- Email: [email protected]
Temporal variations in the methane content of the Cariaco Trench
Deep-SeaResearch,Vol.35, No. 9, pp. 1511-1523, 1988.
0198-0149188 $3.00+ 0.00 ~) 1988PergamonPressplc.
Printedin Great Britain.
T e m p o r a l variations in the methane content of the Cariaco Trench MARY I. SCRANTON*
(Received 22 October 1987; in revised form 10 May 1988; accepted 24 May 1988) A ~ t r a e t - - T h e deep waters of the Cariaco Trench are not in steady state with respect to distributions of either hydrographical or chemical properties. Over the past 15 years, methane concentrations at the bottom of the anoxic zone have increased steadily from about 7-9 to 12.5 ~tM. A simple model for the anoxic waters of the Cariaco Trench suggests that methane distributions may be explained as the result of the flux of methane from the sediments (about 12.5-17.5 ttmol cm-2 y-~), vertical eddy diffusion within the water column, and methane oxidation in the water column. Oxidation rates near the oxic-anoxic interface may be considerably higher than previously measured, although available data are inadequate to determine whether the methane oxidation near the interface occurs under anaerobic conditions or whether it occurs when oxygenated overlying water is mixed down into the top of the anoxic zone. It also is possible that methane fluxes from the shallower sediments of the Cariaco Trench are very low compared to those from the deeper sediments.
INTRODUCTION
METHANE is a terminal product of the anaerobic degradation of organic matter. Therefore, in anoxic water columns, methane concentrations can reach high levels, although concentrations are usually quite low in the overlying oxidized surface layers. Methane production (mostly within the sediments), methane oxidation within the water column and sediments, and physical mixing and diffusion seem to control methane concentrations in most anoxic basins. Previous work (REEBURGH, 1976; RUDD et al., 1976; RUDD and HAMILTON, 1978; HESSLEIN, 1980; WARD et al., 1987) has attempted to quantify the rates of processes controlling the observed distributions using various combinations of direct measurements and modeling. In this study I discuss the temporal variations of methane in the Cariaco Trench using a model that has been used previously to analyse temperature, salinity, hydrogen sulfide and silica distributions (SCRANTONet al., 1987). THE C A R I A C O T R E N C H
The Cariaco Trench is a large, permanently anoxic basin located on the continental shelf of Venezuela at 10°30'N 65°31'W. It is about 200 km long and 50 km wide with a maximum depth of about1400 m. The basin is separated from the Venezuelan Basin and the resi of the Caribbean Sea by a sill at about 150 m, above which water can exchange freely with surface waters further offshore. The deep part of the basin is divided into two sub-basins by a saddle at about 900 m. The oxic-anoxic interface has been found at depths between 240 and 400 m, although details of the variation in interface depth with * Marine Sciences Research Center, State University of New York, Stony Brook, NY 11794-5000, U.S.A. 1511
1512
M.I. SCRANTON Table 1. Sources for methane data
Source ATrdNSONand RICHARDS (1967)
Date Feb. 1965
Location 10040.5'N 65°31.5'W 10o31,N 64o40'W
WIESENBURG(1975)
Feb. 1974
10°36.0'N 65°38.3'W 10°38.3'N 65°37.6'W 10o39.0'N 65°36.5'W 10°38.0'N 65°36.5'W 10°30.0'N 64°40.1'W 10°29.8'N 64°38.9'W
LAMONTAGNE (personal communication)
May 1971 June 1976
10°38'N 65°45'W 10°41'N 65°36'W 10o41'N 65o38'W
SCRANTON(1977)
Feb. 1975
l&29'N 64°44'W
SCRANTON(this report)
Nov. 1982 Feb. 1986
10°31'N 64°46'W 10°40'N 65°31'W
time are not well known. It is unlikely that the depth of the interface has been significantly overestimated, as the human nose is capable of detection of even trace levels of hydrogen sulfide. The first data for methane in the Cariaco Trench were obtained by AaxxNsoN and RICHARDS(1967), who collected samples from stations in the two sub-basins (see Table 1 for station locations and dates of collection for all methane data discussed in this paper). Methane data also are presented in the Master's thesis of WlESENBURG (1975), by LAMOrCrAONEet al. (1973) and LAMOrCrAGNE(personal communication), and by myself and coworkers (SCRArCror~, 1977; WARD et al., 1987; SCRArZroN,this report). Data from Lamontagne and coworkers were collected on two separate cruises (1971 and 1976) and our data were obtained on three cruises (1975, 1982 and 1986). Most of the data available are from the western basin, although Aavar~soN and RICHARDS (1967), WIESENBURG (1975) and SCRANTON(1977, this report) have obtained data from the eastern basin. METHODS Methane data were obtained from water samples collected in Niskin bottles and transferred to sample bottles in the same manner as oxygen samples (filled from the bottom, overflow of at least one bottle volume). AalaNsoN and RaCHARDS(1967) report that all samples to be analysed at sea were stored at I°C. Samples to be run in the lab were collected in ampoules, poisoned and stored in a cool, dark place. Lamontagne and coworkers analysed all samples at sea, but apparently did not poison them, as they also were measuring unsaturated organics (SwINNERTONand LINNENBOM,1969). Wiesenburg's methane data were obtained on samples poisoned with sodium azide and stored for about 1 month before analysis (WIEsE~,~aURO, 1975). (Note: sodium azide does not poison all metabolic processes in anoxic environments.) My 1975 samples were placed in ground glass stoppered bottles and were poisoned immediately after collection with 1 ml saturated mercuric chloride solution per liter. Most of the samples were analysed on board ship, but a few were returned to the
M e t h a n e content of the Cariaco T r e n c h
1513
laboratory and analysed approximately 7 weeks after collection. Samples obtained in 1982 were transferred to 700 ml champagne bottles, poisoned by addition of 1 ml of saturated mercuric oxide solution, and capped using a plastic-lined metal cap and bottle capper. Upon return of the samples to the laboratory, a small bubble was found in each bottle. The bubble volume was estimated and the amount of methane partitioned into the headspace was calculated. Samples were run in the laboratory approximately 6 weeks after collection. The methane samples from the 1986 cruise were collected in ground glass stoppered bottles and were analysed immediately after collection on board ship. Modifications of the extraction and trapping method of SWINNERTONet al. (1962) were used in all studies (ATraNSON and RICHARDS,1967; LAMONTAGNEet al., 1973; WIESENBURG, 1975; SCRANTON, 1977) except in 1982 and 1986 when the vacuum extraction method of SCRANTONet al. (1984) was used to strip gas from solution. All methane analyses were carried out using flame ionization gas chromatography. Calibration for our three cruises was performed by injections of commercially prepared methane-in-nitrogen standard mixtures. The standards used in 1982 and 1986 were calibrated using static dilution techniques, while the commercially quoted standard value was used in 1975. Possible corrections to the 1975 standard value are discussed below. Other workers used similar procedures. RESULTS
AND DISCUSSION
Although many workers have treated the Cariaco Trench as a steady-state system (FANNING and PILSON, 1972; REEBUR~H, 1976), this is not an accurate description of the behavior of chemical or hydrological parameters over the past 20 years (RICHARDS, 1975 ; SCRANTONet al., 1987). In particular, temperature, salinity, hydrogen sulfide and silica have increased markedly in the deep parts of the basin (for example, approximately a 33% increase for sulfide between 1973 and 1982). We recently developed a model that predicts observed trends in hydrographic and chemical parameters in the Cariaco Trench (SCRANTONet al., 1987; Figs 1 and 2). It is a simple non-steady-state box model consisting of 12 boxes, each 92 m thick (50 fathoms). SCRANTONet al. (1987) restricted their model to the eastern basin, largely because the best hydrographic data are available from that end of the Cariaco Trench. However, for methane, no recent data are available from the eastern basin, so the western basin is modeled here instead. In our original study, diffusive fluxes of salinity, temperature, sulfide and silica in the water column out of, or into, the eastern basin across the top of box 8 (915 m) were assumed to be the same as those across the top of the eastern basin. This assumption was reasonable as profiles of most chemical parameters (and therefore chemical gradients) in the two basins are very similar (SCRANTONet al., 1987). Fewer data are available for methane, but examination of the historical data suggest that a small (1.4 ~tM) interbasin difference in methane concentration, beginning at depths as shallow as 600 m, may exist (WIESENaURG, 1975). Thus we have included both deep sub-basins in our model for methane (Fig. 1, Table 2). The model includes transport of chemical species between boxes by eddy diffusion. Supply (or removal) of species also takes place by diffusion from (or into) the sediment and is assumed to be proportional to the area of bottom intersected by the box (thus taking the bathymetry of the basin into account). Values for vertical eddy diffusion coefficients were determined by SCRANTONet al. (1987) by fitting the temperature and
1514
M . I . SCRANTON
\
ANOXIC INTERFACE BOX I 2 5
WESTERN BASIN Fig. 1.
5 6 7 8 _ _ _ 9 _ I0 II 12
EASTERN BASIN
Schematic of geometry of box model (adapted from SCRANTON et al., 1987).
Ci-l, vi-I
BOX i - I
]
AZ F-~
L
Ci-i)
/Ai-i /
Ci ,Vi Ki+l (Ci+l Ci). /Ai+ I BOXi+l I AZ Ci+"Vi+'/ iAi+ 2
Fig. 2. Flux terms and notation for box model calculations (taken from SCRANTON et al., 1987 and Erratum, Deep-SeaResearch,34, 1653). Ai, IT,-and C, are, respectively, the area, volume and methane concentration for box i. Ki is the vertical eddy diffusivity at the top of box i. Az is the thickness of the box (92 m).
salinity data as described. Sensitivity studies have suggested that the error associated using eddy diffusion coefficients derived from temperature changes between 1973 and 1982 (as was done by SCRANTONet al., 1987), rather than between 1974 and 1986, will be small. Hydrographic data from the cruises in which methane was analysed are not sufficiently detailed or of sufficient precision to justify a modification of the earlier constants. Basin volumes (Table 2) are different from those presented in SCRANTONet al. (1987) because an error was made in calculation of the earlier box volumes (see Erratum published in Deep-Sea Research, 34, 1653). Use of the correct volumes results in the use of a lower eddy diffusion coefficient (decreased by 20%), but otherwise the model fits to temperature, salinity, hydrogen sulfide and silica are essentially the same as presented in SCRANTONet al. (1987). Values for the eddy diffusion coefficients used in this study are also presented in Table 2. Earlier work (SCRANTONet al., 1987) suggested that temperature and salinity increases are the result of downward flux of heat and salt from the interface, while hydrogen sulfide concentrations are increasing because flux from the sediments exceeds vertical diffusive loss. Silica concentrations appear to be affected both by diffusion of silica from the sediments and by dissolution of siliceous particles within the water column.
1515
Methane content of the Cariaco Trench
Table 2.
Model parameters for the Cariaco Trench box model
Basin area (10 9 m 2) East 320 412 503 595 686 778 870 961 1052 1144 1236 1327
Basin volume (1011 m 3) West
East
8.22 7.60 6.97 6.31 5.70 5.03 4.40 1.35 1.15 0.92 0.72 0.51
Ki (cm2 s-1) West
7.30 6.70 6.10 5.50 4.90 4.30 3.70 2.40 2.10 1.80 1.50 1.10
1.20 0.95 0.75 0.56 0.47
2.00 1.80 1.50 1.20 1.00
0.49 0.85 1.02 1.42 4.07 4.07 4.07 4.07 4.07 4.07 4.07 4.07
There also is evidence that methane concentrations have increased with time. Data from 1965 (AavaNsoN and RICHAROS, 1967; eastern and western basins), from 1968 and 1969 (LAMotcrAGNEet al., 1973; basin unidentified), from 1971 (LAMorcrAGr~E, personal communication; western basin) and from 1974 (WmSEt~-aURC, 1975; eastern and western basins) all indicate deep methane concentrations of about 7 laM (Fig. 3). Wiesenburg's data suggest some interbasin variability in methane concentrations, with eastern basin bottom water concentrations of about 7.2 IxM and western basin concentrations of about 6.0 IxM. Data from the eastern basin collected by Scranton in 1975 indicated bottom water concentrations of about 9 IxM (although stored samples gave values of 7 IxM; SCaArcroN, 1977). Lamontagne and colleagues also found methane concentrations in the western basin deep waters to be about 9 laM in 1976 (Fig. 4). In contrast western basin samples collected in 1982 and 1986 yielded methane concentrations in bottom samples of 10.8 and 12.5 laM, respectively (Fig. 4). One of the difficulties in comparing historical data sets is that there have been few intercalibrations of standards (and concentrations quoted for commercially obtained standards can be incorrect). Therefore I am unable to determine the absolute accuracy of the early methane data from the Cariaco Trench. In the case of my 1975 data, I can attempt a qualitative evaluation of the standardization by comparison of measured atmospheric methane concentrations (about 1.3 ppmv) with values interpolated for 1975 from the data presented in RASMUSSENand KHALIL(1981). The interpolated atmospheric methane concentration is 1.5 ppmv, somewhat higher than we observed. Thus the standard value used apparently was not too large, as air concentrations were lower than expected while water concentrations were greater than expected [compared, for example, to the data of WIESENBURG(1975)]. I recognize that such a comparison can only be approximate, since methane concentrations in the atmosphere are known to vary with season, with latitude, and with proximity to methane sources (STEELE et al., 1987). The comparison suggests that my standard was adequate to within about +15%, and that deep methane concentrations are not too high due to a standard calibration problem. The 1975 samples are from the eastern basin, found by Wiesenburg to have the higher methane concentrations (WIESENBURG, 1975).
1516
M.I. SCRANTON C H 4 ,/J.M
5
EO
I
I
200 300
f% ,Px°a
40( 50(
£ E
600
i x
•
o
jr3
ex
•g
o
g. x
700
LIJ r~
800
x •
o
•
o
900 •
I000 ItO0
•
o
x •
o
(*)
1200
x
1300
x
o o
(o)
•s
Fig. 3. Methane data for the Cariaco Trench between 1965 and 1975. All data are for the western basin unless otherwise noted and were taken from the following sources: (3, collected in November 1965 (ATKINSON and RtCHARDS, 1967); x, collected in May 1971 (LAMONTAGNE, personal communication); Q, collected in February 1974 (WIESENBURG, 1975; REEBffROH, 1976); D, collected in February 1975 from the eastern basin (SCRA~rroN, 1975). The points in parentheses from ATrdNSON and PdCHARDS(1967) appear to be low, but were not discarded by the original authors.
I have used the model described above to explain the distribution of methane in the Cariaco Trench. In the following discussion, I have used two different data sets for the initial methane distribution. First, I have used the 1975 eastern basin methane profile collected by SCRAWrON (1977). Since no data are available from that cruise from the western basin, I have assumed that the distributions in the eastern basin and western basin are identical. As an alternative, I have used the 1974 data of WIESENBURG(1975) who obtained data from several stations in each basin. In the case of Wiesenburg's data, for depths above saddle depth (about 900 m), I have taken an average methane concentration since I assume that boxes above sill depth are well mixed laterally. Below the depth of the saddle, I have used the actual profiles as initial conditions. Model results do not vary significantly regardless of whether different profiles are used for the two basins or whether the basins are considered to have identical methane profiles (results not shown). With each set of methane data I have used the box areas and volumes for the two basins presented in Table 2. For each time interval, the transports into and out of each box from sediment diffusion and eddy diffusive transport were calculated and were
1517
Methane content of the Cariaco Trench CH4 ,~M
2oo_
IO
15
300~
1%o~ ~
×
40O
o m lo o
5OO
o. xx omm
60O
[3
E"
£
Ii 0.-
70O eI
80O o
9OO
xx
e_m e
Iooo
x o
11oo
xx
•
o
12oo 1300 ° o
Fig. 4. Methane data for the Cariaco Trench between 1975 and 1986. All data are from the western basin unless otherwise noted and were obtained from the following sources: [], collected in February 1975 from the eastern basin (SCRANTON, 1977); O, collected in June 1976 (LAMorrrAGNE, personal communication); x, collected in November 1982 (this work); Q, collected in March 1986 (this work).
added to or subtracted from the values from the previous iteration. Time steps of 0.1 year were used, as shorter periods did not affect the results. In Fig. 5 the 1986 experimental data are presented together with three curves generated for the western basin using the 1975 data as the initial condition, and values for Kv, area and volume from Table 1. In order to calculate model profiles, I made an estimate of the flux of methane across the sediment-water interface. Model profiles for a sediment-water flux of 12.5 lmtol cm-2 y-1 (curve B) agree quite well with 1986 western basin methane data near the bottom. Similar sediment-water fluxes (best fit = 20 pmol cm-2 y-l) apply if I use Wiesenburg's 1974 data as the initial methane distribution. In fact, it does not appear to be possible to generate the observed bottom methane increase, even in the absence of methane oxidation in the water column, without a sediment-water flux of this order of magnitude. Model calculations indicate that the sediment-water flux was not likely to be much less than 5 pmol cm-2 y-1 since fluxes lower than this value would predict decreases in methane concentration with time. Few data exist on methane distributions in the sediments of the Cariaco Trench with which to compare these predicted fluxes, but REEBURCH (1976) indicated pore water values in the top 20 cm of the sediment were on the order of 10--40 pM. If we assume that
1518
M . I . SCRAm'ON CH4,~M 5
I0
15 i
200 30O 40(3 •
\
50(3 E 600
=).-.
a_ 700 80O 9OO 1000 1100 1200 1300
Fig. 5. Model fits with n o / n situ consumption but with different fluxes of methane from the sediments. (A) F = 17.5 ~mol cm -2 y-l; (B) F = 12.5 )unol cm -2 y-l; (C) F = 7.5 gmo1-2 y-a.
the methane content in the surface sediments is 40 ~tM, and noting that the bottom water concentration in 1974 at the time of Reeburgh's study was about 7-9 ~tM, we can estimate the rate of diffusion out of the sediments using dC Flux = ~ Ds-d-~z (BERNER, 1971), where ~p is the porosity (about 0.8). Ds is the diffusion coefficient for methane in sediments [approximately equal to ~p2 D (LERMAN, 1979)]. When D is the free solution diffusion coefficient of approximately 2 x 10-5 cm 2 S- 1 (estimated for 17°C from data for gases of similar molecular weight presented by BROECrmRand PENG, 1984), Ds equals 1.3 X 10-5 cm 2 s-1. The gradient of methane concentration is dC/dz at the interface and can be estimated at (40-9 ixM)/1 cm. Thus the flux is about 10 ~tmol cm-2 31-1. Of course the actual flux will be highly sensitive to details of the shape of the profile of methane concentration with depth near the interface. The data presented by REEBURGrI(1976) are somewhat scattered and may indicate disturbance of the sediments during coting. Since it seems much more likely that methane was lost from the sediments during sampling than that the methane concentration was increased, the above estimate of sediment-water flux may represent a lower limit to the true flux. In his paper, R~EBUR~H (1976) estimated that the flux of methane from the deeper
Methane content of the Cariaco Trench
1519
part of the core to the near-surface zone was on the order of 5-15 I~mol cm -2 y-l, but that this flux was almost completely oxidized by anaerobic methane oxidation before methane could diffuse into the overlying water column. This would suggest that fluxes of methane across the sediment-water interface should be much lower than 5-15 lxmol cm 2 s-1. However, the diffusion coefficient used by Reeburgh (3 x 10-6 cm 2 s-i) is low, suggesting that actual fluxes at depth would be higher than reported in REEBURGH(1976). In addition, my calculation does not imply that methane oxidation within the surficial sediments is unimportant. On the contrary, it is likely that most of the methane produced at depth in the sediments is consumed before diffusing out. It does appear, however, that a significant amount of methane escapes oxidation and reaches the bottom waters of the Trench. The alternative possibility, that large amounts of methane are produced in the water column, seems even less likely than the possibility that sediment-water fluxes are relatively large, considering the fact that methane oxidation is known to be important in the deep waters (WARDe l al., 1987) and that water column processes (such as sulfate reduction) are believed to be quite slow (ScRAWrONet al., 1987). Clearly, we need more detailed and precise data for methane profiles in Cariaco Trench sediment before the relative importance of methane oxidation and loss across the sediment-water interface can be quantified. As discussed below, it is possible that there is no flux of methane from the sediments at some depths in the Cariaco Trench. One way to estimate the magnitude of the sediment-water flux for methane is through use of models such as the one developed to describe distribution of dissolved species in the Cariaco Trench. Although the model results presented in Fig. 5 agree satisfactorily with the data near the bottom of the Cariaco Trench, predicted methane concentrations near the oxicanoxic interface, calculated from all three fluxes, are much higher than observed. One explanation for the poor fit near the interface of the curves in Fig. 5 is the presence of methane oxidation in the anaerobic zone. Figure 6 presents observed methane concentrations together with model fits generated using the "best fit" sediment-water flux from Fig. 5 (12.5 I~mol cm -2 y-l) and different arbitrary patterns of methane oxidation in the anaerobic zone. The interface was assumed to be at 274 m, the precisely determined interface in 1983 and the depth used in previous modeling efforts. Curve A was obtained by making the assumption that methane oxidation (at the rate of 0.82 nmol 1-1 d-1 or 0.3 ~tmol 1-1 y-l) occurred only in the top box, curve B assumes oxidation at the same rate in boxes 1 and 2 and curve C assumes oxidation in boxes 1-4. The details of "best fit" oxidation rates and the depth range over which oxidation occurs are strongly dependent on the choice of initial data set. If, for example, WIESEt~URG'S (1974) western basin data are used (bottom water methane concentrations of 7.5 I~M), oxidation (at a rate of 0.2 ~tmol 1-1 y-l) is required only in the top two boxes to obtain good fits of the model to the data (Fig. 6, curve D). Therefore the values of the model-derived oxidation rates and their variation with depth are dependent on model parameters. For comparison, I have also calculated model profiles using the oxidation rates measured by WARD et al. (1987). Their data suggest that maximum methane oxidation rates were attained in the bottom of the Trench but that oxidation near the interface was negligible. As can be seen in Fig. 7 (curve A), if the sediment-water flux is increased to 17.5 lamol cm -z y-l~ the model profiles agree quite well with the data near the bottom.
1520
M . I . SCRANTON CH 4 ,p.M 5
I0
15
200 3OO 4OO 5OO 6OO E I
7OO 8OO 9OO I000 I100 1200 1300
Fig. 6. Model fits with sediment-water flux of 12.5 ~tmol cm -2 y-1. Consumption of methane is assumed to take place at a rate of 0.3 Ixmol 1-1 y-1 but over varying portions of the water column (line A: box 1 only; line B: boxes I and 2; line C: boxes 1-4). Curve D demonstrates the effect of choice of a different set of initial data (here, data from WIESENBURG, 1977). Curve D was generated using a sediment-water flux of 20 Itmol cm -2 y-i and the assumption that consumption takes place at a rate of 0.2 t~mol 1-1 y-1 in boxes 1 and 2.
However, the predicted methane concentrations near the interface are as high as in the case described in Fig. 6. Since it is possible that methane fluxes could vary with depth, I also have examined the case where methane fluxes from the sediments in boxes 1-4 are zero, while fluxes at depth are high enough to produce the observed bottom water increases. In this case as well, predicted methane concentrations near the interface are in excess of observed values (Fig. 7, curve B) although the discrepancy is less. If oxidation rates throughout the water column are increased to 0.15 l~mol cm 2 y-l, the data can be matched quite well with the model (Fig. 7, curve C). Thus, it seems that long-term average oxidation rates near the interface may be higher than rates measured directly by WARD et al. (1987). Model predicted rates are about 0.15-0.3 I~mol 1-1 y-1 (0.44).8 nmol 1-1 d-1) compared with rates measured near the interface by WARD et al. (1987) of less than about 0.02 ~tmol 1-1 y-1 (0.06 nmol 1-t d-l). The discrepancy between the direct measurements of WARD et al. (1987) and the modelpredicted results may have several explanations. Ward and coworkers measured methane oxidation rates at a season of low primary productivity over a limited period of time (hours) while model results reflect long-term averages of processes. Since the input of
1521
Methane content of the Cariaco Trench
CH4,~M 5 2OO
I0
15
1 i
3O0 4OO 500 6OO E -r 7 0 0 Io.. iJJ a
800
900
I000 IlOO 120o I:.300
Fig. 7. Model fits using the 1974 Scranton data as the initial condition. Curve A was generated with a sediment-water flux at all depths of 17.5 ttmol cm 2 s -1 and methane oxidation rates for each box estimated from the data in WARD e t al. (1987). Curve B also used the Ward e t al. oxidation rate data, but assumed that no methane diffused across the sediment-water interface in boxes 1-4. Curve C assumes a sediment-water flux of methane of zero in boxes 1-4, a flux of 17.5 ttmol cm 2 s-1 in boxes 4-12 and a constant methane oxidation rate of 0.15 ttmol 1-1 y-1.
carbon to the Cariaco Trench deep waters is most probably seasonally variable, water column oxidation rates, especially near the interface, might vary with time. Alternatively, the indication that oxidation rates are highest near the interface may suggest that occasional turbulent mixing of oxygen into the upper part of the anoxic zone accompanied by "aerobic" methane oxidation may actually be the most important process removing methane in this system. There is a suggestion of such an effect in the sulfide data which shows a narrow zone (few tens of meters) of concave upward curvature near the anoxic-oxic interface which cannot be ascribed to the processes included in our 1987 model (SCRANTONet al., 1987). Because of the sensitivity of the model fits to the precise nature of the initial profiles used, errors in the early concentration data may help to explain the fact that one would have expected injection of oxygenated water to occur only in the top box. CONCLUSIONS
Methane concentrations within the deep waters of the Cariaco Trench have increased with time since 1965, while concentrations in waters above about 900 m have remained nearly constant. By applying the non-steady-state box model of SCRA~rrONet al. (1987) to
1522
M . I . SCRArCrON
the data, it appears that the major source of methane to the basin waters is diffusion from the sediments. Temporal variations in methane concentrations at depths greater than 900 m can be explained by the time-dependent box model if the flux from the sediments is about 12.5-17.5 Ixmol cm -2 y-1. The sediment-water flux must be greater than 5 Ixmol cm -2 y-1 or methane concentrations would be decreasing with time, in contrast to observations which suggest concentrations are remaining the same or are increasing (depending on depth). Methane concentrations near the oxic-anoxic interface are lower than would be predicted by a model that includes only a sediment source and transport within the water column by turbulent diffusion. Methane oxidation near the interface seems to be an important process. Model calculations suggest oxidation rates throughout the anoxic zone are on the order of 0.15-0.3 ~tmol 1-1 y-1 (0.4-0.8 nmol 1-1 d-l). Acknowledgements--I wish to thank Paul C. Novelli and Peter A. Loud for their assistance in collecting and analysing Cariaco Trench methane samples and Drs Denis Wiesenburg and Robert Lamontagne for providing me with methane data from their earlier cruises. Financial support for this work was provided by NSF grants OCE-82-07517 and OCE 85-00270 and ONR contract N00014-80-C-0771. Contribution no. 601 from the Marine Sciences Research Center.
REFERENCES ATKINSON L. P. and F. A. RICHARDS (1967) The occurrence and distribution of methane in the marine environment. Deep-Sea Research, 14, 673-684. BERNER R. A. (1971) Principles of chemical sedimentology. McGraw Hill, New York, 240 pp. BROECKER W. S. and T.-H. PENC (1984) Gas exchange measurements in natural systems. In: Gas transfer at water surfaces, W. BRUTSAERT and G. H. JIRKA, editors, D. Reidel, pp. 479-493. FANNING K. A. and M. E. Q. PILSON (1972) A model for the anoxic zone of the Cariaco Trench. Deep-Sea Research, 19, 847-863. HESSLEIN R. H. (1980) Whole lake model for the distribution of sediment-derived chemical species. Canadian Journal of Fisheries and Aquatic Science, 37, 552-558. LAMONTAGNE R. A., J. W. SWINNERTON, V. J. LINNENBOMand W. D. SMITH (1973) Methane concentrations in various marine environments. Journal of Geophysical Research, 78, 5317-5324. LERMAN A. (1979) Geochemicalprocesses: water and sediment environments. Wiley, New York, 481 pp. RASMUSSEN R. A. and M. A. K. KHALIL (1981) Atmospheric methane (CFL): trends and seasonal cycles. Journal of Geophysical Research, 86, 9826-9832. REEBURGH W. S. (1976) Methane consumption in Cariaco Trench waters and sediments. Earth and Planetary Science Letters, 28, 337-344. RICHARDS F. A. (1975) The Cariaco Basin (Trench). Oceanography and Marine Biology Annual Reviews, 13, 11-67. RUDD J. W. M. and R. D. HAMILTON(1978) Methane cycling in a eutrophic shield lake and its effects on whole lake metabolism. Limnology and Oceanography, 23, 337-348. RUDD J. W. M., A. FURtrrANI, R. J. FLETr and R. D. HAMILTON(1976) Factors controlling methane oxidation in shield lakes: The role of nitrogen fixation and oxygen concentration. Limnology and Oceanography, 21, 357-364. SCRANTON M. I. (1977) The marine geochemistry of methane. Ph.D. Thesis. WHOI/MIT Joint Program, Woods Hole, 251 pp. SCRANTON M. I., P. C. NOVELLI and P. A. LOUD (1984) The distribution and cycling of hydrogen gas in the waters of two anoxic marine environments. Limnology and Oceanography, 29, 993-1003. SCRANTON M. I., F. L. SAYLES, M. P. BACONand P. G. BREWER (1987) Temporal changes in the hydrography and chemistry of the Cariaeo Trench. Deep-Sea Research, 34, 945-963. STEELE L. P., P. J. FRASER, R. A. RASMUSSEN, M. A. K. KHALIL, T. J. CONWAY, A. J. CRAWFORD, R. H. GAMMON, K. A. MASARIE and K. W. THONIN6 (1987) The global distribution of methane in the troposphere. Journal of Atmospheric Chemistry, 5, 125-171. SWlr~'ERTON J. W. and V. J. LINNENBOM (1969) LOw molecular weight hydrocarbon analyses of Atlantis II waters. In: Hot brines and recent heavy metal deposits in the Red Sea, E. DEGENS and D. Ross, editors, Springer-Verlag, New York, pp. 251-253.
Methane content of the Cariaco Trench
1523
SWINNERTONJ. W., V. J. LINNENBOMand C. H. CHEEK (1962) Revised sampling procedure for determination of dissolved gases in solution by gas chromatography. Analytical Chemistry, 34, 1509. WARD B. B., K. A. KILPATRICK,P. C. NOVELLIand M. I. SCRANTON(1987) Methane oxidation and methane fluxes in the ocean surface layer and deep anoxic waters. Nature, 327, 22.6-229. WIESENBURGD. A. (1975) Processes controlling the distribution of methane in the Cariaco Trench, Venezuela. MS thesis. Old Dominion University, 106 pp.
0198-0149188 $3.00+ 0.00 ~) 1988PergamonPressplc.
Printedin Great Britain.
T e m p o r a l variations in the methane content of the Cariaco Trench MARY I. SCRANTON*
(Received 22 October 1987; in revised form 10 May 1988; accepted 24 May 1988) A ~ t r a e t - - T h e deep waters of the Cariaco Trench are not in steady state with respect to distributions of either hydrographical or chemical properties. Over the past 15 years, methane concentrations at the bottom of the anoxic zone have increased steadily from about 7-9 to 12.5 ~tM. A simple model for the anoxic waters of the Cariaco Trench suggests that methane distributions may be explained as the result of the flux of methane from the sediments (about 12.5-17.5 ttmol cm-2 y-~), vertical eddy diffusion within the water column, and methane oxidation in the water column. Oxidation rates near the oxic-anoxic interface may be considerably higher than previously measured, although available data are inadequate to determine whether the methane oxidation near the interface occurs under anaerobic conditions or whether it occurs when oxygenated overlying water is mixed down into the top of the anoxic zone. It also is possible that methane fluxes from the shallower sediments of the Cariaco Trench are very low compared to those from the deeper sediments.
INTRODUCTION
METHANE is a terminal product of the anaerobic degradation of organic matter. Therefore, in anoxic water columns, methane concentrations can reach high levels, although concentrations are usually quite low in the overlying oxidized surface layers. Methane production (mostly within the sediments), methane oxidation within the water column and sediments, and physical mixing and diffusion seem to control methane concentrations in most anoxic basins. Previous work (REEBURGH, 1976; RUDD et al., 1976; RUDD and HAMILTON, 1978; HESSLEIN, 1980; WARD et al., 1987) has attempted to quantify the rates of processes controlling the observed distributions using various combinations of direct measurements and modeling. In this study I discuss the temporal variations of methane in the Cariaco Trench using a model that has been used previously to analyse temperature, salinity, hydrogen sulfide and silica distributions (SCRANTONet al., 1987). THE C A R I A C O T R E N C H
The Cariaco Trench is a large, permanently anoxic basin located on the continental shelf of Venezuela at 10°30'N 65°31'W. It is about 200 km long and 50 km wide with a maximum depth of about1400 m. The basin is separated from the Venezuelan Basin and the resi of the Caribbean Sea by a sill at about 150 m, above which water can exchange freely with surface waters further offshore. The deep part of the basin is divided into two sub-basins by a saddle at about 900 m. The oxic-anoxic interface has been found at depths between 240 and 400 m, although details of the variation in interface depth with * Marine Sciences Research Center, State University of New York, Stony Brook, NY 11794-5000, U.S.A. 1511
1512
M.I. SCRANTON Table 1. Sources for methane data
Source ATrdNSONand RICHARDS (1967)
Date Feb. 1965
Location 10040.5'N 65°31.5'W 10o31,N 64o40'W
WIESENBURG(1975)
Feb. 1974
10°36.0'N 65°38.3'W 10°38.3'N 65°37.6'W 10o39.0'N 65°36.5'W 10°38.0'N 65°36.5'W 10°30.0'N 64°40.1'W 10°29.8'N 64°38.9'W
LAMONTAGNE (personal communication)
May 1971 June 1976
10°38'N 65°45'W 10°41'N 65°36'W 10o41'N 65o38'W
SCRANTON(1977)
Feb. 1975
l&29'N 64°44'W
SCRANTON(this report)
Nov. 1982 Feb. 1986
10°31'N 64°46'W 10°40'N 65°31'W
time are not well known. It is unlikely that the depth of the interface has been significantly overestimated, as the human nose is capable of detection of even trace levels of hydrogen sulfide. The first data for methane in the Cariaco Trench were obtained by AaxxNsoN and RICHARDS(1967), who collected samples from stations in the two sub-basins (see Table 1 for station locations and dates of collection for all methane data discussed in this paper). Methane data also are presented in the Master's thesis of WlESENBURG (1975), by LAMOrCrAONEet al. (1973) and LAMOrCrAGNE(personal communication), and by myself and coworkers (SCRArCror~, 1977; WARD et al., 1987; SCRArZroN,this report). Data from Lamontagne and coworkers were collected on two separate cruises (1971 and 1976) and our data were obtained on three cruises (1975, 1982 and 1986). Most of the data available are from the western basin, although Aavar~soN and RICHARDS (1967), WIESENBURG (1975) and SCRANTON(1977, this report) have obtained data from the eastern basin. METHODS Methane data were obtained from water samples collected in Niskin bottles and transferred to sample bottles in the same manner as oxygen samples (filled from the bottom, overflow of at least one bottle volume). AalaNsoN and RaCHARDS(1967) report that all samples to be analysed at sea were stored at I°C. Samples to be run in the lab were collected in ampoules, poisoned and stored in a cool, dark place. Lamontagne and coworkers analysed all samples at sea, but apparently did not poison them, as they also were measuring unsaturated organics (SwINNERTONand LINNENBOM,1969). Wiesenburg's methane data were obtained on samples poisoned with sodium azide and stored for about 1 month before analysis (WIEsE~,~aURO, 1975). (Note: sodium azide does not poison all metabolic processes in anoxic environments.) My 1975 samples were placed in ground glass stoppered bottles and were poisoned immediately after collection with 1 ml saturated mercuric chloride solution per liter. Most of the samples were analysed on board ship, but a few were returned to the
M e t h a n e content of the Cariaco T r e n c h
1513
laboratory and analysed approximately 7 weeks after collection. Samples obtained in 1982 were transferred to 700 ml champagne bottles, poisoned by addition of 1 ml of saturated mercuric oxide solution, and capped using a plastic-lined metal cap and bottle capper. Upon return of the samples to the laboratory, a small bubble was found in each bottle. The bubble volume was estimated and the amount of methane partitioned into the headspace was calculated. Samples were run in the laboratory approximately 6 weeks after collection. The methane samples from the 1986 cruise were collected in ground glass stoppered bottles and were analysed immediately after collection on board ship. Modifications of the extraction and trapping method of SWINNERTONet al. (1962) were used in all studies (ATraNSON and RICHARDS,1967; LAMONTAGNEet al., 1973; WIESENBURG, 1975; SCRANTON, 1977) except in 1982 and 1986 when the vacuum extraction method of SCRANTONet al. (1984) was used to strip gas from solution. All methane analyses were carried out using flame ionization gas chromatography. Calibration for our three cruises was performed by injections of commercially prepared methane-in-nitrogen standard mixtures. The standards used in 1982 and 1986 were calibrated using static dilution techniques, while the commercially quoted standard value was used in 1975. Possible corrections to the 1975 standard value are discussed below. Other workers used similar procedures. RESULTS
AND DISCUSSION
Although many workers have treated the Cariaco Trench as a steady-state system (FANNING and PILSON, 1972; REEBUR~H, 1976), this is not an accurate description of the behavior of chemical or hydrological parameters over the past 20 years (RICHARDS, 1975 ; SCRANTONet al., 1987). In particular, temperature, salinity, hydrogen sulfide and silica have increased markedly in the deep parts of the basin (for example, approximately a 33% increase for sulfide between 1973 and 1982). We recently developed a model that predicts observed trends in hydrographic and chemical parameters in the Cariaco Trench (SCRANTONet al., 1987; Figs 1 and 2). It is a simple non-steady-state box model consisting of 12 boxes, each 92 m thick (50 fathoms). SCRANTONet al. (1987) restricted their model to the eastern basin, largely because the best hydrographic data are available from that end of the Cariaco Trench. However, for methane, no recent data are available from the eastern basin, so the western basin is modeled here instead. In our original study, diffusive fluxes of salinity, temperature, sulfide and silica in the water column out of, or into, the eastern basin across the top of box 8 (915 m) were assumed to be the same as those across the top of the eastern basin. This assumption was reasonable as profiles of most chemical parameters (and therefore chemical gradients) in the two basins are very similar (SCRANTONet al., 1987). Fewer data are available for methane, but examination of the historical data suggest that a small (1.4 ~tM) interbasin difference in methane concentration, beginning at depths as shallow as 600 m, may exist (WIESENaURG, 1975). Thus we have included both deep sub-basins in our model for methane (Fig. 1, Table 2). The model includes transport of chemical species between boxes by eddy diffusion. Supply (or removal) of species also takes place by diffusion from (or into) the sediment and is assumed to be proportional to the area of bottom intersected by the box (thus taking the bathymetry of the basin into account). Values for vertical eddy diffusion coefficients were determined by SCRANTONet al. (1987) by fitting the temperature and
1514
M . I . SCRANTON
\
ANOXIC INTERFACE BOX I 2 5
WESTERN BASIN Fig. 1.
5 6 7 8 _ _ _ 9 _ I0 II 12
EASTERN BASIN
Schematic of geometry of box model (adapted from SCRANTON et al., 1987).
Ci-l, vi-I
BOX i - I
]
AZ F-~
L
Ci-i)
/Ai-i /
Ci ,Vi Ki+l (Ci+l Ci). /Ai+ I BOXi+l I AZ Ci+"Vi+'/ iAi+ 2
Fig. 2. Flux terms and notation for box model calculations (taken from SCRANTON et al., 1987 and Erratum, Deep-SeaResearch,34, 1653). Ai, IT,-and C, are, respectively, the area, volume and methane concentration for box i. Ki is the vertical eddy diffusivity at the top of box i. Az is the thickness of the box (92 m).
salinity data as described. Sensitivity studies have suggested that the error associated using eddy diffusion coefficients derived from temperature changes between 1973 and 1982 (as was done by SCRANTONet al., 1987), rather than between 1974 and 1986, will be small. Hydrographic data from the cruises in which methane was analysed are not sufficiently detailed or of sufficient precision to justify a modification of the earlier constants. Basin volumes (Table 2) are different from those presented in SCRANTONet al. (1987) because an error was made in calculation of the earlier box volumes (see Erratum published in Deep-Sea Research, 34, 1653). Use of the correct volumes results in the use of a lower eddy diffusion coefficient (decreased by 20%), but otherwise the model fits to temperature, salinity, hydrogen sulfide and silica are essentially the same as presented in SCRANTONet al. (1987). Values for the eddy diffusion coefficients used in this study are also presented in Table 2. Earlier work (SCRANTONet al., 1987) suggested that temperature and salinity increases are the result of downward flux of heat and salt from the interface, while hydrogen sulfide concentrations are increasing because flux from the sediments exceeds vertical diffusive loss. Silica concentrations appear to be affected both by diffusion of silica from the sediments and by dissolution of siliceous particles within the water column.
1515
Methane content of the Cariaco Trench
Table 2.
Model parameters for the Cariaco Trench box model
Basin area (10 9 m 2) East 320 412 503 595 686 778 870 961 1052 1144 1236 1327
Basin volume (1011 m 3) West
East
8.22 7.60 6.97 6.31 5.70 5.03 4.40 1.35 1.15 0.92 0.72 0.51
Ki (cm2 s-1) West
7.30 6.70 6.10 5.50 4.90 4.30 3.70 2.40 2.10 1.80 1.50 1.10
1.20 0.95 0.75 0.56 0.47
2.00 1.80 1.50 1.20 1.00
0.49 0.85 1.02 1.42 4.07 4.07 4.07 4.07 4.07 4.07 4.07 4.07
There also is evidence that methane concentrations have increased with time. Data from 1965 (AavaNsoN and RICHAROS, 1967; eastern and western basins), from 1968 and 1969 (LAMotcrAGNEet al., 1973; basin unidentified), from 1971 (LAMorcrAGr~E, personal communication; western basin) and from 1974 (WmSEt~-aURC, 1975; eastern and western basins) all indicate deep methane concentrations of about 7 laM (Fig. 3). Wiesenburg's data suggest some interbasin variability in methane concentrations, with eastern basin bottom water concentrations of about 7.2 IxM and western basin concentrations of about 6.0 IxM. Data from the eastern basin collected by Scranton in 1975 indicated bottom water concentrations of about 9 IxM (although stored samples gave values of 7 IxM; SCaArcroN, 1977). Lamontagne and colleagues also found methane concentrations in the western basin deep waters to be about 9 laM in 1976 (Fig. 4). In contrast western basin samples collected in 1982 and 1986 yielded methane concentrations in bottom samples of 10.8 and 12.5 laM, respectively (Fig. 4). One of the difficulties in comparing historical data sets is that there have been few intercalibrations of standards (and concentrations quoted for commercially obtained standards can be incorrect). Therefore I am unable to determine the absolute accuracy of the early methane data from the Cariaco Trench. In the case of my 1975 data, I can attempt a qualitative evaluation of the standardization by comparison of measured atmospheric methane concentrations (about 1.3 ppmv) with values interpolated for 1975 from the data presented in RASMUSSENand KHALIL(1981). The interpolated atmospheric methane concentration is 1.5 ppmv, somewhat higher than we observed. Thus the standard value used apparently was not too large, as air concentrations were lower than expected while water concentrations were greater than expected [compared, for example, to the data of WIESENBURG(1975)]. I recognize that such a comparison can only be approximate, since methane concentrations in the atmosphere are known to vary with season, with latitude, and with proximity to methane sources (STEELE et al., 1987). The comparison suggests that my standard was adequate to within about +15%, and that deep methane concentrations are not too high due to a standard calibration problem. The 1975 samples are from the eastern basin, found by Wiesenburg to have the higher methane concentrations (WIESENBURG, 1975).
1516
M.I. SCRANTON C H 4 ,/J.M
5
EO
I
I
200 300
f% ,Px°a
40( 50(
£ E
600
i x
•
o
jr3
ex
•g
o
g. x
700
LIJ r~
800
x •
o
•
o
900 •
I000 ItO0
•
o
x •
o
(*)
1200
x
1300
x
o o
(o)
•s
Fig. 3. Methane data for the Cariaco Trench between 1965 and 1975. All data are for the western basin unless otherwise noted and were taken from the following sources: (3, collected in November 1965 (ATKINSON and RtCHARDS, 1967); x, collected in May 1971 (LAMONTAGNE, personal communication); Q, collected in February 1974 (WIESENBURG, 1975; REEBffROH, 1976); D, collected in February 1975 from the eastern basin (SCRA~rroN, 1975). The points in parentheses from ATrdNSON and PdCHARDS(1967) appear to be low, but were not discarded by the original authors.
I have used the model described above to explain the distribution of methane in the Cariaco Trench. In the following discussion, I have used two different data sets for the initial methane distribution. First, I have used the 1975 eastern basin methane profile collected by SCRAWrON (1977). Since no data are available from that cruise from the western basin, I have assumed that the distributions in the eastern basin and western basin are identical. As an alternative, I have used the 1974 data of WIESENBURG(1975) who obtained data from several stations in each basin. In the case of Wiesenburg's data, for depths above saddle depth (about 900 m), I have taken an average methane concentration since I assume that boxes above sill depth are well mixed laterally. Below the depth of the saddle, I have used the actual profiles as initial conditions. Model results do not vary significantly regardless of whether different profiles are used for the two basins or whether the basins are considered to have identical methane profiles (results not shown). With each set of methane data I have used the box areas and volumes for the two basins presented in Table 2. For each time interval, the transports into and out of each box from sediment diffusion and eddy diffusive transport were calculated and were
1517
Methane content of the Cariaco Trench CH4 ,~M
2oo_
IO
15
300~
1%o~ ~
×
40O
o m lo o
5OO
o. xx omm
60O
[3
E"
£
Ii 0.-
70O eI
80O o
9OO
xx
e_m e
Iooo
x o
11oo
xx
•
o
12oo 1300 ° o
Fig. 4. Methane data for the Cariaco Trench between 1975 and 1986. All data are from the western basin unless otherwise noted and were obtained from the following sources: [], collected in February 1975 from the eastern basin (SCRANTON, 1977); O, collected in June 1976 (LAMorrrAGNE, personal communication); x, collected in November 1982 (this work); Q, collected in March 1986 (this work).
added to or subtracted from the values from the previous iteration. Time steps of 0.1 year were used, as shorter periods did not affect the results. In Fig. 5 the 1986 experimental data are presented together with three curves generated for the western basin using the 1975 data as the initial condition, and values for Kv, area and volume from Table 1. In order to calculate model profiles, I made an estimate of the flux of methane across the sediment-water interface. Model profiles for a sediment-water flux of 12.5 lmtol cm-2 y-1 (curve B) agree quite well with 1986 western basin methane data near the bottom. Similar sediment-water fluxes (best fit = 20 pmol cm-2 y-l) apply if I use Wiesenburg's 1974 data as the initial methane distribution. In fact, it does not appear to be possible to generate the observed bottom methane increase, even in the absence of methane oxidation in the water column, without a sediment-water flux of this order of magnitude. Model calculations indicate that the sediment-water flux was not likely to be much less than 5 pmol cm-2 y-1 since fluxes lower than this value would predict decreases in methane concentration with time. Few data exist on methane distributions in the sediments of the Cariaco Trench with which to compare these predicted fluxes, but REEBURCH (1976) indicated pore water values in the top 20 cm of the sediment were on the order of 10--40 pM. If we assume that
1518
M . I . SCRAm'ON CH4,~M 5
I0
15 i
200 30O 40(3 •
\
50(3 E 600
=).-.
a_ 700 80O 9OO 1000 1100 1200 1300
Fig. 5. Model fits with n o / n situ consumption but with different fluxes of methane from the sediments. (A) F = 17.5 ~mol cm -2 y-l; (B) F = 12.5 )unol cm -2 y-l; (C) F = 7.5 gmo1-2 y-a.
the methane content in the surface sediments is 40 ~tM, and noting that the bottom water concentration in 1974 at the time of Reeburgh's study was about 7-9 ~tM, we can estimate the rate of diffusion out of the sediments using dC Flux = ~ Ds-d-~z (BERNER, 1971), where ~p is the porosity (about 0.8). Ds is the diffusion coefficient for methane in sediments [approximately equal to ~p2 D (LERMAN, 1979)]. When D is the free solution diffusion coefficient of approximately 2 x 10-5 cm 2 S- 1 (estimated for 17°C from data for gases of similar molecular weight presented by BROECrmRand PENG, 1984), Ds equals 1.3 X 10-5 cm 2 s-1. The gradient of methane concentration is dC/dz at the interface and can be estimated at (40-9 ixM)/1 cm. Thus the flux is about 10 ~tmol cm-2 31-1. Of course the actual flux will be highly sensitive to details of the shape of the profile of methane concentration with depth near the interface. The data presented by REEBURGrI(1976) are somewhat scattered and may indicate disturbance of the sediments during coting. Since it seems much more likely that methane was lost from the sediments during sampling than that the methane concentration was increased, the above estimate of sediment-water flux may represent a lower limit to the true flux. In his paper, R~EBUR~H (1976) estimated that the flux of methane from the deeper
Methane content of the Cariaco Trench
1519
part of the core to the near-surface zone was on the order of 5-15 I~mol cm -2 y-l, but that this flux was almost completely oxidized by anaerobic methane oxidation before methane could diffuse into the overlying water column. This would suggest that fluxes of methane across the sediment-water interface should be much lower than 5-15 lxmol cm 2 s-1. However, the diffusion coefficient used by Reeburgh (3 x 10-6 cm 2 s-i) is low, suggesting that actual fluxes at depth would be higher than reported in REEBURGH(1976). In addition, my calculation does not imply that methane oxidation within the surficial sediments is unimportant. On the contrary, it is likely that most of the methane produced at depth in the sediments is consumed before diffusing out. It does appear, however, that a significant amount of methane escapes oxidation and reaches the bottom waters of the Trench. The alternative possibility, that large amounts of methane are produced in the water column, seems even less likely than the possibility that sediment-water fluxes are relatively large, considering the fact that methane oxidation is known to be important in the deep waters (WARDe l al., 1987) and that water column processes (such as sulfate reduction) are believed to be quite slow (ScRAWrONet al., 1987). Clearly, we need more detailed and precise data for methane profiles in Cariaco Trench sediment before the relative importance of methane oxidation and loss across the sediment-water interface can be quantified. As discussed below, it is possible that there is no flux of methane from the sediments at some depths in the Cariaco Trench. One way to estimate the magnitude of the sediment-water flux for methane is through use of models such as the one developed to describe distribution of dissolved species in the Cariaco Trench. Although the model results presented in Fig. 5 agree satisfactorily with the data near the bottom of the Cariaco Trench, predicted methane concentrations near the oxicanoxic interface, calculated from all three fluxes, are much higher than observed. One explanation for the poor fit near the interface of the curves in Fig. 5 is the presence of methane oxidation in the anaerobic zone. Figure 6 presents observed methane concentrations together with model fits generated using the "best fit" sediment-water flux from Fig. 5 (12.5 I~mol cm -2 y-l) and different arbitrary patterns of methane oxidation in the anaerobic zone. The interface was assumed to be at 274 m, the precisely determined interface in 1983 and the depth used in previous modeling efforts. Curve A was obtained by making the assumption that methane oxidation (at the rate of 0.82 nmol 1-1 d-1 or 0.3 ~tmol 1-1 y-l) occurred only in the top box, curve B assumes oxidation at the same rate in boxes 1 and 2 and curve C assumes oxidation in boxes 1-4. The details of "best fit" oxidation rates and the depth range over which oxidation occurs are strongly dependent on the choice of initial data set. If, for example, WIESEt~URG'S (1974) western basin data are used (bottom water methane concentrations of 7.5 I~M), oxidation (at a rate of 0.2 ~tmol 1-1 y-l) is required only in the top two boxes to obtain good fits of the model to the data (Fig. 6, curve D). Therefore the values of the model-derived oxidation rates and their variation with depth are dependent on model parameters. For comparison, I have also calculated model profiles using the oxidation rates measured by WARD et al. (1987). Their data suggest that maximum methane oxidation rates were attained in the bottom of the Trench but that oxidation near the interface was negligible. As can be seen in Fig. 7 (curve A), if the sediment-water flux is increased to 17.5 lamol cm -z y-l~ the model profiles agree quite well with the data near the bottom.
1520
M . I . SCRANTON CH 4 ,p.M 5
I0
15
200 3OO 4OO 5OO 6OO E I
7OO 8OO 9OO I000 I100 1200 1300
Fig. 6. Model fits with sediment-water flux of 12.5 ~tmol cm -2 y-1. Consumption of methane is assumed to take place at a rate of 0.3 Ixmol 1-1 y-1 but over varying portions of the water column (line A: box 1 only; line B: boxes I and 2; line C: boxes 1-4). Curve D demonstrates the effect of choice of a different set of initial data (here, data from WIESENBURG, 1977). Curve D was generated using a sediment-water flux of 20 Itmol cm -2 y-i and the assumption that consumption takes place at a rate of 0.2 t~mol 1-1 y-1 in boxes 1 and 2.
However, the predicted methane concentrations near the interface are as high as in the case described in Fig. 6. Since it is possible that methane fluxes could vary with depth, I also have examined the case where methane fluxes from the sediments in boxes 1-4 are zero, while fluxes at depth are high enough to produce the observed bottom water increases. In this case as well, predicted methane concentrations near the interface are in excess of observed values (Fig. 7, curve B) although the discrepancy is less. If oxidation rates throughout the water column are increased to 0.15 l~mol cm 2 y-l, the data can be matched quite well with the model (Fig. 7, curve C). Thus, it seems that long-term average oxidation rates near the interface may be higher than rates measured directly by WARD et al. (1987). Model predicted rates are about 0.15-0.3 I~mol 1-1 y-1 (0.44).8 nmol 1-1 d-1) compared with rates measured near the interface by WARD et al. (1987) of less than about 0.02 ~tmol 1-1 y-1 (0.06 nmol 1-t d-l). The discrepancy between the direct measurements of WARD et al. (1987) and the modelpredicted results may have several explanations. Ward and coworkers measured methane oxidation rates at a season of low primary productivity over a limited period of time (hours) while model results reflect long-term averages of processes. Since the input of
1521
Methane content of the Cariaco Trench
CH4,~M 5 2OO
I0
15
1 i
3O0 4OO 500 6OO E -r 7 0 0 Io.. iJJ a
800
900
I000 IlOO 120o I:.300
Fig. 7. Model fits using the 1974 Scranton data as the initial condition. Curve A was generated with a sediment-water flux at all depths of 17.5 ttmol cm 2 s -1 and methane oxidation rates for each box estimated from the data in WARD e t al. (1987). Curve B also used the Ward e t al. oxidation rate data, but assumed that no methane diffused across the sediment-water interface in boxes 1-4. Curve C assumes a sediment-water flux of methane of zero in boxes 1-4, a flux of 17.5 ttmol cm 2 s-1 in boxes 4-12 and a constant methane oxidation rate of 0.15 ttmol 1-1 y-1.
carbon to the Cariaco Trench deep waters is most probably seasonally variable, water column oxidation rates, especially near the interface, might vary with time. Alternatively, the indication that oxidation rates are highest near the interface may suggest that occasional turbulent mixing of oxygen into the upper part of the anoxic zone accompanied by "aerobic" methane oxidation may actually be the most important process removing methane in this system. There is a suggestion of such an effect in the sulfide data which shows a narrow zone (few tens of meters) of concave upward curvature near the anoxic-oxic interface which cannot be ascribed to the processes included in our 1987 model (SCRANTONet al., 1987). Because of the sensitivity of the model fits to the precise nature of the initial profiles used, errors in the early concentration data may help to explain the fact that one would have expected injection of oxygenated water to occur only in the top box. CONCLUSIONS
Methane concentrations within the deep waters of the Cariaco Trench have increased with time since 1965, while concentrations in waters above about 900 m have remained nearly constant. By applying the non-steady-state box model of SCRA~rrONet al. (1987) to
1522
M . I . SCRArCrON
the data, it appears that the major source of methane to the basin waters is diffusion from the sediments. Temporal variations in methane concentrations at depths greater than 900 m can be explained by the time-dependent box model if the flux from the sediments is about 12.5-17.5 Ixmol cm -2 y-1. The sediment-water flux must be greater than 5 Ixmol cm -2 y-1 or methane concentrations would be decreasing with time, in contrast to observations which suggest concentrations are remaining the same or are increasing (depending on depth). Methane concentrations near the oxic-anoxic interface are lower than would be predicted by a model that includes only a sediment source and transport within the water column by turbulent diffusion. Methane oxidation near the interface seems to be an important process. Model calculations suggest oxidation rates throughout the anoxic zone are on the order of 0.15-0.3 ~tmol 1-1 y-1 (0.4-0.8 nmol 1-1 d-l). Acknowledgements--I wish to thank Paul C. Novelli and Peter A. Loud for their assistance in collecting and analysing Cariaco Trench methane samples and Drs Denis Wiesenburg and Robert Lamontagne for providing me with methane data from their earlier cruises. Financial support for this work was provided by NSF grants OCE-82-07517 and OCE 85-00270 and ONR contract N00014-80-C-0771. Contribution no. 601 from the Marine Sciences Research Center.
REFERENCES ATKINSON L. P. and F. A. RICHARDS (1967) The occurrence and distribution of methane in the marine environment. Deep-Sea Research, 14, 673-684. BERNER R. A. (1971) Principles of chemical sedimentology. McGraw Hill, New York, 240 pp. BROECKER W. S. and T.-H. PENC (1984) Gas exchange measurements in natural systems. In: Gas transfer at water surfaces, W. BRUTSAERT and G. H. JIRKA, editors, D. Reidel, pp. 479-493. FANNING K. A. and M. E. Q. PILSON (1972) A model for the anoxic zone of the Cariaco Trench. Deep-Sea Research, 19, 847-863. HESSLEIN R. H. (1980) Whole lake model for the distribution of sediment-derived chemical species. Canadian Journal of Fisheries and Aquatic Science, 37, 552-558. LAMONTAGNE R. A., J. W. SWINNERTON, V. J. LINNENBOMand W. D. SMITH (1973) Methane concentrations in various marine environments. Journal of Geophysical Research, 78, 5317-5324. LERMAN A. (1979) Geochemicalprocesses: water and sediment environments. Wiley, New York, 481 pp. RASMUSSEN R. A. and M. A. K. KHALIL (1981) Atmospheric methane (CFL): trends and seasonal cycles. Journal of Geophysical Research, 86, 9826-9832. REEBURGH W. S. (1976) Methane consumption in Cariaco Trench waters and sediments. Earth and Planetary Science Letters, 28, 337-344. RICHARDS F. A. (1975) The Cariaco Basin (Trench). Oceanography and Marine Biology Annual Reviews, 13, 11-67. RUDD J. W. M. and R. D. HAMILTON(1978) Methane cycling in a eutrophic shield lake and its effects on whole lake metabolism. Limnology and Oceanography, 23, 337-348. RUDD J. W. M., A. FURtrrANI, R. J. FLETr and R. D. HAMILTON(1976) Factors controlling methane oxidation in shield lakes: The role of nitrogen fixation and oxygen concentration. Limnology and Oceanography, 21, 357-364. SCRANTON M. I. (1977) The marine geochemistry of methane. Ph.D. Thesis. WHOI/MIT Joint Program, Woods Hole, 251 pp. SCRANTON M. I., P. C. NOVELLI and P. A. LOUD (1984) The distribution and cycling of hydrogen gas in the waters of two anoxic marine environments. Limnology and Oceanography, 29, 993-1003. SCRANTON M. I., F. L. SAYLES, M. P. BACONand P. G. BREWER (1987) Temporal changes in the hydrography and chemistry of the Cariaeo Trench. Deep-Sea Research, 34, 945-963. STEELE L. P., P. J. FRASER, R. A. RASMUSSEN, M. A. K. KHALIL, T. J. CONWAY, A. J. CRAWFORD, R. H. GAMMON, K. A. MASARIE and K. W. THONIN6 (1987) The global distribution of methane in the troposphere. Journal of Atmospheric Chemistry, 5, 125-171. SWlr~'ERTON J. W. and V. J. LINNENBOM (1969) LOw molecular weight hydrocarbon analyses of Atlantis II waters. In: Hot brines and recent heavy metal deposits in the Red Sea, E. DEGENS and D. Ross, editors, Springer-Verlag, New York, pp. 251-253.
Methane content of the Cariaco Trench
1523
SWINNERTONJ. W., V. J. LINNENBOMand C. H. CHEEK (1962) Revised sampling procedure for determination of dissolved gases in solution by gas chromatography. Analytical Chemistry, 34, 1509. WARD B. B., K. A. KILPATRICK,P. C. NOVELLIand M. I. SCRANTON(1987) Methane oxidation and methane fluxes in the ocean surface layer and deep anoxic waters. Nature, 327, 22.6-229. WIESENBURGD. A. (1975) Processes controlling the distribution of methane in the Cariaco Trench, Venezuela. MS thesis. Old Dominion University, 106 pp.