Jamaica Bay is a heavily urbanized estuary (∼80 km²) located primarily between the New York City boroughs of Brooklyn and Queens, with a connection to the Atlantic Ocean through Rockaway Inlet (Figure 1). The Bay, consisting of over a dozen isolated marsh islands and a labyrinth of waterways, has been characterized as a eutrophic estuary with water salinity in the range of 20–26 PSU, temperature 1–26°C, and pH 6.8–9, respectively [USFWS (U.S. Fish and Wildlife Service), 1997]. Some of the marsh islands are natural, but many of them have been restored through engineering efforts. The dominant plant species in the low salt marsh is saltmarsh cordgrass, S. alterniflora, and in the high salt marsh is salt meadow cordgrass, Spartina patens. These salt marshes provide critical ecological services, including habitat and food sources for wildlife, shoreline erosion control, and water column filtration, as well as serve as a buffer against storm tides and waves (New York City Department of Environmental Protection, 2007a, b; Marsooli et al., 2017). However, Jamaica Bay’s salt marsh losses have been severe; about 60% of the Bay’s salt marsh has converted into mudflats since 1951, with smaller islands losing up to 78% of their vegetation cover, due to a combination of factors including sediment load reduction, dredging, boat traffic, sea level rise, nutrient enrichment and H₂S concentration increase (Rafferty et al., 2010; Campbell et al., 2017). The nutrient load of the Bay from sewage treatment plant effluent and combined sewer overflows is thought to be an important factor that may contribute to this high rate of marsh loss in recent years (Deegan et al., 2012).

Jamaica Bay has a mean semi-diurnal tidal range of approximately 1.5 m, the investigated marsh sites are all inundated during high tide. Two marsh sites were studied in the present work: one is a natural salt marsh, JoCo (JC), which is considered a “healthy” marsh, and the other is a partially restored salt marsh, Big Egg (BE). Three sampling sites were established in each marsh. One of the three sites at Big Egg (BE1) was previously restored by spraying sandy sediment from the adjacent channel onto the marsh surface as part of a restoration effort in 2003 (Frame et al., 2006), but the remaining two sites (BE2 and BE3) have not yet been restored. The two marshes are dominated by S. alterniflora, while JoCo has a mixture of marsh grasses that includes Spartina patens, a high marsh species. JoCo experienced a marsh loss of approximately 7% between 1974 and 1999, while losses at Big Egg were up to 38% (NYSDEC, 2006).

Pore water samples, solid marsh cores, and sensor deployment in JoCo were taken on October 10, 2014, May 1, 2015 and September 9, 2015 and April 22, 2016, respectively. In Big Egg sites, they were taken on October 8, 2014, April 13, 2015, September 25, 2015, and April 15, 2016, respectively.

Discrete pore water samples were taken using pore water “sippers” at depths of 5, 10, 15, and 25 cm in the marsh peat. The sippers are hollow acrylic rods that end in a small opening. Each sipper is connected to Tygon tubing which can be connected to a 50 ml plastic syringe. A valve connected to the Tygon tubing between the syringe and the sipper facilitates purging of the syringe. Pore water samples of ∼50 ml were slowly drawn into the syringe through the sipper, and immediately filtered in the field through 0.45 μm filters. Aliquots for dissolved sulfide, nutrients, and trace elements were collected. The sulfide aliquots were fixed in the field by adding 0.5 ml of 0.05 M Zn(C₂H₃O₂)₂.2H₂O to each sample, the Fe²⁺ aliquots were stabilized by acidification. The concentrations of dissolved sulfide, nutrients (NH₄⁺ and HPO₄^2–) and dissolved Fe²⁺ in these discrete pore water samples were subsequently measured in lab.

Total dissolved sulfide (as ΣH₂S = [H₂S] + [HS^–] + [S^2–]) and ammonium were measured using techniques described by Kolker (2005) and Cochran et al. (2013). Briefly, total sulfide was measured by spectrometric method (Cline, 1969; Reese et al., 2011) with relative standard deviation (RSD) < 5%. Ammonium and phosphate were measured using a Lachat Nutrient Autoanalyzer with RSD ± 5%, and dissolved Fe²⁺ was determined by spectrometric method described by Stookey (1970). Pore water pH and E_h (ORP; Oxidation-Reduction Potential) were measured in the field with a YSI 1009-1 09C multi-parameter probe that measured pH on the total H⁺ scale and used a Pt electrode for E_h. Salinity values were determined in the laboratory through measurement of chloride or with a refractometer.

2-D distributions of dissolved Fe²⁺ in pore water were measured in situ by deploying optical planar Fe²⁺ sensors in salt marsh. The irreversible planar sensor of dissolved Fe²⁺ was fabricated with ferrizone as the optical indicator that was immobilized in polyurethane hydrogel (D4) membrane, modified from our previous publication (Zhu and Aller, 2012). The blank Fe²⁺ sensor film was colorless and transparent, and it rapidly converted to a violet-red color with maximum absorption wavelength at 562 nm after the sensor film was exposed to dissolved Fe²⁺ solution. The sensor response range depends on the sensor deployment time in Fe²⁺ solution, and the sensor showed a good linear response in the range of 0–200 μmol/L of Fe²⁺ with the limit of detection of 5 μmol/L when a 10 min deployment time was applied. A short sensor deployment time was used when Fe²⁺ concentration was higher than 200 μmol/L, and the sensor could respond to dissolved Fe²⁺ as high as 1 mmol/L when it was deployed in the sample for 1 min. In this work, 1–10 min Fe²⁺ sensor deployment time in salt marsh was used, depending on the Fe²⁺ concentration in the marsh pore water.

The sensor film sheet (14 cm × 20 cm) was cut into five sensor film stripes with dimension of 2.8 cm × 20 cm, four strips were used for in situ deployment and one for calibration. The sensor film strip was mounted on a 5 cm × 40 cm × 1 cm (width × length × thickness) polycarbonate plastic sheet with a beveled end. In situ measurements were performed at low tide by inserting/punching the plastic sheet with attached sensor strip vertically into the salt marsh peat (Figure 2) and allowing it to react with dissolved Fe²⁺ in pore water for 1–10 min. After the reaction, the sensor film was pulled out of the marsh and rinsed with seawater. The color-developed Fe²⁺ sensing film was then wrapped with a paper towel and brought back to the laboratory for imaging. In situ measurements were duplicated at each marsh sample site. After return to the laboratory, the sensor films were calibrated with identical response time by using Fe²⁺ standards prepared in seawater which was collected from the same sample site. The colored Fe²⁺ sensor strips were imaged by a flat scanner (Canon), and all color images were analyzed with Maxim DL image processing software version 2.0X (Diffraction Limited) and Image-Pro plus version 4.1 for Windows (Media Cybernetics). The isolated green bands of the color images were used to calculate the absorbance within individual pixels. In situ measurements of 2-D Fe²⁺ distributions in pore water were conducted at October 2014, April 2015, and September 2015.

2-D distributions of dissolved H₂S in salt marshes were measured by two different optical planar H₂S sensors. The 2-D H₂S distributions in the October 2014 samples were measured by a fluorescence planar sensor in box cores which were collected from each site and incubated/re-equilibrated with seawater (collected from the same site) in the laboratory for 24 h at room temperature (22°C) with constant aeration of overlying water. The fluorescence H₂S planar sensor sheet was prepared by non-covalently immobilizing H₂S fluorescence indicator pyronin in an ethyl cellulose polymer membrane on a transparent polyester sheet, and coating with a layer of gas permeable silicone (Zhu and Aller, 2013). The fluorescence sensor responded well to dissolved H₂S in the range of non-detectable to 3.15 mmol/L with detection limit 40 μmol/L dissolved H₂S. The sensor response time was about 60 s, and it was very suitable for H₂S quantification in highly sulfidic salt marshes. However, it was only a moderately reversible fluorescence sensor, losing its response after 5–6 measurement cycles (Zhu and Aller, 2013). The sensor film was cut into the strips of 3 cm × 15 cm for the following 2-D measurement: an H₂S sensor strip was installed on the inside of the front face of the box corer and sensing membrane contacted to the marsh sediment. The bottom of the box core was tightly sealed. After 24 h incubation, the fluorescence image of the H₂S sensor strip was taken at 577 nm (±10 nm) with excitation wavelength at 554 nm by using our home-made imaging system (Zhu and Aller, 2013). Image analysis and data calculations were performed with Maxim DL image processing software version 2.0X (Diffraction Limited) and Image-Pro plus version 4.1 for Windows (Media Cybernetics). Images were split into blue, green and red bands and the red band was used to calculate the intensities of individual pixels.

2-D H₂S distributions in other seasons, April 2015 (spring) and September 2015 (late summer), were obtained by in situ deployments using an irreversible H₂S colorimetric planar sensor that was prepared from diphenylcarbazone-Zn²⁺ complex in polyurethane hydrogel (D4) on a transparent polyester sheet, covered by a gas permeable silicone membrane to eliminate possible interfering hydrated ions. The blank sensor film showed a dark purple color with a maximum absorbance at 530 nm, and the sensor absorbance was inversely correlated with dissolved H₂S concentration in the range of 5–4,000 μmol/L (Yin et al., 2017). A 3.5 cm × 20 cm H₂S sensor film was mounted on one side of a 5 cm × 40 cm × 1 cm (width × length × thickness) polycarbonate plastic sheet with a beveled end, and the deployment method was the same as 2-D Fe²⁺ in situ measurement as described above. Response time of the H₂S sensor in marsh was 1–10 min (depend on the H₂S concentration in the marsh). After the reaction, the sensor film was pulled out of the sediment and rinsed with seawater. The color-changed sensing film was scanned within 20 min in the field by using a flat scanner (Canon). After return to the lab, the H₂S sensor responses were calibrated for identical response times by using H₂S standard solutions prepared from sodium sulfide in pH < 4 solutions. All color images, including sample sensing images, were analyzed with Maxim DL image processing software version 2.0X (Diffraction Limited) and Image-Pro plus version 4.1 for Windows (Media Cybernetics). The isolated green bands of the color images were used to calculate the absorbance within individual pixels.

Cores for solid phase geochemistry (sulfur and iron) were taken by carefully inserting an aluminum tube (i.d., 7 cm) into the marsh peat. After the tube was positioned on the marsh surface, vertical cuts were made around the perimeter to minimize compaction as the core was inserted. Immediately after return to the laboratory, the cores were frozen. They were later defrosted only enough to permit the sediment to be extruded and then were sectioned into 1–2 cm intervals. The JoCo sediment cores had more peat and roots whereas the Big Egg sediment cores had fewer large roots and more mud. Small aliquots of sediment were removed from each section for solid phase geochemistry and the remainder was weighed, dried, weighed again to determine water content and then ground to a powder. Solid phase reactive iron was measured by leaching ∼50 mg of dried sediment in 1 N HCl for 24 h at room temperature. Total Fe in the leach solution was measured colorimetrically using the ferrozine method (Stookey, 1970). Sulfur was measured in two pools—chromium reducible sulfides (CRS, dominantly pyrite, and FeS₂) and total sulfur, as described in Kolker (2005) and Cochran et al. (2013).

Salt marshes are sites of intense biogeochemical reactions involving anoxic organic matter remineralization, which typically result in dramatic changes in the concentration of H⁺, Fe²⁺, H₂S, and other pore water species in the upper layer of marsh sediments (Morse et al., 1987; Canfield et al., 1992; Brendel and Luther, 1995; Bull and Taillefert, 2001; Zhu and Aller, 2013; Yin et al., 2017; Koop-Jakobsen et al., 2018). Compositional changes of Fe²⁺ and H₂S with depth in sediment pore water are usually assumed to occur in an overall average vertical progression. Significant heterogeneity and complex three dimensional reaction patterns over mm to cm scales can result, however, from bioturbation activity (Zhu and Aller, 2013; Yin et al., 2017). In salt marshes, the growth of wetland plants and oxygen transport through the aerenchyma lacunae of plants from aboveground sources to rhizosphere can also create a complicated thin layer of aerobic environment around roots through the radial oxygen loss (ROL) from roots (Armstrong, 1979; Howes and Teal, 1994; Colmer, 2003; Han et al., 2016; Koop-Jakobsen et al., 2018). Most of the selected study sites in Jamaica Bay are dominated by saltmarsh cordgrass S. alterniflora, in which the roots extensively intergrow into a dense network over at least the top 30 cm of the peat. Thus, it is expected that the 2-D distribution patterns of Fe²⁺ and H₂S in marsh peat are influenced by the plant root growth patterns. We hypothesize that the oxygen transport through plant rhizosphere and the radial oxygen loss can generate a time-dependent 3-D distribution pattern of low level dissolved H₂S and Fe²⁺ (cold-spots) associated with individual plant roots in marsh peat.

In order to test this hypothesis and study Fe²⁺ and H₂S distributions and the associated biogeochemical processes in marsh peat with a high density of living plants, box cores made of clear acrylic plastic were collected from each site in October 2014 and the H₂S sensor deployments were conducted in the laboratory so that the plant root features and sensor foil deployment position can be directly seen from side of the core. The 2-D H₂S distribution measured by a reversible H₂S fluorescence planar sensor was compared to the visible image of peat core (side view). The green color of the visual images (Figures 3A, 4A) was from the visual light source (white is not available in the image system used). The corresponding 2-D H₂S images in Figures 3A, 4A (shown as pseudo-color, reflecting concentrations) revealed that the concentration distribution patterns were directly associated with the plant root structures in marsh peat. The H₂S concentrations around the roots were <1 mM which was five times less than in the surrounding peat. The results in Figures 3A, 4A were measured in the laboratory in incubated box cores with aeration of overlying seawater for 24 h at room temperature without drainage, the changes of environmental conditions could alter oxygen transport and radial oxygen loss in marsh (Howes and Teal, 1994; Colmer, 2003), as well as H₂S and Fe²⁺ distributions around the roots. In order to avoid these possible artifacts, the real-time 2-D H₂S and Fe²⁺ distribution patterns in salt marshes were studied by deploying the irreversible H₂S (Yin et al., 2017) and dissolved Fe²⁺ (Zhu and Aller, 2012) colorimetric planar sensors in situ at different seasons. The real-time data in Figures 5–8A,C showed that the distribution patterns of H₂S and Fe²⁺ in sulfidic marsh pore water were significantly complicated by belowground radial oxygen loss from roots into the surrounding marsh peat. The oxygen leaked from the roots formed a thin oxic layer in which the dissolved H₂S and Fe²⁺ were oxidized to sulfate and Fe-oxide (e.g., iron plaque), resulting in a thin zone of low levels of H₂S and Fe²⁺ surrounding the roots in the sulfidic peat. Furthermore, oxygen and/or nitrate, rather than sulfate, served as the electron acceptors for organic matter remineralization in the thin oxic layers. Such an interface of oxic-anoxic sediment around the roots can be seen in both visible peat images and 2-D H₂S images, indicating heterogeneous remineralization patterns. The low and/or undetectable spots/tracks (cold spots) of H₂S and Fe²⁺ associated with root structures in Figures 3–8A,C was a direct evidence for our hypothesis, and the cold spots of H₂S and Fe²⁺ surrounding individual roots became more pronounced (Figures 7, 8) when more oxygen transport and high oxygen leakage occurred in the summer season (Colmer, 2003; Soana and Bartoli, 2013).

H₂S is a phytotoxin to marsh plants, with a the threshold that can be harmful to S. alterniflora of about 2 mM [see Kolker (2005) and references therein]. Long-term exposure of the marsh plants to high levels of H₂S can cause plant die-off and marsh peat collapse. We noted that the average H₂S levels at JoCo marsh could be >4 mM just below the marsh surface (Figures 4A,B) or in deep peat (Figure 7A, JC1) in the fall and summer seasons, but JoCo is considered as a “healthy” marsh and plants grow well. We hypothesize that the oxic layers formed around roots of S. alterniflora help the plants survive in the high levels of H₂S by reducing sulfide absorption. We tested this hypothesis at the National Synchrotron Light Source-II (Brookhaven National Laboratory; Feng et al., 2018). We used the Hard X-ray Nanoprobe Beamline to obtain nanometer-scale measurements of trace elements in S. alterniflora root tissue. The results showed that when iron concentrations in pore water and root epidermis were high, the root epidermis showed lower concentrations of S and P even though sulfide and phosphate in the pore water were high in late summer (Feng et al., 2018). At the same time, pore water Fe was lower than in the spring sampling, but Fe in the root epidermis was high. This pattern is consistent with the roots producing an oxic microenvironment with oxygen transported into the peat through the roots, such that Fe²⁺ in the pore water is oxidized, producing iron “plaque” (i.e. FeOOH) on the root epidermis. This plaque prevents sulfide from entering the root tissue. Phosphate may also be excluded via adsorption onto the FeOOH or formation of an iron phosphate phase.

There are also many “hot spots” on the 2-D Fe²⁺ and H₂S images, reflecting the heterogeneous distributions of labile organic matter and the microniches of exoenzymes and microbes on sediment particles (Cao et al., 2013). We also noted that large variation of H₂S and/or Fe²⁺ distributions may occur in same site in duplicate in situ measurements over distances of <1 m between two deployed sensor sheets.

In addition to the microscale heterogeneities of Fe²⁺ and H₂S distributions around plant roots, the 2-D H₂S and Fe²⁺ images also showed clear vertically and laterally heterogeneous distribution patterns. Note that no infaunal burrows were found in the measurements, suggesting that bioturbation is not a factor in these sediments. H₂S concentrations near the surface of the Big Egg sediment were relatively low but sharply increased below the water-sediment interface and reached maxima at ∼2 cm in the fall sampling at BE1 and BE3. The H₂S levels at JoCo marsh were elevated to 4–6 mM just below the marsh surface (Figures 4A,B). It should be pointed out that the 2-D H₂S distributions in the fall 2014 season (Figures 3A, 4A) were incubated and measured in laboratory with the box core incubation without water drainage, but in all other sampling seasons, the H₂S measurements were performed in situ in the field. The real-time 2-D H₂S distributions (Figures 5A, 6A, 7A, 8A) showed that H₂S levels in top 4 cm were low even in the summer season, and gradually increased to higher concentrations at depth. The high H₂S levels elevated just below the water-peat interface in Figure 3A (BE3) and Figure 4A (JC) were not found in the 2-D in situ measurements in the field. This phenomenon is likely caused by the pore water drainage in the marshes. The marshes of Jamaica Bay are periodically (tidally) submerged by seawater. An important pathway by which this water drains from marsh islands such as Big Egg and JoCo is vertically, and the two sites have distinctly different drainage velocities of 7.9 and 25.9 cm/d, respectively, as determined by Ra isotopes (Tamborski et al., 2017). Thus the concentrations of H₂S (and other solutes) in marsh peat pore water result from a balance between the rates of biogeochemical processes that produce or consume them and drainage. In summer and fall seasons, higher H₂S concentrations in JC marsh were produced due to the high sulfate-reduction bacteria activities and the relative high organic carbon concentration in this site (Table 3), but the more rapid drainage there produced high flow-through fluxes of H₂S through the marsh peat (Tamborski et al., 2017). As well, the effects of roots on the 2-D H₂S distribution patterns, are especially evident at JC in the summer, a time when marsh plant growth is dense and bacterial activity is high (Figure 7A).

2-D Fe²⁺ distributions were measured only by in situ sensor deployments, so visible images of the sensing marsh peat were not available. The spatially heterogeneous 2-D Fe²⁺ distributions also showed sharp vertical gradients that were closely correlated to the H₂S vertical distributions, especially in the low drainage BE marsh peats. Fe²⁺ concentration sharply increased just below the water-sediment interface and reached Fe²⁺ maximum zone at 1–2 cm deep. At depths below which the reactive particulate Fe-oxide had been depleted, sulfate reduction dominated the metabolism, producing dissolved sulfide species (H₂S, HS^–, and S^2–), and an increase of H₂S with depth in pore water below the Fe²⁺ maximum zone (Bull and Taillefert, 2001; Jørgensen and Kasten, 2006; Johnston, 2011). The free sulfide subsequently scavenged the dissolved Fe²⁺ to form the reactive solid phase FeS, resulting in a Fe²⁺ maximum above the H₂S maximum zone in pore water (Bull and Taillefert, 2001). FeS is unstable and converts to the more stable form pyrite (FeS₂) by a variety of pathways (Howarth, 1979; Berner, 1984).

The in situ zonations of 2-D Fe²⁺ and H₂S distributions in Figures 3–8 clearly showed this sequence: the maximum zones of Fe²⁺ were generally produced at 2–4 cm, and the H₂S levels increased to maxima below 4 cm. The in situ heterogeneous 2-D Fe²⁺ and H₂S distributions are useful for determining redox zonation and understanding the dominant biogeochemical processes occurring within the marsh peat. Both dissolved Fe²⁺ and H₂S can be efficiently removed by FeS₂ precipitation from pore water. When millimolar levels of H₂S occurred in marsh pore water, the dissolved Fe²⁺ concentrations were generally <20 μM at the same depths. On the other hand, the high pore water concentrations of Fe²⁺ were linked to low H₂S levels in all seasons. There were some Fe²⁺ “cold spots” co-distributed with low levels of H₂S, as seen in Figures 7A,C. As noted above, these cold spots are likely produced in the oxic and suboxic layers around roots and caused by O₂ transport down to the deep sediment through the rhizosphere. Unfortunately, H₂S and Fe²⁺ concentrations could not be simultaneously measured by optical sensor in the present study. Fe²⁺ concentrations at JC and BE were in the range of 10–100 μM, which is typical of coastal marsh pore water. High levels of Fe²⁺ in pore water may be harmful to plants because it may co-precipitate or adsorb the nutrients ammonium and phosphate in the marsh, preventing their uptake by the roots.

H₂S distributions in marsh pore water at all sites varied seasonally, generally with higher H₂S levels in summer and fall, and lower or undetectable levels in spring (Table 2 and Figures 3–8). At the JoCo marsh sites, dissolved H₂S in pore water was <0.02 mM found by sensors in the spring sensor deployments but it was elevated to as high as 3–6 mM in summer and fall at the same sampling sites. Thus was likely due to loadings of labile organic matter on the marshes produced in the summer in the eutrophic Bay, coupled with the temperature-dependent variation of the rate of microbial decomposition of organic matter. The discrete pore water samples showed the redox potential at JoCo from 5 to 15 cm was in the range of −41 to −200 mV in spring while it was lower, −180 to −350 mV, in summer and fall seasons, indicating less “reducing” environments in the marsh peat in the spring (Table 1). Our previous study showed that the exoenzymes and microbes in marine sediments have high activities in the summer and fall, but very low in winter and early spring (Cao et al., 2013). Thus, the temperature dependence of H₂S distributions was a direct reflection of the change in sulfate-reduction microbial activity with temperature. However, data in Table 1 also showed a clear spatial heterogeneity of the redox potential at different marsh sites in the spring season. Relative lower redox potentials were found at Big Egg marsh sites in the spring compared to fall, the reason for this unusual phenomenon was not clear, but the stronger “reducing” conditions at BE resulted in high H₂S levels in the late spring (April-May) (Figure 6A), and as a consequence, likely caused more plant die-off and marsh loss. H₂S distributions at BE sites in winter and early spring were not measured.

Typically, the high concentration of H₂S (or ΣH₂S) in anaerobic salt marshes comes from the reduction of sulfate during organic matter decomposition, generating a decreasing redox potential with depth (Hambrick III, DeLaune and Patrick, 1980), and a complicated 3-D pattern associated with the roots of plants due to oxygen transport and radial oxygen loss through roots. The redox potential can be used as an indicator of the degree of oxidation of marsh peat. The results in Table 1 showed that the redox potentials of all sites generally decreased from 5 to 25 cm depth, but exhibited a clear seasonal variation. The redox potential in spring was much higher than that in summer and fall in JC, indicating the lower organic matter oxidation rate by sulfate in spring. The pH values of all sites and depths in summer and fall seasons are in the range of pH 6–7 with vague seasonal variations, implying that the intense oxidation of organic matter by various electron acceptors at warm temperature tend to buffer sediment pore water close to 6–7. Interestingly, the pH values of pore water at JC in spring are in the range of 5–6.5 which are lower than the pH values in other seasons. By integrating other geochemical parameters found at this site in spring, for example the high redox potentials (Table 1), extreme low total sulfide (Table 2) and non-detectable (<0.02 mM) dissolved H₂S (Figure 5A), we conclude that the rate of organic matter decomposition in JC in late spring (April–May) is still high, but oxidants with high redox potentials (such as O₂, nitrate, and Mn/Fe-oxides), rather than sulfate, dominate the redox reactions with organic matter at JC. This might be attributed to the healthy and high-density biomass of S. alterniflora and fast pore water drainage velocity in JC. The intertidal nature and fast drainage of the JC salt marsh could result in surface gas exchange playing a more important role in the sediment oxygen transports than in low drainage marshes like BE, resulting in a relatively higher redox potential in JC than BE in the spring season.

Fe²⁺ concentrations in marsh pore water also show seasonal changes (Table 2 and Figures 3C–8C), but the seasonal changes are not very significant compared to the H₂S distribution variations, partially due to the low Fe²⁺ concentrations in the marsh pore water. But, the distribution patterns of Fe²⁺ are closely associated with H₂S distribution patterns in marsh pore water, as discussed above. It seems that Fe²⁺ concentrations in pore water at most sites of JoCo and Big Egg were relatively high in summer than in spring and fall.

The seasonal variation of H₂S and Fe²⁺ in marsh pore water should also be tightly linked to the plant life cycle. The previous studies have shown that the temporal changes in belowground biogeochemical processes, such as oxygen leakages from roots, sulfate reduction, sulfide oxidation, Fe²⁺ oxidation and precipitation etc., were well correlated with the changes in plant physiology (Hines et al., 1989; Soana and Bartoli, 2013; LaFond-Hudson et al., 2018). We did not measure directly the oxygen leakage from roots in this study, but the microscale “cold spots” of H₂S and Fe²⁺ in the 2-D distribution patterns associated with individual roots (Figures 3–8A,C) clearly indicated radial oxygen loss from roots. In the spring season, low temperatures decreased oxygen consumption in plant rhizosphere by slowing respiration in roots (Armstrong, 1979), resulting in more “excess” oxygen leakage into surrounding marsh peat and the increase of redox potential in sediment. This could be another possible explanation as to why the redox potential in JC marsh is higher in spring than other seasons (Table 1). In the summer season, dissolved organic matter could be released from the roots during active growth of S. alterniflora and rapidly enhance the sulfate reduction rate (Hines et al., 1989). We believe that any additional dissolved organic matter released from roots would also fuel the rate of Fe-oxide reduction, resulting in a relative high dissolved Fe²⁺ in marsh pore water (Figures 7C, 8C). As a consequence, more active FeS was formed and accumulated in summer season in JC (Table 3).

The precipitation of the solid phases FeS and FeS₂ (pyrite) as a result of elevated concentrations of H₂S and Fe²⁺ in the pore water effectively removes H₂S from solution, and is thus a geochemical means of controlling H₂S concentrations (Howarth, 1979; Berner, 1984). The pore water data described in sections “Heterogeneities of Fe²⁺ and H₂S Distributions in Salt Marsh Pore Water” and “Seasonal Variation of H₂S and Fe²⁺ Distributions in Marsh Pore Water” are essentially snapshots of conditions at the time of sampling. The composition of the solid phase can integrate the redox reactions occurring in the sediments over longer periods. We measured fractions of the solid phase S and Fe pools in the peat at all the sites (Table 3). For sulfur, the important pools include Acid Volatile Sulfide (AVS), CRS, and Total Sulfur. AVS is a measure of FeS-associated sulfur, CRS of the pyrite-associated sulfur and total S of the AVS, CRS, and all other forms, including organosulfur compounds. For iron, we measured a fraction termed “reactive” Fe, i.e. that obtained by a 1 N HCl cold leach of the dried sediment. This procedure extracts the portion of Fe that is diagenetically mobile and readily able to react with sulfide to produce the iron sulfide phases. Our prior experience in JoCo (Cochran et al., 2013) suggests that AVS < CRS ≤ Total S. The AVS pool is labile and converts to pyrite, which is a longer term sink for Fe and especially S in the marsh peat. In effect, the formation of pyrite removes dissolved sulfide from the pore water and stores it, lowering the exposure of the plants to this phytotoxin. The solid phase measurements enable calculation of a parameter termed the “Degree of Pyritization” (DOP), that is, the degree to which reactive Fe in the solid phase is associated with pyrite and is thus unable to further remove dissolved sulfide from pore water. DOP is defined as:

where Fe_pyrite is calculated as CRS/2 (i.e. one mole of Fe per two moles of S in FeS₂), and Fe_reactive is the acid leachable Fe. Table 3 gives the solid phase DOP results.

The results in Table 3 show that the total sulfur at JoCo is generally higher than that at Big Egg, but the CRS-sulfur concentrations are comparable at these two sites and the fraction of CRS in the total sulfur pool is much smaller in JoCo. To the extent that total sulfur includes diagenetically produced organosulfur compounds and sulfur associated with the marsh plants, this difference can be explainable by the higher organic content of the marsh sediments at JoCo. Both total sulfur and CRS-sulfur show vertical variations at different depths from surface to deep marsh peat, but the pattern at each site is not consistent even in the same season, implying significant heterogeneities of both organosulfur and pyrite-associated sulfur in the solid marsh peat, probably caused by the plant growth. Similar to dissolved H₂S and Fe²⁺ in pore water, the total sulfur, CRS-sulfur and acid leachable Fe in the peat also show seasonal differences between the summer 2015 and spring 2016 samplings. These reflect the redox cycling of the various pools of Fe and S over the year.

The DOP at JoCo is higher than that at Big Egg. Indeed, the DOP values at JoCo can approach 1, indicating that 100% of the reactive iron is associated with pyrite. These high values suggest that the biogeochemical control of pore water H₂S via FeS₂ precipitation is close to a limit and that high sulfide but low Fe²⁺ concentrations in the pore water can be expected, as observed.

Our group has studied Jamaica Bay marshes since 2004 (Cochran et al., 2013, 2018). In particular, we have pore water ΣH₂S concentrations measured by “sipper” at the two marshes studied here. All samples were taken in the upper 25–30 cm and analyzed for ΣH₂S as described above. Analytical techniques were the same throughout and thus the values are comparable within and between the marshes. However, although sampling sites were in the same general area of each marsh, they were not identical over time. In addition, not all marsh sites were sampled in each year. The record of sampling includes: Big Egg- 2004/05, 2014-2016; JoCo- 2005-2007, 2014-2016. These time series of samples permit us to examine changes in the average concentration of ΣH₂S in pore water of the marsh peat in a depth zone in which root density is high and the plants can be considered susceptible to exposure to elevated levels of his phytotoxin. To do this, we calculated the average ΣH₂S value of the pore water analyses from sippers deployed from ∼5 to 30 cm. Generally, 4–5 sample depths were involved (e.g., 5, 10, 15, and 25 cm). All sites sampled in each marsh at a given time were averaged to produce a single ΣH₂S value, presumed to be representative of that marsh at that time. The results are shown in Figure 9, and the patterns in each marsh are as follows:

Big Egg: a portion of this marsh (BE1) was restored in 2003 by spraying sandy sediment from the adjacent subtidal bottom onto the marsh surface (Frame et al., 2006). The unrestored portion of the marsh was sampled in 2004/05 and again in our recent sampling. We also sampled the restored portion of the marsh in the present study (BE1). The effects of the sediment spray are well documented in the ²¹⁰Pb profile in the core taken in this area of the marsh (Cochran et al., 2018; data not shown). Concentrations of ΣH₂S in the unrestored portion of the marsh are as high as 3.5–4 mM in the summer and are quite similar between 2004/05 and 2014-16, with comparable values in comparable seasons (Figure 9). Samples from the restored portion of the marsh appear to have lower ΣH₂S in the pore water, and the offset is especially clear in the fall 2014, and spring 2015 and 2016 samplings.

JoCo: JoCo presents a conundrum in that it is considered a “healthy” marsh in Jamaica Bay, yet over the period 2005–2016, average ΣH₂S levels were >2 mM in summer and fall, 2006 and summer, 2015 and exceeded 4 mM in fall, 2005 and 2014. Offsetting these high concentrations is the fact that levels of ΣH₂S were low (<0.5 mM) at other sampling times, including in the spring, summer and fall, 2007, and were exceptionally low (∼0 mM) in spring, 2015 and 2016. Several factors controlling ΣH₂S are likely in play at JoCo. The large variations in ΣH₂S from sampling to sampling may in part be due to the fact that most of the reactive iron in the JoCo peat is associated with FeS₂ and is thus not able to readily react with H₂S produced via sulfate reduction. Secondly, although temperature is a strong control on bacterial activity and hence the production of H₂S, there is no clear relationship between mean air temperature during the month or week of sampling and mean pore water ΣH₂S (data not shown). As noted above, drainage rates in JoCo are the fastest seen of the marshes examined in this study (Tamborski et al., 2017). These rates may vary with time as a function of tides (neap vs. spring) and season (warm vs. cold; variation in precipitation). It is possible that at least the portion of JoCo studied is at or near a “tripping point” for degradation.