Limnol.

Oceauogr.,

28(l),

1983, 70-82

Sulfate reduction in the salt marshes at Sapelo Island, Georgia’ Robert W. Howurth The Ecosystems Center, Marine Biological Woods IIole, Massachusetts 02543

Laboratory,

Anne Giblin2 Boston University

Marine

Program,

Marine

Biological

Laboratory

Abstract Sulfate reduction rates were measured in stands of Spartina dternijh-a at Sapelo Island, Georgia, in November 1980 by injecting tracer amounts of ‘%04’- into cores, incubating overnight, and analyzing for the incorporation oi ‘% into reduced sulfur compounds. Qualitatively, sulfate reduction in the Georgia marsh is very similar to that in the Massachusetts marshes we have studied: FeS, (pyrite or marcasite) is the major end product. Lesser amounts of soluble sulfides, iron monosulfides, and elemental sulfur are also formed. The rate of sulfate reduction (determined by the same method) is significantly lower during November in Georgia than in the Great Sippewissett Marsh in Massachusetts, 0.090 vs. 0.27 moles S042-*m--2* cl-l in stands of short Spartina. The lower rates in Georgia may reflect a lower rate of organic carbon input by belowground production. Sulfate reduction appears to be the major form of respiration in the sediments of salt marshes in Georgia as well as in Massachusetts.

The organic-rich sediments underlying productive aquatic ecosystems are typically anoxic below the top few centimeters, since the potential rate of oxygen consumption is greater than the rate of oxygen supply. Much of the organic matter in these systems is decomposed through anaerobic processes (Jorgensen 1977; Rich and Wetzel 1978). Dissimilatory sulfate reduction and associated fermentations rather than methanogenesis dominate decomposition in anoxic marine sediments. This results from the abundance of dissolved sulfate in seawater and the higher relative energy yield for bacteria carrying out sulfate reduction than for methanogenesis. J@rgensen (1982) found that sulfate reduction accounts for about 50% of the organic mineralization in sediments underlying O-20 m of water and about a third in sediments underlying 20-200 m of water. Salt marshes are among the most pro’ Financial support was provided by NSF grants DEB 78-03557 and DEB 81-04701, NOAA Office of Sea Grant grant 04-8-MOI-149, and by the Univcrsity of Georgia Marine Institute. %Present address: Woods Hole Oceanographic Institution, Woods IIole, Mass. 02543.

ductive of marine ecosystems (Valiela et al. 1976; Smith et al. 1979; Pomeroy and Wiegert 1981), and the sediments underlying stands of Spartina alterniflora. are anoxic below l-cm depth (Teal and Kanwisher 1961; Howes et al. 1981). Under these conditions, one would expect sulfate reduction to be very important in organic matter decomposition. Our data from the Great Sippewissett Salt Marsh (Cape Cod, Massachusetts) show that sulfate reduction annually degrades an amount of organic carbon equivalent to >90% of estimates of net primary production there (Howarth and Teal 1979, 1980.) Of the methods for measuring or estimating rates of sulfate reduction in natural sediments only a few are suitable for use in sediments that do not approximate steady state assumptions, such as salt marshes or the surface sediments of other coastal environments. In these sediments, radiotracer methods are the best approach for accurately and precisely determining the rates of sukte reduction (Jorgensen 1977, 1978). Radiotracer methods determine the turnover of sulfate from the ratio of 35S in reduced sulfur compounds to Y304’- af70

Marsh sulfate ter a short incubation period. During this brief period only a tiny fraction of the n5S042- is reduced, so that this pool can be considered constant. The rate of sulfate reduction is estimated by multiplying the fraction of n5S042- reduced by the bulk concentration of sulfate. Obviously, for this approach to work, all of the label in reduced sulfur compounds must be determined. In many subtidal anoxic marine sediments, soluble sulfides (H,S and HS) and acid-labile sulfides (mostly FeS) are the only major products of sulfate reduction formed during short incubations. Elemental sulfur may be important in some subtidal anoxic sediments (Howarth 1979, and unpubl. data), perhaps forming as an artifact during acidification to distill FeS (Berner 1974). Although pyrite (FeS2) is formed in these sediments, it is usually formed only slowly from the gradual reaction of iron monosulfide with elemental sulfur (Berner 1970; Rickard 1975). Thus pyrite does not typically show up as a product of sulfate reduction when short incubations with 35S0,2- are used in anoxic subtidal sediments. In the New England salt marsh sediments we have examined, the situation is markedly different. Pyrite forms very quickly and is the major end product of sulfate reduction formed during the radiotracer measurements (Howarth 1979; Howarth and Teal 1979). Conventional radiotracer methods would not record the inclusion of label into pyrite or any reduced sulfur compounds other than soluble and acid-volatile sulfides because only these compounds are recovered for counting. Failure to include pyrite and other end products in analyses of sediment after incubation from the Great Sippewissett Salt Marsh results in underestimation of rates of sulfate reduction by 70-95%. Sulfate reduction rates for vegetated salt marsh sediments have only been reported for one other site, Sapelo Island, Georgia (Skyring et al. 1979). In that study, soluble and acid-labile sulfides were the only end products assayed. Thus, if pyrite or other compounds were major end products there as in the Massachusetts

reduction

71

marsh, the rates of sulfate reduction reported by Skyring et al. (1979) might have been underestimated. The marshes at Sapelo Island are different in many respects from the Great Sippewissett. For example, the sediments we studied at Sapelo Island were composed largely of clay with an organic content of only 5-10% (dry wt). The sediments at Great Sippewissett are 50-80% organic matter (dry wt) and contain a thick root mat. Our investigation at Sapclo Island provided a good opportunity to determine the generality of the rapid formation of pyrite in marshes. We thank the staff of the University of Georgia Marine Institute for making our stay there pleasant. D. Kinsey, C. Hopkinson, D. Whitney, and R. Christian were particularly helpful. Drafts of this manuscript were reviewed by R. Christian, D. Kinsey, C. Martens, C. Montague, B. Peterson, D. Rickard, E. Sherr, W. Wiebe, and R. Wiegert. H. Sladovich and V. Kijowski helped with chemical analyses. R. Marino measured the trapping efficiency of ““S-labeled sulfides. We thank G. Luther for SEM energy-dispersive X-ray analysis of sediments. Methods All fieldwork was done at Sapelo Island during 12 days in November 1980. We sampled sediments underlying two different stands of short S. alterniflora (a total of 24 cores on 8 days) and one stand of tall S. alterniflora (11 cores on 4 days). One of the stands of short grass was in the “Airport” marsh in the Duplin River watershed, a site previously studied by many workers (e.g. see Christian et al. 1978) and near site No. 14 of Teal and Kanwisher (1961). The other short stand we sampled was also in the Duplin River watershed, near Kenan Field and near site No. 1 of Teal and Kanwisher (1961). The tall stand we sampled was immediately off “Teal’s” boardwalk in Laughing Gull Marsh to the SSE of the Marine Institute laboratories on South End Creek. Cores were obtained by inserting plastic core tubes as previously described (Howarth and Teal 1979; Howarth et al.

72

Hoeuarth

in press). Porewaters were obtained for integrated 5-cm intervals with a Reeburgh (1967) p ress. Sulfides were determined calorimetrically (Howarth et al. in press) and sulfate by indirect titration (Howarth 1978). Oxidation of the sulfide samples was not a serious problem since porewaters were rapidly squeezed into a syringe, filtered from one syringe into another without introducing air, and quickly fixed with ZnOH (IIowarth et al. in press). However, a positive interference in the sulfate measurement undoubtedly occurred since we took no precautions to prevent oxidation of soluble reduced sulfur species in those samples. As discussed later, we feel this is a minor problem for the data reported here. Chloride was determined by automated coulometric titration with silver in a Buchler-Cotlove chloridometer automatic titrator. The alkalinity, pH, and concentrations of ammonia, phosphate, silicate, iron, and manganese were also measured (Giblin 1982) but are reported elsewhere. Total insoluble sediment sulfur was measured by aqua regia digestion in sediments washed with distilled water (Howarth and Teal 1979). Sulfate reduction was measured using 3sS0,2- (JGrgensen 1977, 1978), the method modified to include pyrite as an end product (Howarth and Teal 1979). We have further modified the procedure to simplify the distillation of labeled sulfides by using a passive distillation. We were inspired to try this by the work of Hobbie and Crawford (1969), Munson (1977L and Taylor et al. (1981). Cores were brought back to the lab and subcores taken with cut-off lo-ml plastic syringes. Subcores were 1.5 cm in diameter and had volumes of about 5 ml. They were injected in a line with about 1.25 &i of 35S0,2- in 20 ~1 of solution while still in the syringes. The syringes were plugged with rubber stoppers, taking care to leave no gas spaces. The syringes were then incubated at ambient mud temperatures (16”-20°C) for 20 h or so in a saturated brine solution so as to prevent oxygen diffusion through the plastic. At the end of the incubation, the sam-

and Giblin ples were placed in a specimen jar (4.4cm diam, 7.6 cm high) which had previously been flushed with argon via a syringe needle pushed through a serum stopper in the lid. Each jar contained a stirring bar. A second stopper in each lid had been fitted with a plastic cup containing Chromatographic paper for trapping ‘*CO, (Hobbie and Crawford 1969). After the samples were placed in the jars, the lids were closed and the jars flushed with argon for another 10 min. A syringe was used to add 0.2 ml of phenethylamine to the plastic cup containing the paper, and 20 ml of anoxic 1 N H,SO, was added to the jars. The H,SO, was made anoxic by bubbling with argon and adding a small crystal of pyrogallol. The contents of the jar were vigorously stirred for at least 2 h. Then the Chromatographic papers were removed and placed in scintillation vials. Ten milliliters of Aquassure scintillation cocktail (New England Nuclear) were added to each vial. The activity in these samples represented 35S in soluble and acid-volatile sulfides. These activities were corrected by the efficiency of this method of counting and trapping in such jars, 55% (SE = 4.7; n = 6). This trapping efficiency was determined with Na,““S standards. After doing the work reported here, we improved the passive distillation by using all glass jars, which are less reactive and easier to flush with argon. However, the approach used here is inexpensive and adequate for many purposes. The remaining sediment was filtered through a No. 451 Whatman filter in a Buchner funnel. One milliliter of the filtrate from each sample was added to 10 ml of Aquassure in a scintillation vial. The volume of the filtrate was measured to determine the total activity of 35S left in solution. This was assumed to be ‘30d2-. Our previous work at the Great Sippcwissett indicates that this is a reasonable assumption. The remaining mud from each sample was washed carefully five times with 50 ml of distilled water (a total of 250 ml). This mud was then digested in 15 ml of aqua regia (5 ml of coned HCl and 10 ml

Marsh sulfate Table

1. Percent

Distribution

of end products

of sulfate

reduction. Short Spartina

of end products

site

in 1 N I12S04

TallSpurtino

site

(SE=L8811n=26)

16.1 (SE=4.6; n= 14)

91.9 (SE=1.8; n=26)

83.9 (SE=4.6; n=14)

as II,S, HS-, and FeS insoluble

73

reduction

insoluble in 1 N H&SO, but extracted by hexane

(SE=2.:;?=2)

( :zl)

insoluble in 1 N H,SO, but extracted by 1.5 M NaOII

(SE=O.;:n=2)

( rL21)

of coned IINO,). Digestion was at room temperature and was allowed to continue overnight (about 20 h). Digests were diluted to 60 ml with distilled water and l-ml aliquots added to 10 ml of Aquassure in scintillation vials. For some samples, we extracted the sediments with hexane or 1.5 N NaOH before aqua regia digestion to estimate 35S in elemental sulfur or other end products, The sediments had to be dried before hexane extraction. Subsamples of these extracts were counted, again with Aquassure as the fluor. A few samples were permeated wih 0.1 m Na,MoO, before injection of the nsS0,2to inhibit sulfate reduction (Oremland and Taylor 1978). This resulted in a decrease in the rate of sulfate reduction of QO-97%. The small remaining reduction was probably in zones unreached by the molybdate, although it may also reflect the presence of small amounts of contaminants such as :‘5S-labeled sulfides or thiosulfate in the 35S0,2- we used, Nonetheless, any error introduced by such contaminants would be small, as shown by the molybdate-inhibited controls. Results and discussion End products of sulfate reductionThe distribution of end products of sulfate reduction from our Sapelo Island samples is shown in Table 1. Of the 3”S0,2- reduced, very small amounts ended up as soluble (H,S, HS-) or acid-volatile (FeS) sulfides-about 8% of all the reduced products in the short Spartina stand and 16% in the tall Spartina stand. In over 40 analyses, we never found

>38% of the reduced products as soluble plus acid-volatile sulfides. Despite the variance in the efficiency of trapping of H235S by the method used, the conclusion that soluble and acid-volatile sulfides were minor products is inescapable. The major part of the reduced sulfur, from 62 to 99.4% in individual samples, ended up as insoluble particulate sulfur compounds (Table 1). Most of this insoluble particulate sulfur we believe to be FeS,, as we previously reported for the Great Sippewissett (IIowarth 1979; Howarth and Teal 1979). The product was insoluble in 1 N H2S04, so we can conclude that it was not amino acid or sulfate-ester sulfur (Roy and Trudinger 1970; Howarth 1979). Hexane and NaOH extractions accounted for ~10% of this insoluble product (Table 1). There were no obvious differences between the hexane and NaOH extractions, and we believe that both solvents were probably extracting elemental sulfur (Howarth 1979). Thus, we conclude that FeS, (either pyrite or marcasite) is the major end-product of sulfate reduction in the Georgia marshes, with lesser amounts of elemental sulfur, soluble sulfides, and acid-volatile sulfides. Some of what we are calling pyrite or marcasite may in fact be a refractory organic sulfur of some sort, and this deserves further study. Even if some is organic sulfur, however, it is still reduced sulfur and represents sulfate reduction (Howarth 1979). Scanning electron microscopy and energy-dispersive X-ray analysis indicate an abundance of FeS2 in these sediments (Luther et al. in press). Single crystals of FeS, of a diameter of

74

Howarth

0.2 to 2.0 pm predominate, This is further evidence for a rapid formation of pyrite since small single crystals are believed to occur when pyrite forms rapidly, framboids when it forms slowly (Rickard 1975; Goldhaber and Kaplan 1974). Most of the sediments we studied were gray or brown, not black. Black coloration of sediments indicates either organic pigmentation or the presence of acid-volatile sulfides (FeS). The lack of black coloration strongly suggests that the concentrations of acid-volatile sulfides at Sapelo Island are low, a conclusion supported by the results of Oshrain (1977) who found acid-volatile sulfide concentrations of only 0.1-6.7 prnol. cm-” of sediment at these same sites. However, some of our sediment samples, particularly from the tall Spartina site, were quite black, possibly indicating that these were relatively rich in acid-volatile sulfides. Our 35S0,“- reduction measurements showed that FeS, was still the major end product in these black sediments, although soluble and acid-volatile sulfides made up a slightly higher percentage of the end products than in the gray sediments from the short sites (Table 1). If the black color of some of the sediments does indeed indicate relatively high concentrations of acidvolatile sulfides and not merely black organic pigmentation, then the presence of acid-volatile sulfides should not be taken as evidence of lack of rapid pyrite formation, as we had earlier suggested (Howarth and Teal 1979). Clearly, a black color is not definitive evidence of lack of rapid FeS2 formation. To determine whether pyrite is a major product of sulfate reduction, one must directly measure the amount of radiolabel in the pyrite pool. It is not sufficient to estimate pyrite formation by difference, nor can one adequately infer the absence of pyrite formation merely from a high percent of recovery of radiolabel from all of the other pools (as is implied by Pomeroy and Wiegert 1981). In the experiments reported here, an average of 97% (SE = 0.5%; n = 28) of the added 35S0,2remained unreduced at the end of the incubation. The percentage recovery of ra-

and Giblin diolabel is largely a measure of recovery of unreduced sulfate. Rates of sulfate reduction-The fraction of 90d2reduced for each sample was calculated by summing the 9 in the sulfide pool and soluble plus acid-volatile the insoluble particulate sulfur (pyrite) pool and dividing by the 35Sin the sulfate pool. We made no corrections for changes in the sulfate pools during the incubations because these were small. Sulfate reduction rates were then calculated by multiplying fractions of 35S0,2- reduced by bulk sulfate concentrations. We measured sulfate concentrations per volume of porewater (see Table 3) but converted the rates of sulfate reduction to rates per volume of sediment using our porosity data. Porosities were measured by weight loss on drying of a known volume of sediment. Although the dry densities of sediments varied among sites, the porosities were quite constant at 0.77 ml. crne3 (SE = 0.02; n = 14). The rates of sulfate reduction per volume of sediment underlying short and tall stands of Spartina are shown in Fig. 1. The data from the two stands of short Spartina we studied are combined since there were no significant differences between them in the rate of sulfate reduction. It is apparent from the small standard errors, even after pooling the two short stand sites, that day-to-day variations and spatial variations were small. Previous workers in the Sapelo Island marshes have hypothesized that the sediments there contain microzones in which free sulfate is absent or present in only very low concentrations (Oshrain and Wiebe 1979; King and Wiebe 1978). Skyring et al. ( 1979) were concerned that the existence of such microzones might lead to an overestimation of the rate of sulfate reduction. Actually, the existence of low-sulfate or sulfate-free microzones should not lead to any bias, negative or positive, in the method used here since 35S042- is added in only tracer amounts (Jorgensen 1978). However, such microzones might increase the variance in our microinjection technique: for example, an injection into a low-sulfate microzone

Marsh sulfute SHORT

Spartina

olternif/ora

gyy=y,

(

0.3

0

Spartina

SO,‘- cmu3 d -’

o/ternl’f/ora

K[;,,~;

, 0

0.9

0.6

pmoles

TALL

,y;

0.3 ,umoles

,‘, 0.6

,

( 0.9

SOi- crnm3 d - ’

Fig. 1. Rates of sulfate reduction over depth in short and tall Spurtinn uZterni~Rorn. sites. Two sites are combined for short Spnrlina site data. Standard errors arc plotted.

would give a faster turnover time for the same rate of sulfate reduction. When multiplied by the average bulk sulfate concentrations, this sample would give an apparent rate of sulfate reduction which is too high. On the other hand, injections of :i5S0,2- outside of low-sulfate microzones would give apparent rates of sulfate reduction which are too low, since the actual sulfate concentration there would be greater than the average bulk concentration for the whole sediment (including the microzones). Thus, given enough replication, the mean value determined by the YSO,“- tracer method should be accurate despite the presence of microzones. But microzones would tend to increase the variance. The variance in the rates we report here is quite low (Fig. l), so we conclude that low-sulfate or sulfate-free microzones are either absent or rare in the sediments we studied, or that they are small relative to the volume of sediment exposed to 35S0,2-. Sulfate reduction rates are significantly higher in the short grass sites than in the tall grass site (95% level, paired t-test). Integrating the rates presented in Fig. 1 over depth, we estimate mean rates of sulfate reduction of 90 mmoles *rnd2. d-l in short grass sites and 30 in tall grass

reduction

75

sites. This finding is contrary to the argument which has been developed by other workers at Sapelo Island that microbial activity is greater in tall stands than in short stands (Christian and Wiebe 1978; Skyring et al. 1979; Pomeroy and Wiegert 1981). However, as we have argued elsewhere (Howarth and Hobbie I982), we do not find the evidence previously presented on the comparison of microbial activities in short and tall grass sites to be conclusive. For example, Christian and Wiebe (1978) measured faster glucose turnover in tall grass sites, but their samples had been screened to remove macroscopic organic matter, were exposed to air, and were slurried. All of these manipulations may have had major effects on microbial activity. Also, even if glucose mineralization is greater in tall grass stands, it does not necessarily follow that total heterotrophic activity is greater. Leakage of ethanol and other plant metabolites from the grass rhizosphere may be important in fueling microbial heterotrophy, particularly in short grass sites, and this would not be measured by glucose turnover (Mendelssohn et al. 1981; Howarth and Hobbie,1982). On the other hand, our finding that sulfate reduction rates are greater in tall than in short grass sites is based on data from only one tall grass site and two short grass sites. We feel that the data from the short grass sites are probably representative; the two sites differed markedly in pH and other sediment characteristics (Giblin 1982), and yet sulfate reduction rates were quite similar. But we have no basis for determining if our tall grass site is representative. We feel that the question of microbial activities in tall grass sites vs. short grass sites should be studied further. Comparison with previous studies of sulfur dynamics -Skyring et al. (1979), in the only previous study of sulfur dynamics at Sapelo Island, did not adequately measure sulfate reduction because FeS, and other nonvolatile products were ignored. They measured only the net formation of soluble plus acid-volatile sulfides. This underestimates the true rate

76

Howarth

of sulfate reduction by the extent that FeS, and other reduced products are formed during the incubation, which, for their measurements, is not known. So although they report their data as rates of sulfate reduction, this is erroneous and their rates are not comparable to our total rates. However, we can compare their results with our estimate of the net formation rate of soluble plus acid-volatile sulfides. This rate for our data is about 8% of the total rate of sulfate reduction in short Spartina stands and 16% in tall Spartina stands (Table 1). Thus, below 5 cm in short stand sites our measured net formation rate for soluble plus acid-volatile sulfides decreased with depth and varied from 2 x lo+’ to 2.8 x lo-” mol. cm-:’ *d-’ , Skyring et al. (1979), at one of the same sites we studied, found rates which varied from 3 x lOmg to 1.1 x 1O-7 mol*cm-“~d-L. These are quite comparable to our rates even though their measurements were made in June at temperatures ~10°C higher than ours, 27”-35°C vs. 16”-20°C. In their samples from the top 5 cm of the short grass stands, Skyring et al. (1979) found a net formation of soluble plus acidvolatile sulfides of 1.22 x lo+ mol. crnM3. d-‘, a rate some 15-fold greater than ours. If their reported value is accurate, it may or may not reflect higher rates of sulfate reduction. It may merely reflect a difference in the proportion of the sulfate reduced that is going to these particular products. However, there are also reasons to suspect their estimate for the top 5 cm. They reported very low recoveries of added tracer at that depth, and, if correct, their value implies a turnover of soluble and acid-volatile sulfides of only 1 day. Such a turnover is much finster than in any of our samples, or than in any of their other samples, and seems unrealistically fast. In our tall grass stand, we found a net rate of formation of soluble plus acid-volatile sulfides of 1.7 x IO-’ to 2.1 X 10e7 mol.cm-3-d-l in the top 5 cm and about lo+ mol. crnd3* d-l at greater depths. Skyring et al. (1979) reported rates 6-70 times greater for the top 5 cm and 2-10 times greater at depth. Again there is no

and Giblill Table 2. Range of concentrations of acid-volatile sulfides (FeS) in top 25 cm of sediment reported for Sapelo Island marshes. Short Time ot’ year (rcfcrence)

(mol’g-’

August through December (Oshrain 1977) June (Skyring 1979) *

Tall

Spartina sit&

dry wt X 10-l;)

3-12

0.5-19

et al.

to have a dry-wt our data for Airport marsh. I’ ASSLI~I~CI to IXLVC B dry-wt our data for Teal’s marsh. ASSLI~WCI

Spartina sites*

2-10 density

of 0.33 g.cm

density

of 0.37 g.cm-I’,

60-90 8, the avcrqc the average

of of

way to determine if these differences reflect differences in the total rate of sulfate reduction. The concentration of acid-volatile sulfides at the tall grass site when they made their measurements was unusually high in comparison with the sites and times studied by Oshrain (1977). On the other hand, the short grass site of Skyring et al. (1979) had low concentrations of acid-volatile sulfides, with values typical of those previously measured (Table 2). This suggests that Skyring et al. may have measured an unusually high rate of acid-volatile formation in their tall grass site. It should be noted that Skyring et al. added their 35S042- dried on a glass rod, not injected in a liquid phase as we and others (Jergensen 1977, 1978) have done. They suggested that incorporation of 2% sodium silicate in the Na23”S0,2- solution reduces the amount of smearing along the insertion path of the rod, although they did not use silicate during the time they took the reported data. The use of sodium silicate is questionable since it may buffer the microzone exposed to 35S042- at an abnormally high pH. This change in pH might change the shott term products of sulfate reduction, favoring acid-volatile sulfides such as FeS over pyrite or marcasite during the short time of the 35S0,2incubation (Howarth 1979; Berner 1970; Hickard 1975; Goldhaber and Kaplan 1974). Importance of sulfute reduction in

Marsh sulfute Sulfate reduction and mursh respirationrelated fermentative processes are apparently the major form of organic mineralization in the sediments of the Sapelo Island marshes. This is in agreement with our findings at Great Sippewissett (Howarth and Teal 1979, 1980; Howarth and Hobbie 1982). Th e reduction of 1 mole of sulfate mineralizes 2 moles of organic carbon to carbon dioxide (Jorgensen 1977; Howarth and Teal 1979). Thus the rates of sulfate reduction we report here account for the respiration of 2.2 g C* rn-“. d-l in short grass stands and 0.72 in tall grass stands. Oxygen uptake data are not available for the Sapelo Island marshes in November. But in summer, when we would expect oxygen uptake to be at its highest, Teal and Kanwisher (1961) estimated the uptake by mud to be 7.9-9.2 mm3 0, *cm-2 *h-r. Not all of this reflects respiration since some oxygen is undoubtedly consumed in chemical oxidation of sulfides and other substances. Nevertheless, if oxygen uptake were all respiration, it would represent the respiration of only 1 g C*m-2.d-1. Teal and Kanwisher (1961) found no difference in oxygen uptake between tall and short grass sites. Other forms of organic matter mineralization are apparently even smaller. Denitrification rates for Sapelo Island have not been measured in November, but for the period from December through June, Haines et al. (1977) estimated a mean rate of 2.3 mmoles N .rne2* d-l. Such a rate would account for the mineralization of 0.035 g C *me2 *d-‘, one to two orders of magnitude less than the mineralization of carbon by sulfate reduction in November. Methane losses from the Sapelo Island marshes during November hadve been estimated as roughly 0.002 g C-m-“*d-l for tall grass stands (King and Wiebe 1978) and from 0.002 to 0.09 for short grass stands (Atkinson and Hall 1976; King and Wiebe 1978). Again, sulfate reduction rates account for the mineralization of one to two orders of magnitude more organic matter, Comparison of Sapelo Island with Great Sippewissett Marsh-In Fig. 2 we

77

reduction

0.5 -

GREAT

SIPPEWISSETT

0.4 7 . D N ‘E 0.3A f :: 0.2 t s z

0.0,

’ ’ ’ JFMAMJJASOND

’

’

’

’

’

’

’

’

Months

Fig. 2. Comparison of rates of sulfate reduction in stands of short Spartina aZternijZoru at Sapelo Island and at Great Sippewisset. Calculation of error bar explained in text.

compare the rate of sulfate reduction in short Spartina stands at Sapelo Island during November with rates for all seasons in short Spartina stands at Great Sippewissett. The data for Sapelo Island include both of our sites. As stated above, both of these sites had similar rates of sulfate reduction (as evidenced by the low standard error in Fig. 1) despite large differences in such sediment characteristics as pH and soluble metals (Giblin 1982). This gives us some confidence that sulfate reduction rates from these sites are Some of the data from representative. Great Sippewissett have been presented before (Howarth and Teal 1979), but the error limits are new. These error limits were determined as follows. We took all of the data for rates of sulfate reduction per unit volume of sediment for each month and integrated these over depth; this gives the mean rate of sulfate reduction per unit surface area of marsh for each month. We then took these same data for rate per unit volume, subtracted the standard errors for each depth, and reintegrated. This gives a minimum estimate for sulfate reduction per unit surface area. Adding the standard errors and reintegrating gives a maximum estimate. We handled the data from Sapelo Island in the same way. Sulfate reduction rates are clearly lower during November at Sapelo Island than

78

Howarth

anytime during the fall at Great Sippewissett. In November, sulfate reduction rates in the Great Sippewissett were three times greater, or 0.27 mol. rnp2. d-’ (Howarth and Teal 1979). Since grasses senesce later in Georgia than in Massachusetts and since temperatures at Sapelo Island in November are more like those at Great Sippewissett in September or October than in November, it is perhaps better to compare the Sapelo Island rate of sulfate reduction in November with that at Great Sippewissett in September or October. By this comparison, reduction rates at Great Sippewissett are 3.84.6 times greater or 0.34-0.41 mo1.m-2* d-’ (Howarth and Teal 1979). At Great Sippewissett we find rates of sulfate reduction lower than the November rate at Sapelo Island only from January through March, when sediment temperatures are low (