Deposition rates, mixing intensity and organic content in two contrasting submarine canyons

Page 1

Available online at www.sciencedirect.com

Progress in Oceanography 76 (2008) 192–215

www.elsevier.com/locate/pocean

Progress inOceanography

Deposition rates, mixing intensity and organic content intwo contrasting submarine canyons

R. Garcıa a,*, D. van Oevelen b, K. Soetaert b, L. Thomsen a,H.C. De Stigter c, E. Epping c

a Jacobs University Bremen, Campus Ring 1, 28759 Bremen, Germanyb The Netherlands Institute of Ecology (NIOO-KNAW), Centre for Estuarine and Marine Ecology,

Korringaweg 7, 4401 NT Yerseke, The Netherlandsc Royal Netherlands Institute for Sea Research (NIOZ), 1790 AB Den Burg, Texel, The Netherlands

Available online 17 January 2008

Abstract

The hydrographically different conditions characterising the Western Iberian Margin (NE Atlantic) and the Gulf ofLions (Mediterranean) may play an important role in determining the biogeochemical characteristics of the sediments.To investigate this, we compared the Nazare and Cap de Creus canyons, and their respective adjacent open slopes in termsof the organic carbon (Corg) contents, chlorophyll-a (chl-a) concentrations, C:N and chl-a:phaeopigment ratios, and also interms of modelled mixing intensities, chl-a and 210Pb deposition and background concentrations in sediments. Chloro-phyll-a and 210Pb profiles were fitted simultaneously with a reactive transport model to estimate mixing intensity, deposi-tion and background concentrations. Further, to account for the possibility that the decay of chl-a may be lower in thedeep sea than in shallow areas, we estimated the model parameters with two models. In one approach (model 1), the tem-perature dependent decay rate of chl-a as given by Sun et al. [Sun, M.Y., Lee, C., Aller, R.C. (1993) Laboratory Studies ofOxic and Anoxic Degradation of chlorophyll-a in Long-Island sound sediments. Geochimica et Cosmochimica Acta, 57,147–157] for estuaries was used. In the other approach (model 2), an extra parameter was estimated to derive the chloro-phyll-a degradation rate. An F-test, taking into account the different number of parameters in the models, was used tosingle out the model that significantly fitted the data best. In most cases, the model parameters were best-explained withmodel 1, indicating the empirical relationship by Sun et al. (1993) is a valid means to estimate the chlorophyll-a degrada-tion rate in deep sea sediments. To assess the robustness with which the model parameters were estimated we provide a firstapplication of Bayesian analysis in the modelling of tracers in sediments. Bayesian analysis allows calculating the mean andstandard deviation for each model parameter and correlations among parameters. The model parameters for stations forwhich 210Pb and chlorophyll-a profiles were available were robustly fitted as evidenced by an average coefficient of vari-ation of 0.22. Corg contents, chl-a concentrations, chl-a:phaeo ratios, mixing intensities, depositions and background con-centrations of chl-a and 210Pb indicated that the Cap de Creus canyon and adjacent slope were less active in terms oforganic matter accumulation and burial than the Nazare canyon and respective open slope.� 2008 Elsevier Ltd. All rights reserved.

Keywords: Mixing intensity; Deposition; Reactive transport model; chl-a; 210Pb; Organic contents; Submarine canyons

0079-6611/$ - see front matter � 2008 Elsevier Ltd. All rights reserved.

doi:10.1016/j.pocean.2008.01.001

* Corresponding author. Tel.: +49 421 200 3507; fax: +49 421 200 3229.E-mail address: [email protected] (R. Garcıa).

Page 2

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 193

1. Introduction

Submarine canyons are major transport systems that transfer sedimentary organic matter from the shelf tothe deep ocean (Durrieu de Madron, 1994; Mullenbach and Nittrouer, 2000; Canals et al., 2006; Palanqueset al., 2006a), and accumulate high amounts of sediments and organic matter (Carson et al., 1986; Monacoet al., 1990; Epping et al., 2002; Van Weering et al., 2002). Downward transport and the redistribution of sed-iments in canyons are controlled by hydrodynamic processes, such as down and along slope bottom currents,internal waves and gravity flows interacting with the bottom topography. Thus, hydrographically differentconditions in contrasting continental margins will affect the transport, deposition and accumulation of organicmatter in submarine canyons and surrounding slopes. For example, submarine canyons dominated by internaltides will focus organic material within the canyon walls, while canyons dominated by down canyon circula-tion will mostly transport organic matter to greater water depths. These processes may also affect the move-ment and feeding activity of benthic organisms, which in turn may affect the concentration and lability of theorganic matter in the sediment.

Benthic organisms enhance the transport of the organic particles from the sediment–water interface into thesedimentary column in a process called bioturbation (Meysman et al., 2006). Bioturbation is importantbecause it affects the physical structure of the sediment (Aller, 1982, 1994; Wheatcroft et al., 1990) and exertsa direct control on carbon mineralisation and burial efficiency (Sun et al., 1993; Green et al., 2002). Organicmatter deposition and bioturbation rates (Db) have been determined from modelling depth profiles of singletracers, such as 210Pb, 234Th and chl-a (chlorophyll-a), which are strongly associated with organic particles (i.e.Soetaert et al., 1996a; Boon and Duineveld, 1998; Turnewitsch et al., 2000; Schmidt et al., 2001; Green et al.,2002). The derived Db values however depend on the used tracer. Short-lived tracers like 234Th and chl-a tendto yield higher Db values than longer-lived tracers like 210Pb (Smith et al., 1993, 1997;Turnewitsch et al., 2000;Reed et al., 2006). Smith et al. (1993) hypothesised that short-lived tracers are associated with labile organicmatter, while long-lived tracer with refractory organic matter. Organisms tend to preferably handle fresh overrefractory material, hence short-lived tracers would be preferentially ingested and mixed at higher rates thanlong-lived tracers; a process called age-dependent mixing. Consequently, a single Db value cannot be assignedto a sediment unit and biogeochemical modelling becomes difficult. Reed et al. (2006) challenged this hypoth-esis indicating that long-lived tracers estimate the true mixing intensity of the sediment, while the Db values ofthe short-lived tracers are a modelling artefact.

To estimate a single mixing intensity from chl-a (short-lived tracer) and 210Pb (long-lived tracer) sedimentdepth profiles, and the respective deposition rates, a simple single tracer model has been used. We use a dualtracer approach, in which depth profiles of chl-a and 210Pb were fitted simultaneously and we apply Bayesiananalyses to quantify the model parameters and their associated uncertainties. Subsequently, we investigate theinfluence of hydrographically different conditions on the magnitudes of the model parameters, concentrationsof chl-a, chl-a:phaeo ratios (chlorophyll to phaeopigment ratio), organic carbon contents, and C:N ratios (car-bon to nitrogen ratio) in the contrasting Nazare canyon (Western Iberian Margin) and Cap de Creus canyon(Gulf of Lions, Western Mediterranean), and respective adjacent slopes. Part of the chl-a data and chl-a:phaeoratios of the Nazare canyon and adjacent open slope were discussed in Garcıa and Thomsen (in press), wherethe spatial distribution of bioavailable organic matter in surface sediments (0–0.5 cm) was investigated. Here,we present and compare vertical profiles for two submarine canyons and their respective adjacent slopes fromtwo hydrographically distinct areas.

Marine environments are heterogeneous in terms of organic matter composition, reactivity, oxic/anoxicconditions, benthic fauna composition and activity. The quantity and quality of the particulate organic matterdeposited on the sea floor depend on the origin and the degree of diagenetic transformation while residing inthe benthic boundary layer (Cowie et al., 1992; Thomsen et al., 2002). In general, at larger water depthsorganic matter has had more time to decay before deposition compared to shallower stations. Thus, wehypothesised a priori that the decay rate of chl-a decreases with water depth and demonstrates site-specific(canyon-slope and interregional) differences. To account for the anticipated differences in decay rate, we fittedtwo models to the data. In one model the chlorophyll decay rate (k) was a priori derived from the empiricalrelationship ln k ¼ 18:3� 6160 � T�1 (Sun et al., 1991, 1993), which describes the oxic degradation of chl-a as afunction of temperature in estuarine sediments (T, degrees Kelvin). In the second model, the chl-a decay rate

Page 3

194 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

was modelled with an additional free parameter, which was also fitted against the data. An F-test, taking intoaccount the different number of model parameters in the models, was used to single out the model that sig-nificantly fitted the data best. Finally, we provide a novel application of Bayesian analyses in tracer modellingby estimating the uncertainty of fitted model parameters and correlations among them.

2. Study areas

The Gulf of Lions (Western Mediterranean) is a micro-tidal and river dominated continental margin char-acterized by a broad shelf (�70 km) and incised by several submarine canyons (Palanques et al., 2006b). TheCap de Creus canyon is located in the most western part of the Gulf of Lions, where the shelf narrows andwhere the wind driven coastal circulation and the Liguro-Provenc�al or Northern current converge (Millot,1990) (Fig. 1). The water column along this margin is not strongly stratified with yearly mean temperaturesof �13 �C. The water column is characterized by three water masses (Font, 1987; Millot, 1987, 1990): thewater of Atlantic origin on the surface, the Levantine Intermediate Water (LIW) and the Deep Western Med-iterranean Water (DWMW). The canyon represents an important area of sediment transport for the entireGulf of Lions because it funnels a larger amount of suspended sediments towards deeper water than othercanyons in the area (Palanques et al., 2006b). Primary production in the Gulf of Lions ranges between 78and 142 g C m�2 y�1 (Lefevre et al., 1997).

In contrast, the Western Iberian Margin is a tidal margin characterized by a narrower shelf (�50 km) and asteep irregular slope incised by deep gullies and canyons. The Nazare canyon is located in the middle part ofthis margin (Fig. 1), is the largest canyon in the area, and intersects the entire continental shelf (Vanney andMougenot, 1981). The water column along the Western Iberian Margin is stratified, grading from relativelywarm (14–18 �C) and saline (35.4–35.8) water at the surface (North Atlantic Central Water) to cold (2 �C)and less saline (34.8) water at 5000 m depth (North Atlantic Deep Water). A distinct salinity maximum ofup to 36.2 found between 600 and 1500 m is associated with a vein of Mediterranean water flowing alongthe margin (Maze et al., 1997; Fiuza et al., 1998). The upper and middle canyon captures suspended partic-ulate matter from the adjacent shelf and is affected by internal tide circulation of water with strong bottomcurrent speeds (De Stigter et al., 2007). Intermittent sediment gravity flows have been registered in the canyon,which, coupled to the internal tide circulation produce a net downward canyon transport of particulate mate-rial (De Stigter et al., 2007). Surface water productivities at the Western Iberian Margin vary between 154 and556 g C m�2 y�1 (Alvarez-Salgado et al., 2003). These higher primary productivities when compared with theGulf of Lions are the result of the prevailing summer upwelling regime in the Western Iberian Margin (Vit-orino et al., 2002).

3. Materials and methods

3.1. Sampling

Sediment cores were collected along a depth gradient inside the canyons and their adjacent slopes duringcruises RV Pelagia 64PE225 and 64PE236 of Royal NIOZ in May 2004 and May 2005 respectively. TheNazare canyon was sampled at 6 stations inside the canyon (St 41, St 26, St 34, St 24, St 22 and St 13)and at three stations on the adjacent open slope (St 39, St 27 and St 25). The Cap de Creus canyon was sam-pled at four stations (St 52, St 50, St 47 and St 48) and the adjacent open slope at three stations (St 54, St 51and St 49) (Table 1, Fig. 1).

Sediment samples were taken with the MUC 8+4 multiple-corer developed by Oktopus GmbH. The MUCis equipped with eight small (6 cm in i.d.) and four large (10 cm in i.d.) core tubes of 61 cm length. At eachstation, three small cores were used for chlorophyll-a and phaeopigment analysis, four small cores were usedfor organic carbon and nitrogen analysis, and one large core was used for 210Pb analysis. Cores were stored atin-situ temperature in temperature-controlled room and processed within 3 hours. The cores for chlorophyll-aand phaeopigment analysis were sub-sectioned at 0.5 cm intervals down to 1 cm, at 1 cm intervals from 1down to 5 cm, at 2 cm intervals from 5 down to 9 cm, and at 3 cm intervals from 9 down to 15 cm. The cores

Page 4

Fig. 1. Maps of the Gulf of Lions and Western Iberian Margin showing the stations sampled in a) the Cap de Creus canyon (St52, St50,St47 and St48) and adjacent open slope (St 54, St51 and St49), and in b) the Nazare canyon (St41, St26, St13, St34, St24 and St 22) andadjacent open slope (St39, St27 and St25).

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 195

for organic carbon and nitrogen were sub-sectioned at 0.25 cm intervals down to 1 cm, 0.5 cm intervals from 1down to 3 cm, 1 cm intervals from 3 down to 7 cm and at 2 cm intervals the from 7 cm until bottom of thecore. These samples were stored in plastic bags at �20 �C until further analysis. The cores for 210Pb analysis

Page 5

Table 1Accessory parameters used in the model to estimate bioturbation rates and fluxes through the sediment water interface

Stations Latitude (N) Longitude (W) Temperature (oC) /o (V/V) /1 (V/V) x/ (cm) x1 (cm d�1)

Cruise 64PE225 May 2004

Gulf of LionsOpen slope

St 54 42o13.90 3o35.50 13.2 0.757 0.618 3.558 2.5E�04St 51 42o18.00 3o54.00 13.2 0.879 0.640 2.275 4.9E�05St 49 42o07.50 4o03.50 13.1 0.742 0.555 2.890 3.5E�05

Cap de Creus canyon

St 52 42o23.30 3o19.80 13.2 – – – –St 50 42o15.90 3o40.00 13.2 0.843 0.620 2.159 1.7E�04St 47 42o10.40 4o04.20 13.1 0.741 0.625 1.229 1.5E�04St 48 42o14.90 4o20.70 13.2 0.790 0.579 0.639 8.4E�05

Western Iberian MarginOpen slopeSt 39 39o39.90 9o35.90 11.9 0.583 0.461 0.062 1.2E�04a

St 27 39o33.90 9o40.90 9.0 0.819 0.655 6.751 7.9E�05St 25 39o46.50 10o59.90 2.0 0.949 0.733 0.589 4.1E�06b

Nazare canyon

St 41 39o34.80 9o09.90 11.8 0.864 0.762 5.074 1.3E�03St 26 39o35.90 9o23.90 8.6 0.829 0.362 5.074 2.3E�03St 34 39o30.00 9o45.90 4.3 0.791 0.408 1.697 –St 24 39o48.00 10o37.90 2.0 0.898 0.784 1.610 –St 22 39o53.90 11o10.00 1.8 0.834 0.681 6.180 5.2E�06

Cruise 64PE236 May 2005

Nazare canyon

St 13 39o35.80 9o24.30 9.3 0.710 0.783 1.007 6.2E�04

/o = porosity at the sediment–water interface; /1 = porosity at infinite depth; x/ = attenuation porosity coefficient; x1 = burial velocity.a x1 from station 64PE218-57 at 151 m, a shallower station sampled previous year.b x1 from station 64PE218-08 at 4806 m, same station sampled the previous year.

196 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

were stored in upright position at +5 �C until analysis, and sub-sectioned in the laboratory at 0.5 cm intervaldown to 5 cm, and at 1 cm interval from 5 down to 30 cm.

3.2. Chlorophyll-a and phaeopigment analysis

Pigments were extracted following the method of Yentsch and Menzel (1963). Sediment samples werefreeze-dried and homogenised in a mortar and pigment extracts therefore represent the total extractable pig-ment pool of free and bound fractions (Sun et al., 1991). Pigments were extracted (�1 g dried sediment) in10 ml of a 90% acetone solution, stored in the dark at �6 �C for �24 h (shaken twice) and centrifuged at4500 rpm for 20 min. The supernatant was measured in a TD-700 Turner fluorometer following Shumanand Lorenzen (1975).

3.3. Organic carbon and nitrogen analysis

Sediment organic carbon and nitrogen were measured using a ThermoFinnigan flash element analyser fol-lowing the procedures described by Lohse et al. (2000). Sediment was freeze-dried and homogenized in a mor-tar mill. Subsequently, 10–25 mg of sediment was weighed into silver-cups and decarbonated with 2 N HCl.The samples were dried at 60 �C for 20 to 60 min and the acid addition was repeated until termination of bub-bling. Silver cups were transferred into a second silver cup, in which they were pinched, closed and compactedto spherical balls. These samples were analysed for organic carbon and nitrogen. A second set of samples was

Page 6

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 197

treated the same way but was not acidified. When analysed, these unacidified samples estimate identical con-tent of nitrogen to the samples in the previous run, plus the organic and inorganic fractions of carbon.

3.4. 210Pb analysis

The activity of 210Pb was measured indirectly from its granddaughter 210Po with half-life 138.4 days, asdescribed in Boer et al. (2006). 100 to 500 mg of freeze-dried and ground sediment samples were spiked with209Po and leached with concentrated HCl. After leaching, Po-isotopes were deposited by suspending a silverdisk in the solution, which was heated at 80 �C for 4 h. Silver plates were left in solution overnight at roomtemperature, and the activity of 210Pb was measured via a-spectrometry with Canberra Passivated ImplantedPlanar Silicon (PIPS) detectors.

3.5. Sediment porosity

Sediment porosity was determined from the loss of weight of known volumes of sediment after freeze-dry-ing. Depth profiles of porosity for the different stations were fitted with the equation:

/ðxÞ ¼ ½/1 þ ð/0 � /1Þ� � exp � xx/

� �ð1Þ

in which /1 is the porosity at infinite depth (v/v), /0 is the porosity at the sediment–water interface (v/v), x isdepth in sediment (cm) and x/ is the depth attenuation coefficient for porosity (cm). When porosity profileswere truncated at certain sediment depth, porosity profiles were fit with the dual-exponential equation:

if x < xbreak ; /ðxÞ ¼ ½/11 þ ð/0 � /11Þ� � exp

�� x

x/1

�;

if x > xbreak ; /ðxÞ ¼ ½/12 þ ð/11 � /12Þ� � exp � x� xbreak

x/2

! ð2Þ

in which x is sediment depth (cm), xbreak is sediment depth (cm) at the truncation of the porosity profile, /o isthe porosity at the sediment–water interface (v/v), /1

1 and /21 are the porosity at infinite depth (v/v) before

and after xbreak respectively, and x1/ and x2

/ are the depth attenuation coefficients for porosity (cm) before andafter xbreak, respectively. The estimated values for /1, /0 and x/ obtained from fitting the observed porositydata to Eqs. (1) and (2) are used in the reactive transport model (Table 1). The best estimates for these param-eters were obtained by minimising the sum of the residuals, which indicates the distance between the observedporosity and modelled porosity at a certain depth. The sum of the residuals was defined as

Xx¼0

x¼1ð/mod

x � /obsx Þ

2 ð3Þ

in which x is sediment depth, /modx is the modelled porosity values at depth x, and /obs

x is the observed porosityvalues at depth x. The average minimized sum of residuals between modelled and observed porosity data of allstations was 0.6% indicating good empirical fits of the porosity data.

3.6. Tracer model formulation

Assuming steady state conditions and first-order degradation, bioturbation can be obtained from the reac-tive transport equation (Boudreau, 1997; Meysman et al., 2005), which describes the concentration or activitychange of a particulate tracer due to advection, mixing and reaction:

oð1� /ÞCi

ot¼ 0 ¼ o

oxð1� /ÞDbðxÞ

oCi

ox

� �� �� x1 ð1� /Þ oCi

ox

� �� kið1� /ÞCi ð4Þ

Page 7

198 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

where Ci is the concentration or activity of the tracer i per cm3 of solid sediment (chl-a or 210Pb respectively inthis study), t is time (d), x is depth in the sediment (cm), / is the porosity (v/v), Db(x) is biodiffusion coefficient(cm2 d�1), x1 is the burial velocity (cm d�1), and k is the first-order decay constant of the tracer (d�1).

(1 � /)Ci is the excess concentration or activity of the tracer i per cm3 of bulk sediment. Excess tracer isdefined as the concentration or activity of a tracer in excess of a background concentration or activity. There-fore, for comparison with measured chl-a concentrations and 210Pb activities, the background (refractory)concentrations or activities are added to the excess.

The biodiffusion coefficient (Db) was assumed to be constant in an upper well-mixed layer of 10 cm thick-ness, below which it declines exponentially at a rate of 1 cm thus vanishing at a depth of �16 cm.

At the sediment–water interface (x = 0), an advective input flux Ji is imposed:

J i ¼ �ð1� /0ÞDbð0ÞoCi

ox

����x¼0

þ x1 ð1� /1Þ Ci x ¼ 0j ð5Þ

Whilst at infinite depth, it is assumed that the excess tracer concentration is zero:

Cjx¼1 ¼ 0

To obtain the biodiffusion coefficient (Db(0)) and deposition rates across the sediment water interface, chl-aconcentrations (lg cm�3) and 210Pb activities (dpm cm�3) over sediment depth were fitted simultaneously.The decay rate was fixed for 210Pb (0.031 yr�1) and either imposed (model 1) or estimated (model 2) for chlo-rophyll-a.

The values for porosity (/x) and sediment burial velocity (x1) are given in Table 1. For model 1, k for chl-awas estimated with an empirical temperature-dependent equation (Sun et al., 1993):

ln k ¼ K� 6160 � T�1 ð6Þ

in which T is temperature (degrees Kelvin) and K = is a parameter that gives an indication about the degra-dation history of the chl-a arriving to a given location. Model 1 has the empirically derived value of 18.3 for K(see Sun et al., 1993), which was obtained from estuarine sediment incubations. Model 2 has K as a freeparameter because a priori there is no reason to suppose that K is universally constant. We assume that a devi-ation in K represents a deviation in the lability of chlorophyll-a as compared to the lability of the algal mate-rial used by Sun et al. (1993). These two approaches will be referred to as model 1 and model 2, respectively.

3.7. Model calibration

Overall, the following parameters were estimated by fitting the model(s) to data: the two deposition ratesJchla, JPb210, and background concentrations/activities of chl-a and 210Pb, and the biodiffusion coefficient(Db(0)). Model 2 has the decay rate of chlorophyll (kchl-a) as additional parameter. We only report on modelfits for those stations that could be fitted with diffusive mixing conditions.

As neither the porosity nor the biodiffusion coefficient was constant with sediment depth, the diageneticequations were approximated by finite differencing, using the scheme as given in Soetaert et al. (1996b).The model was implemented on a personal computer, in the simulation environment FEMME (Soetaertet al., 2002). This environment takes care of solving the model and performs the calibration and Bayesiananalysis. The steady-state profiles were estimated by solving the system of equations using the Newton–Raph-son method (Soetaert et al., 2002). As the resulting equations are linear, convergence is achieved in one New-ton–Raphson step. Model calibration was performed using the Levenberg–Marquardt algorithm. Duringcalibration, a model cost function, which quantifies the difference between data and model output, isminimised.

The cost function was defined as

Min ¼Xvar

i

Xobs

j

ðxmodi;j � xobs

i;j Þ2

r2i;j

!ð7Þ

Page 8

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 199

where xmodi;j is the modelled value, xobs

i;j is the corresponding mean observed value and rI,j is the observed stan-dard deviation. This weighting of the residuals by the standard deviation serves two purposes. First of all, itensures that the model cost function is a-dimensionalised, (and therefore residuals consisting of chl-a and210Pb can be meaningfully merged), and secondly, it gives less weight to less certain data. However, this pro-cedure requires robust estimates of standard deviations. To this end, standardized relative and absolute errorsfor chl-a observations were obtained by plotting the standard deviation of the chl-a data against the respectivemean values. The relative error and absolute error were taken as the slope and the intersection of the linear fitof the data (Fig. 2a). When no significant linear fit was obtained the relative error was set to zero and the abso-lute error was calculated as the mean of the standard deviations (Fig. 2b). This is a more robust fitting pro-cedure than using standard deviations based directly from the data, as the latter may skew the model fitconsiderably to fitting those values where standard deviations were (perhaps erroneously or by chance) esti-mated to be very small.

The 210Pb observations were not replicated and thus the relative and absolute errors could not be obtained.However, long-lived tracers such as 210Pb (half-life = 22.3 years) have much less variability than short-livedtracer such as chl-a (half-life = 23 days) (Reed et al., 2006). The coefficient of variation (the standard deviationdivided by the mean) obtained for the chl-a data was �20%, thus we set smaller errors for 210Pb at a coefficientof variation of 10%.

Graphs representing model fits and Bayesian results were created with the R-software (R DevelopmentCore Team, 2005).

3.8. Model complexity

The two models described in the previous section differ in only in one respect: model 1 has a fixed K,whereas K is a free parameter in model 2. The addition of one fit parameter to a model (model 2) implies

y = 0.5094x - 0.0163R2 = 0.7441

-0.02

0.00

0.02

0.04

0.06

0.08

0.00 0.05 0.10 0.15 0.20

Mean chl a concentration (ug cm-3)

SD

a

y = -0.0209x + 0.0102R2 = 0.0075

0.000

0.004

0.008

0.012

0.016

0.020

0.00 0.02 0.04 0.06 0.08 0.10 0.12Mean chl a concentration (ug cm-3)

SD

b

Fig. 2. Means of chl-a versus standard deviations (SD) plotted to obtain the standardized relative and absolute errors; (a) example of plotshowing correlation in station St 54 where the relative error and absolute error were taken as the slope and the intersection of the linear fitof the data, and (b) example of plot showing lack of correlation in station St 50 where then the relative error was set to zero and theabsolute error was calculated as the mean of the standard deviations.

Page 9

200 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

that the model fit is always better than the fit obtained with the simpler model (model 1). The question iswhether the improvement of the fit outweighs the cost of adding an additional parameter. In order to testfor this we used a one-tailed F-test (Sokal and Rohlf, 1995), similarly as in Soetaert et al. (1996a). The nullhypothesis is that there is no difference between the residual variance between modelled and observed datain the more elaborated approach (model 2) compared to the simpler one (model 1). The alternative hypoth-esis is that the complex approach significantly reduces the residual variance. The null hypothesis is rejectedwhen the calculated F-value > F-distribution with (df1 � df2)/df2 degrees of freedoms. The F-value is cal-culated as

F ¼SSR1�SSR2

df1�df2

� �SSR2

df2

� � ð8Þ

where SSR1 and SSR2 are the sum of the squared residuals of the simple and elaborated model respectively,and df1 and df2 are the respective degrees of freedom (number of observations � number of parameters � 1).The number of data points in the chl-a and 210Pb depth profiles is the number of observations.

3.9. Parameter uncertainty analysis

Parameter estimates derived by model fitting always have an associated error, because parameters arefitted to uncertain data with an imperfect model. Bayesian analysis provides a means to quantify the result-ing parameter uncertainty (Gelman et al., 2003). Here, a prior parameter uncertainty (e.g. minimum andmaximum values from the literature) is updated with the likelihood of finding model output given theobservations. The prior distribution times the likelihood gives the posterior probability distribution, whichdefines the uncertainty of the model parameters and the relation between parameters. Bayesian inferencewas run with a Markov Chain Monte Carlo (MCMC) simulation using the simple, yet effective Metropolisalgorithm. The MCMC was initiated with the parameter combination that gave the optimal fit (see above)to avoid a burn-in period, and with a non-informative prior probability distribution. We choose a uniformdistribution within a minimum value of 1E�10 and a maximum value of 1E+6 for all parameters, thusmaking sure that this prior probability distribution encompasses the ranges of each parameter. The largeprior parameter ranges ensures that the posterior distribution is effectively determined only by the informa-tion in the data.

Starting from a particular parameter combination, a new combination is selected by a random step (jump)in parameter space by the MCMC. If this parameter combination gives a better fit to the data than the pre-vious one, the new parameter set is accepted and used as new starting point for a next random step. If the newparameter combination provides a worse fit, it is accepted with a probability equal to the probabilities of thetested parameter combination divided by the probabilities of the existing parameter combination. The steplength for each parameter in the MCMC was (manually) optimised such that around 20% of the runs wereaccepted. The model was run 100,000 times, which typically gave 10,000–20,000 accepted runs. Convergenceof the MCMC was checked by visual inspection. The distribution of parameter values in the set of acceptedruns represents the posterior probability distribution of each parameter, from which the mean and standarddeviation of each parameter can be calculated and correlations among them. This technique has only foundrecent application in the field of biogeochemistry and ecology (i.e. Andersson et al., 2006; van Oevelen et al.,2006).

4. Results

4.1. Sediment composition

The open slope of the Gulf of Lions and Western Iberian Margin had similar concentrations of chl-a, con-tents of Corg, chl-a:phaeo and C:N ratios at equivalent water depths (Fig. 3, Table 2). Concentrations of chl-atended to decrease with increasing water depth, while Corg contents did not show a clear pattern with depth.

Page 10

Average 0-15 cm

0

5

10

15

20

010

0020

0030

0040

0050

0060

00

Water depth (m)

C:N

h

Average 0-15 cm

0.0

0.5

1.0

1.5

2.0

010

0020

0030

0040

0050

0060

00

Water depth (m)

%C

org

d

0-0.5 cm

0

1

2

3

4

010

0020

0030

0040

0050

0060

00

Water depth (m)

chla

(ug

cm-3

)

a Inventory

0

4

8

12

16

010

0020

0030

0040

0050

0060

00

Water depth (m)

chla

(ug

cm-2

)

b

0-0.5 cm

0.0

0.5

1.0

1.5

2.0

010

0020

0030

0040

0050

0060

00

Water depth (m)

%C

org

c

Average 0-15 cm

0.00

0.03

0.06

0.09

0.12

0.15

0.18

010

0020

0030

0040

0050

0060

00

Water depth (m)

chla

:pha

eo

f0-0.5 cm

0.00

0.03

0.06

0.09

0.12

0.15

0.18

010

0020

0030

0040

0050

0060

00

Water depth (m)

chla

:pha

eo

e

0-0.5 cm

02468

101214

010

0020

0030

0040

0050

0060

00

Water depth (m)

C:N

g

GoL Slope Cap de Creus canyonWIM Slope Nazare canyon

Fig. 3. (a) Chlorophyll-a concentration 0–0.5 cm, (b) chlorophyll-a inventory 0–15 cm, (c) organic carbon content 0–0.5 cm and (d)average 0–20 cm, (e) chlorophyll-a to phaeopigment ratio 0–0.5 cm and (f) average 0–15 cm, (g) atomic C:N ratio of the organic matter for0–0.5 cm and (h) average for 0–20 cm in the Cap de Creus canyons (Gulf of Lions, GoL), in the Nazare canyon (Western Iberian Margin,WIM) and in respective adjacent slopes.

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 201

The Cap de Creus canyon stations had lower chl-a concentrations and Corg contents than the Nazare canyonsstations at equivalent water depth (Fig. 3, Table 2). The upper/middle parts of the Nazare canyon had highestchl-a concentrations, and stations St 41 and St 26 showed the highest chl-a:phaeo ratios (Fig. 3, Table 2). The

Page 11

Table 2Characteristics of the sampling stations, listing station numbers, water depths, the chlorophyll-a concentrations, the chlorophyll-a tophaeopigments ratios, the organic carbon contents, and the atomic C:N ratios of the organic matter in the upper 0.5 cm of the sedimentand in the upper 15 cm (chlorophyll) or upper 20 cm (organic matter)

Stations Depth(m)

Chl-a 0–0.5 cm(lg cm�2)

Chl-a 0–15 cm(lg cm�2)

Corg 0–0.5 cm (%)

Corg 0–20 cm (%)

Chl-a:phaeo0–0.5 cm

chl-a:phaeo0–15 cm

C:N 0–0.5 cm

C:N 0–20 cm

Gulf of LionsOpen slope

St 54 343 0.12 0.61 0.55 0.55 0.05 0.03 8.72 8.33St 51 1209 0.09 0.06 – – 0.03 0.01 – –St 49 1874 0.07 0.06 0.44 0.39 0.03 0.03 6.18 6.99

Cap de Creus canyon

St 52 344 0.72 1.62 0.55 0.41 0.05 0.04 11.08 8.29St 50 1215 0.10 0.39 0.86 0.67 0.03 0.03 8.21 9.20St 47 1801 0.08 0.12 0.68 0.44 0.02 0.02 11.50 9.34St 48 2112 0.08 0.09 – – 0.04 0.03 – –

Western Iberian MarginOpen slope

St 39 307 0.11 0.61 0.34 0.30 0.02 0.02 8.43 8.57St 27 1000 0.13 0.50 0.98 1.02 0.03 0.03 8.48 8.80St 25 4798 0.02 0.05 0.69 0.49 0.02 0.02 7.36 7.05

Nazare canyon

St 41 332 4.20 9.18 1.85 1.66 0.16 0.08 8.58 9.56St 13 927 0.64 2.21 – – 0.04 0.04 – –St 26 1121 0.75 14.71 1.73 0.96 0.05 0.16 9.44 10.94St 34 2847 0.25 1.68 1.63 1.00 0.02 0.03 10.65 9.22St 24 4810 0.05 0.62 1.55 1.03 0.02 0.03 11.09 11.70St 22 4969 0.03 0.82 1.36 1.42 0.02 0.04 10.59 17.73

202 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

remaining stations from both canyons had similar chl-a:phaeo ratios. C:N ratios in both canyons at equivalentwater depths were similar, but in the Nazare canyon there was a clear increase in C:N ratios with increasingwater depth (Fig. 3, Table 2).

When comparing canyons against open slopes, canyon stations generally recorded higher concentrations ofchl-a and of Corg than adjacent open slope stations (Fig. 3, Table 2). This was particularly true for the WesternIberian Margin, where the Nazare canyon stations had markedly higher chl-a concentrations and Corg con-tents than open slope stations at equivalent depth. In the Gulf of Lions, the Cap de Creus canyon had onlyslightly higher chl-a concentrations than open slope stations, and Corg contents were similar in both systems.The chl-a:phaeo ratios in the upper regions of the Nazare canyon were higher than on the adjacent slope atequivalent depths (Fig. 3, Table 2). The deeper canyon and slope stations had similar ratios. In the Cap deCreus canyon and adjacent slope chl-a:phaeo ratios were always similar. C:N ratios were always higher in bothcanyon systems compared to the slopes (Fig. 3, Table 2).

4.2. Model output

210Pb data were not available for the stations St 52, St 25, St 34 and St 24 (Fig. 4) and it appeared thatmodel fits based on chlorophyll data alone did not result in well-constrained parameter estimates (resultsnot shown). Stations St 26, St 34, St 24 and St 22 had subsurface peaks of chl-a, and station St 26 also of210Pb, that could be indicative of non-local mixing (Fig. 4). The model we adopted accounts for diffusive mix-ing and useful to fit exponential-like tracer profiles, but sub-surface peaks cannot be accurately reproduced.Since the mechanisms that may account for these subsurface peaks were not known for the canyons, werefrained from modelling these stations.

The porosity profiles for all stations were readily fitted with the empirical exponential function (Fig. 5).The profiles of chl-a and 210Pb also show a good data versus model correspondence, except in the 210Pb

Page 12

St 52

0

5

10

15

20

25

0.0 0.5 1.0chl a (ug cm-3)

Sedi

men

tdep

th(c

m)

St 34

0

5

10

15

20

25

0.0 0.2 0.4 0.6 0.8chl a (ug cm-3)

Sedi

men

tdep

th(c

m)

St 24

0

5

10

15

20

25

0.0 0.1 0.2chl a (ug cm-3)

Sedi

men

tdep

th(c

m)

St 25

0

5

10

15

20

25

0.00 0.01 0.02 0.03chl a (ug cm-3)

Sedi

men

tdep

th(c

m)

St 26

0

5

10

15

20

25

0 1 2 3 4 5 6chl a (ug cm-3)

Sedi

men

tdep

th(c

m)

St 26

0

5

10

15

20

25

30

35

0 25 50 75 100

210Pb (dpm cm-3)

Sedi

men

tdep

th(c

m)

St 22

0

5

10

15

20

25

0.0 0.1 0.2chl a (ug cm-3)

Sedi

men

tdep

th(c

m)

St 22

0

5

10

15

20

25

30

35

0 50 100 150

210Pb (dpm cm-3)

Sedi

men

tdep

th(c

m)

Fig. 4. Stations showing chl-a and 210Pb profiles with subsurface maxima that could not be explained with diffusive mixing conditions (Eq.(4)), and stations lacking 210Pb data. Station St 52 is in the Cap de Creus canyon (Gulf of Lions), station St 25 is in the slope adjacent tothe Nazare canyon (Western Iberian Margin) and stations St 26, St 34, St 24 and St 22 are in the Nazare canyon.

Nazare - Canyon sites

Nazare - Slope sites

Capde Creus - Canyon sites

Capde Creus-Canyon sites

0

5

10

15

20

250.6 0.7 0.8

Porosity (v/v)

Sedi

men

t dep

th (c

m)

St. 54 - 343 m0

5

10

15

20

250.4 0.5 0.6

Porosity (v/v)

Sedi

men

t dep

th (c

m)

St. 39 - 307 m0

5

10

15

20

250.6 0.7 0.8 0.9

Porosity (v/v)

St. 51 - 1209 m

0

5

10

15

20

250.5 0.6 0.7 0.8

Porosity (v/v)

St. 48 - 2112 m0

5

10

15

20

250.4 0.5 0.6 0.7 0.8

Porosity (v/v)

Sedi

men

t dep

th (c

m)

St. 50 - 1215 m0

5

10

15

20

250.5 0.6 0.7 0.8

Porosity (v/v)

St. 47 - 1807 m0

5

10

15

20

250.7 0.8 0.9

Porosity (v/v)

Sedi

men

t dep

th (c

m)

St. 41 - 332 m0

5

10

15

20

250.7 0.8 0.9

Porosity (v/v)

St.13 - 927 m

0

5

10

15

20

250.6 0.7 0.8 0.9

Porosity (v/v)

St. 27 - 1000 m0

5

10

15

20

250.4 0.5 0.6 0.7 0.8

Porosity (v/v)

St. 49 - 1847 m

Fig. 5. Modelled (solid line) versus observed porosities (dots) for all stations explained with diffusive mixing conditions (Eq. (4)).

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 203

Page 13

Nazare – Canyon sites Cap de Creus – Canyon sites

St. 41 – 332 m St. 13 – 927 m St. 50 – 1215 m St. 47 – 1807 m St. 48 – 2112 m

Sed

imen

t dep

th (

cm)

0 1 2 3 4 5

2520

1510

50 ●●

●●●

●

●

●

●

●

Chl a (μg cm−3 )

0.1 0.3 0.5 0.7

2520

1510

50 ●●

●●

●●

●

●

●

●

Chl a (μg cm−3 ) Sed

imen

t dep

th (

cm)

0.02 0.06 0.10

2520

1510

50 ●●

●●

●●

●

●

●

●

Chl a (μg cm−3)

0.00 0.04 0.08

2520

1510

50 ●●

●●

●●

●

●

●

●

Chl a (μg cm−3

0.00 0.04 0.08

2520

1510

50 ●●

●●

●●

●

●

●

●

Chl a (μg cm−3)

Sed

imen

t dep

th (

cm)

30 35 40 45 50 55

2520

1510

50 ●● ●

●●

●

●

●

●

Pb210 (dpm cm−3)

65 70 75 80

2520

1510

50 ● ● ●

●●

●●

●

●

●

●

●Pb210 (dpm cm−3) Sed

imen

t dep

th (

cm)

0 10 30 50 70

2520

1510

50 ●● ●

●●

●

●

●

●

●

● Pb210 (dpm cm−3)

5 10 20 30

2520

1510

50 ●● ●

●●

●

●

●

●

●

●

Pb210 (dpm cm−3)

0 10 20 30 40 50

2520

1510

50 ●●●

●●

●●

●

●

●

●

Pb210 (dpm cm−3)

Nazare – Slope sites Cap de Creus – Slope sites

St. 39 – 307 m St. 27 – 1000 m St. 54 – 343 m St. 51 – 1209 m St. 49 – 1847 m

Sed

imen

t dep

th (

cm)

0.04 0.08 0.12

2520

1510

50 ●●

●●

●

●

●

●

●

Chl a (μg cm−3)

0.05 0.10 0.15

2520

1510

50 ●●

●●

●●

●

●

●

●

Chl a (μg cm−3) Sed

imen

t dep

th (

cm)

0.00 0.10 0.20

2520

1510

50 ●●

●●●

●

●

●

●

●

Chl a (μg cm−3)

0.02 0.06 0.10

2520

1510

50 ●●

●●●●

●

●

●

●

Chl a (μg cm−3)

0.00 0.02 0.04 0.06

2520

1510

50 ●●

●●●●

●

●

●

●

Chl a (μg cm−3)

Sed

imen

t dep

th (

cm)

10 20 30 40 50

2520

1510

50 ●

●●

●●

●

●

●

●

●

● Pb210 (dpm cm−3)

10 30 50

2520

1510

50 ●●●

●●

●

●

●

●

●

● Pb210 (dpm cm−3) Sed

imen

t dep

th (

cm)

5 10 15 20 25

2520

1510

50 ● ●●

●●

●

●

●

●

● Pb210 (dpm cm−3)

0 20 40 60 80

2520

1510

50 ●●●

●●

●

●

●

●

●

● Pb210 (dpm cm−3)

10 20 30 40

2520

1510

50 ●●●

●●

●

●

●

●

●

● Pb210 (dpm cm−3)

)

Fig. 6. Modelled (solid line) versus observed mean (dots) +-standard deviation (horizontal dash) chl-a and 210Pb profiles. Model profilesof stations St 41, St 48 and St 49 were obtained with model 2, the remaining profiles were obtained with model 1.

204 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

profiles for stations St 41, St 13 and St 39 (Fig. 6). For stations St 41 and St 48 the more parameter-richmodel (model 2) resulted in a significantly better fit of the chl-a and 210Pb data. The F-test of station St 49showed that model 2 is close to being significantly better than model 1 (a = 0.06, Table 3). Degradable chl-aconcentrations profiles obtained with model 1 vanish in the top 0.5 cm of the sediment (data not shown),whereas model 2 predicts a penetration of degradable chl-a down to 1.6 cm, being more in accordance withthe observed chl-a profile at St 49 (Fig. 6). Below 2 cm depth the two models are indistinguishable sinceboth predict background chl-a concentration. Hence model 2 can only improve the fit of only 3 out of10 data points and therefore cannot strongly improve the total explained variance. Moreover, fitted chl-adeposition by model 1 is considerably higher than for the two stations hundreds of meters higher up theslope, which is not very realistic. For these reasons we have chosen the results from model 2 as being mostrepresentative for this station. The chl-a deposition rates based on model 2 are a factor 20 to 40 lower in

Page 14

Table 3F-test comparing the residual variance between the simpler model 1 and the more complex model 2, and the mean ± standard deviation of the posterior parameter probabilitydistribution, as obtained with the Bayesian analysis

Stations Depth(m)

F- test(1 vs 2)

Bestmodel

Pb210 Depo(dpm cm�2 y�1)

Pb210background

(dpm cm�3)

Db (cm2 y�1) chl-a Depo(lg cm�2 y�1)

chl-abackground

(lg cm�3)

K k (d�1)

Gulf of LionsOpen slope

St 54 343 0.91 1 1.80 ± 0.11 2.65 ± 0.21 1.83 ± 0.43 0.36 ± 0.20 0.06 ± 0.02 18.3 0.04St 51 1209 0.57 1 1.37 ± 0.09 3.21 ± 0.22 0.50 ± 0.09 0.13 ± 0.02 0.01 ± 0.001 18.3 0.04St 49 1874 0.06 2 0.52 ± 0.04 4.11 ± 0.20 0.03 ± 0.01 0.004 ± 0.0003a 0.002 ± 0.001a 14.53 ± 0.03a 0.001 ± 4.6E�10a

Cap de Creus canyonSt 50 1215 0.78 1 3.12 ± 0.16 3.38 ± 0.27 0.09 ± 0.02 0.30 ± 0.07 0.04 ± 0.003 18.3 0.04St 47 1801 0.15 1 0.50 ± 0.05 2.55 ± 0.22 0.12 ± 0.06 0.40 ± 0.22 0.01 ± 0.001 18.3 0.04St 48 2112 0.001* 2 1.05 ± 0.06 2.90 ± 0.21 0.12 ± 0.03 0.007 ± 0.003a 0.003 ± 0.0004a 14.70 ± 0.36a 0.001 ± 6.5E�10a

Western Iberian MarginOpen slope

St 39 307 0.60 1 7.25 ± 0.48 9.98 ± 0.81 160.92 ± 71.92 1.58 ± 0.60 0.05 ± 0.01 18.3 0.037St 27 1000 0.90 1 1.99 ± 0.19 4.29 ± 0.29 0.50 ± 0.06 0.12 ± 0.04 0.04 ± 0.004 18.3 0.029

Nazare canyon

St 41 332 0.05* 2 3.03 ± 1.16 29.45 ± 4.88 35.21 ± 53.47 1.71 ± 0.66 0.25 ± 0.12 16.13 ± 0.22 0.004 ± 5.1E�10St 13 927 0.97 1 0.73 ± 0.52 67.19 ± 4.29 44.87 ± 18.74 1.53 ± 0.39 0.17 ± 0.02 18.3 0.030

Db = biodiffusion coefficients, chl a Depo and Pb210 Depo = deposition of chl-a and Pb210, chl-abackground = background concentration of chl-a, Pb210background ¼ background

concentration of Pb210, K = degradation parameter in Eq. (6) and k = decay rate of chl-a. When F-test is not significant the reported parameters are from model 1, and when issignificant (*) the reported parameters are from model 2.

a Refers to model 2 parameters whose accuracy is considered dubious (see Section 5).

R.

Ga

rcıaet

al./P

rog

ressin

Ocea

no

gra

ph

y7

6(

20

08

)1

92

–2

15

205

Page 15

Fig. 7. Posterior parameter probability distribution (resulting from the Bayesian analysis) for station St 51, as generated with model 1,where Pb bckgrnd = 210Pb background concentration (dpm cm�3), Pb depo = 210Pb deposition (dpm cm�2 y�1), Db = biodiffusioncoefficient (cm�2 y�1), Chl bckgrnd = chl-a background (lg cm�3), and Chl depo = chl-a deposition (lg cm�2 y�1). Each dot in a plotrepresents an accepted parameter combination; the histograms on the diagonal represent the posterior probability distribution for eachsingle parameter; the values in the upper panel reflect the correlation coefficient (r).

206 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

stations St 48 and St 49 as compared to stations hundreds of meters shallower. The profiles of the otherstations were not significantly better fitted with model 2 and therefore the results from model 1 are pre-sented (Table 3, Fig. 6).

The Bayesian analysis for stations fitted with model 1 or model 2 shows a similar pattern, but some dif-ferences between the model 1 and model 2 are apparent (Figs. 7 and 8). The posterior probability distribu-tion of the parameters in model 1 (station St 51 is taken as representative example) shows that modelparameters are normally distributed and are generally not correlated, except for 210Pb deposition and thebiodiffusion coefficient (Db), which are positively correlated (Fig. 7). This means that a combination of high

Page 16

Fig. 8. Results of the Bayesian analysis for station St 48 fitted with model 2, where Pb bckgrnd = 210Pb background concentration(dpm cm�3), Pb depo = 210Pb deposition dpm cm�2 y�1), Db = biodiffusion coefficient (cm�2 y�1), Chl bckgrnd = chl-a background(lg cm�3), and Chl depo = chl-a deposition (lg cm�2 y�1), Chl decay = K factor in (Eq. (4)). See also Fig. 7.

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 207

Db and high 210Pb deposition may produce a similarly good fit as a combination of low Db and low 210Pbdeposition.

The posterior distributions for the model parameters in model 2 are more complex (station St 48 taken asrepresentative example), but most striking is the strong positive correlation (0.94) between chl-a depositionand degradation (Fig. 8). Other weaker correlations exist between the parameters 210Pb background, 210Pbdeposition and biodiffusion coefficient (Db) and between the parameters chl-a background and degradation(Fig. 8).

Most model parameters were well constrained by the fitting as shown by low coefficients of variation(CoV = standard deviation/mean) that range from 1.3E-07 to 1.52 for all stations, with an average value of0.22 (Table 3). Also, the parameter ranges are significantly constrained with respect to the initial uncertaintyrange that was set for each parameter (see Material and Methods for initial ranges).

Page 17

210Pb deposition

012345678

332m

927m

307m

1000

m

1215

m

1801

m

2112

m34

3 m

1209

m

1874

m

Nazarécanyon

Open slope Cap Creuscanyon

Open slope

Western Iberian Margin Gulf of Lions

dpm

cm-2

y-1

chl a deposition

0.00.20.40.60.81.01.21.41.61.8

332m

927m

307m

1000

m

1215

m

1801

m

2112

m34

3m12

09m

1874

m

Nazarécanyon

Open slope Cap Creuscanyon

Open slope

Western Iberian Margin Gulf of Lions

ì gcm

-2y-1

chl a background

0.00

0.05

0.10

0.15

0.20

0.25

0.30

332 m

927 m

307 m

1000

m

1215

m

1801

m21

12m

343 m

1209

m

1874

m

Nazarécanyon

Open slope Cap Creuscanyon

Open slope

Western Iberian Margin Gulf of Lions

ì gcm

-3

Db

020406080

100120140160180

332 m

927 m

307 m

1000

m

1215

m

1801

m

2112

m34

3 m

1209

m

1874

m

Nazarécanyon

Open slope Cap Creuscanyon

Open slope

WesternI berian Margin Gulf of Lions

cm2

y-1

210Pb background

01020304050607080

332 m

927 m

307 m

1000

m

1215

m

1801

m

2112

m34

3 m

1209

m

1874

m

Nazarécanyon

Open slope Cap Creuscanyon

Open slope

Western Iberian Margin Gulf of Lions

dpm

cm-3

a

b

c

d

e

Db

0.00.40.81.21.62.0

1215

m

1801

m

2112

m34

3m12

09m

1874

m

CapC reus Open slope

Gulf of Lions

cmy

Fig. 9. Estimated biodiffusion coefficients (cm2 y�1), deposition rates of chl-a (lg cm�2 y�1) and of 210Pb (dpm cm�2 y�1), andbackground concentrations of chl-a (lg cm�3) and of 210Pb (dpm cm�3) in the Nazare canyon (Western Iberian Margin), in the Cap deCreus canyons (Gulf of Lions), and in the respective adjacent slopes.

208 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

4.3. Biodiffusion coefficient, chl-a and 210Pb deposition rates, background concentrations and decay rates

Estimated biodiffusion coefficients, deposition rates of chl-a and of 210Pb, and background concentrationsof chl-a and of 210Pb were generally higher in the Western Iberian Margin than in the Gulf of Lions (Fig. 9,Table 3).

The biodiffusion coefficient in the Nazare canyon and adjacent slope did not show any specific trendwith water depth or between canyon and slope stations. The slope station at 307 m (St 39) showed a clearpeak in the biodiffusion coefficient (Fig. 9a, Table 3). Differently, biodiffusion coefficients in the Cap deCreus canyon stations were similar, while at the adjacent slope they decreased with increasing waterdepth.

Deposition rates of chl-a in the Nazare canyon were higher than on the adjacent slope, although thedifference was small for the stations at 332 and 307 m (Fig. 9b, Table 3). In the Cap de Creus canyon chl-a

Page 18

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 209

deposition was higher than on the adjacent slope at equivalent water depths, with chl-a deposition ratesdecreasing with increasing water depths in both areas.

210Pb deposition was lower in the Nazare canyon than at the adjacent slope at equivalent water depth(Fig. 9c, Table 3). In the Cap de Creus canyon 210Pb depositions tended to be higher than on the adjacentslope.

The Nazare canyon had higher background concentrations of chl-a and of 210Pb than the adjacent slope(Fig. 9d and e, Table 3). In the Cap de Creus canyon and adjacent slope background concentrations ofchl-a and of 210Pb were similar, with chl-a background concentrations decreasing with increasing waterdepth.

Model 2 fitted significantly better for the stations St 41, St 48 and St 49, and the derived decay rates of chl-awere 10 and 40 times lower than in the stations that were fitted with model 1 (Table 3). For those cases wheremodel 2 is not significantly better than model 1, the decay rates estimated with model 2 were similar to thedecay rates of model 1. The only exceptions were stations St 13 and St 27 that had respectively 40 and 6 timeshigher decay rates, and St 47 with 8 times lower decay rates.

5. Discussion

Using a combined observation and dual tracer modelling approach, two submarine canyon ecosystems withhydrographically different regimes, and their respective adjacent slopes, have been found to differ in terms ofseveral sedimentary characteristics. The steady-state tracer model fitted simultaneously chl-a and 210Pb sedi-ment profiles and allowed (1) to single out those sediments that are subject to other transport processes thandiffusive mixing, (2) to test chlorophyll-a decay rates are only temperature-dependent or that they underwentsignificant alteration before deposition, and (3) to characterize sediments in terms of deposition rates, decayrates, and biodiffusion coefficients.

5.1. Model results

Most of the stations in the Nazare canyon (4 out of 6) had pronounced subsurface peaks in chl-a and/or210Pb (Fig. 4) that could not be explained with the biodiffusion process adopted in the tracer model (Eq. (4)).Subsurface peaks were also discernable in some of the Cap de Creus Canyon sites, but not to the extent thatthey could not be represented with a simple biodiffusion model approach. Such subsurface peaks can becaused either by non-local mixing effectuated by large benthic fauna (Meysman et al., 2003) or by physicalprocesses such as turbidity flows that may deposit low chl-a sediment on top of sediments with high chl-a con-centrations (Fig. 4). From the available data we cannot identify the process that has caused these non-diffusivechl-a profiles and we have therefore not attempted to model these profiles. It is however noteworthy that mostof these deviating profiles were associated to canyon sites.

Three stations in the Iberian margin (St 13, St 41 and St 39) showed straight 210Pb profiles, which deviatesfrom the exponential decrease with sediment depth as a result of diffusive mixing (Fig. 6). As with the profileswith subsurface peaks mentioned above, it is also noteworthy that two of the stations (St 13 and St 41) were inthe canyon. Hence, all Nazare canyon stations had the common trait of having chl-a or 210Pb profiles deviat-ing from the traditional exponential decrease. The Nazare canyon is an active canyon prone to turbidity flows,and influenced by internal tide circulation that resuspends the sedimentary material during each tidal cycle (DeStigter et al., 2007), which may result in deviating profiles. Macrofauna mixing activity may also play a role,but it is difficult to conclude whether the origin of these profiles is fully due to bioturbation. The only faunaldensities available for the same stations are on meiofauna (Garcıa et al., 2007), which are not the expectedbioturbators. At other stations however, macrofauna densities (0.5 mm-sieve) show a clear decrease withincreasing water depth with values of �20,000 ind m�2 at �400 m (Flach, 2003), 9273 ind m�2 at �2894 m,595 ind m�2 at 3514 m and 408 ind m�2 at 4141 m (Curdia et al., 2004). The upper and middle Nazare canyonclearly shows high macrofauna densities when compared to other continental margins (Flach, 2003). This sug-gests the presence of potential bioturbators in the area. However, the tracers’ profiles along the canyonbathymetry showed similar degree of disturbance (Figs. 4 and 6). If bioturbation was originating theseprofiles, the disturbance degree should decrease with water depth in line with the decrease in macrofauna

Page 19

210 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

densities. Hence, physical disturbance in the Nazare canyon plays a more important role shaping the sedimen-tary record. When modelling the straight 210Pb profiles in the slope station St 39, an extremely high mixingrate was obtained (160 cm y�1) as compared to the other stations (Table 3, Fig. 9) or other studies (e.g.Turnewitsch et al., 2000; Wheatcroft, 2006). A straight 210Pb profile can only be reproduced with a diffusivemodel when a very high Db is adopted. The lack of macrofaunal densities for the adjacent slope means that wecannot evaluate whether biological mixing can explain these straight profiles, however given the extremelyhigh Db values, we believe that physical mixing likely plays a role here.

The stations exhibiting diffusive mixing conditions in the sediments were fitted with two models. In onemodel chl-a decay rate was fixed based on in situ temperature and using the empirical degradation relation-ship derived by Sun et al. (1993). In model 2, the same relation was used but with K as additional parameterto be fitted. This permitted to test if the chl-a degradation rate is only temperature dependent, or whetherthe other site/depth specific factors that influence its lability play a role. We used an F-test to establishwhether the additional free model parameter significantly improved the fits of the data. Surprisingly, in mostof the cases (7 out of 10), the extra fitting parameter did not significantly improve the fitting. This falsifiedour initial assumption that the chlorophyll-derived matter was strongly altered (aged) before settling.Apparently, using the chl-a decay rate as derived by Sun et al. (1993) from incubations of shallow estuarinesediments, and widely used for different environments (e.g. Boon and Duineveld, 1998; Gerino et al., 1998;Green et al., 2002) is a good-enough estimate for most continental slopes and submarine canyons investi-gated in this study.

In three stations (St 41, St 48 and St 49) though, the data were (quasi-) significantly better fitted (seeResults) by model 2, because the penetration of chlorophyll was significantly underestimated by model 1.The coefficient that governs exponential decay of a tracer with sediment depth is �

ffiffiffiffiffiffiffiffiffiffiffik=Db

p, hence the fit of

a given profile can be improved by decreasing the degradation rate of chlorophyll (k) with a certain factoror increasing the mixing intensity (Db) with the same amount. The underestimated penetration can thus beexplained by either an overestimated degradation (and thus chl-a disappears too fast), or underestimated mix-ing (and thus chl-a was not diluted fast enough into the sediment).

Within our data set, the mixing coefficient is predominantly determined by the 210Pb data, because of thedeep penetration of the exponential profile of 210Pb. This was verified by fitting Db against only the 210Pbdata, which gave very similar results as presented here (Table 3). The mixing rate of chlorophyll-a can there-fore be underestimated if organisms favourably handle fresh over more refractory material in a processcalled age-dependent mixing (Smith et al., 1993). This concept was recently challenged by Reed et al.(2006), who argue that short-lived tracers do not undergo sufficient bioturbation events to give a profile thatcan be reliably used to derive a biological mixing coefficient. Long-lived tracers in contrast, do integratesufficient bioturbation events and should therefore provide a better estimate of mixing. Our mixing coeffi-cient is determined primarily by the long-lived 210Pb tracer and should thus provide be a reliable estimate.However, we have insufficient data to exclude the possibility of independent mixing of fresh and aged mate-rial at the stations.

The obtained decay constant of chlorophyll-a is determined primarily by the exponential decay in theupper few cm of the sediment (Fig. 6). Because of that the chlorophyll-a degradation rates are based oncomparatively few data points, because sampling in these upper cm is limited due to practical issues of slic-ing. The low number of relevant data points explains why it proves difficult to distinguish model 1 frommodel 2 (see also Section 4). Only when chlorophyll-a penetrates deep into the sediment it becomes neces-sary to alter model 1. These considerations show the difficulties in obtaining reliable degradation rates forchlorophyll-a and an assessment of fitting robustness provides therefore a good means to quantify theresulting uncertainty.

The robustness of the diagenetic model parameters was assessed with a Bayesian analysis (Andersson et al.,2006). This technique provides a means to determine the probability distribution of the model parameters andtheir correlations after fitting to observational data (Figs. 7 and 8). Most parameters were well constrained bythe data as indicated by the small coefficients of variation (CV) (Table 3) and the reduced posterior parameterranges obtained with the Bayesian analysis (not shown). Two noteworthy exceptions are the mixing coefficientand 210Pb deposition rates of stations St 13 and St 41, which were poorly constrained (CV 0.71, 1.52). This isdue to the fact that the 210Pb profile does not show the traditional decrease with sediment depth (Fig. 6). Most

Page 20

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 211

model parameters are normally distributed and uncorrelated, especially when obtained with model 1 (Fig. 7).However, chl-a deposition and decay are strongly positively correlated in model 2 (Fig. 8). This correlation iseasily explained, because a vertical profile of chl-a will not change drastically when a higher deposition of chl-ais accompanied by a higher degradation rate of chl-a. The negative correlation between fitted 210Pb back-ground concentration and deposition rates (Fig. 7) can be explained by the following reasoning. A lower210Pb deposition should be accompanied with a higher 210Pb background value in order to keep the 210Pb con-centration in the mixed layer near the observed concentration.

5.2. Patterns in the biogeochemical parameters

The Cap de Creus canyon and open slope sediments seemed to be less active than the Nazare canyon andrespective open slope in terms of mixing intensities and chl-a and 210Pb deposition (Fig. 9, Table 3). This wasconfirmed by the chl-a and Corg contents (Fig. 3, Table 2) that were respectively, 3–30 times lower and 1.5–4times lower in the Cap de Creus canyon than in the Nazare canyon. The respective open slopes shared similarchl-a and Corg contents, but the markedly lower chl-a and 210Pb background concentrations in the Cap de Cre-us canyon and open slope sediments confirmed they were less active in terms of organic matter deposition andburial (Fig. 9, Table 3). These differences may be partly explained by the generally higher primary productivityin the Western Iberian Margin (Alvarez-Salgado et al., 2003) than in the Gulf of Lions (Lefevre et al., 1997).

The Cap de Creus and Nazare canyons also showed different organic matter accumulation patterns.When compared to the respective open slopes, the Nazare canyon had 5–30 times higher chl-a contents,2–5 times higher Corg contents (Fig. 3, Table 2) and higher chl-a and 210Pb background concentrations(Fig. 9, Table 3). The high 210Pb background concentrations in the canyon were a sampling artefact. Toreach typical 210Pb background values <10 dpm cm�3, deeper piston core samples are needed (De Stigteret al., 2007). Differently, the Cap de Creus canyon and respective adjacent open slope were similar withrespect to their chl-a and Corg contents (Fig. 3, Table 2), their chl-a and 210Pb background concentrationsand the chl-a deposition rates (Fig. 9, Table 3). In contrast, mixing intensities were lower in the canyon, and210Pb depositions were slightly higher. In the Nazare canyon more than half of the chl-a and 210Pb profileshad clear subsurface peaks indicating a high sediment disturbance (Fig. 4). Overall, the Nazare canyon eco-system is more active in terms of organic matter accumulation and burial than the Cap de Creus canyon.While the Cap de Creus canyon may trap some bulk organic matter, only little labile phytodetritus accu-mulates within the canyon. Only the upper part traps some phytodetritus, as chl-a concentrations wereslightly higher (2–6 times) than on the adjacent slope, but still considerably lower than in the Nazare can-yon. The higher primary productivities in the Western Iberian Margin alone may not fully explain the dif-ference in organic matter accumulation in the Nazare and Cap de Creus canyons. The transportmechanisms of suspended particles triggered by different current regimes are more likely to be responsiblefor the observed differences in the sediments.

The Western Iberian Margin is characterised by tide driven currents, internal waves and upwelling regime(McCave and Hall, 2002; Vitorino et al., 2002) that favours the formation of nepheloid layers transportingsuspended material offshore (Oliveira et al., 2002; Van Weering et al., 2002). The presence of dense nepheloidlayers in the upper Nazare canyon indicates transport of shelf material into the canyon (De Stigter et al.,2007). Tide driven currents within the canyon resuspend and deposit sedimentary material in cycles; this mate-rial is transported up and down canyon with the ebb and flood tides, producing a net down canyon transportthat can be coupled with sporadic turbidity flows (De Stigter et al., 2007). This oceanographic regime favoursthe sedimentation of suspended material and burial, which would explain the high organic contents, deposi-tions and background concentrations in the Nazare canyon. Faster sediment accumulation has previouslybeen reported for the Nazare canyon than for the Iberian continental slope (Schmidt et al., 2001; Van Weeringet al., 2002; Epping et al., 2002; De Stigter et al., 2007). In contrast with the previous studies, 210Pb depositionwas lower in the Nazare canyon than at the adjacent slope and chl-a deposition was similar (Fig. 9, Table 3).In combination with the higher background values of 210Pb and chl-a found in the canyon, this may suggestthat continuous deposition may not be the case. The profiles of 210Pb in the Nazare canyon (Fig. 6) do notshow the traditional decrease with sediment depth, which suggests mixed conditions. Intense animal reworkingor sediment slides falling on top of each other may produce this type of profile. As mentioned before, the latter

Page 21

212 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

seems more probable because all the canyon stations showed disturbed chl-a and/or 210Pb profiles. In the can-yon, the observed 210Pb concentrations are close to background values because the sediment was not sampleddeep enough to reach typical 210Pb background values (De Stigter et al., 2007). Therefore, the model estimatesthat a small deposition is necessary to maintain the observed 210Pb concentrations because it assumes constantdeposition (Eq. (4)).

A hydrographically different regime characterises the Gulf of Lions. The convergence of the shelf cycloniccirculation of water and the Liguro-Provenc�al current nearby the Cap de Creus canyon causes intense down-welling of shelf water on the slope along with down-canyon sediment transport (Palanques et al., 2006b).Nepheloid layers produced by along slope contour currents disperse suspended material offshore (Durrieude Madron et al., 1990; Durrieu de Madron, 1994). Most of the time, fine-grained sediment and organic mat-ter may be spread out evenly over the canyon and adjacent slope by settling without much focusing of depo-sition in the canyon. All chl-a and 210Pb profiles give evidence of low deposition and low mixing intensity, wellin accordance with the relatively low primary productivity and lack of tidal transport mechanism. In addition,dense shelf water cascading events seasonally occur in the Gulf of Lions (Durrieu de Madron et al., 2005), andcan intensify the down-welling of shelf water along the slope and canyons. Sporadically, the downslope cur-rents in Cap de Creus canyon can be large enough to resuspend sand, which subsequently erodes fine-grainedcohesive sediments and form giant furrows field and transport large amounts of organic material to the deepocean (Canals et al., 2006). This could explain the low and homogenous organic contents and phytodetrituslabilities, low depositions, mixing intensities, and background concentrations in the Cap de Creus canyon.Organic material would be eroded away. Whereas this may be the case in some parts of the canyon, the studiedsediments do not show evidence of an erosion regime (coarse grain sediment or decapitated 210Pb profiles)resulting from dense water cascading. Hence, the observed low depositions and organic contents are the resultof low primary productivity, lack of focusing mechanism in the canyon coupled with intensified down-wellingevents would intermittently transport accumulated material down canyon.

Several rivers (e.g. Rhone, Tet) and creeks discharge lithogenic material in the Gulf of Lions. Due to thecirculation pattern in the Gulf of Lions, this material is funnelled through the Cap de Creus canyon. Inagreement, C:N ratios indicate a strong terrestrial signal in the organic carbon in this canyon. The Nazarecanyon shares similar C:N ratios with the Cap de Creus canyon, but in this case, the canyon is far awayfrom the influence of any river discharge. The Nazare canyon is very close to land and the strong terrestrialsignature of the organic carbon in this canyon may have its origins from the weathering of the semi-aridcliffs along the coast.

6. Conclusions

The chl-a decay rate as obtained from the empirical temperature-dependent relation by Sun et al. (1993)(Eq. (6)) is a good-enough estimate for modelling chl-a degradation in most continental slopes and canyonstations investigated herein.

The Nazare canyon and adjacent slope are more active in terms of organic matter deposition and burial,and generally show higher organic contents and background concentrations than the Cap de Creus canyonand respective slope. Most of the vertical tracer profiles in the Nazare canyon sediment show evidence ofstrong perturbations or predominance of transient effects, which cannot be explained by steady-state diffusivemixing and decay.

Acknowledgements

This study was supported by EUROSTRATAFORM Project, EC Contract EVK3-CT-2002-00079 fundedby the European Commission, DG-XII, and the HERMES Project, EC Contract OCE-CT-2005-511234funded by the European Commission’s Sixth Framework Programme under the priority Sustainable Develop-ment, Global Change and Ecosystems. Shiptime on RV ‘Pelagia’ was provided by Royal NIOZ, Texel, TheNetherlands. The authors thank all of the crew from the R V ‘Pelagia’ for their work, and Dr. Erica Koning,Dr. Tanja Kouwenhoven and Karoliina Koho for valuable help provided during core deployment, retrievaland sample processing. We thank A. Palacz for assisting with chloropigment measurements, Wim Boer and

Page 22

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 213

Sharyn Crayford from NIOZ for 210 Pb and C–N analysis respectively, and two reviewers for their commentsthat helped to improve the manuscript. This is publication number 4218 of the Netherlands Institute of Ecol-ogy (NIOO-KNAW), Yerseke.

References

Aller, R.C., 1982. The effect of macrobenthos on chemical properties of marine sediments and overlying water. In: McCall, P.L., Tevesz,

M.S.S. (Eds.), Animal–Sediment-Relations. Plenum Publishing Co., New York, pp. 53–102.

Aller, R.C., 1994. Bioturbation and remineralization of sedimentary organic-matter – effects of redox oscillation. Chemical Geology 114,

331–345.

Alvarez-Salgado, X.A., Figueiras, F.G., Perez, F.F., Groom, S., Nogueira, E., Borges, A.V., Chou, L., Moncoiffe, G., Rıos, A.F., Miller,

A.E.J., Frankignoulle, M., Savidge, G., Wollast, R., 2003. The Portugal coastal counter current off NW Spain: new insights on its

biogeochemical variability. Progress in Oceanography 56, 281–321.

Andersson, J.H., Middelburg, J.J., Soetaert, K., 2006. Identifiability and uncertain analysis of bio-irrigation rates. Journal of Marine

Research 64, 407–429.

Boer, W., van den Bergh, G.D., de Haas, H., de Stigter, H.C., Gieles, R., van Weering, T.C.E., 2006. Validation of accumulation rates in

Teluk Banten (Indonesia) from commonly applied 210Pb models, using the 1883 Krakatau tephra as time marker. Marine Geology

227, 263–277.

Boon, A.R., Duineveld, G.C.A., 1998. Chlorophyll-a as a marker for bioturbation and carbon flux in southern and central North Sea

sediments. Marine Ecology Progress Series 162, 33–43.

Boudreau, B.P., 1997. Diagenetic Models and Their Implementation: Modelling Transport and Reactions in Aquatic Sediments. Springer-

Verlag, Heidelberg, 414pp.

Canals, M., Puig, P., Durrieu de Madron, X., Heussner, S., Palanques, A., Fabres, J., 2006. Flushing submarine canyons. Nature 444,

354–357.

Carson, B., Baker, E.T., Hickey, B.M., Nittrouer, C.A., Demaster, D.J., Thorbjarnarson, K.W., Snyder, G.W., 1986. Modern sediment

dispersal and accumulation in quinault submarine-canyon – a summary. Marine Geology 71, 1–13.

Cowie, G.L., Hedges, J.I., Calvert, S., 1992. Sources and relative reactivities of amino acids, neutral sugars, and lignin in an intermittently

anoxic marine environment. Geochimica et Cosmochimica Acta 56, 1963–1978.

Curdia, J., Carvalho, S., Ravara, A., Gage, J.D., Rodrigues, A.M., Quintino, V., 2004. Deep macrobenthic communities from Nazare

submarine canyon (NW Portugal). Sciencia Marina 68, 171–180.

De Stigter, H.C., Boer, W., De Jesus Mendes, P.A., Jesus, C.C., Thomsen, L., Van den Bergh, G., Van Weering, T.C.E., 2007. Recent

sediment transport and deposition in Nazare Canyon, Portuguese continental margin. Marine Geology. doi:10.1016/j.mar-

geo.2007.04.01.

Durrieu de Madron, X., Nyffeler, F., Godet, C.H., 1990. Hydrographic structure and nepheloid spatial distribution in the Gulf of Lions

continental margin. Continental Shelf Research 10, 915–929.

Durrieu de Madron, X., 1994. Hydrography and Nepheloid structures in the Grand-Rhone Canyon. Continental Shelf Research 14, 457–

477.

Durrieu de Madron, X., Zervakis, V., Theocharis, A., Georgopoulos, D., 2005. Comments on ‘‘Cascades of dense water around the world

ocean”. Progress in Oceanography 64, 83–90.

Epping, E., van der Zee, C., Soetaert, K., Helder, W., 2002. On the oxidation and burial of organic carbon in sediments of the Iberian

margin and Nazare Canyon (NE Atlantic). Progress in Oceanography 52, 399–431.

Fiuza, A.F.G., Hamann, M., Ambar, I., Diaz del Rio, G., Gonzalez, N., Cabanas, J.M., 1998. Water masses and their circulation off

western Iberia during May 1993. Deep Sea Research Part II Topical Studies in Oceanography 45, 1127–1160.

Flach, E., 2003. Factors controlling soft bottom macrofauna along and across european continental margins. In: Wefer, G., Billet, D.,

Hebbeln, D., Jørgensen, B.B., Schluter, M., van Weering, T. (Eds.), Ocean Margin Systems. Springer-Verlag, Berlin, pp. 351–363.

Font, J., 1987. The path of the Levantine intermediate water to the Alboran sea. Deep-Sea Res I 34, 1745–1755.

Garcıa, R., Koho, K.A., De Stigter, H.C., Epping, E., Koning, E., Thomsen, L., 2007. Distribution of meiobenthos in the Nazare Canyon

and adjacent slope (western Iberian Margin) in relation to sedimentary composition. Marine Ecology – Progress Series 240, 207–220.

Garcıa, R., Thomsen, L., in press. Bioavailable organic matter in surface sediments of the Nazare canyon and adjacent slope (Western

Iberian Margin). Journal of Marine Systems. doi:10.1016/j.jmarsys.2007.11.004.

Gelman, A., Carlin, J.B., Stern, H.S., Rubin, D.B., 2003. Bayesian Data Analysis. In: Texts in Statistical Science Series. Chapman and

Hall/CRC.

Gerino, M., Aller, R.C., Lee, C., Cochran, J.K., Aller, J.Y., Green, M.A., Hirschberg, D., 1998. Comparison of different tracers and

methods used to quantify bioturbation during a spring bloom: 234-thorium, luminophores and chlorophyll-a. Estuarine Coastal and

Shelf Science 46, 531–547.

Green, M.A., Aller, R.C., Cochran, J.K., Lee, C., Aller, J.Y., 2002. Bioturbation in shelf/slope sediments off Cape Hatteras, North

Carolina: the use of Th-234, Ch1-a, and Br – to evaluate rates of particle and solute transport. Deep-Sea Research Part II – Topical

Studies in Oceanography 49, 4627–4644.

Lefevre, D., Minas, H.J., Minas, M., Robinson, C., Williams, P.J.L.E.B., Woodward, E.M.S., 1997. Review of gross community

production, primary production, net community production and dark community respiration in the Gulf of Lions. Deep-Sea Research

44, 801–832.

Page 23

214 R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215

Lohse, L., Kloosterhuis, R.T., de Stigter, H.C., Helder, W., van Raaphorst, W., van Weering, T.C.E., 2000. Carbonate removal by

acidification causes loss of nitrogenous compounds in continental margin sediments. Marine Chemistry 69, 193–201.

Maze, J.P., Arhan, M., Mercier, H., 1997. Volume budget of the eastern boundary layer off the Iberian Peninsula. Deep-Sea Research Part

I – Oceanographic Research Paper 44, 1543–1574.

McCave, I.N., Hall, I.R., 2002. Turbidity of waters over the Northwest Iberian continental margin. Progress in Oceanography 52, 299–

313.

Meysman, F.J.R., Boudreau, B., Middelburg, J.J., 2003. Relations between local, nonlocal, discrete and continuous models of

bioturbation. Journal of Marine Research 61, 391–410.

Meysman, F.J.R., Bernard, P., Boudreau, P., Middelburg, J.J., 2005. Modeling reactive transport in sediments subject to bioturbation and

compaction. Geochimica et Cosmochimica Acta 69, 3601–3617.

Meysman, F.J.R., Middelburg, J.J., Heip, C.H.R., 2006. Bioturbation: a fresh look at Darwin’s last idea. Trends in Ecology and Evolution

21, 688–695.

Millot, C., 1987. Circulation in the western Mediterranean Sea. Oceanologica Acta 10, 143–149.

Millot, C., 1990. The Gulf of Lions’ hydrodynamics. Continental Shelf Research 10, 885–894.

Monaco, A., Biscaye, P., Soyer, J., Pocklington, R., Heussner, S., 1990. Particle fluxes and ecosystem response on a continental-margin –

the 1985–1988 Mediterranean ecomarge experiment. Continental Shelf Research 10, 809–839.

Mullenbach, B.L., Nittrouer, C.A., 2000. Rapid deposition of fluvial sediment in the Eel Canyon, northern California. Continental Shelf

Research 20, 2191–2212.

Oliveira, A., Vitorino, J., Rodrigues, A., Jouanneau, J.M., Dias, J.M.A., Weber, O., 2002. Nepheloid layer dynamics in the northern

Portuguese shelf. Progress in Oceanography 52, 195–213.

Palanques, A., Martin, J., Puig, P., Guillen, J., Company, J.B., Sarda, F., 2006a. Evidence of sediment gravity flows induced by trawling in

the Palamos (Fonera) submarine canyon (northwestern Mediterranean). Deep-Sea Research Part I – Oceanographic Research Paper

53, 201–214.

Palanques, A., Durrieu de Madron, X., Puig, P., Fabres, J., Guillen, J., Calafat, A., Canals, M., Heussner, S., Bonnin, J., 2006b.

Suspended sediment fluxes and transport processes in the Gulf of Lions submarine canyons. The role of storms and dense water

cascading. Marine Geology 236, 43–61.

R Development Core Team, 2005. R: A language and environment for statistical computing. R Foundation for Statistical Computing,

Vienna, Austria. ISBN 3-900051-07-0, URL <http://www.R-project.org>.

Reed, D.C., Huang, K., Boudreau, B.P., Meysman, F.J.R., 2006. Steady-state tracer dynamics in a lattice-automaton model of

bioturbation. Geochimica et Cosmochimica Acta 70 (23), 5855–5867.

Schmidt, S., de Stigter, H.C., van Weering, T.C.E., 2001. Enhanced short-term sediment deposition within the Nazare Canyon, North-

East Atlantic. Marine Geology 173, 55–67.

Shuman, F.R., Lorenzen, C.F., 1975. Quantitative degradation of chlorophyll by a marine herbivore. Limnology and Oceanography 20,

580–586.

Smith, C.R., Pope, R.H., DeMaster, D.J., Magaard, L., 1993. Age-dependent mixing of deep-sea sediments. Geochimica et Cosmochimica

Acta 57, 1473–1488.

Smith, C.R., Berelson, W., DeMaster, D.J., Dobbs, F.C., Hammond, D., Hoover, D.J., Pope, R.H., Stephens, M., 1997. Latitudinal

variations in benthic processes in abyssal equatorial pacific: control by biogenic particle flux. Deep Sea Research Part II: Topical

Studies in Oceanography 44, 2295–2317.

Soetaert, K., Herman, P.M.J., Middelburg, J.J., Heip, C., de Stigter, H.S., van Weering, T.C.E., Epping, E., Helder, W., 1996a. Modeling

Pb-210-derived mixing activity in ocean margin sediments: Diffusive versus nonlocal mixing. Journal of Marine Research 54, 1207–

1227.

Soetaert, K., Herman, P.M.J., Middelburg, J.J., 1996b. Dynamic response of deep-sea sediments to seasonal variations: a model.

Limnology and Oceanography 41, 1651–1668.

Soetaert, K., de Clippele, V., Herman, P., 2002. FEMME, a flexible environment for mathematically modelling the environment.

Ecological Modelling 151, 177–193.

Sokal, R.R., Rohlf, F.J., 1995. Biometry, third ed. In: The Principles and Practice of Statistics in Biological Research Freeman, New York,

887pp.

Sun, M.Y., Aller, R.C., Lee, C., 1991. Early diagenesis of chlorophyll-a in long-island sound sediments – a measure of carbon flux and

particle reworking. Journal of Marine Research 49, 379–401.

Sun, M.Y., Lee, C., Aller, R.C., 1993. Laboratory studies of oxic and anoxic degradation of chlorophyll-a in long-island sound sediments.

Geochimica et Cosmochimica Acta 57, 147–157.

Thomsen, L., van Weering, T., Gust, G., 2002. Processes in the benthic boundary layer at the Iberian continental margin and their

implication for carbon mineralization. Progress in Oceanography 52, 315–329.

Turnewitsch, R., Witte, U., Graf, G., 2000. Bioturbation in the abyssal Arabian Sea: influence of fauna and food supply. Deep Sea

Research Part II: Topical Studies in Oceanography 47, 2877–2911.

Van Oevelen, D., Middelburg, J.J., Soetaert, K., Moodley, L., 2006. The fate of bacterial carbon in an intertidal sediment: modeling an

in situ isotope tracer experiment. Limnology and Oceanography 51, 1302–1314.

Van Weering, T.C.E., De Stigter, H.C., Boer, W., De Haas, H., 2002. Recent sediment transport and accumulation on the NW Iberian

margin. Progress in Oceanography 52, 349–371.

Vanney, J.R., Mougenot, D., 1981. La plateforme continentale du Portugal et les provinces adjacentes: analyse geomorphologique.

Memorias dos Servic�os Geologicos de Portugal 28, 145.

Page 24

R. Garcıa et al. / Progress in Oceanography 76 (2008) 192–215 215

Vitorino, J., Oliveira, A., Jouanneau, J.M., Drago, T., 2002. Winter dynamics on the northern Portuguese shelf. Part 1: physical processes.

Progress in Oceanography 52, 29–153.

Wheatcroft, R.A., Jumars, P.A., Smith, C.R., Nowell, A.R.M., 1990. A mechanistic view of the particulate biodiffusion coefficient: step

length, rest period and transport directions. Journal of Marine Research 48, 177–207.

Wheatcroft, R.A., 2006. Time-series measurements of macrobenthos abundance and sediment bioturbation intensity on a flood-dominated

shelf. Progress in Oceanography 71, 88–122.

Yentsch, C.S., Menzel, D.W., 1963. A method for the determination of phytoplankton chlorophyll and phaeophytin by fluorescence.

Deep-Sea Research 10, 221–231.