Biogeochemical cycles in the coastal ocean are changing and will continue to change in response to a changing climate. Effects on the oxygen and carbon cycles are particularly important, as either episodic or permanent shifts toward lower oxygen and/or higher inorganic carbon conditions can impact coastal ecosystems negatively. Here we study the sensitivity of these cycles to changes that may occur in the coastal ocean, focusing on a summer wind-driven upwelling region off southern Vancouver Island shelf. We use a quasi 2-D configuration of the Regional Ocean Modeling System (ROMS) to perform six sensitivity experiments. Results indicate that carbon and oxygen cycles in this region may be significantly affected by an altered upwelling season, a shallower offshore Oxygen Minimum Zone, and a carbon-enriched environment. Combinations of these scenarios suggest a potentially increasing risk for the development of coastal hypoxia and corrosive conditions in the region.
 Future climate change will affect every aspect of the ocean, from its physics to its chemistry and biology. In particular, biogeochemical cycles will not only be altered directly by the rising partial pressure of carbon dioxide (pCO2) in the atmosphere, but also by indirect effects, such as higher temperatures and changes in wind strength and patterns [e.g., Bakun, 1990; Falkowski et al., 2000]. For instance, the increase in sea surface temperatures and the strengthening of stratification as global climate warms is expected to be sufficient to intensify existing hypoxia and generate hypoxia in new areas [Rabalais et al., 2010]. In addition, if the biotic carbon to nitrogen ratio (C:N) increases under elevated CO2 conditions (as observed in a mesocosm experiment [Riebesell et al., 2007]), anthropogenic CO2 emissions may extend tropical oxygen (O2) deficient zones [Oschlies et al., 2008]. Models project a decline in O2 concentrations in the oceans during the 21st century [Frölicher et al., 2009] and over the next 100,000 years [Shaffer et al., 2009]. Recent warming in the 1990s has already produced an estimated global oceanic O2 outgassing of 0.3 ± 0.4 1014 mol-O2 yr−1 [Keeling and Garcia, 2002].
 Oxygen Minimum Zones (OMZ) are permanently hypoxic regions of the open ocean (O2 < 60 mmol m−3), typically along the continental margins of the eastern Pacific, eastern Atlantic, and Indian oceans [Helly and Levin, 2004; Stramma et al., 2008]. Time series of O2 show a persistent decline of concentrations in different OMZs as well as a shoaling of their upper boundaries [Whitney et al., 2007; Stramma et al., 2008]. These changes could impact shallower waters in regions where OMZs affect outer continental shelves and upper slopes [Helly and Levin, 2004], especially where wind-driven upwelling brings deep waters closer to or onto the shelves (e.g., western North America). The shoaling of the OMZ in the northeast Pacific [Whitney et al., 2007] may be a cause of recently observed hypoxic events in the coastal waters off Oregon and California [Grantham et al., 2004; Bograd et al., 2008; Chan et al., 2008], which in some cases turned the shelves into “dead zones” due to the widespread mortality of benthic animals.
 Cycling of carbon in the ocean is also expected to change as pCO2 increases in the atmosphere and climate changes. A decline in pH and an increase in surface pCO2 have already been observed at several oceanic time series sites: stations ALOHA in the central North Pacific [Dore et al., 2009] and BATS, off Bermuda in the North Atlantic (both time series longer than 20 years [Bates, 2007]), as well as station ESTOC (100 km north off Gran Canaria Island in the North Atlantic [Santana-Casiano et al., 2007]), which has a 10 yearlong record. These stations show that surface ocean pCO2 has increased at rates indistinguishable from the atmospheric increase (1.5 to 1.9 μatm yr−1 [Bindoff et al., 2007]). On the west coast of the US, model simulations indicate a pH decrease of ∼0.1 since pre-industrial times [Hauri et al., 2009]. Moreover, corrosive waters have been observed on shelves off western North America [Feely et al., 2008]. There, the undersaturation horizon of aragonite (the less stable form of calcium carbonate, CaCO3, found in corals for example) was observed at depths between 40 and 120 m, even reaching the surface along one transect off northern California.
 In contrast to other coastal regions of western North America, wide spread hypoxic events and corrosive waters have not yet been observed along the west coast of Vancouver Island. This region represents the northern limit of the California Current System (CCS) in Pacific Canada. The outflow from the Juan de Fuca Strait, relatively fresh and nutrient-rich due to intense tidal mixing in the Strait, generates a buoyancy-driven coastal current, known as the Vancouver Island Coastal Current (VICC) [Freeland et al., 1984; Thomson et al., 1989; Crawford and Dewey, 1989]. This O2-rich current and the relatively wide shelf protect the shallower waters from developing hypoxia and aragonite undersaturation [Bianucci et al., 2011]. Moreover, wind-driven upwelling only occurs during summer [Strub et al., 1987a, 1987b], such that the influence of carbon-rich and O2-depleted deep waters from offshore is limited to that season (winds are downwelling-favorable during winter). However, it is uncertain if hypoxia and undersaturation may develop in the future, given potential alterations to environmental forcing with climate change. Projections for the 21st century from an earth system model have predicted a decrease in O2 concentrations and increase in ocean acidification in the entire CCS [Rykaczewski and Dunne, 2010]. Therefore, we use a coastal circulation model to investigate the sensitivity of the carbon and O2cycles to different forcings in our region of interest. A quasi 2-D configuration represents summer upwelling over the southern Vancouver Island shelf. Six experiments (details insection 2) investigate the model sensitivity to changes in strength of upwelling, depth of the OMZ, and inorganic carbon concentration. We introduce to our analysis the concept of the Respiration Index [Brewer and Peltzer, 2009a, section 3], which simultaneously considers the effects of changes in both O2 and pCO2 on marine organisms. This index is calculated in the different simulations and compared with the traditional definitions of hypoxia.
2. Design of Sensitivity Experiments
 We have developed a quasi 2-D application of the Regional Ocean Modeling System (ROMS,www.myroms.org [Haidvogel et al., 2008]) to model upwelling on the Vancouver Island shelf in response to time variable wind-forcing. The domain is a transect perpendicular to the isobaths (i.e., cross-shore distance versus depth) at ∼49°N (Figure 1) and the small alongshore dimension (∼5 km) has uniform conditions. A summary of the model configuration is provided in Table 1. This quasi 2-D approach precludes the inclusion of an alongshore pressure gradient that varies with cross-shelf distance and depth, preventing the modeling of a dynamic VICC [Masson and Cummins, 1999]. Hence, to model the VICC, water properties in the nearshore ∼6 km are restored toward a fixed vertical profile representative of the core of the VICC. The restoring is strongest at the inshore boundary and decreases as exp(−x2) toward offshore, where x is the distance from the inshore boundary (for more details, see Bianucci et al. [2011, Appendix B]. The model configuration and evaluation, as well as the biological and sediment modules coupled to the physical model, are described in detail elsewhere [Bianucci, 2010; Bianucci et al., 2011]. Vertical profiles represented correctly the observed vertical structure of temperature, nitrate, and other variables. Moreover, the statistical properties of the model and observations were in agreement. Modeled time series of sea surface height captured some of the surface dynamics observed at a meteorological buoy in the region. The model also represented the observed cross-shore gradient of DIC [Ianson et al., 2003], such that VICC waters are enriched in DIC with respect to shelf waters at comparable depths.
Table 1. Summary of Model Configuration
Cross-shore (Lx) = 185 km.
Alongshore (Ly) = 5 km.
Vertical (H) = from 40 to 1500 m.
Cross-shore (dx) = 953 m
Alongshore (dy) = 1.67 km.
Vertical (dz) = from 1.3 to 72.8 m (30 σ-layers).
dt = 320 sec.
Total length of experiments
 Coastal upwelling ecosystems are intrinsically 3-D systems, where mesoscale phenomena play an important role [e.g.,Gruber et al., 2006; Lathuilière et al., 2010]. However, a 2-D model is able to represent locally forced upwelling and facilitates extensive sensitivity analysis. Previous studies on the Oregon shelf have successfully used 2-D models [Allen et al., 1995; Federiuk and Allen, 1995; Spitz et al., 2003]. Here, a set of model experiments helps to determine the sensitivity of the carbon and O2 cycles to different forcing and changing conditions. By changing one aspect of the model at a time, we compare the differences with respect to a control simulation (experiment 1). The sensitivity experiments are described below and summarized in Table 2. Surface incoming shortwave radiation and net heat fluxes are specified in all experiments from NCEP reanalysis as daily values for the period 27 May to 29 September 1993, at 48.57°N, 125.62°W. Wind stress is calculated from observed hourly winds at meteorological buoy 46206 (48.83°N, 126.00°W, Figure 1) following Smith , then filtered with a 6-hour low-pass Fast Fourier Transform (FFT) filter. Atmospheric pCO2 concentration (pCO2atm) is set as a constant boundary condition (370 ppmv, except in experiment 6). To provide sufficient time for spin-up, since sediments take more than a month to approach equilibrium [Bianucci et al., 2011], the experiments start on 27 May 1993 (day 0) and analyses begin on 16 July (day 50).
Table 2. Description of Model Sensitivity Experiments
Values in parentheses represent conditions in control experiment 1.
Control experiment with 1993 late spring and summer forcing from 27 May 1993 (day 0) to 29 September 1993 (day 125).
Like 1, but with 1993 wind stress during spin-up (day 0 to 50) and 2002 wind stress (increased upwelling) after day 50 (16 July).
Mean τy = −0.028 N m−2 (−0.018 N m−2)
Like 1, but with 2002 wind stress from day 0.
Mean τy = −0.023 N m−2 (−0.018 N m−2)
Like 1, but initial conditions for O2 with a shallower OMZ and lower O2 concentrations.
Min. initial O2 at ∼800 m = 1.3 mmol m−3 (10.6 mmol m−3)
OMZ hypoxic threshold depth = 280 m (380 m)
Like 4, but with 2002 wind stress from day 0.
Min. initial O2 at ∼800 m = 1.3 mmol m−3 (10.6 mmol m−3)
OMZ hypoxic threshold depth = 280 m (380 m)
Mean τy = −0.023 N m−2 (−0.018 N m−2)
Like 1, but with higher pCO2atm and DIC initial conditions (expected 2000 to 2050 changes from CanESM 1.1 A2 simulation). CanESM 1.1 is the Canadian Earth System Model version 1.1 [Arora et al., 2009; Christian et al., 2010].
Max. initial DIC = 2291 mmol m−3 (2273 mmol m−3)
pCO2atm = 513 ppmv (370 ppmv)
2.1. Control Experiment (Experiment 1)
 Wind and surface heat forcing corresponds to late spring and summer 1993, a year representing a “normal upwelling summer”: upwelling indices for July and August (38 and 26 m3 s−1 per meter of coastline, respectively) were close to the climatological monthly averages (34 and 22 m3 s−1 per meter of coastline; upwelling indices from the Environmental Research Division, Pacific Fisheries Environmental Laboratory). Observed deep ocean summer profiles from the study region are used to create average depth profiles to initialize scalar properties (i.e., horizontally uniform distributions), and initial velocities are set to zero.
2.2. Stronger Upwelling Experiments (Experiments 2 and 3)
 Increasing trends in upwelling intensity have been observed in some major coastal upwelling systems of the world during the 20th Century [e.g., Bakun, 1990; Schwing and Mendelssohn, 1997; McGregor et al., 2007]. Intensified summer upwelling due to increased atmospheric pCO2 is predicted from a regional climate model with high resolution over the coastal region of California [Snyder et al., 2003]. Off the Vancouver Island shelf, an ensemble of 18 climate models predicts increased upwelling summer winds in the 21st century [Merryfield et al., 2009]. Therefore, we test the sensitivity of the O2and carbon cycles in the model to increased upwelling. The summer of 2002 experienced higher than normal upwelling-favorable winds off Vancouver Island. Mean alongshore wind stress was 28 % greater in 2002 than in 1993 for the period 27 May to 29 September (−0.018 N m−2 in 1993 versus −0.023 N m−2 in 2002; see Figure 2a). For the period 16 July to 29 September (the period of analysis following 50 days of spin-up), mean alongshore wind stress almost tripled in 2002 (−0.026 N m−2) relative to 1993 (−0.009 N m−2). The simulations with stronger upwelling are forced with 2002 wind stress in two ways: experiment 2 has the same spin-up as the control experiment (first 50 days with 1993 wind stress), while experiment 3 is forced from the start with 2002 winds. The first approach allows a comparison of results after a common forcing during the 50 day spin-up period; the second allows examining the effect of using different forcing during spin-up.
2.3. Shallower OMZ Experiments (Experiments 4 and 5)
 Although not all species have the same tolerance to low O2 concentrations [Vaquer-Sunyer and Duarte, 2008], hypoxia is commonly defined as waters with O2 < 60 mmol m−3 (= 60 μM ∼ 60 μmol kg−1 ∼ 1.4 mL L−1) [Gray et al., 2002; Whitney et al., 2007; Stramma et al., 2008]. The hypoxic threshold at Ocean Station Papa (OSP) has shoaled roughly by 100 m (from 400 to 300 m depth) between 1956 and 2006 [Whitney et al., 2007]. The southern CCS has experienced a shoaling of up to 90 m in the period 1984–2006 [Bograd et al., 2008]. Moreover, O2concentrations in the OMZ are decreasing with observed rates of 0.18 mmol-O2 m−3 yr−1at a depth of 800 m at OSP and 0.15 mmol-O2 m−3 yr−1 at a depth of 500 m in the southern CCS region [Whitney et al., 2007; Bograd et al., 2008]. To analyze the effects of a shallower and more intense OMZ on the southern Vancouver Island shelf system (experiment 4), we modify the initial O2field such that the hypoxic threshold is 100 m shallower than in the control simulation. Moreover, the minimum concentration is changed to 1.3 mmol-O2 m−3, ∼9 mmol-O2 m−3 lower than the minimun in the control experiment (Figure 2b). This O2 decrease would be achieved in 50 years at the observed rate at 800 m at OSP. In experiment 5, in addition to the shallower OMZ, the model is forced with the stronger upwelling winds from 2002.
 As CO2 increases in the atmosphere due to anthropogenic activities, roughly one third of emissions are absorbed by the ocean [Sabine et al., 2004; Sabine and Feely, 2007]. Experiment 6 examines the effect of increased pCO2atm and ocean dissolved inorganic carbon (DIC) on biogeochemical cycles in the model. The increments correspond to the predicted changes in DIC and pCO2atm over the period 2000 to 2050 from a global climate model (Figures 2c and 2d). We use the Canadian Earth System Model (CanESM 1.1), with emission scenario A2 [Arora et al., 2009; Christian et al., 2010] at the model location closest to the Vancouver Island shelf (50°N, 130°W). DIC concentrations in the VICC were assumed to increase by the same amount as those over the shelf. pCO2atm, which is set as a constant boundary condition, was increased by 143 ppmv relative to the control experiment (from 370 to 513 ppmv).
3. Respiration Index (RI) and Aragonite Saturation State (ΩA)
 Recently, Brewer and Peltzer [2009a] argued that elevated pCO2 may impose a physiological strain on higher animals, and that the use of an O2 limit alone to define a dead zone implicitly assumes that pCO2 levels are low and inversely proportional to O2. They defined a Respiration Index that is linearly related to available energy in basic oxic respiration (RI = log(pO2/pCO2)). The RI reflects the thermodynamic energy yield of aerobic respiration as the concentration ratio of substrate and product changes. This concept suggests that, as atmospheric pCO2 rises and more carbon is absorbed by the ocean, dead zones could expand even if O2 levels were not affected [Brewer and Peltzer, 2009a]. RI = 1 can be considered as a general boundary for aerobic stress, although Brewer and Peltzer [2009a]pointed out that the actual limits will be species-dependent.
 Some of the assumptions behind this index aroused controversy. For instance, the calculation assumes a closed thermodynamic system, while living organisms are essentially open systems [Seibel et al., 2009]. These authors also were concerned about the use of environmental gas partial pressures, arguing that the intracellular concentrations are regulated independently by kinetic and physiological mechanisms. Despite these criticisms, to which Brewer and Peltzer [2009b] responded, we use the RI in the context of our modeling study to evaluate the combined effect of O2 and CO2 as a threshold for habitable shelf environments, compared with the more typical O2 thresholds.
 Aragonite is the less stable form of CaCO3 in the ocean and its degree of saturation can be approximated as ΩA ∼ [CO32−]/[CO32−]sat where [CO32−] and [CO32−]sat represent the concentrations of the carbonate ions in ambient seawater and at saturation. When ΩA > 1 (<1), seawater is supersaturated (undersaturated) with respect to aragonite and favors calcification (dissolution). We will use ΩA as well as pCO2 to describe the carbon state of the system.
 ΩA is calculated from modeled temperature (T), salinity (S), DIC, and total alkalinity (TA) using the CO2SYS software [Lewis and Wallace, 1998]. When comparing changes in ΩA between two simulations, we calculate the total change as
where the subscripts 1 and n indicate the control experiment and any of the sensitivity simulations, respectively. To evaluate the role that DIC alone plays in we compute where ΩA∗ is calculated with T, S, and TA from experiment 1 and DIC from experiment n. allows us to quantify the effect of changing only DIC on The same procedure can be performed for every variable (e.g., computation of etc) or for combinations of variables (e.g., the combined role of TA and DIC leads to
4.1. Effect of Increased Upwelling
 Temporal and vertical averages of O2 and pCO2 across the shelf are compared for the control experiment (experiment 1) and experiment 2, which has increased upwelling (Figure 3, blue and red lines, respectively). The upper 30 m of the water column experiences higher O2 concentrations under intensified upwelling over most of the shelf (Figure 3a), while below 30 m depth these wind conditions decrease O2 concentrations (Figure 3c). The O2 changes are up to +7 and −11% in the upper and lower region of the water column, respectively. The inner shelf concentrations are similar in both simulations due to the restoring to VICC properties. On average, pCO2 in the upper ocean is lower than pCO2atm (dashed grey line in Figure 3b) over the mid- and outer shelf in both experiments, so air-sea CO2 fluxes are from the atmosphere to the ocean in those regions (surface pCO2 has a similar pattern but lower values than the average over the upper 30 m). However, the high DIC concentrations specified in the VICC lead to pCO2 concentrations higher than pCO2atm over the inner shelf. CO2 outgassing occurs on the inner 4 km of the model domain in experiment 1 and within 6 km in experiment 2.
 In experiment 2, pCO2 is higher in the bottom layers (Figure 3d) and lower in the upper 30 m over most of the model domain (Figure 3b), with changes reaching up to +4 and −7% in each case. However, there is a region of the inner shelf (∼10–25 km from inshore boundary) where pCO2 is higher in the upper layer under intensified upwelling; moreover, approximately in the same region (∼10–18 km) O2 is lower under stronger upwelling relative to the control experiment 1. These changes respond to the position of phytoplankton blooms in both simulations, since phytoplankton are advected offshore near the surface during intense upwelling events (Figure 4). Regression analysis predicts a ∼20 km displacement of the maximum primary production from the inshore boundary for a mean alongshore equatorward wind stress of 0.05 N m−2 in the previous 9 days [Bianucci, 2010]. The bulk of the primary production (which increases by 12%) occurs farther offshore in experiment 2, over the mid- and outer shelf (between 10 and 40 km,Figure 4). Moreover, high primary production over the mid-shelf leads to more organic matter reaching the seafloor at those depths, contributing to enhanced exchange of O2 and DIC with the sediments. Between 20 and 30 km from the inshore boundary (i.e., at bottom depths between 90 and 115 m) O2 and DIC exchanges with the sediments increase up to 50% (the increment is 8% on average).
 As the development of hypoxia and/or corrosive conditions first occurs in the bottom waters over the shelf, we focus on the spatial and temporal distributions of O2 and ΩAin the near-bottom layer of the model (Figure 5). This layer accurately represents the bottom boundary layer, except within the shallowest 3 km of the domain where restoring to the VICC is strongest [Bianucci et al., 2011]. With increased upwelling as in experiment 2, low O2concentrations develop in the bottom layer inshore of the shelf break, reaching a minimum of 34 mmol-O2 m−3 (Figure 5c). In particular, the hypoxic boundary (60 mmol-O2 m−3, bold black contour) migrates inshore across the shelf after the intense upwelling-favorable wind event centered on day 80(seeFigure 2a). The onset of hypoxia in shallower waters is triggered by advection of low oxygen during that event (red line in Figure 6). Once upwelling intensity decreases, biological consumption (especially in the sediments; black line in Figure 6) maintains the low O2 levels achieved by advection. The enhanced advection of low O2 results partly from the stronger upwelling circulation and partly from the lower O2 content of upwelled waters (the depth of upwelling increases ∼40 m in experiment 2). Since advection and local biological consumption generate the lowest O2 concentrations in bottom waters, vertical mixing generates a downward flux of O2 to the bottom layers from the overlying, more oxygenated waters (blue line in Figure 6).
 The wind-forcing used during the first 50 days of the experiments with stronger upwelling affects the results. An intense upwelling event between days 35 to 45 in 1993 (∼first week of July) advects the hypoxic threshold closer to the shelf break. Therefore, if 1993 winds are used during spin-up (experiment 2,Figure 5c), waters on the shelf become more hypoxic than if 2002 winds are used from day 0 (experiment 3, Figure 5e). From day 50 to 125, hypoxia covers 43% of the near-bottom waters over the shelf in experiment 2 compared with 18% in experiment 3. Moreover, the minimum O2concentration is lower in experiment 2 relative to experiment 3 (34 versus 41 mmol-O2 m−3). Hence, the timing of the onset of the upwelling season is another factor determining both the timing and magnitude of hypoxia (and analogously, of events with lower aragonite saturation state).
 Dissolved carbon concentrations in the near-bottom layer over the shelf respond to the same processes as dissolved O2: DIC increases due to advection of high concentrations from offshore, while sediment remineralization maintains high DIC over the mid-shelf during relaxation and periods with weak winds (DIC budget terms not shown). In experiment 2, ΩAdecreases in the near-bottom layer over the shelf near day 90 (Figure 5d) relative to the control experiment 1 (Figure 5b). As ΩA depends on T, S, DIC and TA, we evaluate the contribution of each variable to the total change in ΩA between both simulations as explained in section 3 (Figure 7). The increase in DIC by upwelling is primarily responsible for the drop in ΩA (Figure 7a), while the increase in TA cancels out about 40% of the DIC effect (Figure 7b). Upwelling also brings colder waters from offshore that reduce the temperature near the bottom over the shelf by up to 3°C (the mean cooling over the near-bottom shelf is 0.63°C), which tends to decrease ΩA by increasing [CO32−]sat (section 3). The reduction in ΩA due to cooling (Figure 7c) is small compared with the change due to increasing DIC (Figure 7a). The changes in salinity due to stronger upwelling do not significantly affect ΩAover the near-bottom of the shelf outside the VICC region (Figure 7c). The combined effect of DIC and TA (Figure 7d) accounts for essentially all of the change in near-bottom ΩA between experiments 1 and 2 (black bars in Figure 7).
4.2. Effect of a Shallower Offshore OMZ
 On average (from day 50 to 125), the bottom waters over the shelf have lower O2 in the experiment with a shallower OMZ (experiment 4) compared with the control experiment 1 (Figure 8). Most of the shelf has bottom concentrations below 80 mmol-O2 m−3in experiment 4, while in experiment 1 shelf waters are mainly above that level (waters between 60 and 80 mmol-O2 m−3 are shaded in Figure 8). The upper layers of both simulations remain similar, since their initialization is the same (Figure 2b) and the changes in deeper waters do not greatly modify biology in the upper ocean or air-sea O2exchange. The mean hypoxic threshold (60 mmol-O2 m−3, bold black line in Figures 8a and 8b) reaches the shelf break in experiment 4, penetrating up to depths of ∼135 m. During the 75 days of analysis, the near-bottom hypoxic threshold in experiment 4 is either near the shelf break or over the outer shelf (bold black contour inFigure 5g). The threshold penetrates onshore after strong upwelling events (after days ∼40 and 108, see Figure 2a). The lower O2 concentrations with respect to the control experiment 1 respond to two factors: (1) the initial conditions, which are already lower for depths greater than 130 m (Figure 2b), and (2) the enhanced advection of lower O2from offshore onto the mid- and outer shelf (see O2 budget in Figure 9). Biological sinks (both in the water column and the sediments) remain essentially unchanged over the shelf between experiments 1 and 4 (Figure 9), which differ only in the initial profiles of O2 concentrations. Since the lower bottom O2 concentrations increase the vertical gradient of O2, vertical mixing is enhanced in experiment 4 and partly compensates the decrease in O2 by advection.
 When the modified initial conditions with a shallower OMZ are used in combination with stronger upwelling-favorable winds (experiment 5), O2concentrations decrease even more in the near-bottom layer over the shelf (Figure 5h). Most of the bottom waters over the shelf become hypoxic after the strong upwelling event around day 80, except for the area influenced by the O2-rich VICC. O2concentrations drop to 21 mmol-O2 m−3near the shelf break, the lowest over the shelf for all the experiments presented here. Concentrations lower than 20 mmol-O2 m−3 (or 0.5 mL L−1) are usually considered to represent “severe hypoxia” [e.g., Monteiro et al., 2006; Chan et al., 2008; Diaz and Rosenberg, 2008].
 In experiments 4 and 5 (Figures 5g and 5h), the region with RI ≤ 1 expands greatly compared with experiments 1 to 3 (Figures 5a, 5c, and 5e; bold yellow contours indicate the RI = 1 contour). The decrease in RI is mainly due to the lower O2concentrations in the near-bottom layer of the open ocean, since pCO2 does not change significantly in those waters. In particular, the shoaling of the OMZ does not significantly affect the carbon cycle, e.g., differences between experiments 1 and 4 for pCO2 and ΩA are < 0.6% and 0.4%, respectively (not shown).
4.3. Effect of Higher Inorganic Carbon
 Higher DIC initial conditions (experiment 6, see Figures 2c and 2d) increase pCO2 throughout the model domain (Figure 10). The bold black contours in the averaged pCO2 vertical sections correspond to the atmospheric pCO2atm values in experiments 1 (370 ppmv, Figure 10a) and 6 (513 ppmv, Figure 10b). Over most of the shelf, the surface ocean is on average undersaturated in both simulations and absorbs CO2 from the atmosphere. However, surface VICC waters within 4 km from the inshore boundary are oversaturated in the control experiment 1 (releasing CO2 to the atmosphere), in contrast to experiment 6. Aragonite undersaturation expands considerably over the shelf in the latter scenario. The average saturation horizon (ΩA = 1) coincides with averaged pCO2 values of 710 and 800 ppmv in experiments 1 and 6, respectively (Figure 10). The sensitivity of ΩA to pCO2 varies between these experiments due to the different effect of DIC on ΩA and pCO2. We can approximate the former using the carbonate ion concentration in seawater and at saturation [Sarmiento and Gruber, 2006]: ΩA ∼ [CO32−]/[CO32−]sat∼ (TA-DIC)/[CO32−]sat. Hence, an increase in DIC leads to lower ΩA. However, pCO2 can be approximated by (2 × DIC − TA)2/(TA-DIC), such that higher DIC increases pCO2.
 In experiment 1, the saturation horizon in the open ocean is below 250 m (not visible in Figure 10a). However, in experiment 6, the saturation horizon shoals ∼250 m in the open ocean with respect to experiment 1 and extends over the shelf up to ∼30 m above the seafloor (Figure 10b). Moreover, the saturation state in experiment 6 is below unity (reaching down to 0.78) after day 50 in the near-bottom layer over the whole shelf seaward of 8 km from the inshore boundary (Figure 5f). There, the VICC carries waters with lower DIC relative to near-bottom waters over the shelf [Bianucci et al., 2011, and references therein]. However, in the control experiment 1, near-bottom waters with ΩA ≤ 1 are found only offshore from ∼45 km from the inshore boundary (Figure 5b).
 The area with RI < 1 expands modestly in experiment 6 compared with the control experiment 1 (yellow bold contour in Figure 5f). In waters deeper than 1000 m, the modified initial conditions do not show a large increase in DIC (Figure 2f). Consequently, pCO2increases only moderately in the near-bottom waters of the slope (where RI < 1, at ∼60 km from the inshore boundary). If RI is calculated using pO2 from experiment 4 instead, the area with indices values <1 is almost identical to that of experiment 4 (only slightly larger). The combination of pO2 from experiment 4 and pCO2from experiment 6 is consistent with an scenario of low-O2and high-DIC, since T and S are the same in both simulations and changes in TA are minimal.
5. Discussion and Conclusions
 The experiments described here provide insight into the sensitivity of the carbon and O2 cycles to changes in forcing or initial conditions in a seasonal upwelling region. Both cycles respond to changes in upwelling in our model experiments. However, only O2 concentrations are affected by the shoaling of the OMZ and only the carbon cycle responds to changes in the oceanic DIC inventory. Nonetheless, changes in the depth of the OMZ and carbon content of the ocean will not be independent of each other, as both are expected to respond to the increase in atmospheric CO2 from human activities. Moreover, these changes will be coupled to modifications in the cycling of other nutrients. Given the complexity of the feedbacks between biogeochemical cycles, the introduction of a single change per experiment allows a clearer understanding of the responses to any given change.
 This study extends that of Ianson and Allen , which described results from a compartment model. Our quasi 2-D model [Bianucci et al., 2011] can resolve the spatiotemporal response to wind-forcing in both the vertical and across-shelf directions. Moreover, the current work focuses on sensitivities to climate-related change of interactions between the carbon system and O2, which was not included in the model of Ianson and Allen .
 Strong upwelling events (such as the one at day ∼80 in experiment 2) result in onshore advection of low O2 and high DIC from offshore that triggers low O2 and ΩAevents in the near-bottom layers over the shelf. The ΩA decrease is partially compensated by the onshore advection of high TA (advection of cooler waters has a much smaller effect). In addition to the intensified onshore advection under increased upwelling scenarios, O2consumption and DIC production in the sediments are enhanced over the mid-shelf due to offshore transport of near-surface phytoplankton blooms. The latter increases deposition of organic matter to the seafloor over the mid-shelf, enhancing the exchange of O2 and DIC between the water column and the sediments. Therefore, low O2 and ΩA values are maintained after the upwelling events by remineralization within the sediments. O2concentrations in particular are demonstrated to be sensitive to intensified upwelling, since hypoxia develops in the near-bottom waters on the shelf in both simulations with stronger upwelling (experiments 2 and 3).
 Differences between experiments 2 and 3 (which differ only in the wind conditions during the spin-up period of 50 days) demonstrate the effect of the timing of the onset of the upwelling season. If summer upwelling is stronger and starts earlier (experiment 2), O2-poor and DIC-rich near-bottom waters are advected upwards closer to the shelf break earlier in the season. Thus, subsequent upwelling events further reduce O2 and ΩAin the near-bottom layers over the shelf. A regional climate model of the northern California shelf, with 40 km horizontal resolution, projects a 1-month delay of the onset of seasonal upwelling for increased atmospheric CO2 (560 to 686 ppmv), as well as intensified upwelling [Snyder et al., 2003]. In the context of the present study, a delayed upwelling season would reduce the potential for hypoxia and acidification on the Vancouver Island shelf. In contrast, Barth et al. reported on the negative consequences of a one-month delay in the 2005 transition to upwelling-favorable wind stress on the California and Oregon shelves (warm waters, low nutrient levels, low primary productivity, and low recruitment of rocky intertidal organisms).
 The effect of a shallower OMZ (experiment 4) is straightforward: it moves the hypoxic threshold closer to the shelf break, so upwelling events are more likely to advect O2-poor waters onto the shelf. Therefore, the combination of stronger upwelling and a shallower OMZ (experiment 5) would further reduce O2 concentrations on the shelf, representing a potential increasing stress on ecosystems. The present model represents the shelf off Vancouver Island: despite the considerable shelf width, the influence of a shallower OMZ reaches the inner shelf (except for the region influenced by the O2-rich VICC). The effect could be more severe on narrower shelves, where upwelling can transport offshore waters to even shallower depths onshore. The expected overall decrease in O2 concentration in the subsurface open ocean would lead to an expansion of the region with RI < 1, implying a larger area where aerobic marine life would be under stress [Brewer and Peltzer, 2009a]. In none of the experiments performed did RI reach values below unity over the shelf (although O2levels did decline below the 60 mmol-O2 m−3hypoxic threshold). These results suggest that RI is an indicator of more severe aerobic stress than the 60 mmol-O2 m−3 threshold.
 RI calculated for a hypothetical scenario with lower O2 and higher DIC showed a higher sensitivity of this index to O2 than to CO2 changes. This difference results from the changes in the initial conditions in experiments 4 and 6, since the fractional decrease of initial O2 is larger than the fractional increase of DIC (both changes aim to represent a projection 50 years into the future).
 Under elevated inorganic carbon conditions (experiment 6), near-bottom waters over the shelf (beyond the area of influence of the VICC) become corrosive after 50 days, i.e., ΩA drops below unity. In the current model configuration, no biological process depends on DIC concentration (e.g., calcification is not included), so the increase of DIC in the ocean does not affect the O2or nitrogen cycles directly. Decreased calcification in a carbon-rich ocean may affect photosynthesis and phytoplankton community structure [Sikes et al., 1980; Paasche, 2001], providing a potential link between O2 and carbon. Furthermore, the model is missing a negative feedback in the carbon cycle that could potentially reduce acidification: a reduction of calcification (increase of CaCO3 dissolution) reduces the production (enhances the consumption) of CO2 in seawater through the reaction 2HCO3− + Ca2 + ⇌ CO2 + CaCO3 + H2O. Most published works agree on the reduction of calcification rates with higher pCO2, although Iglesias-Rodriguez et al.  reported the opposite and aroused some controversy [see Riebesell et al., 2008]. Blooms of calcifiers (e.g., coccolithophorids) are not frequent over the Vancouver Island shelf, although they have been observed (D. Ianson, personal communication, 2010).
 We need to understand how biogeochemical cycles will respond to climate change in the coastal ocean, since we depend on its resources. The present work contributes toward that goal by focusing on the sensitivities of biogeochemical cycling to environmental factors, individually and collectively, that may change (in most cases, certainly will) as the anthropocene evolves. The perturbations analyzed here (increased upwelling, shallower OMZ, and higher DIC and pCO2atm) consistently drive the system toward lower O2, pH, and ΩA states. These results emphasize the potential negative impact that these perturbations may have on benthic ecosystems, along the lines of the consequences first hypothesized by Bakun  for increased upwelling. Therefore, this sensitivity analysis suggests that a region such as the Vancouver Island shelf could develop hypoxia and corrosive conditions (ΩA< 1) in the future. Full 3-D biogeochemical modeling as well as continuous and simultaneous observations of carbon and O2 are needed to assess the current state of these cycles in the coastal ocean and how they vary in time and space.
 We thank K. Fennel for her constructive comments on an earlier version of this manuscript, as well as the valuable input from two anonymous reviewers. L.B. acknowledges several sources of graduate support: a UVic Fellowship, a Maritime Award Society of Canada Scholarship, and a Bob Wright Fellowship. Computing support was provided by the Canadian Centre for Climate Modeling and Analysis, Environment Canada. K.D. acknowledges funding from an NSERC Discovery Grant.