Aragonite saturation state dynamics in a coastal upwelling zone

Authors


Abstract

[1] Coastal upwelling zones may be at enhanced risk from ocean acidification as upwelling brings low aragonite saturation state (ΩAr) waters to the surface that are further suppressed by anthropogenic CO2. ΩAr was calculated with pH, pCO2, and salinity-derived alkalinity time series data from autonomous pH and pCO2 instruments moored on the Oregon shelf and shelf break during different seasons from 2007 to 2011. Surface ΩAr values ranged between 0.66 ± 0.04 and 3.9 ± 0.04 compared to an estimated pre-industrial range of 1.0 ± 0.1 to 4.7 ± 0.1. Upwelling of high-CO2 water and subsequent removal of CO2 by phytoplankton imparts a dynamic range to ΩAr from ~1.0 to ~4.0 between spring and autumn. Freshwater input also suppresses saturation states during the spring. Winter ΩAr is less variable than during other seasons and is controlled primarily by mixing of the water column.

1 Introduction

[2] The oceanic uptake of anthropogenic CO2 and the resultant ocean acidification have been well-documented [Haugan and Drange, 1996; Caldeira and Wickett, 2003; Feely et al., 2009; Midorikawa et al., 2012], with a decrease in seawater pH of ~0.13 since pre-industrial times [Dore et al., 2009] and an accompanying decrease in the saturation state of calcium carbonate minerals. For aragonite, the more soluble form of CaCO3, saturation state (ΩAr) is calculated using the relationship:

display math(1)

where K*spAr is the apparent solubility product constant for aragonite. When ΩAr < 1, waters are undersaturated and dissolution is thermodynamically favored. Although there are some exceptions [e.g., Byrne et al., 2009; Dupont et al., 2010; Pansch et al., 2012], most studies have determined that constant exposure to undersaturated waters can impair the health of many marine calcifiers [Shirayama and Thornton, 2005; Gazeau et al., 2007; Fabry et al., 2008; Barton et al., 2012; Comeau et al., 2012].

[3] Coastal upwelling zones, such as the one along the U.S. Pacific coast, naturally experience low ΩAr values as Ekman transport pumps subsurface water onto the shelf during periods of equatorward winds [Hales et al., 2005a; Feely et al., 2008; Cao et al., 2011; Fassbender et al., 2011]. These upwelling events also bring nutrients into the euphotic zone [Hales et al., 2005b]; and as a result, coastal upwelling zones are some of the most biologically productive in the world [Muller-Karger et al., 2005]. The resultant commercial fisheries are economically important [Chavez and Messie, 2009] and therefore ocean acidification could have important implications for both the marine ecosystem and the fisheries economy in these areas [Barton et al., 2012]. In the spring of 2007, Feely et al. [2008] found evidence of aragonite-undersaturated waters reaching the surface during strong upwelling along the coast of Northern California. A shellfish hatchery on the Oregon Coast also observed low ΩAr during the summer of 2009 [Barton et al., 2012]. With limited shipboard data, Hales et al. [2005a] showed that productivity on the shelf led to widespread depletions of surface water partial pressure of CO2 (pCO2) following upwelling events; however, Evans et al. [2011] found with moored time series measurements that offshore displacement of the productivity bloom associated with strong summer upwelling led to persistent high-pCO2 in inner-shelf surface waters, suggesting that upwelling can depress surface ΩAr over extended periods and spatial areas.

[4] Shipboard studies have been important for mapping spatial distributions in ΩAr [e. g., Hales et al., 2005a; Feely et al., 2008; Bates et al., 2009; Jiang et al., 2010; Cao et al., 2011; Fassbender et al., 2011], but the seasonal variability and magnitude of ΩAr have not been well-characterized. Juranek et al. [2009, 2011] and Alin et al. [2012] have made important progress in estimating ΩAr using the more widespread oxygen and temperature (T) data, but these empirical relationships are only useful for waters >30 m depth. Autonomous sensors for two carbonate system parameters, pCO2 and pH, are now available [DeGrandpre et al. 1995, 1999; Seidel et al, 2008], and extended time series studies can now be conducted on ocean moorings [Gray et al., 2011, 2012]. These data make it possible to calculate and characterize the daily, seasonal, and annual variation in ΩAr of highly dynamic regions like the Oregon shelf.

[5] As a follow-up to the air-sea CO2 flux time series study by Evans et al. [2011], we present long-term time series of ΩAr values in the Oregon coastal upwelling zone. Additionally, we use high temporal resolution data to examine the processes controlling short-term changes in ΩAr and compare the current ΩAr range to the estimated pre-industrial ΩAr range.

2 Methods

[6] Seawater pCO2 and pH were measured with Submersible Autonomous Moored Instruments for pCO2 and pH (SAMI-CO2 and SAMI-pH) [DeGrandpre et al., 1995, 1999; Seidel et al., 2008]. The data presented here are from instruments deployed at 2 m depth on the NH-10 buoy located on the 80 m isobath ~20 km from shore (124.304°W, 44.633°N) and at 116 m depth on the shelf break (SB) mooring located on the 120 m isobath ~35 km offshore (124.500°W, 44.641°N) along the Newport Hydrographic (NH) Line. From 2007 to 2008, only SAMI-CO2 instruments were deployed. From 2009 to 2011, SAMI-CO2 and SAMI-pH pairs were deployed on NH-10. Due to various problems, combined pCO2 and pH data were obtained only for spring 2009, autumn 2009, and from spring to early summer 2011. Saturation states, dissolved inorganic carbon (DIC), and other carbonate parameters were calculated in CO2SYS [Pierrot et al., 2006] using a combination of pH or pCO2 data and salinity-derived alkalinity (ATsalin). ATsalin can be calculated using the relationship reported by Gray et al. [2011] for the U.S. Pacific coast. As shown by Gray et al. [2011], ATsalin combined with pH or pCO2 can provide better estimates of ΩAr than the pH-pCO2 combination. The validity of the ATsalin relationship is checked by comparing the measured pH with that calculated from the ATsalin-pCO2 combination when both pH and pCO2 data were available. Based on this comparison, we found that the relationship in Gray et al. [2011] predicts AT with reasonable accuracy. The average differences for all periods when pH and pCO2 were available ranged from 0.01 ± 0.02 for pH to 7 ± 12 µatm for pCO2, which led to an average difference in calculated ΩAr of 0.04 ± 0.08. Further discussion of these uncertainties can be found in the supporting information.

[7] The processes controlling ΩAr dynamics were analyzed using property:property plots of ΩAr versus DIC or salinity (S) with reaction pathways super-imposed for net community production (NCP), air-sea gas exchange, water mass mixing, and dilution by river run-off. In brief, the NCP reaction pathway in the ΩAr:DIC plot was derived by varying DIC and AT in proportion to the Redfield ratio (106:18). Changes in ΩAr due to mixing of shelf water with the Columbia River (CR) plume were represented by a simple two end-member mixing model. The effect of air-sea gas exchange was modeled by incrementing the mean DIC by the air-sea CO2 flux keeping AT constant.

[8] The ΩAr time series were compared to pre-industrial ΩAr estimated using the difference between current and pre-industrial pCO2 based on a method modified from Feely et al. [2008]. The supporting information provides additional details on the methods used for this study.

3 Results

[9] The multi-year time series for ΩAr, S, and T are plotted in Figure 1, along with the North-South component of the winds from 2007 to show seasonal changes in upwelling and downwelling. The pCO2 and pH data, along with winds from 2008 to 2011, can be found in the supporting information (Figure S1). Surface ΩAr ranged from 0.66 to 3.9 (Figure 1a, Figure 2). The mean ΩAr was 2.2 ± 0.5, significantly below the estimated average for the North Pacific Ocean (3.3 ± 0.7) [Feely et al., 2009]. ΩAr was highly variable in all seasons except winter. The greatest temporal variability occurred during the summer upwelling season (Figure 1a, June through August) as periods of strong upwelling-favorable winds alternated with relaxation events (Figure 1e). In one event, surface water ΩAr changed by ~3 over a diel period (mid-June 2008, Figure 1a, point A). There were seven intervals of ΩAr < 1 (Figure 1a) on the surface at NH-10 that persisted from 6 h to 3 days, although the frequency and duration of upwelling events differed inter-annually. In their study of Netarts Bay (2 m average depth) on the Oregon coast, Barton et al. [2012] found frequent periods of undersaturation during summer 2009, a result of the stronger influence of upwelling near shore, as well as contributions from nighttime net respiration in shallow sediments of the bay.

Figure 1.

ΩAr time series at (a) NH-10 (2 m depth) and (b) shelf break (SB) (116 m depth). Refer to the legend in Figure 1a for all Figure 1 color schemes. ΩAr = 1 (dashed black). The filled red triangles represent discrete samples taken from surface waters near NH-10 (see supporting information). (c) Temperature at NH-10 (solid) and SB (dashed). (d) Salinity at NH-10 (solid) and SB (dashed). (e) 10 day averaged N-S component of wind for 2007 to show seasonal upwelling (negative values, equatorward winds) and downwelling (positive values, poleward winds). The wind data from 2008 to 2011 are found in the supporting information (Figure S1). Seasons are indicated in Figure 1e according to the division of seasons on the shelf used by Smith et al. [2001].

Figure 2.

Frequency distribution of ΩAr for (a) NH-10 and (b) the shelf break (SB). Data from 2007 to 2011 are represented by the unfilled bars, and pre-industrial ΩAr values are represented by the cross hatched bars.

[10] Average ΩAr values in SB bottom boundary layer waters were 1.4 ± 0.5, nearly one unit lower than values seen at the surface during the same time period (Figures 1b and 2). Due to remote wind forcing, low ΩAr waters (Figure 1b, point B) appeared long before local winds indicated upwelling-favorable conditions. The periods of low saturation support the seasonal dynamics of ΩAr predicted for SB waters by Juranek et al. [2009]. Unlike the near-Gaussian distribution seen in surface ΩAr (Figure 2), the deep SB water was constrained by the lower ΩAr of the California Undercurrent end-member, estimated to be 0.9 ± 0.3 from mean California Undercurrent T, S, DIC, and alkalinity (AT) (see supporting information).

4 Discussion

[11] Several processes contribute to ΩAr variability on the shelf including upwelling and other water mass changes (e.g., riverine inputs), NCP, and air-sea gas exchange. Variability from CaCO3 formation and dissolution is expected to be minimal in this area [Fassbender et al., 2011]. We plotted ΩAr versus DIC and S (Figure 3) to discern relative contributions from each mechanism, calculated as described in section 2 and supporting information.

Figure 3.

The relationships between ΩAr and DIC (calculated as described in the supporting information) and salinity during each season (labeled at right). (a,c,e,g) The Redfield (gray), (a,c,e,g) gas exchange (red), and (a–h) Columbia River dilution (black). The magnitude of the arrows corresponds to each relationship's potential effect on ΩAr. Years are identified by the color legend in lower right corner. Winter 2009 includes data from November 2009 through March 2010. The SB data are from November 2009 to March 2010.

[12] Several of these processes were important but varied widely between seasons and from year to year (Figure 3). Spring data are discussed first with interannual variability when applicable. Spring (April–May, 2008–2010) variability generally followed NCP and gas exchange relationships (Figure 3a). Surface ΩAr values extended along the Redfield relationship from low ΩAr (0.8–1.3) and high DIC indicative of the SB end-member to high ΩAr (>3.0) and low DIC as nutrient-rich upwelled waters triggered biological DIC uptake and drove ΩAr up. Biological production uses up the excess preformed nutrients that accompany the upwelled water [Hales et al., 2005a, 2005b], and surface water pCO2 can be driven well below atmospheric saturation, thereby increasing ΩAr and extending the original Redfield signature of the upwelled SB water (Figure 3a, SB data in 3g). Spring 2011 exhibited strong effects of CR dilution relative to the previous springs as ΩAr was depressed due to dilution of [Ca2+] and [CO32−] by the CR plume (Figures 3a–3b) [Salisbury et al., 2008; Chierici and Fransson, 2009]. The CR plume regularly extends southward from the river mouth during spring and summer, although its width and distance from shore vary based on wind conditions and water discharge [Hickey et al., 2005; Burla et al., 2010]. Water gauge records for the Columbia River (Table S3, supporting information) indicate that the spring 2011 river discharge was 130–140% greater than that during spring 2008 and 2009. The higher discharge and weaker upwelling (as measured by upwelling index, Table S3) during this period resulted in a stronger influence of the CR plume relative to other years. The 2011 data do not exactly follow the CR dilution trend, and it is likely that NCP increased ΩAr (up to ~0.75) above the ΩAr predicted by the CR dilution trend (Figures 3a and 3b). Fassbender et al. [2011] examined DIC dynamics on the U.S. West Coast using cruise data from two Northern California shelf transects during May 2007. They used a box model to estimate the effect of primary productivity on inorganic carbon dynamics and found that full utilization of available nitrate would result in a sea surface ΩAr of ~3. This value is comparable to the maximum spring sea surface ΩAr values observed at NH-10 (Figure 1a, points C and D; Figure 3a, 2008, 2009), supporting the importance of NCP in regulating ΩAr on the shelf.

[13] Relationships between ΩAr, DIC, and S during summer (June–August, 2008 and 2011) (Figures 3c–3d) are similar to those observed for the spring. In 2008 and 2011, the strong influence of the CR plume due to weaker upwelling and higher CR discharge (compared to 2010, Table S3) additionally suppressed ΩAr already lowered by upwelling.

[14] Trends in ΩAr in autumn (September–October) (Figures 3e–3f) and early winter (Figures 3g–3h) closely followed the predicted trends in biological productivity and gas exchange (Figures 3e and 3g) with little evidence of the dilution effects seen in spring and summer. In autumn 2009, the ΩAr was high (>3.0, Figure 1a, point E) but began to decline, likely due to convective mixing of the water column. In mid- to late November 2009, the mixed layer depth (MLD, not shown) deepened from near 25 m to >70 m, mixing surface and subpycnocline water and lowering surface ΩAr. Winter SB and NH-10 ΩAr relationships overlapped as a result of the homogenized water column (Figures 1a and 1b, point F; Figures 3g and 3h). After this period, surface ΩAr varied little for the remainder of winter 2009, with an average of 2.2 ± 0.4. Autumn 2010 ΩAr was significantly lower than in 2009 as a result of the prolonged and intense summer upwelling (wind data shown in Figures S1e and S1f).

[15] Winter SB ΩAr (the only season when SB data were collected) was mainly influenced by vertical mixing, downwelling, and movement of upwelled water onto the shelf driven by remote wind forcing [Hickey et al., 2006] (Figures 3g and 3h). There was little short-term variability in comparison to the surface data because these bottom waters were not influenced by gas exchange, heating or cooling, or the CR plume to any significant extent.

5 Implications

[16] These inorganic carbon data were used to back-calculate the natural (i.e., pre-industrial) ΩAr range on the shelf (see supporting information). This analysis indicates that mean contemporary ΩAr is 0.52 less than mean pre-industrial levels (Figure 2). Pre-industrial ΩAr was rarely undersaturated whereas contemporary surface values occasionally drop as low as 0.66. At the shelf break, contemporary ΩAr is undersaturated ~30% of the time, whereas pre-industrial undersaturation occurred only ~10% of the time. These changes in ΩAr from pre-industrial levels are consistent with the findings of the modeled simulations of Hauri et al. [2013] which also concluded that contemporary ΩAr observations in the California Current System have already shifted substantially from the pre-industrial range. It is unclear to what extent the shift to lower saturation in surface waters is affecting local organisms, all of which evolved in pre-industrial conditions. Reductions in ΩAr, even for waters that remain supersaturated, have been shown to harm many aragonite-forming organisms, resulting in shell dissolution for pteropods [Bednaršek et al., 2012; Comeau et al., 2012], decreased larval and mid-stage growth rates in bivalves [Gazeau et al., 2007; Barton et al., 2012; G. G. Waldbusser et al., A developmental and energetic basis linking larval oyster shell formation to acidification sensitivity, accepted by Geophysical Research Letters, 2013], and lower developmental rates in echinoderms [Shirayama and Thornton, 2005; O'Donnell et al., 2010]. Researchers that conduct ocean acidification-related organismal studies should consider the range of variability in this study when developing experimental designs. In addition, this study suggests that ΩAr can be calculated with reasonable accuracy from previously collected high-frequency pH or pCO2 data if salinity was also measured and the relationship with alkalinity is known.

[17] With atmospheric CO2 concentrations continuing to rise and an increasing trend in upwelling wind strength and duration [Bakun, 1990; Schwing and Mendelssohn, 1997; McGregor et al., 2007], it is likely that periods of undersaturation will increase in both frequency and intensity. Therefore, long-term monitoring of ocean saturation states will continue to be necessary.

Acknowledgments

[18] We thank C. Beatty (The University of Montana), C. Holm, M. Levine, D. Hubbard, D. Langner, and C. Wall (Oregon State University) for help with the SAMI deployments, and the OrCOOS program for the buoy. We thank the two reviewers for their constructive and detailed comments. This research was funded by the National Science Foundation (OCE-0836807 and OCE-0628569) and a NASA Space Grant Consortium Fellowship to K. E. Harris.

[19] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.

Ancillary