Reduced carbon sequestration in a Mediterranean seagrass ( Posidonia oceanica ) ecosystem impacted by fish farming

We studied the relationship between sediment nutrient enrichment and carbon sequestration, using the ratio of gross primary production to respiration (P/R), in a fishfarming impacted and an unaffected Mediterranean seagrass (Posidonia oceanica) ecosystem in the Aegean Sea, Greece. Carbon (C), nitrogen (N) and phosphorus (P) sedimentation, nutrient pools in sediment and dissolved nutrients in pore water were significantly and positively intercorrelated, indicating close linkage between sedimentation and sediment nutrient pools in seagrass meadows. C, N and P sediment pools were significantly enhanced in the impacted meadow throughout the year, even during winter when fish farming activity was low. In the impacted sediment, the increase in C and N was higher than P, reflecting a faster remineralization and uptake of P than C and N. The ecosystem P/R ratio decreased exponentially with sediment nutrient enrichment. Threshold values are given for C, N and P sedimentation rates and sediment pools, and for N and P concentrations in pore waters, after which P/R ratio in the seagrass meadow decreases below 1, indicating a shift from autotrophy to heterotrophy with sediment nutrient enrichment. Such a regime shift indicates a loss of storage capacity of the seagrass ecosystem, jeopardizing the key role of P. oceanica as a carbon sink in the Mediterranean.


INTRODUCTION
Seagrass ecosystems are being increasingly considered to be amongst the largest carbon (C) sinks in the ocean (Nellemann et al. 2009).It has recently been estimated that the global mean gross primary production (GPP) to respiration (R) ratio (GPP/R, hereafter simply P/R) of a seagrass meadow is 1.55 (± 0.13 SE), with 63% of available data on seagrass meta bolism pointing to P/R ratios >1 (Duarte et al. 2010).Since a P/R ratio equal to 1 indicates metabolic balance, the majority of seagrass ecosystems sequester C in excess of consumption.A significant part of this excess is stored in seagrass sediments (Duarte & Cebrian 1996), where it can remain for millennia (Mateo et al. 2006), and seagrass meadows bury on average 48 to 112 Tg C yr −1 , which renders them hotspots for carbon sequestration (Kennedy et al. 2010).Duarte et al. (2010) estimated that the mean P/R ratio of a Posidonia oceanica meadow is 1.78, indicating that the dominant seagrass in the Mediterranean (Gobert et al. 2006) represents a crucial C sink for the basin.P. oceanica meadows constitute 1 to 2% of the Mediterranean sea bottom, usually extending up to 40 m depth and covering a total area of 50 000 km 2 (Bethoux & Copin-Montegu 1986).These meadows are often millenary years old (Mateo et al. 1997) and are considered very important for coastal protection and conservation, being hotspots for biodiversity (Hemminga & Duarte 2000).
Undisturbed sediments colonized by Posidonia oceanica are characterized by high carbon but low nitrogen (N) and phosphorus (P) deposition, with maximum values reported in spring (Gacia et al. 2002), and slow, mainly aerobic, mineralization rates (Barron et al. 2006).However, the annual sedimentary input of N and P in undisturbed P. oceanica meadows has been reported to equal the annual demand for growth (Gacia et al. 2002).Hence, sedimentation constitutes a significant part of nutrient input to the meadow, through mineralization of sedimentary organic matter which supplies the seagrasses with pore water nutrients necessary for growth.P. oceanica seagrass seems to buffer nutrient limitation on an annual basis by nutrient storage and translocation within the clone (Alcoverro et al. 2000), being well adapted to the oligotrophic regime in the Mediterranean.
Nutrient enrichment of the coastal zone (Nixon 1995) has resulted in gross decline of Posidonia oceanica meadows across the Mediterranean (Boudou resque et al. 2009).Fish farm discharges, in particular, cause severe damage to P. oceanica meadows growing nearby (Delgado et al. 1999, Dimech et al. 2002, Marbà et al. 2006, Pergent-Martini et al. 2006, Holmer et al. 2008, Apostolaki et al. 2009, Ruiz et al. 2010).When extending near fish cages, P. oceanica meadows receive a significant amount of particulate organic matter discharged from these cages (Holmer et al. 2008).Available studies show that excessive nutrient inputs are deposited beneath the cages (Sarà et al. 2004, Holmer et al. 2007), increasing the sediment (Apostolaki et al. 2007) and the pore-water (Cancemi et al. 2003) nutrient pools.Redox potential is usually decreased, while sulfate reduction rate is stimulated (Holmer et al. 2003a), leading to an in crease of sulfide pools in both the sediment (Pérez et al. 2007) and pore water (Calleja et al. 2007).Sulfides can invade seagrass roots (Frederiksen et al. 2008), inducing seagrass decline (Borum et al. 2005).
Díaz-Almela et al. ( 2008) have related sediment nutrient enrichment from fish farm effluents to an increase in shoot mortality and a decrease in shoot recruitment in Posidonia oceanica meadows across the Mediterranean.It is expected that nutrient enrichment from fish farm discharges also leads to a decrease in seagrass metabolism.A decrease in the metabolism of the seagrass ecosystem has been shown in case of organic loading (Holmer et al. 2004), low light availability (Gacia et al. 2005, Calleja et al. 2006) and fertilization from seabirds (Herbert & Fourqurean 2008).P. oceanica ecosystems (i.e.composed of the seagrass and the associated epiphytes, as well as benthic microalgae, benthic fauna, bacteria and phytoplankton) growing near fish cages also show a decrease in both GPP and R (Apostolaki et al. 2010b).However, it is still unknown whether this decrease is related to nutrient enrichment of the sediment.In fact, the relationship between sediment nutrient enrichment from fish farm effluents and ecosystem meta bolism has never been studied before for any seagrass species.
The temporal variation of organic matter from fish farms released and deposited on nearby sediment on an annual scale is still unknown.Most studies have been conducted only during summer (Pergent et al. 1999, Ruiz et al. 2001, Cancemi et al. 2003, Holmer et al. 2004, Sarà et al. 2004, Apostolaki et al. 2007, Holmer & Frederiksen 2007, Holmer et al. 2008).
Research effort has been focused on the warm period, as fish farm production and feed supply are in creased during the summer, causing enhancement of fish farm effluents in the same period (Karakassis et al. 1998).Yet, fish farm production continues through out the year, constantly releasing effluents to the surrounding environment.It can be expected that seasonal fluctuation in sediment nutrient enrichment might interfere with the annual cycle of seagrass metabolism, jeopardizing the key role of Posidonia oceanica ecosystems in carbon sequestration and cycling in the Mediterranean.
The present paper aims at examining whether the capacity of a Posidonia oceanica ecosystem to sequester carbon, using the ecosystem P/R ratio as proxy, is modified when sediment is nutrient enriched from fish farm effluents.To this end, the temporal variation of C, N and P sedimentation rate, of C, N and P pools in the sediment and of N (ammonium, NH 4 + ) and P (phosphate, PO 4 −3 ) concentration in the pore water were studied in a fish-farming impacted and in an unaffected P. oceanica meadow, during one year.C, N and P sedimentation rates, nutrient pools in sediment and dissolved nutrients in pore water were correlated with the ecosystem P/R ratio.Ecosystem P/R ratios were calculated from GPP and R previously published by Apostolaki et al. (2010b).

Sampling strategy
The study was conducted in a Posidonia oceanica meadow in Sounion (37°39.55'N, 23°58.24'E), Aegean Sea, Greece, from June 2006 until April 2007.The site was characterized by coarse sand (0.45 to 0.90 mm mean diameter pore size), a low percentage of silt and/or clay (4.63 to 4.83%), oxic conditions (292 to 353 mV redox potential) (Apostolaki et al. 2007) and 5.5 cm s −1 mean bottom current speed (Holmer et al. 2007).At the time of the study, the fish farm had been operating for 10 yr, producing on average 400 tons yr −1 of gilthead sea bream Sparus aurata and sea bass Dicentrarchus labrax.Ten years before the present study, in 1996 when the fish farm had just been set up, the meadow ex tended up to the edge of the cages (I.Karakassis pers.obs.), whereas in 2006, the meadow had re gressed ~20 m from the edge of the cages, showing clear signs of disturbance due to fish farm effluents.Indeed, a series of abiotic (water column and sediment nutrients) and biotic (seagrass shoot density, biomass and growth) variables revealed that fish farming activities had affected the meadow (Holmer et al. 2008 and references therein), which showed decreased production and increased mortality 20 m away from the cages (Marbà et al. 2006, Díaz-Almela et al. 2008, Apostolaki et al. 2009).Hence, we selected one station (hereafter called 'cages') which was located as close to the fish cages as possible (around 20 m away) at a depth of 16 m.Although Ruiz et al. (2010) detected a dispersion of farm wastes over several kilometres, a dialysis bag experiment at the study site showed that nutrient concentration in the water column decreased significantly 500 m away from the fish cages (Pitta et al. 2009), implying that the spatial extent of fish-farming impact in our case is smaller.Furthermore, Holmer et al. (2008) analyzed a series of variables in the water column, sediment, seagrass and benthic fauna from 4 sites, including our study site, and suggested that fish-farming impact is evident up to 400 m from the cages.Hence, we selected a reference station (hereafter called 'control') outside the area affected by fish farm waste, located 1 km away from the cages and at a depth of 14.5 m.The site was visited every 2 mo (from June 2006 until April 2007).
Benthic sediment traps (5 replicates) were deployed by divers during each sampling event for 48 h, as described by Gacia et al. (1999), for the determination of particulate C, N and P sedimentation rates.The traps were deployed just above the seagrass canopy and along the direction of the main water currents.Each trap consisted of a 20 ml cylindrical centrifugation tube with an aspect ratio of 5 (i.d.= 16 mm), in order to prevent internal resuspension.Sediment cores (3 replicates, i.d.= 4.5 cm each) were collected for the determination of surface (0 to 2 cm) total organic C, N and P concentrations.Pore water samples (3 replicates, 60 ml each) were taken using an acid-washed syringe with a perforated sipper which collected pore water from a sediment depth of 1 to 10 cm for determination of ammonium (NH 4 + ) and phosphate (PO 4 3− ).

Laboratory analysis
In the laboratory, the particulate content of each sediment trap was filtered with pre-combusted and pre-weighted filters (Whatman GF/F).Each filter was dried at 60°C to constant weight and weighted.C and N in filters and sediment samples were measured using a Perkin Elmer 2400 CHN Analyzer (Hedges & Stern 1984).Total P concentration in filters and sediment samples was determined using hydrochloric acid (Burton & Riley 1956) and then analysed colorimetrically using molybdate solution (Strickland & Parsons 1972).Pore water samples were filtered (Whatman GF/F), kept frozen and later analyzed for NH 4 + (Ivan<i< & Deggobis 1984) and PO 4 −3 (Strickland & Parsons 1972).Particulate carbon, nitrogen and phosphorus sedimentation rate SR i (mg m −2 d −1 ) in each sampling event (i) was estimated as: where W i is the weight of the particulate material collected in the sediment trap (in mg), C i is the concentration of carbon, nitrogen or phosphorus (as % dry wt), d is the internal diameter (in m) of the tube and t is the time of deployment (in d).
Concurrent with sediment sampling, ecosystem P/R was calculated for each sampling event and station.Estimates for GPP and R were obtained from Apostolaki et al. (2010b).Briefly, GPP and R were determined following oxygen evolution for 24 h using in situ benthic chambers.Benthic chambers (~5 l) consisted of a PVC cylinder, a gas-tight polyethylene plastic bag and a sampling port to collect water samples.These chambers are widely used in metabolism studies (Duarte et al. 2002, Barrón et al. 2004) because they allow conduction of time-course experiments without dilution, continuous stirring, or gaseous head-space (Hansen et al. 2000), while they allow turbulence in the interior of the bag ensuring mixing (Barrón et al. 2006).The cylinders were slowly and carefully inserted in the sediment without cutting the rhizomes but rather pushing them down to the sediment, causing minimal sediment disturbance in comparison to other methods like laboratory incubations.Water samples were analyzed for dissolved oxygen concentration using the Winkler titration method.Oxygen units were converted to carbon units by assuming photosynthetic and respiratory quotients of 1.2 and 1.1, respectively (Oviatt et al. 1986).

Data analysis
Differences in C, N and P sedimentation rates, C, N and P concentration in the sediment, NH 4 + and PO 4 −3 pore water concentrations and ecosystem P/R ratio between the cages and control stations in different sampling events were assessed by means of a 2-way ANOVA.If significant differences (p < 0.05) occurred among sampling events, a Tukey's post hoc test was used to identify which events were different from each other.Regression analysis was used in order to detect possible relationships between the tested variables (which were log 10 transformed when necessary).

RESULTS
C sedimentation rate and sediment C concentration, N sedimentation rate, sediment N and pore water NH 4 + , and P sedimentation rate, sediment P and pore water PO 4 −3 were significantly and positively intercorrelated (Table 1).
Station (i.e.proximity to the cages) had a significant effect on C sedimentation rate, but the effect depended significantly on time of year (Table 2).The difference was due to the higher sedimentation rate at the cages station during the warm months (June, August and October; Table 2).In each sampling event, C sedimentation rate and sediment C were higher at the cages station than at the control station (Fig. 1).On an annual basis, C sedimentation rates were 4 times higher in sediments lo cated at the cages station than in those at the control station.Mean C concentration in the sediment beneath the cages was 14 times higher than that measured at the control station.
N sediment concentrations were significantly enhanced at the cages station compared to the control station throughout the year (Fig. 2, Table 2).N sedimentation rate differed significantly between stations, depending on time, which differentiated October from December samplings (Table 2).Annual N sedimentation rate was 3 times higher at the cages station than at the control station.Average N concentration in the sediment was 9 times and in the pore water 2 times higher compared to the control.
P sediment inputs and sediment concentrations were significantly higher at the cages station than at the control station throughout the year (Fig. 3, Table 2).Station significantly affected pore water P, but the effect depended on time; June differed from August, October and April samplings (Table 2).On an annual scale, P in sedimentation rate was 4-fold higher at the cages station compared to the control.The sediment and pore water concentrations were on average 1.3-and 3-fold, respectively, higher at the cages compared to the control.
Ecosystem P/R ratio (Table 3) was significantly lower at the cages station compared to the control station (2-way ANOVA, Station: F = 4.94, p = 0.036, Time: F = 1.39, p = 0.266, Station × Time: F = 0.49, p = 0.778).Annually, P/R ratio was 1.3 and 1.7 at the cages and control station, respectively.P/R ratio of the seagrass ecosystem decreased exponentially with increasing C, N and P sedimentation rates, C, N and P sediment concentrations and increasing N and P pools in pore water (Fig. 4).

DISCUSSION
The temporal fluctuation in C, N and P sedimentation rates (higher from June until October and lower in December) may reflect the seasonality in fish farm effluent release.Release of effluents is usually higher during the warm season in the Mediterranean (summer−autumn) as a result of the enhancement in fish farm production during this period (Karakassis et al. 1998).Similar studies in the Mediterranean basin have also reported higher sedimentation during summer, driven by loss of waste products from the fish cages, and subsequent sediment and pore water enrichment (Cancemi et al. 2003, Sarà et al. 2004, Apostolaki et al. 2007, Holmer & Frederiksen 2007, Holmer et al. 2008).In the present study, however, we found that also during winter, nutrient de position and accumulation in the sediments and pore waters were higher at the cages station compared to the control station.This suggests a continuous nutrient enrichment around the cages through out the year, or at least that the reduction in feed supply during the cold period is not enough to allow a recovery of the benthic environment.
C pools in sedimentation and sediment, N pools in sedimentation, sediment and pore water and P pools in sedimentation, sediment and pore water were positively correlated, indicating a close linkage be tween C, N, P inputs and sediment nutrient pools in seagrass meadows, as suggested by de Boer ( 2007).The increase in C and N sediment pools at the cages station was higher than the increase in P, possibly reflecting a faster remineralization and uptake of P than C and N. The lack of P accu mulation compared to C and N suggests that seagrass nutrient dynamics respond differently to nutrient enrichment.This result supports previous findings showing that nutrient limitation in seagrasses depends on specific plant features and local conditions (Romero et al. 2006).N concentration in seagrass leaves at the control station (1.7% dry wt, obtained by Apostolaki et al. 2009) was close to the threshold of 1.8% dry wt for no N limitation in seagrass growth (Duarte 1990), and N enrichment from fish farming would probably not stimulate uptake of N by the seagrasses.On the contrary, P concentration in seagrass leaves at the control station (0.08% dry wt, obtained by Apostolaki et al. 2009) was substantially lower than 0.2% dry wt, the critical P content below which seagrasses are P deficient (Duarte 1990).It can thus be expected that Posidonia oceanica acquires P rapidly when this nutrient is available (Agawin et al. 1996, Alcoverro et al. 1997, Udy & Dennison 1997).Indeed, increased P availability is often measured under nu trient enriched conditions (Apostolaki et al. 2010a,b).Under such conditions, sulfate reduction rates in the sediments are often intensified by the organic enrichment (Holmer & Frederiksen 2007).Increased sediment sulfide production releases P bound to oxidized iron (Heijs et al. 2000) and P can then be taken up from pore water or the water column by the sea- ) between cages and control stations, at different sampling events grasses.Moreover, dissolution of carbonate sediments by organic acid exudates from seagrasses liberates P from the sediments (Long et al. 2008).Therefore, the increased P availability in fish-farm sediments possibly induced uptake of P by seagrasses in excess of their metabolic demands, a process known as 'luxury consumption' (Romero et al. 2006) which has been reported for seagrasses (Ferdie & Fourqurean 2004) and other primary producers (Romero et al. 2006).Luxury uptake and storage of P in seagrass tissue possibly resulted in lower accumulation of P in sediment at the cages station.
The increase in seagrass leaf P concentration at the cages station did not lead to increased seagrass standing crop (annual mean 38.05 and 166.62 g m −2 at cages and control station, respectively; obtained from Apostolaki et al. 2009).On the contrary, seagrass standing crop decreased with increasing leaf P content (linear regression, intercept = 511, slope = −5803, R 2 = 0.73, p < 0.001), implying that the excess nutrient supply was not capitalized into shoot growth.This excess was not capitalized into ecosystem production either, as shown by the reduced ecosystem P/R ratio upon nutrient enrichment.Holmer et al. (2004) have measured a similar decrease in net community production (NCP) of a Posidonia oceanica meadow impacted by organic inputs.Deposition and accumulation of organic matter and nutrients in seagrass sediments (Díaz-Almela et al. 2008, Pérez et al. 2008) reduce seagrass photosynthetic capacity and subsequent production of carbon re serves (Burkholder et al. 1994, Ruiz et al. 2001, Invers et al. 2004).An increase in sulfide pools in the sediments along with a decrease in NCP has been shown in trop ical seagrass meadows (Gacia et al. 2005, Calleja et al. 2006), suggesting interference of sulfides in seagrass metabolism.Increasing sulfide pools in the sediment (Pérez et al. 2007) and pore water (Calleja et al. 2007), lead to seagrass decline (Borum et al. 2005, Frederiksen et al. 2007, 2008, Mascaró et al. 2009).
Excess C originated from fish farm discharges is not likely to counterbalance the loss of C production in the impacted meadow.The biochemical composition of organic matter accumulated in the sediment be neath fish cages is different, being enriched in organic C and lipids (Pusceddu et al. 2007).Bacterial consumption of C was probably enhanced (Danovaro et al. 2003), accelerating C cycling (López et al. 1995(López et al. , 1998)).Epiphytes, which are known to contribute sig- Sediment C (% dry wt) Fig. 1.Carbon pool (mean ± SD) in sedimentation and sediment at each station and sampling event nificantly to seagrass ecosystem production (Borowitzka et al. 2006, Ouisse et al. 2010), were not related to either C production or consumption (Apostolaki et al. 2011).In contrast, NCP decreased with decreasing shoot density (annual mean 145 and 312 shoot m −2 at cages and control station, respectively; ob tained from Apostolaki et al. 2009) close to the fish cages (linear regression, intercept = −1.22,slope = 0.10, R 2 = 0.52, p < 0.01), indicating declining metabolic activity of the seagrass ecosystem upon sediment nutrient enrichment.It is suggested, therefore, that the decrease in P/R ratio is mainly driven by the decline of the seagrass, which is the key species of the ecosystem.
Our results indicate that ecosystem P/R ratio decreased exponentially in response to increasing nutrient enrichment, and was below 1 at the cages station, causing a threshold response with an ecosystem shift from autotrophy to heterotrophy.Several ecosystems show threshold responses to severe environmental change and modulate between alternative states (Andersen et al. 2009).The present results suggest that at this particular site a shift from autotrophy to heterotrophy occurs when sedimentation rates exceed 391, 34 and 34 mg m −2 d −1 of C, N and P, respectively, when sediment pools exceed 10, 0.4 and 0.03% dry wt of C, N and P, respectively, and when NH 4 + and PO 4 −3 in pore water surpass 6 µM.These thresholds should be extrapolated with caution to other Posidonia oceanica ecosystems, since they derive from measurements in a single meadow.Nonetheless, Holmer at al. (2008 and references therein) studied the effect of fish farming on this and 3 more P. oceanica meadows (in Cyprus, Italy and Spain) by measuring a series of abiotic (water column and sediment nutrients) and biotic (benthic fauna and seagrass) variables.They concluded that fish farm discharges had similar impacts on the seagrass ecosystem near the fish cages, despite the fact that site and fish farm characteristics differed (depth, temperature, bottom current, farm size, annual fish production, food input and waste production).Moreover, Díaz-Almela et al. ( 2008) suggested a similar threshold for P sedimentation (50 mg P m −2 d −1 ), after which shoot mortality accelerates and shoot recruitment declines substantially.Therefore, despite a lack of site replication, our results appear indicative of fish-farming impact on carbon sequestration in P. oceanica communities.
A regime shift to heterotrophy with the continuation of sediment nutrient enrichment from fish farm discharges would eventually cause carbon sequestration to cease in impacted Posidonia oceanica sediments.Healthy P. oceanica ecosystems are known to ) Fig. 3. Phosphorus pool (mean ± SD) in sedimentation, sediment and pore water at each station and sampling event se quester annually more carbon than unvegetated (i.e void of seagrass or macroalgae) communities (Barrón et al. 2006, Barrón & Duarte 2009), implying that considerably more carbon is stored in P. oceanica than in bare sediments.In our case, the control meadow showed 9 times higher annual NCP than the adjacent unvegetated community (Apostolaki et al. 2010b), emphasizing the high C sink capacity of this meadow.Duarte et al. (2010) suggested that 1/3 of the global carbon sink capacity of seagrasses has been lost due to a global decrease of 29% in seagrass area since seagrass areas were initially recorded in 1879 (Waycott et al. 2009).Furthermore, P. oceanica is a slow-growing species with a very slow recovery rate (Marbà et al. 2002), meaning that its full recovery is usually considered irreversible in human time scale (Marbà et al. 2005).It is therefore essential to restore and protect P. oceanica meadows, in order to fulfil their key role in carbon cycling and storage in the Mediterranean, by adopting efficient regulation of human pressures such as fish farming.Monitoring of sedimentation, as also suggested by Holmer et al. (2008), as well as sediment and pore water nutrient pools are effective and reliable tools for managing seagrasses near fish farms.This is particularly important to consider in order to achieve sustainable fish farming activities along the Mediterranean coasts, since the future of aquaculture is expected to increase significantly (Duarte et al. 2009) to meet the increasing demand of animal proteins for the rapidly increasing human population.

Fig. 2 .
Fig. 2. Nitrogen pool (mean ± SD) in sedimentation, sediment and pore water at each station and sampling event

Table 1 .
Linear regression coefficients and significance levels between carbon, nitrogen (N) and phosphorus pools in sedimentation, sediment and pore water.The relationship between variables is described by the equation: log 10 y = a + b × log 10 x, except the N sedimentation rate and pore water relationship which is described by the equation: log 10 N sedimentation rate = 1/ (a + b/log 10 Pore water N)