Contrasting nutrient availability between marine and brackish waters in the late Mesoproterozoic: Evidence from the Paranoá Group, Brazil

Understanding the delayed rise of eukaryotic life on Earth is one of the most fundamental questions about biological evolution. Numerous studies have presented evidence for oxygen and nutrient limitations in seawater during the Mesoproterozoic era, indicating that open marine settings may not have been able to sustain a eukaryotic biosphere with complex, multicellular organisms. However, many of these data sets represent restricted marine basins, which may bias our view of habitability. Furthermore, it remains untested whether rivers could have supplied significant nutrient fluxes to coastal habitats. To better characterize the sources of the major nutrients nitrogen and phosphorus, we turned to the late Mesoproterozoic Paranoá Group in Brazil (~1.1 Ga), which was deposited on a passive margin of the São Francisco craton. We present carbon, nitrogen and sulphur isotope data from an open shelf setting (Fazenda Funil) and from a brackish‐water environment with significant riverine input (São Gabriel). Our results show that waters were well‐oxygenated and nitrate was bioavailable in the open ocean setting at Fazenda Funil; the redoxcline appears to have been deeper and further offshore compared to restricted marine basins elsewhere in the Mesoproterozoic. In contrast, the brackish site at São Gabriel received only limited input of marine nitrate and sulphate. Nevertheless, previous reports of acritarchs reveal that this brackish‐water setting was habitable to eukaryotic life. Paired with previously published cadmium isotope data, which can be used as a proxy for phosphorus cycling, our results suggest that complex organisms were perhaps not strictly dependent on marine nutrient supplies. Riverine influxes of P and possibly other nutrients likely rendered coastal waters perhaps equally habitable to the Mesoproterozoic open ocean. This conclusion supports the notion that eukaryotic organisms may have thrived in brackish or perhaps even freshwater environments.


| INTRODUC TI ON
The Mesoproterozoic era (ca. 1.6-1.0 Ga), which was bracketed by major biological, chemical and climatic perturbations associated with global oxygenation events in the Paleo-and Neoproterozoic , was crucial for the evolution of our biosphere because it saw the appearance and initial radiation of eukaryotic life as indicated by numerous occurrences of eukaryotic microfossils (e.g. Adam et al., 2017;Beghin et al., 2017;Buick & Knoll, 1999;Javaux et al., 2001;Knoll & Nowak, 2017;Pang et al., 2020;Strother et al., 2011). These fossils tend to occur mostly in shallow marine and possibly lacustrine environmental settings, which may be due to persistently anoxic conditions at depth. Deep marine anoxia has been invoked to explain the delayed rise of metazoans (Reinhard et al., 2016), and it probably also hindered the expansion of the first more complex eukaryotic organisms. Several data sets indicate that throughout most of the mid-Proterozoic only land surfaces and the surface ocean were mildly oxygenated (Gutzmer & Beukes, 1998;Hardisty et al., 2017;Sindol et al., 2020), while the deep ocean was largely ferruginous with a low sulphate reservoir (Gilleaudeau & Kah, 2015;Kah et al., 2004;Luo et al., 2014;Planavsky et al., 2011;Poulton et al., 2010). Sulphidic (euxinic) conditions likely existed in upwelling zones along continental margins, and occasional entrainment of sulphide into surface waters may have further restricted the habitat of complex eukaryotic organisms. In a globally anoxic ocean, several essential nutrients may have been limiting, because remineralization of organic matter was suppressed, base metals were trapped in sulphide minerals, phosphorus was efficiently scavenged by ferrous iron, and nitrate was more rapidly reduced to N 2 gas (Derry, 2015;Koehler et al., 2017;Reinhard et al., 2013Reinhard et al., , 2017Stüeken, 2013). To address this question, we turned to the late Mesoproterozoic Paranoá Group in central Brazil. This unit contains interbedded stromatolitic carbonate and siliciclastic intervals in which previous workers identified diverse assemblages of microfossils, including cyanobacteria and acritarchs (Fairchild et al., 1996). Trace elements and cadmium isotope data from the carbonate units indicate deposition under an oxygenated water column . Importantly, the strata include facies from an open shelf setting that was well connected to the open ocean as well as stromatolites formed in a brackish waters with limited seawater exchange and evidence of significant riverine influence on water chemistry (Section 2). The Paranoá Group thus allows us to compare and contrast living conditions for eukaryotic life under the influences of marine and riverine input. To do so, we analysed samples of stromatolitic carbonate beds for organic carbon, nitrogen and sulphur isotopes. Samples were taken from two field localities: Fazenda Funil and São Gabriel, which are thought to represent an open subtidal and an intertidal lagoon environment, respectively .  (Alvarenga et al., 2014;Pimentel et al., 1999Pimentel et al., , 2011. Age constraints suggest a late Mesoproterozoic to earliest Neoproterozoic depositional age for the Paranoá Group based on Sr-C isotope chemostratigraphy (Alvarenga et al., 2014). The maximum deposition age is further constrained to 1.54 Ga by U-Pb dating of detrital zircons (Matteini et al., 2012). Diagenetic xenotime overgrowth of zircons, dated with Lu-Hf, indicates a minimum depositional age of ~1.04 Ga (Matteini et al., 2012). Biostratigraphy, specifically microfossil assemblages and the occurrence of the Conophyton metulum Kirichenkio in the upper Paranoá Group and point to an age bracket of 0.9-1.2 Ga (Dardenne et al., 1976;Fairchild et al., 1996). Taken together, the age is therefore best constrained to a range from 1.0 to 1.2 Ga.

| G EOLOG IC AL S E T TING
The sedimentary strata of the Paranoá Group are approximately 1000 m thick and divided into nine lithostratigraphic units that were deposited during two transgressive and one regressive cycle (Alvarenga et al., 2014;Campos et al., 2012). Most of the Paranoá Group is made up of siliciclastic rocks including sandstones, siltstonesandstone rythmites and shales. Dolo-and limestone lenses, which were the focus of this study, occur within the rythmite. These lenses are several tens of metres thick and extend over hundreds of metres along strike (Alvarenga et al., 2014). They are interfingered with pelitic and psammitic rocks, which also occur as intraclasts (Campos et al., 2012). In outcrop, the carbonate units are classed as mudstones and intraclastic grainstones, packstones and floatstones with local conical or columnar and rarely dome-shaped stromatolites. Campos et al. (2012) proposed that the carbonate lenses represent topographic highs during the time of deposition. Tidal indicators within the siliciclastic sedimentary rocks, as well as micritic rip-up clasts from some of the carbonates (Campos et al., 2012) suggest that-with important exceptions highlighted below-sediment deposition largely occurred on an open platform environment with frequent storm events and sporadic subaerial exposure (de Morrison Valeriano, 2016). The stromatolitic strata investigated in this study come from the upper part of the Paranoá Group and were deposited during the second transgression cycle (Campos et al., 2012). Samples were collected at two localities ( Figure 1 The conical morphology of the stromatolites suggests a water depths of around 20 m (Alvarenga et al., 2014;Campos et al., 2012). In contrast, the dome-shaped stromatolites and planar microbialites from São Gabriel are thought to represent shallower water depth (Campos et al., 2012). Carbonates from this setting are flat-laminated, and storm indicators are absent (Fairchild et al., 1996). The depositional setting of São Gabriel has therefore been interpreted as a restricted lagoon. Eukaryotic microfossils have been documented at this locality (Fairchild et al., 1996). The entire sedimentary package has been only weakly metamorphosed (below greenschist facies, Fuck et al., 1988).
The environmental contrast between the two settings is further supported by trace elements and cadmium isotope data from the carbonates (excluding detrital siliciclastic contributions), as shown by Viehmann et al. (2019). These data show a typical marine signature for Fazenda Funil with depletion in shale-normalized (SN) light rare earth elements (REE, Yb SN /Pr SN = 2.1-3.9), strongly superchondritic Y/Ho ratios (37.9-46.2) and fractionated Cd isotopes (ε 112/110 Cd = −3.5 to +3.8 units, where ε 112/110 Cd = [( 112 Cd/ 11 0 Cd) sample /( 112 Cd/ 110 Cd) standard −1] × 10,000, and the 'zero reference' standard is NIST SRM 3108. Note that this ε-notation can be converted to the more familiar δ-notation by dividing by 10, meaning that δ 112 Cd in these rocks falls between −0.35‰ and +0.38‰). In contrast, domal stromatolites from São Gabriel show light REE enrichment (Yb SN /Pr SN < 0.84) and weakly developed superchondritic Y/ Ho ratios (33.6-39.3), Gd SN anomalies (Gd/Gd* = 1.1 to 1.3) and near crustal ε 112/110 Cd values (−0.54 to −0.17 ε-units). We note that normalizing the REY inventory of São Gabriel stromatolites to local, ambient hinterland rocks instead of to the commonly used PAAS (Post Archean Australian Shale) results in typical seawater-like REY patterns for these rocks ; however, 'seawater-like' REY patterns can already develop in freshwater environments even with minute amount of seawater (5%-10%, Tepe & Bau, 2016). The data are therefore best explained by a scenario where the São Gabriel stromatolites grew in a setting that experienced significant riverine input from the hinterland paired with occasional flooding by seawater . Overall, these geochemical trends are thus strong evidence for two distinct water masses. In other words, the geochemistry supports the sedimentological interpretation that São Gabriel was deposited further inland, perhaps in a lagoon or estuarine setting, compared to Fazenda Funil. This study site is therefore ideal for comparing nutrient sources to the late Mesoproterozoic biosphere. We used carbon and sulphur isotopes to gain further insights into the depositional settings and nitrogen isotopes to reconstruct nitrogen metabolisms. The previously published Cd isotope data can serve as an indirect proxy for phosphorus availability (discussed below). For a detailed petrographic description of these samples,  Table 1, where secondary indices (lower to higher numbers) correspond to interior, middle and upper parts of individual stromatolites. The subsamples were cut from horizontal stromatolite slabs, using a diamond-bearing saw. Additionally, we analysed pieces of carbonate infill that had formed between stromatolite domes at the São Gabriel location ( Table 1). The stromatolite pieces were hammered into mm-sized chips on a steel plate, and the chips were washed in glass beakers with methanol (reagent grade), 1N HCl (reagent grade) and 18 MΩ cm −1 DI-H 2 O. Each solvent was applied for about 5-10 s and then decanted. The clean chips were dried in a closed oven at 60°C and then pulverized in an agate ball mill.

| Decarbonation for organic carbon, nitrogen and reduced sulphur analyses
The powder was decarbonated with 2N HCl (reagent grade) in glass centrifuge tubes at 60°C overnight. The next day, the samples were centrifuged and the acid was decanted. A few drops of fresh acid were added to ensure that the samples were no longer reactive, indicating that all carbonate had been dissolved.
The residues were then washed with 18 MΩ cm −1 DI-H 2 O three times and dried at 60°C. The dried residues were stored in scintillation vials. All glassware used during the sample preparation was pre-combusted at 500°C overnight to remove organic matter.
With regard to sulphur, this decarbonation protocol removed all carbonate-associated sulphate and any trace sulphate evaporite minerals, leaving behind organic-bound and pyrite-bound sulphide phases. We will therefore refer to it as total reduced sulphur, abbreviated as TRS in the following.

| Isotopic analyses
For the C-N-S isotope analysis, the appropriate amount of powder (2-10 mg for decarbonated residues, 0.15-0.20 mg for barite) were weighed into a tin capsule, and 0.5-1.5 mg of V 2 O 5 (Elemental Microanalysis) was added as a combustion aid. The capsules were sealed and combusted in an elemental analyser (EA Isolink, Thermo Fisher) that was coupled via a Conflo IV to a gas source isotope-ratio mass spectrometer (MAT253, Thermo Fisher). The EA was equipped with a combustion reactor filled with WO 3 and electrolytic copper, a water trap filled with magnesium perchlorate, and a ramped gas chromatograph oven, which allows measuring all three elements (carbon, nitrogen and sulphur) in one combustion (Sayle et al., 2019;Stüeken et al., 2020). Isotopic data are expressed in delta

| Sulphur cycling
Sulphur isotopes are primarily fractionated during microbial sulphate reduction under anoxic conditions, where the resulting sulphide becomes depleted in δ 34 S by up to 70 ‰ relative to sulphate (Canfield, 2001;Sim et al., 2011). If sulphate reduction occurs in sedimentary pore waters, it may become diffusion-limited and express strong Rayleigh distillation effects, especially if sulphate concentrations in the overlying water column are less than a few mM (Fike et al., 2015;Gomes & Hurtgen, 2013, 2015. Under these conditions, the residual porewater sulphate becomes progressively enriched in The two slightly higher δ 34 S CAS values of +30.1‰ and +33.5‰ may be part of the isotopic variability that is seen throughout the Proterozoic, when seawater was relatively sulphate-poor (Kah et al., 2004;Luo et al., 2014;Shen et al., 2002); however, it is also possible that these subtle enrichments reflect the effects of Rayleigh  (Gomes & Hurtgen, 2013, 2015. In Figure 3a, the δ 34 S red values from São Gabriel are thus probably best represented by cumulative sulphides at the endpoint of the xaxis (0% SO 4 2− remaining).
This interpretation would imply that the total sulphate reservoir at the São Gabriel site was smaller (and thus more rapidly con- These constraints allow for the possibility that Mesoproterozoic F I G U R E 2 Scatter plots of (a) total nitrogen versus total organic carbon; (b) nitrogen isotopes versus organic carbon isotopes; (c) sulphur isotopes versus organic carbon to sulphur ratios; (d) nitrogen isotopes versus organic carbon to nitrogen ratios. Dark grey circles = São Gabriel, light grey triangles = Fazenda Funil river water and seawater were not as different as they are today in terms of sulphate content; however, our data are best explained if a significant difference existed already at that time. Further support for this interpretation comes from organic carbon to sulphur ratios, which are known to scale with salinity, because less saline waters tend to contain less dissolved sulphate (Berner & Raiswell, 1984;Wei & Algeo, 2019). The higher C/S ratios at São Gabriel relative to Fazenda Funil (Figure 2c) thus probably point towards a less saline water column. An alternative interpretation could be that Fazenda Funil experienced water-column euxinia and is therefore enriched in S relative to TOC (Leventhal, 1983); however, this is unlikely because

| Carbon cycling
Carbon isotopes are fractionated during CO 2 fixation of primary producers where biomass becomes depleted in δ 13 C relative to the CO 2 source (Figure 3c). The magnitude of the fractionation differs between different metabolic pathways, but typically falls around −27 ± 7‰ for average marine biomass (Schidlowski, 2001). If anaerobic organisms are present, especially during diagenesis, they

| Background on nitrogen isotope fractionation
The main source of nitrogen to the biosphere is biological N 2 fixation to ammonium, which typically imparts a small isotopic fractionation of −2‰ to +1‰ (Zhang et al., 2014 (Figure 3b), possibly along a redox cline in deeper waters further offshore ( Figure 4); however, the preservation of the isotopic signature of nitrate in the rock record indicates that the residual amount of nitrate was high enough to sustain a significant fraction of the local ecosystem (Kipp et al., 2018). This residual nitrate pool would have been washed up onto the continental shelf.
Similar to sulphate, the marine nitrate reservoir with a δ 15 N of +8‰ probably also periodically spilled into the lagoon at São Gabriel. However, the comparatively lower δ 15 N values at this site (mean 3.7 ± 0.7 ‰) suggest that an additional source of isotopically light nitrogen was present. The most likely source is biological N 2 fixation, which would have contributed biomass with a composition around −2‰ to +1‰. Similar to the modern Cariaco basin (Thunell et al., 2004), mixing of biomass from N 2 -fixers and nitrate assimilators probably led to an average composition of +3‰ to +4‰. It is also possible that isotopically light nitrate was washed in by rivers, derived from oxidation of microbial mats on land (Thomazo et al., 2018  , the eukaryotic ecosystem that has been described from this locality (Fairchild et al., 1996) may have been sustained by a combination of prokaryotic N 2 fixation and possibly riverine nitrate input.

| Considering diagenesis
Diagenesis can alter sedimentary δ 15 N values by a few permil (Altabet et al., 1999;Freudenthal et al., 2001;Lehman et al., 2002;Moebius, 2013;Robinson et al., 2012), but the effect is biologically driven and largest in oxic sediments where organic ammonium is partially oxidized to nitrate. In this case, the residual ammonium becomes isotopically enriched in the heavy isotope. Under anoxic conditions, growth of N 2 -fixing anaerobes can lower the net δ 15 N value (e.g. Lehman et al., 2002). In other words, diagenetic alteration of sedimentary δ 15 N is dependent on the environment and therefore part of the environmental signature that we are trying to extract. If the samples from Fazenda Funil were affected by partial ammonium oxidation to nitrate during diagenesis and therefore isotopically enriched, it would only support our overall conclusion that nitrate was present in the water column at this site. However, the relatively lower C/N ratios and the high TN intercept at Fazenda Funil argue against significant aerobic ammonium oxidation, because this process removes organic carbon and nitrogen simultaneously. In contrast, anaerobic biomass degradation coupled to sulphate reduction oxidizes carbon to CO 2 but cannot oxidize ammonium; ammonium oxidation coupled to sulphate reduction is thermodynamically unfeasible, allowing ammonium to build up in pore waters (Stüeken et al., 2016). Subsequent adsorption of this pore water ammonium to clay minerals would thus explain the enrichment in TN in the Fazenda Funil samples (Figure 2a). At São Gabriel, where C/N ratios are higher and the TN enrichment is lower, sulphate-driven biomass degradation was likely less important, consistent with evidence for a smaller sulphate reservoir (Section 5.1) and higher rates of biomass burial (Section 5.2). Hence prior to diagenesis, δ 15 N values at São Gabriel may have been lower than what was measured, while at Fazenda Funil pre-diagenetic values were possibly slightly higher or similar to the final archived value. Diagenetic alteration does therefore not impact our overall conclusions.

| Comparison to other mid-Proterozoic basins
The isotopic properties of the two sampling sites present a selfconsistent environmental scenario for the Paranoá basin ( Figure 4): Oxic waters rich in nitrate and sulphate flushed around the stromatolites that were growing offshore at Fazenda Funil; nitrate acted as a major nitrogen source to living organisms at this site; sulphate underwent partial reduction during anaerobic diagenesis within microbial mats. In contrast, the more restricted lagoon at São Gabriel only received occasional seawater input and therefore had a limited reservoir of marine nitrate and sulphate. Sulphate was more rapidly and quantitatively consumed during diagenetic reduction, and the marine nitrate limitation was offset by biological N 2 fixation or possibly by an influx of nitrate from land. With regard to sulphur, similar basinal gradients with higher sulphate availability offshore and lower availability onshore have previously been described from other mid-Proterozoic basins. In the Roper basin in northern Australia (~1.4 Ga), δ 34 S values of reduced sulphur are depleted down to −20‰ in deep basinal facies and enriched up to +50‰ on the inner shelf ( Figure 5) (Shen et al., 2003). In the Taoudeni basin in NW Africa (~1.1 Ga), values within a euxinic wedge offshore along the coast (within the chemocline) drop down to around −20‰ while values from intermittently euxinic deeper waters range from −5 to +35 and samples from shallow waters on the shelf plot between 0 and +20 (Figure 5b) (Gilleaudeau & Kah, 2015). In both cases, the very light values were interpreted as evidence of sulphate reduction within the water column, while the heavy values likely suggest diagenetic sulphate reduction under closed-system conditions. Water-column sulphate reduction is not observed in the Paranoá Group, but the relative enrichment in 34 S at both Fazenda Funil and São Gabriel is comparable to the enrichments seen in non-euxinic facies in the Roper and Taoudeni basins and probably explained by the same mechanism of diffusion-limited diagenetic sulphate reduction within sediments.
However, the nitrogen data from the Paranoá Group differ somewhat from those of other mid-Proterozoic basins (Table 2).

F I G U R E 5
Comparison of sulphur isotope data between the Paranoá Group in Brazil (this study), the Taoudeni Basin in NW Africa (Gilleaudeau & Kah, 2015) and the Roper Group in northern Australia (Shen et al., 2003) TA B L E 2 Comparison of δ 15 N values between Mesoproterozoic basin.   Supergroup (1.2 Ga) typically fall around +3‰ to +4‰ in intertidal facies and between 0‰ and +2.5‰ in subtidal facies (Hodgskiss et al., 2020;Koehler et al., 2017;Stüeken, 2013). Subtidal sedimentary rocks from the Vindhyan Supergroup (+1.8‰, 1.05 Ga) also fit into this pattern (Gilleaudeau et al., 2020). Sedimentary lithology is unlikely to play a role, because nitrogen is primarily introduced into sediments via burial of organic matter, independently from the carbonate content. As noted above, some ammonium partitions into clay minerals during diagenesis, but the isotopic fractionation associated with that is small. Hence, the ratio of clays to carbonates is not expected to impart a systematic change on the preserved isotopic value. Therefore, this basinal gradient of nitrogen isotope ratios has been interpreted as evidence that nitrate bioavailability was restricted to shallow waters during the mid-Proterozoic, imposing an additional throttle on the radiation of eukaryotic life (Koehler et al., 2017;Stüeken, 2013 (Schmitt et al., 2009). Cadmium concentrations and its isotopes are gaining increasing traction in biogeochemistry, because they can indirectly record the availability of the macronutrient phosphorus (Abouchami et al., 2014;Boyle et al., 1976;Gault-Ringold et al., 2012;Xie et al., 2019). It has been shown that in highly productive surface waters of the modern ocean, the uptake of Cd by phototrophic organisms leads to an overall depletion of seawater in dissolved Cd concentrations and a shift in isotopic composition of residual Cd to higher values (δ 112 Cd of +0.3‰ to +0.5‰) due to the preferential (biological) uptake of the light isotope species into biomass (Conway & John, 2015;Guinoiseau et al., 2019;Lacan et al., 2006;Xie et al., 2019). Under nutrient limited conditions, the remaining dissolved Cd pool becomes isotopically enriched due to Rayleigh fractionation, and this residual enriched Cd reservoir can be preserved in shallowwater carbonates (Hohl et al., 2017). A regeneration of organic-  (Table 1). This approach was taken, because different amounts of sample material were used for the study of Viehmann et al. (2019) and this study; however, samples of both studies originate from the exact same hand specimens and region of the various stromatolites.
The In contrast, at the São Gabriel locality, authigenic δ 112 Cd values plot close to crustal values, reflecting a strong influence by riverine nutrient influx that was not significantly affected by subsequent isotopic fractionation. However, we know from the carbon isotope data (Section 5.2) that this setting experienced significant biological activity. Hence, the lack of isotopic fractionation in Cd is unlikely to reflect limited primary productivity. Instead, it is more likely that the dissolved Cd reservoir was too large (and too frequently refreshed) to experience significant isotopic alteration by partial biological uptake of Cd. If correct, this interpretation would imply that other major nutrients, particularly P, whose behaviour appears to mirror that of dissolved Cd in the modern ocean (see above), were relatively abundant in the São Gabriel lagoon. Considering the significant freshwater input to São Gabriel, as indicated by REY data  and sulphur geochemistry (this study), our results therefore suggest that freshwaters were relatively enriched in P, potentially more so than the Mesoproterozoic open ocean, which is thought to have been P-depleted (Reinhard et al., 2017). We stress that the Cd isotope proxy is still in its infancy and more data are needed to corroborate its applicability to Proterozoic settings.
However, our conclusion is consistent with modelling studies suggesting that Precambrian river waters were P-enriched compared to today and compared to Precambrian seawater (Hao et al., 2017(Hao et al., , 2020. Brackish and freshwater settings may therefore have offered important niches for early life.

| CON CLUS IONS
The Paranoá Group presents itself as an open marine shelf that was probably well-oxygenated. Sulphate reduction appears to have been limited to sedimentary pore waters, and nitrate was sufficiently bioavailable to sustain a large proportion of the ecosystem. This F I G U R E 6 Average Cd isotopic compositions obtained on carbonate leachates compared to average nitrogen (a) and sulphur isotopic compositions (b) obtained on same stromatolite hand specimen (error bars represent 1SD within subsamples). Light grey triangles represent Fazenda Funil, and dark grey circles represent Saõ Gabriel environment was thus hospitable to eukaryotic life, as supported by the presence of eukaryotic microfossils. However, the Paranoá Group provides two important nuances to our view of eukaryotic habitats: First, the passive margin environment captured by these rocks may have been better mixed and perhaps more deeply oxygenated than restricted basins; and second, brackish waters such as those represented by São Gabriel were probably equally habitable to complex organisms, as evidenced by eukaryotic acritarchs at this site (Fairchild et al., 1996). Phylogenetic data suggest that some of the first eukaryotes thrived in non-marine environments during the mid-Proterozoic (Sánchez- Baracaldo et al., 2017), but this hypothesis has been difficult to test with geological samples because of the low preservation potential of non-marine sedimentary rocks (Peters & Husson, 2017). The Paranoá cannot fill this gap, but it demonstrates that settings with only intermittent seawater input were in-

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in this manuscript are presented in Tables 1 and 2