A sediment trap intercomparison study in the Santa Barbara Basin

Earth and Planetary Science Letters, 53 (1981) 409-418 Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands

409

[61

A sediment trap intercomparison study in the Santa Barbara Basin Jack Dymond, Kathy Fischer, Milo Clauson, Richard Cobler Oregon State University, Corvallis, OR ¢7331 (U.S.A.)

Wilford Gardner, Mary Jo Richardson Lamon t-Doherty Geological Observatory of Columbia University, Palisades, N Y 10964 (U.S.A.)

Wolfgang Berger, Andrew Soutar and Robert Dunbar Scripps Institution of Oceanography, University of California at San Diego, La Jolla, CA 92093 (U.S.A.)

Received April 21, 1980 Revised version received February 2, 1981

Four sediment traps of radically different design were deployed in the Santa Barbara Basin for approximately 45 days. The measured fluxes ranged from 370 to 774 g m-2 yr I for the different designs. These values lie within flux measurements previously determined for the basin. Compared to the 25-year record (920 g m -2 yr-i), however, all fluxes determined in this experiment are somewhat low. Because this experiment was conducted during a general period of high storm activity and runoff, measurement of greater than average flux was expected. It is probable that the higher flux recorded by the sediments results from a significant input of detritus into the basin by near bottom transport. The chemical composition of trapped material was nearly identical in all four trap designs. The deep cone design, however, had a significantly lower Mn content. Since this trap was the only one in which reducing conditions were produced in the sample container, reduction and mobilization of manganese after collection is believed to have occurred. In spite of the very different designs tested, the factor of two agreement in flux determination and the compositional similarity of the material collected is encouraging for future attempts to directly measure the flux of particulates in the ocean.

1. Introduction I n recent years the d e p l o y m e n t of sediment traps has b e c o m e technically feasible in the open ocean. Traps have been used to collect material which is falling through the water c o l u m n as a m e a n s of m e a s u r i n g the flux of solid phases a n d associated elements [ 1 - 6 ] . Recent laboratory studies suggest the t r a p p i n g efficiency of different trap designs can vary greatly [7,8]. I n order to directly compare the variations in b o t h mass a n d c o m p o s i t i o n of material collected by traps of dif-

ferent designs, we deployed four radically different traps for 48 days i n the Santa Barbara Basin. The designs used were: (1) paired cones, (2) folding box or "suitcase", (3) cylinder, a n d (4) closing box (Fig. 1). I n a d d i t i o n to the trap deployment, current meters a n d a recording n e p h e l o m e t e r were deployed for the d u r a t i o n of the experiment. The S a n t a B a r b a r a Basin was chosen for the sediment trap i n t e r c o m p a r i s o n because previous studies of sediment flux to the seafloor a n d sedim e n t a c c u m u l a t i o n provide a data base which could be helpful in the i n t e r p r e t a t i o n of our re-

0012-821X/81/0000-0000/$02.50 © Elsevier Scientific Publishing Company

410 suits. The basin has a maximum depth of 580 m and a sill depth of 470 m. A low oxygen environment, which inhibits benthic life, exists below the sill depth. This low oxygen environment is due to the high productivity of the overlying waters and the location of the oxygen minimum at the sill depth. Consequently, the sediment, which accumulates at a rate of approximately 4 mm y r - l or 92 mg cm 2 y r - 1 [9,10], is not disturbed by burrowing of benthic organisms and is finely varved. The varves represent alternations between sediments rich in biogenic material that accumulate during the spring and summer, and those rich in terrigenous clays and silts that accumulate primarily during the winter storm season [2].

2. Trap designs 2.1. Soutar cone trap The cone trap is a design used for several years in studies of the Santa Barbara Basin [2,10]. The particular configuration used in this experiment was a modification of traps originally designed for hydrocarbon studies and consists of paired funnels 120 cm long and 57 cm in diameter ( 1 / 4 m 2 collecting area for each cone). The funnel tapers to a 6-cm-diameter bottom opening which is attached to a removable collection cup 25 cm long. The funnels and collection cup are teflon-coated stainless steel. A honeycomb grid which acts as a false bottom interceptor [2] of particles is placed in the mouth of the cones. The grid is composed of epoxy-coated nylon material and has cells 1 cm in diameter and 5 cm deep. Laboratory flume studies indicate a grid of these dimensions prevents the penetration of turbulent eddies into the cones that are induced by the flow of water over the traps. The main advantages of this design are: (1) the preconcentration of sample into the collecting cup allows easy sample processing after recovery; (2) the deep cone design is amenable to in situ poisoning, although this was not done during this experiment; (3) there are no moving parts or closures, which eliminates mechanical or electronic malfunctions.

2.2. Gardner cylinder trap The cylinder trap design was based upon laboratory flume studies which suggest sediment traps collect particles through fluid exchange [7]. Under the influence of horizontal currents, it has been shown that water enters the trap near the downstream wall and particles settle from the water during the flow through the trap. Gardner [7] found that cylinders with height to width ratios between 2:1 and 3:1 collect at approximately 100% efficiency (i.e., collection rate equals vertical flux in currents less than 10 cm s - l ) . The cylinder trap used in this intercalibration was a PVC cylinder 25 cm in diameter and 62 cm high (0.05 m 2 area). An internal butterfly valve powered by surgical tubing and triggered by an external electronic timer provided a closure. The small size of this trap allows numerous traps to be deployed and recovered easily. The all-plastic construction is appropriate for trace metal studies.

2.3. Berger box ("suitcase") trap This trap consists of four identical 0.17-m 2 chambers in a fiberglass housing which are covered by a honeycomb grid. The grid, intended to act as an artificial seafloor and damper of eddies, is composed of the same epoxy-coated nylon material used in the Soutar cone trap. The cell dimensions of the grid, however, are 10 cm deep by 1 cm diameter, giving twice the height to width ratio for the baffling in this trap. In the chamber beneath the grid are located nylon bags with drawstrings connected to individual spring motors. These motors are released by an electronic timer to close the bags prior to recovery. In addition, the entire trap hinges at the center allowing the two chambers of each side to fold shut during launch and recovery. The relatively large surface area (0.68 m 2) makes this trap useful in pelagic areas with low sedimentation rates. The suitcase-like closure reduces contamination during deployment and greatly decreases the drag of the trap through the water during deployment and recovery. Each of the four chambers provide a presumably identical sample for use in various types of analyses.

411

2.4. OS U box trap

DEPTH Om--

SEDIMERTTRAPMOORINGS SOUTAR-BERGER

DY~ND-~RDRER

ARRAYI

AR~Y2

I I

~

8 - 17" glass s~heres B -

17"

glass spheres

OSUbox trap (162 m)

k Current ~ter (166m)

Gardner cylinder trap (213

15 inch glass spheres

m)

Nephelometer (254m) 17 inch glass spmeres

An earlier version of this trap was successfully deployed for eight months on the Galapagos spreading center [4]. The present design is a box with a square meter of collecting surface. Plastic baffles 10 cm square and 30 cm deep were placed in the top of the trap to inhibit turbulence and loss of sediment from resuspension. A waterproof nylon curtain, driven by titanium springs and released by a melting wire connected to a timer circuit, sealed the lower half of the trap prior to recovery. This trap features a large collecting area (1 m 2) and all-plastic or fiberglass construction.

OSUbox trap (292 m) I

1

]

Suitcase trap (341 m)

Conetrap

(381

Gardner cylinder trap (328 Current meter (330m)

m)

m)

Dual acoustic releases 400 Ib anchor

•

0 Ir~

Dual acoustic

|

releases

L

1500 Ib anchor

~///// ////~X~(////////W///'i~~ ~~7/////y/////////////////~~~(//////////!////~:~Hi:/; Fig. 1. Mooring configuration for the two moorings used in the experiment.

3. Mooring The traps were deployed on two moorings at depths between 162 and 381 m (Fig. 1; Table 1) from 2 March to 19 April 1978. Two-in-one braided nylon line, glass sphere flotation, and dual A M F Model 200 acoustic releases were used with both moorings. In addition to the traps, two Aanderaa current meters and a recording nephelometer were deployed on one of the moorings. The mooring with the suitcase box trap and the paired cones had an excess buoyancy of approximately 200 lbs. The mooring which held the cylinders and OSU box traps had an excess buoyancy of approximately 600 lbs.

TABLE1 Sedimenttrap flux data Trap

Berger suitcase Soutar cone Gardner cylinder upper lower b OSU box upper lower a,b

Depth (m)

Deployment period (hours)

Area (m z)

Mass recovered (g)

Flux (g m 2 yr

341 381

1062 1148

0.17 0.54

13.75 50.484

668 714

213 328

1092 1092

0.05 0.05

4.816 2.894

774 465

162 292

1020 not closed

1.0 1.0

43.072 8.51

370 73

a Because this trap did not close, most of the sediment was lost. b Poisoned using a combination of antibiotics.

l)

412

4, Sample processing

four-way splitter design modified after McManus et al. [11]. Two of the four splits were refrigerated and retained for freeze drying. Within five days after their collection these splits were washed with distilled water and centrifuged to remove salts and then freeze dried. The weight of each dry split provided an estimate of the precision of the splitting process, and these values were used to calculate the total flux. In all cases the weight of these two splits agreed to within 7%. Further splits of the dry material for chemical analyses were made using a dry splitter after grinding the material to < 200 /~m. One of the remaining wet splits was treated with buffered formalin and retained for

Immediately upon recovery each trap sample was passed through a 1-mm nylon sieve to remove larger organisms which Soutar has observed may swim into the trap and contribute erroneously to the flux. There was negligible material in this size fraction in any of the traps. The few coarse particles were poisoned with buffered formalin and kept for microscopic examination. The material from the paired cones was combined prior to processing and only one of the four compartments in the folding box trap was used for analyses in this paper. Each sample was wet split using a SANTA

~AR~A~A

~A~ZN

I00

so

I0

20

30

MARCH

I0 APRIL

2976

2O

CD i,1 (D

I

I

I

I

~

I

I

F

I

I-I

I

I

I

I

t

I

I0

I

I

I

I

I

~

I

P

I

I

I

20

I

F

I

I

[

I

I

I

I

SO

MARCH

F

I

I

I

I

I

I

P

I

I

I

I

I

I

I

I

I0 APRIL

197B

7.5

£z) ILl cr CO LU E3

7.0

6.5

I

~

I

I

I

I

I

i

I

I

I

I

I

I

I .I

i

I

I

I

I

Fig.

2. Light meter

scattering placed

data

from

at 330 m.

I

I

the

recording

n

~

I

I

I

I

+- I

30

MARCH

current

I

20

I0

nephelometer

placed

at 254 m (top)

I

I

I

I ~

I

I0 APRIL

297B and

the current

speeds

and

temperature

from

the

413 •56

microscopic evaluation. The remaining wet split was retained by each of the investigators as an archive and was frozen. Each sample was analyzed for A1, Si, Mg, K, Ca, Mn, Fe, Ni, Cu, Zn, and Ba using HF, aqua regia dissolution in teflon bombs and flame atomic absorption analysis [12]. Organic carbon and carbonate carbon were measured by LECO combustion with thermal conductivity detection of CO 2.

•48

~APRIL

.40

5

"~2

20

25

-24 5. Current meter data The Aanderaa current meter attached to the D y m o n d / G a r d n e r mooring at 330 m recorded currents ranging from less than 0.8 cm s-~ (stall speed for the rotor) to 17.5 cm s -~ with a mean speed of 5.2 ± 2.8 cm s - ~ (Fig. 2). For 94% of the time, currents were less than 10 cm s-~, which was the maximum velocity of flume calibrations of sediment traps by Gardner [7]. The current meter at 166 m failed to record data due to an electronic problem. The current at 330 m was dominated by the semidiurnal tide at 12.42 hours, but also had an energy peak at the 21-hour inertial frequency. There were several significant (at the 95% confidence interval) energy peaks between 4 and 9 hours which were probably caused by seiches and internal waves within the basin. The mean flow at 3 3 0 m (Fig. 3) was to the northeast (subject to direction errors reported for Aanderaa current meters [13]). Therefore, at this depth sediment transport came mostly from the direction of Santa Rosa Island rather than from the mainland. At different depths sediment is transported into the basin from the California coast [ 14,15]. During the 48 days of the experiment, the mean temperature at 330 m decreased by 0.2°C from 7.0 to 6.8°C, probably as a result of upwelling. Temperature fluctuations occurred on a regular tidal cycle with excursions from 0.2 to 0.6°C. Based on XBT profiles in the vicinity from the same season in previous years, this temperature differential represents a vertical excursion of the water column of 25-45 m at 0.2°C and 70-85 m at 0.6°C. If this change is real, it represents a vertical current

ARCH

18

-8

Fig. 3. Progressive vector d i a g r a m of currents m e a s u r e d at

330 m, demonstrating that the dominant current direction at that depth was from the southwest. velocity of 0.11-0.36 cm s -~. Since this is only 2-7% of the mean horizontal current, it is not expected to affect the collection characteristics of the traps significantly, although the relative influence may increase during periods of slack currents. All of the traps were also below the depth where high frequency surface waves could cause resuspension in the traps due to orbital motion at the trap bottom.

6. In situ poisoning Attempts were made to poison the lower OSU box trap and the lower cylinder trap using a combination of five antibiotics: chloramphenicol (100 parts), nalidixic acid (1 part), rifamycin sv (1 part), streptomycin sulfate (50 parts), and tetracycline hydrochloride (100 parts). These antibiotics were chosen because they have been found to

414

be very effective inhibitors of marine bacterial activity, even at the low concentrations permitted by their solubility in water (K. Nealson, personal communication). A silica gel matrix which suspended the powdered antibiotic mixture was contained in vials having nylon mesh-covered holes [16]. These vials were located inside the trap. The effectiveness of this poisoning scheme for the two relatively open traps poisoned in this experiment is unclear. Unfortunately, the box trap which was poisoned failed to close before recovery and most of the material was lost. Based on the uptake rates of ~4C-labeled glutamic acid, the material which was retained in the poisoned trap was found to have bacterial activity as high as that in the unpoisoned trap material. However, a qualitative observation was made that the poisoned box and cylinder traps had very little bacterial slime on the floor and walls relative to the unpoisoned traps. The organic carbon contents of the traps are consistent with at least partial effectiveness of the antibiotics for preventing decomposition. The organic carbon contents of the poisoned cylinder and box traps are 7% and 11% greater, respectively, than that of the unpoisoned traps of the same design (Table 2). Unfortunately, the failure in the closure of the poisoned box trap prevents this comparison from being taken too seriously, since some fractionation of the trap material may have taken place in the open box trap during ascent of the mooring.

7. Discussion

7.1. Measured fluxes and trapping efficiency The total mass of the sediment recovered in each trap and the computed flux is listed in Table 1 and is shown in Fig. 4. The fluxes measured by our traps (599 ± 173 g m -2 yr - I ) were all within the range of fluxes measured during previous deployments of the paired cone trap.(689,± 451 g m 2 y r - 1 ) [2,10], although the past measurements were all made in the upper 150 m and bottom 30 m of the basin. There are no previous flux measurements in midwater. The traps spanned more than a 200-m depth range (162-381 m), but it was

PARTICLE CONCENTRATION (Fg/I) 25 I

0

50 I

,oo

r I

"'-

~i

i

400

200 I

// %.. \ ; " ":::" 2

20C

I

T.AP,LUXES

- - - -

OSUBOX

GARDNERO

/~ ~i... SILL / ././~.~. ~ ~.-z.'" DEPTH 200 -~'r~ "~'-':: .... % ~

1600 I

GARDNER

""~';

500

800 I

:

~00

400

I00 I

.

• BERGER

' 1 OSO~TAR 400 600 800 FLUX g/m2/yr)

:

NEPHEL "//////////////////////////E[O

y

i

~ :..,,,...

T "rb'M ~ / / / / / / / / / / / / / / / / / / / / / /

Fig. 4. Nephelometer profiles made during both the deployment leg in March (heavy solid line) and the recovery leg in April (all other lines) using the nephelometer calibration of Biscaye and Eittreim [18]. The measured trap fluxes as a function of depth are shown in the inset.

hoped they would be sufficiently far removed from the particle sources of surface productivity and resuspended bottom sediments to be in the most homogeneous region of vertical flux. Excluding from the discussion the box trap which did not close, these traps, which differ greatly in design, record fluxes which agree within approximately a factor of two. We view these results as encouraging regarding the intercomparison and future application of traps in addressing questions about the flux of material falling through the water column. The problem of establishing the absolute efficiency of the traps used in this study, however, is difficult. A comparison between our 48-day experiment and the long-term accumulation rate of sediment is fraught with uncertainties, but it is the only independent comparison available. Bruland et al. [10] have measured sediment accumulation rates in the Santa Barbara Basin by 2~°pb dating of box cores and obtained an average rate of 920 g m 2 yr -I for the past 25 years. Based on the production of 2~°pb from the U parent in the overlying water column and the atmospheric flux in this region, however, there is an eight-fold excess of 2~°pb in the basin sediments [17]. The excess is most likely associated with sediment advected in from the basin margins and from rivers

415

[14,15]. From the work of Drake [15] it appears that much of the transport occurs near the bottom, below the level of the traps. However, his profiles from previous years, as well as our own profiles taken on the deployment and recovery cruises (Fig. 4) with the LDGO-Thorndike nephelometer [18], show mid-water maxima where particle concentrations change in single profiles by more than a factor of two. These profiles suggest that horizontal advection may be introducing particles to the basin at mid depths as well. In addition, the LDGO-Thorndike recording nephelometer at 254 m (Fig. 2) indicated periods of several days where the particle concentration increased by a factor of two, with some peaks 3.5 X the minimum concentrations. In addition to problems of advection, there are other reasons why the flux of material to the traps would differ from the accumulation rate on the seafloor. Solution and oxidation of material falling on the seafloor could return a portion of the settled material to the seawater. Because most of the sediment being deposited in the Santa Barbara Basin is relatively refractory continental detritus, this effect may be small, but some organic carbon, opal and Mn solubilized by the low oxygen environment is undoubtedly lost to the water col-

umn. Studies of sediment transport to the Santa Barbara Basin indicate that most of the detritus is brought at least to the adjacent shelf during the high runoff winter season [14]. The 1978 winter was clearly one of the most active storm periods in Southern California's recent history. Consequently, our trap experiment may have been conducted during a period of unusually high sediment accumulation in the basin. Since the flux measured by all traps was less than the 25-year average accumulation flux to the sediments, either all four traps have trapping efficiencies less than 100% or a significant but unknown portion of the sediment accumulating in the basin is introduced by bottom transport.

7.2. Composition of trapped material Except for Ca the major dement composition of the material from different traps varies from the mean by only 4% (Table 2). The Ca content of the lower cylinder and OSU box trap is approximately 25% lower than that of the other traps. Since the lower OSU box trap did not close prior to recovery, the material remaining in the trap may have been fractionated which could account for the low Ca. The high Fe, Cu and Zn concentra.

i

TABLE 2 Major element composition of the trapped material Element a

A1 Si Mg K Ca Fe Mn Ni Cu Zn Ba Core Cca¢o ~

Soutar cones

Berger box

Gardner upper cylinder

7.23 24.77 1.70 1.73 3.91 4.43

7.04 24.17 1.75 1.76 3.86 4.62

7.35 24.43 1.79 1.78 3.63 4.78

360 59 47 142 596 3.71 0.89

460 55 50 171 635 3.55 0.96

530 58 46 160 593 3.49 0.88

Gardner lower cylinder b 7.29 23.47 1.89 1.83 3.09 4.77 470 57 36 164 697 3.74 0.55

Upper OSU box

Lower OSU box b

Surface sediment c

7.27 24.07 1.76 1.85 3.86 4.76

7.29 23.97 1.80 1.92 3.03 4.95

6.74± 24.45-+ 1.93 ± 1.98± 4.50-+ 3.6 -+

510 62 43 160 59,4 3.18 0.94

520 61 61 355 702

330 52 37 119 600

0.37 1.0 0.52 0.51 0.49 0.185

-+ 10 -+ 2 ± 2

3.54 0.82

a AI through Fe, Corg and Ccaco 3 given as weight percent; Mn through Ba given as ppm by weight. b Poisoned with antibiotics, c Average of 0-1 cm for 4 gravity cores taken in the central basin during trap deployment.

416

tions of the material in this trap further support the possibility of fractionation. Consequently, this trap will not be considered in the following discussion. It should be noted, however, that both the lower cylinder trap and the lower OSU box trap were poisoned in situ and it is possible that the Ca variations noted in these samples are related in an unknown way to the poisoning. In general, the minor element compositions of the different traps differ by less than 10% (Table 2). Manganese concentrations of material collected in the cone trap, however, are 37% lower than the average of the other traps. The 360 ppm of manganese measured in the cone trap sample is similar to the 330 ppm value measured in surface sediments. We believe that Mn is lost from the particulates which reach the bottom as well as those which fall into the cone trap because a low oxygen environment within the cone trap and on the bottom reduces MnO 2 to soluble Mn 2+ . The cone trap was the only design in which the sample recovered smelled of HzS. This experiment suggests that much of the Mn carried by particulates is in an easily reduced and sohibilized form and can be returned to the water column very quickly upon encountering a reducing environment. These results also reinforce the need to poison traps in situ in order to study the transport and diagenetic processes which affect transition metals. If the traps are poisoned, the bacterial activity which decomposes organic matter and makes the trap interior strongly reducing would be slowed greatly. The need for poisoning in this case is most important for cone traps and other traps in which the sample is concentrated into a small volume with o osu

Box

a CONES

•

BERGER

o GARDNER-UPPER

0/02oL t.5

I.O

m

0.6

w

t

t

AI

Si

t

Co

t

Mn

L

i

Fe

Ni

i

Cu

l

Zn

j

Ba

Fig. 5. Composition of trap materials normalized to the average surface sediment composition.

little exchange with seawater. The generally close similarity in composition of the material collected by all four designs is demonstrated by Fig. 5, which compares trap compositions to surface sediment compositions. This similarity suggests that even with a factor of two variations in trapping efficiency there is no fractionation of the material by the different trap designs.

7. 3. Particle trapping models The four trap designs used in this expeirment are based on two conceptual models of how sediment traps collect particles. The Soutar and Berger traps are designed so that a minimum disturbance of current flow occurs over the mouth of the trap, a model developed by Soutar. The Gardner and OSU box traps are based on Gardner's model of optimal fluid exchange. These two models and the design features involved are described below. A trap in a current generates turbulent eddies around it. Soutar chose design features to reduce the eddies over the opening of the trap so that particles could settle in an undisturbed fashion. From laboratory flume studies, he found that small diameter, relatively deep (honeycomb) baffles very effectively damp the large eddies which are formed by the disruption of current flow over the trap. To further decrease the turbulent flow over the trap, Soutar uses a thin lip which extends approximately 10 cm beyond the collecting area of the baffles. This lip, which also holds the honeycomb baffles, deflects approaching currents downward, and its sharp leading edge helps maintain a laminar flow over the baffles' surface. Thus, Soutar believes this trap ideally has only diffusional exchange between the water in the trap and the outside. Some, though not all, particles upon reaching the trap surface drop into the trap and are collected. The suitcase trap, which uses baffles twice as deep as the cone trap and an even larger rim around the baffles, was designed around the same conceptual model. The conceptual model for particle trapping in the open cylinder and OSU box trap with its larger baffles is one of fluid exchange. In laboratory flume studies involving traps of many shapes, Gardner [7] observed with the help of dyes and

417

settling particles that eddies which are generated at the leading edge of the trap plunge into the trap at the downstream edge, circulate through the trap and emerge at the upstream edge. During the passage of eddies through the trap, particles settle and are effectively trapped when they encounter the most stagnant parts of the trap. Gardner suggests the residence time of the water flowing through the trap effectively determines the trapping efficiency, and the most important factor for residence time is trap geometry. This residence time for cylinders can be increased by increasing the height to width ratio, and empirically Gardner found a height to width ratio of cylinders of between 2" 1 and 3:1 produced approximately 100% trapping efficiency in currents less than 10 cm s -l. Very deep traps and those with small openings relative to their internal volume had trapping efficiencies which could be much greater than 100% (e.g., an Erlenmeyer flask under the same circumstances has a trapping efficiency of greater than 500%). It is noteworthy that in spite of these very different conceptual models for particle trapping, there is only a factor of two difference in trapping efficiencies among all the designs. On the basis of Gardner's flume studies relating a trap's collection efficiency to the residence time of water flowing through the trap and Soutar's belief that the honeycomb baffle and sharp leading edge cause a very slow exchange rate, the cone trap and suitcase box trap might be expected to overtrap relative to traps with more open baffles or without baffles or a lip. Our data indicate the cone trap is 92-150% efficient relative to the cylinder trap and 190% efficient relative to the OSU box trap. The suitcase trap, in spite of its 10:1 honeycomb baffles, is 94% efficient relative to the cone trap. However, the suitcase trap has a smaller fraction of its volume below the baffles than the cone trap, which may decrease the water residence time in the suitcase trap. The observation that the collecting efficiencies of different trap designs cannot be related in any clear fashion to trap exchange rates may indicate inadequacies in our conceptual models. Alternatively, it is likely that trapping efficiency for particles with rapid settling velocities is not so

design-sensitive as it is for smaller particles. Dunbar and Berger [19] have observed that 60% and possibly as much as 90% of the material trapped by the suitcase trap in this experiment was fecal pellet aggregate. Their measured fecal pellet settling velocities ranged from 71 m day -I to 1694m day -l, with 89% of the estimated fecal pellet flux being in the form of large tabular forms, thought to be produced by salps. Dunbar and Berger compute an average fecal pellet settling velocity of 710 m d a y - i. Particles of this size have relatively steep fall trajectories in the slow currents measured during this experiment and upon entering a trap are not likely to be advected out. It is possible that the lower efficiency of the OSU box trap resulted from decomposition and mechanical breakdown of fecal pellets in this trap. Because of the relatively open design and probable higher exchange rates in this trap, resuspension of the fine particles released from decomposed fecal pellets could be lost during the deployment period. On the basis of the fluid exchange model of particle trapping, traps with long residence times might be expected to trap more fine particles than traps with rapid internal mixing as seen with small trap models [7]. Preferential trapping of finer particles could strongly fractionate the composition of the collected material unless all particles have similar compositions. This could happen if the large particles as fecal pellets are aggregates of fine particles. The observed fecal pellets [19] were dominated by finer grained clays and quartz, attesting to the ability of zooplankton to filter terrigenous material and aggregate it into rapidly settling particles. Therefore, it is not surprising that all traps collected material of essentially the same composition (Table 2, Fig. 5). In conclusion, this intercomparison study of four distinct trap designs provides encouragement that various trap designs do not greatly differ in their trapping efficiency and will collect particles of similar compositions. These observations may be linked to the dominance of large particle flux through the water column. If the results can be generalizecl to other areas of the ocean, it emphasizes the importance of trap designs which favor preservation, ease of sample handling, mechanical reliability, and lack of contamination.

418

Acknowledgements We wish to thank Charlotte Muratli who did the chemical analyses and Roxanne Roderick who prepared the manuscript. This research was funded by the National Science Foundation as part of MANOP.

References 1 P.H. Wiebe, S.H. Boyd and C.J. Winget, Particulate matter sinking to the deep-sea floor at 2000 m in the Tongue of the Ocean, Bahamas, with a description of a new sediment trap, J. Mar. Res. 34 (1976) 341-354. 2 A. Soutar, S.A. Kling, P.A. Crill, E. Duffin and K.W. Bruland, Monitoring the marine environment through sedimentation, Nature 266 (1977) 136- 139. 3 D.W. Spencer, P.G. Brewer, A. Fleer, S. Honjo, S. Krishnaswami and Y. Nozaki, Chemical fluxes from a sediment trap experiment in the deep Sargasso Sea, J. Mar. Res. 36 (1978) 493-523. 4 R. Cobler and J. Dymond, Sediment trap experiment on th e Galapagos spreading center, equatorial Pacific, Science 209 (1980) 801-803. 5 G.A. Knauer, J.H. Martin and K. Bruland, Fluxes of particulate carbon, nitrogen, and phosphorus in the upper water column of the northeast Pacific, Deep-Sea Res. 26A (1979) 97-108. 6 G.T. Rowe and W.D. Gardner, Sedimentation rates in the slope water of the northwest Atlantic Ocean measured directly with sediment traps, J. Mar. Res. 37 (1979) 581-600. 7 W.D. Gardner, Sediment trap dynamics and calibration: a laboratory evaluation, J. Mar. Res. 38 (1980) 17-39.

8 B.T. Hargrave and N.M. Burns, Assessment of sediment trap collection efficiency, Limnol. Oceanogr. 24 (1979). 9 M. Koide, A. Soutar and E.D. Goldberg, Marine geochronology with 2~°pb, Earth Planet. Sci. Lett. 14 (1972) 442446. 10 K.W. Bruland, R.P. Franks, W.M. Landing and A. Soutar, Southern California inner basin sediment trap calibration. Earth Planet. Sci. Lett. 53 (1981) 400-408. 11 D.A. McManus, L.G. Hanson and J.V. Baker, A sample splitter for moist, saturated, or suspended sediment, J. Sediment. Petrol. 36 (1966) 632-635. 12 J. Dymond, J.B. Corliss, G.R. Heath, C.W. Field, E.J. Dasch and H.H. Veeh, Origin of metalliferous sediments from the Pacific Ocean, Geol. Soc. Am. Bull. 84 (1973) 3355-3371. 13 R.M. Hendry and A.S. Hartling, A pressure-induced direction error in nickel-coated Aanderaa current meters, DeepSea Res. 26 (1979) 327-355. 14 D.E. Drake, R.L. Kolpack and P.J. Fleischer, Sediment transport on the Santa Barbara-Oxnard shelf, Santa Barbara Channel, California in: Shelf Sediment Transport, Swift, Duane and Pilkey, eds. (Dowden, Hutchinson and Ross, Stroudsburg, Pa., 1972). 15 D.E. Drake, Distribution of suspended matter, Santa Barbara Channel, California, Dissertation, University of Southern California (1972) 349 pp. (unpublished). 16 K. Fischer and R. Cobler, Antibiotic poisoning of sediment traps, Fall Meet., Am. Geophys. Union (1979). 17 W.S. Moore, K.W. Bruland and J. Michel, Fluxes of uranium and thorium series isotopes in the Santa Barbara Basin, Earth Planet. Sci. Lett. 53 (1981) 391-399. 18 P.E. Biscaye and S.L. Eittreim, Suspended particulate loads and transports in the nepheloid layer of the abyssal Atlantic Ocean, Mar. Geol. 23 (1977) 155-172. 19 R.B. Dunbar and W.H. Berger, Fecal pellet flux to modern bottom sediment of Santa Barbara Basin (California) based on sediment trapping, Geol. Soc. Am. Bull. (in press).