Seasonal Changes in Carbonate Saturation State and Air-Sea CO₂ Fluxes During an Annual Cycle in a Stratified-Temperate Fjord (Reloncaví Fjord, Chilean Patagonia)

Abstract

Changes may be occurring in the carbonate chemistry of fjords due to natural and anthropogenic disturbance of major freshwater sources. We present a high-frequency time series study of seasonal pH and CO₂ partial pressure (pCO₂) in a north Patagonian fjord with a focus on changes in freshwater inflows and biological processes. To do this, we monitored pH and pCO₂ in situ, along with river streamflow, salinity, temperature, and dissolved oxygen (DO) in the Reloncaví Fjord (41.5°S) for a full year (January to December 2015). Strong seasonal variability was observed in the pCO₂, pH, and DO of the fjord's surface waters. During the summer, pCO₂ reached its annual minimum (range: 187–571 μatm) and pH its maximum (range: 7.98–8.24), coinciding with lower freshwater inflows (204–307 m³/s) and high DO (280–378 μmol/kg), as well as aragonite saturation states (Ω_Arag) higher than 1. In contrast, in winter, pCO₂ ranged from 461–1,008 μatm and pH from 7.57–8.03, coinciding with high freshwater inflows (1,049–1,402 m³/s), lower oxygen (216–348 μmol/kg), and constant undersaturation of Ω_Arag. Reloncaví Fjord had an annual air-water CO₂ flux of 0.716 ± 2.54 mol·m⁻²·year⁻¹ during 2015 and thus acted as a low emission system. The annual cycle was mainly governed by seasonal changes in biological processes that enhanced the shift from a CO₂ sink in late spring and summer, caused by high primary production rates, to a CO₂ source during the rest of the year caused by high community respiration due to allochthonous organic carbon inputs.

Key Points

• North Patagonian fjords are aquaculture-heavy used environments, and there are not available data for understanding major biogeochemical process like air-sea CO₂ fluxes and aragonite saturation
• A high-frequency time series of the annual cycle of CO₂, pH, and dissolved oxygen in a north Patagonian fjord is evaluated
• The fjord inorganic carbon cycle is very dynamic, primarily driven by varying contributions from primary production and respiration of allochthonous organic carbon

1 Introduction

Fjords are areas of special importance for research due to the magnitude of the biogeochemical processes that are present there, for example, they are important areas of organic carbon burial (Hinojosa et al., 2014; Skei, 1983; Smith et al., 2015), and their role in ecosystem services such as aquaculture (Brattegard, 1980). The characterization of the CO₂ system variability is particularly important in Patagonian fjords because many of them are important nursery areas for marine calcifiers, including species utilized for intensive aquaculture (e.g., mussels). Early life stages of calcifiers invertebrates may be particularly vulnerable to perturbation in the carbonate system induced by high levels of CO₂ (Duarte, 2015; Ellis et al., 2016; Navarro et al., 2013, 2016; Wang et al., 2015). The CO₂ system of coastal waters can respond to a myriad of factors and processes (natural or anthropogenically driven) in a wide range of spatiotemporal scales. However, due latitudinal distribution of fjord systems, seasonal biogeochemical cycle is a dominant source of variability but modulated by long-term fluctuation including those associated to the global change (e.g., ocean acidification, perturbation in the hydrological cycles, and global warming).

Highly seasonal primary production (PP) of Patagonian fjords has been attributed to a combination of physical (e.g., stratification and solar irradiation) and chemical (e.g., nutrient input) processes (Iriarte et al., 2007; Jacob et al., 2014). PP is dominated by highly silicified, chain-forming diatoms (Iriarte & González, 2008) supported by nutrients and resources of marine and continental origin; the latter is particularly true in fjords with direct rich freshwater inflows like Reloncaví Fjord (Silva, 2008; Torres et al., 2014; Vandekerkhove et al., 2015). Patagonia archipelago inner waters have been suggested to be net sinks of atmospheric CO₂ during the warm season (Torres et al., 2011); a small pCO₂ data during the cold season suggest that this tendency could be reduced or inverse during winter time (Torres et al., 2011); however, the low frequency of those measurements preclude any annual balance. These authors argue that the haline-stratified fjords of Patagonia can have a particularly strong capacity to remove atmospheric CO₂ since haline stratification not only enhance light and continental nutrients (e.g., silicic acid and iron) availability for surface water phytoplankton but also enhance its efficiency (in term of nutrients) to drop pCO₂, due the low alkalinity (low buffer capacity) of surface waters. In other hand, they pointed out that while a strong halocline precludes the vertical fluxes of respiration products (CO₂) from below the halocline, it cannot prevent the rain of particulate organic matter from surface to subhalocline waters, therefore intensifying the effects of the vertical segregation between productivity (low surface water pCO₂) and respiration (high subsurface water pCO₂).

Surface water pCO₂ data display considerable variability in high-latitude coastal regions, with relatively high surface pCO₂ occurring in riverine-influenced coastal areas, probably associated to organic matter discharges (net metabolism) to this coastal regions; however, areas of sea ice melt coincided with low-surface pCO₂ (Atamanchuk et al., 2015; Burgers et al., 2017; Mørk et al., 2014). Fjords in the Northern Hemisphere have been more intensively studied (Andersson et al., 2017; Atamanchuk et al., 2015; Borges et al., 2004; Feely et al., 2010; Meire et al., 2015; Murray et al., 2015; Omar et al., 2016; Reisdorph & Mathis, 2014; Reum et al., 2014; Ruiz-Halpern et al., 2010; Rysgaard et al., 2012). In most of these studies, seasonal changes due to PP, ice formation, ice melting, and freshwater inputs have been pointed out as the main drivers for changes in air-sea CO₂ fluxes and the carbonate system.

Patagonian fjords occurred from 41°S to 56°S have been an interesting location for studying the carbonate system; however, just a few studies have focused on the CO₂ cycle within this vast region (Alarcón et al., 2015; Torres et al., 2011).

The Reloncaví Fjord is the northern most fjord of Patagonia and has many freshwater sources like the Puelo River (León-Muñoz et al., 2013) that has been reported having a high interannual variability, driven by seasonal rainfall and snowmelt regimes. In the fjord, the magnitude of the PP during seasons when blooms occur has been associated with interannual changes in streamflow patterns (Iriarte et al., 2016). The productivity of Reloncaví Fjord has an strong seasonal variability, with relatively high PP in spring–summer (~170 mmol C·m⁻²·day⁻¹) and very low in winter (<8 mmol C·m⁻²·day⁻¹; González et al., 2010) likely playing a major role in driving the seasonal carbonate system in this fjord.

Furthermore, as the northernmost fjord in South America, Reloncaví does not experience some other physical processes like ice formation or melting from glaciers at the head of the fjord (Iriarte et al., 2014). This makes it a unique and interesting system offering a clear view of the impacts of freshwater changes on the surface chemistry and biology of the fjord system. Importantly, recent climatic predictions for the Patagonian fjord ecosystem lead us to expect decreasing freshwater supply to the fjord's surface water as a result of decreased atmospheric precipitation in the coming decades (Boisier et al., 2016; Garreaud et al., 2013).

The aim of this study was to describe the annual fluctuations of surface water inorganic carbon chemistry and air-sea CO₂ fluxes in the fjord, through the deployment of high performance sensors and field sampling campaigns. We discuss the role of the biological and physical processes driving the inorganic carbon chemistry and CO₂ fluxes and their dependence on the local hydrological cycle.

2 Materials and Methods

2.1 Study Area

Reloncaví Fjord (Figure 1), located between 41°22′S and 41°44′S, is approximately 55 km long by 3 km wide and is divided into three basins. The maximum depths are 450 m at the mouth and 50 m at the head. The fjord is dominated by fresh estuarine surface waters and modified Sub-Antarctic Water with salinities ranging from 31–33 in the subsurface (Castillo et al., 2016; Silva et al., 2009). It presents a relatively shallow (<5 m), continuously stratified buoyant layer and has a tidal regime that is mainly semidiurnal, with a tidal range between 6 and 7 m during spring tides decreasing to ~1 m at neap tides (Valle-Levinson et al., 2007).

Figure 1

(A) South America. (B) Location of the fjord area in the South American Continent. (C) Reloncaví Fjord area map. The buoy was anchored adjacent to the river mouth (North Patagonia Buoy: http://portal.goa-on.org/Explorer). The legend indicates the position of the instruments used in the study. The samples were taken at exactly the same place as the buoy.

The fjord has a three-layer vertical circulation pattern in which the surface (<5 m) and deepest (>100 m) layers tend to move toward the mouth (outflow layers), whereas the intermediate inflow layer (>5 and 100 m) moves toward the head of the fjord (Castillo et al., 2016; Valle-Levinson et al., 2007). There is exchange of dissolved inorganic nutrients between the more saline (29–32), nutrient-rich (N and P) subsurface layer, and the low-salinity (2–20), low-nutrient (except for silicic acid) surface layer. The surface layer rarely extends below 5 m, but there is a marked seasonal variability in its inorganic nutrient load between spring and winter (Castillo et al., 2016; González et al., 2010). Specifically, the upper 5 m of the surface layer is low in NO₃⁻ (<5 μM) and PO₄³⁻ (<0.7 μM) in summer through autumn but high in silicic acid (>100 μM); while in winter, silicic acid, NO₃⁻ and PO₄³⁻ concentrations are all high (30–138, 10–22, and >1 μM, respectively; González et al., 2010).

The sampling site was located close to the mouth of the Puelo River (Figure 1; 41°35′S, 72°20′W). The Puelo River is the primary source of riverine freshwater inflow into the Pacific Ocean in northern Patagonia (Dávila & Figueroa, 2002; León-Muñoz et al., 2013) with a mean discharge of 600 m³/s. Therefore, this station offered a representative view of the total inflow contributed by the river basin.

2.2 Field Sampling and Measurements

Samples were taken from the upper part of the surface layer (1-, 5-, 10-, and 15-m depths) from January to December 2015 during eight field campaigns. Water samples for total alkalinity (A_T) and pH determination were collected with a 10-L Go-Flo bottle and stored in the dark at low temperature (<7 °C) in a 250-ml gas-tight container. Seawater samples for A_T analysis were poisoned with HgCl₂; pH was analyzed after collection (<12 hr) using impure m-cresol purple as an indicator at 25.0 °C with an OceanOptics STS-Vis (350–800 nm) spectrophotometer (Byrne et al., 1988). When possible, potentiometric pH was determined at the same time as a comparison, calibrated by using standard pH tris buffer (pH = 8.089 at 25.0 °C; Dickson et al., 2007; Riebesell et al., 2010; DOE, 1994). All pH samples were measured in duplicate giving a reproducibility better than 0.001 pH units when the spectrophotometric methods was used. Total alkalinity (A_T) was determined at the laboratory using an automatic potentiometric titration system (Haraldsson et al., 1997). We used certified reference material supplied by Andrew Dickson (Scripps Institution of Oceanography) to verify A_T accuracy or correct A_T values; reproducibility was typically less than 2 μmol/kg.

For inorganic dissolved nutrients, 500 ml of seawater were collected from the upper layer (surface to 15 m) during the synoptic campaigns. The water was filtered through glass fiber filters (Whatman GF/F) and stored frozen (−20 °C) until analysis (Parsons et al., 1984). The analysis of nutrients was done in a certified laboratory (CERAM of the Universidad Austral de Chile at Puerto Montt). For the autotrophic biomass (chlorophyll-a), 250 ml of seawater were filtered through a 0.7-μm glass fiber filter (Micro Filtration System), extracted with 90% acetone, and measured with a fluorometer (Turner P700) as recommended by Parsons et al. (1984).

The discrete A_T, pH, inorganic dissolved nutrients, and chlorophyll-a measurements are available in the supporting information Table S1.

2.3 Sensor-Based Measurements

High-resolution pCO₂, pH, depth, temperature, conductivity, and dissolved O₂ (DO) measurements were recorded simultaneously in situ using autonomous Submersible Autonomous Moored instrument (SAMI)-CO₂, SAMI-pH sensors (DeGrandpre et al., 1995; Seidel et al., 2008, Sunburst Sensors, LLC), and SBE 37 MicroCA_T CTD-ODO (SeaBird Electronics), respectively. Measurements started in the austral summer (January 2015). All the sensors were placed at exactly the same depth mounted in one single steel frame with sensors water intakes at same vertical position. The absolute measurement depth of the pressure sensor varied between 1.0 and 3.5 m below the surface during the entire period due to the extreme Puelo River flow and the high tidal variation. DO values were used to compute the apparent oxygen utilization (Emerson & Hedges, 2008). All sensors recorded data hourly throughout 2015, with the exception of January and February when data were recorded every 20 min. Gaps of 2 or 3 weeks in the time series were due to routine maintenance and calibrations. Sensors were cleaned at 2, 7, and 10 month after deployment to prevent biofouling formation, and the intake tubings were cleaned with deionized water. At month 10 postdeployment we run an extended sensor maintenance and calibration using sensor's manufactures recommended protocols. The SAMI-CO₂ sensor was calibrated using standard CO₂ concentration gas tanks up to 800 ppm of CO₂. Values above the calibration range are likely to have an error (up to 5% at 1,500 ppm) due to nonlinearity and insensitivity of the response at these high pCO₂ levels (DeGrandpre et al., 1999). SAMI-pH instruments use an accuracy test instead of a calibration procedure (Seidel et al., 2008) with a tris pH buffer sample using purified m-Cresol Purple at 25 °C, giving an accuracy of ~0.005 pH units (Sunburst Sensors, LLC). However, there could be inaccuracy in the indicator equilibrium constant over the broad pH and salinity range found in the fjord. We assume that the pH values obtained at lower salinities are subject to an error of up to ±0.02 pH units at S < 5, and pH values above 7 (Mosley et al., 2004).

The salinity and A_T values from the bottle sampling presented a linear relation (r² = 0.976; supporting information Figure S1), and this regression was used to predict surface A_T based on salinity (A_Tsal). The SAMI-pH data were compared with discrete bottle data in order to evaluate in situ sensor performance. SAMI-pH and A_Tsal data were used to compute the other carbonate parameters at the ambient temperature and salinity, because that pair of carbonate system parameters has shown good accuracy (Cullison-Gray et al., 2011). The pCO₂ computed from the SAMI-pH and A_Tsal was compared with the in situ pCO₂ data. All inorganic carbon parameters were calculated with CO2SYS_v2.1-2018 software (Orr et al., 2015; Pierrot et al., 2006) modified using the K₁ and K₂ constants in Table 5 of Dickson and Millero (1987) for salinity = 0–40. Nutrient data were not included in the computations, due to the lack of continuous measurements during the study. The calculated pCO₂ changes by <10 μatm when the highest observed levels of total phosphate (2 μM) and silicate (100 μM) are included in CO2SYS.

Previous studies have shown that SAMI-pH and SAMI-CO₂ sensors are stable over a time period of several months, showing no significant drift during a deployment period (Omar et al., 2016), although the pCO₂ sensors require in situ validation during deployment (Cullison-Gray et al., 2011; DeGrandpre et al., 1995). The differences between discrete and sample pCO₂ and pH, along with the comparison of in situ and calculated values, are shown and discussed below. In order to address the differences obtained between the estimated values (linear regression for A_Tsal) and computed values from CO2SYS using measured pH and pCO₂, we plot the delta A_T correlated to salinity (Figure S4). As is known that enclosed coastal areas with high influence of major freshwater sources, like Reloncaví Fjord and the Puelo River, the dissolved organic matter could act as organic bases enhancing A_T (Cai et al., 1998; Hernández-Ayon et al., 2007; Yang et al., 2015); there is no empirical evidence to suggest that the Reloncaví Fjord could have allochthonous organic alkalinity addition. However S-A_T relationship have been found to be highly correlated in Patagonian waters (Alarcón et al., 2015; Torres et al., 2011), suggesting that most of A_T variability is driven by mixing in this estuarine systems.

2.4 Meteorological Data

A meteorological station (HOBO-U30; 41°41′S; 71°23′W) close to the Puelo River mouth measured air temperature, solar radiation, wind speed and direction, rain, and barometric pressure every 5 min. This information was synchronized with the buoy data in order to make air-water CO₂ flux estimates. For the atmospheric pCO₂, values of atmospheric measurements from the Earth System Research Laboratory database (National Oceanic and Atmospheric Administration Marine Boundary Layer Reference 53.1°S to 17.5°S; www.esrl.noaa.gov/gmd/ccgg/mbl/data.php) were interpolated for the year 2015 and corrected with local barometric pressure. These values assume clean marine air, and therefore, a possible error is created due to terrestrial effects on pCO₂. The error is not readily quantified because no regional pCO₂ data are available.

Streamflow information for the Puelo River was obtained from the Carrera Basilio hydrological station, the station closest to the mouth of the river (41.6°S; 72.2°W), run by Dirección General de Aguas de Chile (http://snia.dga.cl/BNAConsultas/reportes; Figure 1). The data consisted of hourly streamflows for the 2015 hydrological year (January to December). Because the Puelo is one of Patagonia's major rivers, this station offered a representative view of the variability in the freshwater contributions from the Reloncaví Fjord basin.

2.5 Flux Determinations

Air-sea flux was calculated using the diffusive boundary layer model, from the bulk flux equation expressed in terms of CO₂ partial pressure (equation 1):

(1)

where F is the air-water flux (moles per area per time) and k is the gas transfer velocity (length per time) that accounts for gas diffusion at the air-sea boundary and was estimated using a wind speed relationship adjusted for in situ conditions using the updated equation in Wanninkhof (2014):

(2)

where <U²> is average squared wind speed adjusted to 10-m height above sea level and Sc is the Schmidt number, which accounts for differences in molecular diffusivity between gases. A positive F value represents a flux from the ocean to the atmosphere. K₀ is the solubility of the gas expressed in units of concentration/partial pressure, and pCO_2w and pCO_2a are the partial pressures of CO₂ in the surface water and in equilibrium with the overlying air, respectively, as described in Wanninkhof (2014). Because different months of the year have different total amounts of data points, the monthly average flux was determined for each month and these values were used to calculate the annual mean flux. The values for atmospheric pCO₂ obtained from interpolation (section 2.4) were used as input data. Atmospheric pCO₂ in wet air was then calculated by including water vapor pressure according to the following formula (Dickson et al., 2007):

(3)

where P_w is the water vapor of seawater at in situ salinity and temperature of equilibration (Forstner & Gnaiger, 1983).

2.6 Temperature Effect on pCO₂

Because seawater pCO₂ is strongly affected by changes in temperature due to the temperature dependence of CO₂ equilibria and solubility, we used the equation of Takahashi et al. (2009); equation 4) to determine the effect of seasonal cooling/warming on CO₂ variability in the fjord.

(4)

where the annual mean temperature (T_mean) was 13.33 °C, the annual mean pCO₂ (pCO_2mean) was 643 μatm, and T_obs is the measured temperature in degrees Celsius. This equation uses the temperature coefficient for pCO₂ to determine to what extent the pCO₂ will vary from a fixed pCO₂ (in this case the mean of the entire time series) over the observed range of temperatures.

3 Results

3.1 Hydrographic Conditions

The Puelo River had highly contrasting flows: streamflow was at its lowest (monthly means: 204–307 m³/s) in summer–autumn (January–April) and highest (monthly means: 713–1,402 m³/s) in winter (May–August). During the year, the river presented several massive freshwater pulses, reaching up to >4,000 m³/s in early June (Figure 2d). Accordingly, salinity fluctuated significantly at the buoy mooring point due to the effect of freshwater and the mooring line changing depth from tides and current (Figure 2e). In summer the variation of salinity was due to a signal in the tidal cycle, while in winter the precipitous decline in salinity corresponded with the massive freshwater pulses from the Puelo River (Figure 2d). During summer, salinity values fluctuated dramatically with the tidal cycle, likely because sensors were sampling through the stratified layer during different tidal states due to the proximity of the mooring to the river mouth. During the annual cycle the atmospheric temperature varied from −0.8 to 29 °C, while the sea surface temperature was more stable and varied from 8.7 up to 19.4 °C (Figure 2f). Seawater temperature changes occurred gradually, with the highest monthly average sea surface temperature in February and the lowest in August (Table 1).

Figure 2

Annual dynamics of the primary measured parameters at the Reloncaví buoy (Figure 1). Time series of pH (a), pCO₂ (b), dissolved oxygen (c), Puelo River streamflow (d), salinity (e), and air (black) and sea surface (blue) temperatures (f). A shows data sampled at different discrete depths; the horizontal red line in (b) shows the pCO₂ atmospheric value. Depth variation from (e) is explain above.

Table 1. Monthly Mean Statistics of the Carbonate System (pH and pCO₂), Salinity, Temperature, Dissolved Oxygen (DO), and CO₂ Flux in 2015

Month	pH	Salinity	pCO2 (μatm)	Temp (°C)	DO (μmol/kg)	Flux CO2 (mmol·m^−2·hr^−1)	Wind (m/s)
Jan	8.24	20.41	187	16.3	378	−0.422	4.17
Feb	8.19	21.65	243	16.9	335	−0.296	3.66
Mar	7.98	28.42	571	14.4	280	0.104	2.41
Apr	7.81	29.85	912	12.5	229	0.259	1.66
May	7.63	29.35	1,308	11.7	176	0.446	1.86
Jun	7.57	23.37	1,008	11.2	216	0.521	2.44
Jul	7.72	27.32	970	10.6	223	0.358	1.77
Aug	7.75	27.66	805	10.6	236	0.171	1.80
Sep	8.03	25.98	461	10.9	348	0.069	2.25
Oct	8.09	28.05	471	11.9	339	−0.045	2.47
Nov	8.10	25.03	330	13.1	317	−0.094	2.73
Dec	8.16	20.15	297	14.7	348	−0.185	3.57
Max	8.24	29.85	1,308	16.9	378	0.521	4.17
Min	7.57	20.15	187	10.6	176	−0.422	1.66
Mean	7.94	25.60	630	12.9	285	0.074	2.57
SD	0.23	3.45	361	2.2	67	0.295	0.83
Median	8.01	26.65	521	12.2	299	0.086	2.44

• Note. Annual values are calculated from monthly mean data.

3.2 Data Quality and Carbonate System Consistency

The full-year time series obtained by the pH sensors showed a good match with the field sample data (Figure 2a), with a mean difference of 0.05 pH units (SD ± 0.059, salinity range 23.9–32.2, n = 7) using the 5-m depth as contrast. As stated above, the relationship between total alkalinity and salinity was determined using discrete samples collected throughout the period for the surface layer (top 15 m). The resulting linear model for the observations (A_Tsal = 53.107 × salinity + 385.97 in micromoles per kilogram; R² = 0.976, n = 46, Table S2 and Figure S1) is an ideal tool for assessing the total alkalinity based only on salinity values. The intercept (386 ± 34 μmol/kg) is a reasonably good match with the Puelo River end member, which had an average A_T of ~340 μmol/kg (n = 3, salinity of 0.3 PSU) in the spring.

During the annual cycle, pH and pCO₂ were correlated in a negative exponential relationship (Figure 3), with the most corrosive (lowest pH) conditions occurring in winter (Figure 3b). Data collected in higher salinity water seem to present a better fit to the exponential model than the data collected in low salinity water (Figure 3a); points that fall well below the fit are those found during winter (Figure 3b), specifically during the months when there were massive freshwater flood pulses from the Puelo River, with correspondingly low salinities (Figure 3a). The steep relationship between pH and pCO₂ for the low salinity (<7) data can be reproduced in CO2SYS using low constant A_T (e.g., 500 μmol/kg), whereas seawater conditions fall around the general exponential trend (Figure S3). Therefore, the large scatter shown in Figure 3 makes sense for the large range in salinity observed during the study. Note also that the vertical or horizontal deviations in pH or pCO₂, respectively, could be fouling of the SAMI-pH or SAMI-CO₂. Regarding pH, same steep trend is observed in December and January (Figure 3b) and follow the same trend that modeled data from Figure S3.

Figure 3

The relationship for pH and pCO₂ for the data from Figure 2 colored by salinity (a) and date (b).

The A_Tsal can be used with in situ pH data to provide an in situ data evaluation (Cullison-Gray et al., 2011). In this case, the in situ pCO₂ data were compared with pCO₂ computed from in situ pH and A_Tsal using CO2SYS with in situ temperature and salinity. Although there is a high linear correlation between the computed and measured values observed for pH (R² = 0.92; Figure S4a) and pCO₂ (R² = 0.91; Figure S4b), the slope of the pH model was closer to 1 (0.98) than the lower slope obtained for the pCO₂ model (0.78), showing lower calculated values. The mean difference between measured pCO₂ and computed pCO₂ was 76 μatm + 131 μatm (n = 10,227), while the mean difference between measured pH and computed pH (in situ pH − computed pH) was 0.051 ± 0.078 pH units.

The deviations obtained from the CO₂ system calculations appear large, but they form only a small (<4%) portion of the observed pH (~1.3 pH unit) and pCO₂ (~2,000 μatm) ranges. Deviations may be caused by several factors such as biofouling, deviation of A_Tsal from the linear ratio, organic sources of A_T, systematic errors at high pCO₂ due to nonlinear calibration, inaccuracy of the calculated pH and CO₂ due to uncertainty of the different pKa and equilibrium constants at low salinity, and errors due to rapid changes and extreme stratification around the individual sensors on the buoy mooring so that they measure different water due to slightly different measurement times (Cullison-Gray et al., 2011; Lai et al., 2016).

3.3 Year-Round System Dynamics

The carbonate system exhibited a large range of variability over the year-long time series (Figures 2a, 2b, and 3 and Table 1). The pH presented maximum values during midsummer (>8.0). January was the month with the highest average pH (pH = 8.24), which then started to decrease during the autumn and reached its minimum in early winter (June average pH = 7.58). The opposite is observed for pCO₂ where the lowest values were recorded in midsummer (January average pCO₂ = 200 μatm); pCO₂ then increased in autumn up to its highest point in May (May average pCO₂ = 1,312 μatm; Figure 2b and Table 1). This trend is followed by the seasonal changes in DO that occur during spring–summer, where the fjord system drives high DO values in the warmer months (>16 °C, >350 μmol/kg). During early fall and spring, DO had the highest seasonal correlation found for pH and pCO₂ (Table S4). Toward winter, the DO values gradually decreased in the surface layer (<200 μmol/kg; Figure 2c and Table 1); in fact, the DO peak and lowest point coincide with the extreme values for pH recorded at the end member of the Reloncaví Fjord system (Figures 2a and 2c, red arrows). The pH and DO minima occurred at the same time that the temperature was low and river discharge was high (Figures 2d–2f). An abrupt reduction of DO (from 300 to 180 μmol/kg; Figure 2c), an increase in pCO₂ (to 500–1,000 μatm; Figure 2b), and pH dropping consistently below 7.8 (Figure 2a) were associated with the lowest observed river streamflow (early May, Figure 2d, red arrow, and Table 1). Salinity and DO shows significant correlation values with pCO₂ during winter months (Table S4), while temperature does not show a clear patter, and at the same time changes in temperature (equation 2) explained less than 15% of the annual pCO₂ variability (not shown).

Aragonite and calcite saturation are important variables of the carbonate system within the Reloncaví Fjord, particularly because it is an area of intense aquaculture production—principally mussel farming (Molinet et al., 2017). The aragonite saturation state (Ω_Arag) was computed using the pH and A_Tsal in combination with the measured temperature and salinity. There is a clear seasonal trend (Figure 4) with a marked decrease in the saturation state from summer to autumn and then a period of consistent undersaturation during late autumn and winter. Puelo River A_T values at zero salinity are approximately 380 μmol/kg based on the linear regression of alkalinity data at S = 0, as stated above. The near-zero values of Ω_Arag in late autumn mark the system minimum; at the same time, the undersaturated values found in winter, even in high salinity waters (Figure 2e), are clearly driven by low pH and high CO₂ water (Figures 2a and 2b and Table S1). There was a considerable period of time when the system was undersaturated even at high salinities. Ω_Arag moved above the saturation line again in the spring to remain saturated most of the time and at high salinity.

Figure 4

Annual in situ aragonite saturation (Ω_Arag) computed from total alkalinity (A_Tsal) and in situ pH. The solid line shows an aragonite saturation threshold of 1, while the black arrows indicate the evolution of the system from saturated to undersaturated and vice versa.

3.4 Air-Sea CO₂ Fluxes

The air-sea CO₂ fluxes varied widely throughout the entire deployment period, with values ranging from −2.1 mmol·m⁻²·hr⁻¹ (net uptake) in summer to 9.3 mmol·m⁻²·hr⁻¹ in winter (Figure 5 and Table 1). CO₂ uptake started in the spring (October–December: monthly mean flux of −0.013 to −0.13 mmol·m⁻²·hr⁻¹) and became more intense during the summer (January–February: monthly mean flux of −0.44 and −0.29 mmol·m⁻²·hr⁻¹). Outgassing occurred during the rest of the year (March–September: monthly mean flux of 0.27 mmol·m⁻²·hr⁻¹; Figure 5 and Table 1). The data exhibit prevalent outgassing activity in the fjord during the year, with a marked shift in the flux behavior in early autumn, from sink to source, and again, in spring, suggesting that the study area behaves as a net source for atmospheric CO₂ (Figure 5). The central portion of Reloncaví Fjord had an annual flux of 0.71 mol·m⁻²·year⁻¹ ± 2.48 mol·m⁻²·year⁻¹.

Figure 5

Annual dynamics of the air-sea CO₂ flux in Reloncaví Fjord.

4 Discussion

In coastal areas, the carbonate system of surface waters is modulated mainly by heating and cooling (e.g., DeGrandpre et al., 2002), production and respiration (Burgers et al., 2017; Cox et al., 2015; Feely et al., 2010; Gattuso et al., 1998; Hales et al., 2005), freshwater runoff (Harris et al., 2013; Jacquet et al., 2017; Krauss et al., 2018), transport of allochthonous nutrients and organic matter (Bhatia et al., 2013; Silva et al., 2011; Tiwari et al., 2018), surface stratification, and advective processes such as upwelling (Beitzel Barriquand et al., 2015; Burgers et al., 2017). Some of these factors play important roles in fjord systems (Jolivet et al., 2015).

In Reloncaví Fjord, the increase to high dissolved oxygen values observed during spring were due to high PP, being consistent with the higher pH and lower pCO₂ values observed during this study (Table S4). High PP rates and chlorophyll-a (up to 3.8 g C·m⁻²·day⁻¹) have been reported for the spring–summer period in this region (González et al., 2010; Montero et al., 2011). The decrease in CO₂ is associated with the negative CO₂ fluxes observed in Reloncaví Fjord (Figure 5). If we take into account only the CO₂ flux in summer and winter, it appears that the two seasons almost cancel each other out. During the spring, pH increased and pCO₂ decreased, which based on the DO correlation was due to the onset of spring–summer phytoplankton blooms in the region (Figure 2). In the northern section of the Patagonian marine system, the classic spring–summer diatom blooms account for the greater part (80%) of the total annual PP (Iriarte et al., 2007). Diatoms assemblages are efficient at absorbing and exporting carbon and therefore capable of contributing to the seasonal decline of pCO₂ levels in the surface layers of Patagonian fjords. Recently, experimental approaches through microcosms experiments have shown high growth rates coincident with high pH values (~8.0) during an artificially triggered phytoplankton bloom in north Patagonian waters (Iriarte et al., 2013; Olsen et al., 2014). Following the low pCO₂ values in summer, pCO₂ start a marked and steady increase to levels above the atmospheric equilibrium in autumn (March–May), which persisted through late winter and early spring (June–August; Figure 2b).

On the other hand, the combination of lower pH and higher pCO₂, as well as the drop in DO observed during the winter months, could be attributed to a number of processes: (a) enhanced postbloom heterotrophic processes (Montero et al., 2011; Olsen et al., 2014), (b) the mixing of oxygen-poor deep water from below the halocline up to the surface layer, as has been suggested for Reloncaví Fjord during winter months (Castillo et al., 2016; González et al., 2010; León-Muñoz et al., 2013), and (c) allochthonous organic matter transported to the fjord (Rebolledo et al., 2015) by runoff from heavy rain events during winter months.

Evidence of short-term variations in pCO₂ and O₂ have revealed that strong winds can overcome stratification in semienclosed basins, mixing water types with different pCO₂ levels (Atamanchuk et al., 2015; Bates et al., 1998; Turk et al., 2013). In Reloncaví Fjord, we noted that periods of maximum winds have an important role in the air-sea pCO₂ exchange, especially in winter when the dominant northerlies (>5 m/s) affect the surface layer CO₂ flux dynamic (Castillo et al., 2016). During our study, the yearly mean wind speed was 2.6 m/s (from monthly mean values), with a maximum monthly average wind speed in January (4.29 m/s) and minimum in August (1.76 m/s). The highest wind speed was found to occur during a winter storm in early June (13.4 m/s).

Reloncaví Fjord, at the study area, was found to be a smaller CO₂ source than fjords in the Northern Hemisphere at higher latitudes, like Roskilde Fjord, North Zealand, Denmark, which has been reported to be an annual CO₂ source (3.9 mol CO₂·m⁻²·year⁻¹; Mørk et al., 2014), and Koljo Fjord, Sweden, which also showed a marked transition in pCO₂ dynamics from autumn to spring (Atamanchuk et al., 2015). In Friday Harbor in the glacially formed San Juan Archipelago (Washington, USA), pCO₂ ranged from ~300 to ~1,100 μatm over a 2-year period, mostly with high positive ΔpCO₂, suggesting that the sea is a source for atmospheric CO₂ in this area (Murray et al., 2015). On the other hand, Kongsfjorden and Tempelfjorden Arctic fjords (Svalbard, Norway) appeared to be less variable systems (in terms of the carbonate dynamics and saturation state of Ω_Arag) and they apparently act as CO₂ sink systems (Fransson et al., 2016, 2017). In the case of the Godthåbsfjord system (SW Greenland), the input of fresh glacial meltwater produces a strong CO₂ uptake giving a net negative CO₂ flux within this fjord system (Meire et al., 2015). Because of the high heterogeneity between these fjords-estuary ecosystems and their middle- to high-latitude occurrence, is hard to establish a direct CO₂ source-sink gradient, in more cases the autotrophic-heterotrophic nature of them and the proximity to densely populated areas could play a major role, particularly under the present scenario of high and increasing atmospheric CO₂ (Feely et al., 2010; Gattuso et al., 1998; Vieira Borges et al., 2004).

A year-long, high-temporal resolution record of parameters such as Ω_ara, A_Tsal, pH, pCO₂, and CO₂ fluxes provides a valuable tool for present and future assessment of ocean acidification in studies focused on processes involved in the natural variability of the carbonate system. It is therefore especially necessary to obtain Ω_Arag saturation values, given the importance of this mineral in the life cycles of calcifying organisms in natural habitats, as well as in intensive mussel aquaculture zones like Reloncaví Fjord. The dynamics and changing conditions of the carbonate system can affect different aspects of organisms, from their ecological behavior to their physiological functioning, especially in fisheries and aquaculture development (Broitman et al., 2017; Haigh et al., 2015; Vargas et al., 2017), the two most economically important industries occurring in Patagonian coastal waters.

The saturation state of seawater in terms of Aragonite (Ω_Arag) in Reloncaví Fjord was maximum in summer and minimum in winter consistently with previously for the adjacent Reloncavi sound (Alarcón et al., 2015). Ω_Arag can be highly variable in coastal areas (Harris et al., 2013; Xu et al., 2017), some regions at the open ocean (Hauri et al., 2015; Jiang et al., 2015), and inland seas like fjords and bays from different regions (Feely et al., 2010; Fransson et al., 2015, 2016; Jantzen et al., 2013; Kapsenberg et al., 2015), driving challenging conditions for calcifying species inhabiting those systems. Water undersatured in calcium carbonate in the form of aragonite (Ω_Arag < 1) can have deleterious effects growth and survival implications for organisms dwelling near the surface across a broad spectrum of taxa (Cornwall et al., 2013; Duarte et al., 2013; Gattuso et al., 2015; Turley & Gattuso, 2012). Experimental work in the context of ocean acidification, using the mussel Mytilus chilensis, suggest that under high pCO₂ conditions the scope for growth (SfG is the subtraction of energy spent in feces and metabolism from the energy consumed, Naylor et al., 1989), metabolism, and calcification of this mussel decreases (Duarte et al., 2014; Navarro et al., 2016, 2013). Therefore, changes in this inorganic carbon speciation can have effects on the aquaculture industry and coastal communities by reduced SfG. Additionally, the net rate of calcium deposition and total weight has been found to be negatively affected by high pCO₂ to low pH (Navarro et al., 2016). Particularly, the mussel industry in Chile depends on the harvesting of seeds from the natural environment, where the Reloncaví Fjord plays a very important role by providing most of the seeds (≈80% of all seeds collectors; Molinet et al., 2017), at the same time, as harvesting extends its period from February up to August (middle winter; Viviana Videla unpublished data, 2019), exist a broad time frame were critical production activities (e.g. mussel seed harvesting) have place under harsh acidic conditions (Ω_ara < 1), situation that has not been studied and may be of crucial interest for mussels farmers. Considering that it is projected to have extended periods of up to 6 months/year of Ω_ara undersaturation in surrounding seas by the end of the century (Hauri et al., 2015), we should expect to see an expansion in the winter Ω_ara undersaturated period, highlighting the urgency to count with stable and continuous environmental monitoring of carbonate system parameters in order to have the ability to properly study the scope from the different drivers and stressors (local and global). In the other hand, the most important aquaculture resource by value in Chile, the Atlantic salmon (Salmo salar L), the heavily farmed salmon in Chile, is also seriously impacted by high pCO₂ levels by decreased growth and deterioration of important physiological parameters such as reduction in plasma chloride (Fivelstad et al., 2015). These conditions should be of concern to the fjord region of northern Patagonia, considering the extent of mussel and salmon farms in this large marine system.

5 Conclusions

These time series data enabled us to make an initial estimate of the CO₂ flux over seasonal periods with significantly contrasting biological and hydrological processes. A low CO₂ efflux (0.65 mol·m⁻²·year⁻¹) was estimated, characterizing this fjord system as a net source. The gradual changes and marked trends of pCO₂, pH, Ω_Arag, and O₂ can mostly be accounted for by biological processes, especially PP in the spring–summer seasons and respiration in the autumn–winter seasons. Temperature was not an important driver within the carbonate system. Phytoplankton production was certainly a significant and major factor in the low pCO₂ observed in spring–summer, as evidenced by the high oxygen and pH values. During the study period, hydrological-meteorological phenomena were responsible for the rapid, abrupt changes in the fjord system, especially regarding the massive freshwater pulses from the Puelo River (up to 4,000 m³/s). Recent climatic predictions for the Patagonian fjord ecosystem lead us to expect decreasing freshwater supply to the fjord's surface water as a result of decreased atmospheric precipitation in the coming decades (Boisier et al., 2016; Garreaud et al., 2013). Furthermore, Reloncaví Fjord is under intensive pressure from an aquaculture industry that depends entirely on the ecosystem services that the fjord provides (natural mussel seeds collection site, protected and brackish coast that helps with the sea lice and gives an ideal smoltification environment for salmonids; Soto et al., 2019).

Acknowledgments

This research was funded by CONICYT-FONDECYT 1141065 (J. L. Iriarte) and is partially included in the framework of Research Program 1 of the IDEAL Center (Grant 15150003). Partial funding was provided by CONICYT-FONDECYT 1140385 (R. Torres), Ocean Certain EU-FP7 603773 (J. L. Iriarte), and DID-UACh. Special thanks to Manuel Díaz and Dr. Carlos Molinet for making the map and providing the meteorological data. Data presented are part of the PhD Thesis of M. V. J. at UACh. During this study, M. V. J. was receiving financial support from a CONICYT Scholarship (Beca Doctorado Nacional 2015 # 21150285). M. DeGrandpre received funding from the U.S. National Science Foundation (Grant OCE-1459255). L. A. Cuevas was supported by Millennium Nucleus Project MUSELS funded by MINECON NC1200286. The authors gratefully acknowledge the insightful comments and suggestions of two anonymous reviewers that helped to improve this manuscript. All sensor data (SBE 37 MicroCAT CTD-ODO, SAMI-pH, and SAMI-CO₂) used for this study are publicly available and can be accessed at https://figshare.com/articles/Puelo_Bouy/7754258. Finally, as this is M. V. J.'s first research paper, he would like to dedicate it to his wife Leslie and sons Camilo and Nacho and his marine biologist father, Toño, and Pepa, his mother, in gratitude for their support.