Water quality, isoscapes and stoichioscapes of seagrasses 1 indicate general P limitation and unique N cycling in shallow 2 water benthos of Bermuda

19 Striking spatial patterns in stable isotope ratios (isoscapes) and elemental ratios (stoichioscapes) 20 of seagrass leaves and the water column nutrients indicate general P-limitation of both water 21 column and benthic primary productivity on the Bermuda Platform, and they highlight the role of 22 the Bermuda Islands as a source of N and P. We found consistent differences among the four 23 seagrass species ( Syringodium filiforme , Thalassia testudinum , Halodule sp. and Halophila 24 decipiens ) in the N, P, δ 13 C and δ 15 N of leaf tissues. The δ 15 N of seagrass leaves was especially 25


Introduction
Water quality in coastal environments can be episodic and ephemeral, such that it requires high frequency sampling or a long time series to adequately characterize spatial and temporal patterns in water quality.However, the long life span of seagrass tissues, relative to the temporal scale of environmental variability, allows seagrasses to integrate this variable signal of environmental quality through time.Single or low frequency samples of seagrasses can clearly indicate prevailing water quality conditions as well as indicate the effects of extreme events.Thus, seagrasses are biological sentinels of conditions in the coastal zone, or "coastal canaries" (Orth et al., 2006).
Seagrasses are rooted plants; for a single location they can provide an integrated picture of water quality attributes, such as nutrient availability, chlorophyll a (Chl a), and turbidity that affect their growth.For example, ratios of the abundance of the major elements of carbon (C), nitrogen (N) and phosphorus (P) in seagrass tissues provide indications of the relative availability of light, N and P (Atkinson and Smith, 1983;Abal et al., 1994;Grice et al., 1996).The stable C isotopic composition of seagrass leaves serves as an indicator of light availability (Campbell and Fourqurean, 2009;Cooper, 1989;Cooper and DeNiro, 1989;Udy and Dennison, 1997) and stable N isotopic composition is used as a tracer of envi-Published by Copernicus Publications on behalf of the European Geosciences Union.ronmental δ 15 N values, which can indicate human sewage pollution (Cabana and Rasmussen, 1996) as well as internal cycling within the ecosystem (Fourqurean et al., 1997(Fourqurean et al., , 2005;;Lamb et al., 2012).Landscape-scale spatial patterns in stable isotopic composition (termed the "isoscape"; Bowen, 2010) and in elemental content (which we can similarly call "stoichioscapes") of the seagrass "canaries" can be very useful in elucidating ecosystem-scale patterns in sources of materials and processes affecting elemental cycling (Burkholder et al., 2013;Fourqurean et al., 1992aFourqurean et al., , 1997Fourqurean et al., , 2005Fourqurean et al., , 2007;;Fourqurean and Zieman, 2002).
Seagrasses are a polyphyletic group of plants and therefore it should not be surprising that they have different morphology, growth rate and biochemical pathways that can lead to interspecific differences in stoichiometry and elemental content (Campbell and Fourqurean, 2009;Fourqurean et al., 2007), so differences in species distribution could complicate interpretations of isoscapes and stoichioscapes if these species differences are not appreciated.Additionally, interspecific difference in elemental and isotopic composition of seagrasses can complicate stable isotope-based food web studies, since even when such studies are conducted in areas with relatively high seagrass species diversity, a few seagrass samples are often assumed to represent the possible range in isotopic content of all seagrasses within the range of the herbivores being studied (e.g., Davis et al., 2014, and Vander Zanden et al., 2013, as recent examples).The spatial pattern in the isoscape of a single seagrass species and interspecific variation among seagrasses could actually provide refined signals to be used to determine foraging location and food preferences of herbivores.Such approaches have proven to be informative in terrestrial ecosystems (Rubenstein and Hobson, 2004) and, when coupled with knowledge of marine species distributions and isoscapes, could prove equally useful in the ocean (McMahon et al., 2013).
Since the 1990's, there has been a loss of about 23 % of the spatial extent of seagrass meadows across the Bermuda Platform, and the density of seagrass in many of the remaining meadows has decreased from historic conditions (Murdoch et al., 2007).Interestingly, most of this decline has occurred in the offshore areas, distant from any direct anthropogenic nutrient sources.While it has become clear that grazing by the large population of green sea turtles that use Bermuda as developmental juvenile habitat has contributed to the decline (Fourqurean et al., 2010), the role that nutrient availability plays in controlling seagrass distribution and health across the platform is still unknown.Globally, the majority of seagrass declines documented since the beginning of the 20th century have been related to decreases in water quality caused by anthropogenic nutrient loading (Short and Wyllie-Echeverria, 1996).This suggests that understanding the current state of water quality and its relationship to seagrass vitality is important for a more complete understanding of the recent trajectory of Bermuda's seagrasses.In addition, given their restriction to areas of very high bottom light availability compared to other more tropical areas (Manuel et al., 2013), the potential exists for even small changes in water quality to impact the distribution of seagrasses.The stoichioscapes and isoscapes of Bermuda's seagrasses should reflect the mean state of the environment and provide clues about the factors controlling their distribution and trajectory.Furthermore, species-specific isotopic differences and isoscapes could also provide tools for understanding the movements of the green turtle, herbivores that may be contributing to seagrass declines.
In this paper, we present an extensive data set describing the water quality of the Bermuda Platform.If land-based sources of nutrients are impacting the marine environment of the Bermuda Platform, then we hypothesize there will be a distinct spatial pattern of decreasing concentrations of inorganic nutrients, organic matter and Chl a with distance from the shoreline.Furthermore, we test whether the isoscape and stoichioscape of seagrasses across the Bermuda Platform are representative of this long-term water quality data set.Specifically, we hypothesize that the C : N : P ratio of seagrass will approach seagrass Redfield ratios (SRR, defined as N : P of 25-30 for plants growing with a relative abundance of both N and P compared to light availability; Atkinson and Smith, 1983;Duarte, 1990;Ferdie and Fourqurean, 2004) and, given the importance of light availability in determining seagrass distribution across the Bermuda Platform (Manuel et al., 2013), we also expect seagrass tissues to have C : N : P near SRR and a decrease in δ 13 C of the seagrass in low light environments.Because of the influence of human wastewater on the isotopic content of nitrogen in sewage effluent (Cabana and Rasmussen, 1996), we expect water quality to be a determinant of δ 15 N of seagrasses and for the isoscape of seagrass δ 15 N to trend from high values close to shore to lower values further from human influence.Also, given our knowledge of the distribution of seagrass species across the Bermuda Platform (Manuel et al., 2013) we expect that species identity could influence our interpretations of seagrass isoscapes and stoichioscapes.

Study site
Despite its high latitude, the shallow platform surrounding the mid-Atlantic island chain of Bermuda (32.4 • N) has coral reefs, seagrasses, and macroalgal beds formed by tropical species (Coates et al., 2013).Coral cover on Bermuda's reefs is very high compared to most locations in the Caribbean (Sheppard, 2013;Riegl and Dodge, 2008).Seagrass meadows are widespread on the Bermuda Platform and are found at about 24 % of randomly selected sites inside the fringing coral reef (Manuel et al., 2013); many of the larger beds are in nearshore locations.Syringodium filiforme is the most often encountered seagrass, followed in order of prevalence by Halophila decipiens, Thalassia testudinum and Halodule sp.(there is uncertainty about the taxonomic identity of the Halodule that is currently found in Bermuda; Manuel et al., 2013).Owing to the relatively high latitude of Bermuda compared to the rest of the range of these tropical plants, seagrasses are restricted to areas of shallower water and higher mean light transmittance to the bottom in Bermuda; this is likely a consequence of low insolation during winter months (Manuel et al., 2013).
Phytoplankton primary production (PP) and biomass in the Sargasso Sea surrounding Bermuda are limited by the supply of P to the photic zone (Ammerman et al., 2003;Lomas et al., 2004;Salihiglu et al., 2008).It is believed that PP in the shallow water benthos on the Bermuda Platform is also P-limited.Given the nutrient-limited state of the marine ecosystem surrounding Bermuda, there exists the potential for land-based sources of nutrients to influence those marine ecosystems.In one embayment on the north shore of Bermuda, experimental work indicates a general P-limitation of seagrass growth offshore with a lessening of that P-limitation inside the eutrophic embayment (McGlathery et al., 2001) while at the same locations macroalgae can be N-limited (McGlathery et al., 1992).
The total land area of Bermuda is quite small, just 53.6 km 2 , but the shoreline of the more than 150 islands is relatively long, ca.296 km (Meyer et al., 2015), and the nearshore area is extensive.Bermuda is densely populated with no comprehensive wastewater collection system; many households dispose of wastewater in septic systems or cesspits that sit in a highly vuggy karst bedrock.Molecular markers of human enterococci and coprostanol indicate that some sewage makes its way through the groundwater to the nearshore ocean (Jones et al., 2011).This contaminated groundwater contains high concentrations of nutrients (Simmons and Lyons, 1994) and some eutrophication of enclosed, P-limited bays has been demonstrated (Lapointe and O'Connell, 1989).

Seagrass surveys, collection and analysis
We surveyed seagrass abundance and collected seagrass leaves for analysis of elemental content and stable isotope ratios of C and N from 516 sites distributed across the Bermuda Platform during the summer months of 2006-2008 using a stratified-random protocol to locate the sites (see Manuel et al. (2013) for a detailed description of seagrass abundance on the Bermuda Platform).In addition, we used summer 2007 data from 17 permanent monitoring sites established in existing seagrass beds (Fig. 1).We estimated seagrass abundance at each of these 533 sites using modified Braun-Blanquet surveys composed of 10 randomly placed 0.25 m 2 quadrats along a 50 m transect (Fourqurean et al., 2001).
Seagrasses were present at all of the 17 permanent monitoring sites, but only at 168 of the 516 random sites.If seagrasses were present, multiple intact short shoots of each sea-grass species were haphazardly collected along the 50 m transect line.In order to assure sufficient samples for all subsequent analyses, we collected five short shoots of the larger T. testudinum, 30 short shoots of S. filiforme, 40 short shoots of Halodule sp. and ca.300 leaves of H. decipiens.In the lab, seagrass shoots were washed free of sediment, and all green leaves from each species at each site were pooled into a common sample.Leaf material was gently cleaned of epiphytes using a razor blade, dried to a constant weight at 50 • C and ground to a fine powder.C and N content of leaf material was analyzed in duplicate using a CHN analyzer (Fisons NA1500).Leaf P content was determined through a dry oxidation and acid hydrolysis extraction followed by a colorimetric analysis (Fourqurean et al., 1992b).Elemental content was calculated on a dry weight basis (i.e., mass of element/dry weight of sample × 100 %); all elemental ratios were calculated on a molar basis.Stable isotope ratios of C and N were analyzed using standard elemental analyzer isotope ratio mass spectrometer (EA-IRMS) procedures.The EA was used to combust the organic material and to reduce the formed gases into N 2 and CO 2 , which were measured on a Finnigan MAT Delta C IRMS in a continuous flow mode.Isotopic ratios (R) were reported in the standard delta notation (‰ ): δ (‰) = [(R sample /R standard )−1] × 1000.These results are presented with respect to the international standards of atmospheric N (AIR, N 2 ) and Vienna Pee Dee Belemnite (V-PDB) for C. Analytical reproducibility of the reported δ values, based on sample replicates, was better than ±0.2 ‰ for δ 15 N and ±0.08 ‰ for δ 13 C.

Water quality measurements
Water quality parameters were measured on a monthly basis at the 17 permanent monitoring sites, over a period of 1-3 days.For a detailed description of the water quality methods and data, please see the Supplementary Material.As water quality data were collected monthly at each site for the period 2007-2012, the period of water quality data collection began 1 year later and continued 3 years after our collection of seagrass elemental and isotopic content data.We tested the water quality data for long-term monotonic trends from 2006 to 2012; when these trends were lacking, we assumed that the means of the water quality data could be applied to the analyses of the relationships among seagrass and water quality variables and we then subsequently compared the average water quality data to seagrass data from the three years of record.

Submarine light regime in seagrass beds
At each site from which seagrass leaves were collected, we estimated the fraction of solar radiation reaching the bottom, as a percentage of sunlight entering the water column (%I 0 ), by interpolating the maps of %I 0 in Manuel et al. (2013).These maps were generated using site-specific water depth and an interpolated map of attenuation coefficients, k d .As %I 0 is a function of both k d as well as depth, it more accurately represents the influence of light availability on seagrasses than would be evident by water depth alone.

Statistical methods
Prior to statistical analyses, data were tested for their adherence to assumptions of the statistical techniques; we used parametric methods if the data were normally distributed and homoscedastic.If data did not meet these assumptions we used non-parametric methods.We tested for long-term monotonic trends in the water quality data using linear regression with time as the independent variable.Relationships between water quality measurements and seagrass elemental and stable isotopic content were elucidated by analyzing bivariate correlations; a general lack of normality in the water quality data resulted in using Spearman's ρ rank correlation coefficient as a test of monotonic relationships among water quality variables.Seagrass elemental and isotopic content, as well as the water quality parameters N : P and %I 0 met the assumptions for parametric statistical analysis, so we assessed relationships with Pearson's correlation coefficient, r.Analysis of variance (ANOVA) with Student-Newman-Keuls post hoc tests were used to assess differences in elemental and isotopic content among seagrass species.Differences in elemental and isotopic content among co-occurring seagrasses were tested using paired t tests, pairing species within sites.
Spatial patterns in measured variables were visualized by creating contour maps using a point kriging algorithm assuming a linear variogram and isotropic conditions (SURFER 9, Golden Software, Inc, Golden, Colorado, USA).We used the water quality contour plots to estimate water quality at sites from which seagrass was collected, which were often different from the water quality stations, by sampling the water quality fields at the locations of the seagrass sampling.Relationships between the mean value of water quality variables, as assessed by this spatial interpolation of the mean data fields, and seagrass elemental and isotopic content was further explored using principal component analysis (PCA) with varimax rotation of the solution to facilitate interpretation of the composite variables (IBM SPSS Statistics 21, IBM Corporation, Armonk, New York, USA).Before PCA, all data were standardized by computing Z scores.

Water quality on the Bermuda Platform
Reflecting its small size and position in the central Atlantic Ocean, water quality surrounding Bermuda reflected the influence of the ocean endmember, with generally high water clarity and low nutrient concentrations (See Supplement).Linear regressions yielded no significant trends in time in platform-averaged average water quality measurements.
In comparison to N, P concentrations were very low; soluble reactive phosphorous (SRP) concentrations averaged 0.048 ± 0.001 µM and total phosphorus (TP) concentrations averaged 0.136 ± 0.002 µM.Both the dissolved inorganic nitrogen (DIN) : SRP and N : P ratios were much higher than Redfield, and the total organic carbon (TOC) : N ratio lower than Redfield, suggesting that, in general, the waters around Bermuda are P-limited.Chl a concentrations were positively correlated with temperature (Table 1), suggesting seasonal, summertime increases in phytoplankton biomass; high Chl a was also correlated with higher NO − 2 , NO − 3 , TP, SRP, TOC, dissolved organic carbon (DOC) and SiO 2 .On the contrary, there was no correlation between total nitrogen (TN) and Chl a (Table 1), further suggesting N availability does not generally limit phytoplankton around Bermuda.
The islands of Bermuda exhibited a marked influence over water quality on the platform.Station-averaged TN and TP generally declined with distance from the island (Fig. 1).The pattern in station-averaged N : P was more complex, with generally lower N : P (ca.130 : 1) in the enclosed waters of the sounds and harbors, compared to higher values in the more open waters of the platform.There were also strong gradients in TOC concentrations and Chl a across the platform, with higher values close to shore that decreased with distance north of the island (Fig. 2).

Seagrass elemental content and stoichiometry
Seagrass leaves exhibited a wide range of values for N and P content (Fig. 3, Table 2).N content ranged from 1.40 to 3.91 % and P content ranged from 0.058 to 0.465 %.When data were pooled from all sites by species, there were significant differences among species in N content (ANOVA, F 3.376 = 11.8, p < 0.001) and P content (ANOVA, F 3.376 = 18.4,p < 0.001).Post hoc tests indicated that there were no significant differences in N content between S. filiforme and H. decipiens, but T. testudinum and Halodule sp. were different from all other species.In contrast, there was more overlap among the species in P content: there was no significant difference between S. filiforme and Halodule sp.nor between Halodule sp. and T. testudinum, while P content of H. decipiens was distinct.While there was a significant difference in N : P among species (ANOVA, F 3.375 = 5.8, p = 0.001), H. decipiens was the only species with different N : P ratios from the other species.N : P of the larger-bodied seagrasses (S. filiforme, T. testudinum and Halodule sp.) were greater than 30 : 1, indicative of P limitation.N : P of the smaller, faster-growing, and low light-adapted H. decipiens indicated P limitation in approximately 80 % of the samples, but N : P was < 30 in about 20 % of the samples, suggesting potential N limitation at some sites.
Comparisons of global means of elemental content data (Table 2) do not fully capture the actual differences among species driven by their physiology and morphology, since the distributions of the species across the platform are dif-  ferent (Manuel et al., 2013) and there also were gradients in availability of N and P in the water column across the platform (Fig. 1).To control for this variation, we compared species pairwise based on co-occurrence at individual sampling sites.At the 66 sites where they co-occurred, S. filiforme had N content 0.23 % ± 0.06 % higher than T. testudinum (paired t = 3.9, df = 65, p < 0.001).S. filiforme had N content 0.22 % ± 0.06 higher (p < 0.001) than Halodule sp. and 0.39 % ± 0.14 % higher (p = 0.013) than H. decipiens.T. testudinum had significantly lower N content (by 0.14 % ± 0.06 %, p = 0.038) than Halodule sp.On a single occasion, T. testudinum was found co-occurring with H. decipiens.There were no significant differences in the N content of Halodule sp. and H. decipiens at the 15 sites where they co-occurred.There were no significant differences in the P content of T. testudinum, nor of S. filiforme (co-occurred at 66 sites, p = 0.151) or Halodule sp.(co-occurred at 39 sites, p = 0.139); however the P content of S. filiforme was 0.012 % ± 0.005 % higher than Halodule sp.(69 sites, p = 0.015).S. filiforme had a P content 0.046 % ± 0.015 % higher (19 sites, p = 0.005) than H. decipiens.Halodule sp., had a P content on average 0.045 % ± 0.013 % higher than H. decipiens (15 sites, p = 0.004).
For all possible two-species comparisons except S. filiforme and Halodule sp., seagrass species had distinct N : P ratios (p ≤ 0.02 for all comparisons).N : P of S. filiforme was 4.9 ± 1.4 greater than N : P of T. testudinum and 7.1 ± 2.7 less than H. decipiens.N : P of Halodule sp. was 6.6 ± 1.8 higher than T. testudinum, and N : P of Halodule sp. was 7.7 ± 2.9 less than H. decipiens.As there were no systematic differences in the N : P of leaves of S. filiforme and Halodule sp. and given the widespread distribution of these two species across the Bermuda Platform, we mapped the spatial pattern in N : P of these two species.When both species were present at a site, we averaged their N : P ratios.The resultant spatial pattern (Fig. 4) indicates broad-scale P limitation across the platform for these species (i.e., N : P > 30), with only a few locations very close to shore near the west end of the island and in an enclosed cove near the northeastern part of the island with N : P ratios indicative of light limitation.Interestingly, no sites where S. filiforme or Halodule sp.occurred had N : P < 25 that would indicate N limitation.
Given that there were no differences in the δ 15 N of S. filiforme, T. testudinum or Halodule sp., the spatial pattern in the δ 15 N from all sites where any of these species occurred indicates a striking spatial pattern in the δ 15 N of the seagrasses from across the platform (Fig. 6).δ 15 N on the platform distant from the Bermuda Islands was distinctly depleted in 15 N, with some areas exhibiting δ 15 N values as low as −10.1 ‰.As distance from land decreased, δ 15 N values increased with values higher than 8 ‰ in the semi-enclosed Hamilton Harbor and in the enclosed Harrington Sound.The spatial gradients in δ 15 N were very steep, with δ 15 N changing as rapidly as 1.8 ‰ km −1 from Hamilton Harbor to the center of the platform.

Relationships among seagrass elemental content and δ 15 N and δ 13 C values
Given the differences among species in N : P ratios and δ 13 C, we analyzed the relationships among elemental and stable isotopic composition of seagrasses by species (Table 3).For all seagrass species there was a significant positive correlation between C : N and C : P, but N : P was in all cases correlated with C : P and only weakly correlated with C : N for Halodule sp.(r = 0.243, p = 0.032).There were also strong positive correlations between C : N and δ 15 N for all species, which indicated that plants with low N content had relatively enriched δ 15 N values.With one exception, δ 13 C was positively correlated with both C : P and N : P, indicating that plants with low P content tended to have stable carbon composition more enriched in 13 C; for H. decipiens there was no significant correlation between δ 13 C and N : P.There were some exceptions to the general correlation patterns: for T. testudinum only, there was a significant positive correlation between C : P and δ 15 N, and N : P was significantly, negatively correlated with δ 15 N for S. filiforme and H. decipiens, but not for T. testudinum or Halodule sp.

Water quality as a predictor of seagrass elemental content and δ 15 N and δ 13 C values
There were no simple linear relationships between seagrass elemental and isotopic content and water quality that applied across all four seagrass species.There were differences in the  ranges of light availability at sites where the different species occurred: %I 0 at sites supporting T. testudinum averaged 70 ± 1 %, compared to 59 ± 2 %, 55 ± 2 % and 44 ± 2 % for S. filiforme, Halodule sp., and H. decipiens, respectively.At sites that supported S. filiforme and T. testudinum, there was a significant negative relationship between water column N : P and the amount of light reaching the bottom (%I 0 , a function of water quality and depth), but there was no significant relationship between these environmental parameters at sites that supported Halodule sp. and H. decipiens (Table 3, italic font).%I 0 was significantly and positively correlated with the C : N, C : P, and N : P, and δ 13 C of S. filiforme and Halodule sp., suggesting that high light availability leads to decreases in N and P content in seagrass leaves, an increase in the relative amount of N with respect to P, and enrichment with 13 C relative to 12 C for these species.The δ 15 N of S. filiforme and H. decipiens were both significantly positively correlated with %I 0 .Interestingly, there were only weak relationships between the water column N : P and sea-grass elemental and isotopic content.Halodule sp.C : N was higher when water column N : P was higher, but H. decipiens C : P and N : P were higher when the water column N : P was higher.δ 15 N of S. filiforme and T. testudinum were correlated with water column N : P, but there were no similar significant correlations for Halodule sp and H. decipiens.Interspecific differences in the distributions of species and elemental content and isotopic composition necessitated that analyses of the relationships between water quality and seagrass elemental and isotopic content be done individually for each species.The multiple bivariate correlations among the water quality and seagrass variables (Tables 2, 4) resulted in robust extraction of composite variables in PCA that explained between 73.1 and 82.7 % of the original variation in the species-specific data sets (Table 4).For all four species, PCA revealed strong, positive relationships between the water quality variables Chl a, DOC, TP and TN and the δ 15 N of seagrass leaves which loaded strongly onto the first principal component (PCI) which explained between 31.3 and 36.7 %  of the variation in the original data set.There were speciesspecific differences in the subsequently extracted PCs.For S. filiforme, PCII was strongly related to high %I 0 and seagrass N : P and δ 13 C; PCIII was strongly correlated with high water column TP and low water column TN : TP; and PCIV was related to high C : N and C : P in seagrass leaves.For T. testudinum, PCII was positively correlated with high C : P, N : P and δ 13 C of seagrass leaves; PCIII was negatively correlated with %I 0 but positively correlated with water column TN and TN : TP; and PCIV was associated with high seagrass C : P and C : N.Only three significant PCs were extracted for Halodule sp.; with PCII strongly correlated with high %I 0 and seagrass C : P, C : N and N : P; and PCIII positively correlated with water column TN and TN : TP.For H. decipiens, PCII was positively correlated with seagrass C : P and N : P; PCIII was associated with high seagrass C : N and δ 13 C; and PCIV was influenced strongly by high water column TN : TP.

Discussion
The striking spatial pattern in water quality on the Bermuda Platform clearly demonstrates the impact of the island and platform on water quality, and strong relationships between the long-term water quality data and the one-time sampling of seagrass tissues indicates that seagrasses really do serve as a long-term, integrating "canary" for environmental conditions.Stoichiometry of both the water column and the seagrasses that grow across the platform are consistent with general P-limitation of primary producers close to the island.However, the lack of clear relationships between the N : P in the water column and the N : P in the seagrasses suggests that benthic processes act to modify the N : P available to the benthos compared to the water column.Strong spatial patterns in the δ 15 N of seagrasses are consistent with a signature of sewage-derived N influencing waters very close to shore, but large areas of severely depleted 15 N signatures in the seagrasses of the Bermuda Platform suggest a very active role of the nearshore benthos in processing land-based sources of N.

"Stoichioscapes" as indicators of ecosystem processes
C : N : P stoichiometry of the water column strongly suggests general P limitation of the nearshore waters of Bermuda, with the vast majority of the measurements of water column N : P well in excess of the 16 : 1 of the Redfield ratio (Redfield, 1958).Furthermore, the spatially coherent patterns in water column TP, TOC and Chl a clearly demonstrate that the islands of Bermuda are a source of the limiting nutrient to the nearshore marine waters.Our findings based on C : N : P stoichiometry of the P-limited nature of the phytoplankton community on the Bermuda Platform agree with recent studies finding generalized P limitation of plankton production in the ocean surrounding Bermuda -the source waters for the Bermuda Platform (Ammerman et al., 2003;Lomas et al., 2004;Salihiglu et al., 2008).Stoichiometry of seagrass tissues is also consistent with P limitation across much of the Bermuda Platform, with most N : P values greater than the "seagrass Redfield ratio" (SRR) of 30 : 1 (Atkinson and Smith, 1983;Duarte, 1990).
The strength of seagrass P limitation, as assessed by the degree of N : P deviation from SRR, is not as severe as doc-umented in other parts of these species range, like south Florida (Fourqurean and Zieman, 2002;Fourqurean et al., 1992a), and seagrass N : P values are not as far from the stoichiometric balance between N and P availability as in the water column.In fact, there was no significant relationship between the water column and seagrass N : P across the Bermuda Platform.These conditions likely arise because of relatively low light availability on the bottom compared to the more tropical locations at the center of the ranges of these seagrass species (Manuel et al., 2013).When growing in light-limited conditions, a consequence of strong seasonality of insolation in Bermuda, nutrients are not limiting to seagrass production (Fourqurean and Zieman, 2002); hence, the seagrass elemental content tends towards the SRR.Light limitation could ameliorate the degree of P limitation in Bermuda's seagrasses as suggested by other studies for the relationship between light availability and nitrogen limitation (Abal et al., 1994;Grice et al., 1996;Zimmerman et al., 1987).This is consistent with our results for Halophila decipiens, which grew in the deepest and lowest light environments on the Bermuda Platform (Manuel et al., 2013) and had a mean N : P that was closest to the SRR of the four Bermuda seagrasses.
In addition to the differences in the mean N : P of H. decipiens compared to the other species that arise as a consequence of differences in the mean depths at which these species are found, biological differences among species led to consistent differences in N : P within sites.N : P is an integrated measure of the balance of light, N, and P availability for seagrass growth (Fourqurean and Rutten, 2003), and availability is determined by the balance between resource supply and the demand for that resource by the plants.Making the assumption that the supply of resources is constant within sites, then the species with the highest demand, i.e., the highest relative growth rate, should experience the lowest relative availability and exhibit greatest resource limitation (Tilman, 1980), as has been shown by comparing N : P of co-occurring seagrasses in other locations (Campbell and Fourqurean, 2009;Fourqurean et al., 1995).The relative N : P of Bermudian seagrasses likewise follow the pattern expected by comparing their relative growth rates.The relative growth rates (%d −1 ) are higher for H. decipiens, intermediate for Halodule sp. and S. filiforme, and lowest for T. testudinum (Duarte, 1991;Gallegos et al., 1994;Kenworthy et al., 1989), resulting in, when co-occurring, a relative N : P of seagrasses of H. decipiens > (S. filiforme and Halodule sp.) > T. testudinum.Since S. filiforme and Halodule sp. were similar in N : P and widely distributed across the platform, we were able to use their common stoichiometry to indicate the nutrient status of the different environments.
Despite our observations of general P limitation of seagrasses across the platform, on occasion our sampled seagrasses had N : P < SRR, indicative of N limitation (Fig. 4).Because the island is karstic with no natural streams, nutrients emanating from the island that could influence wa-ter quality and phytoplankton abundance in nearshore waters travel through groundwater to the sea.Recent studies of the importance of N and P in groundwater in controlling the primary production of benthic macroalgae in cave pools point to spatial and species-specific differences, with some macroalgae species being P-limited at some sites and other species being N-limited or light-limited at others (Maloney et al., 2011).Localized differences in N : P in groundwater may be responsible for this variation in the identity of the limiting nutrient in cave pools and thus the groundwater.This could help explain why N limitation of benthic marine plants very close to shore has been observed (McGlathery et al., 1992).

Isoscapes as indicators of ecosystem processes
The striking spatial pattern in δ 15 N serves as an informative isoscape (Bowen, 2010) for generating hypotheses and testing models about biogeochemical processes on the Bermuda Platform.Human wastewater can contain DIN with δ 15 N values as high as 40 ‰ (Tucker et al., 1999), and plants in close proximity to wastewater outflows often exhibit high δ 15 N (Cabana and Rasmussen, 1996).The very nearshore, 15 Nenriched δ 15 N signatures of seagrasses, and the rapid decline of seagrass δ 15 N offshore at the rate of 1.8 ‰ km −1 , suggest that sewage-influenced groundwater transports nutrients into the nearshore marine environment and that those nutrients are rapidly processed by both the water column and the benthos.Hence, the influence of sewage-derived nutrients on the benthos is currently limited in spatial extent in Bermuda; it is clear that the zone of δ 15 N of seagrasses > 0 corresponds with the zones of enhanced TOC, TN, TP and Chl a in the water column.
To our knowledge, the δ 15 N values we report from all four seagrass species occurring in Bermuda greatly expand the known range of values previously reported for each species.For example, the published range of δ 15 N values for T. testudinum encompasses −4.3 to 9.4 ‰ (Fourqurean et al., 2005); our data expands this range to a new minimum of −10.1 ‰. δ 15 N values < 0 have rarely been reported for seagrasses, and the mechanisms that would lead to plant tissues more depleted than atmospheric N need to be elucidated.Large spatially contiguous zones of δ 15 N values between −4.3 and 0 ‰ in the Florida Keys are hypothesized to be a result of either an isotopically depleted source pool of DIN, or maximum fractionation of DIN on uptake in regions of very high N availability (Fourqurean et al., 2005).Additionally, the process of ammonification produces NH + 4 that has δ 15 N values lower than the values of the source organic matter (Mobius, 2013); furthermore, when DIN is available in excess compared to plant demand, plants discriminate against 15 N uptake (Fourqurean et al., 2005;Yoneyama et al., 1991;Evans, 2001;Koba et al., 2010).Indeed, experimental N additions have been shown to increase fractionation of nitrogen isotopes upon uptake when nitrogen supply exceeds demand (Udy and Dennison, 1997) , 1997).
The strong correlation we found between δ 15 N and C : N for all four species of seagrass is consistent with enhanced fractionation of the source pool in regions of high N availability, but such a fractionation may not be sufficient to explain the large negative δ 15 N values.It is likely that benthic processing of N in the middle of the Bermuda Platform is creating uniquely isotopically depleted DIN that is being incorporated into seagrasses.The depleted 15 N on the Bermuda Platform is consistent with the input of new N to the region by N 2 fixation (Montoya et al., 2002), but additional processes must be involved to generate the observed negative δ 15 N. High rates of ammonification could partly explain why DIN would be depleted in 15 N (Mobius, 2013).

δ 13 C of seagrasses in Bermuda
We expected to find that δ 13 C of seagrass tissues would be a function of light availability, because in high light environments there is less fractionation of dissolved inorganic carbon (DIC) upon photosynthetic uptake.In essence, seagrasses are more isotopically enriched in 13 C if light availability does not limit photosynthesis (Campbell and Fourqurean, 2009;Fourqurean et al., 2005;Hu et al., 2012) and therefore C demand is high relative to availability.We did find this expected relationship across the platform for S. filiforme and Halodule sp., but not for T. testudinum or H. decipiens.The lack of these relationships for T. testudinum and H. decipiens could be owing to the more restricted ranges of light availability at which these species were found: T. testudinum is generally restricted to high light environments, H. decipiens to low light environments, and the other two species are found from high light to low light environments.Additionally, lower light availability at Bermuda's latitude could decrease the dependence of δ 13 C on light compared to more tropical locations with higher light.Interestingly, for all four seagrasses, plants with higher δ 13 C had a lower P content, similar to previous reports from south Florida (Fourqurean et al., 2005), suggesting that plants from high light environments experience greater P limitation than plants from low light environments.
Differences in morphology and carbon uptake produce species-specific differences in the δ 13 C of co-occurring seagrasses.It has been shown that, when occurring together in south Florida, the relative δ 13 C of Halodule sp.< T. testudinum < S. filiforme (Campbell and Fourqurean, 2009).Our data corroborate these relative differences and add information about Halophila decipiens, which we frequently found growing in the same sites with other seagrasses in Bermuda, whereas in south Florida it rarely is found growing together with other seagrasses (Fourqurean et al., 2002).H. decipiens in Bermuda has δ 13 C values intermediate between Halodule sp. and T. testudinum.

Stoichioscapes and isoscapes as important tools to study marine ecological processes
The spatial patterns of elements and their ratios, which we can call a stoichioscape, and the stable isotopic composition of widespread seagrasses (an isoscape) specifically point to (1) generalized P limitation of both the water column and the benthos, (2) land-based sources of both N and P, and (3) locations within the more generally P-limited seascape that are light-or N-limited.The substantial impact of light availability on seagrass distribution, elemental content and stable isotopic composition did not obscure the spatial patterns.Furthermore, our results have generated new hypotheses about the nature of N cycling on the Bermuda Platform that will guide research of local processes but also of the role of the coastal ocean in global biogeochemical cycles.
Bermuda presents a unique opportunity to examine the use of elemental and isotope content of organisms to help explain ecological processes, in particular the large decline of seagrasses on the outer platform.Although the isoscape of nutrients implicates land-based sources of nitrogen and phosphorus, we cannot attribute the decline of seagrasses to degraded water quality and have hypothesized that seagrass losses are the result of overgrazing by green turtles (Fourqurean et al., 2010).Our isoscapes and stoichioscapes can be used as tools in ecogeochemistry where the elemental and isotope content are used as natural markers to help understand trophic ecology and population structure.Ecogeochemistry is providing new insights on animal movements in marine ecosystems (see McMahon et al., 2013 for review) and has long been used to study food web structure.Pronounced patterns in isotopic composition of plants across an area like the Bermuda Platform can help in interpreting the distribution of stable isotope ratios among individuals in a population of consumers.For example, large variation in δ 13 C and δ 15 N of green sea turtles within a population has been interpreted as individual diet specialization, with some individuals feeding on seagrasses, others on macroalgae and still others with a carnivorous diet (e.g., Burkholder et al., 2011).Given the very large ranges in isotopic composition of seagrasses we found on the Bermuda Platform, a similar variability in the isotopic composition of green sea turtles could be driven by spatial variation in the isotopic composition (and species identity) of seagrasses and restricted home ranges of individual turtles in the population, and indeed variation in stable isotopic composition of sea turtles has been used to infer differences in provenance of sea turtles in other locations (Pajuelo et al., 2012).Only with knowledge of the relationship between the isoscapes and stoichioscapes and the home ranges of individuals can variation of ecogeochemical tags within a population be properly interpreted.

Figure 1 .
Figure 1.Isopleths of period-of-record averages for total nitrogen (TN; µM), total phosphorus (TP; µM) and the molar ratio of N : P in the water column on the Bermuda Platform.The blue lines are the 20 m isobaths on the rim of the platform.Red dots indicate locations of the 17 permanent sampling locations.

Figure 2 .
Figure 2. Isopleths of period-of-record averages for total organic carbon (TOC; µM) and chlorophyll a (Chl a; µg L −1 ) in the water column on the Bermuda Platform.The blue lines are the 20 m isobaths on the rim of the platform.Red dots indicate locations of the 17 permanent sampling locations.

Figure 3 .
Figure 3. Frequency distribution of the measured carbon, nitrogen and phosphorus content of seagrasses collected on the Bermuda Platform, by species.The cross-hatched region on the molar N : P plots indicates the seagrass Redfield ratio of 25-30; N : P less than this indicates N limitation, while N : P greater than this indicates P limitation.

Figure 4 .
Figure 4. Isopleths of the N : P of leaves of the seagrasses Syringodium filiforme and Halodule sp.Black dots indicate sampling locations.

Figure 5 .
Figure 5. Frequency distribution of the measured δ 13 C and δ 15 N of seagrasses collected on the Bermuda Platform, by species.

Figure 6 .
Figure 6.δ 15 N of seagrass leaves from the Bermuda Platform.Black dots indicate sites where at least one of the seagrasses Syringodium filiforme, Thalassia testudinum or Halodule sp.occur.If more than one species occurred at a site, the average value for all co-occurring species is shown.

Table 1 .
Pearson's correlation of water quality data, correlation coefficient (ρ) above the diagonal, and two-tailed test of significance of the correlation (p) below the diagonal.The number of observations for each quantity are in Supplement TableS1; in general n ≈ 1000.* Indicates p ≤ 0.05; * * indicates p ≤ 0.01.

Table 2 .
Nitrogen and phosphorus content, N : P, δ 13 C and δ 15 N values of seagrass leaves from the Bermuda Platform.Superscripts indicate homogenous groups as indicated by Student-Newman-Keuls post hoc tests in ANOVA of data grouped by species; significant differences at α = 0.05.

Table 3 .
Relationships among elemental content and stable isotope ratios for seagrasses from the Bermuda Platform.Values above the diagonal are Pearson's correlation coefficients (r), below the diagonal is the two-tailed test of the significance of the correlation (p).Italics indicate water column parameters, boldface type indicates statistically significant relationships with p ≤ 0.05.

Table 4 .
Principal component analysis of the relationship among water quality and seagrass environmental and isotopic composition, by seagrass species.Values are the correlations between the original variables and the extracted principal components.Values in bold type have the largest loadings within a principal component.

www.biogeosciences.net/12/6235/2015/ Biogeosciences, 12, 6235-6249, 2015 grass
. In fact, the only report of sea-δ 15 N values as low as we observed in the center of the Bermuda Platform are from experimental N addition experiments with very fast growing, high N-demanding seagrasses(Udy and Dennison