows the importance of calcifying phytoplankton in carbon cycling in the region. In the following autumn months, increases in A_T* up to 40 μmol kg^–1 (positive ΔC_{T alk}) indicated A_T inputs from terrestrial weathering and perhaps from CaCO₃ dissolution in a decaying coccolithophore bloom, likely from sampling events that captured particulate inorganic carbon (CaCO₃ shells) in the water column.

The water column across Kaldfjorden remained saturated with respect to aragonite (and calcite, not shown) during the study period. The annual Ω range in Kaldfjorden compares very well (1.3–2.5) to that reported for fjords of western Norway (Omar et al., 2016). Seasonality in surface water Ω from minima in winter/spring to maximum in late summer was driven by changes in C_T from the production (photosynthesis) and decay (respiration/remineralisation) of organic matter. Over the growing season, biological production reduced C_T by 100 μmol kg^–1 in surface waters, leading to concomitant increases in Ω to 2.26–2.33 from March to September. Increases in ΔΩ _{bio C} (and ΔΩ _{bio N}) of 0.25–0.40 between April to June dominated the monthly changes Ω to show the importance of biological carbon uptake on the surface water acidification state (Figure 12). Removal of A_T and subsequent lowering of Ω during calcification constituted a minor competing effect (ΔΩ _CaCO3 of –0.09) that slightly counteracted the biologically-driven increases in Ω from C_T drawdown in summer coccolithophore blooms. The effect of freshwater dilution reducing C_T largely offset the parallel effect of A_T reduction, resulting in increases in ΔΩ _sal up to 0.11 across the fjord during the period of increasing freshwater fractions (April to June). Photosynthetically driven increases in Ω that counteract effects of freshwater dilution have been reported in other high latitude coast and fjord regions (Chierici and Fransson, 2009; Chierici et al., 2011; Fransson et al., 2015, 2016; Ericson et al., 2019a). The decrease in Ω to ~1.9 from May to June 2018 coincided with an episodic drop in salinity during a period of precipitation and peak freshwater fractions. Concomitant decreases in ΔΩ _{bio C} and ΔΩ _flux suggested that suppression of Ω had occurred through increased C_T from the degradation of organic matter and atmospheric CO₂ uptake in the freshwater flowing into the fjord, as previously reported (Anderson et al., 2009; Shadwick et al., 2011; Evans et al., 2014; Meire et al., 2015; Ericson et al., 2018). In addition, changes in water temperature affect Ω as colder waters have lower carbonate saturation through thermodynamic controls; increases in Ω by 1% arise due to warming of 1°C (Mucci, 1983). Coldest surface waters (2.17°C) with Ω of 1.59 in March warmed by 10.1°C to seasonal maximum (12.27°C) with Ω of 2.10 in July; thus, a thermodynamic increase in Ω of 0.16 could be expected and contribute to positive ΔΩ _total during this time. Saturation states steadily decreased during the autumn, winter and early spring in post-bloom conditions driven by organic matter remineralisation and net respiration (ΔΩ _{bio C} of –0.50) with effects of seasonal cooling. Impacts of mixing with subsurface carbon-rich Norwegian Coastal Water and air–sea CO₂ exchange played minor roles in monthly changes in Ω in surface waters.

Throughout each season, surface waters remained undersaturated with respect to the atmosphere and showed that Kaldfjorden was a sink for atmospheric CO₂ on an annual basis. Previous studies have also reported net atmospheric CO₂ sinks for sub-Arctic coast and fjord systems (Omar et al., 2016; Tynan et al., 2016; Yasunaka et al., 2016). The mean atmospheric CO₂ uptake of 0.86 ± 0.63 mmol m^–2 day^–1 is similar to that of 0.73 ± 0.40 mmol m^–2 day^–1 estimated for a marginal Arctic coastal environment of Hudson Bay, Canada (Else et al., 2008b). Wind speeds in Kaldfjorden were dampened due to orographic steering, compared to measurements off the shelf (Nordby et al., 1999); thus, atmospheric CO₂ uptake in Kaldfjorden of 2.7 mmol m^–2 day^–1 is weak compared with the CO₂ influx of >15 mmol m^–2 day^–1 for the Norwegian Sea (Yasunaka et al., 2016) and the non-ice-covered Arctic shelf seas that generally absorb CO₂ at between 1 and 15 mmol m^–2 day^–1 (Omar et al., 2005; Cai et al., 2006; Else et al., 2008a, 2013; Ericson et al., 2018; Chierici et al., 2019). Enhanced CO₂ uptake in higher latitude waters results from substantial blooms, greater biological productivity and subsequent export of organic matter to deeper waters, coupled to strong winds enhancing oceanic CO₂ uptake. These results highlight the large spatial variability of the oceanic CO₂ sink and emphasize the need to resolve key processes on regional scales.

Strong seasonality in hydrography, carbonate chemistry, and macronutrients in the marine environment of Kaldfjorden, northern Norway, was driven by freshwater inputs, biological production, and mixing with subsurface coastal water. High latitude fjord and coastal environments may be particularly sensitive to future changes in ocean chemistry and the effects of ocean acidification, compared to the open ocean, due to the greater influence and spatial-temporal variations of freshwater sources and terrestrial influences. Fjords are dynamic ecosystems that naturally experience large ranges in carbonate chemistry and thus may exhibit a degree of resilience to future changes but may also be vulnerable to extreme values. Freshwater inputs play a key role in monthly changes in the acidification state (Ω) of surface waters through dilution effects on C_T and A_T. Increases in air and seawater temperatures and greater freshwater fluxes will have consequences for the seasonal stratification and mixing of the water column, as well as phytoplankton species composition, bloom development, and biogeochemical cycling. Freshwater entering the fjord changes the seawater chemistry so that A_T, carbonate ion concentrations and pH decrease due to dilution. In addition to continued uptake of anthropogenic CO₂, these processes reduce the buffering capacity of the fjord water and increase the vulnerability of surface waters to acidification. Increased acidification is expected to have adverse effects on marine life, such as the pelagic calcifiers, coccolithophores, which contribute to monthly C_T deficits during summer blooms and form an integral part of the food web (Andersson et al., 2015). Terrestrial inputs of organic matter and weathered minerals, such as carbonates and silicates from the surrounding bedrock (Fransson et al., 2015), enhance C_T and provide a minor source of A_T to surface waters. Effects of dilution from freshwater were strongly counteracted by primary production, with intense C_T and nitrate drawdown from the period of the spring bloom to early autumn dominating the Ω seasonality. Estimates of NCP were modest but in accordance with other high latitude fjord and coastal regions: Kaldfjorden is an annual sink for atmospheric CO₂. High latitude fjords have been regarded as regions of high organic carbon sequestration, provided strong stratification does not prevail during the productive season, and if elevated fluxes of organic matter from productive surface waters and surrounding terrestrial sources can be exported and stored in fjord sediments (Smith et al., 2015). Changes in pH may also influence other chemical processes such as bioavailability of metals and toxins (Millero et al., 2009; Breitbarth et al., 2010).

The time-series data presented here emphasise the need for year-round sampling to better understand the natural variability in the marine environment. Addressing some of the remaining uncertainties could include resolving the characteristics (C_T, A_T, nutrients, δ¹⁸O, salinity, organic matter) of the oceanic (coastal) and freshwater endmembers to elucidate the main sources of inorganic and organic carbon, nutrients and minerals to Kaldfjorden. Phytoplankton community composition and net primary production, as well as drawdown of carbon and uptake of atmospheric CO₂, will ultimately depend on how the fjord marine ecosystems will respond to climatically induced changes. Observations of the biogeochemical dynamics on seasonal timescales further our understanding of carbon and nutrient cycling in these important marine systems and serve as benchmarks against which future changes can be compared and evaluated.

The data collected during this study are publicly available at the Norwegian Marine Data Centre with the following reference Angelika Renner (2020) Hydrography in Kaldfjorden, Troms, Norway..

The authors gratefully acknowledge the Captains and crews of RV Helmer Hanssen, RV Johan Hjort, RV Johan Ruud and RV Kronprins Haakon for the logistical support and opportunity to use the vessels to conduct fieldwork. Particular thanks to Z Walker (UW/UiT), K Dunlop (IMR), P Renaud (Akvaplan-niva), C Ballantine (Akvaplan-niva) and KØ Gjelland (NINA) for boat logistics and use of Dytiscus and numerous others for assistance with boating and sample collection. We would like to thank the Editors and two reviewers for valuable comments that have helped to greatly improve the manuscript.

This work was part of the project “Impact of massive Winter Herring Abundances on the KaLdfjorden Environment (WHALE; grant number 201914747042018), which was funded by the flagship “Effects of climate change on sea and coastal ecology in the north” and the flagship “Ocean acidification and effects in northern waters” of the FRAM – High North Research Centre for Climate and the Environment. I Wiedmann was funded by ARCEx, the Research Centre for Arctic Petroleum Exploration (Norwegian Research Council #228107 and industry partners).

The authors have no competing interests to declare.

• Contributed to conception and design: AR, IW, MC, MB

• Contributed to acquisition of data: AR, EJ, MB, HHL

• Contributed to analysis and interpretation of data: EJ, HHL, AR, IW, MC

• Drafted and/or revised the article: EJ, AR, IW, MC

• Approved the submitted version for publication: EJ, AR, IW, MC, HHL, MB