Alkalinity Production Coupled to Pyrite Formation Represents an Unaccounted Blue Carbon Sink

Blue carbon ecosystems, including mangroves, saltmarshes, and seagrasses, mitigate climate change by storing atmospheric carbon. Previous blue carbon research has focused on organic carbon stocks. However, recent studies suggest that lateral inorganic carbon export might be equally important. Lateral export is a long‐term carbon sink if carbon is exported as alkalinity (TAlk) produced via sulfate reduction coupled to pyrite formation. This study evaluates drivers of pyrite formation in blue carbon ecosystems, compares pyrite production to TAlk outwelling rates, and estimates global pyrite stocks in mangroves. We quantified pyrite stocks in mangroves, saltmarshes, and seagrasses along a latitudinal gradient on the Australian East Coast, including a mangrove dieback area, and in the Everglades (Florida, USA). Our results indicate that pyrite stocks were driven by a combination of biomass, tidal amplitude, sediment organic carbon, sediment accumulation rates, rainfall, latitude, temperature, and iron availability. Pyrite stocks were three‐times higher in mangroves (103 ± 61 Mg/ha) than in saltmarshes (30 ± 30 Mg/ha) and seagrasses (32 ± 1 Mg/ha). Mangrove pyrite stocks were linearly correlated to TAlk export at sites where sulfate reduction was the dominant TAlk producing process. However, pyrite generation could not explain all TAlk outwelling. We present the first global model estimating pyrite stocks in mangroves, giving a first‐order estimate of 197 Mg/ha (RMSE = 24 Mg/ha). In mangroves, estimated global TAlk production coupled to pyrite formation (∼3 mol/m2/y) is equal to ∼24% of their global carbon burial rate, highlighting the importance of including TAlk export in future blue carbon budgets.

Dissolved carbon export from blue carbon ecosystems can be considered a long-term atmospheric carbon sink if carbon is exported as total alkalinity (TAlk), since exported TAlk has a residence time in the ocean of ∼1 × 10 5 years (Emerson & Hedges, 2008;Middelburg et al., 2020). Flushing of porewater TAlk followed by lateral export to the coastal ocean is a major carbon sink in mangroves and saltmarshes Wang & Cai, 2004). Sedimentary processes that couple organic matter degradation and TAlk production include denitrification, manganese reduction, iron reduction and sulfate reduction (Burdige, 2011;Krumins et al., 2013). However, oxidation of reduced compounds, for example, sulfide (HS − ), may consume neo-formed TAlk: ). This reaction releases a proton (H + ) that titrates the bicarbonate alkalinity back to CO 2 , thus offsetting the initial blue carbon sequestration. Therefore, only a permanent spatial decoupling of anaerobic remineralization products from TAlk, such as loss via nitrogen gas from denitrification and precipitation of reduced sulfur as sedimentary pyrite, can contribute to net TAlk production (Hu & Cai, 2011). Due to nitrogen limitation, denitrification rates in pristine blue carbon ecosystems are usually low (Bianchi, 2007). As such, sulfate reduction coupled to pyrite formation is often the dominant net TAlk production process coupled to organic matter degradation in pristine blue carbon ecosystems ( Figure 1).
In coastal sediments, pyrite (FeS 2 ) is formed via several, complex pathways. A common pathway involves iron(II) sulfide (FeS) as an initial precursor, which is formed from the reaction of sulfide, produced by microbially mediated sulfate reduction, with ferrous iron (Fe 2+ ) or with reactive iron(III) oxides (Schoonen & Barnes, 1991). Since sulfur in pyrite is slightly more oxidized (S-I) than in sulfide (S-II), pyrite is produced via intermediate redox reactions, including partial oxidation of FeS. Pyrite formation mainly occurs in upper sediment layers, where reactive iron concentrations and sulfate reduction rates are sufficiently high due to mixed redox conditions, adequate sulfate supply from surface waters, and available labile organic matter (Burdige, 2011;Wada & Seisuwan, 1986). Within this biogeochemically dynamic layer, pyrite can form via reaction of FeS with polysulfide (S 2 x− ) or via a greigite intermediary (Fe 3 S 4 ) (Goldhaber, 2003;Rickard & Luther, 2007). Alternatively, FeS can react with hydrogen sulfide to form pyrite and hydrogen gas (Rickard, 1997), especially within deeper sediment depths where sulfide concentrations are often comparatively high. Howarth (1979) observed that in saltmarshes, pyrite could form rapidly without FeS as an initial precursor. In situ rates of pyrite formation depend on a wide range of environmental conditions (Burton et al., 2011).
The availability of sulfate, organic carbon, and reactive iron are key factors controlling pyrite formation in coastal sediments (Berner, 1970(Berner, , 1984 Figure 1). In blue carbon ecosystems, tidal seawater inundation resupplies the sulfate required for sulfate reduction. Therefore, sulfate is usually only limiting in upper estuarine freshwater reaches (where the salinity drops below 2.5-4), where sediments are impermeable, or in extremely carbon-rich sediments (Burdige, 2011). Sulfate reduction also requires a source of organic carbon. Even though blue carbon ecosystems are generally carbon-rich, their carbon content is variable and depends on climatic factors, sediment accumulation rates (SAR), geomorphology, hydrology, vegetation species composition, and nutrient availability (Kristensen et al., 2008;Macreadie et al., 2014;Ouyang & Lee, 2014;Sasmito et al., 2020). In carbon-rich systems (>15% sediment organic carbon content), reactive iron availability can also limit pyrite formation (Berner & Raiswell, 1984). The availability of reactive iron depends on regional geology, sediment texture, SAR, and weathering rates (Kendall et   & Canfield, 2012). Whether sulfate, organic carbon or reactive iron limits pyrite formation is highly site specific (Morse et al., 2007), hence modeling these processes is a challenge.
This study aims to evaluate TAlk production coupled to pyrite formation in blue carbon ecosystems and to examine whether pyrite stocks represent a significant blue carbon sink. To achieve this, we investigated pyrite stocks and controlling factors from local to regional to global scales. We hypothesized that pyrite formation is subject to local-scale variability in controlling factors (e.g., soil carbon) and that large-scale disturbance, such as mass mangrove mortality would affect mangrove pyrite stocks. To test these hypotheses, we conducted two case studies: one within Everglades National Park, which protects the largest contiguous mangrove forest in North America (144,447 ha), and a second in an Australian mangrove forest that experienced the most severe dieback ever documented in the literature. We tested whether pyrite stocks can be used as a proxy for TAlk generation locally by comparing previously measured TAlk export rates to pyrite stocks. We quantified pyrite stocks in mangroves, saltmarshes and seagrasses along a latitudinal gradient and investigated relationships between pyrite stocks and key drivers. Based on these key drivers we developed the first global model for mangrove pyrite stocks. This model, coupled with global averaged sediment accumulation rates, was then used to provide a first-order estimate of global mangrove TAlk generation through pyrite generation.

Study Sites
We investigated pyrite stocks in three areas that span the general range of climatic, biomass, diversity, and sediment carbon stocks observed in mangroves globally, as well as covering a range in catchment geology ( Figure 2, Table 1). First, at Everglades National Park in Florida, USA, sampling was conducted along the Shark River estuary. The Shark River estuary is dominated by mangroves, transitioning to freshwater marsh areas in the freshwater reaches.
Second, we sampled an Australian mangrove dieback area located in the Gulf of Carpentaria near Karumba. During late 2015 and mid-2016, more than 7,400 ha of mangrove forest died due to high temperatures, drought, and a rapid drop in sea levels during the pre-monsoon dry season (Duke et al., 2017;Lovelock et al., 2017). The Norman River separates a dead mangrove stand from an adjacent living mangrove area, allowing direct comparison to test the effect of the dieback event on pyrite stocks in an identical climate zone.
Third, we conducted sampling along a latitudinal gradient at the Australian East Coast ( Figure 2, Table 1). Sampling areas along the latitudinal gradient from 12° to 38°S included mangroves, saltmarshes and seagrasses located in Darwin, Hinchinbrook Island, Rockhampton, Seventeen Seventy, River Heads, Sunshine Coast, Jacobs Well, Ballina, Coffs Harbour, Queens Lake, Newcastle, Sydney, and Barwon Heads. The Australian study sites cover a large range of climate zones, ranging from low average annual temperature (15 °C) and rainfall (666 mm) in temperate Barwon Heads to high average annual temperature (27 °C) and rainfall (1,694 mm) in tropical Darwin.
(AVS) and chromium-reducible sulfur (CRS) analysis, together known as reduced inorganic sulfur (RIS), were either subsampled on-site into plastic bags with all air-removed, and kept frozen or subsampled from sediment cores, which were sealed with plastic shrink wrap, stored frozen and transported to the laboratory. Additional sediment samples were taken for dry bulk density, reactive iron, and organic carbon measurements. At Darwin, Hinchinbrook Island, Seventeen Seventy, Jacobs Well, Newcastle, and Barwon Heads subsamples were taken to analyze SAR.
At selected Everglades and Karumba sites, porewater profiles were analyzed for TAlk, Fe 2+ , total aqueous iron (Fe Tot ), and aqueous sulfide (S(-II)), which includes H 2 S, HS − , and S 2− . Porewater extraction was conducted as described by Johnston et al. (2016). Briefly, sediment cores were collected in PVC cores, and 10 cm long Rhizon samplers were inserted into the cores at increasing depths. Porewater was extracted under vacuum via tubing and Luer-Lock connectors through a needle into N 2 filled, O 2 -free 10 ml glass vials, which were sealed with rubber septa. Porewater samples were analyzed on the day of collection.
At Karumba, groundwater was sampled for TAlk analysis 1 m adjacent to the sediment core collection site. Boreholes were dug with a post-hole digger and purged three times with a peristaltic pump. Groundwater was allowed to completely recharge before sampling. Groundwater TAlk samples were stored cool and measured within 1 day. REITHMAIER ET AL.

Sample Analysis
Sediment AVS, which includes intermediaries such as FeS and greigite, was analyzed by the diffusion method outlined by Burton et al. (2009), extracting AVS with hydrochloric acid/ascorbic acid and trapping H 2 S in an alkaline zinc solution. CRS was quantified sequentially after AVS analysis, using the method developed by Burton et al. (2008). Both AVS and CRS concentrations were determined via iodometric titration of the zinc traps, with a measurement error of 11% (based on repeat duplicate analysis) and a detection limit of 2 µmol/g. Sediment organic carbon of cores that were collected in 2019 (Table 1) was analyzed, as described by Radabaugh et al. (2018). In brief, the bulk density was determined by mass loss, drying the samples at 105 °C. Subsequently, organic carbon was determined by the loss-on-ignition (LOI) method, combusting samples at 550 °C. To convert LOI (%) into sediment organic carbon (%), an ecosystem specific conversation factor was used for mangroves (0.42), saltmarshes (0.50), and seagrasses (0.40) (Fourqurean et al., 2012;Radabaugh et al., 2018). Organic carbon of cores collected before 2019 was analyzed using a Flash Elemental Analyzer coupled to a Thermo Fisher Delta V isotope ratio mass spectrometer.
Groundwater TAlk was determined by a Gran titration using a titrator (Metrohm 888 Titrando with Tiamo light) with a precision better than 5 µM. Drifts and deviations in the acid concentration were corrected using certified reference materials (CRM batch 175 and CRM batch 178), as described by Dickson (2010). SAR was determined by examining the nuclear fallout signature of plutonium-239 + 240 as described by Sanders et al. (2016b) with an uncertainty of 16%. In brief, samples were dry-ashed, spiked with Pu-242, and REITHMAIER ET AL. Note. Annual temperature and rainfall were averaged between 1982 and 2012 (Climate-Data.org, 2020).

Calculations and Statistics
Pyrite concentrations were calculated by dividing CRS concentrations by two, assuming that the stoichiometry of pyrite is FeS 2 and that elemental sulfur, which may contribute to CRS, is negligible. These assumptions are reasonable since elemental sulfur is often negligible in blue carbon ecosystems and thus FeS 2 is the dominant form of CRS (Ding et al., 2014;Holmer et al., 2006;Johnston et al., 2016). Pyrite and carbon stocks were calculated by summing the pyrite and sediment organic carbon density (concentrations multiplied by the dry bulk density) over the first meter of each sediment core. If cores were shorter than 1 m, the deepest measured pyrite concentration and bulk densities were extrapolated to 1 m. We chose to integrate pyrite stocks to 1 m depths to enable a direct comparison with sediment carbon stocks, which are commonly integrated over 1 m in the blue carbon literature.
The degree of pyritization (DOP), a proxy for iron limitation, was calculated according to Berner (1970): At sites where SAR was determined, TAlk production rates coupled to pyrite formation, as well as carbon burial rates were estimated. Mass accumulation rates (MAR) were calculated as the product of SAR and dry bulk density. Multiplying MAR with the carbon concentration at each depth, carbon accumulation rates were calculated and averaged per core. Similarly, pyrite accumulation rates were calculated by multiplying pyrite concentrations with MAR per depth and averaged per core.
The TAlk production during pyrite formation, which is coupled to the stoichiometry of sulfate reduction, was estimated. Different examples of the stoichiometry of TAlk production during pyrite formation exist. According to Sheoran et al. (2010), approximately 2-3 TAlk equivalents are produced per mol of reduced sulfate (SO 4 2− ), with the ratio depending on the organic substrate, that is, acetate (CH 3 COO -), lactate (3CH 3 CHOHCOO -), or propionate (CH 3 CH 2 COO -) (Sheoran et al., 2010): Given that 2 mols of reduced sulfur are required for a single mol of pyrite (Blodau, 2006;Johnston et al., 2012), the resulting conversion factors for TAlk production per pyrite formation range from 4 to 5. Alternatively, it has been argued that sulfate reducers utilize formaldehyde (CH 2 O), which derives from fermenters breaking down complex organic carbon substrates. Furthermore, TAlk production per pyrite formation is impacted by the oxidant used to partially oxidize sulfide in FeS (Burdige, 2006). If oxygen is used for sulfide oxidation the stichometry can be described as (adapted from Berner, 1970): This results in a conversion factor of 4. Another possible sulfide oxidant is clay bound Fe 3+ : In theory, this reduces the conversion factor for TAlk production during pyrite formation to 3. We consider Equation 6 as a moderate estimate of pyrite associated TAlk production and one that is likely, since mangrove roots are known to create oxidized microzones (see Section 4.1). Therefore, in this study, we used a conversion factor of 4 per mol of pyrite to estimate TAlk production rates coupled to pyrite formation. Uncertainty in calculated TAlk production rates was estimated by propagating the error related to the conversion factor (±25%) as well as the measurement errors of the pyrite (±11%) and SAR (±16%) analysis.
Statistical analysis was conducted in R-3.6.2. Pearson coefficients for correlations between pyrite stocks and drivers (average annual temperature, average annual rainfall, average annual tidal amplitude, average salinity, sediment organic carbon stocks, aboveground biomass, reactive iron within the catchment and SAR) were determined. Probability levels for Pearson correlations are indicated as * if p < 0.05, ** if p < 0.01, and *** if p < 0.001.
We developed a global model predicting global pyrite stocks in mangroves. A detailed description of the modeling process can be found in the Supporting Information S1. In brief, the best model was chosen in R-3.6.2 based on the lowest Akaike information criterion (AICc). The model data equation was developed based on data presented in Table 3: The root mean square error (RMSE) of the model was determined to provide some indication of model accuracy (Li, 1988). Global maps published in the literature were used to construct the model in ArcMap 10.7.1, including maps for sediment organic carbon (Sanderman et al., 2018), reactive iron (Rossel et al., 2016), aboveground biomass (Simard et al., 2019), and average tidal amplitude (Vestbo et al., 2018). The modeled average global pyrite stock (Mg/ha) in mangroves was normalized to the global mangrove area reported by Giri et al. (2011) to estimate the total global pyrite stock (Pg) in mangroves.

Everglades National Park
At Everglades National Park, bulk density ranged between 0.09 and 0.94 g/cm 3 and AVS was below the detection limit in all samples. Pyrite concentrations (3.4-408 µmol/g) were higher at the polyhaline downstream (E1) and mesohaline midstream (E2, E3) sites compared to the oligohaline upstream site (E4) and generally increased with depth ( Figure 3). There was no significant difference in the pyrite concentrations with increasing distance from the river.
On average, porewater Fe Tot (0.62-25 µM) was two-times higher at the upstream site (E4) compared to the mid and downstream sites (E1, E3) and decreased with depth. The share of porewater Fe 2+ (0.17-6.8 µM) on Fe Tot increased with depth and was highest at the midstream and downstream sites (up to 100% at E1 and E3). Increasing with depth, porewater TAlk (1,345-14,840 µmol/kg) was two-times higher at the downstream and midstream sites (E1 and E3) compared to the upstream site (E4).

Mangrove Dieback Near Karumba
At the Karumba mangrove site, the bulk density varied between 0.37 and 1.7 g/cm 3 . Concentrations of AVS ranged from 0 to 5.3 µmol/g and were below the detection limit in the majority of the samples. Pyrite reached higher values (0-217 µmol/g) at the living (K4-K6) than at the dead mangrove area (0-127 µmol/g, K1-K3) (Figure 4). Pyrite concentrations decreased with distance from the shore. However, this effect decreased with depth, particularly at the dead area. The redox potential indicated anoxic conditions at close proximity to the ocean (−321 to −109 mV), but partly suboxic conditions at mid (−215 to 12 mV) and distal (−145 to 231 mV) cores. The redox potential did not differ significantly between dead and living mangrove areas.
Decreasing with depth, reactive iron was on average two times lower at the living (0.25%) than at the dead (0.41%) area. The DOP was almost twice as high at the living (0.72) than at the dead (0.43) area. Porewater Fe 2+ ranged from 1 to 57 µM. In the living area, Fe 2+ accounted for an average of 24% of Fe Tot (1.6-83 µM), which showed two distinct peaks at 20 and 40 cm. At both areas, porewater S(-II) (0.11-8,543 µM) and TAlk (1,400-52,000 µmol/kg) increased simultaneously with depth. At the living area, S(-II) increased abruptly just below the lower Fe Tot peak.
On average, pyrite stocks were 20% lower at the dead (range: 21-74 Mg/ha, mean ± error: 50 ± 15 Mg/ ha) compared to the living area (   correlated with groundwater TAlk (ranging from 7,400 to 48,000 µmol/kg) in the living area (R 2 = 0.59*), but showed no significant correlation at the dead area.
Pyrite stocks were three-times higher in mangroves (103 ± 61 Mg/ha) than in saltmarshes (30 ± 30 Mg/ ha) and seagrass (32 ± 1 Mg/ha) sites ( Figure 6, Table 2). In saltmarshes, pyrite stocks were on average, threefold higher in cores taken closer to the water edge compared to cores taken at a higher elevation in the tidal frame, whereas at mangroves, no difference was found. Carbon stocks, calculated for 1 m depth, in mangroves (249 ± 58 Mg/ha) and saltmarshes (261 ± 86 Mg/ha), were three-times higher than in seagrasses (72 ± 15 Mg/ha). On average, carbon stocks were two-times higher than pyrite stocks at mangroves REITHMAIER ET AL.
Along the latitudinal gradient, average SAR ranged from 0.13 to 0.63 cm/y and MAR from 0.08 to 0.27 g/ cm 2 /y (Table 4). Pyrite accumulation and resulting TAlk production were lowest at temperate Barwon Heads (0.04 mmol/m 2 /d and 0.2 mmol/m 2 /d, respectively) and highest at tropical Darwin (0.41 mmol/m 2 /d and 13 mmol/m 2 /d, respectively), which showed the highest MAR and pyrite concentrations. TAlk production coupled to pyrite formation is equal to 1%-33% of the carbon burial rates, which ranged between 7 and 39 mmol/m 2 /d.

Pyrite Formation in Blue Carbon Ecosystems
Sulfate reduction rates and reactive iron availability are the key drivers impacting pyrite formation and thus the development of pyrite stocks (Berner, 1984; Figure 1). Seagrasses typically inhabit subtidal environments, where sulfate from overlying seawater is transported into their sediments via porewater diffusion and bioturbation (Chanton et al., 1987). In mangroves and saltmarshes, sulfate is resupplied during each flood tide, whereby tidal pumping drives effective porewater exchange, encouraging flushing of TAlk and transporting sulfate into deeper sediment layers, facilitating sulfate reduction (Hemond et al., 1984;Sadat-Noori et al., 2017). At the mangrove sites, average surface water salinities were well above the level (>15) at which sulfate availability can limit sulfate reduction (2.5-4) (Burdige, 2011), suggesting that sulfate availability was not a limiting factor for pyrite formation at our study sites.

Table 2 Average Pyrite Stocks, Sediment Carbon Stocks, and Degree of Pyritization (DOP) for 1 m Depth at Mangrove, Saltmarsh, and Seagrass Sites
In addition to sulfate supply, sulfate reduction requires organic carbon loading in excess of the aerobic respiration capacity. Blue carbon ecosystems store a large amount of organic carbon in their sediments (Mcleod et al., 2011), typically fueling high sulfate reduction rates. Berner and Raiswell (1984) stated that usually only in extremely carbon-rich systems (>15%), factors other than organic carbon, for example, reactive iron, limit pyrite formation. In our study, sediment organic carbon exceeded 15% only at mangroves on Hinchinbrook Island, whereas at most sites, sediment organic carbon was below 10%, thus suggesting that organic carbon might be a major factor controlling pyrite formation at blue carbon ecosystems. Accordingly, pyrite was significantly correlated with sediment organic carbon in both mangroves and seagrasses, underlining the importance of organic carbon on pyrite formation in these systems. Similarly, previous mangrove studies found correlations between organic carbon and pyrite (Crémière et al., 2017;Ding et al., 2014;Ferreira et al., 2007b;Sherman et al., 1998). The positive correlation between mangrove pyrite stocks and SAR suggests that high SAR may limit the degradation of organic matter via aerobic carbon mineralization pathways REITHMAIER ET AL. Notes. Annual temperature and rainfall were averaged between 1982 and 2012 (Climate-Data.org, 2020). Average annual tidal amplitude was retrieved from Bureau of Meteorology (2020). Reactive iron was averaged for a 50 km radius around the sites (Rossel et al., 2016). a Reithmaier et al. (2020b). b Sippo et al. (2016). c Brown et al. (2018). d Maher et al. (2015). e Reithmaier et al. (2020a). f Lee and Patterson (2002). g Jerath et al. (2016). h Sanders et al. (2016a). i Sippo et al. (2020a). j Simard et al. (2019). k Smoak et al. (2013). l Logan et al. (2011). m Conrad et al. (2017). n Macreadie et al. (2012). Note. Values for SAR and carbon burial rates in the Everglades were retrieved from Smoak et al. (2013). Abbreviations: MAR, mass accumulation rates; SAR, sediment accumulation rates; TAlk, total alkalinity.  (Berner, 1984;Raiswell & Canfield, 2012), thus providing labile carbon for sulfate reduction. At the seagrass sites, sediment carbon stocks were three times lower than that at mangrove and saltmarsh sites, which may contribute to the comparatively low pyrite stocks in seagrasses. Sediment organic carbon and pyrite concentrations showed no significant correlation at saltmarsh sites, suggesting that other limiting factors, such as less favorable redox conditions, might be more relevant controls on pyrite formation at the saltmarsh sites.
In addition to sedimentary organic carbon, vegetation itself can impact pyrite formation. Mangrove pyrite stocks were significantly correlated with aboveground biomass (r = 0.65*). Biomass might be a more conservative proxy for carbon availability than sediment carbon stocks, which are highly heterogeneous . Furthermore, aboveground biomass can be regarded as a proxy for belowground biomass (Komiyama et al., 2008), which actively influences pyrite formation. Roots may facilitate pyrite formation by lowering the pH, releasing exudates, stimulating sulfate-reducing bacteria and creating local oxidizing conditions required for partial FeS oxidation (Ferreira et al., 2007a;Giblin, 1988;Holmer et al., 1994Holmer et al., , 2006Morse, 1999). Previous studies have found higher pyrite content in vegetated areas than adjacent unvegetated areas, highlighting the importance of vegetation on pyrite formation (Andrade et al., 2012;Ferreira et al., 2007a;Giblin, 1988;Holmer et al., 2003;Otero et al., 2009). The degree to which the vegetation impacts pyrite formation varies between species, depending on root characteristics and primary productivity (Holmer et al., 2006;Sherman et al., 1998).
Reactive iron is also a critical ingredient for pyrite formation. Previous studies observed that reactive iron availability impacted pyrite formation in mangroves (Ferreira et al., 2007b;Otero et al., 2009), saltmarshes (Giblin, 1988;Morse et al., 2007), and seagrasses (Holmer et al., 2003;Morse, 1999), which were characterized by a sandy sediment texture or calcareous sediments and were thus iron-poor. Reactive iron was negatively correlated with pyrite at all blue carbon ecosystems. This can be explained by the biogeochemical zonation of coastal sediments (Froelich et al., 1979;Johnston et al., 2011). Pyrite and S(-II) increased with depth, whereas reactive iron(III) oxides are consumed, and Fe 2+ decreased under more reducing conditions due to pyrite formation. The DOP is a more effective measure to evaluate the effect of reactive iron on pyrite formation than in situ iron concentrations. The DOP was considerably higher at mangroves (up to 0.97) than at saltmarsh (up to 0.36) and seagrass (up to 0.59) sites, suggesting that only at some mangrove sites pyrite formation was limited by reactive iron. At Hinchinbrook Island (DOP = 0.97) and Darwin (DOP = 0.92), sites characterized by high sediment organic carbons stocks, DOP was particularly high, suggesting iron limitation. Furthermore, mangrove pyrite stocks showed a positive correlation with average reactive iron within a 50 km radius around the sites, suggesting that the catchment geology also impacted pyrite formation.
Sedimentary pyrite accumulation requires mixed redox conditions to form polysulfides and for partial oxidation of FeS (Luther et al., 1982). In blue carbon ecosystems, roots and bioturbation can create oxidized microzones promoting pyrite formation. However, intense bioturbation, high rates of primary production, and oxygen release from roots can also lead to oxidation of pyrite (Ferreira et al., 2007a;Giblin, 1988;Holmer et al., 1994Holmer et al., , 2006Luther et al., 1982). In intertidal mangroves and saltmarshes, redox conditions are strongly moderated by the water level. Consequently, pyrite formation can vary considerably between different geomorphological settings and locations within an ecosystem (Ferreira et al., 2007b;Giblin, 1988;Machado et al., 2014;Sherman et al., 1998).
At Karumba, pyrite decreased slightly with increasing distance to the ocean, which was accompanied by increasing redox potential. However, the proximity to the water edge (i.e., tidal channel) did not generally affect measured mangrove pyrite stocks. In contrast, saltmarsh pyrite stocks decreased noticeably with increasing distance from the water edge. The fact that saltmarsh pyrite stocks were threefold lower than mangrove pyrite stocks, despite equal carbon stocks and lower DOP, is likely due to more oxidizing conditions in the upper sediment layers, since saltmarshes occupy spaces higher in the tidal frame than mangroves and have therefore a shorter hydroperiod. Oxidizing conditions in the upper sediment layers are supported by the change in pyrite concentrations going down core in saltmarshes, where pyrite was close to zero in the upper sediment layers and increased abruptly after around 40-80 cm ( Figure 6).
Climatological and physical factors indirectly influence pyrite formation via their impacts on sulfate reduction and reactive iron availability. For example, mangrove pyrite stocks were significantly correlated with the tidal amplitude (r = 0.57*). Higher tidal amplitude increases porewater exchange , thereby resupplying sulfate, removing aqueous reaction products, and transporting allochthonous organic carbon and reactive iron through the sediments. Furthermore, mangrove pyrite stocks were also positively correlated with average annual temperature and rainfall, which affect sedimentary organic carbon and aboveground biomass (Hutchison et al., 2014;Sanders et al., 2016a). Moreover, microbial sulfate reduction is highly temperature dependent (Robador et al., 2016;Westrich & Berner, 1988). Increasing rainfall drives weathering rates and sediment transport, and thus influences reactive iron supply from surrounding catchments (Kendall et al., 2012).

Mangrove Pyrite Formation Represents an Overlooked Blue Carbon Sink
The blue carbon paradigm has focused primarily on the role of coastal vegetated ecosystems being hotspots of organic carbon burial and subsequently large sedimentary carbon stocks. In contrast, TAlk production has been largely overlooked as an atmospheric carbon sink (Maher et al., 2018). During sulfate reduction, organic carbon is converted into bicarbonate ions, some of which are exported from mangroves to the ocean via tidal porewater exchange (Krumins et al., 2013;Sippo et al., 2016), whereas sedimentary pyrite is sequestered in situ. This spatial decoupling of two key products of sulfate reduction (pyrite and bicarbonate ions) leads to a net generation of TAlk and generates a long-term carbon sink in the form of marine bicarbonate. As such, we argue that the blue carbon framework should incorporate a more holistic biogeochemical perspective, which encompasses not only in situ sedimentary organic carbon sequestration, but also pyrite formation and its attendant TAlk export. This is supported by our results that show TAlk production associated with pyrite formation can account for up to 33% of the organic carbon burial rate and represents, therefore, a quantitatively relevant carbon sink.
At the Everglades and Karumba, porewater TAlk increased simultaneously with S(-II) and pyrite, highlighting the importance of pyrite formation for TAlk production. Sherman et al. (1998) also observed a significant correlation between pyrite and TAlk in mangrove sediments, but the correlation was restricted to deeper sediments, suggesting that TAlk in surface sediments was either exported by tidal flushing or driven by other processes. At Everglades National Park and at mangrove sites near Karumba, porewater TAlk was on average six-times higher than average TAlk of seawater (2,300 µmol/kg; Millero et al., 1998), indicating that mangrove sediments can act as a TAlk source to coastal waters, since tidal pumping drives effective porewater exchange .
Rates of TAlk production coupled to pyrite formation (determined as described in Section 2.4) were compared to published lateral TAlk export rates in Figure 7. A significant linear correlation was found for Barwon Heads, Jacobs Well, Hinchinbrook Island, and Darwin. In contrast, Newcastle, Seventeen Seventy, and the Everglades had disproportionally high TAlk export rates despite apparently low TAlk production coupled to pyrite formation. The high TAlk export rates might be due to a range of alternative TAlk producing processes. For example, in the Everglades, mangroves populate a large carbonate platform, and therefore carbonate dissolution contributes substantially to TAlk production (Ho et al., 2017). At Newcastle, high denitrification rates caused by high nutrient freshwater inputs  may decouple lateral TAlk export and pyrite formation. Being located in the dry tropics, Seventeen Seventy had the lowest carbon stocks and slowest SAR, which might have favored other carbon decomposition processes over sulfate reduction. Overall, our results and analysis suggest that pyrite formation is a reasonable lower limit proxy for total TAlk production and export if sulfate reduction is the dominant TAlk producing process.
An important observation is that lateral TAlk export was ∼10-fold higher than the estimated TAlk production coupled to pyrite formation (Figure 7). This discrepancy might be a consequence of the contrasting REITHMAIER ET AL.

10.1029/2020GB006785
14 of 20 Figure 7. Mangrove total alkalinity (TAlk) export is plotted as a function of TAlk production coupled to pyrite formation. A linear regression is shown for Barwon Heads, Jacobs Well, Hinchinbrook Island, and Darwin. Export values have been retrieved from Reithmaier et al. (2020b) for the Everglades and from Sippo et al. (2016) for the remaining sites. time scales of TAlk production and export. Production of TAlk coupled to pyrite formation integrates several decades, whereas the measured TAlk export rates presented here for comparison, which were determined over two tidal cycles, have been found to vary seasonally (Ho et al., 2017;Maher et al., 2013) and over spring-neap cycles (Taillardat et al., 2018), suggesting short-term measurements likely do not capture long-term rates. This suggestion is supported by the findings of Faber et al. (2014), who estimated TAlk production associated with pyrite formation in two temperate mangroves based on external iron input. The TAlk production associated with pyrite formation was two orders of magnitudes lower than measured TAlk export in summer. The authors argued that this discrepancy is most likely caused by periods of reduction resulting in net TAlk production in summer following periods of oxidation, causing net TAlk consumption in winter.
Additionally, the discrepancy between TAlk production coupled to pyrite formation and TAlk export might be due to additional TAlk production processes, such as denitrification, manganese reduction, iron reduction, and carbonate dissolution (Krumins et al., 2013). It is also very likely that a proportion of sulfate reduction is not coupled to pyrite formation (Berner, 1984), with some oxidation of exported reduced compounds occurring outside of the mangroves, and therefore not accounted for in the lateral exchange method of Sippo et al. (2016). Consequently, future carbon sink estimates based on lateral TAlk exports should also quantify carbonate dissolution (e.g., by measuring calcium) and examine lateral TAlk export over extended time periods to provide a net estimate.
Sulfate reduction only contributes to net TAlk production if the reduced sulfur is spatially decoupled from bicarbonate ions and prevented from re-oxidizing in a manner that consumes produced bicarbonate. In addition to pyrite formation, other processes can theoretically remove sulfide from mangrove sediments and result in pyrite being an underestimate of net TAlk production rates. For example, mangroves can remove reduced sulfur by assimilating and storing sulfide in their plant tissue (Fry et al., 1982;Okada & Sasaki, 1995). To the best of our knowledge, sulfide uptake rates by mangrove has not yet been quantified. Alongi et al. (2003) compiled mangrove tree sulfur content in Avicennia marina and Rhizophora stylosa stands and found an average of 0.6 and 2.1 Mg S/Ha, which, assuming all sulfur was associated with sulfide uptake, would only increase net TAlk production by between 0.4% and 11% for our sites.
In addition to plant uptake, reduced sulfur may be removed from the sediment by outgassing as hydrogen sulfide gas (H 2 S) (Castro & Dierberg, 1987;Ganguly et al., 2018). There are limited forest-scale estimates of H 2 S emissions. However, a study in the Sundarbans mangrove forest found emissions of 0.3 g S/m 2 /y (Ganguly et al., 2018). If these emissions are representative of mangrove forests more generally, such gaseous H 2 S flux would only result in a TAlk production (unaccounted for by pyrite) of 0.1 mmol/m 2 /d an order of magnitude smaller than our pyrite-based estimates of TAlk production. Currently, sulfur removal by plant uptake and hydrogen sulfide emissions are not sufficiently quantified to constrain their effect on net TAlk production. However, this analysis suggests that their effect is likely to be minor compared to pyrite stocks.
Production and export of TAlk coupled to pyrite formation represent not only a blue carbon sink, but might also buffer coastal acidification . Conversely, when pyrite is oxidized, net TAlk production can be reversed, resulting in acidity release that offsets blue carbon sinks. At Karumba, pyrite stocks at the living area were significantly correlated to groundwater TAlk, suggesting that pyrite formation was linked to TAlk generation, whereas in the dead area, groundwater TAlk was unrelated to pyrite stocks. Sippo et al. (2020b) observed that iron in wood chronologies at the dead area rapidly increased approximately 2 years prior to the dieback. The authors suggested that a combination of low sea level and low rainfall resulted in decreased water availability, which likely caused some degree of pyrite oxidation. Our results showed that pyrite stocks were 20% lower in the dead mangrove area compared to the living mangrove area. Assuming similar pyrite concentrations in living and dead areas prior to the dieback, the dieback reversed the TAlk generation, releasing 43 mol/m 2 of acidity through the oxidation of pyrite. If pyrite was gradually lost over time, then 21 mmol/m 2 /d of TAlk was lost as CO 2 emissions between the start of the observed iron increase in wood chronologies (Sippo et al., 2020b) and our sampling. This rate might represent a minimum estimate since similar redox potentials between dead and living forests at the time of our sampling indicated that reducing conditions, not conductive to pyrite oxidation, had returned some time before our sampling campaign. Overall, this highlights that mangrove sediment disturbance leading to pyrite oxidation can reverse the TAlk carbon sink coupled to pyrite formation, releasing carbon to the atmosphere.

Global Pyrite Stocks and Resulting Alkalinity Production in Mangroves
From our data, we constructed a global model for mangrove pyrite stocks and the associated TAlk production. Although sampling covered a broad range of climatic zones, it was limited to sites in Australia and the USA. Therefore, future research is encouraged to validate our model by quantifying pyrite stocks in other parts of the world, since to our knowledge, no other mangrove pyrite stocks have been published. Despite the limitations of our model, it is a useful first-order estimate to determine the general importance of alkalinity production coupled to pyrite formation on a global level.
Our model explained 80% (p = 0.007) of the variability in the observed mangrove pyrite stocks (Table S1). By combining our model with published global data sets for key model parameters (see Section 2.4 and Supporting Information S1 for details), we estimate that mangroves store in total 2.7 Pg of pyrite globally to a depth of 1 m with an average of 197 Mg/ha (RMSE = 24 Mg/ha), resulting in a permanent TAlk production ( Figure 8). Assuming that the RMSE represents the uncertainty of the model, uncertainty ranges of the global pyrite model range from 173 to 221 Mg/ha. However, these values should be viewed with caution since more research is required to validate our model.
Currently, no high-resolution global maps for SAR in mangroves exist, therefore, the TAlk production rate is limited to an empirical calculation. Multiplying the global average pyrite stocks by the global average SAR for mangroves (0.49 cm/y) determined by Alongi (2012), and converting it to TAlk production (calculation details available in the supporting information S2), yields a total TAlk production that is coupled to global pyrite formation of 0.45 Tmol/y, assuming a conversion factor of 4 for TAlk production per pyrite formation. Using alternative conversation factors reported by the literature (3-6) results in a TAlk production range of 0.34-0.68 Tmol/y, underlining the importance of considering this uncertainty in future research. This TAlk production rate is within the broad range estimated by Hu and Cai (2011), who used sediment C/S ratios and carbon burial rates to estimate that pyrite formation in blue carbon ecosystems results in a net TAlk production of 0.1-1.1 Tmol/y. Our estimated global average TAlk production rate (3.3 mol/m 2 /y; range: 2.5-4.9 mol/m 2 /y) is equivalent to 24% (range: 18%-36%) of the global organic carbon burial rate (14 mol/ m 2 /y) within mangrove sediments (Breithaupt et al., 2012), highlighting the importance of including this mode of TAlk generation into future blue carbon budgets.

Conclusion
The majority of blue carbon research has focused on organic carbon stocks, whereas investigations of TAlk production and export are comparatively limited. By making quantitative estimates of pyrite stocks and the associated TAlk production in blue carbon ecosystems, we found that alkalinity production coupled to pyrite formation likely represents a significant blue carbon sink. However, the long-term production rates of pyrite are an order of magnitude smaller than short-term measured TAlk export rates, suggesting that the timescales of measurement, and methodology used need to be adequately integrated and further refined. While more research is required to understand the drivers and relevance of this process, we argue that pyrite accumulation should be included in the blue carbon paradigm as it represents a long-term carbon sink.