Carbon stocks and accumulation rates in Red Sea seagrass meadows

Seagrasses play an important role in climate change mitigation and adaptation, acting as natural CO2 sinks and buffering the impacts of rising sea level. However, global estimates of organic carbon (Corg) stocks, accumulation rates and seafloor elevation rates in seagrasses are limited to a few regions, thus potentially biasing global estimates. Here we assessed the extent of soil Corg stocks and accumulation rates in seagrass meadows (Thalassia hemprichii, Enhalus acoroides, Halophila stipulacea, Thalassodendrum ciliatum and Halodule uninervis) from Saudi Arabia. We estimated that seagrasses store 3.4 ± 0.3 kg Corg m−2 in 1 m-thick soil deposits, accumulated at 6.8 ± 1.7 g Corg m−2 yr−1 over the last 500 to 2,000 years. The extreme conditions in the Red Sea, such as nutrient limitation reducing seagrass growth rates and high temperature increasing soil respiration rates, may explain their relative low Corg storage compared to temperate meadows. Differences in soil Corg storage among habitats (i.e. location and species composition) are mainly related to the contribution of seagrass detritus to the soil Corg pool, fluxes of Corg from adjacent mangrove and tidal marsh ecosystems into seagrass meadows, and the amount of fine sediment particles. Seagrasses sequester annually around 0.8% of CO2 emissions from fossil-fuels by Saudi Arabia, while buffering the impacts of sea level rise. This study contributes data from understudied regions to a growing dataset on seagrass carbon stocks and sequestration rates and further evidences that even small seagrass species store Corg in coastal areas.

even these small seagrasses have been documented to significantly enhance sediment stabilization and accumulation 19 and potentially contributing to C org storage 10,20 .Given that there are over 370 km 2 of seagrasses in Saudi Arabia 18,21 , there is potential for these meadows to contain substantial stores of C org .This study aims to provide estimates of soil C org stocks and accumulation rates in seagrass meadows from the Red Sea, thus contributing to achieve a more balanced representation of variation among seagrass habitats by considering underrepresented regions 10 , and to place our results within a larger context by comparing these data to estimates from global datasets.
We combine estimates of soil C org density down to 1 m depth with soil chronologies derived from 14 C age dating to estimate (a) C org stocks within the top meter of soil, and (b) the accumulation rate of C org over the last millennia.We also estimate the contribution of autochthonous and allochthonous sources to the seagrass soil C org pool and determine soil grain-size composition to examine their relationships with soil C org pools.
The δ 13 C values of sedimentary organic matter in seagrass soils averaged −16 ± 0.2‰ (Table 1).The δ 13 C values of potential organic sources into seagrass soils collected at the four study sites are presented in Table 3.The mixing models applied indicated that seagrass detritus was the most important source of soil C org in Red Sea seagrass meadows (41%), while mangrove plus halophytes and seaweed plus seston inputs were less important (32 and 27%, respectively; Fig. 1

and Supporting Information Table A).
Soil biogeochemical characteristics (dry bulk density, C org content (in % and mg cm −3 ), δ 13 C of soil organic matter, and sediment grain-size composition) differed significantly among study sites and meadows with distinct species composition (P < 0.05; Table 4).Soil accretion rates (cm yr −1 ) and C org accumulation rates (g m −2 yr −1 ) also differed significantly among locations (P < 0.01; Fig. 2A) but did not differ among the seagrass species tested (P > 0.05; Fig. 2B), while C org stocks (kg m −2 in 1 m-thick soils) did not differ either among locations or meadows with distinctive species composition (P > 0.05; Table 4; Fig. 2A,B).
Seagrass soils at Thuwal Island had higher amounts of fine soil particles (78% of clay and silt and very fine sands) compared to the other locations (ranging from 51 to 63%; Fig. 2A).Meadows at Economic City had a relatively larger proportion of coarse sands (25%) compared to the other locations (ranging from 1 to 14%).Meadows at Thuwal Island and Petro Rabigh were 13 C-depleted (averaging −17.8 and −22.7‰, respectively) compared to the other locations (δ 13 C ranging from −12.7 to −15.4‰; Fig. 2A).The mixing models applied indicated that seagrass detritus was a relevant source of soil C org in meadows at Economic City (51%) and Khor Alkharar (45%), but a relatively minor contributor at the rest of locations (ranging from 12 to 31%; Fig. 1A).Mangrove plus halophyte constituted the main soil C org sources in Petro Rabigh meadows (73%) and Thuwal Island (38%), while the contribution of seston ranged between 15 and 32% among the study sites.
Meadows formed by T. hemprichii, E. acoroides and H. uninervis were 13 C-enriched (ranging from −11.2 to −14‰) compared to H. stipulacea (−16.0‰) and T. ciliatum (−20.0‰)meadows (Fig. 2B).The trends in δ 13 C values with soil depth remained stable in T. hemprichii, E. acoroides and H. uninervis meadows (Supporting Information Fig. A), indicating that either the organic matter decomposition in the soils or the inputs of organic matter remained stable.However, the δ 13 C values become more negative with depth/ageing in H, stipulacea and T. ciliatum meadows, in particular below ~60 cm soil depth.The 13 C-depletion in soil organic matter could either indicate the lack of seagrass matter inputs below cm ~60 in the cores or the decomposition of carbohydrates and proteins with ageing, which have more positive δ 13 C values than the remaining organic matter (e.g.lignin 22 ).In order to constrain the uncertainties mentioned above, we run the mixing models using average δ 13 C values within the top 60 cm of the cores only.Despite the δ 13 C values only remained stable within the top 60 cm in E. acoroides (R 2 = 0.05) and H. stipulacea (R 2 = 0.14) meadows (i.e. the δ 13 C values significantly increased with soil depth in T. hemprichii meadows (R 2 = 0.60) and significantly decreased with soil depth in H. univervis (R 2 = 0.30) and T. ciliatum (R 2 = 0.34)), their variability is relatively small within the top 60 cm of soil compared to 1 m-thick soils,  Continued thereby constraining the uncertainties associated with diagenetic effects.The mixing models applied indicated that seagrass detritus was a significant source of soil C org in all meadows studied: T. hemprichii (58%), E. acoroides (59%), H. uninervis meadows (45%) and H. stipulacea (40%), except T. ciliatum (21%; Fig. 1B).Mangrove plus halophyte constituted the main soil C org sources in T. ciliatum meadows (54%), while the contribution of seston ranged from 24 to 29% in the meadows among meadows with different species composition.

Discussion
Seagrass meadows represent one of the most important vegetated communities in the otherwise arid and oligotrophic Red Sea region.Seagrasses are widely distributed along the Kingdom of Saudi Arabia Red Sea coast 18 , and the soils underneath those seagrass meadows contain considerable C org stocks.).The relatively low C org sink capacity of Red Sea seagrasses could be due to the extreme environmental conditions such as nutrient limitation and high temperature, reducing the growth rates of the seagrasses and increasing the rate of respiration in the soil 21,23 .This disconnection between Red Sea seagrass C org stocks and the global estimates are likely linked to the very limited data set used to produce global estimates, which was biased by the extremely high C org content of soils from Mediterranean Posidonia oceanica meadows 6 .Recent estimates of soil C org stocks in low biomass seagrass species from Abu Dhabi (H.uninervis, Halophila ovalis and H. stipulacea; ranging from 0.2 to 10.9 kg C org m −2 24 ), Asia (Zostera spp.and T. hemprichii; ranging from 3.8 to 8.9 kg C org m −2 20 ), and Australia (H. uninervis and T. hemprichii; ranging from 2.3 to 5.0 kg C org m −2 10 ) are within the range of C org stocks estimated for Red Sea seagrasses.Hence, it is likely that C org stocks in Red Sea seagrass soils tend to be in the lower range but are not necessarily below those in seagrass soils in all other meadows, suggesting a need to update the global estimate of C org content in seagrass soils using a more balanced geographical distribution of seagrass meadows, but also accounting for habitat variability (i.e.diversity of morphological traits across species and geomorphology).
Moreover, long-term C org accumulation rates in Red Sea seagrass (7 g C org m −2 yr −1 on average) are lower than previous estimates for large and long-living Posidonia spp.meadows in the Mediterranean (84 ± 20 g C org m −2 yr −1 ) and Australia (12 ± 7 g C org m −2 yr −1 ) 25 , but similar to estimates from low biomass and fast-growing seagrass species (i.e.Zostera spp.and T. hemprichii) from Japan (ranging from 1.8 to 10.1 g C org m −2 yr −1 20 ).
Nevertheless, owing to the low C org storage in desert land areas and the relatively large seagrass habitat in the Red Sea coast of Saudi Arabia (370 km 2 18 ), seagrasses constitute hotspots of C org storage in this extremely arid region.Multiplying this area by the average C org stocks in 1 m-thick soils and the average C org accumulation rates, yields a total estimate of 1.2 ± 0.1 Tg C org at 1,700 ± 337 Mg C org yr −1 in Saudi Arabia meadows.The total C org stored in Saudi Arabia's seagrass meadows is roughly equivalent to 7 years of total CO 2 emissions from fossil-fuel burning, cement production, and gas flaring by Saudi Arabia, while seagrass meadows sequester annually around 0.8% of these emissions (Saudi Arabia emissions estimated at 0.16 Tg C at 2014 rates 26 ).
Long-term (i.e. based on 14 C) soil accumulation rates in Red Sea seagrass (ranging from 0.2 to 16 mm yr −1 ; 2.4 mm yr −1 on average) are within the range of 14C-derived values reported in previous studies from Australia (ranging from 0.15 to 2.5 mm yr −1 25,27 ), Japan (ranging from 0.37 to 1.3 mm yr −1 20 ), Spain and Italy (ranging from 0.6 to 4.9 mm yr −1 5, 25,28,29 ).The capacity of seagrass to elevate the seabed through sediment accretion has been previously recognized as a major component of their role in climate change adaptation 4 , as it helps mitigate against sea level rise.The results obtained in this study confirm that seagrass meadows in the Red Sea play a significant role in climate change adaptation through the protection against sea level rise, despite this being an arid region with very limited supply of terrestrial sediment via run-off.The sea level rise in the coast of Saudi Arabia has been estimated at 2.2 ± 0.5 mm yr −1 30 , hence seagrass ecosystems along the Saudi Arabian coast have been playing a key role in offsetting relative sea level rise.
The large variability in C org concentrations, stocks and accumulation rates among seagrass habitats (i.e.species composition and location) support the hypothesis that C org storage in seagrass soils is influenced by interactions of biological, chemical and physical factors within the meadow 14,15,31 .Despite that no significant differences in C org stocks among locations existed (at 95% confidence), the biogeochemical characteristics of the cores allowed the reconstruction of the processes and drivers involved in C org storage (Fig. 3A).Soil C org was negatively  correlated to soil dry bulk density (R 2 = 0.34), as previously shown in a range of sediments including seagrass soils 32,33 , which could explain the significant differences found in C org concentration (%) and the lack of differences in C org stocks (kg m −2 ).The relatively high soil C org concentrations (%) at Economic City could be related to the relatively high accumulation of seagrass detritus and abundance of fine sediments.These results support the hypothesis that the seagrass plants themselves play a key role in determining the amount of C org available for burial 14 , while the presence of fine sediments tends to reduce remineralization rates due to lower oxygen exchange and redox potentials 11,34,35 .The mechanisms behind C org accumulation and preservation in seagrass meadows at Petro Rabigh appear to be mainly related to the relative high soil accumulation rates together with large fluxes of C org from adjacent mangrove and tidal marsh ecosystems.Previous studies have shown that high soil accumulation rates in seagrass meadows, linked to the capacity of their canopy to tap and retain sediment particles 12,36 , the hydrodynamic energy and the production of biogenic carbonates within the meadow 37,38 , contribute to higher accumulation and preservation of C org after burial 14 .Petro Rabigh is an enclosed environment surrounded by mangrove forests, which has been shown to largely contribute to soil C org storage in adjacent seagrass meadows 15,39 .The relatively low soil C org storage of seagrass meadows at Thuwal Island could be explained by the relatively low contribution of seagrass detritus to the soil C org pool and the low soil accumulation rates (Fig. 3A).
Clear differences were observed among meadows with distinct species composition, with the highest soil C org concentrations (%) found in meadows composed of the largest seagrass species T. hemprichii and E. acoroides (Fig. 3B).However, the relatively low soil dry bulk density found in these meadows led to similar C org stocks among all meadows studied.The results obtained in this study show that soil C org concentration was influenced by the relative contribution of seagrass detritus to the soil C org pool and the amount of fine sediments, which support the results obtained in previous studies 14,15,25 .The relatively high soil C org concentration and seagrass contribution to the soil C org pools in T. hemprichii and E. acoroides could be explained by the highest above-and below ground biomass of stands formed by these species (ranging from 72 to 87 g DW m −2 and 210 to 392 g DW m −2 , respectively) compared to the other seagrass species studied (ranging from 2.3 to 27 g DW m −2 and 2.6 to 61 g DW m −2 40 .This study supports previous research reporting that the intrinsic properties of the seagrass themselves (e.g.canopy structure, below-and above-ground biomass, and productivity) can influence soil C org storage 14,15,31 .Moreover, the relative constant C stable isotope signatures along the cores confirm the stability of organic sources to soil C org pools, except for H. stipulacea and T. ciliatum meadows (i.e.δ 13 C values decreased below cm 60), which may indicate that seagrass meadows have only been present for the last centuries at these locations.The presence of coarse soil fibers throughout ~14 cores indicated that seagrasses were present at the coring sites throughout the period reconstructed or the soil depth studied.However, in half of the cores seagrass fibers disappeared at 25-60 cm depth, which could be either due to seagrass absence or the decomposition of coarse organic matter with ageing.Indeed, with the proxies analyzed here it is not possible to assure that the seagrass species occurring at present have remained the same through time.
Our results contribute to gaps in the existing global database on seagrass meadow C org stocks and accumulation rates, which were thus far lacking information from seagrass species in arid environments and suggest that even meadows comprised of ephemeral seagrass species can play an important role in C org sequestration.

Material and Methods
Study site and sampling.This study was conducted in four locations (Thuwal Island, Petro Rabigh, Economic City and Khor Alkharar) along 80 km of the Kingdom of Saudi Arabia coastline in the Central Red Sea (Fig. 4).Seagrass meadows are found along the Saudi coast, mainly composed by H. stipulacea, T. hemprichii, E. acoroides, T. ciliatum and H. uninervis 18 .
Seagrass meadows at Thuwal Island grow on shallow soil of weathered coral and are located near the fisherman city of Thuwal 41 .Petro Rabigh is a major industrial and petrochemical complex, whereas the Economic City is a newly developed city and harbor complex subject to intense coastal development 41,42 .The Khor Alkharar lagoon encompasses a relatively undeveloped coastal plain and is permanently connected to the Red Sea.
Twenty-seven soil cores were sampled in 1 to 7 m-deep mono-specific seagrass meadows using manual percussion and rotation (PVC pipe with an inner diameter of 60 mm; Supporting Information Table B).Three to four replicate cores were sampled within 100 m 2 of each mono-specific seagrass meadow at each site (three cores at Thuwal Island, 10 cores at Economic City, four cores at Petro Rabigh and 10 cores at Khor Alkharar).One third of the cores collected at each site were kept inside the PVC corers and transported to the laboratory (hereafter referred to as 'whole cores').The other cores from each study site were sampled in the field using a corer consisting of a PVC pipe with pre-drilled holes in the sidewall (3 cm wide and 3 cm apart; hereafter referred to as 'port cores'), allowing sub-sampling of soil samples along the core in the field by inserting 60 ml syringes into the pre-drilled holes along the PVC pipes (Supporting Information Table B).The total length of the core barrel used, the empty space inside the barrel before retrieval, the length of barrel outside the soil before retrieval, and the length of retrieved seagrass soil were recorded in order to correct the core lengths for compression effects and all variables studied here are referenced to the corrected, uncompressed depths (Supporting Information Table B).The volume of each subsample retrieved from the port cores was recorded in the field.The whole cores were sealed at both ends, transported vertically and stored at 4 °C before processing in the laboratory.
Laboratory procedures.The whole cores were opened lengthwise and cut into 1 cm-thick slices, and each slice together with the sub-samples from the port cores were oven-dried at 60 °C until constant weight to determine the dry bulk density (g cm −3 ).All samples from the port cores and every second slice of the whole cores were then grounded in an agate mortar and subdivided for analysis.
For the analyses of soil organic carbon (C org ) and stable isotope composition (δ 13 C), 1 g of ground sample was acidified with 4% HCl to remove inorganic carbon, centrifuged (3400 revolutions per minute, for 5 min), and the supernatant with acid residues was carefully removed by pipette, avoiding resuspension.The sample was then washed with Milli-Q water, centrifuged and the supernatant removed.The residual samples were re-dried and then encapsulated for C analyses using a Thermo Delta V Conflo III coupled to a Costech 4010 at the UH Hilo Analytical Laboratory, USA.The content of C org was calculated for the bulk (pre-acidified) samples.Carbon isotope ratios are expressed as δ values in parts per thousand (‰) relative to the Vienna PeeDee Belemnite standard.Replicate assays and standards indicated measurement errors of 0.01% for C org content and 0.06‰ for δ 13 C.
A total of 58 radiocarbon analyses were conducted in the 27 cores sampled (1-5 analyses per core) at the AMS Direct Laboratory (USA) following standard procedures 44 .Samples consisted of pooled shells and bulk soil (Supporting Information Table C).Shells were partially digested with 10% HCl, rinsed in ultrapure MQ water in order to remove fine sediment particles, inspected under a stereomicroscope for absence of attached reworked materials, and dried at 60 °C to a constant weight before radiocarbon dating.The 14 C age-depth models were produced using the R routine "Bacon" for Bayesian chronology building 45 , after 14 C calibration using the marine13 radiocarbon age calibration curve 46 taking into account a local Delta R of 110 ± 38 years 47 .From the Bacon routine output, the mean age was used to produce an age-depth weighted regression model forced through 0 (0 cm is cal.BP: 1950), using as weight the sum of the Euclidean distance of the minimum and maximum ages.In four cores, the 14 C results indicated either that the samples dated were modern (younger than ~400 years) or that the core was mixed (Supporting Information Table C), and therefore we did not produce age-depth models for these four cores.The relatively unknown marine reservoir effects at our study sites (and changes through time) remains a big assumption when calibrating 14 C ages 48 .All 14 C results used to model core age-depth chronologies in this study are older than ~400 years, and therefore, the burning of fossil fuels did not affect our 14 C-derived soil accumulation rates.All dates reported in this paper are expressed as radiocarbon calibrated years.
Numerical procedures.C org density (g C org cm −3 ) was calculated for each soil depth in each core by multiplying the sediment dry bulk density (g cm −3 ) by the C org concentration (%).For soil depths where C org content (%) was not analyzed, we extrapolated the %C org (i.e. by averaging the %C org between above and below depths) and multiplied the %C org by the dry bulk density (g cm −3 ) to obtain C org density (g C org cm −3 ).To allow direct comparisons among locations, the soil C org standing stocks per unit area (cumulative stocks; kg C org m −2 ) were standardized to 1 m-thick deposits.The total soil depth sampled was higher than 100 cm in 13 cores out of 27 cores sampled and therefore, no extrapolation was required for these cores.However, the soil C org stocks in 1 m-thick soil deposits were inferred in 14 cores (soil depths sampled ranged from 44 to 64 cm) to 1 m, by extrapolating linearly integrated values of C org content (cumulative C org stock; kg C org cm −2 ) with depth.Correlation between extrapolated C org stocks from 44 cm to 1 m and measured C org stocks in 1 m soil cores was r = 0.80 (P < 0.001; Supporting Information Fig. B).Note that scaling C org stocks to 1 m using this method could either lead to over-or underestimates of C org stocks.
Soil accretion rates (expressed in cm yr −1 ), soil accumulation rates (expressed in g DW m −2 y −1 ) and soil C org accumulation rates (expressed in g C org m −2 y −1 ) for the last millennia were estimated using 14 C age-depth models (Table 2).Accumulation rates of C org were calculated in 24 out of the 27 cores sampled by multiplying the C org inventories in 1 m-thick soil by the average 14 C soil accretion rate.
Analyses to test for differences in the variables studied among sites were performed using General Linear Model procedures in SPSS v. 14.0.General Linear Models were used to test for differences in dry bulk density (g cm −3 ), soil accretion rates (cm yr −1 ), soil C org concentration (in %), soil C org density (mg cm −3 ), soil C org stocks (kg m −2 in 1 m-thick soils) and soil C org accumulation rates (g m −2 yr −1 ), δ 13 C signatures of organic matter, and sediment grain size fractions among study sites and species composition (Table 4), followed by Tukey HSD posthoc tests to assess pairwise differences (Fig. 3).All response variables were square-root transformed prior to analyses and had homogenous variances.Study site (Thuwal Island, Petro Rabigh, Economic City and Khor Alkharar) and seagrass species (H.stipulacea, T. hemprichii, E. acoroides, T. ciliatum and H. uninervis) were treated as fixed factors in all statistical models (probability distribution: normal; link function: identity).
Stable Isotope Mixing Models were used to estimate the proportion of the autochthonous and allochthonous C org to the seagrass soil C org pool using δ 13 C and a one-isotope three-source mixing model 49,50 .The δ 13 C signatures within the top 60 cm of each core were pooled and analysed for the probability of relative organic matter contribution to soil stocks using Stable Isotope Mixing Models in R ('simmr' and 'rjags' packages) 51 .The δ 13 C signatures of potential C org sources (seagrass was considered as autochthonous C org , while mangroves plus halophytes, and seaweed plus seston were considered allochthonous C org ) in the four study sites were obtained from Almahasheer et al. 17 .The 'simmr_mcmc()' function works by repeatedly producing potential values of the proportional contribution of source material through a Markov chain Monte Carlo, with initial burn-in iterations (1,000) discarded and subsequent iterations (10,000) stored for use in the posterior distribution and analyses of the data 51,52 .Model convergence was confirmed using diagnostic plots and upper confidence intervals, while no overlap between source δ 13 C signature means ± standard deviations were observed.The simmr package allows for the incorporation of δ 13 C uncertainty into mixing models, while producing a Bayesian quantification of the most likely source contributors where there is a greater than n +1 sources when matching against n isotopes 52 .The dataset generated for this manuscript is provided as Supporting Information.

Figure 3 .
Figure 3. Biplots showing the relationships among the variables studies in the seagrass cores from the Red Sea based on study site and seagrass species.<0.125 mm (%) indicates the percentage of clay and silt and very fine sands within the bulk soil.

Figure 4 .
Figure 4. Location of seagrass meadows sampled in Saudi Arabia, Central Red Sea.The map was produced with ArcMap Version 10.2. Background map credits: the World Administrative Divisions layer provided by Esri Data and Maps, and DeLorme Publishing Company.Redistribution rights are granted http://www.esri.com/~/media/Files/Pdfs/legal/pdfs/redist_rights_103.pdf?la=en.The seagrass species present within the meadows surveyed at each study site are indicated.

Table 1 .
Average ± standard error (SE) of (a) dry bulk density (in g cm −3 ), organic carbon (C org ) content (in % and mg cm −3 ), δ 13 C signatures of soil organic matter and (b) sediment grain-size content at Red Sea seagrass soil cores (for the total length of core sampled; see Supporting Information TableBfor further details).The total number of data (N) provides an indication of the core length.Cores T1 to T3 were sampled at Thuwal Island, cores EC1 to EC10 at Economic city, cores PR1 to PR4 at Petro Rabigh and cores KA1 to KA10 at Khor Alkarar.
The soil C org content of Red Sea seagrass in 1 m-thick soils (3.4 kg C org m −2 on average) is well below the values from global estimates (ranging

Table 2 .
Soil accretion rates (cm yr −1 ) and organic carbon (C org ) stocks (kg m −2 ) and accumulation rates (g m −2 yr −1 ) in Central Red Sea seagrass meadows (in 1 m-thick soils).The C org stocks were extrapolated to 1 m in 14 out of 27 cores studied (see methods section for further details).Mean ± SE values are reported.n/a indicates variables that were not measured.
from 12 to 83 kg C org m −2 6

Table 3 .
Mean (±SE) of isotopic carbon values (‰) of potential organic sources into seagrass soils collected at the four study sites.N indicates the number of samples analyzed.n/a indicates variables that were not measured.Mean ± SE values are reported.

Table 4 .
Results of General Linear Models.Soil dry bulk density, soil accretion rates, organic carbon (C org ) concentration, stable carbon signatures (δ 13 C) of sedimentary organic matter, and sediment grain size fractions in response to Site (Thuwal Island, Economic City, Petro Rabigh and Khor Alkharar) and Seagrass species (Halophila stipulacea, Thalassia hemprichii, Enhalus acoroides, Thalassodendrum ciliatum and Halodule uninervis).The degrees of freedom (d.f.) for each term in the mixed model analysis are indicated.