Expanding Greenland seagrass meadows contribute new sediment carbon sinks

The loss of natural carbon sinks, such as seagrass meadows, contributes to grenhouse gas emissions and, thus, global warming. Whereas seagrass meadows are declining in temperate and tropical regions, they are expected to expand into the Arctic with future warming. Using paleoreconstruction of carbon burial and sources of organic carbon to shallow coastal sediments of three Greenland seagrass (Zostera marina) meadows of contrasting density and age, we test the hypothesis that Arctic seagrass meadows are expanding along with the associated sediment carbon sinks. We show that sediments accreted before 1900 were highly 13C depleted, indicative of low inputs of seagrass carbon, whereas from 1940’s to present carbon burial rates increased greatly and sediment carbon stocks were largely enriched with seagrass material. Currently, the increase of seagrass carbon inputs to sediments of lush and dense meadows (Kapisillit and Ameralik) was 2.6 fold larger than that of sparse meadows with low biomass (Kobbefjord). Our results demonstrate an increasing important role of Arctic seagrass meadows in supporting sediment carbon sinks, likely to be enhanced with future Arctic warming.


Results and Discussion
The stocks of organic carbon (C org ) within the top 10 cm of the sediments ranged from 197 g C m −2 to 595 g C m −2 , and the inorganic carbon (C inorg ) stocks were similar (Ameralik) or about 2 times lower (Kapisillit and Kobbefjord) than those of C org ( Table 1). The sediments under the seagrass meadows presented 210 Pb concentration profiles that allowed establishing robust age models (Fig. 1), despite mixing may have occurred in the upper 2 cm sediment at Kobbefjord. The presence of a relatively thin mixed layer has limited impact when applying the CRS and the CF:CS (below the mixed layer) models 15 . Coupling of sediment chronologies with organic carbon concentration revealed recent increases in the C org concentration of eelgrass sediments, particularly so for the dense seagrass meadow at Kapisillit over the past 40 years (Fig. 2).
The corresponding C org burial rates since 1900 ranged 10 fold from 1.30 g C m −2 year −1 for that at Ameralik to 10.53 g C m −2 year −1 for the meadow at Kapisillit, with a large increase in the ratio of C org to C inorg towards present in all three meadows (Fig. 3), likely reflecting changes in sedimentary supply over time as well as differential diagenesis with sediment depth/age. Burial rates of C org increased greatly from 1940's to present in the two meadows (3.5 fold at Kobbefjord; 9.1 fold at Kapisillit) where rates could be resolved over this time period (Fig. 4). For Ameralik, only a few layers could be dated (since 1940) and, hence, we cannot resolve any potential change in carbon burial rates. Hence, the evidence for increased sedimentation rates toward present time is based on limited data and should be, therefore, considered to carry considerable uncertainty.
Analysis of changes in Corg with sediment depth suggested a recent change in sources of C org to the Greenland sediments examined. In particular, sediments accumulated before 1900 were characterized by highly 13 C depleted C org pools (mean ± SE = −30.44 ± 0.38‰ across all three sites), with C org pools subsequently becoming  progressively 13 C-enriched (Fig. 5). The C org source dominating sedimentary inputs before 1900 could be a combination of phytoplankton (δ 13 C = −24.7 ± 1.12‰) with land-derived C org , both from terrestrial vegetation and fossil organic carbon released with glacial melting, which have similar isotopic composition (δ 13 C extending to −35‰ in both cases) 16 . However, these sources cannot account for the 13 C-enriched C org pools stored in the sediment since 1900 (Fig. 5), which requires either a novel source more enriched in 13 C or an increase in the contribution of an existing source enriched in 13 C. This is likely to be eelgrass-produced carbon, as the average (±SE) δ 13 C of present-day eelgrass is −7.24 ± 0.21‰, indicative of highly 13 C-enriched organic carbon. The mixing model using the carbon sources before 1900, i.e. business as usual carbon source scenario, and eelgrass carbon as end members indicated an increased contribution of eelgrass to sediment C org after 1900 in all meadows (Fig. 5). Indeed, the surface sediments in the eelgrass meadow at Kapisillit contain the largest contribution of eelgrass (see   (Table 1) 14 . Moreover, the contribution of eelgrass material to sediment C org pool has been increasing since 1900 at Kapisillit and Kobbefjord (see small plots in Fig. 5). It would be unlikely that the depletion of 13 C observed in older sediment would result from diagenesis, since decomposition rate of eelgrass is about 8 times slower than that of phytoplankton 17 . The C org stocks over the top 10 cm of sediment, where isotope mixing models unambiguously support the contribution of eelgrass, are low when compared to global seagrass C org stocks within a similar soil thickness (2.0 to 6.0 Mg C org ha −1 in the meadows studied here compared to 9.6 ± 0.7 Mg C org ha −1 , on average ± SE, recalculated from the global compilation of Fourqurean et al. 6 ).
Warming and reduced ice cover in Greenland fjords have been proposed to be conducive to a poleward expansion of marine macrophytes 9,14,18,19 , as both light and temperature thresholds become more favourable to support macroalgal and seagrass growth. Indeed, experimental evidence indicates that Greenland current and projected (under IPCC scenarios of greenhouse gas emissions) warming conditions enhance eelgrass growth 20 . The presence of eelgrass in Greenland fjords was first documented in the Godthåbsfjord system in 1830 14 , and in Kapisillit and Ameralik in 1916 (Herbarium specimens from Greenland Herbarium, Botanical Museum, University of Copenhagen) and 1921 21 . However, eelgrass in Kobbefjord, the population with the smallest extent and biomass among those studied here, wasn't reported until 2009 14 . Hence, we speculate that while eelgrass meadows have been present in Greenland for at least 180 years, they appear to be expanding and increasing their productivity. This is supported by the rapid growth in the contribution of seagrass-derived carbon to the sediment C org pool, from less than 7.5% at the beginning of 1900 to 53% at present, observed in the studied meadows. Expansion and enhanced productivity of eelgrass meadows in the subarctic Greenland fjords examined here is also consistent with the on average 6.4-fold acceleration of C org burial in sediments between 1940 and present.
Seagrass meadows have been shown to rank amongst the most intense carbon-sink ecosystems of the biosphere 6,22,23 with conservation and restoration programs aimed at protecting and restoring the carbon stocks and sink capacity lost with global seagrass decline 24,25 . In contrast, seagrass meadows in Greenland seem to be expanding, propelled by warmer seawater temperature and higher doses of submarine irradiance with reduced ice cover 14 . The expansion of seagrass in Greenland fjords represents a novel carbon sink, with limited significance at present due to the small size of the meadows. However, the potential for further expansion is huge, as the convoluted Greenland coastline represents about 12% of the global coastline. The poleward latitudinal limit of eelgrass is located at far higher latitudes (70°N) than those studied here (64°N), suggesting that eelgrass, along with other boreal macrophyte species, is likely to expand poleward with decreasing ice cover and higher temperatures 9,14,26 . Hence, whereas the carbon sink associated with sediments under Greenland eelgrass meadows is likely to be very modest at present, it may reach significant levels along the 21 st century.
Whereas the concern globally is in slowing down or stopping altogether further losses of seagrass 24,27 , we provide evidence here for an increasingly important role of sediments under seagrass meadows in Greenland as a carbon sink, whose significance is likely to increase with further climate warming. Propelling this emerging carbon sink requires protection of extant meadows, as these do not only represent carbon sinks already in operation, but also supply propagules essential for further expansion of this valuable ecosystem. The fjords are 17 km to 100 km long, with the inner parts of the branches covered with sea ice from November to May during severe winters and no sea ice in the outer bays 14 . The tidal amplitude in the region ranges from 2 to 5 m. Two of the studied seagrass meadows (Ameralik, 64°15′N, 51°35′W; and Kapisillit, 64°28′N, 50°13′W) were located at the inner branches of the Ameralik-Itivleq and Kapisigdlit Kangerdluat fjords whereas the third meadow grew at the middle part of Kobbefjord (64°09′N, 51°33′W). All studied seagrass meadows were permanently submerged, at water depths ranging between 2 and 4 m, and exposed to maximum summer water temperatures of 9 °C at Kobbefjord, to 14-15 °C, at Ameralik and Kapisillit 14 14 . Z. marina at Ameralik and Kapisillit formed lush and extensive meadows, with 100% cover and high above-and belowground biomass (Table 1). Conversely, Z. marina at Kobbefjord developed only few patches of vegetation, with less than 10% cover and lower total summer biomass ( Table 1). Olesen et al. 14 provide a detailed description of Z. marina dynamics and environmental conditions at the studied sites.
Sampling. In August 2012, we collected sediment cores (5.2 cm diameter) in seagrass meadows of the 3 study sites. We retrieved 3 cores from Kobbefjord and 4 cores from Ameralik, each being 36-42 cm long, and a single 16 cm long core from the meadow at Kapisillit. The sediment at Kapisilit was extremely rich in clay and it prevented to insert the cores deeper; despite trying we could only collect one sediment core longer than 10 cm. We collected the sediment cores from a boat using a manual sampling device to insert the corer as deep as possible into the sediment. Because of the method to collect the cores, we could not measure sediment compaction due to sampling, but it should be negligible, at least, for Kobbefjord and Ameralik sediments, since they were highly sandy and organic poor. The seagrass meadows sampled extended along the suitable habitat. However, we did not collect samples in bare sediments adjacent to vegetated ones, because their current bare nature may be transient and there is no guarantee that these sediments would not have supported seagrass at some point along the 100 years of sediment carbon burial reconstructed here. We transported the sediment cores vertically to the Greenland Institute of Natural Resources, Nuuk. In the laboratory, we sliced one core from Ameralik and Kobbefjord every centimeter and the remaining cores every 2 cm. We weighed all sediment samples after oven-drying them at 60 °C for 48 h, and estimated their sediment dry bulk density by dividing sediment dry weight by wet sediment volume. We ground the dried sediment and stored the samples until analysis of organic matter, inorganic carbon concentrations, δ 13 C and 210 Pb concentrations.
Along with the collection of sediment cores we also collected 3 samples of eelgrass leaves, each containing material from different shoots, from each meadow, dried at 60 °C for 48 h and ground them for subsequent analysis of δ 13 C (see below). 210 Pb sediment dating and sediment accretion rates. The cores were dated by means of 210 Pb.
Concentrations of 210 Pb along the upper 10-20 cm of one core at each site were determined by alpha spectrometry through the measurement of the activity of its granddaughter, 210 Po, following Sanchez-Cabeza et al. 28 . Briefly, 200 mg aliquots of each layer were spiked with a known amount of 209 Po, acid digested and dissolved into a 100 mL 1 M HCl solution, from which the polonium isotopes were autoplated onto pure silver disks. Polonium emissions were measured by alpha spectrometry using Passivated Implanted Planar Silicon, PIPS detectors (CANBERRA, Mod. PD-450.18 A.M.). Reagent blanks were run in parallel and found to be comparable to the detector backgrounds. Supported 210 Pb concentrations were determined by averaging total 210 Pb concentrations at the base of each profile. These were comparable to the 226 Ra concentrations obtained at selected depths in each core by measuring the emissions of its decay products 214 Pb (295 and 352 keV peaks) and 214 Bi (609 keV peak) using a high-purity germanium detector (CANBERRA, mod. GCW3523) in calibrated geometries, sealed for 21 days.
The concentration profiles of excess 210 Pb used for the age modelling were determined by subtraction of the supported 210 Pb from total 210 Pb concentrations along each core (Fig. 1), showing a decrease in concentration from the surface to constant concentrations at various depths depending on each core (excess 210 Pb horizon). Sediment accumulation rates were calculated by applying the constant flux: constant sedimentation (CF:CS) 29 and the constant rate of supply (CRS) 30 models, which rendered equivalent results (Table 1).
Organic and inorganic carbon stocks, burial rates and δ 13 C. We measured the concentration of organic matter (OM, % DW) using the loss of ignition technique along the sediment depth profile at 2 cm resolution, and in subsamples of 3 sediment cores per meadow, except for Kapisillit where only one core was collected. We combusted pre-weighed dry (60 °C) sediment samples at 550 °C for 5 hours and estimated the concentration of organic matter (OM) as: where W 60 is the dry weight of the sample at 60 °C and W 550 the weight of the sample after combustion at 550 °C. We estimated the sediment organic carbon concentrations (C org , %DW) from measured organic matter concentrations (OM, % DW) using the relationship described for seagrasses by Fourqurean et al. 6 . We also analyzed the concentration of inorganic carbon (C inorg , %DW) by conducting a second combustion of the sediment samples at 1000 °C for 2 h to release the CO 2 of the carbonate and subsequently calculating the concentration of C inorg as: where W 1000 was the weight of the sediment sample after the second combustion and 0.27 is the ratio of the atomic weight of carbon (12 g) to the molecular weight of CO 2 (44 g) released during carbonate combustion. Along the sediment profiles of each core, we calculated the ratio between organic and inorganic carbon concentrations (C org : C inorg ) as well as the pools of C org (g C org cm −3 ) and C inorg (g C inorg cm −3 ) by multiplying, respectively, C org and C inorg concentrations with the sediment dry bulk density of each sediment sample. We calculated the stocks of C org and C inorg (in g C m −2 ) within the top 10 cm Z. marina sediments by integrating the C org and C inorg pools within the top 10 cm sediment layer and over one meter square of seagrass meadow. Similarly, we calculated the C org and C inorg stocks (in g C m −2 ) accumulated since year 1900 by integrating, respectively, C org and C inorg pools in sediments younger than 112 years, using the sedimentation rates obtained from the We tested if the relative contribution (f) of seagrasses to sediment organic carbon increased after year 1900. We did so by applying a two source-mixing model, 13 sed after 1900 13 seagr 13 sed before1900 that considered Z. marina (δ 13 C seagr Ameralik = −7.31 ± 0.02‰, δ 13 C seagr Kapisillit = −6.58 ± 0.33‰, δ 13 C seagr Kobbefjord = −7.83 ± 0.15‰) and a business as usual carbon source scenario, represented by the average δ 13 C sed observed in sediments accreted before year 1900 (δ 13 C sed after 1900 = −30.44 ± 0.38‰), as end members. We corrected for the historical change in the δ 13 C source signatures due to 13 C depletion in the atmospheric CO 2 and oceanic DIC δ 13 C signature towards present derived from the burning of fossil fuels (i.e. Suess effect) 32 . This was done by applying the model described by Schelske  to estimate the δ 13 C of atmospheric CO 2 (δ 13 C atm ) over time (years, Y) since year 1840. These values were subsequently normalized to δ 13 C atm in year 1840, and the resulting time-dependent depletion in δ 13 C since1840 was subtracted from the measured δ 13 C sed for each dated sediment section.
Data analysis. Average (and standard error) values of carbon parameters were calculated from measurements of 3 replicated sediment cores from Ameralik and Kobbefjord while values from a single core were applied from Kapisillit. We used JMP to fit linear trends to sediment profiles of enhanced eelgrass contribution (relative to "business as usual" organic carbon input sources, i.e. those prior 1840) to sediment organic carbon after year 1900.