The aim of this work was to explore the feasibility of using plant functional
traits to identify differences in sediment organic carbon (OC) storage within
seagrass meadows. At 19 sites within three seagrass meadows in the coastal
waters of Zanzibar, Tanzania, species cover was estimated along with three
community traits hypothesized to influence sediment OC storage (above and
belowground biomass, seagrass tissue nitrogen content, and shoot density).
Sediments within four biogeographic zones (fore reef, reef flat, tidal
channel, and seagrass meadow) of the landscape were characterized, and
sediment cores were collected within seagrass meadows to quantify OC storage
in the top 25 cm and top meter of the sediment. We identified five distinct
seagrass communities that had notable differences in the plant traits, which
were all residing within a thin veneer (ranging from 19 to 78 cm thick) of
poorly sorted, medium to coarsely grained carbonate sands on top of carbonate
rock. One community (B), dominated by
Seagrasses influence key ecological functions within coastal ecosystems through their productivity and by trapping sediment, altering hydrodynamics, and modifying biogeochemical processes in the water column and sediment (Duarte and Chiscano, 1999; Marba et al., 2006). Through their effects on ecosystem processes, seagrasses provide numerous ecosystem services including sediment stabilization, coastline protection, nutrient cycling, pathogen reduction, support of fisheries, and enhancement of biodiversity (Duffy, 2006; de la Torre Castro and Rönnbäck, 2004; Lamb et al., 2017; Orth et al., 2006). In the last 2 decades, seagrasses have been recognized as important “blue” carbon (organic carbon sequestered by vegetated coastal ecosystems) sinks, adding climate regulation to their list of well-established ecosystem services (Duarte et al., 2005; Fourqurean et al., 2012a; Macreadie et al., 2014; Mateo et al., 1997; Nellemann et al., 2009; Pergent et al., 1994; Romero et al., 1994).
A surge in research efforts has revealed the wide range (up to 18-fold) of sediment organic carbon (OC) storage within seagrass sediments, with OC stocks varying with seagrass species (Gullström et al., 2017; Lavery et al., 2013; Serrano et al., 2014, 2016a), plant characteristics (Dahl et al., 2016; Jankowska et al., 2016; Samper-Villarreal et al., 2016), meadow attributes (Armitage and Fourqurean, 2016; Samper-Villarreal et al., 2016; Serrano et al., 2014, 2016b), sediment characteristics (Campbell et al., 2014; Dahl et al., 2016; Miyajima et al., 2017; Röhr et al., 2016; Serrano et al., 2016a), landscape configurations (Campbell et al., 2014; Gullström et al., 2017; Lavery et al., 2013; Phang et al., 2015), and climatic zones (Fourqurean et al., 2012a; Lavery et al., 2013; Miyajima et al., 2015). The potential for high variability in OC stocks presents a formidable obstacle for reliably valuing the ecosystem service of OC storage because baseline stock estimates are needed before conservation or restoration can be incentivized under a blue carbon framework (Barbier et al., 2011; Costanza et al., 1997; 2014; Herr et al., 2012; Macreadie et al., 2014). To achieve IPCC tier 3 standards of accuracy for OC stock inventories, considerable sampling effort is required (Howard et al., 2014; Macreadie et al., 2014).
A potential solution is to utilize easy-to-measure functional traits that can be linked to ecosystem functions underlying the service of OC storage as a proxy for sediment OC content (de Bello et al., 2010; de Chazal et al., 2008; Grime, 2001; Kremen, 2005). Plant functional traits have been shown to be valuable tools for assessing and managing ecosystem services (de Bello et al., 2010; Díaz et al., 2007). An important trade-off of plant traits linked to OC cycling, known as the “fast–slow plant economic spectrum”, consists of a suite of coordinated characteristics that either promote fast carbon acquisition and decomposition or promote the conservation of resources within well-protected tissues with inherently slower decomposition rates (Conti and Díaz, 2012; Díaz et al., 2004; Freschet et al., 2012; Grime, 2001; Reich et al., 1997; Wright et al., 2004). Acquisition traits such as high specific leaf area, high nutrient content, and low tissue longevity and density are consistently associated with high OC inputs via photosynthesis and high OC losses through decomposition (Cornwell et al., 2008; Díaz et al., 2004; Grime et al., 1997; Herms and Mattson, 1992; Reich, 2014; Wright et al., 2004). Conservative traits include the opposite of the above characteristics and promote slow-growing, long-lived biomass with low OC losses via decomposition. At the ecosystem level, acquisition traits promote high carbon fluxes, while conservation traits are conducive for the retention of OC stocks (Conti and Díaz, 2012; De Deyn et al., 2008; Díaz et al., 2009; Wardle et al., 2004).
Several seagrass traits have been proposed to be influential on OC
sequestration and storage within seagrass sediments (Duarte et al., 2011).
Canopy characteristics, such as high leaf density and complexity, have been
shown to enhance the OC burial flux into the sediment by filtering and
trapping particles from the water column and promoting sediment deposition
and retention (Peterson et al., 2004; Duarte and Chiscano, 1999; Duarte et
al., 2005; Gacia and Duarte, 2001; Gacia et al., 1999; Hendriks et al., 2008;
Samper-Villarreal et al., 2016). Furthermore, seagrass sediment OC stocks have been
positively correlated with shoot density, both directly (Dahl et al., 2016)
and indirectly (Serrano et al., 2014). Seagrass tissue stoichiometry has been
correlated with decomposition rates, with tissues containing relatively
higher nitrogen and phosphorus content decomposing faster (Enriquez et al.,
1993), at least in the initial phase of decomposition (Berg and McClaugherty,
2014). Low concentrations of nitrogen (C : N ratio above 20–25) within
tissues indicate the potential for microbial nitrogen limitation,
necessitating nitrogen immobilization from the environment and resulting in
low carbon-use efficiency during litter decomposition (Berg and McClaugherty,
2014; Hessen et al., 2004; Sinsabaugh et al., 2013). Furthermore, the
nutrient content of tissues covaries with other structural and chemical
properties that reflect the plant species' ecological strategy and can serve
as a proxy of tissue quality and decomposability (Birouste et al., 2012;
Cornwell et al., 2008; Freschet et al., 2012; Zechmeister-Boltenstern et al.,
2015). Seagrass biomass has been positively correlated to OC storage
(Armitage and Fourqurean, 2016; Gullström et al., 2017; Serrano et al.,
2016a). Belowground (BG) production of seagrass roots and rhizomes places OC
directly into sediments, which can be stabilized on mineral surfaces, within
aggregates, or if microbial activity is suppressed due to lack of oxygen
(Belshe et al., 2017; Duarte et al., 2011). In addition, the binding of the
sediment by the root–rhizome system (Christianen et al., 2013) and the high
lignin content of BG tissues (Klap et al., 2000) can promote OC
storage. Larger plants disproportionately contribute to OC accumulation
by shedding more biomass per unit of ground area (Garnier et al., 2004; Lavorel
and Grigulis, 2011). Seagrass interspecies variation in these traits places
them within the continuum of the fast–slow plant economic spectrum, with
small-bodied, ephemeral species, such as
The aim of this study was to determine whether seagrass community traits can be linked to differences in sediment OC content within meadows residing in the open coastal waters of Zanzibar, Tanzania. Our sites were located within three meadows that contained up to eight co-occurring seagrass species, with a wide breadth of functional traits (Gullström et al., 2002), all residing within a landscape with similar abiotic conditions (Shaghude et al., 2002). Our goal was to add to the growing body of evidence investigating where, and to what extent, plant community traits can be used to determine the size of and variability in OC storage within seagrass sediments.
Study sites were located within three meadows (M1, M2, M3) in open
coastal waters adjacent to coral cays west of the main city, Zanzibar City,
Unguja Island (
This study was conducted within the tropical Indo-Pacific bioregion (Short et
al., 2007) in diverse seagrass meadows of the western Indian Ocean (WIO),
specifically in coastal waters of Unguja Island (
Sample sites were established within three seagrass meadows (M1, M2, M3) in open coastal waters adjacent to coral cays west of the main city, Zanzibar City (Fig. 1). These meadows were chosen because they contained a range of seagrass species (up to eight species) with different life-history strategies, and at the same time had similar landscape positions and abiotic properties, such as shallow water depth, carbonate sediments, and negligible terrestrial inputs (Shaghude et al., 2002). M1 is located in shallow waters (70–380 cm in depth) to the southeast of Kibandiko Island and encompasses an area of 15 ha, which includes several small intermittent patch reefs. M2 is also located 1.5 km to the west of M1, and encompasses an area of 4.8 ha. M2 resides within a shallow lagoon (50–320 cm in depth) adjacent to a sand spit and fringing reef on the northeastern side Changu Island. M3 covers 4.6 ha and is located in shallow waters (50–375 cm in depth) north of Chumbe Island, adjacent to patch reefs and a sand spit. M3 resides 16 and 17 km south of M1 and M2, respectively. The seagrass within these meadows grows within a shallow sediment layer on top of an uplifted Pleistocene carbonate platform (Kent et al., 1971; Short et al., 2007). The sediments are biogenic, with the major constituents being benthic foraminifera, molluscs (pelecypods and gastropods), and coral, with negligible terrigenous inputs (Gita R. Narayan, unpublished; Shaghude et al., 2002).
During October 2013, physical properties (temperature, pH, dissolved oxygen,
and conductivity) of the water column were measured using a WTW 3430
multiparameter probe (Weilheim, Germany) within the three meadows. Light
levels at the surface (
Between 17 September and 17 October 2013, 19 sample sites were established
across the three meadows (M1, M2, M3) to capture the zonation of species
assemblages found across the extent of each meadow. A snorkeling survey was
conducted at each meadow, consisting of five 50 m transects (perpendicular
to the coastline) throughout each meadow. Based on this initial survey, six
to seven distinct vegetation zones were identified for each meadow. The
pattern of zonation within the meadows was a mosaic of patches, following
both the depth gradient and running parallel to the coastline. Within each
zone, a 0.25 m
At each site, seagrass species composition was quantified within six
randomly tossed 0.25 m
To quantify traits of each seagrass community, three biomass cores and five
ramets of each seagrass species present were collected at each of the
19 sites. Biomass cores encompassing both seagrass shoots and the entire
rhizosphere (ranging from 10 to 30 cm of depth) were taken by placing a 13 cm
diameter PVC ring on top of the sediment and using a knife and garden trowel
to remove all plant biomass within the ring. This methodology was utilized
because of the coarse, shallow carbonate sediments at our sites. Plant
material was placed directly into a mesh bag (2 mm mesh size), rinsed free
of sediment in the field, stored in plastic bags, and frozen for subsequent
analysis. After thawing in the lab, seagrasses were sorted by species, and
short-shoot density (shoot m
To determine if carbon storage within sediments varied among different seagrass communities, sediment cores were taken within the five seagrass communities (determined from the multivariate analysis) during October of 2014. Three sediment cores were taken with a hand-driven, 7.6 cm internal diameter corer on a SCUBA, within each of the five identified seagrass communities and on bare sediment adjacent to, but outside of, the seagrass meadows. Within each community, cores were distributed among the three meadows, resulting in one core extracted per community per meadow. Due to the shallow and variable sediment accumulation on top of the carbonate platform at our sites, the depth of penetration of sediment cores varied from 19 to 78 cm. The presence of the impenetrable carbonate layer was verified manually after the core was extracted by hand or by inserting a metal rod. Core compaction was not measured in this study but was assumed to be minimal due to the coarse sediment composition. We also assumed that there were no historic differences in community composition, plant traits, or meadow extent during past carbon deposition because there were no historic data available at our sites, which is a limitation of this study.
In the lab, cores were sectioned into 3 cm slices. From each slice, a
15 cm
To characterize the sediments of the four landscape zones (fore reef, reef
flat, tidal channel, seagrass), we applied the Udden–Wentworth scale
(Wentworth, 1922) as follows: gravel (> 2000
Multivariate analyses were used to describe and categorize the patterns in seagrass species assemblages found at the 19 sample sites. First, Braun-Blanquet cover categories were converted to the midpoint of the cover range (Wilkum and Shanholtzer, 1978) and then square-root transformed to down-weight the influence of abundant species, and finally cover estimates were relativized to the total abundance of each site. A Bray–Curtis similarity index was then calculated based on the similarity of species composition and cover among sites (Bray and Curtis, 1957). Then, based on this similarity matrix, both nonmetric multidimensional scaling (NMDS) and hierarchical cluster analysis (average linkage) were preformed to group sites by similarity in seagrass species composition and cover (ter Braak, 1995; Kent and Coker, 1992; Legendre and Legendre, 1998). These categorizations were used to identify the seagrass species assemblages (communities) present in the sampled meadows. The vegan package (version 2.2-0; (Oksanen et al., 2014) in R (R Core Development Team, 2016) was used for all multivariate analysis.
For all analysis of trait differences among communities, the unbalanced sample design created from the unequal grouping of the original 19 sites into communities (based on similarity of species and cover) necessitated special attention in regards to model appropriateness and validation of assumptions. Model residuals were tested for homogeneity of variance and normality with Levene's and Shapiro–Wilk tests, respectively. Selected models were also validated visually with plots of model residuals (fitted values vs. absolute residuals (homogeneity of variance), a qqplot comparing the distribution of the standardized residuals to the normal distribution (normality), and a lag plot of the raw residuals vs. the previous residual (independence; Zuur et al., 2009). Further, spatial independence was confirmed with variogram plots of model residuals using the gstat package (Zuur et al., 2009).
Differences in AG and BG biomass among communities and meadows
were determined using a two-way ANOVA with post hoc Tukey HSD at the
Differences in short-shoot density among the seagrass communities and meadows
were determined using a generalized linear model, specifically a negative
binomial model (link
The %N of each community was estimated by calculating the mean and
standard deviation of the %N weighted by the abundance of each species
present within the community specific to each meadow. Because of our unequal
sample sizes and variance heterogeneity, communities and meadows within a
community were simply compared visually and considered different when there
was no overlap between 95 % confidence intervals, which were calculated
as the weighted mean
To explore how OC varied among communities and across our sites, models that
included both community and meadow as direct effects were evaluated. Because
of our relatively small sample size (
Landscape sediment characterization of the four biogeographic zones. Summary statistics (mean, standard deviation, skewness, kurtosis) are based on the logarithmic Folk and Ward method and are shown on both micrometer and phi scales. The distribution spread (D10–D90) of grain sizes is calculated as the difference between the grain size (D10) at which 10 % of grains are coarser and the grain size (D90) at which 90 % of the grains are coarser. D50 is the median grain size.
Physical properties of the water column were similar among meadows, with pH
ranging from 8.19 to 8.31 (
Average percentage of sediments within each sediment grain size class (top left) for the four biogeographic zones (tidal channel, fore reef, reef flat, and seagrass meadow) found at our study sites. Within each zone, the grain size distributions of each sample are also shown (bottom and left). Grain size classes are based on the Udden–Wentworth grade scale.
Five distinct seagrass species assemblages were identified, hereafter
referred to as communities A, B, C, D, and E (Fig. 3). The first two
communities, A and B, are monospecific, composed 100 % of
Seagrass communities were determined by grouping the 19 sites from
three meadows (M1, M2, M3) based on their similarity in seagrass species
composition and cover using two methods. First,
There was a significant effect of both community (
Mean seagrass above- (AG) and belowground (BG) biomass
(g DW m
BG biomass followed a similar pattern with significant effects
of community (
Meadow-specific estimated mean (
There was a significant effect of community on seagrass short-shoot density
(
Percent nitrogen (N) in rhizome
The nitrogen content within seagrass leaves varied among seagrass communities
(determined by no overlap in 95 % CI), with community D having the
highest percentage of nitrogen (M1:
Percent organic carbon at different depths (cm) down each sediment core taken within the five seagrass communities (A–E) and bare sediment (F).
The nitrogen content within seagrass rhizomes did not significantly vary
among communities or meadows, with the weighted mean percentage of nitrogen in all
communities across all meadows ranging from 0.42 % to 0.67 % (Fig. 6)
and rhizome C : N ratios ranging from 78 to 97. However, within community D
there was notably higher variability in rhizome nitrogen content within M2,
which had the highest tissue %N (0.92 %) due to the high relative
abundance (74 %) of
Organic carbon storage of
The percentage of OC within the sediment was low within all communities
(A–E), varying from a maximum of 0.75 % in surface sediments to a
minimum of 0.15 % down core (Fig. 7). There were no differences in %OC
in the top 25 cm (where all cores had data) among seagrass communities
(A–E;
The OC stored within the top 25 cm of sediment did not differ among seagrass
communities (A–E,
In three seagrass meadows off the coast of Zanzibar City, Tanzania, we
identified five distinct seagrass communities. Even with the natural
variation across meadows there were still notable differences among
communities in key plant traits shown to influence ecological processes
linked to OC sequestration and storage in other ecosystems (Aerts and Chapin,
2000; Chapin, 2003; Díaz et al., 2004). However, these trait differences
did not translate into differences in sediment OC stocks among seagrass
communities. The OC storage in the top 25 cm (
A clear contrast emerges when comparing OC storage within community B,
dominated by the large-bodied, persistent species
On the one hand, water flow at our sites is energetic with moderate to high
current velocities (ranging from 0.25 to 2 ms
These results fit within the emerging framework that the stabilization of OC within soils and sediments is a whole-ecosystem property (Lehmann and Kleber, 2015; Schmidt et al., 2011). This view posits that all organic matter can decay quickly if conditions are right (Gramss et al., 1999; Hamer et al., 2004; Hazen et al., 2010; Wiesenberg et al., 2004). However, decomposition can be altered by ecosystem properties that impede the microbial access to, or remineralization of, certain molecules (Lehmann and Kleber, 2015; Schmidt et al., 2011). For example, in sediments with low oxygen concentrations, the decomposition of complex, recalcitrant OC can be impeded (due to a lack of electron acceptors or enzyme cofactors that require oxygen), resulting in the selective preservation of oxygen-sensitive OC (Arnarson and Keil, 2007; Burdige, 2007; Burdige and Lerman, 2006; Hedges and Keil, 1995; 1999; Keil and Mayer, 2014). Likewise, sediment mineralogy and aggregation can reduce the bioavailability and accessibility of OC to microbes and enzymes (Arnarson and Keil, 2001; 2007; Hedges and Keil, 1999; Keil and Mayer, 2014; Mikutta et al., 2006; Schrumpf et al., 2013; Six et al., 2004; 1998; Sollins et al., 1996; Tisdall and Oades, 1982). Alternatively, in the absence of ecosystem controls, even low-quality, chemically complex compounds such as lignin can be degraded relatively quickly (Dittmar and Lara, 2001). This view shifts plant input quality into an auxiliary role, with the persistence of sediment OC ultimately determined by geophysical properties of the sediment.
The modulation of the role of plant traits in OC storage by sediment
properties is seen when comparing our sites on the western coast of Unguja
Island, Zanzibar, to meadows located on the south and east coasts of the
island. At these other locations, sediment OC storage is 2 to 3 times
higher than what was measured at our sites (40.7 to 73.8 Mg C ha
However, this study does add a key piece to the growing body of evidence showing
that geophysical conditions of the sediment modulate the importance of plant
traits in regards to retention of OC within blue carbon ecosystems (Alongi et
al., 2016; Armitage and Fourqurean, 2016; Campbell et al., 2014; Dahl et al.,
2016; Miyajima et al., 2017; Röhr et al., 2016; Samper-Villarreal et al.,
2016; Serrano et al., 2016a). Here we show that once sediments become very
coarse and shallow, large inputs of low-quality seagrass OC are not
necessarily stabilized against microbial decay. This extends and contrasts
previous work from sites without high sediment loading and fine sediments,
which show plant traits (biomass, density, and cover) became better
predictors for OC storage as sediments become more coarse (Dahl et al.,
2016). This increase in explanatory power by plant characteristics as
sediments become coarser was also shown for large-bodied, persistent species
(
This study, placed into the context of the growing body of evidence of the large variation in OC storage in seagrass ecosystems (Campbell et al., 2014; Dahl et al., 2016; Lavery et al., 2013; Miyajima et al., 2015; Röhr et al., 2016; Samper-Villarreal et al., 2016; Serrano et al., 2014, 2016a, b), illustrates the complexity of controls and mechanisms that govern OC storage in seagrass sediments. Even within meadows with similar environmental conditions, data on plant traits or carbon sources (as a proxy for OC input quality) cannot alone provide a full picture of the location or magnitude of sediment OC; therefore, we caution against their singular use as proxies for OC storage. Future efforts should focus on quantifying the interactions among properties of OC inputs (quantity and quality) and a suite of geophysical sediment properties, including mineralogy, structure, and the full range of the grain size distribution. Once these interactions can be quantified, spatial information on sediment parent material (Hartmann and Moosdorf, 2012) and composition can be integrated with data on seagrass characteristics and extent to better model the spatial variability in OC storage within seagrass sediments.
In this study, we were unable to link variations in plant traits to differences in sediment OC stocks within diverse seagrass meadows off the coast of Zanzibar City, Tanzania. Geophysical constraints of the sediment outweighed any effects of trait differences on OC stabilization and resulted in low OC storage across all seagrass communities despite high inputs of low-quality OC within some communities. In spite of being constrained within the particular environment, seagrasses still managed to store twice as much OC as bare sediment. This highlights the importance of seagrass habitats for OC cycling in coastal marine ecosystems; however, further research is needed to identify under which geophysical conditions seagrass traits can be linked to the ecosystem function of OC storage.
The data are available from the first author upon request.
The supplement related to this article is available online at:
All authors contributed to this paper; specifically, EFB contributed to the design, data acquisition, analysis, and interpretation and wrote the first draft of the paper. DH contributed to the design, data acquisition, and analysis and provided a critical review of the paper. NH contributed to the design, data acquisition, and analysis and provided a critical review of the paper. MM contributed to the design and data acquisition and critically revised the paper. MT contributed to the design, and data interpretation and critically revised the paper. All authors read and approved the final version of the paper.
The authors declare that they have no conflict of interest.
This project was carried out within the framework of and through funding provided by the Leibniz Graduate School SUTAS (Sustainable Use of Tropical Aquatic Systems; SAW-2013-ZMT-4) and the Leibniz Centre for Tropical Marine Research (ZMT) based in Bremen, Germany. Elizabeth Fay Belshe was supported by the German Academic Exchange Service (DAAD) with funds from the German Federal Ministry of Education and Research (BMBF) and the People Programme (Marie Curie Actions) of the European Union's Seventh Framework Programme (FP7/2007-2013) under REA grant agreement no. 605728 (P.R.I.M.E. – Postdoctoral Researchers International Mobility Experience). Dieuwke Hoeijmakers and Natalia Herran were supported by SUTAS and Mirta Teichberg was supported by the German Research Foundation (DFG) within the Individual Grants Program, project SEAMAC, TE 1046/3-1. We thank the following people and organizations that supported this work: Dr. Christopher Muhando and Dr. Narriman S. Jiddawi at the Institute of Marine Sciences (IMS), University of Dar es Saalam, and Uli Kloiber and team from the Chumbe Island Coral Park Ltd. (CHICOP). Lastly, we would like to thank Dr. Benjamin Bolker (McMaster University, Canada) for support and advice on modeling procedures and statistical analysis. Edited by: Peter van Bodegom Reviewed by: two anonymous referees