Benthic alkalinity and DIC fluxes in the Rhône River prodelta generated by decoupled aerobic and anaerobic processes

Estuarine regions are generally considered a net source of atmospheric CO 2 as a result of the high 2 organic carbon (OC) mineralization rates in the water column and their sediments. Yet, the intensity 3 of anaerobic respiration processes in the sediments tempered by the reoxidation of reduced 4 metabolites controls the net production of alkalinity from sediments that may partially buffer the 5 metabolic CO 2 generated by OC respiration. In this study, a benthic chamber was deployed in the 6 Rhône River prodelta and the adjacent continental shelf (Gulf of Lions, NW Mediterranean) to assess 7 the fluxes of total alkalinity (TA) and dissolved inorganic carbon (DIC) from the sediment. 8 Concurrently, in situ O 2 and pH microprofiles, electrochemical profiles, pore water and solid 9 composition were measured in surface sediments to identify the main biogeochemical processes 10 controlling the net production of alkalinity in these sediments. The benthic fluxes of TA and DIC, 11 ranging between 14 and 74 mmol m -2 d -1 and 18 and 78 mmol m -2 d -1 , respectively, were up to 8 times 12 higher than the DOU fluxes (10.4 ± 0.9 mmol m -2 d -1 ) close to the river mouth, but their intensity 13 decreased offshore, as a result of the decline in OC inputs. Low nitrate concentrations and strong pore 14 water sulfate gradients indicated that the majority of the TA and DIC was produced by sulfate and 15 iron reduction. Despite the complete removal of sulfate from the pore waters, dissolved sulfide 16 concentrations were low due to the precipitation and burial of iron sulfide minerals (12.5 mmol m -2 17 d -1 near the river mouth), while soluble organic-Fe(III) complexes were concurrently found 18 throughout the sediment column. The presence of organic-Fe(III) complexes together with low sulfide 19 concentrations and high sulfate consumption suggests a dynamic system driven by the variability of the organic and inorganic particulate input originating from the river. By preventing reduced substances from being reoxidized, the precipitation and burial of iron sulfide decouples the iron and 22 sulfur cycles from oxygen, therefore allowing a flux of alkalinity out of the sediments. In these conditions, carbonate with associated determination

sulfur cycles from oxygen, therefore allowing a flux of alkalinity out of the sediments. In these 23 conditions, the sediment provides a source of alkalinity to the bottom waters which mitigates the 24 effect of the benthic DIC flux on the carbonate chemistry of coastal waters. 25

Introduction 27
As a link between continental and marine environments, the coastal ocean plays a key role in the 28 global carbon cycle (Bauer et al., 2013). In particular, large fluxes of dissolved and particulate organic 29 carbon (POC) are delivered by rivers to neighbouring continental shelves (Bianchi and Allison, 2009). 30 In fact, even though shelf regions only occupy around 7 % of the global ocean surface area (Jahnke,31 2010), they account for more than 40 % of POC burial in the oceans of which about half is buried in 32 river deltas and estuaries (Hedges and Keil, 1995 (Chen andBorges, 2009, Cai, 2011). In these river-dominated 39 margins, high sedimentation rates of material containing large concentrations of POC decrease the 40 residence time of organic carbon in the oxic sediment layers (Hartnett et al., 1998) and increase the 41 relative contribution of anaerobic compared to aerobic degradation pathways of organic carbon 42 (Canfield et al., 1993a). Anaerobic respiration processes, including denitrification, dissimilatory 43 nitrate reduction to ammonium (DNRA), manganese reduction, iron reduction, and sulfate reduction 44 create total alkalinity (TA) (Berner, 1970;Dickson, 1981 Burdige, 2011). In turn, the precipitation of carbonate species, such as calcite and aragonite, 49 with 1cm diameter corers made of cut 10-ml syringes every 5 cm through pre-drilled holes. The 193 content of these subsamples was carefully inserted in gas tight vials containing deionized water and 194 HgCl2 solution and kept at 4°C until methane analysis. Dissolved methane was quantified after 195 degassing of the pore waters into the headspace and quantified by gas chromatography with a relative 196 uncertainty of ± 5 % (Sarradin and Caprais, 1996). The position of the sulfate-methane transition zone 197 (SMTZ) was determined as the zone around the depth where [SO4 2-] = [CH4] (Komada et al., 2016). 198 Finally, acid volatile sulfur (AVS) for the determination of FeSs was extracted from the same sediment 199 used for the pore water extractions and conducted in triplicate by cold acid distillation of H2S (g) 200 under anoxic conditions that was trapped by NaOH and quantified voltammetrically (Henneke et al., 201 1991). 202

Calculations of oxygen uptake and AVS burial rates 218
Diffusive oxygen uptake (DOU) fluxes were calculated using Fick's first law (Berner, 1980, Eq. 219 4), 220 where ϕ is the sediment porosity, Ds is the apparent diffusion coefficient in the sediments, and 221 is the oxygen gradient at the SWI. The Ds coefficients were adjusted for diffusion in a porous 222 environment according to: = 0 (1+3•(1−∅)) with the diffusion coefficient in free water (D0) chosen 223 according to Broecker and Peng (1974) and recalculated to in situ temperature by the Stokes-Einstein 224 relation (Li and Gregory, 1974). 225 AVS burial fluxes were estimated using available sedimentation rates (ω from Charmasson et al. 226 (1998) and Miralles et al. (2005)), average AVS concentrations and porosities of each sediment core, 227 according to Eq. 5, 228 where ϕ is the sediment porosity, ω the sedimentation rate, and ρ the sediment dry bulk density. 229

Stoichiometric ratios 230
To determine the relationship between net TA and DIC production and to establish whether 231 sulfate reduction represents the main source of TA and DIC in these sediments, stoichiometric ratios 232 of the relative production of TA compared to DIC (rAD), as well as TA (rAS) and DIC (rDS) compared 233 to sulfate consumption, were calculated from the pore water data and compared to theoretical ratios 234 from the reaction stoichiometries (Table 1). Experimental stoichiometric ratios were obtained from 235 the slope and standard deviation of the linear regression of TA, DIC, and sulfate property-property 236 plots of concentration changes with respect to bottom water concentrations at each depth in the pore 237 waters (ΔTA, ΔDIC and ΔSO4 2-) relative to each other after correcting for differences in TA, DIC and 238 sulfate diffusion in the sediments (Berner, 1980, Eq. 6), 239 where i is the concentration of either TA or DIC, j the concentration of SO4 2or DIC and Di and Dj 241 are the corresponding diffusion coefficients. At the pH of the pore waters (pH ~ 7.5), more than 95 % 242 of DIC and carbonate alkalinity are composed of bicarbonate ion (HCO3 -). Given the relatively small 243 difference in the diffusion coefficients of HCO3and CO3 2-(11.8 and 9.55 x 10 -6 cm 2 s -1 at 25°C, Li 244 and Gregory, 1974) and the high proportion of HCO3relative to CO3 2-, the diffusion coefficient of 245 HCO3was adopted for both TA and DIC diffusion. 246 The effect of the precipitation or dissolution of calcium carbonate on TA and DIC variations was 247 also accounted for by considering the Ca 2+ concentration gradients in the pore waters. For these 248 calculations, the absolute value of the Ca 2+ concentration relative to its bottom water concentration 249 (ΔCa 2+ ) was added to the ΔTA or ΔDIC after taking the corresponding diffusion coefficients into 250 account (DTAΔ TA + 2DCa|ΔCa 2+ | for alkalinity and DDICΔ DIC + DCa|ΔCa 2+ | for DIC) and plotted 251 against DSO42-ΔSO4 2-. The calculated slope provided a stoichiometric ratio corrected for the 252 precipitation of calcium carbonate (rIJc). Pore water saturation states, regarding Calcite (ΩCa), were 253 calculated according to the equation proposed by Mucci (1983) and Millero (1995). 254

Bottom water and surface sediment characteristics 256
At all stations, bottom water salinities ranged from 37.5 to 38.0 and temperatures varied from 257 14.7 to 20.6 °C (Table 2). Total alkalinity and DIC concentrations (average TA = 2.60 ± 0.01 mM and 258 average DIC = 2.30 ± 0.02 mM, Table 2) were relatively high compared to the Mediterranean Sea 259 average, but common for the Gulf of Lions (Cossarini et al., 2015). The pHT of the bottom waters 260 varied from 8.05 to 8.09 with the highest value observed at station AK and the lowest at station E. 261 Although the oxygen concentration decreased with water depth, bottom waters were always well 262 ventilated, with dissolved O2 concentrations higher than 220 µmol L -1 . Sediment porosity ranged 263 between 0.7 and 0.8 at the SWI, and they were similar at all stations between 20 and 400 mm depth 264 (Table 2). 265

Benthic total and diffusive fluxes 266
The in situ pH and O2 microprofiles reflected the differences between the three study domains 267 under the influence of the Rhône River plume (Fig. 2). In the proximal zone (stations A and Z), the 268 oxygen penetration depth was only 1.5 to 2.5 mm into the sediment as also indicated by separate 269 voltammetric measurements (Fig. 5). The oxygen penetration depth increased from 2 to 6 mm at 270 station K and reached 8 to 11 mm at the most offshore station E. As a result of bad weather conditions, 271 no exploitable in situ microprofiles were recorded at stations AK and B, though ex situ voltammetric 272 profiles determined oxygen penetration depths of 4 and 2 mm, respectively (Fig. 5). All pH microfiles 273 indicated a pH minimum between 7.2 and 7.4 just below the OPD followed by an increase to between 274 7.5 and 7.6 in the manganous/ferruginous layers of the sediment around 5 mm inshore and below 12 275 mm offshore (Fig. 2). Below this depth, pH stabilizes. 276 The benthic chamber was deployed once at stations A and E and twice at station Z (Z' is the 277 replicate). Total alkalinity and DIC concentrations increased linearly with time in the chamber, but 278 concentration changes decreased along the nearshore-offshore transect (Fig. 3). The highest benthic 279 fluxes were recorded for the two deployments at station Z, with TA fluxes of 73.9 ± 20.6 and 56.0 ± 280 17.8 mmol m -2 d -1 and DIC fluxes of 78.3 ± 10.9 and 37.2 ± 7.2 mmol m -2 d -1 (Fig. 4, Table 2). At ± 0.9 mmol m -2 d -1 at stations A and Z and decreased offshore to 5.9 ± 1.0 mmol m -2 d -1 at station K 285 and 3.6 ± 0.6 mmol m -2 d -1 at station E (Fig. 4, Table 2). Although the relative importance of DOU 286 compared to TA and DIC fluxes increased offshore, the TA and DIC fluxes were always between 2 287 and 8 times larger than the DOU fluxes (Fig. 4). FeSaq was only detected below 15 cm at station K, ΣH2S was produced in low concentrations (< 5 300 µM) around 6.5 cm at stations B and K (Fig. 5). A peak of Fe 2+ was initially formed in the top 5 cm 301 of the distal domain (station E) but decreased to a minimum value with depth and did not correlate 302 with the organic-Fe(III) voltammetric signals, which also remained low throughout the profile (Fig.  303 5). Finally, station E displayed generally low concentrations of ΣH2S in the pore waters (< 6 µM), 304 though the onset of ΣH2S production was much shallower (2.5 cm) and ΣH2S concentrations were 305 consistently higher throughout the profile than at any other stations. 306 Both TA and DIC concentrations increased rapidly within the pore waters (Fig. 6) was much smaller at station E with a minimum concentration of 28 mM (bottom water sulfate 321 concentration was 31.4 mM). As a result, TA and DIC changes in concentration at a given depth were 322 highly inversely correlated (r 2 > 0.97) with sulfate changes in concentration at stations A, Z, AK, B, 323 and K (Table 3). At station E, sulfate variations in the observed depth were in the same order of 324 magnitude as the measuring uncertainties. Simultaneously, TA and DIC demonstrated strong 325 correlations (r 2 > 0.97) at stations A, Z, AK, B, and K (Table 3). In the proximal domain (stations A 326 and Z), ammonium increased with sediment depth to concentrations > 3 mM (Fig. 6). At station B, 327 ammonium reached concentrations > 2 mM with depth, whereas ammonium concentrations did not 328 exceed 1.5 mM at station AK, 0.6 mM at station K, and 0.3 mM at station E. At all stations nitrite 329 plus nitrate concentrations were less than 20 µM (data not shown). Significant methane 330 concentrations (> 50 µM) were detected at the bottom of the sediment core at stations A, Z, and AK 331 ( Fig. 6), and a SMTZ was identified between 28 and 39 cm at station A and between 19 and 39 cm at 332 station Z. As methane was < 50 µM throughout the profile at station K and sulfate was not completely 333 consumed inside the sediment core at station AK, the SMTZ was not determined at these two stations. 334 Methane analyses were not carried out for the other stations. large increase in concentration (up to 160 µM at station Z) was observed at station AK, K and Z 341 between 15 and 22 cm. In turn, ΣPO4 3production was minimal throughout station E pore waters (< 342 10 µM). Dissolved phosphate was not measured at station B. Sediment samples were analyzed for 343 AVS as a function of depth at stations A, AK, and E to assess one station in each domain ( Fig. 6). At 344 station A, a peak in AVS (65 µmol g -1 ) was measured around 8.0 cm followed by a second, smaller 345 peak (22 µmol g -1 ) at 14 cm, after which AVS decreased with depth. The AVS concentrations were 346 low in the top portion of the sediment at station AK but increased with depth to 100 µmol g -1 around 347 15 cm. At station E, only a small AVS peak of 20 µmol g -1 was observed at 14 cm. Finally, large 348 concentrations of FeS nanoparticles (FeS0) were found in the proximal and prodelta stations, 349 including two broad peaks and maximum concentrations around 1 mM at stations A and Z and a large 350 subsurface maximum up to 6 mM at 145 mm at station AK. These FeS0 concentrations increased as 351 a function of depth to a relatively constant 0.5 mM below 4.5 cm at station B and below 12 cm at 352 station K, whereas they remained mostly negligible at station E (Fig. 6). approach. Finally, the link between inputs to the sediment, carbon mineralization processes, sulfide 361 mineral burial, and the benthic TA flux is provided using a conceptual model.  such, denitrification would account for < 10% of the TA flux in the proximal zone where substantial 386 fluxes were measured by in situ benthic chambers (Fig. 4). Furthermore, the only net production of 387 TA by denitrification must be related to external nitrate sources as nitrification (overall oxidation of 388 ammonium to nitrate) consumes 2 moles of TA per mole of ammonium transformed into nitrate (Table  389 1, Eq. 2; Hu et al., 2011a). As coastal sediments mostly display coupled nitrification-denitrification, 390 this process does only represent a small source of TA to the bottom waters (Brenner et al., 2016). It 391 can therefore be concluded that the contribution of denitrification to TA fluxes is minimal in the 392 proximal zone and could be proportionally more important on the shelf where TA fluxes are much 393 lower. 394

DIC and TA produced by sulfate reduction 395
Sulfate reduction typically represents a major organic carbon mineralization pathway in organic-396 rich sediments that simultaneously produces two moles of total alkalinity (TA) and two moles of DIC 397 per mole of sulfate ( in two moles of TA produced per mole of Fe. As these two processes equally produce two moles of 400 TA per mole of terminal electron acceptor (Table 1, Eq.6 for SO4 2and Eq. 7 for Fe(OH)3), they can 401 both contribute significantly to the bulk alkalinity production in sediment pore waters. The low 402 concentration of nitrate, relatively low production of reduced metals in the pore waters (Fig. 5), and 403 intense ammonium and DIC production in parallel with sulfate consumption at depth (Fig. 6)  relative production of DIC and TA compared to sulfate consumption may indicate the dominant 407 reaction pathways responsible for the high alkalinity generated in these sediments (Burdige and 408 Komada, 2011). Factoring carbonate precipitation using the pore water Ca 2+ data, the rDSc were 409 determined to range between -2.05 and -1.86, except for one value at -1.37 (station B), whereas the 410 rASc ratios ranged between -2.35 and -1.89 with the exception of station B at -1.58 (Table 3). 411 Theoretically, the rDS and rAS should equal -2.0 if sulfate reduction is the only control on DIC and TA 412 production (Table 1, Eq. 6), suggesting that, except at station B, the influence of other diagenetic 413 processes on rASc and rDSc is limited. At station B, however the higher rDSc ratio (

Formation of iron sulfide species 418
Although the complete depletion of sulfate in the first 30 cm of the sediment at stations A, Z, and 419 B implies an equivalent production of dissolved sulfide (ΣH2S) (Table 1, Eq. 6), pore waters displayed 420 little to no ΣH2S (Fig. 5). If all of the produced ΣH2S diffused upward and reacted in the oxic sediment 421 layer, the alkalinity produced by sulfate reduction would be consumed by the oxidation of ΣH2S by 422 dissolved O2 and the pH should be lowered significantly given the large acidity generated by this 423 reaction (Table 1, Eq. 4). Although ΣH2S was nearly absent of the pore waters (Fig. 5), the pH 424 minimum was never lower than 7.2 and the observed alkalinity fluxes across the SWI were substantial 425 (Fig. 4), indicating that ΣH2S was removed from the pore waters below the oxic layer. Abiotic 426 reduction of Fe(III) oxides by ΣH2S ( The observed range of rADc (1.06 to 1.15) and rDSc (-2.05 to -1.86) ratios in the proximal and 443 prodelta stations, except at station B (Table 3), is fully compatible with sulfate reduction coupled to 444 iron reduction and FeS precipitation (possibly followed by pyritization), though rADc and rDSc ratios 445 are not able to distinguish abiotic and microbial pathways of iron reduction. The occurrence of 446 dissimilatory iron reduction in the proximal and prodelta domains, however, is substantiated by 447 several other pieces of evidence. First, the production of soluble organic-Fe(III) complexes deeper 448 than the oxygen penetration depths (Fig. 5) indicates that these species did not result from the 449

FeS precipitation 464
The discrepancy between sulfate consumption and the low concentration of ΣH2S along with the 465 high TA fluxes clearly suggest that much of the sulfur was precipitated in the solid phase. Indeed, 466 AVS measurements show precipitation of FeS in the proximal and prodelta domains (Fig. 6). In 467 addition, the large phosphate concentrations observed at depth in the proximal and prodelta domains 468 they were rarely observed in the Rhône River delta (Fig. 5). Indeed, the ion activity products (pIAPs) 478 calculated at most stations indicate that pore waters were either undersaturated, as a result of the low 479 concentrations (stations AK, B, and K) or complete absence (stations A and Z) of dissolved sulfides, 480 or close to the solubility of amorphous FeS or mackinewite (Fig. 8) (Table 1, Eq. 15). In this calculation, the alkalinity production flux was estimated from 497 the average AVS burial fluxes using Eq. 5, with the caveat that these flux comparisons are made 498 assuming steady-state which is questionable in such a dynamic system. Nonetheless, the average AVS 499 concentration of the proximal station (station A) was used, as the sedimentation rate at this station is 500 so high (>30 cm y -1 ) that the entire sediment layer investigated is buried rapidly in a year. The 501 calculated AVS burial flux provides an alkalinity-equivalent flux of 25.0 ± 7.7 mmol m -2 d -1 in the 502 proximal domain (Table 4), which falls within the range of benthic alkalinity fluxes measured by 503 benthic chamber at stations A and Z (14.3 -73.9 mmol m -2 d -1 ; Fig. 4 and Table 4). In the prodelta, 504 the alkalinity-equivalent flux is estimated at 9.8 ± 2.8 mmol m -2 d -1 at station AK (Table 4) average AVS burial flux at station E. This flux is much lower than the 3.7 ± 0.9 mmol m -2 d -1 flux 508 measured by benthic chamber (Fig. 4), a difference that could be due to denitrification and shallow 509 carbonate dissolution. CaCO3 precipitation at depth in proximal zone sediments. Yet, the intense consumption of dissolved 518 oxygen in the first millimeters below the sediment-water interface generates a large pH decrease (Fig.  519 2) that may induce carbonate dissolution at this scale. Calcium carbonate saturation states at a 520 millimeter scale near the SWI were calculated from pH profiles and an interpolation of the centimetre-521 scale DIC profiles using the SeaCarb software (Fig. 9). They show that in the proximal zone, the 522 saturation state with respect to calcite, which is the most abundant detrital carbonate in these 523 sediments (Rassmann et  the TA-consuming reoxidation of reduced metabolites (i.e., NH4 + , ΣH2S, Fe 2+ ) is not important in the 548 oxic sediment layers, and a significant fraction of the anaerobically-produced TA is transferred across 549 the SWI (Fig. 10, red dashed line). In these conditions, anaerobic and aerobic processes are 550 decoupled, and the consumption of oxygen no longer reflects the overall respiration rates within these 551 sediments (Pastor et al., 2011) as observed by the relatively lower contribution of DOU fluxes 552 compared to TA and DIC fluxes in the proximal domain (Fig. 4). 553 In contrast, sedimentation rates (Table 2), overall respiration rates (Fig. 4), and the intensity of 554 iron and sulfate reduction (Fig. 6)  findings likely reflect the fact that less riverine Fe(III) oxides were available for FeS precipitation. 559 With low sedimentation rates (0.1 to 1 cm yr -1 ) and thus low input of organic matter and Fe(III) 560 oxides, the overall carbon turnover is decreased and the reduced by-products of sulfate and/or iron 561 reduction may be transported back to the oxic sediment layers to be reoxidized by dissolved oxygen. 562 In this case, the alkalinity generated by anaerobic respiration processes is consumed by reoxidation 563 of the reduced metabolites, and the flux of alkalinity near the SWI decreases to weak values at station 564 E ( Fig. 4 and Fig. 10, black line). 565 The strong TA flux to the overlying waters measured in the Rhône River delta, may contribute, 566 along with riverine inputs, to the overall high alkalinity of the Gulf of Lions waters compared to the 567 Mediterranean average (Cossarini et al., 2015). However, the influence of the benthic TA flux on the 568 water column pH and ultimately on the absorption of atmospheric CO2 depends mainly on the TA to 569 DIC benthic flux ratio (FTA/FDIC), vertical mixing in the water column, and thus the residence time of 570 the bottom waters (Hu and Cai, 2011b, Andersson and Mackenzie, 2012). The FTA/FDIC ratios, ranging 571 between 0.8 and 1 in the proximal and prodelta zones of the Rhône River delta (Fig. 11), are in the 572 high range of a compilation of TA to DIC flux ratios obtained in different coastal systems and 573 continental shelves (expanded from Hu and Cai, 2011b). As these ratios do not exceed 1, alkalinity 574 generated in the sediments will not decrease pCO2 in the bottom waters and thus not draw atmospheric 575 CO2 into the coastal ocean. Yet, the large benthic TA fluxes generated from deltaic sediments and the 576 elevated FTA/FDIC (>0.8), which were unknown in the Rhône River prodelta before this study, may 577 modify the carbonate cycle paradigm in these coastal regions. 578

Conclusion 579
In this study, benthic respiration, as well as benthic alkalinity and DIC fluxes were quantified in 580 the Rhône River delta using benthic landers. These measurements demonstrated that sediments from 581 the proximal and prodelta domains represent a strong source of alkalinity to the water column. much stronger than fluxes of dissolved oxygen, indicating the decoupling of oxic and anoxic 584 biogeochemical processes. As pore water oversaturation with respect to calcite prevented carbonate 585 dissolution to occur over the entire sediment column, the high benthic alkalinity fluxes resulted from 586 the high intensity of anaerobic respiration processes, mainly via sulfate reduction and precipitation 587 of iron sulfide minerals, but also with some contributions from dissimilatory iron reduction and AOM. 588 The intensity of sulfate reduction in the proximal domain also resulted in the consumption of a 10-589 20% fraction of the alkalinity and DIC by the precipitation of authigenic carbonates. As the reduced 590 metabolites Fe 2+ and ∑H2S produced by the mineralization of organic matter were buried in the solid 591 phase, alkalinity was not consumed by their reoxidation in the oxic sediment layers. Consequently, a 592 significant fraction of the total alkalinity generated in the pore waters was transferred to the bottom 593 waters (benthic flux of 14-74 mmol m -2 d -1 ). Although sulfate reduction dominated the proximal and 594 prodelta domains, evidence for dissimilatory reduction of Fe(III) oxides was simultaneously observed 595 in the depth profiles, suggesting that anaerobic processes in the Rhône River prodelta are dynamic 596 and potentially controlled by pulsed sediment accumulations. The intensity of the alkalinity and DIC 597 fluxes decreased offshore as the sedimentation rate and the relative importance of anaerobic 598 mineralization pathways compared to aerobic processes decreased. In these conditions the more 599 "classical" coupling between aerobic and anaerobic reactions occurs, hence producing much lower 600 benthic alkalinity fluxes. Overall, these findings suggest that deltaic sediments exposed to large 601 riverine inputs of inorganic and organic material may provide a large source of alkalinity to the 602 overlying waters and thus weaken the increase in pCO2 more significantly than previously thought in 603 coastal waters. 604 and Cai (2011b) which was corrected in their later publication (Hu and Cai, 2013). 922              Table 3. Diffusion-corrected stoichiometric ratios rAD , rDS, and rAS and their corresponding ratios corrected for carbonate precipitation (rADc , rDSc, and rASc) along with their associated determination coefficients (r 2 ) from linear regression; n.d = not determined.