Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit

Marine Pollution Bulletin xxx (2013) xxx–xxx

Contents lists available at ScienceDirect

Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul

Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit Matteo Vacchi a,d,⇑, Monica Montefalcone b, Chiara F. Schiaffino b,d, Valeriano Parravicini c,d, Carlo Nike Bianchi b, Carla Morri b, Marco Ferrari b a

Aix-Marseille Université, CEREGE CNRS-IRD UMR 34, Aix en Provence, France Department of Earth, Environmental and Life Sciences, University of Genoa, Italy UR 227 – ‘‘CoReUs2’’ IRD – Institut de Recherche pour le Développement, Laboratoire Arago, Banyuls-sur-Mer, France d SEAMap Ltd., Environmental Consulting, Borghetto Santo Spirito, Italy b c

a r t i c l e

i n f o

Keywords: Seagrass Nearshore hydrodynamics Ecological modelling Reference conditions Posidonia oceanica Mediterranean Sea

a b s t r a c t The upper portion of the meadows of the protected Mediterranean seagrass Posidonia oceanica occurs in the region of the seafloor mostly affected by surf-related effects. Evaluation of its status is part of monitoring programs, but proper conclusions are difficult to draw due to the lack of definite reference conditions. Comparing the position of the meadow upper limit with the beach morphodynamics (i.e. the distinctive type of beach produced by topography and wave climate) provided evidence that the natural landwards extension of meadows can be predicted. An innovative model was therefore developed in order to locate the region of the seafloor where the meadow upper limit should lie in natural conditions (i.e. those governed only by hydrodynamics, in absence of significant anthropogenic impact). This predictive model was validated in additional sites, which showed perfect agreement between predictions and observations. This makes the model a valuable tool for coastal management. Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction The problem of sliding baselines represents a major concern for the evaluation of global change impacts on ecosystems (Dayton et al., 1998; Hobday, 2011). Lack of information on pristine (or at least historical) conditions to be used as references impairs our understanding of the effects of global change on natural ecosystems and their consequences on the amount of resources a healthy ecosystem can provide. Although marine systems may seem less accessible to human uses and impacts than terrestrial ecosystems, global assessments revealed that pristine areas are lost also in the marine realm (Jackson and Sala, 2001; Stachowitsch, 2003). Seagrass meadows are among the most important elements for the functioning of marine coastal ecosystems, and represent a major focus for research and conservation (Koch et al., 2006). This paper is focused on the endemic seagrass Posidonia oceanica, the most important and abundant of the Mediterranean Sea. Its meadows form a key coastal habitat and strongly influence coastal features in terms of wave reduction and nearshore sedimentary patterns (Boudouresque et al., 2012). Anthropogenic pressure causes the regression of P. oceanica meadow upper and lower limits (i.e. the

⇑ Corresponding author at: Aix-Marseille Université, CEREGE CNRS-IRD UMR 34, Aix en Provence, France. Tel.: +33 4 42971665. E-mail address: [email protected] (M. Vacchi).

shallowest and the greatest depths reached by the plant, respectively), which implies a reduction in its extension and therefore a functional loss for the coastal ecosystems (Montefalcone et al., 2010a, 2010b; Boudouresque et al., 2012). Assessing the natural position, i.e. the baseline, of seagrass meadow limits is crucial to define their reference status in order to distinguish the impact of natural processes as opposed to anthropogenic factors (Pergent et al., 1995; Montefalcone et al., 2010b). Recent efforts have been invested worldwide for the preservation of coastal ecosystems, also from a legislative point of view (Ricketts and Harrison, 2007; Barnes and McFadden, 2008; Borja et al., 2008). In Europe, both the Water Framework Directive (WFD, 2000/60/EEC) and the Marine Strategy Framework Directive (MSFD, 2008/56/EEC) recognize as mandatory the maintenance of seafloor integrity, which has to be assessed by comparison with reference conditions (Duarte et al., 2008). In general, reference conditions can be defined in three ways: (i) historical information, when available and reliable, which is not always the case (Leriche et al., 2004; Montefalcone et al., 2013; Lyons et al., in press); (ii) data collected in pristine areas, still scarce worldwide and often insufficiently enforced (Stachowitsch, 2003; Montefalcone et al., 2009); (iii) modelling (Valle et al., 2011; Parravicini et al., 2012; Downie et al., in press). The P. oceanica meadow upper limit usually occurs within the most dynamic region of the seafloor (Boudouresque et al., 2012). Recent literature highlighted that coastal dynamics strongly

0025-326X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

2

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

influence the shallowest meadow portions (Folkard, 2005; Infantes et al., 2009; Vacchi et al., 2010). This paper explores the relationship between the position of the meadow upper limit and the morphodynamic domain of the beach (i.e. the distinctive type of beach created by the interplay of topography, wave climate and sediment composition, Benedet et al., 2004) on a wide spatial scale in the NW Mediterranean Sea. We hypothesized that, in absence of major human pressures, the structure (i.e. shoot density and seagrass cover) of the upper limit of P. oceanica meadows and its position along the region of the underwater beach profile mostly reworked by wave action, are mainly controlled by local nearshore hydrodynamics. An innovative predictive model was developed: it can accurately identify the seafloor portion where the meadow upper limit should lie in natural conditions (i.e. conditions defined by hydrodynamics alone) and could provide the proper tool to define reference conditions for healthy meadows. 2. Methods 2.1. Study area The study was carried out in Liguria, an administrative Region of NW Italy, where virtually all existing P. oceanica meadows have been included within Sites of Community Importance (SCIs) according to the European Community Directive of 1992 (EEC, 1992). We investigated 10 coastal areas (Fig. 1) where P. oceanica meadows occur on sedimentary seafloor, according to the detailed cartography provided by Liguria Region (Diviacco and Coppo, 2007). Selected coastal areas show different geomorphologic setting and wave exposure and were historically subject to low anthropic pressure (Diviacco and Coppo, 2007; Montefalcone et al., 2010a, 2010b). The 350 km long Ligurian coastline (Fig. 1) is mainly characterized by rocky headlands alternating with sandy to gravelly beaches, especially along the Eastern Riviera: soft coasts are relatively less developed and are typically found adjacent to small coastal plains along the Western Riviera (Rovere et al., 2010, 2011). Ligurian coastline wave regime is mainly influenced by winds blowing from southwest, with small differences between the two Rivieras (Fig. 1, Corsini et al., 2006; www.idromare.it). The SW (220–240° N) is the dominant wave direction, with a fetch longer

than 800 km and an offshore wave height of more than 4 m. The SE (130–150° N) and the S (180° N) wave directions, both characterized by fetches of about 200 km and waves of about 2 m height, have comparatively lower impact. 2.2. Hydrodynamic study Two hydrodynamic limits along the underwater beach profile represent important boundaries to be considered in coastal engineering design (Sorensen, 2006): (i) the breaking depth, i.e. the depth where waves break (Smith, 2003); and (ii) the closure depth, i.e. the depth where wave action on the seafloor becomes negligible (Dean and Dalrymple, 2004). Intense alongshore and onshore– offshore transports take place only in depths shallower than the closure, whereas horizontal water particle velocity reaches its maximum values at breaking depth and increases its movement towards the shoreline. In each area a detailed bathymetric survey was carried out with a single-beam echo-sounder (single frequency, error ± 0.1 m, 1 point every 5 s) and differential GPS to define the morphology of seafloor where meadows grow. A detailed 2D bathymetric map (1:5000) was produced for each of the 10 areas; together with the local wave parameters (Table 1), it allowed identifying the two hydrodynamic boundaries along the underwater beach profile. The breaking depth (db) was calculated using the formula (Smith, 2003)

db ¼ Hb =cb

ð1Þ p

where Hb ¼ H0 K sh ðuo =ub Þ; (Ksh = shoaling coefficient, uo and ub = offshore and nearshore waves approach angle, respectively); and cb ¼ b  ðaHb Þ=ðgT 20 Þ (a and b being empirical coefficients depending on the beach slope, Smith 2003). Annual offshore wave parameters (return time 1 year) were preferred to daily average waves as the latter could underestimate the effect of annual extreme events on the meadow (Infantes et al., 2009; Vacchi et al., 2012a). The closure depth (dc) was computed using the following formula (Sorensen, 2006)

dc ¼ 6:75Hs

ð2Þ

where Hs is the mean annual significant wave height. 2.3. Underwater surveys We assessed the structure of the upper portion of the P. oceanica meadows and the sedimentary features of the seafloor by scuba diving surveys carried out on three different sites (A, B and C), which were selected at least 100 m apart from each other in each coastal area (Fig. 2). In each site, five stations were located along the underwater beach profile in order to encompass both breaking and closure depth (Fig. 2); in particular, the location of station 1 Table 1 H0, T0, and L0 values (return time 1 year) in the 10 study areas (Corsini et al., 2006, www.idromare.it). See Fig. 1 for the codes of each areas. H0 is the offshore wave height, T0 is the offshore wave period and L0 is the offshore wave length.

Fig. 1. Study areas along the Ligurian coastline. Black dots individuate the 10 Posidonia oceanica meadows investigated: Alassio (ALS), Arenzano (ARE), Arma di Taggia (ARM), Camogli (CAM), Framura (FRA), Imperia (IMP), Monterosso al Mare (MTM), Spotorno (SPO), Capo Nero (CPN) and Ceriale (CER). White dots represent the 4 meadows where the predictive model was tested: Albenga (TALB), Bussana (TBSN), Pieve Ligure (TPVL) and Ospedaletti (TOSP). The annual wave climate is also indicated (data from Corsini et al. (2006), modified). HS0 is the significant offshore wave height (m) recorded by the La Spezia buoy (43.92917° N, 9.82778° E).

Area

H0 (m)

T0 (m)

L0 (m)

ALS ARE ARM CAM CER CPN FRA IMP MTM SPO

2.6 4.0 4.0 4.2 2.6 4.0 4.2 4.0 4.2 2.6

5.8 7.5 7.5 7.8 5.8 7.5 7.8 7.5 7.8 5.8

52.4 87.7 87.7 94.9 52.4 87.7 94.9 87.7 94.9 52.4

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

3

Fig. 2. Explicative scheme of the sampling design used in this study. The right panel explains the abbreviation of the different variables recordend in each of the 5 stations. This sampling design was carried out in Alassio (ALS), Arenzano (ARE), Arma di Taggia (ARM), Camogli (CAM), Framura (FRA), Imperia (IMP), Monterosso al Mare (MTM), Spotorno (SPO), Capo Nero (CPN) and Ceriale (CER).

was shallower than the computed breaking depth (db), the location of station 5 was located deeper than the computed closure depth (which never exceeded 10.5 m in this study), and stations 2, 3 and 4 were regularly spaced between breaking and closure depth (dc). A total of 150 stations were investigated (see Appendix 2). At each station, one sediment sample was collected by hand coring (Fig. 2). Grain size was analyzed by dry sifting at 1/2U intervals (Wentworth, 1922); median (grain size value at 50% in the cumulative distribution expressed in U scale), sorting (standard deviation of the grain size expressed in U scale), and mud percentage (grain size > 4 expressed in U scale) percentage were determined according to Folk and Ward (1957). Moreover, four descriptors of P. oceanica meadow structure were measured for each station (Fig. 2): (1) shoot density; (2) living P. oceanica cover; (3) ‘dead matte’ extent, which is represented by the interlaced remnants of roots, rhizomes and entangled sediment (Giovannetti et al., 2008); (4) bare bottom, which corresponded to the portion of seafloor without evidence of living or dead meadow. The former descriptor was measured using a 40 cm  40 cm PVC frame in five replicate counts (Buia et al., 2004), whilst each of the latter three descriptors was estimated visually on a surface of about 25 m2 independently by two divers swimming 3 m above the bottom, their value being expressed as a percentage (Montefalcone, 2009). 2.4. Statistical analysis In order to test whether the structure of the P. oceanica meadows was influenced by its hydrodynamic position along the beach profile (i.e. the relative position of each station with respect to the two hydrodynamic boundaries), by sediment features, or by a combination of both factors, the multivariate extension of general linear models proposed by Warton (2011) was employed. Compared to other common multivariate analyses, such as Anosim (Clarke, 1993) or Permanova (Anderson, 2001), multivariate general linear models have the advantage of clearly showing the effect of the explanatory variables on each of the response variables. Moreover, such models also verify the significance of an overall multivariate effect (Warton et al., 2012). The multivariate statistics is, in fact, built after fitting univariate linear models for each response variable. As shoot density showed no significant variation along the underwater beach profile (see Section 3), the response

variables chosen to describe the P. oceanica meadow structure were the cover of living P. oceanica, the dead matte extent, and the amount of bare bottom, all expressed as a percentage. Before the analysis, percent data underwent the transformation arcp sin (x/100) in order to reduce skewness (Legendre and Legendre, 1998). The difference between station depth and breaking and closure depth (hereafter Dbreak and Dclos, respectively) was employed to measure the hydrodynamic position of each station (see Fig. 2). Median, sorting, and mud % were used as descriptors of sediment features (Folk and Ward, 1957). Multi-collinearity was tested by examining the Pearson correlation coefficient between each pair of explanatory variables (see Appendix 1). Collinearity was detected between Dbreak and Dclos (descriptors of the hydrodynamic position) and between median and sorting (descriptors of sediment features). Given the high correlation found, it was decided to keep only the variables that explained more variation as single predictors and to drop the others (see Appendix 1). The explanatory variables retained have been therefore Dbreak, sorting, and mud percentage. Backward model selection was then conducted starting from the full model, thereby dropping non significant variables with the stepwise approach described in Zuur et al. (2009). Correlation among the response variables was accounted by an unstructured correlation matrix, and multivariate p-values were obtained by a Monte Carlo resampling approach with 10,000 iterations. This method ensures approximately valid inference even when the correlation structure or the mean–variance relationship has been miss-specified (Warton et al., 2012). In order to test whether the shoot density of P. oceanica was influenced by hydrodynamic position, sediment features, or by a combination of both, we employed the same analytical framework used for the multivariate test. An univariate generalized linear model was adopted assuming a quasi-Poisson distribution. Also in this case, backward selection of explanatory variables was employed and p-values were obtained by a Monte Carlo resampling approach with 10,000 iterations. In order to avoid any confounding effect due to the presence/absence of P. oceanica in the sampling stations, the analysis was performed only on those samples where P. oceanica was present. Statistical analysis results indicated breaking depth as the hydrodynamic boundary of the meadow upper limit development

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

4

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

towards the shoreline (see Section 3). For this reason, it was tested how accurate a prediction could be made concerning the meadow upper limit natural position with respect to the breaking depth, based on the morphodynamics of a given coastal area. In each site (A, B, C in each of the 10 coastal areas), underwater surveys were used to take two spatial measures between the breaking depth and the meadow upper limit (Fig. 3): (i) the linear distance along the bottom profile jobs was measured using a centimetre-marked line laid on the bottom; (ii) the difference in depth zobs was measured by averaging two accurate electronic depth gauges with a final precision 6 0.3 m. In each site, the surf scaling index e (Jackson et al., 2005) was calculated combining wave parameters with information derived from the specifically created bathymetric maps (Table 1). This index is controlled by wave steepness and beach slope (Dean and Dalrymple, 2004). The following formula was employed:

e ¼ ax2 =g  tan2 b

ð3Þ

where a (breaker amplitude) = H0/2; x (incident wave radian energy) = 2p/T0, T0 = period; g = acceleration of gravity; b = the slope of the beach in the surf–zone. Surf scaling index defines three distinct domains (Jackson et al., 2005): reflective (e < 2.5), intermediate (2.5 < e < 20), dissipative (e < 20). This index already proved effective in defining the morphodynamic domains of the Ligurian coastline (Montefalcone et al., 2010b; Vacchi et al., 2010). The relationship between e and the two measured distances (jobs and zobs) was assessed by means of a general linear model. The predictive performance of each linear model was assessed by 10-folds cross validation (CV). This technique divides the original datasets into 10 subsamples while retaining only 9 of them to fit the model, whose errors are then evaluated on the subsample not used in the model. The procedure is repeated 10 times and allows estimating a more parsimonious R2 (i.e. the cross-validated R2), representing the expected performance of the model when fitted to new data (Efron, 1983).

2.5. Test of the predictive model We used the equation derived from the linear model showing the best predictive performance (i.e. the best CV–R2) to predict the theoretical natural position of the meadow upper limit with respect to the breaking depth. Uncertainty of model intercept and slope was estimated as 95 % confidence interval of the parameters obtained by fitting the model on 500 bootstrap replicates of the original dataset. Adequacy of this model was tested in 4 further study sites along the Liguria coastline (Fig. 1): Albenga (TALB) and Pieve Ligure (TPVL), where anthropic pressure is relatively low (Diviacco and Coppo, 2007); Bussana (TBSN) and Ospedaletti (TOSP), which were instead subject to significant anthropic pressure in the last 15 years (Diviacco and Coppo, 2007; Fierro et al., 2011; Vacchi et al., 2012b).

Fig. 3. Linear (j) and vertical (z) distances of the meadow upper limit with respect to the breaking depth position db (white square) measured in each site.

3. Results Living P. oceanica meadow was never found in station 1 (i.e. shallower than the breaking depth) of any site; only scattered dead matte (never exceeding 5% of seafloor) was found in these stations. Both shoot density and living seagrass cover always exhibited a seawards increase from station 2 to station 5 (Fig. 4). Mean shoot density was similar in all stations 4 and 5, whereas mean cover reached the highest values (>80%) in station 5 of all sites. Dead matte did not follow the same pattern: the highest mean value of its extent was found in station 2 and then it decreased with the increasing depth (Fig. 4). Sediments in the 150 stations ranged from fine sand to large pebbles with no clear pattern of the mean median value along the underwater beach profile (Fig. 4). Mean sorting values were similar in all stations. Mud percentage was generally very low, rarely exceeding 10%. Multivariate analyses indicated that the structure of P. oceanica meadow was mainly controlled by the hydrodynamic position along the underwater beach profile. No sediment-related value was found significant, therefore they were all excluded from the final model, which accounts for the relative depth difference between each station and breaking depth (Dbreak) (F = 1.03, p < 0.0001, multivariate Hooper’s R2 = 0.28) only. Moreover, the univariate models from which the multivariate test was computed (Fig. 5) revealed that Dbreak is related to an increase in living P. oceanica cover (F = 0.69, p < 0.0001) and to a parallel decrease of bare bottom (F = 0.68, p < 0.0001). On the contrary, no relationship was found for dead matte (F = 0.03, p < 0.54). The shoot density of P. oceanica was independent from both sediment characteristics and hydrodynamic position along the beach profile. Surf scaling index (e) ranged between 3.5 and 438 (Table 2), respectively indicating low intermediate to highly dissipative domains. Meadow upper limit was very close to the breaking depth in those sites characterized by lower e values, whereas higher e values corresponded to larger distances (Table 2). The parameters j and z were significantly correlated with the surf scaling index e (Fig. 6). However, j showed a much higher correlation (F = 286.45, p < 0.001, cross-validated R2 = 0.88) than z (F = 21.54, p < 0.001, cross-validated R2 = 0.32). The addition of higher order polynomial terms to the dependence of jobs and jpred on e was also examined, but these did not improve the predictive performance of the model (Fig. 6). According to the 95% confidence interval of model parameters obtained by bootstrapping, the seafloor region where the predicted upper limit of P. oceanica meadow should be located was identified according to the following two equations

jmin ¼ 5:94 þ 0:29e

ð4Þ

jmax ¼ 17:83 þ 0:41e

ð5Þ

where jmin and jmax represent the position of the meadow upper limit as predicted using respectively the 2.5% and the 97.5% of the model parameter estimates obtained by the bootstrapping procedure, and e is the surf scaling index. When combined, the two equations provide the range of distances from the breaking depth within which the upper limit is expected to be located. Testing this predictive model in the 4 selected study sites allowed to calculate a jmax–jobsvalue (Table 3), which represents the difference between the predicted seawards boundary of the seafloor region where the upper limit was expected (i.e. the jmax) and the observed upper limit position (i.e. jobs). In the two sites with little human pressure, Pieve Ligure and Albenga, jobs–jmax values were positive and the meadow upper limits were found between the calculated jmin and jmax (Fig. 7a and b). On the contrary,

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

5

Fig. 4. Mean values (±s.d.) of the structural descriptors of the Posidonia oceanica meadows and of the sediment features (expressed in U scale) sampled in the 150 stations along the underwater beach profile.

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

6

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

Fig. 5. Relationship between the Dbreaking and the bottom percentages of living seagrass, dead matte and bare bottom. Percent values are arcsin transformed. Black lines represent the regression fit.

Table 2 Linear (jobs) and vertical (zobs) distances between the breaking depth (db) and the meadow upper limit measured in the 30 sites. Site

db (m)

Surf scaling (e)

Distance (m)

jobs

zobs

ALS 1 ALS 2 ALS 3

4.6 4.6 4.7

249.5 308.0 331.1

115.2 125.6 145.3

6.3 5.4 5.5

ARE 1 ARE 2 ARE 3

3.8 2.9 3.0

67.4 25.1 28.5

26.1 29.9 18.9

0.6 0.8 0.5

ARM 1 ARM 2 ARM 3

3.0 4.3 4.8

27.5 28.1 78.5

6.8 7.8 29.4

0.5 1.2 0.4

CAM 1 CAM 2 CAM 3

4.8 4.0 3.1

43.0 28.2 18.2

8.3 8.2 18.1

0.8 0.2 2.2

CER 1 CER 2 CER 3

4.4 4.1 4.0

438.1 403.3 309.9

131.2 125.5 122.1

3.5 3.5 2.9

FRA 1 FRA 2 FRA 3

4.4 4.2 4.1

40.2 20.2 16.5

7.2 9.1 9.5

0.2 0.7 0.3

IMP 1 IMP 2 IMP 3

5.6 5.6 5.0

115.8 117.6 50.4

75.7 63.5 43.0

2.4 2.3 1.5

MTM 1 MTM 2 MTM 3

4.0 3.2 5.4

11.7 3.4 19.3

8.0 10.5 35.9

0.3 0.5 2.5

OSP 1 OSP 2 OSP 3

6.8 6.7 4.5

129.4 93.9 91.8

58.9 48.6 46.6

3.4 0.3 2.5

SPO 1 SPO 2 SPO 3

3.0 2.1 3.8

16.8 17.8 16.0

18.7 20.1 16.7

1.0 0.3 2.3

in the two sites affected by significant human pressures, Bussana and Ospedaletti, jobs–jpred values were negative and meadow upper limits were located 68.6 m and 43.3 m seaward from jmax (Fig.7c and d). 4. Discussion We explored the structure of the portions of the P. oceanica meadow occurring in the region of the underwater beach profile shallower than the closure depth (i.e. the depth where wave action on the seafloor becomes negligible). We demonstrated that the breaking depth (i.e. the still-water depth at the point where a wave

breaks) represents the major constraint for the landward development of the meadows occurring on sedimentary bottoms. As the linear distance (j) from the breaking depth showed a significant correlation with the beach morphodynamics (expressed by the surf scaling index e), we developed a predictive model to locate the upper limit position, providing evidence that the natural position of the meadow upper limit, in absence of major human pressures, can be predicted on the basis of physical parameters alone. Uncertainty of the predictive equation (i.e. jmin and jmax) delineates a seafloor region that can be identified as the baseline, i.e. the reference condition zone of development for a given P. oceanica meadow under natural conditions (Fig. 7). The adequacy of this model was supported by the results obtained in the 4 test sites. In sites subject to low human pressure (Albenga and Pieve Ligure), the position of the meadow upper limits was found within the reference condition zone, and did not show any sign of regression (Fig. 7a and b). On the contrary, in the sites affected by human pressures (Bussana and Ospedaletti), meadow upper limits lay deeper than the reference condition zone, being positioned several tens of meters seaward from jmax (Fig. 7c and d). Seaward distance from jmax should be interpreted as meters of regression suffered by the P. oceanica meadow upper limit, and thus the linear loss of meadow extent, caused by anthropogenic activities. The coastal area of Bussana underwent considerable modifications due to the construction of low crested defence structures and coastal discharge (Fierro et al., 2011). At Ospedaletti, repeated large beach nourishments strongly modified the original underwater beach profile (Vacchi et al., 2012b). Defining baselines for ecosystems and habitats is always difficult. The predictive model presented in this paper allows quantifying the amount of seagrass meadow regression, thus providing the theoretical reference condition requested by the European Directives (WFD and MSFD). Knowing where the upper limit of a specific P. oceanica meadow should naturally occur is fundamental to quantify any suspected or observed regression caused by anthropic factors. Predicting species distribution and habitat suitability is extremely useful in supporting implementation of environmental legislation, protection and conservation of marine waters and ecosystem-based management (Valle et al., 2011). Recent studies (Downie et al., 2013; Lyons et al., in press) highlighted the increasing importance of predictive models to define seagrass distribution. However, we must stress that the concept of a ‘‘baseline’’ for seagrass meadows is not simple. In fact, seagrass systems are dynamic, and subject to all sorts of natural and anthropogenic stressors, which also interact (Leriche et al., 2011). For example, the position of the seagrass will vary over time, perhaps in response to seasonal changes, and to events such as big storms

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

7

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

Fig. 6. Linear and polynomial models between the two distances (jobs and zobs) and the morphodynamic domain (expressed by the surf scaling index e). CV–R2 is the cross validated R2 value estimated by the 10-fold procedure. Grey dotted lines represent the fits of the individual ten folds used for cross-validation.

Table 3 Results of the test of the predictive model in the 4 study sites (see Fig. 1 for area codes). e is the surf scaling index, db is the breaking depth, jobs is the observed linear distance between the db and the meadow upper limit. jmin and jmax are the minimum and the maximum predicted distances between db and meadow upper limit. jobs–jmax is the linear difference between the observed upper limit position (i.e. the jobs) and the predicted seawards boundary of the seafloor region where the upper limit was expected (i.e. the jmax). Test area

Human pressure

e

db (m)

jobs (m)

jmin (m)

jmax (m)

jmax–jobs (m)

TALB TPVL TBSN TOSP

Low Low High High

312.8 14.2 38.0 84.1

4.3 6.1 3.1 4.8

131.6 12.2 102.0 95.6

96.2 7.3 16.5 29.9

146.1 20.3 33.4 52.3

+14.5 +8.1 68.6 43.3

Fig. 7. Results of the predictive model in the 4 test areas (see Fig. 1 for codes). White squares indicate the breaking depth position, whereas dashed lines define the two boundaries (jmin and jmax) of the predicted reference condition zone (pale grey). The white-striped portion of the seafloor represents the quantitative assessment of the upper limit regression in the areas affected by high human pressures.

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

8

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx

(Boudouresque et al., 2012). P. oceanica is a long-lived species and the position of its meadows is therefore likely to be controlled more by extreme events than by the mean wave climate (Infantes et al., 2009). For this reason, we adopted the annual wave data being aware that effects of storm waves on the meadow could be under-estimated using short period of observations. Our model has been developed for sedimentary seafloor and its adequacy was tested at a wide spatial scale (the whole Liguria Region). Further testing on other Mediterranean areas, could lead to a more general applicability. Once tuned at larger scale, this model could become a very useful tool for management and conservation of the most important seagrass habitat of the Mediterranean. Even if the predictive model represents the most innovative aspect of this paper, our result also provide a better definition of nearshore hydrodynamics influence on the shallow portion of P. oceanica meadows. No significant differences in sedimentary features were observed along the investigated beach profile area, and the multivariate analysis results identified the hydrodynamic position of the P. oceanica meadow (compared with the two hydrodynamic boundaries, breaking and closure depth) as the main factor affecting its structure. Thus, even if sediment features of the seafloor are known to control meadow development (Boudouresque et al., 2012), their role is less important in the shallowest seafloor portion where wave breaking is the dominant hydrodynamic process (Smith, 2003). Multivariate analysis results indicated that the seaward deepening of the meadow from the breaking depth mainly influences P. oceanica cover. In contrast with previous results obtained from a lower number of meadows (Vacchi et al., 2010), shoot density seems to be less influenced by the hydrodynamic position along the beach profile, and probably is more correlated to light availability along the depth gradient (Pergent et al., 1995; Della Via et al., 1998) or to other environmental factors (Leriche et al., 2011). Wave energy was already recognized as a significant limiting factor for P. oceanica growth at shallow depths (Molinier and Picard, 1952; Koch et al., 2006; Infantes et al., 2009). Generally, water motion plays a key role in the initial life stages of seagrass because seedlings are vulnerable to burial or dislodgement (Infantes et al., 2012). Results of our study indicate that wave-related effects significantly control the status of the meadow shallow portion during the whole life-cycle of the plant, which is always in dynamic equilibrium with the surrounding hydrodynamic system. This has important consequences for coastal zone management, as the status and the depth of the upper limit of P. oceanica meadows is part of current protocols for seagrass monitoring activities (Montefalcone, 2009; Boudouresque et al., 2012), which need reference conditions to track their evolution under natural and anthropogenic influences.

Acknowledgments This study was carried out within the project GIONHA, Governance and Integrated Observation of marine Natural HAbitats (EU programme ‘‘Interreg IV Marittimo’’). We are grateful to ‘‘Settore Ecosistema Costiero’’ of Liguria Region for its support during the project. MV contributes to the Labex OT-Med (n° ANR-11-LABX-0061) funded by the French Government «Investissements d’Avenir» program of the French National Research Agency (ANR) through the A*MIDEX project (n° ANR-11-IDEX-0001-02). The authors would like to thank Dr Francesca Baggio (Genova) for English revision. We thank Mariachiara Chiantore (University of Genoa) for suggestions on statistical analyses, and all participating in field-data collection: Stefano Bellati, Carlo Bertora, Alberto Demergasso, Giulia Gatti and Aurora Truccolo (University of

Genoa), Nicolas Mannarino Fernandez (University of Cadiz) and Alessio Rovere (Columbia University).

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.marpolbul. 2013.09.038.

References Anderson, M.J., 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology 26, 32–46. Barnes, C., McFadden, K.W., 2008. Marine ecosystem approaches to management: challenges and lessons in the United States. Marine Policy 32, 387–392. Benedet, L., Finkl, C.W., Campbell, T., Klein, A., 2004. Predicting the effect of beach nourishment and cross-shore sediment variation on beach morphodynamic assessment. Coastal Engineering 51, 839–861. Borja, A., Bricker, S.B., Dauer, D.M., Demetriades, N.T., Ferreira, J.T., Forbes, A.T., Hutchins, P., Jia, X., Kenchington, R., Marques, J.C., Zhu, C., 2008. Overview of integrative tools and methods in assessing ecological integrative in estuarine and coastal systems worldwide. Marine Pollution Bulletin 56, 1519–1537. Boudouresque, C.F., Bernard, G., Bonhomme, P., Charbonnel, E., Diviacco, G., Meinesz, A., Pergent, G., Pergent-Martini, C., Ruitton, S., Tunesi, L., 2012. Preservation et Conservation of Posidonia Oceanica Meadows. RaMoGe and RAC/ SPA publishing, Tunis. Buia, M.C., Gambi, M.C., Dappiano, M., 2004. Seagrass systems. Biologia Marina Mediterranea 11, 133–183. Clarke, K.R., 1993. Non-parametric multivariate analysis of changes in community structure. Australian Journal of Ecology 18, 117–143. Corsini, S., Inghilesi, R., Franco, L., Piscopia, R. 2006. Italian Waves Atlas. APAT Università degli Studi di Roma 3, Roma. Dayton, P.K., Tegner, M.T., Edwards, P.B., Riser, K.L., 1998. Sliding baselines, ghosts, and reduced expectations in kelp forest communities. Ecological Applications 8 (2), 309–322. Dean, R.G., Dalrymple, R.A., 2004. Coastal processes with Engineering Applications. Cambridge University Press. Della Via, J.D., Sturmbauer, C., Schonweger, G., Sotz, E., Mathekowitsch, S., Stifter, M., Rieger, R., 1998. Light gradients and meadow structure in Posidonia oceanica: ecomorphological and functional correlates. Marine Ecology Progress Series 163, 267–278. Diviacco, G., Coppo, S., 2007. Atlante degli habitat marini della Liguria: descrizione e cartografia delle praterie di Posidonia oceanica e dei principali popolamenti marini costieri. Regione Liguria, Genova. Downie, A., von Numers, M., Boström, C., in press. Influence of model selection on the predicted distribution of the seagrass Zostera marina. Estuarine, Coastal and Shelf Science. http://dx.doi.org/10.1016/j.ecss.2012.12.020. Duarte, C.M., Conley, D.J., Carstensen, J., Sánchez-Camacho, M., 2008. Return to Neverland: Shifting baselines affect eutrophication restoration targets. Estuaries and Coasts 32 (1), 29–36. EEC, 1992. Council Directive 92/43/EEC on the conservation of natural habitats and of wild fauna and flora. Official Journal of the European Communities. No L 206 of 22 July 1992. Efron, B., 1983. Estimating the error rate of a prediction rule: improvement on cross-validation. Journal of the American Statistical Association 78, 316–331. Fierro, G., Berriolo, G., Ferrari, M., 2011. Le spiagge della Liguria occidentale: analisi evolutiva. Regione Liguria, Genova. Folk, R.L., Ward, W.C., 1957. Brazos river bar: a study in the significance of grain size parameters. Journal of Sedimentary Petrography 27, 3–27. Folkard, A.M., 2005. Hydrodynamics of model Posidonia oceanica patches in shallow water. Limnology and Oceanography 50 (5), 1592–1600. Giovannetti, E., Lasagna, R., Montefalcone, M., Bianchi, C.N., Albertelli, G., Morri, C., 2008. Inconsistent responses to substratum nature in Posidonia oceanica meadows: an integration through complexity levels? Chemistry and Ecology 24 (S1), 145–153. Hobday, A.J., 2011. Sliding baselines and shuffling species: implications of climate change for marine conservation. Marine Ecology 32, 392–403. Infantes, E., Orfila, A., Bouma, T.J., Simarro, G., Terrados, J., 2012. Posidonia oceanica and Cymodocea nodosa seedling tolerance to wave exposure. Limnology and Oceanography 56 (6), 2223–2332. Infantes, E., Terrados, J., Orfila, A., Cañellas, B., Álvarez-Ellacuria, A., 2009. Wave energy and the upper depth limit distribution of Posidonia oceanica. Botanica Marina 52, 419–427. Jackson, J.B.C., Sala, E., 2001. Unnatural oceans. Scientia Marina 65, 273–281. Jackson, D.W.T., Cooper, J.A.G., Del Rio, L., 2005. Geological control of beach morphodynamic state. Marine Geology 216, 297–314. Koch, E.W., Ackerman, J.D., Verduin, J., Van Keulen, M., 2006. Fluid dynamics in seagrass ecology - from molecules to ecosystems. In: Larkum, A.W.D., Orth, R.J., Duarte, C.M. (Eds.), Seagrasses: Biology, Ecology and Conservation. Springer, pp. 193–225. Legendre, P., Legendre, L., 1998. Numerical ecology, 2nd English ed. Elsevier.

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038

M. Vacchi et al. / Marine Pollution Bulletin xxx (2013) xxx–xxx Leriche, A., Boudouresque, C.F., Bernard, G., Bonhomme, P., Denis, J., 2004. A onecentury suite of seagrass bed maps: can we trust ancient maps? Estuarine, Coastal and Shelf Science 59 (2), 353–362. Leriche, A., Boudouresque, C.F., Monestiez, P., Pasqualini, V., 2011. An improved method to monitor the health of seagrass meadows based on kriging. Aquatic Botany 95 (1), 51–54. Lyons, M.B., Roelfsema, C.M., Phinn, S.R. in press. Towards understanding temporal and spatial dynamics of seagrass landscapes using time-series remote sensing. Estuarine, Coastal and Shelf Science, http://dx.doi.org/10.1016/ j.ecss.2013.01.015. Molinier, R., Picard, J., 1952. Recherches sur les herbiers de phanérogames marines du littoral méditerranéen français. Annales Institut Océanographique, Paris 27, 157–234. Montefalcone, M., 2009. Ecosystem health assessment using the Mediterranean seagrass Posidonia oceanica: a review. Ecological Indicators 9, 595–604. Montefalcone, M., Albertelli, G., Morri, C., Parravicini, V., Bianchi, C.N., 2009. Legal protection is not enough: Posidonia oceanica meadows in marine protected areas are not healthier than those in unprotected areas of the northwest Mediterranean Sea. Marine Pollution Bulletin 58, 515–519. Montefalcone, M., Albertelli, G., Morri, C., Bianchi, C.N., 2010a. A. Pattern of widescale substitution within Posidonia oceanica meadows of NW Mediterranean Sea: invaders are stronger than natives. Aquatic Conservation, Marine and Freshwater Ecosystems 20, 507–515. Montefalcone, M., Parravicini, V., Vacchi, M., Albertelli, G., Ferrari, M., Morri, C., Bianchi, C.N., 2010b. B. Human influence on seagrass habitat fragmentation in NW Mediterranean Sea. Estuarine, Coastal and Shelf Science 86, 292–298. Montefalcone, M., Rovere, A., Parravicini, V., Albertelli, G., Morri, C., Bianchi, C.N., 2013. Evaluating change in seagrass meadows: a time-framed comparison of Side Scan Sonar maps. Aquatic Botany 104, 204–212. Parravicini, V., Rovere, A., Vassallo, P., Micheli, F., Montefalcone, M., Morri, C., Paoli, C., Albertelli, G., Fabiano, M., Bianchi, C.N., 2012. Understanding relationships between conflicting human uses and ecosystem status: a geospatial modeling approach. Ecological Indicators 19, 253–263. Pergent, G., Pergent-Martini, C., Boudouresque, C.F., 1995. Utilisation de l’herbier à Posidonia oceanica comme indicateur biologique de la qualité du milieu littoral en Méditerranée: état des connaissances. Mesogée 54, 3–27. Ricketts, P., Harrison, P., 2007. Coastal and Ocean Management in Canada: moving into the 21st century. Coastal Management 35, 5–22.

9

Rovere, A., Parravicini, V., Vacchi, M., Montefalcone, M., Morri, C., Bianchi, C.N., Firpo, M., 2010. Geo-environmental cartography of the Marine Protected Area ‘‘Isola di Bergeggi’’ (Liguria, NW Mediterranean Sea). Journal of Maps 6 (1), 505– 519. Rovere, A., Vacchi, M., Firpo, M., Carobene, L., 2011. Underwater geomorphology of the coastal tracts between Finale Ligure and Vado Ligure. Quaternary International 232 (1–2), 187–200. Smith, M.J., 2003. Surf zone hydrodynamics. In: Demirbilek, Z. (Ed.), Coastal Engineering Manual, Part II, Coastal Hydrodynamics, Engineer manual 1110–21100. US Army corps of Engineers, Washington, DC (Chapter II-4). Sorensen, R.M., 2006. Basic coastal engineering. Springer, New York. Stachowitsch, M., 2003. Research on intact marine ecosystems: a lost era. Marine Pollution Bulletin 46, 801–805. Vacchi, M., Montefalcone, M., Bianchi, C.N., Morri, C., Ferrari, M., 2010. The influence of coastal dynamics on the upper limit of the Posidonia oceanica meadow. Marine Ecology 31, 546–554. Vacchi, M., Montefalcone, M., Bianchi, C.N., Morri, C., Ferrari, M., 2012a. Hydrodynamic constraints to the seaward development of Posidonia oceanica meadows. Estuarine, Coastal and Shelf Science 97, 58–65. Vacchi, M., Rovere, A., Schiaffino, C.F., Ferrari, M., 2012b. B. Monitoring the effectiveness of re-establishing beaches artificially: methodological and practical insights into the use of video transects and SCUBA-operated coring devices. Underwater Technology 30 (4), 201–206. Valle, M., Borja, Á., Chust, G., Galparsoro, I., Garmendia, J.M., 2011. Modelling suitable estuarine habitats for Zostera noltii, using Ecological Niche Factor Analysis and Bathymetric LiDAR. Estuarine, Coastal and Shelf Science 94 (2), 144–154. Warton, D.I., 2011. Regularized sandwich estimators for analysis of high dimensional data using generalized estimating equations. Biometrics 67 (1), 116–123. Warton, D.I., Wright, S., Wang, Y., 2012. Distance-based multivariate analyses confound location and dispersion effects. Methods in Ecology and Evolution 3 (1), 89–101. Wentworth, C.K., 1922. A scale of grade and class terms for clastic sediments. Journal of Geology 30, 377–392. Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G.M., 2009. Mixed Effects Models and Extensions in Ecology with R. Springer, New York, NY, USA.

Please cite this article in press as: Vacchi, M., et al. Towards a predictive model to assess the natural position of the Posidonia oceanica seagrass meadows upper limit. Mar. Pollut. Bull. (2013), http://dx.doi.org/10.1016/j.marpolbul.2013.09.038