Primary producers may ameliorate impacts of daytime CO₂ addition in a coastal marine ecosystem

Study location and characteristics

We conducted a CO₂ addition experiment on a rocky intertidal shoreline on the northern side of Horseshoe Cove in the Bodega Marine Reserve (BMR), Sonoma County, California, USA (NAD83 38.31672, −123.0711). This work was conducted with the approval of the California Department of Fish and Wildlife (Scientific Collecting Permit number SCP-13405). Our work in the BMR was performed under Research Application # 32752 to the University of California Natural Reserves System to conduct measurements of carbonate chemistry in tide pools. At the site, we marked and surveyed 10 tide pools in the mid-intertidal zone. For each pool, we determined tidal elevation (mean = 1.54 ± 0.05 [s.e.] m above mean lower-low water; range = 1.22 to 1.73 m), volume (mean = 22.3 ± 5.0 L; range = 6.7 to 47.3 L), and bottom surface area (mean = 0.35 ± 0.08 m²; range = 0.16 to 1.00 m²). Tidal elevations were determined using a self-leveling rotary laser level (CST/berger, Watseka, Illinois, USA), with reference to published tidal predictions for Bodega Harbor Entrance (NOAA, 2016). Pool volumes were determined by adding 2 ml of non-toxic blue dye (McCormick, Sparks, Maryland, USA) to each pool. Water samples were read at 640 nm on a benchtop spectrophotometer (UV-1800, Shimadzu, Carlsbad, California, USA) and compared to a standard curve relating absorbance to an equivalent amount of dye added to known volumes of seawater (Pfister, 1995). The abundances of primary producers and sessile invertebrates in each tide pool were determined by spreading a flexible mesh grid (10 cm × 10 cm mesh; Foulweather Trawl Supply, Newport, Oregon, USA) over the bottom of each pool and measuring the cover of macrophytes (seaweeds and surfgrass) and sessile invertebrates (barnacles, mussels, sea anemones, and sponges) and the total surface area of each tide pool (Bracken & Nielsen, 2004) (see Table S1). Mobile invertebrates (particularly turban snails) were also counted in each pool. Tide pool volumes, primary producer abundances, and surface areas were measured immediately after our experiment to avoid disturbing the pools prior to water sampling.

Experimental design and sampling

These pools included some of the same tide pools used in other studies of carbonate chemistry under ambient conditions: Kwiatkowski et al. (2016) measured calcification during the spring of 2014 and 2015, and Silbiger & Sorte (2018) quantified bio-physical feedbacks during the summer of 2016. Our CO₂ addition experiments were conducted on two consecutive days: 31 March and 01 April 2016. On those days, the receding tide isolated pools in the morning (08:30–09:30 on day 1, 09:00–10:00 on day 2), and they remained isolated for approximately 7.5 hr. Pools were randomly assigned to experimental treatments: control or +CO₂, with n = 5 pools assigned to each treatment. There were no differences in tidal elevation (t = 1.4, df = 8, p = 0.210), volume (t = 0.6, df = 8, p = 0.570), or surface area (t = 0.8, df = 8, p = 0.453) between control and +CO₂ pools. Treatments were switched between day 1 and day 2, so that pools assigned to +CO₂ treatments on day 1 were assigned to control treatments on day 2, and vice versa. This switching allowed us to assess the effect of CO₂ addition on carbonate chemistry within individual pools by quantifying the difference between day 1 and day 2. We assume that because CO₂ additions were within the natural variability of the system, there were minimal carryover effects of day 1 CO₂ additions to day 2 control tide pools. Average air temperature (day 1: 11.2 ± 0.3 °C; day 2: 11.8 ± 0.4 °C), wind speed (day 1: 14.7 ± 1.3 km hr⁻¹; day 2: 16.8 ± 0.5 km hr⁻¹), and photosynthetically active radiation (day 1: 871±188 µmol m⁻² s⁻¹; day 2: 1,058 ± 173 µmol m⁻² s⁻¹) values, measured at the Bodega Ocean Observing Node weather station adjacent to our study site, did not differ substantially between day 1 and day 2.

CO₂ was delivered to +CO₂ pools using yeast reactors consisting of watertight plastic boxes (Drybox 2500, OtterBox, Fort Collins, Colorado, USA) containing 500 mL of warm water, 2 g of NaHCO₃ (to buffer the internal pH of the reactor), and 0.7 g of baker’s yeast, an amount calculated to generate pH levels and pCO₂ concentrations that were within the predicted range of 21st-century ocean acidification scenarios (Bopp et al., 2013). Tubing from each reactor led to an air stone anchored within each +CO₂ tide pool. CO₂ is the most abundant gas produced by baker’s yeast, representing ∼99.2% of headspace volume (Daoud & Searle, 1990), though additional volatile organic compounds such as ethanol, 2-methylpropanol, and ethyl acetate are present in low concentrations (Smallegange et al., 2010).

We calibrated the yeast reactors and determined the amount of yeast to use in our experiment by running a series of trials in buckets across a range of temperatures and yeast amounts. Buckets (n = 24) contained 12 L of saltwater at typical ocean concentrations (Instant Ocean^® Sea Salt, Blacksburg, Virginia, USA). Temperature was controlled by placing buckets in growth chambers (Percival Scientific, Perry, Iowa, USA) set to 4°, 10°, 20°, or 30 °C. We measured pH (total scale) in each bucket every hr for 6 hr and monitored temperature using TidbiT^® v2 datalogers (Onset Computer Corporation, Bourne, Massachusetts, USA). These trials also represent controls, allowing us to evaluate the effects of yeast and temperature on tide pool pH in the absence of organisms.

Seawater chemistry and other environmental conditions were measured throughout the low tide period on each day. Water samples were collected from pools every 1.5 hr for 6 hr, beginning as soon as each pool was isolated by the receding tide. However, due to nonlinearity in the carbonate parameters across time—potentially due to carbon limitation as pCO₂ declined over time (Maberly, 1990)—we calculated rates of change in pH, pCO₂, net community production, and net ecosystem calcification using only the initial and final samples for each pool, which represented the overall net change over the sampling period. Temperature, dissolved oxygen and conductivity were simultaneously measured in each pool using sensors (ProODO and Pro30, YSI, Yellow Springs, Ohio, USA). Samples were collected by pumping water (400 mL) into a plastic Erlenmeyer flask using a separate piece of tubing anchored in each pool to minimize disturbance and off-gassing. Anchoring the tubing in each pool also ensured that our samples were collected from the same location every time and that they were not collected adjacent to the airstones delivering CO₂ to +CO₂ pools. Each water sample was immediately divided into subsamples for analyses of pH, total alkalinity, and dissolved inorganic nutrients. Prior to sample collection, all containers were washed in 10% HCl, rinsed 3× with DI water, and rinsed 3× with ocean water. We measured pH by measuring voltage (mV) and temperature (°C) of a 50 mL subsample immediately after collection using a multiparameter pH meter (Orion Star, Thermo Fisher Scientific, Waltham, Massachusetts, USA) with a ROSS Ultra glass electrode (Thermo Scientific, USA; accuracy ± 0.2 mV, resolution ± 0.1, drift <0.005 pH units per day) and a traceable digital thermometer (5-077-8, accuracy =0.05°C, resolution = 0.001 °C; Control Company, Friendswood, TX, USA). pH (total scale) was calculated using a multipoint calibration to a Tris standard (Marine Physical Laboratory, Scripps Institution of Oceanography, La Jolla, California, USA), as described in SOP 6a (Dickson, Sabine & Christian, 2007). Subsamples for determination of total alkalinity were fixed with 100 µL of 50% saturated HgCl₂ and stored in 250 mL brown HDPE bottles. Subsamples for dissolved inorganic nutrients (NO${}_3^{-}$, NO${}_2^{-}$, NH${}_4^{+}$, and PO${}_4^{3-}$) were filtered (GF/F, Whatman, Maidstone, UK) into 50 ml centrifuge tubes and frozen at −20 °C prior to analyses.

Sample processing

Subsamples for total alkalinity were analyzed using open-cell titrations (T50, Mettler-Toledo AG, Schwerzenbach, Switzerland), as described in SOP 3b (Dickson, Sabine & Christian, 2007). A certified reference material (Marine Physical Laboratory, Scripps Institution of Oceanography, La Jolla, California, USA) was run daily. Our measurements of the reference material never deviated more than ±0.4% from the certified value, and alkalinity calculations were corrected for these deviations. NO${}_3^{-}$ + NO${}_2^{-}$ and PO${}_4^{3-}$ concentrations (µmol L⁻¹) were measured on a QuickChem 8500 Series Analyzer (Lachat Instruments, Loveland, Colorado, USA), and NH ${}_4^{+}$ concentrations (µmol L⁻¹) were measured using the phenolhypochlorite method (Solórzano, 1969) on a UV-1800 benchtop spectrophotometer (Shimadzu, Carlsbad, California, USA). In situ pH values and other carbonate parameters were calculated using the seacarb package in R v. 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria) (Gattuso et al., 2017) (Table S2). We note that error propagation for calculating Ω_arag based on pH and TA is ∼3.6% (Riebesell et al., 2011).

Calculation of net community production and net ecosystem calcification in control tide pools

Differences in dissolved inorganic carbon (DIC) were used to calculate net community production (NCP) rates (mmol C m⁻² hr⁻¹), as described in Gattuso, Frankignoulle & Smith (1999): ${\displaystyle \mathrm{NCP}=\frac{\Delta \mathrm{DIC}\cdot \rho \cdot \mathrm{V}}{\mathrm{SA}\cdot \mathrm{t}}-\mathrm{NEC}-\mathrm{F}{\mathrm{CO}}_2}$ where ΔDIC is the difference in salinity-normalized DIC between the first and last time points (in this case 0 and 6 hr; mmol kg⁻¹), ρ is the density of seawater (1,023 kg m⁻³), V is the volume of water in the tide pool at each time point, SA is the bottom surface area of each pool (m²), and t is the time interval between sampling points (6 hr).

NEC is the rate of net ecosystem calcification (mmol CaCO₃ m⁻² hr⁻¹), which was calculated as follows: ${\displaystyle \mathrm{NEC}=\frac{\Delta \mathrm{TA}\cdot \rho \cdot \mathrm{V}}{2\cdot \mathrm{SA}\cdot \mathrm{t}}}$ where, in addition to variables defined above, ΔTA/2 is the difference in total alkalinity (TA, mmol kg⁻¹) between the time points. TA was normalized to a constant salinity and corrected for dissolved inorganic nitrogen and phosphorus to account for their small contributions to the acid–base system (Wolf-Gladrow et al., 2007), including the potential for primary producers to change alkalinity via nutrient uptake (Brewer & Goldman, 1976; Stepien, Pfister & Wootton, 2016). One mole of CaCO₃ is formed per two moles of TA, hence the divisor of 2.

Finally, FCO₂, the air-sea flux of CO₂ (mmol m⁻² hr⁻¹), was calculated as follows: ${\displaystyle F{\mathrm{CO}}_2=k\cdot s\cdot \left({\mathrm{CO}}_{2\left[\mathrm{water}\right]}-{\mathrm{CO}}_{2\left[\mathrm{air}\right]}\right)}$ where k is the gas transfer velocity (m hr⁻¹), and s is the solubility of CO₂ in seawater, which was calculated based on in situ measurements of temperature and salinity (Weiss, 1974). The concentration of CO₂ in air was assumed to be 400 µatm (Tans & Keeling, 2017). The transfer velocity of CO₂ was based on wind velocities measured at the Bodega Ocean Observing Node weather station located 100 m from our study location. Calculated FCO₂ values (mmol kg⁻¹ hr⁻¹) were converted to mmol CO₂ m⁻² hr⁻¹ based on the volume of the tide pool, the density of seawater, and the bottom surface area. All data for NCP, NEC, and FCO₂ are provided in the electronic supplementary materials (Table S3).

Statistical analyses

Changes in pH in buckets were evaluated as a function of temperature (^∘C) and yeast (g) using multiple linear regression (PROC GLM) in SAS v. 9.4 (SAS Institute, Cary, North Carolina, USA). Changes in pH and pCO₂ in field tide pools over time (i.e., pH units hr⁻¹ or µatm hr⁻¹) were calculated by subtracting the initial value (0 hr) from the final value (6 hr) and dividing by the elapsed time. The difference between +CO₂ and control pools was evaluated for each individual pool (the experimental unit in all analyses). We quantified the effect of macrophytes (seaweeds and surfgrass) on carbonate parameters by calculating the cover of non-calcifying macrophytes in each tide pool. We excluded calcifying species because of low photosynthetic biomass relative to non-calcifying species from this location (Bracken & Williams, 2013). On average, less than 5% of algal cover in the pools was composed of calcifying seaweeds. We also quantified the effect of invertebrates by estimating total invertebrate cover based on our surveys. For mobile invertebrates (mostly turban snails), we estimated cover from count data based on the number of individuals of each species in a 10 cm × 10 cm quadrat (=0.01 m²). We then evaluated relationships between macrophyte and invertebrate abundances and net community production (NCP) using linear regression (PROC GLM) in SAS. To account for the effects of primary producers on pH and pCO₂, we used linear regression to evaluate the difference between control and +CO₂ tide pools as a function of the abundance of macrophytes in the pools. We calculated these differences by subtracting the rates of change in pH (total scale hr⁻¹) or pCO₂ (µatm hr⁻¹) in each pool under ambient CO₂ conditions from the rates when CO₂ was added to those same pools. Over time, pH increased and pCO₂ decreased due to photosynthesis. Thus, a negative effect of CO₂ addition on pH indicated that CO₂ addition reduced pH in +CO₂ pools relative to control pools. Conversely, a positive effect of CO₂ addition on pCO₂ indicated that CO₂ addition enhanced pCO₂ in +CO₂ pools relative to control pools. We assessed how pH, pCO₂, net ecosystem calcification (NEC), and O₂ responded to changes in NCP under ambient CO₂ conditions using linear regression (PROC GLM in SAS). Similarly, we evaluated whether NEC was related to cover of calcifiers (mussels, turban snails, and coralline algae) using linear regression. Prior to running linear regressions, we verified that data met the assumptions of normality and homogeneity of variances.