Impact of mooring activities on carbon stocks in seagrass meadows

Boating activities are one of the causes that threaten seagrass meadows and the ecosystem services they provide. Mechanical destruction of seagrass habitats may also trigger the erosion of sedimentary organic carbon (Corg) stocks, which may contribute to increasing atmospheric CO2. This study presents the first estimates of loss of Corg stocks in seagrass meadows due to mooring activities in Rottnest Island, Western Australia. Sediment cores were sampled from seagrass meadows and from bare but previously vegetated sediments underneath moorings. The Corg stores have been compromised by the mooring deployment from 1930s onwards, which involved both the erosion of existing sedimentary Corg stores and the lack of further accumulation of Corg. On average, undisturbed meadows had accumulated ~6.4 Kg Corg m−2 in the upper 50 cm-thick deposits at a rate of 34 g Corg m−2 yr−1. The comparison of Corg stores between meadows and mooring scars allows us to estimate a loss of 4.8 kg Corg m−2 in the 50 cm-thick deposits accumulated over ca. 200 yr as a result of mooring deployments. These results provide key data for the implementation of Corg storage credit offset policies to avoid the conversion of seagrass ecosystems and contribute to their preservation.

Among the multiple ecosystem services seagrass meadows provide, their capacity to sequester carbon dioxide has generated considerable interest among scientists and policy makers due to its potential role in helping mitigate climate change [1][2][3] . The organic carbon (C org ) sequestered in seagrass ecosystems, along with that sequestered in other coastal vegetation such as mangroves and salt marshes, has been termed as 'blue carbon' 4 . 'Blue carbon' strategies build on the opportunity to avoid or mitigate CO 2 emissions through the conservation and restoration of seagrass meadows, which rank among the most threatened habitats in terms of global loss rates. Seagrass losses have been estimated at 29% of their global extent since 1880 and an average net decline in area of 7% yr −1 since 1990 5 .
Among other consequences, loss of seagrass meadows results in erosion of sediments and, potentially, the re-mineralization of the sedimentary C org accumulated over millennia, which may then contribute to increasing atmospheric CO 2 6 . Fourqurean et al. 1 estimated that, considering the carbon accumulated within the top meter of seagrass meadows to be lost and remineralized upon loss of the seagrass, current rates of seagrass loss could generate emissions of up to 300 Tg carbon per year. Therefore, the destruction of seagrass habitats may contribute to the land-use component of anthropogenic CO 2 4 . However, the calculations above are not well constrained and there is considerable uncertainty as to the loss of C org stores after meadow disturbances, a major of the bottlenecks hindering the application of seagrass conservation schemes as a low-cost method to mitigate climate change 3,7,8 . Marbá et al. 6 reported that loss of a seagrass meadow in Oyster Harbor (SW Australia), led, within two decades from the loss, to the loss of the C org stored in the preceding 60 years. However, this is, to the best of our knowledge, the only estimate thus far available on the loss of C org stocks following seagrass loss. Hence, there is urgency in expanding the observational basis on the loss and fate of C org stores following disturbances to seagrass meadows, a critical step for seagrass conservation efforts to be assigned carbon mitigation credits 8 .
Seagrasses losses derive largely from coastal eutrophication, but mechanical disturbance and destruction, such as dredging, anchoring, and coastal development are also responsible for significant losses worldwide [9][10][11][12] . Growing human populations and activities across many of the world's coastlines are mirrored by an increase in the demand for mooring areas, which are often located in the sheltered locations inhabited by seagrass 13 . Moorings impact seagrass by the erosion caused by the mooring chain dragged around the meadows following wind direction and wave action, resulting in un-vegetated scars around the mooring 14 .
Even though the spatial impact of mechanical damage might not be as large-scale as other human-induced disturbances (i.e. eutrophication), moorings and anchor scars (or blowouts) can persist for decades in seagrass meadows. While seagrasses may regrow into small blowouts after moorings are removed, the un-vegetated scars can persist even decades after the original disturbance 14 . When the environmental conditions are unsuitable for seagrass re-colonization, such as in areas exposed to wave action, mooring scars act as focus for erosion, expanding in size to coalesce with adjacent ones to lead to un-vegetated blowouts 14,15 . This process results in seagrass fragmentation at the landscape scale, thus threatening the persistence of seagrass meadows 14,16 .
Australia, one of the nations supporting the highest seagrass area worldwide 17 also has extensive mooring areas, such as that in Rottnest Island, 9 miles offshore from the city of Perth, Western Australia. The first moorings in Rottnest Island were installed in Thomson Bay in the 1930s, expanding to reach 893 moorings distributed within Rottnest embayments, with Thomson Bay supporting the largest number of moorings (316 in total; 14,18 ). The seagrass meadows at Rottnest Island are heavily fragmented as a result of heavy use of permanent moorings 14,19 . To mitigate these impacts, new types of seagrass-friendly moorings have been installed from 1980s onwards (e.g. replacement of original swing moorings to single three chain cyclone moorings) that minimize the impact on seagrass, with the area of mooring-induced scars at Thomson Bay declining by 65% due to seagrass regrowth between 2004 and 2006 18 .
Whereas the impacts of the moorings on seagrass cover have been assessed 14,19 , the impacts on C org sequestration and stocks has not been yet evaluated anywhere. Rottnest Island provides an ideal location to examine the effect of seagrass loss on the C org sequestration function of seagrasses; it allows a comparison of C org stores in meadows that have never been disturbed against patches that have been affected by moorings in otherwise identical environmental conditions. Here we test the hypothesis that seagrass loss and the sustained action of mooring chains lead to losses of C org stocks from seagrass sediments relative to those in adjacent, undisturbed meadows.

Results
Thomson Bay, 101 ha in area, supported 80 ha of seagrass meadows, 17.2 ha of bare sand and 3.4 ha of reef. At Stark Bay, 36 ha in area, seagrass meadows covered 17.4 ha, while bare sands and reefs covered 8.5 ha and 10 ha, respectively ( Fig. 1). Based on the aerial imagery obtained in 2009, a total of 316 and 95 moorings were identified at Thomson Bay and Stark Bay, respectively. The interpretation of the aerial imagery within this study showed that moorings (deployed from 1930s onwards) led to the loss of 1.3 ha and 0.9 ha of seagrass meadows at Thomson Bay and Stark Bay, respectively ( Table 1). The dimensions of individual mooring scars in Stark Bay were larger and more variable (95 ± 11 m 2 ) than those at Thomson Bay (41 ± 3 m 2 ). Taking an average area of mooring scars of 54 ± 4 m 2 (for both study sites) and 893 moorings in total, an estimated total area of 4.8 ha of seagrass has been lost in Rottnest Island due to moorings ( Table 1).
The cores from seagrass meadows at both sites consisted of dark, brown sediments with coarse seagrass detritus (rhizomes, roots and sheaths) and shells of bivalves and gastropods embedded within the sediments. In contrast, the cores from mooring scars contained white carbonate sediments, lacking seagrass fibers in the top 15 to 25 cm, while below these depths sparse seagrass fibers were found.
Overall, the dry bulk density was significantly lower (P < 0.001) in seagrass cores (mean ± SEM; 0.82 ± 0.03 g cm −3 ) compared to the cores from the mooring scars (1.03 ± 0.03 g cm −3 ; Fig. 2; see Supplementary Table S1a and S2). The C org content was five-fold higher in seagrass sediments (1.60 ± 0.06%) compared to scars (0.34 ± 0.03%; P < 0.001). The C org content in seagrass meadows at Thomson Bay decreased from 4% in the top 2 cm to 2% at 10 cm depth to 1-2% between 10 and 50 cm (Fig. 2). The C org content in seagrass meadows at Stark Bay was constant with depth (1.41 ± 0.1%), while the C org content in the sediments in mooring scars was lower and constant (0.30 ± 0.04%) through the core in both sites (Fig. 2). The δ 13 C signatures of organic matter in the seagrass cores (− 14.0 ± 0.3‰ at Thomson Bay and − 13.8 ± 0.4‰ at Stark Bay) were heavier than those in the scars (− 15.0 ± 0.4‰ at Thomson Bay and − 14.6 ± 0.4‰ at Stark Bay; see Supplementary Table S1a and S2), but differences were not significant (P = 0.064; Table S2). No clear pattern in δ 13 C with depth was observed (Fig. 2).
Sediments were carbonate-rich (82-86%), and the CaCO 3 content in seagrass cores (84%) was slightly but significantly higher than in scars (83%; P < 0.001; Fig. 2; see Supplementary Table S1a and S2). All the sediments were dominated by fine and medium sands (39% and 38% in seagrass meadows, and 36% and 43% in scars, respectively) but those from seagrass meadows had a significantly higher percentage of very fine sand and mud (14 ± 1%) compared to the scars (8 ± 0.4%; P < 0.001; Fig. 3; see Supplementary Table S2). In both sites, medium and coarse sands were significantly more abundant in the mooring scars compared to the seagrass beds (P < 0.001; Fig. 3), while fine sands were more abundant in seagrass meadows (P < 0.01; see Supplementary  Table S2).
No significant effects of sediment depth were observed for any of the variables studied (see Supplementary  Table S2). Significant two-way interactions were observed for density, C org and CaCO 3 inventories, very fine sand and mud, and medium sands. Excess 210 Pb horizons in seagrass meadows in Thomson Bay and Stark Bay were reached at 26 cm and 35 cm depth, yielding average sedimentation rates of 0.25 ± 0.01 and 0.27 ± 0.08 cm yr −1 , respectively (Fig. 4). However, the seagrass core at the Stark Bay site showed evidence of mixing in the upper 15 cm, and thus the estimated sedimentation rate must be considered an upper limit. Sediment cores collected     . This agreement support the hypothesis that sediments under seagrass meadows contained all deposited excess 210 Pb for the last 100-150 years. The C org inventories (normalized to 50 cm-thick deposits) in the mooring scars (1.4 ± 0.2 at Thomson Bay and 1.8 ± 0.9 kg C org m −2 at Stark Bay) were four-to five-fold lower than in adjacent seagrass meadows (6.6 ± 0.8 at Thomson Bay and 6.2 ± 0.9 kg C org m −2 at Stark Bay; Fig. 5; see Supplementary Table S1b and S2). Based on the estimated sedimentation rates, the seagrass meadows at Thomson Bay and Stark Bay accumulated 34 ± 3 and 34 ± 10 g C org m −2 yr −1 , respectively (see Supplementary Table S1b). The seagrass meadows at both study sites stored somewhat lower amounts of CaCO 3 in their sediments (312 ± 18 kg CaCO 3 m −2 and 391 ± 46 kg CaCO 3 m −2 , respectively) compared to mooring scars (384 ± 65 kg CaCO 3 m −2 and 486 ± 26 g CaCO 3 m −2 ; see Supplementary Table S1b and S2). Seagrass meadows at both study sites accumulated, on average, 1.6 ± 0.1 kg CaCO 3 m −2 yr −1 and 2.0 ± 0.6 kg CaCO 3 m −2 yr −1 , respectively (see Supplementary Table S1b).

Discussion
Mechanical destruction of seagrass meadows by moorings at Rottnest Island from the 1930s onwards has led to the loss of their C org sequestration capacity by precluding further accumulation of C org , and through the erosion of the centenary C org stores underneath seagrass meadows. The deployment of moorings since the 1930s had led to an average loss of at least 4.8 kg C org m −2 that had accumulated in 50 cm-thick deposits over the last ca. 200 years. Because the undisturbed meadows contained up to 2-fold higher amounts of fine sediments (< 0.125 mm)  compared to mooring scars, it appears that the loss of C org resulted from both direct scouring of sediments and re-suspension and subsequent loss of fine-grained sediments. The presence of fine-grained sediments in the substrate contributes to the accumulation and preservation of C org , due to their higher capacity to retain organic matter compared to coarse sediments 21 , and because fine-grained sediments contribute to the creation of anoxic conditions in the sediment 22 and, ultimately, to the higher preservation of C org through reduced remineralization rates 23,24 . The seagrass canopy has the ability to trap fine sediments suspended in the water column, reducing wave energy and resuspension 25 . Furthermore, the presence of seagrass rhizomes and roots stabilizes sediments and reduces resuspension 26 . Mooring scars with no seagrass fibers in the sediment, living or dead, tend to be unstable, leading to further erosion of the scar edges while creating deeper scouring areas 15,27 and ultimately, contribute to greater loss of sedimentary C org stores.
The enriched δ 13 C signatures of organic matter present in the seagrass meadows indicate a smaller contribution of seagrass-derived organic matter to the sediments in mooring scars. The uniform concentration of excess 210 Pb and its inventory indicate that sediments had not been accumulated at the scar site in Thomson Bay, but likely eroded and thoroughly mixed. Given that the accumulated stocks of C org at mooring scars in Thomson Bay were significantly lower than in seagrass meadows, it is likely that intense sediment reworking through disturbance probably led to the net loss of C org stocks. On the other hand, the depletion of excess 210 Pb in the scar site at Stark Bay suggests that erosion of sediments, and loss of the associated C org , were significant after seagrass meadows had been lost due to mooring activities. Therefore, the lack of sediment accumulation as a consequence of mooring deployment and scouring of the living seagrass biomass may also lead the loss of their C org sequestration capacity by precluding further accumulation of sediments and C org , estimated at 34 g C org m −2 yr −1 (based on 210 Pb age-models). Over an average 50-year period of mooring presence at Rottnest Island 18 , the loss of C org accumulation following disturbance has been estimated as 1.7 kg C org m −2 . Note that the C org stores in the living biomass were also loss due to mooring activities, but were not accounted for in this study.
Examination of the vertical profile of C org relative to the estimated sediment date shows that loss of seagrass cover by the mooring activity led to erosion and subsequent loss of the C org stored over the past ca. 200 years. This is a far more severe erosive loss than that reported by Marbà et al. 6 in Oyster Harbor (SW Australia) following eutrophication, where 60 years of C org accumulation was lost 25 yr after seagrass loss occurred. Hence, the results presented here show that the movement of mooring chains induce sediment erosion and a considerable loss of C org stores, which are then exposed to oxic conditions that in turn may lead to remineralization of organic matter 28 , returning sedimentary C org stocks to the ocean and atmosphere in the form of CO 2 .
The analysis of aerial imagery from 2009 showed that moorings produced scars averaging 41 m 2 at Thomson Bay and much larger, 95 m 2 , at Stark Bay. The differences in the sizes of scars at the two sites may reflect a number of processes. The length of chain on the mooring and scouring from propeller wash from large boats may lead to initially different sized scars. Subsequent environmental processes (i.e. regrowth or erosion) will then act on the scar and can vary depending on the geomorphology and the hydrodynamic energy at each site 13,29,30 . A third factor is the period of time the scar has been present (or deployment time of the mooring), since this will allow greater opportunity for erosional processes to act. Thomson Bay is more sheltered than Stark Bay 14 , where the edges of mooring-induced scars would erode further due to wave action resulting in coalescence between adjacent mooring scars. It is possible to determine from the 1941 and 2003 aerial images that the mooring scars sampled at Thomson Bay have always been present since 1941. However, two out of the four mooring scars sampled at Stark Bay were seagrass back in 1941, but scars in 2003. The resolution of the 1941 imagery is very low (not shown), and the lack of information about the history of mooring deployments at Rottnest Island precludes further interpretations on the effects of time and type of mooring on C org loss.
According to Walker et al. 19 , scars originated from mooring activities in the 1980s at Thomson Bay and Stark Bay averaged 71 and 17 m 2 in diameter and caused a loss of 2.4 ha and 0.18 ha, respectively (Table 1). Though the size of mooring-induced scars observed here contrasted with those reported by Walker et al. 19 , the apparent decrease in the average scarring at Thomson Bay could indicate recovery at this site. According to Rottnest Island Authority 18 , the area of mooring-induced scars at Thomson Bay experienced a 65% decrease (i.e. seagrass growth) during the period of 2004 to 2006. Although the methods and imagery used to estimate the seagrass area loss were not the same, our results seem to indicate that the efforts undertaken by Rottnest Island Authority to reduce the impacts of moorings by replacing some moorings for more environmentally-friendly ones 18 is contributing to reduce the area of mooring scars at Thomson Bay. On the other hand, the scar size measured in this study at Stark Bay is 5 to 6-fold higher than those measured by Walker et al. 19 (Table 1), pointing at the coalescence and blowout expansion of initial patches originated by moorings 13,15 . In particular, this phenomenon is more evident in the northern, deeper parts of Stark Bay, which are further offshore and more exposed to hydrodynamic energy.
Although our sample size was small (i.e. 4 cores per site and treatment), the results presented demonstrate that loss of seagrasses by mooring activities in Rottnest Island led to significant loss of 75%, on average, of the sedimentary C org stocks accumulated in the top 50 cm of sediment, in addition to loss of the seagrass C org sequestration capacity. These losses were associated with erosion of fine sediments leading to the loss of much of the C org stock accumulated over the past 200 yr, thereby involving the loss of C org accumulated about 160 years prior to the loss of seagrass. Although the fate of C org lost after disturbances remain unknown, the C org stores are eroded and at risk to be re-mineralized in the water column or surface sediments onto which they may be re-deposited. Our results add to the report by Marbà et al. 6 , for seagrass loss following eutrophication, to demonstrate that seagrass loss leads to the loss of much of the C org stored over decades and even centuries.
Considering the impact of mooring activities on coastal areas and its capacity to disturbing the sequestration and preservation of C org in seagrass meadows, among other ecosystem services that they provide, there is an urgent need for better management focused on mooring areas. Since seagrass rehabilitation is costly and relatively unsuccessful in southern Australia 31 , the use of environmentally-friendly moorings might be a plausible solution to the problem.  Fig. 6). Rottnest Island is a marine reserve, surrounded by limestone reefs and extensive sub-tidal sand and seagrass habitats 18 . Rottnest Island is a popular tourism destination, with approximately 350,000 visitors per year in the 1990s, of which 70,000 arrived on private vessels 32 . By 2009, the number of visitors increased to approximately half a million, with 150,000 arriving by private vessel 29 . The Australian Army set up moorings for explosives barges during World War II. Some of these original moorings were converted to recreational moorings in the 1950s and 1960s. An increase in the number of moorings occurred over the 1970s and early 1980s, resulting in the establishment of a moratorium on moorings. From the 1980s onwards, new environmentally friendly moorings were installed. Damage produced by mooring chains dragged over the meadows denuded these areas of seagrass, resulting in sandy patches within the seagrass beds 19 , which were first reported from an aerial photograph in 1941 14 .
In order to assess the impacts of moorings on C org stocks, a total of 16 sediment cores were sampled: four sediment cores in mooring scars and four cores in adjacent and undisturbed mixed meadows of Posidonia sinuosa and Amphibolis griffithii at each study site (Thompson Bay and Stark Bay; Fig. 1). Sediments were sampled using PVC cores (80 cm long, 6 cm in diameter) that were gently hammered into the sediment at 2 to 4 m water depth.
PVC cores were cut lengthwise, and sediments inside the cores were cut into 1 cm-thick slices. Each slice was weighed before and after oven drying at 60 °C until constant weight (dry weight; DW). Then, every second slice was divided into two subsamples by quartering (i.e. 18-22 slices per core). One subsample was ground and analyzed for organic carbon (C org ) and stable isotope composition (δ 13 C), and calcium carbonate (CaCO 3 ) content. The other subsample was used for sediment grain-size analyses.
For C org and δ 13 C analyses, 1 g of ground sample was acidified with HCl 4% until bubbling stopped to remove inorganic carbon, centrifuged (3400 rpm for 5 minutes), and the supernatant with acid residues removed carefully by pipette, avoiding resuspension of the sediment. Then, the sample was washed with Milli-Q water, centrifuged, and the supernatant removed. The residual samples were re-dried and encapsulated for analysis using a Micro Cube elemental analyzer (Elementar Analysensysteme GmbH, Hanau, Germany) interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK) at the University of California Davis. Content of C org was calculated for bulk (pre-acidified) samples.
The CaCO 3 content was analyzed with a Calcimeter (Pressure Gauge Model 432 (Fann ® ); ASTM D 4373-84 Standard) by adding 10% HCl to the sample in a sealed reaction cell. The pressure build up due to the CO 2 was measured with a bourdon tube pressure gauge that was pre-calibrated with reagent grade calcium carbonate.
For sediment grain-size analysis, a Mastersizer 2000 laser-diffraction particle analyzer was used following digestion of the samples with hydrogen peroxide to remove organic matter. Sediments were classified as coarse sand (< 1 mm and > 0.500 mm) medium sand (< 0.500 mm and > 0.250 mm), fine sand (< 0.250 mm and > 0.125 mm), and very fine sand plus mud (< 0.125 mm), according to a scale adapted from Brown and McLachland 33 .
Two sediment cores per site (i.e. one core from a mooring scar and one from adjacent meadows) were analyzed for 210 Pb concentration to determine recent (ca. 100 years) sedimentation rates. 210 Pb was determined through the analysis of 210 Po by alpha spectrometry after addition of 209 Po as an internal tracer and digestion in acid media using an analytical microwave 34 . The concentrations of excess 210 Pb used to obtain the age models were determined as the difference between total 210 Pb and 226 Ra (supported 210 Pb). Concentrations of 226 Ra were determined for selected samples along each core by low-background liquid scintillation counting method (Wallac 1220 Quantulus) adapted from Masqué et al. 35 . These concentrations were found to be in agreement with the concentrations of total 210 Pb at depth below the excess 210 Pb horizons. Mean sedimentation rates over the last 100 years were determined using the CF:CS model 36 .
Data analyses. Buoys attached to boat moorings at Stark Bay and Thomson Bay were digitized from high-resolution aerial imagery obtained in October 2009 (NearMap, http://au.nearmap.com) using the Geographic Information System (GIS) software ESRI ArcGIS. The bays were initially divided into a 20 × 20 m grid. Three coverage types were identified within each grid cell: seagrass meadows, bare sediments and rock/reef structures, which were digitized at a 1:25 scale. The total area of seagrass meadows, bare sediments, reef structures, and mooring induced scars was then calculated for each bay (Fig. 1).
The area of mooring-induced scarring was calculated for each digitized buoy to a Euclidean buffer of 10 meters. A radius of 10 m was chosen based on the average drifting lengths of the buoys with respect to their anchoring point in the seafloor at our study sites, and encompasses the length of mooring chain and possible dragging into the sediment. Currents and waves displace the buoy from its anchoring point (i.e. the chain attached to the buoy is not perpendicularly aligned to the seafloor) and we estimated (from aerial imagery) that the drifting lengths within the study area digitised average 10 m radius. As such there is an assumption that bare areas within this buffer are mooring induced, and that bare areas outside this buffer are not mooring induced. The spatial analyst toolbox within ArcMap was used to calculate the 10 m intersect area and determine the portion of any 'bare sediment' polygon that was located within 10 m of the digitized buoys (Fig. 7).
To allow direct comparisons among treatments and sites, the standing stocks per unit area (cumulative stocks; kg DW m −2 ) were standardized to 50 cm-thick deposits (i.e. average length of the sediment cores sampled; extrapolated when necessary). The length of core barrel inserted into the sediment and the length of retrieved seagrass sediment were recorded in order to correct the core lengths for compression effects and all variables studied here are referenced to the corrected, uncompressed depths. The mooring-induced loss of C org stocks was determined from the difference between the average accumulated stocks of C org in seagrass meadows and scar areas. Sediment accumulation rates (g DW m −2 y −1 ) for the last century were estimated from the 210 Pb age models. Accumulation rates of C org were estimated by dividing the inventories in the 50 cm-thick sediments by the average sediment accretion rate derived from 210 Pb (i.e. assuming constant sediment accretion rates through the core). For the three replicate cores, which were not 210 Pb dated, it was assumed they had the mean accumulation rate of the dated core at the site and treatment (i.e. mooring scar and undisturbed meadow).
All analyses were performed using Generalized Linear Mixed Model procedures in SPSS v. 14.0. A Generalized Linear Mixed Model was used to take into account the potential non-independence of samples taken at different depths within the same core, since depth is a proxy for time in the cores. Given the spatial separation of cores within meadows (hundreds of meters) we considered the cores themselves to be spatially independent. All response variables (bulk density, C org and CaCO 3 inventories, δ 13 C signatures and sediment grain size fractions were square-root transformed prior to analyses and had homogenous variances. Treatment (meadows vs scars) and sediment depth (18 depths along 50 cm-thick sediment cores) were treated as fixed factors and study site (Thompson Bay vs Stark Bay) was included as a random effect in all statistical models (probably distribution: normal; link function: identity).