geophysical constraints for organic carbon sequestration capacity of zostera marina seagrass meadows and surrounding habitats

Geophysical constraints for organic carbon sequestration capacity of Zostera marina seagrass meadows and surrounding habitats Toshihiro Miyajima,1* Masakazu Hori,2 Masami Hamaguchi,2 Hir

Geophysical constraints for organic carbon sequestration capacity of Zostera marina seagrass meadows and surrounding habitats

Toshihiro Miyajima,1* Masakazu Hori,2 Masami Hamaguchi,2 Hiromori Shimabukuro,2Goro Yoshida2

1Atmosphere and Ocean Research Institute, The University of Tokyo, Kashiwa, Chiba, Japan

2National Research Institute of Fisheries and Environment of Inland Sea, Japan Fisheries Research and Education Agency, Hatsukaichi, Hiroshima, Japan

Abstract

To elucidate the factors determining the organic carbon (OC) sequestration capacity of seagrass meadows,

the distribution of OC and the fraction of seagrass-derived OC in sediments of the temperate cosmopolitan

sea-grass Zostera marina meadows and surrounding habitats were investigated in relation to physical properties of

sedimentary materials On average, seagrass meadow sediments showed OC levels twofold higher than other

shallow nearshore habitats However, offshore sediments often showed greater OC concentrations than average

seagrass meadow sediments According to estimations of OC sources based on carbon isotope ratios, 8–55%

and 14–24% of OC in nonestuarine seagrass meadow sediments and < 30 m deep offshore sediments,

respec-tively, were assigned to seagrass origin The OC concentration in seagrass meadow and offshore sediments

closely correlated to the specific surface area (SSA) of sediment (r250.816 and 0.755, respectively; p < 0.0001),

with an average OC loading per sediment surface area of approximately 60 lmol m22 In seagrass meadow

sedi-ments, the fraction of derived OC was also greater in samples with a larger SSA, and the

seagrass-derived OC occurred preferentially in sediment grains that had a specific gravity exceeding 2.0, namely, in a

form closely associated with sediment minerals The OC concentration, the fraction of seagrass-derived OC,

and the SSA were positively correlated to the logarithm of areal extent of individual seagrass meadows

(p < 0.01) These findings suggest that the OC sequestration capacity of nearshore vegetated habitats is under

the primary control of geophysical constraints such as sediment supply rate and depositional conditions

Vegetated shallow coastal ecosystems, including intertidal

salt marshes, mangroves, and seagrass meadows have been

ranked among the most efficient biotic systems for

accumu-lating organic carbon (OC) on an areal basis (McLeod et al

2011; Fourqurean et al 2012) It is estimated that these

eco-systems may contribute almost half of OC burial in the

glob-al ocean even though they cover < 2% of the ocean surface

(Duarte et al 2005) Recent interest has focused on the

potential to incorporate these ecosystems, called “blue

for-ests,” into policies for reducing carbon dioxide (CO2)

emis-sions At the same time, there is increased concern about the

possibility of CO2 emissions caused by the decline of blue

forest ecosystems, including the seagrass meadows

(Pendle-ton et al 2012; Grimsditch et al 2013)

High rates of OC accumulation in seagrass meadows are likely the result of specific ecosystem functions such as (1) extremely high primary productivity of seagrasses and associ-ated microalgae, (2) efficient trapping of organic particles within the meadow sediment via its flow-regulation and bottom-stabilization effects, and (3) slowness of remineraliza-tion of OC within the meadow sediment due to the anoxic conditions that prevail (Duarte et al 2013) Most of the OC stored in seagrass meadows exists as detrital OC derived from seagrasses and attached algae, seston, and terrestrial organic matter in the underlying sediment (Duarte et al 2013) On average, approximately half of OC stored in sea-grass meadow sediment is derived from the primary produc-tion of seagrasses and seagrass epiphytes, with the rest being derived from allochthonous sources such as phytoplankton and terrestrial organic matter (Kennedy et al 2010; Miyajima

et al 2015) Both the concentration of OC and the fraction

of seagrass-derived OC can vary widely depending on geo-graphical and oceanographic settings and seagrass species composition (Kennedy et al 2004, 2010; Serrano et al 2014; Miyajima et al 2015) However, the mechanisms through which these external conditions control OC sequestration in

*Correspondence: miyajima@aori.u-tokyo.ac.jp

This is an open access article under the terms of the Creative Commons

Attribution-NonCommercial-NoDerivs License, which permits use and

distribution in any medium, provided the original work is properly cited,

the use is non-commercial and no modifications or adaptations are

made.

and

OCEANOGRAPHY V C 2017 The Authors Limnology and Oceanography published by Wiley Periodicals, Inc.Limnol Oceanogr 00, 2017, 00–00

on behalf of Association for the Sciences of Limnology and Oceanography

doi: 10.1002/lno.10478

seagrass meadow sediments remain poorly understood The

lack of this mechanistic understanding has hampered more

precise estimation of the geographical distribution and the

reliable prediction of future trends of OC stocks in seagrass

meadows and other coastal habitats

Seagrass meadows and macroalgal beds also export a large

fraction of their net primary production to the outer oceans

(Duarte and Cebrian 1996; Heck et al 2008) OC is exported

from seagrass meadows mainly through washout of detached

aged leaves under normal growing conditions, seasonal

bio-mass loss (particularly in annual seagrass populations), and

removal of OC stored in the surface sediment by storm surges

and tsunamis The exported OC may be transported and

stored over the long term in offshore sediments under

favor-able depositional conditions (Sugimatsu et al 2015) Once

appropriately quantified, this process could contribute to

car-bon sequestration, in terms of an ecosystem service performed

by seagrass meadows However, reliable information on the

fate of OC exported from seagrass meadows is still lacking

The concentration of OC in coastal marine sediments has

traditionally been considered to be under the control of

sev-eral environmental factors, such as productivity in the

over-lying water column, quality (accessibility and availability to

bacteria) of OC, sediment accumulation rates, capacity of

the sediment mineral matrix to stabilize organic matter, and

the availability of oxygen to benthic heterotrophs (reviewed

by Hedges and Keil 1995) It has been considered that the

strong correlation that is often observed between the OC

concentration and the specific surface area (SSA) of coastal

marine sediments suggests an essential role of physical

sorp-tion of OC in sediment mineral matrices for stabilizasorp-tion

and sequestration of OC (reviewed by Keil and Mayer 2014)

A typical OC loading per surface area (or OC/SSA ratio) of

m22) has been reported for unvegetated coastal and

conti-nental margin sediments (Keil and Hedges 1993; Mayer

1994a) The OC/SSA ratio tends be lower than the typical

range when the supply rate of mineral particles is relatively

high (Mayer 1994b) or the oxygen exposure period of

sedi-ment after deposition is very long (Aller 1998; Hartnett et al

1998) The close association between OC and mineral surface

has been confirmed by electron microscopy (Ransom et al

1997; Bennett et al 1999) and density-fractionation

techni-ques (Bock and Mayer 2000) The key role of sorption in the

stability of OC has also been demonstrated experimentally

(Keil et al 1994b; Zimmerman et al 2004) The empirical

global relationship between OC content and the SSA

indi-cates that the capacity of coastal ecosystems to sequester OC

in the underlying sediment is globally constrained by the

supply rate of the mineral surface available for sorption of

OC Other factors, such as the supply rate of OC and oxygen

exposure period, would be of local or secondary importance

in enhancing and attenuating the OC/SSA ratio

In this study, we compared sedimentary OC stocks in tem-perate cosmopolitan seagrass Zostera marina (eelgrass)

unvegetated tidal flats, macroalgal beds, and shallow offshore sediments The relationships of sediment OC to physical properties, such as SSA and density of sediment materials, were used to characterize the properties of OC stored in differ-ent habitats Using the stable isotope technique, we also examined how the fraction of seagrass-derived OC in the total sediment OC depended on sedimentological factors

Based on the obtained results, we examined the following questions and hypotheses: Is the strong correlation between

OC content and the mineral surface also present in sedi-ments of vegetated ecosystems such as seagrass meadows? Is the average OC loading per surface area in these habitats within the range typical of continental margin sediments? If

so, can we hypothesize that the capacity of OC sequestration

in these habitats is largely determined by external geophysi-cal factors, such as the delivery rate and the granulometric properties of mineral sediment, and that the role of ecosys-tem functions would be of secondary importance? What are the roles of specific ecosystem functions in OC sequestra-tion? To address the latter question, we compared several subtidal sediments collected from seagrass meadows of con-trasting areal extent to demonstrate the role of the ecosys-tem functions of seagrass meadows on the granulometry and

OC of the sediment Finally, using the results from this and other related studies, we discuss the potential of offshore sediment as a remote sink of OC exported from nearshore vegetated habitats

Materials and methods Study sites, sample collection, and initial sample processing

The Seto Inland Sea located in central Japan (Fig 1a) is a shallow water body (average depth, 37 m; surface area, ca 22,000 km2) with abundant seagrass meadows, of mainly Z marina L (Komatsu 1997) It is located in a warm temperate region of the western North Pacific, enclosed by three large islands of Japan, and connected via narrow straits to both the Pacific Ocean and the Sea of Japan The average sea sur-face temperature is between 188C and 208C, with seasonal variations being much larger in the inner sections (10–288C) than at the mouth of the strait (17–258C; Yanagi 1984) The tidal range of the spring tide is 3.5–4.0 m at most of the study sites

We collected surface sediment cores from intertidal and subtidal seagrass meadows (11 sites), bare areas adjacent to seagrass meadows (2 sites), subtidal macroalgal beds (8 sites), and unvegetated estuarine intertidal flats (5 sites) using a hand-operated knocking corer or a piston corer (Adachi et al 2010) during 2009–2012 Collected cores were immediately cut into 10 cm sections, and in this study only the top

30 cm sections were used for further analyses We processed

only one core from each site, except at ZmE2, where three

cores were collected from different microhabitats (seagrass

coverage) For some of the samples, the full-length OC

pro-files were described in Miyajima et al (2015) (Table 1) In

addition, surface sediments were collected from 23 offshore

stations (depth, 9.7–82 m) using a gravity corer during two

cruises of the R/V Shirafuji-maru (Japan Fisheries Research

and Education Agency) in 2009 and 2011 The top 5-cm

sec-tions of all of the offshore cores and subsurface secsec-tions (25–

30 cm) of the 2009 cores were used for further analyses All

samplings were performed in the western part of the Seto

Inland Sea (Fig 1b) Detailed information about the

sam-pling stations is provided in Table 1 Although triplicate

cores were collected and analyzed for the 2009 offshore

sta-tions (Os15–20, Os22, and Os23), data from only one core

from the triplicate are shown here because the difference in

the OC concentration within the site was small compared to

that among sites

All of the sectioned samples were packed in screw-capped

polypropylene containers and frozen at 2208C for temporal

storage and transportation At the laboratory, the samples

were freeze-dried, and the water content was evaluated from

the weight lost on drying The dried samples were gently

crushed by hand using a mortar and pestle, and passed

through a 1 mm mesh stainless sieve to remove large gravels

and seagrass rhizomes that were occasionally found in the

samples (< 20% and < 1% of bulk weight, respectively) The

fraction that passed through the sieve was further homogenized

by grinding in an automatic mortar for 25 min (ALM-200, Nitto Kagaku Ltd., Nagoya, Japan) The influence of grinding on the granulometric properties of samples such as specific surface area and mesopore distribution was insignificant because the area of new surfaces created by grinding was negligible compared to the original surface area of the sediments The final samples were stored in tightly capped glass vials under < 40% relative humidity

Elemental and isotopic compositions The dried and homogenized sediment samples were sub-jected to acid treatment to remove inorganic carbon Approximately 1 g of dried sample was placed into screw-capped glass tubes (10 mL), and 2.0 N hydrochloric acid (HCl) solution was added to the sediment dropwise until all

(without caps) were placed in a vacuum desiccator with a 50-mL beaker containing about 10 g of NaOH pellets as an acid absorber and another beaker containing about 20 mL of concentrated H2SO4 as a desiccant, and kept in vacuo until the sediments were completely dehydrated The NaOH pel-lets were replaced regularly when they became liquefied due

to absorption of HCl and water The drying process normally took 7–10 d The dried sediment was weighed again to check for weight changes due to the acid treatment

The concentrations and isotope ratios of OC and total nitrogen (TN) in the treated samples were determined simul-taneously by EA-IRMS (FLASH 2000/Conflo IV/DELTA V Fig 1.Location of the sampling sites of the surface sediment in the Seto Inland Sea Detailed information can be found in Table 1 [Color figure can

be viewed at wileyonlinelibrary.com]

Table 1. Location of sampling sites and sampling dates of the sediment cores used in this study.

Station name in Miyajima et al

Seagrass meadows (ZmC)

ZmC6 34.29671 132.91541 3.0 06 Oct 2010 Z3 Ikunoshima Island

Estuarine seagrass meadows (ZmE)

ZmE1 34.32424 132.89444 Intertidal 05 Oct 2010 Z1 Hachi tidal flat,

Takehara ZmE2 34.32377 132.89385 Intertidal 19 Jun 2012 Z2 Hachi tidal flat,

Takehara ZmE3 33.60731 131.23482 Intertidal 17 Oct 2009 Zj Nakatsu tidal flat (Z.

japonica bed) Bare areas adjacent to meadow (Ba)

Unvegetated estuarine tidal flat (Tf)

Tf1 34.32536 132.89594 Intertidal 05 Oct 2010 B1 Kamo river mouth,

Takehara Tf2 34.32459 132.89485 Intertidal 05 Oct 2010 B2 Kamo river mouth,

Takehara Tf3 33.60619 131.23761 Intertidal 16 Oct 2009 Bj Yamaguni river mouth,

Nakatsu Tf4 33.62406 131.19647 Intertidal 15 Oct 2009 E1 Yamaguni river mouth,

Nakatsu Tf5 33.62583 131.17417 Intertidal 18 Oct 2009 E2 Yamaguni river mouth,

Nakatsu Macroalgal beds (Mb)

Mb5 33.44535 132.22660 1.5 21 Jun 2012 Ma Sadamisaki Peninsula

Offshore stations (Os)

Advantage, ThermoFisher Scientific, Bremen, Germany).

Three to five standard materials of different d13C (233.8& to

210.2&) and d15N (27.8& to 113.8&) values (SI Science

Ltd., Saitama, Japan, and Iso-Analytical Ltd., Crewe, UK)

were used for daily calibration Samples and standards were

weighed into tin capsules (S€ANTIS Analytical AG, Teufen,

Switzerland) before analysis The measured isotope ratios

were represented using conventional d-notation (d13C and

d15N, in &) with Vienna Pee-Dee Belemnite and atmospheric

N2 as the reference materials The instrumental analytical

precision was normally within 61% for the OC and TN

con-centrations and 6 0.1& for d13C and d15N However, the

sub-sampling errors for both concentrations and d-values were

sometimes twofold to threefold as large as the instrumental

errors

Specific surface area and particle size distribution

Measurement of the SSA of the sediment was performed

by the multipoint Brunauer–Emmett–Teller (BET) method

based on N2 gas adsorption under reduced pressure Using

the same data, the mean diameter of mesopores (MMD) on

sediment grains could also be estimated The dried and

homogenized sediment samples were treated at 3508C for

12 h under normal atmosphere to remove most of the

organ-ic coating (Keil et al 1997) Weight loss on heating was

determined for each individual sample Between 0.5 g and

2.0 g of the treated samples were weighed into glass flasks

specific to the gas adsorption measurement, and desiccated

further in vacuo (< 1022kPa) at 3508C for 3 h Immediately

after cooling, a multipoint BET measurement was performed

with N2 (purity, > 99.9995%) as the adsorbate, using a

BELSORP mini II (MicrotracBEL, Osaka, Japan) surface area analyzer The slope of the BET plot at the inflection point

used for estimating the SSA

The particle size distribution analysis of sediment samples was conducted for several seagrass meadow- and tidal flat sediments by GeoAct, Ltd (Kitami, Japan) using core sam-ples that were collected separately The size distribution was determined by a combination of the standard methods, such

as sedimentation analysis (for  0.075 mm particles) and dry sieving methods (for > 0.075 mm particles)

Density fractionation Selected dried and homogenized sediment samples were subjected to density fractionation by the polytungstate heavy solution method (Sollins et al 2009) A series of heavy solutions (specific gravity of: 1.50, 1.75, 2.0, 2.2, 2.4) were prepared with sodium polytungstate SPT0 (TC-Tungsten Compounds, Grub am Forst, Germany) and ultrapure water The density of the prepared solutions was adjusted using a DMA 35 densitometer (Anton Paar, Graz, Austria) Approxi-mately 2.0 g of the sample was weighed into a 50 mL screw-capped polypropylene centrifuge tube and 25 mL of the lightest heavy solution was added The tube was shaken by a reciprocal shaker at 90 rpm for 30 min to disperse and homogenize the sample, and then centrifuged at 2000 3 g and 208C for 30 min using a swing-bucket rotor The super-natant was filtered through pre-weighed 25 mm glassfiber

Pennsylvania) under reduced pressure The filters were washed three times with 10 mL of ultrapure water and

TABLE 1 Continued

Station name in Miyajima et al

* Directly measured depth for offshore stations; depth below datum level for the other sites.

stored frozen at 2258C until analysis When the volume of

particles separated in the supernatant was so large that more

than two filters were required to collect all the suspended

particles, the pellet was resuspended in 25 mL of the same

heavy solution and the separation process was repeated once

more to ensure recovery of the low-density particles The

pel-let was then resuspended in the second-lightest heavy

solu-tion and homogenized in the shaker This cycle was repeated

using successively heavier heavy solutions, although a

centri-fugation time of 60 min was used for solutions  2.0 g cm23

The wet weight of the pellet confirmed that the amount of

heavy solution carried over to the next step was usually

<5% of the added amount The pellet resulting from the

centrifugation using the heaviest heavy solution was washed

by suspending it twice in 30 mL of ultrapure water and

cen-trifugation, and was then stored frozen at 2258C

The filters and the pellet were freeze-dried and weighed to

determine the weights of the respective density fractions

The samples (including the glassfiber filters) were crushed

and homogenized by the agate mortar and quantitatively

transferred to 10 mL screw-capped glass tubes The contents

were acidified by 5 mL of 1.0 M HCl to remove inorganic

carbon as CO2 The tubes were then centrifuged at 1000 3 g

and at 158C for 30 min The pellets were washed by

suspend-ing them three times in 10 mL of 0.1% NaCl solution in

ultrapure water and centrifugation, and then freeze-dried

and re-weighed The final samples were analyzed for the

con-centrations and the isotopic ratios of OC and TN, as

described above

Data processing

The results of the OC and TN concentrations and the SSA

were expressed as lmol and m2 per unit weight of salt-free

bulk dry sediment The salt content in the sediment sample

was estimated from the water content of the original wet sediment assuming a pore-water density of 1.025 (i.e., a salinity of 35) To calculate the OC stock per unit area of habitat, the dry bulk density estimated following the

meth-od of Miyajima et al (2015) was multiplied by the weight-based concentration determined as above to obtain the con-centration per unit volume of original sediment

Analysis of the sources of organic matter based on the car-bon isotopic composition was performed by the stochastic approach using the IsoSource model developed by Phillips and Gregg (2003) A detailed protocol can be found in Miyajima

et al (2015) In the present study, the following endmember

d13C values were assumed; for seagrass-derived carbon, the average d13C (210.11& 6 0.23&, n 5 5) of Z marina leaves col-lected in seagrass meadows near Sta Mb6, where terrestrial influence was the lowest of all our sites, was used As the

(226.76& 6 1.71&) of soil samples (n 5 3) collected at the Yamaguni River beds (near Sta Tf5) and wood debris (n 5 9) found in the sediment cores collected at Sta Tf1 and Ba1 was used For phytoplankton-derived OC, the asymptotic conver-gence point (221.58& 6 0.11&) of the exponential fitting line for the OC–d13COCplot of offshore sediments (Os1–23; Fig 2a) was assumed to represent the endmember value For the calcu-lation, the source increment and mass balance tolerance parameters were set to be 1% and 0.1&, respectively

Statistical tests (ANOVA, ANCOVA, and multiple regres-sion) and curve fitting (linear, logarithmic, and exponential models) were conducted using the commercial software packages Aabel NG1 (ver 4; Gigawiz, Tulsa, Oklahoma) and pro Fit (ver 7; QuantumSoft, Uetikon am See, Switzerland), respectively The confidence interval of the exponential curve fitting was evaluated by the Monte Carlo method built into pro Fit with an iteration of 1000, assuming appropriate

0 400 800 1200 1600 2000

ZmC

Os

Salt-corrected OC [µmol C g ]

COC

0.06 0.08 0.10 0.12 0.14 2

4 6 8 10

TN/OC ratio [µmol µmol ]

NTN

] ZmC: r = 0.7188, p = 0.0014

Os: r = 0.4384, p = 0.0252 c

Non-estuarine seagrass meadow (ZmC) Estuarine seagrass meadow (ZmE)

Bare areas adjacent to meadow (Ba) Macroalgal bed (Mb)

Offshore sediment (Os)

0 400 800 1200 1600 2000 2

4 6 8 10

Salt-corrected OC [µmol C g ]

NTN

Os ZmC

Fig 2.The concentrations and the isotope ratios of organic carbon (OC) and total nitrogen (TN) in sediments collected from various habitats of the Seto Inland Sea a, b: plots of d13C OC (a) and d15N TN (b) against the concentration of OC c: plot of d15N TN against the atomic ratio of TN/OC Iso-tope ratios described beside the curves in a and b are the convergence values (C) of the exponential functions, d13C OC 5 A exp ([OC] / B) 1 C, for the data plots of non-estuarine seagrass meadow (ZmC, solid circle), offshore (Os, 1), and unvegetated estuarine tidal flat samples (Tf, open triangle) Lines in c are the linear regression lines for the data plots of ZmC and Os samples with the correlation coefficients (r) and the probabilities for the null hypotheses (p) being denoted beside the lines [Color figure can be viewed at wileyonlinelibrary.com]

magnitudes of errors for both independent and dependent

variables (explained in the figure caption)

Satellite image analysis

Of the seagrass meadows studied, we selected eight

subti-dal sites (ZmC in Table 1) where the influence of river-borne

terrestrial inputs was relatively small, and the areal extent of

each seagrass meadow was estimated by satellite image

anal-ysis (Sagawa et al 2008) We used multispectral satellite

images AVNIR2 (10 m resolution) taken from the Advanced

Land Observing Satellite (ALOS1) during the growing season

of Z marina (winter to early summer) of 2010 and 2011

Image processing and analysis were performed using the

geo-graphic information system (GIS) software ArcGIS 10 (ESRI,

Redlands, California) Water bodies deeper than 10 m, where

no seagrasses could grow, were masked based on bathymetric

maps available from the Japan Hydrographic Association

Unmasked shallow coastal areas were classified further into

seven classes: unvegetated sandy or muddy areas; patchy

sea-grass distribution (< 50% coverage); dense seasea-grass

distribu-tion ( 50% coverage); rocky macroalgal beds; land; clouds;

and other water surfaces, by comparison with available

infor-mation, such as direct observation, low-altitude aerial

pho-tos, and acoustic mapping Approximately 30 polygons in

the ALOS images that were known to belong to one of the

seven classes were chosen, and RGB brightness bands in the

polygons were extracted to determine the spectra typical to

each class Next, supervised classification with a maximal

likelihood method was employed to identify areas with

patchy and dense seagrass distribution on the satellite

images The habitat classification obtained by this method

was generally consistent with direct underwater observation

The areal extent of seagrass meadows was determined by

summing the pixel counts (1 pixel 5 100 m2)

Results

Organic carbon and total nitrogen in sediments

The concentration of OC in the sediments tested varied

widely from 35 lmol C g21to 1890 lmol C g21 It was, on

average, more than twofold higher in Z marina meadows

(697 lmol C g21667 lmol C g21, mean 6 SE, n 5 39) and

offshore sediment samples (group Os; 922 6 104, n 5 31)

than in the bare areas adjacent to meadows (Ba; 256 6 70,

n 5 8), the unvegetated intertidal flats (Tf; 193 6 32, n 5 13),

macroalgal beds (Mb; 156 6 23, n 5 23), and a Zostera

japoni-ca meadow (ZmE3; 344 6 64, n 5 3) The OC in either of the

former two habitats was significantly higher than either of

the latter four habitats (p < 0.05; ANOVA with Scheffe’s post

hoc test) There was no significant difference between Z

marina meadows and Os samples, or between the bare area

(Ba, Tf) and Mb samples The OC concentration in sediments

of estuarine seagrass meadows (group ZmE including the Z

japonica meadow, 922 6 85, n 5 15) was significantly higher

than the other Z marina meadows that were distant from

river mouths (group ZmC, 532 6 75, n 5 27; p 5 0.0161) Sim-ilar trends and statistically significant differences were also detected for the concentration of TN, except that the differ-ence between ZmE and ZmC was not significant for TN The OC/TN atomic ratio was significantly lower in the

Os samples (9.3 6 0.2) than in any other habitats (p < 0.05) The ZmE samples showed a significantly higher OC/TN ratio (13.0 6 0.3) than the ZmC (10.9 6 0.3) and Mb (10.5 6 0.4) samples The OC/TN ratio of the bare area sam-ples (Ba, 11.9 6 1.3; Tf, 12.7 6 0.5) ranged in between these values

The stable isotope ratio of sediment OC (d13COC) ranged from 225& to 216& (Fig 2a) For the samples with an

OC < 300 lmol g21, the d13COC varied widely and was not apparently habitat-specific However, the d13COC converged

in habitat-specific ranges with increasing OC concentration

in the ZmC, Os, and Tf groups The convergence values for

225.0& 6 0.46&, which are typical for marine phytoplank-ton and terrestrial organic matter, respectively The conver-gence value for the ZmC group (218.5& 6 0.33&) was significantly higher than those for Os and Tf, but was still

(210.11& 6 0.23&) The d13COC of the Ba and Mb groups showed a decreasing trend with increasing OC, although no clear convergence values could be determined The d13COC

of the ZmE group varied widely, even when the OC concen-tration was > 1000 lmol g21

The fraction of seagrass-derived OC estimated by the Iso-Source model was 8–55% (mean, 31%) and 0–52% (mean, 15%) for ZmC and ZmE habitats, respectively A small contri-bution of seagrass-derived OC was also determined for Os (mean, 19%), Ba (12%), and Tf (9%) habitats The fraction of terrestrial OC was higher for Tf (59%) and Ba (46%) habitats The IsoSource model was not applied to Mb samples because the endmember d13COCof macroalgal OC was not

sufficient-ly constrained from existing data

The stable isotope ratio of TN (d15NTN) varied from 12&

to 110& for the samples with an OC < 300 lmol g21 (Fig 2b) It converged within a relatively narrow range between 15& and 17& as the OC concentration increased In con-trast to d13COC, difference in the convergence value between habitats was not clear (15.5 6 0.48& for Os, 14.6 6 0.79& for Tf, 15.6 6 0.37& for ZmC) In the samples of the ZmC and Os groups, there was a weak but significant positive cor-relation between the d15NTN and the TN/OC ratio (Fig 2c) For reference, the d15N of Z marina leaves collected around Sadamisaki Peninsula (Sta Mbx) and Hiroshima Bay (near Sta Os8) ranged between 13.9& to 15.2& and 17.2& to 18.5&, respectively, and that of macroalgae (Sargassum spp.) collected around Sadamisaki Peninsula was 15.0& 6 1.60& (n 5 13) (unpublished data) The TN/OC ratio of these macrophytes was mostly lower than in the sediment samples (0.033–0.070 for Z marina leaves, 0.046 6 0.017 for Sargassum spp.;

unpublished data) No significant correlation between the

d15NTN and the TN/OC ratio was detected for the groups

ZmE, Mb, Ba, and Tf

Large-scale spatial variations were evaluated for the data

of the Os samples Since the sampling stations were scattered

from east to west (Fig 1) and the water depth varied widely

(Table 1), longitude and water depth were used as spatial

var-iables, and the multiple regression analysis was applied to

the concentrations of OC, the OC/TN ratio, d13COC, and

d15NTN (Table 2) Only the data for the top layer of the core

samples was used here The values for OC, OC/TN, and

d13COC strongly depended on the water depth but not on

the longitude The values for OC and OC/TN were higher

and d13C was more negative at the shallower stations In

contrast, d15NTN was strongly dependent on the longitude

(higher in eastern stations) but did not depend on the water

depth The interaction between water depth and longitude

was statistically insignificant (p > 0.18) for all variables

Specific surface area and grain size distribution

The SSA evaluated by the BET method for Os samples

(20.9 6 1.86 m2g21, n 5 31) was, on average, more than

two-fold higher than any habitats in the shallower nearshore

areas (p < 0.0001) A significant negative correlation was

found between the SSA and water depth of the offshore sites

(r250.4551, p 5 0.0011), and sediments with a relatively

large SSA (> 25 m2g21) occurred only in depth ranges of 9–

30 m (Fig 3a) Of the shallower nearshore habitats, seagrass

9.19 6 0.77, n 5 15 for ZmE) had a larger SSA than

non-meadow sediments (3.02 6 0.49, n 5 8 for Ba; 4.73 6 0.19,

n 5 23 for Mb; 5.21 6 0.36, n 5 13 for Tf), although only the

difference between ZmE and Ba was marginally significant

according to an ANOVA (p 5 0.0441) The difference in the

SSA between seagrass meadow and nonmeadow sediments

was apparently related to the grain size distribution of the

sediment (Fig 4) In particular, the fraction of size range

between 0.002 and 0.1 mm, i.e., silt size class, increased

from < 15% in an unvegetated tidal flat sediment (SSA,

3.97 m2 g21) to > 80% in a Z marina meadow sediment

(15.8 m2 g21) The fraction of clay size class (< 0.002 mm) was also more than twofold higher for the seagrass meadow than the tidal flat sediments In contrast, fine to medium sand fraction (0.1–0.5 mm) dominated in the tidal flat and

Z japonica meadow sediments (Fig 4)

The mean mesopore diameter (MMD) of sediment grains estimated by the BET method was similar among the sedi-ments of the seagrass meadows and adjacent bare areas (Fig 3b) with no significant difference between these groups (p > 0.277 by ANOVA) However, the MMD for Mb, Tf, and

Os was significantly smaller than that of ZmC, ZmE, and Ba (p < 0.002) Among the Os samples, those collected from depths of > 30 m commonly showed a MMD < 14 nm How-ever, Os samples collected from a depth of < 30 m exhibited 14–18 nm MMD, except for Os23 (10.6 nm), which over-lapped with the range for seagrass meadow sediments (Fig 3b) There was no significant relationship between MMD and SSA in any of these habitats

The OC concentration showed a linear correlation to SSA for both ZmC and Os (p < 0.0001; Fig 3c) The average OC loading per measured surface area, as evaluated from the slope of the linear regression model, was 61.5 6 5.5 and

m22) for ZmC and Os, respectively The difference of the slope was not significant (p 5 0.557 by ANCOVA) though the offset was significantly higher for ZmC than for Os (p 5 0.0010) OC was dependent on SSA also for Mb (100.4 6 14.6 lmol C m22; p 5 0.0010), while the range of variation of SSA was very small (3.4–6.5 m2 g21; Fig 3b) Although the OC loading per surface area was apparently high for Mb compared to Os and ZmC, differences in the slope between Mb and Os and between Mb and ZmC were not significant (p 5 0.294 and 0.092, respectively) No signifi-cant linear correlation was detected between OC and SSA for the samples of ZmE, Ba, or Tf (p 5 0.129, 0.051, and 0.177, respectively) The sediment samples with SSA < 10 m2 g21 showed highly variable OC/TN ratios from < 7 to 16

Howev-er, the OC/TN ratio of seagrass meadow sediments (ZmC and ZmE) and Os converged within a narrow range around 9 as the SSA increased beyond 12 m2g21(Fig 3d)

Table 2. Dependence of OC, the OC/TN ratio, d13COC, and d15NTN of Os samples on water depth and longitude of sampling sites (n 5 23) evaluated by multiple regression analysis

Predictors*

OC/TN [mol/mol] 20.037 <0.0001 20.175 0.2398 0.669 <0.0001

*Depth and longitude of the sampling sites were not significantly correlated (p 5 0.2193).

Fig 3.Relationship of the specific surface area (SSA) of offshore sediment samples and the water depth from which the samples were collected (a), mean mesopore diameter (MMD) with standard deviation of sediment mineral particles from each habitat (b), and plots of the OC concentration (c) and the OC/TN atomic ratio (d) against SSA Lines in a and c are linear regression lines for offshore (Os, 1) and non-estuarine seagrass meadow (ZmC, solid circle) samples, with the correlation coefficients (r) and the probabilities for the null hypotheses (p) being denoted beside the lines In b, the MMD of Os samples was averaged separately for shallow (< 30 m) and deep (> 30 m) sediments [Color figure can be viewed at wileyonlineli-brary.com]

Fig 4.Particle size distribution of surface sediment samples collected from four different habitats The specific surface area (SSA) of each sample is described in the legend Inset: plot of SSA against weight-average grain size Data were fitted to a 2/3-power-law curve (dashed line): SSA 5 1.87 (average grain size)22/3 [Color figure can be viewed at wileyonlinelibrary.com]

algal bed (j), unvegetated tidal flat (k, l), and offshore sites (m – p) The density fraction is denoted below the abscissa (g cm23) OC is represented as a concentration per unit weight of bulk sediment (i.e., before density separation) [Color figure can be viewed at wileyonlinelibrary.com]