Coral Records of Reef-water Ph across the Central Great Barrier Reef, Australia: Assessing the Influence of River Runoff on Inshore Reefs

The boron isotopic (δ 11 B carb) compositions of long-lived Porites coral are used to reconstruct reef-water pH across the central Great Barrier Reef (GBR) and assess the impact of river runoff on inshore reefs. For the period from 1940 to 2009, corals from both inner-and mid-shelf sites exhibit the same overall decrease in δ 11 B carb of 0.086 ± 0.033 ‰ per decade, equivalent to a decline in seawater pH (pH sw) of ∼ 0.017 ± 0.007 pH units per decade. This decline is consistent with the long-term effects of ocean acidification based on estimates of CO 2 uptake by surface waters due to rising atmospheric levels. We also find that, compared to the mid-shelf corals, the δ 11 B carb compositions of inner-shelf corals subject to river discharge events have higher and more variable values, and hence higher inferred pH sw values. These higher δ 11 B carb values of inner-shelf corals are particularly evident during wet years, despite river waters having lower pH. The main effect of river discharge on reef-water carbonate chemistry thus appears to be from reduced aragonite saturation state and higher nutrients driving increased phytoplankton productivity, resulting in the drawdown of pCO 2 and increase in pH sw. Increased primary production therefore has the potential to counter the more transient effects of low-pH river water (pH rw) discharged into near-shore environments. Importantly, however, inshore reefs also show a consistent pattern of sharply declining coral growth that coincides with periods of high river discharge. This occurs despite these reefs having higher pH sw , demonstrating the overriding importance of local reef-water quality and reduced aragonite saturation state on coral reef health.


Introduction
Coral reefs are under threat, not only from the global effects of CO 2 -driven climate change but also from direct local impacts, in particular disturbed river catchments and degraded water quality (McCulloch et al., 2003;Brodie et al., 2010b).Changing land-use practices since the arrival of European settlers to the central Queensland region has produced increased discharge of terrestrial material from the Burdekin River into the Great Barrier Reef (GBR; McCulloch et al., 2003;Lewis et al., 2007).The increase in discharge of terrestrial material into the GBR has resulted in a decrease in inshore water quality, mainly through increased nutrient loads and decreased water clarity (Brodie et al., 2010b;Fabricius et al., 2014).Known impacts include the promotion of intense and extensive phytoplankton blooms and the increase in abundance of macro-algae (Devlin and Brodie, 2005;Brodie et al., 2010b).Changes in water quality within inner-shelf environments of the GBR have also been linked to a decrease in coral calcification (D'Olivo et al., 2013), coral biodiversity (De'ath and Fabricius, 2010), decreased coral cover (Sweatman et al., 2011), and crown-of-thorns starfish outbreaks (Brodie et al., 2005).
Despite the mounting evidence for the negative impacts of increased terrestrial discharge into the GBR, the effect of river flood plumes on the carbonate status of reef waters, a fundamental property controlling calcification, remains largely unknown.It is commonly assumed that, because both the salinity and pH of plume waters (pH pw ) are generally much lower than ocean waters, a decrease in aragonite saturation state ( arag ) might be expected (Salisbury et al., 2008), with consequent negative effects for coastal calcifying organisms such as corals (e.g., Kleypas, 1999;Doney et al., 2009;McCulloch et al., 2012).The effect of lower seawater pH (pH sw ) could, however, be offset by the input of nutrients associated with river plumes, as in semiisolated environments (e.g., enclosed lagoons) and/or highly productive areas where biological processes actively modify the local seawater chemistry (e.g., Hinga, 2002;Andersson et al., 2005;Bates et al., 2010;Drupp et al., 2011;Falter et al., 2013;Duarte et al., 2013).For example, increased productivity during phytoplankton blooms can cause pH sw to rise significantly (Hinga, 2002).Nutrient-enhanced photosynthetic activity has been shown to amplify the seasonal pH cycle by more than 0.5 pH units in experiments within marine enclosures in Narragansett Bay, Rhode Island (Frithsen et al., 1985), and to increase pH sw by 0.7 units in the Peruvian coastal upwelling zone (Simpson and Zirino, 1980).In the GBR, it is not known to what extent terrestrial runoff and the associated phytoplankton blooms influence pH sw and hence coral calcification, partly due to the sparseness of especially longer-term pH sw and other seawater carbonate system records.To overcome this, here we use the skeletons of longlived massive Porites corals as archives of changing environmental conditions.Temporal changes in the boron isotope (δ 11 B) composition of the skeletons provide a time series of seawater pH, which, together with instrumental data, help deconvolve the competing impacts of climate, ocean acidification, and water quality on coral calcification.
The δ 11 B composition of biogenic carbonates (δ 11 B carb ) is an established paleo-proxy for pH sw , first developed in corals by Vengosh et al. (1991) and Hemming and Hanson (1992) and more recently refined by Trotter et al. (2011).The method relies on the preferential incorporation of the isotopically lighter B(OH) − 4 over the B(OH) 3 species into marine carbonate skeletons, with the relative boron species concentration and isotopic compositions being pH dependent (Vengosh et al., 1991;Hemming and Hanson, 1992;Hönisch et al., 2004).Its application to corals has been validated (Hönisch et al., 2004;Reynaud et al., 2004;Trotter et al., 2011) and used for long-term pH sw reconstructions using massive Porites corals (Pelejero et al., 2005;Liu et al., 2009;Wei et al., 2009;Shinjo et al., 2013).Although coral δ 11 B carb compositions closely parallel variations in pH sw (Hönisch et al., 2004;Krief et al., 2010;Trotter et al., 2011;McCulloch et al., 2012), there is a consistent species-specific positive offset of coral δ 11 B carb compositions above the borate δ 11 B value for ambient pH sw (Trotter et al., 2011;McCulloch et al., 2012).This elevation of pH was recently shown to be consistent with the physiological up-regulation of pH at the site of calcification to promote aragonite precipitation (Trotter et al., 2011;McCulloch et al., 2012).Trotter et al. (2011) quantified this internal pH offset and consequently derived ambient seawater values based on the systematic relationships In this study we present annual-resolution δ 11 B carb data obtained from cores of massive Porites heads collected from two inner-shelf and two mid-shelf reefs of the central GBR (Fig. 1).The δ 11 B carb data are used to reconstruct the variability in surface pH sw on annual timescales for the period from 1940 to 2009.These results are then compared with measurements of calcification rates (i.e., linear extension) obtained from the same cores, together with a more extensive database from the central GBR (D'Olivo et al., 2013).This sampling regime enables comparative analysis of the dynamic inner-shelf reef environments, which are subject to terrestrial and anthropogenic influences, with the more stable conditions of mid-shelf reefs that are less exposed to terrestrial runoff and pollutants (Lough, 2001;Furnas, 2003;Brodie et al., 2012).Collectively these coral records of δ 11 B (this study) and linear extension (D'Olivo et al., 2013) provide a unique data set giving insight into the long-term variability of pH sw in a natural coastal system and how these changes interrelate to other important environmental parameters (temperature, river discharge and nutrient flux) and their overall influence on coral calcification.

Instrumental river discharge and sea surface temperature records
Monthly records of river water discharge and pH (pH rw ) for both the Burdekin and Herbert rivers were obtained from the State of Queensland, Department of Environment and Resource Management (DERM; http://watermonitoring. derm. qld.gov.au, 2011).Annual rainfall data are defined from October to September based on the rainfall seasonal pattern in Queensland, with October marking the start of the warmer summer wet season (Lough, 2007(Lough, , 2011)).Monthly sea surface temperature (SST) records were obtained from the Met Office Hadley Centre's sea ice and sea surface temperature data set (HadISST1) centered at 18 • S and 147 • E with a spatial resolution of 1

Water samples
Water samples were collected in order to characterize the δ 11 B composition of the plume waters (δ 11 B pw ) present in the inner-shelf area during flood events, and hence facilitate the interpretation of the δ 11 B carb coral signal.In February 2007, a total of 29 water samples were collected (by Stephen Lewis, James Cook University) along a northward transect from the mouth of the Burdekin River to Magnetic Island (Fig. 1).In February 2009, a second suite of water samples were collected (by J. P. D'Olivo) from the Burdekin River bridge, located between the towns of Ayr and Home Hill, as well as seven seawater samples along a northward transect from the mouth of the Burdekin River to Pandora Reef.The presence of discolored water indicated that the river plume had reached all of the sampling sites.After collection, 125 ml of each sample was filtered through a 0.45 µm Teflon membrane and then acidified using 2-3 drops of ∼ 7 M HNO 3 .Samples were then stored in acid-cleaned, low-density polyethylene bottles in a cool room.Salinity data were provided by S. Lewis (2009).Water samples were analyzed for B by solution quadrupole ICP-MS using a Varian 820 ICP-MS at the Research School of Earth Sciences at the Australian National University (ANU). 10Be was used as an internal standard spiked at a concentration of 4 ppb.Water samples were diluted to a final salinity of 0.035.

Sclerochronology and coral sampling for δ 11 B analysis
Five cores were drilled from massive Porites coral heads from four sites in the central GBR (Fig. 1): Pandora Reef (PAN02) and Havannah Island (HAV06A and HAV09_3) in the inner shelf, and Rib Reef (RIB09_3) and Reef 17-065 (1709_6) in the mid-shelf.All cores exhibit clear and regular annual growth banding.Subsamples representing annual growth increments, from the beginning of each high-density band (for mid-shelf corals) or luminescent band (for innershelf corals), were milled along the axis of maximum growth of 0.7 cm thick slabs from each core.

Boron isotope methodology
The boron isotopic compositions of water samples and annual coral subsamples of cores HAV06A, PAN02, and RIB03 were analyzed at the ANU by positive thermal ionization mass spectrometry (PTIMS) using a Thermo Finnigan TRI-TON.Annual coral subsamples of cores HAV09_3 and 1709_6 were analyzed at the University of Western Australia (UWA) by multi-collector inductively coupled plasma mass spectrometry (MC-ICPMS) using a NU Plasma II.Prior to analysis, samples were cleaned in H 2 O 2 (for PTIMS) or in NaOCl (for MC-ICPMS) to remove organic material and then dissolved in HNO 3 , after which the boron was purified using ion chromatography.The boron separation technique used for the PTIMS is based on methodology by Wei et al. (2009) and refined by Trotter et al. (2011).For MC-ICPMS, a combined cation-anion ion-exchange technique was employed as described by McCulloch et al. (2014).
In both methods, the boron is collected in relatively large fractions, ensuring 100 % collection efficiency.The δ 11 B composition of water samples was analyzed by PTIMS following a simplified purification procedure (cf.coral samples) that omitted the H 2 O 2 cleaning step and employed a single AGW50-X8 cation column elution followed by an IRA743 column elution (i.e., the final cation column was omitted).The amount of water sample required to extract and purify 1 µg of B was estimated from the relationship between the measured [B] and salinity (S) in the flood plume B = 0.1299(S) + 0.1188 and B = 0.1302(S) − 0.0374; 2007 and 2009, respectively (Fig. 2).The amount of water subjected to the boron extraction and purification procedure varied from 250 µl (S = 35) to 5000 µL (S = 0.7), while 30 000 µl was processed for the river water sample (S = 0).All boron isotopic ratios are expressed in the conventional delta notation (δ) relative to the NBS951 boric acid international standard.Over the course of this study, PTIMS analysis of the SRM951 standard yielded a mean value of δ 11 B = +0.05‰ relative to a reference value for this standard of 4.054, with an external precision (2 SD) of ±0.35 ‰ (n = 43) and an internal precision of ±0.07 ‰ .Repeated analyses of the samples (n = 46) gave an aver- age reproducibility of ±0.20 ‰ (2 SD).A modern coral from Papua New Guinea (NEP B) was used as a secondary working standard.Repeated NEP B analyses gave an average value of 26.35 ± 0.44 ‰ (2 SD, n = 33) using PTIMS, and 25.96 ± 0.32 ‰ (2 SD, n = 70) using MC-ICPMS.The δ 11 B carb values of coral samples (n = 8) analyzed by PTIMS are consistently offset by +0.45 ‰ relative to the MC-ICPMS data.To maintain consistency between the two data sets, a +0.45 ‰ correction was therefore applied to the data measured by MC-ICPMS.Measurements of the international carbonate standards JCp-1 by MC-ICPMS gave a δ 11 B value of 24.35 ± 0.34 ‰ (2 SD), identical to the 24.33 ± 0.11 ‰ (SE) reported by Foster et al. (2013) and 24.22 ± 0.28 ‰ (2 SD) reported by Wang et al. (2010).
Conversion of δ 11 B carb to pH of the calcifying fluid (pH cf ) values was undertaken using the relationship pH cf = pK B − log (1) , where δ 11 B sw is the B isotope composition of seawater (δ 11 B sw = 39.61 ‰ ; Foster et al., 2010) and the B isotope fractionation factor (α B3-B4 ) of 1.0272 is taken from Klochko et al. (2006).The B dissociation constant (pK B ) was adjusted to the ambient temperature and salinity (Trotter et al., 2011), the latter especially relevant for corals from the innershelf reef region, which is diluted by fresh water during flood events.Seasonal variation in salinity was estimated based on the linear relationship observed between the magnitude of past flood events and corresponding salinity values reported by King et al. (2001) and Walker (1981) for these reefs (Supplement Fig. S1).
Average annual salinity values were estimated using the equation given in Fig. S1.For each year, a dilution factor derived from a 2-month period of maximum river discharge was applied to an initial seawater value of 35.5.Despite limitations with this approach, as flood events are spatially and temporally variable (King et al., 2001;Furnas, 2003), the effect after correcting the pK B for salinity and temperature on the estimated pH sw value is small (<∼ 0.01 pH units).
The external pH sw value was estimated following the method for Porites sp. as described by Trotter et al. (2011): pH sw = (pH cf − 5.954)/0.32. (2) All measured and reconstructed pH sw values are reported relative to the total pH scale.River water collected from the Burdekin River bridge (23 km upstream from the river mouth) during the 2009 flood event had a δ 11 B pw value of +28 ‰.Despite the large difference in the δ 11 B pw values of waters near the river mouth in the two flood events, δ 11 B pw of samples taken close to the innershelf reefs are very similar, with an average δ 11 B value of 39.80 ± 0.34 ‰ (2 SD; n = 2).This value is identical to seawater samples from Lady Musgrave Island in the southern GBR (δ 11 B sw = 40.09± 0.37 ‰; 2 SD; n = 2), and is consistent with previously reported seawater values (Foster et al., 2010).

Coral δ 11 B records
Coral δ 11 B carb compositions for fives cores give average values that range from 23.6 ± 0.37 to 25.2 ± 0.57 ‰ over the common period of 1973 to 2002 (Table 2).These values are consistent with previously reported values for Porites corals (Hönisch et al., 2004;Pelejero et al., 2005;Wei et al., 2009;Krief et al., 2010).Using Eqs. ( 1) and (2), these δ 11 B values translate to pH sw values of 7.88 ± 0.07 to 8.19 ± 0.11 (Table 2).The mid-shelf corals have slightly lower δ 11 B carb (and pH) values that are less variable (±0.37 to ±0.44 ‰; 2 SD) than the inner-shelf corals (±0.50 to ±0.66 ‰; 2 SD).Differences in δ 11 B carb compositions between coral cores are significant (Kruskal-Wallis one-way ANOVA on ranks).Pairwise multiple comparison procedures (Tukey test; p < 0.05) indicate that these differences are significant between the three inner-shelf cores and the mid-shelf core 1709_6; midshelf core RIB09_3 is only significantly different from inner- shelf core HAV06A.No significant correlation was found between the different δ 11 B carb coral records.Linear regressions applied to the reconstructed annual pH sw time series for all five coral records show a decrease in time over the full length of each coral records (Fig. 3).The decrease in pH sw is equivalent to −0.017 ± 0.008 units per decade for the inner-shelf corals and −0.018 ± 0.007 units per decade for the mid-shelf corals.These rates are consistent with global estimates of 0.017 to 0.019 pH units decrease per decade from 1984 to 2011 based on instrumental data (Dore et al., 2009;Santana-Casiano et al., 2007;Bates et al., 2012), but are lower than the 0.041 ± 0.017 pH unit decrease per decade for the period of 1940 to 2004 shown by the δ 11 B carb coral records from Arlington Reef (mid-reef) in the central GBR previously reported by Wei et al. (2009).However, the Arlington Reef core exhibited a marked decrease in δ 11 B carb composition in association with the effects of the severe 1998 bleaching event.
Composite records for δ 11 B carb were obtained for the inner-shelf and mid-shelf corals (Fig. 4) for the periods of 1963 to 2005 and 1973 to 2009, respectively.These are the periods common to more than one core for each region.The data from each core were first normalized according to the following equation: where x t is a δ 11 B carb value at a certain point in time, x c is the mean δ 11 B carb value for a given coral.The composite records were then obtained by calculating the average from the normalized data at a certain point in time.To preserve the units, the composite records were "rescaled" according to the equation where Z t is a composite δ 11 B carb value at a certain point in time and x R is the average mean value for all the cores from a specific region.The δ 11 B carb annual composite for innershelf and mid-shelf records are plotted relative to time and compared with annual discharge from the Burdekin River, SST from HadISST1, and coral linear extension rates from D' Olivo et al. (2013;Fig. 4).The composite δ 11 B carb records for inner-shelf corals show significant correlation with river discharge, SST, and coral extension rates (Table 3).The mid- shelf records shows a weaker but still significant correlation with SST.
The 8-year low-pass filter applied to the composite reconstructed pH sw data derived from the three inner-shelf δ 11 B carb coral records reveals semi-decadal variation (Fig. 5), which is not observed in the mid-shelf δ 11 B carb data.Similar variability is observed in inner-shelf coral extension rates and calcification when periods of slower growth coincide with higher terrestrial runoff (D'Olivo et al., 2013).For innershelf corals it follows that higher δ 11 B carb (higher pH) signals coincide with periods of increased river discharge and slower extension rates.

Variations in river and seawater δ 11 B during flood events
The data from DERM (2011) show that the Burdekin River have an average pH rw of 7.58 ± 0.46 (2 SD; Fig. S2); this indicates that low-pH fresh water, relative to pH sw , is introduced into the coastal area of the GBR during wet periods.
The pH rw shows no significant difference between summer (7.56 ± 0.41; 2 SD) and winter (7.65 ± 0.35; 2 SD) values, with a decrease in pH during some high-discharge events, but there is no consistent seasonal pattern.The variability in pH rw indicates that factors other than the amount of discharge or rainfall determine the pH rw , and may reflect variability that is related to the nature of the material being car-ried by the river and the catchment supplying the water.The δ 11 B pw near the mouth of the Burdekin River can vary significantly between different flood events (Fig. 2).Given that the δ 11 B rw is likely influenced by various factors, including both the type and amount of terrigenous material carried by the river during flood events as well as the source of the river runoff and nature of the catchment, more work is needed to characterize B dynamics and isotope fractionation mechanisms during flood events.Although the δ 11 B pw values during 2007 and 2009 show a large variation close to the Burdekin River mouth, the δ 11 B pw values near the inner-shelf reefs are typical of ocean waters (i.e., δ 11 B ∼ 40 ‰).

Origin of interannual δ 11 B carb variability in corals
The results shown in Fig. 4 and corresponding correlations suggest a relationship between coral δ 11 B carb and both ambient seawater temperature and terrestrial runoff, particularly for the inner-shelf region of the GBR.However, the interannual pH sw variability of less than ±0.01 pH units, which is directly attributable to temperature and salinity changes due to river runoff, contrasts with the interannual variability of more than ±0.07 pH units reconstructed from the δ 11 B carb compositions for the inner-shelf and mid-shelf corals.Possible explanations for the interannual variability observed in δ 11 B carb (reconstructed pH sw ) are examined below.

Boron adsorption onto clays and sediments
The interannual δ 11 B carb variations for inner-shelf corals could be explained by the adsorption of B onto sediments and clays that are delivered to the inner-shelf region by rivers.Clays preferentially remove the lighter isotope 10 B from seawater (Palmer et al., 1987;Barth, 1998).This selective removal results in the respective depletion of 11 B in marine clays but enrichment in seawater, and is the accepted explanation for the heavy isotopic composition of seawater (e.g., 39.61 ± 0.04 ‰; Foster et al., 2010) relative to average continental crust (Spivack and Edmond, 1986;Palmer et al., 1987;Barth, 1998).Given the large silt and clay wash load transported from the Burdekin River (Belperio, 1979), fractionation of δ 11 B between the dissolved and adsorbed B phases could have a significant effect on the δ 11 B of seawater.However, the conservative mixing of B along the transect (Fig. 2) indicates that boron is not being quantitatively removed from the plume waters by clays, and that the clay material is already in equilibrium with the river water before entering the ocean.Similar results have been reported by Barth (1998) and Xiao et al. (2007).Furthermore, the oceanic δ 11 B pw values near the reefs during flood events and the low B concentration of river waters require that, at the reef sites, the δ 11 B signal be dominated by seawater.Finally, large δ 11 B carb interannual variations occur on the mid-shelf reefs where there is no clay-dominated terrestrial runoff.

Effect of nutrient enrichment and biological productivity on pH sw
River discharge is an important source of particulate and dissolved nutrients as well as sediments to the inner-shelf area of the GBR (King et al., 2002;Devlin and Brodie, 2005), being responsible for ∼ 90 % of the particulate and dissolved nutrients introduced during flood events (Mitchell and Bramley, 1997;Furnas, 2003).Most particulate matter and sediments are deposited within a few kilometers (∼ 10 km) of river mouths where salinity is <10 (Wolanski and Jones, 1981), whereas dissolved nutrients are carried greater distances of up to ∼ 200 km along the coast (Devlin and Brodie, 2005).Once the turbidity decreases and low light levels are no longer limiting, the dissolved nutrients are rapidly taken up by primary producers, resulting in phytoplankton blooms (Furnas, 2003;Devlin and Brodie, 2005;Brodie et al., 2010b (Devlin and Brodie, 2005), which is where the inner-shelf coral reefs of this study are located.
If we consider NO − 3 as the main form of nitrogen sourced from the Burdekin River into the inner-shelf area of the GBR (Furnas, 2003), plankton productivity will result in the decrease in seawater CO 2 , increase in alkalinity, and uptake of H + (Gattuso et al., 1999).The strong coupling between CO 2 dynamics and large phytoplankton bloom events observed at Kane'ohe Bay, Hawaii, changes the reef system from being a source of CO 2 to a sink of CO 2 (Drupp et al., 2011).At Kane'ohe Bay, bloom events are fueled by nutrient inputs following rainfall and terrestrial runoff events.The enhanced productivity is reflected by increased phytoplankton biomass from ∼ 2 to ∼ 6 µg L −1 Chl a, which draws down pCO 2 by ∼ 100 ppm (Drupp et al., 2011).Similar large nutrient-fueled changes in phytoplankton biomass occur in the central GBR, where Chl a increases from between 0.3 and 0.7 µg L −1 Chl a (Brodie et al., 2007) to up 1 to 20 µg L −1 Chl a within flood plumes (Devlin and Brodie, 2005;Brodie et al., 2010a).For rivers in the central GBR, including the Burdekin, typical dissolved inorganic nitrogen (DIN) (chiefly NO − 3 ) concentrations during flood conditions vary between 20 and 70 µM, while dissolved inorganic phosphorus (DIP) varies between 0.15 and 1.3 µM (Furnas, 2003).These are much higher than typical seawater concentrations of 0.02 µM for DIP and 0.03 µM for NO − 3 ) during non-flood periods in the inshore region of the central GBR (Furnas et al., 2011).
The uptake of CO 2 from nutrient-enhanced productivity and the associated changes to the carbonate parameters during flood events in the central GBR were estimated using Redfield ratios and river flood plume nutrient loads (Fig. 6).For these calculations we assumed a conservative dilution process (Wooldridge et al., 2006).Three scenarios are included.In the first scenario only the changes due to the mixing of plume waters with seawater are considered (Fig. 6a).In the second scenario the input of nutrients from plume waters is included, with N assumed as the limiting nutrient (Fig. 6b).This scenario is consistent with observations that N appears to be the dominant control for new phytoplankton biomass formation in the GBR (Furnas et al., 2005).For the final scenario P is considered the limiting nutrient (Fig. 6c).When the effect of nutrient-enhanced productivity is not included (Fig. 6a), there is a small increase in pH sw of up to 0.03 units under the most extreme flood conditions (e.g., salinity 24 near the reef).The scenarios with nutrient input from floodwaters in Fig. 6b and c result in a more marked increase in pH sw due to the drawdown of pCO 2 associated with nutrient-enhanced productivity.In the case of the scenario with a salinity of 24 and DIN concentration of 23 µM, there is a drawdown of pCO 2 of up to 300 ppm and a corresponding increase in pH sw of ∼ 0.33 units (Fig. 6b).It follows that the higher δ 11 B carb (and reconstructed pH sw ) values from the inner-shelf reefs during periods of high river discharge and high nutrient input are consistent with decreased pCO 2 and increased pH sw due to the stimulation of phytoplankton production.
In addition to the changes in pH sw , there is a decrease in arag , under the flood scenario with no nutrient input (Fig. 6a), which is associated with the low TA and dissolved inorganic carbon (DIC) content of the floodwaters.The decrease in arag can be of up to 1 unit under extreme flood conditions (e.g., salinity 24).When nutrient-enhanced productivity is included, the negative effect from the plume waters on the arag is reduced (compare Fig. 6a with b and c) as the CO 2 uptake by the phytoplankton results in an increase in CO 2− 3 .Under the scenario with the highest DIN concentration (23 µM), the effect from the enhanced productive is sufficient to counteract the dilution effect in arag , resulting in a net increase of up to 0.6 units (Fig. b).However, the reduction in arag that is expected under most flood conditions is likely to negatively affect coral calcification (McCulloch et al., 2012).One limitation for the present model is that it does not allow for air-sea equilibration, which is likely to reduce the effects from the enhanced productivity on the carbonate parameters.

Cross-shelf differences in coral δ 11 B carb and reconstructed pH sw
A systematic cross-shelf pattern is observed in δ 11 B carb (and reconstructed pH sw ), with both average values and interannual variability being greater for inner-shelf corals compared to mid-shelf corals (Table 2).This cross-shelf pattern is consistent with the available, yet limited, pH sw data that comprise summer (2007) values of 8.14 ± 0.003 (Gagliano et al., 2010) measured at Magnetic Island in the inner-shelf area, and summer 2012 values of 8.03 ± 0.03 measured at Davies Reef (Albright et al., 2013).It is not surprising to find differences in reconstructed pH sw values between reefs given that the GBR is characterized by significant spatial gradients in seawater parameters (e.g., temperature and water quality), especially across different shelfal environments (D'Olivo et al., 2013;Cantin and Lough, 2014;Fabricius et al., 2014).These spatial differences could, for example, account for the lower interannual variation of the coral δ 11 B carb records from the mid-shelf reefs compared to inner-shelf reefs through the decreased influence of river runoff on mid-shelf reefs as previously suggested by Wei et al. (2009).
The reason for the lack of correlation between the δ 11 B carb records at nearby reefs, specifically between Pandora and Havannah Island, is unclear.Differences in biologically driven effects (e.g., species or gender related) or local variability in environmental parameters (e.g., light regime) are possible explanations.Nevertheless, the overall good agreement between the composite δ 11 B carb coral records and the terrestrial runoff indices is encouraging, suggesting that multi-core replication (Lough, 2004;Jones et al., 2009) and consideration of ambient environmental conditions are es-1231 Figure 6.(a) Estimated seasonal pH sw and arag changes based on in situ SST (AIMS, 2011) with an average pCO 2 of 431 ppm and a seasonal variation of 60 ppm (Uthicke et al., 2014).Seawater pH values were estimated considering a seawater end member with summer total alkalinity (TA) of 2270 mmol kg sw −1 and winter TA of 2256 mmol kg sw −1 (Uthicke et al., 2014); dissolved inorganic carbon (DIC) changes were estimated from TA and pCO 2 .The grey lines define the range of historic maximum and minimum SST values for any given month.Shown here are changes in pH sw and arag resulting from a hypothetical flood event characterized by a decrease in salinity from 35 to values of 24 or 30 in February and increasing to 33 in March.For flood conditions the river water end member was estimated considering TA of 787.7 mmol kg sw −1 and DIC of 811.1 mmol kg sw  2 is included.The minimum and maximum slopes obtained for the corals in Table 2 are included as confidence intervals.Calculations were made using CO 2 SYS with carbonate constants K1 and K2 from Merhbach et al. (1973) refitted by Dickson and Millero (1987), and for sulfate from Dickson (1990) with 0 dbar pressure.sential when interpreting δ 11 B carb records.Further analyses of additional records from the same area should help clarify the uncertainties and improve our understanding of the δ 11 B carb seawater proxy in dynamic reefal systems that characterize inshore environments.A multi-proxy approach should also prove helpful in confirming the results from this study, as well as determining the response of corals to specific environmental parameters such as SST, nutrients, sediment, or pH sw .

Relationship between δ 11 B carb (reconstructed pH sw ) and coral growth rates
Calcification and linear extension rates of inner-shelf corals from the GBR show a long-term decrease from the period of 1930 to 2008 (Lough, 2008;D'Olivo et al., 2013).The decrease in coral growth has been attributed to factors ranging from thermal stress, bleaching, eutrophication, and ocean acidification (Cooper et al., 2008;Lough, 2008;De'ath et al., 2009;D'Olivo et al., 2013).The present study reveals that decadal-scale wet periods with increased terrestrial runoff in the central GBR coincide with periods of reduced inner-shelf coral growth (Fig. 5).The decrease in coral growth occurs despite higher ambient pH sw , as determined from δ 11 B carb .
Given the apparently more favorable pH sw for coral growth during wet periods, other factors such as degraded water quality or reduced arag must clearly be responsible for the decline in extension rates observed during wet periods.For example, nutrient-fueled increase in phytoplankton biomass has been a longstanding explanation for the drowning of coral reefs throughout the geological records (Hallock and Schlager, 1986).More in situ monitoring of seawater carbonate system parameters (e.g., Uthicke et al., 2014), especially during wet periods, would greatly help in understanding the response of the complex inner-shelf systems to changes in water quality.

Physiological controls on coral calcification
The coral pH cf values calculated from the measured δ 11 B carb compositions (Table 2) indicate that corals elevate the pH at the site of calcification, in agreement with previous studies (Al-Horani et al., 2003;Venn et al., 2011;McCulloch et al., 2012).This elevation of pH cf for massive Porites grown in the natural environment is estimated to be 0.41 ± 0.02 for the inner-shelf corals and 0.43 ± 0.03 for the mid-shelf corals.The latter values were estimated using the coral pH cf values in Table 2 and the directly measured pH sw summer value of 8.14 from Gagliano et al. (2010) as an independent reference value for the inner-shelf region, and an average pH sw annual value of 8.06 from Albright et al. (2013) for the mid-shelf region.
Aside from pH up-regulation at the calcification site (Mc-Culloch et al., 2012), other processes involved in promoting aragonite precipitation include the transport of ions to the mineralization site and the synthesis of an organic matrix (Allemand et al., 2004;Venn et al., 2011).Inhibition or reduced activity of these processes has been associated with significant reduction of calcification in other studies (Tambutte et al., 1996;Allemand et al., 1998Allemand et al., , 2004;;Al-Horani et al., 2003).The reduction in growth of inner-shelf coral may thus be explained by the effects of river discharge that likely include a decrease in arag that is accompanied by an increase in shading, turbidity, sedimentation, or competition for carbon from the photosynthetic activity of zooxanthellae.These factors can affect the availability of DIC, enzyme activity, or synthesis of the organic matrix involved in the calcification process (Tambutte et al., 1996;Allemand et al., 1998Allemand et al., , 2004;;Al-Horani et al., 2003), because energy and DIC required by these processes is reallocated into cleaning or mucus production (Riegl and Branch, 1995;Telesnicki and Goldberg, 1995;Philipp and Fabricius, 2003).Therefore, it should not be surprising to find reduced growth during wet periods and hence degraded water quality despite conditions of higher pH.Critically, this demonstrates the overriding importance of local reef-water quality relative to the subordinate longer-term effects of ocean acidification.
In contrast to the inner-shelf reefs, the δ 11 B carb records of the mid-shelf corals shows no significant relationship to river discharge, consistent with the reduced effects of river flood plumes.Extreme flood events can, however, occasionally reach the mid-shelf reefs, especially when offshore winds occur (King et al., 2002).The presence of subdued luminescent bands in coral records from Rib Reef that coincide with some large river discharge events confirms the minor effect of flood events in the mid-shelf region.Nevertheless, coral linear extension and calcification at both mid-and outer-shelf reefs increase over the last ∼ 50 years, coincident with the rise in temperature over this period (D'Olivo et al., 2013).This observed increase in coral calcification, despite the decrease in the reconstructed pH sw , indicates that ocean acidification has so far played a secondary role in impacting coral calcification in the mid-shelf region.

Summary and conclusions
Coral δ 11 B carb values show cross-shelf variability with higher average and amplitude values characteristic of the corals closer to the coast.The reconstructed pH sw values calculated from the coral δ 11 B carb indicate that, in their natural environment, massive Porites up-regulate pH cf by ∼ 0.4 units.Variability in coral δ 11 B carb shows an interannual range in reconstructed pH sw from ∼ 0.07 pH units in mid-shelf corals to ∼ 0.11 pH units in inner-shelf corals, compared to a much smaller long-term (1940 to 2009) trend of ∼ 0.017 pH unit decrease per decade.This rate of change is consistent with previous estimates of decreasing surface pH sw from pCO 2 -driven ocean acidification.Results from δ 11 B carb and coral growth indicate that terrestrial runoff has a significant 1233 effect on inner-shelf reef environments.We propose that phytoplankton blooms, fueled by increased nutrient inputs from river plume waters, drive the drawdown of dissolved CO 2 and thus increase the pH in surface seawaters.Consequently, on a local scale the inner-shelf reefs of the GBR exhibit high rates of nutrient-driven production, and following river discharge events temporarily counter the effects of ocean acidification.Despite the higher pH sw , we observe an associated decrease in coral linear extension and calcification (D'Olivo et al., 2013), consistent with expectations of a coral calcification decrease with decreasing arag (McCulloch et al., 2012).The incompatible relationship of higher pH sw and decreased coral growth suggests that the effects of large flood events on lowering arag and degrading water quality (e.g., increased shading, turbidity, sedimentation, or competition for carbon by up-regulated photosynthetic activity of zooxanthellae) are the dominant cause of reduced coral growth.This study demonstrates the value of coral δ 11 B as a paleo-proxy for reconstructing past pH sw changes, as well as the importance of disentangling the effects of changing local water quality from ocean acidification and global warming, which is relevant not only to the inner-shelf region of the GBR but also other coral systems worldwide.
The Supplement related to this article is available online at doi:10.5194/bg-12-1223-2015-supplement.

Figure 1 .
Figure 1.Map of the central area of the GBR showing locations of water sample and coral core collection sites: Pandora Reef and Havannah Island in the inner shelf, Rib Reef and 17-065 Reef in the mid-shelf region, and water sample collection sites during the flood events of 2007 and 2009.

Figure 2 .
Figure 2. (a) Boron concentration plotted against salinity of waters from the flood events of 2007 and 2009.A linear regression through the data is compared to the theoretical conservative mixing relationship based on a seawater end member with 4.52 mg B / L at S = 35.(b) Boron isotope composition of waters along salinity transects from the 2007 and 2009 flood events.
, δ 11 B ratios, and pH during flood events Measured [B] and δ 11 B pw are plotted against salinity of the water samples collected during the flood events of 2007 and 2009 (Fig. 2).Salinity ranges from 0 at the river mouth to 30 at Havannah Island and 26.5 at Pandora Reef.Salinity and [B] concentration during the flood events of 2007 and 2009 show a linear relationship (r 2 = 0.9834; p < 0.0001; n = 10 for 2007 and r 2 = 0.9915; p < 0.0001; n = 8 for 2009), consistent with conservative mixing of boron between seawater and floodwaters.The flood events sampled in 2007 and 2009 show large differences in the δ 11 B pw composition of the low-salinity floodwaters collected close to the river mouth (Fig.2).The low δ 11 B pw (+14.13 ‰) measured during the larger 2009 flood event (29 TL) contrasts with the less variable and higher δ 11 B pw (+42.8 ‰) of the smaller 2007 flood event (10 TL).

Figure 3 .
Figure 3. Annual time series and linear regressions for δ 11 B carb values and corresponding pH cf of coral cores from (a) the innershelf reefs of Havannah Island and Pandora Reef, and (b) the midshelf reefs of Rib Reef and 17-065.

Figure 4 .
Figure 4. Annual composite δ 11 B carb records for the inner-shelf corals and mid-shelf corals compared to annual discharge from the Burdekin River (a), SST from HadISST1 (b), and coral linear extension rates from D'Olivo et al. (2013) for the corresponding innershelf or mid-shelf regions (c).Note that the temperature and linear extension axis has been reversed to facilitate comparisons.

Figure 5 .
Figure 5. Normalized annual (grey) and smoothed (colored 8-year low-pass filter) time series for (a) the inner-shelf composite reconstructed pH sw obtained from δ 11 B carb coral records, (b) averaged linear extension of inner shelf in the central GBR (D'Olivo et al., 2013), terrestrial influx indicated by (c) discharge from the Burdekin River, and (d) coral Ba/Ca data from Pandora and Havannah Island (McCulloch et al., 2003).
Figure 6.(a) Estimated seasonal pH sw and arag changes based on in situ SST(AIMS, 2011)  with an average pCO 2 of 431 ppm and a seasonal variation of 60 ppm(Uthicke et al., 2014).Seawater pH values were estimated considering a seawater end member with summer total alkalinity (TA) of 2270 mmol kg sw −1 and winter TA of 2256 mmol kg sw −1(Uthicke et al., 2014); dissolved inorganic carbon (DIC) changes were estimated from TA and pCO 2 .The grey lines define the range of historic maximum and minimum SST values for any given month.Shown here are changes in pH sw and arag resulting from a hypothetical flood event characterized by a decrease in salinity from 35 to values of 24 or 30 in February and increasing to 33 in March.For flood conditions the river water end member was estimated considering TA of 787.7 mmol kg sw −1 and DIC of 811.1 mmol kg sw −1 (DERM, 2014).(b) Estimated seasonal pH sw and arag changes as in (a), including DIC changes associated with the CO 2 drawdown due to enhanced phytoplankton production during flood events based on DIN river inputs of 20 and 70 µM in February and assuming conservative dilution.Nutrient inputs are estimated to reduce by half in March and back to seawater levels by April.(c) Estimated seasonal pH sw and arag changes as in (b) considering DIP river inputs of 0.15 and 1.3 µM in February and assuming conservative dilution.(d) Estimated annual pH sw changes with time based on reconstructed SST (HadISST1), changes in atmospheric CO 2 (NOAA), and salinity changes associated with flood events obtained from the equation in Fig. S1.A comparison based on the average pH sw slope obtained from all the coral δ 11 B carb data in Table2is included.The minimum and maximum slopes obtained for the corals in Table2are included as confidence intervals.Calculations were made using CO 2 SYS with carbonate constants K1 and K2 fromMerhbach et al. (1973) refitted byDickson and Millero (1987), and for sulfate fromDickson (1990) with 0 dbar pressure.
et al.: Coral records of reef-water pH across the central Great Barrier Reef

Table 1 .
Periods covered by coral core samples analyzed.

Table 2 .
Average values and variability (2 SD) for coral δ 11 B carb , pH cf , and reconstructed pH sw calculated over the common period of 1973-2002.The reconstructed pH sw values were estimated using Eq.(2) to correct for the pH offset at the site of calcification.Slopes were obtained from the linear regression of the full length of each core (see Table1) and are used to indicate the annual rate of change for the reconstructed pH sw records.The uncertainty for the slopes is based on the external reproducibility for the standard.

Table 3 .
Olivo et al. (2013) coefficients (r) and corresponding p values for correlations of annual composite coral δ 11 B carb records from the inner shelfand mid-shelfwith Burdekin River runoff, SST from HadISST1, and coral linear extension rates for the corresponding region taken from D'Olivo et al. (2013).