Multi-scale spatial variability in fish assemblages associated with Posidonia oceanica meadows in the Western Mediterranean Sea

Abstract

Fish assemblages associated with Posidonia oceanica from three locations of the western Mediterranean (Mallorca, Formentera and Alacant, Spain) were sampled in order to assess their spatial variability at three different scales ranging from <1 km to >100 km. Sampling was carried out using a beam trawl. Simultaneous sampling at these three spatial scales with the appropriate number of replicates implies a huge effort which is rarely possible to achieve. Consequently, we propose an arrangement of data coming from different sampling programs, after testing and making sure the specificities of each program cannot be confounded with the spatial variability. The two preliminary experimental designs adopted in order to combine the datasets were: (1) differences between two consecutive years were tested by sampling the same meadow (Formentera) in June 2001 and June 2002; and (2) differences attributable to the meadow structure were evaluated by sampling two meadows (Mallorca 2000 and Formentera 2001). The absence of any significant correlation pattern for univariate community descriptors and multivariate species-specific densities (i.e. individuals per hour) related to both between-year and within-location structural differences of the meadows, allow us to combine data from the three locations (Formentera 2001, Mallorca 2000 and Alacant 2000) in a single analysis aimed at determining how much variability is explained by the three spatial scales considered. The between-location scale (>100 km) is the most variable scale for species-specific densities (multivariate approach). Spatial variability at the smallest scale (<1 km) is also considerable, but the variability corresponding to the intermediate scale (<10 km) was found to be non-significant. This is an expected result for fishes given that the spatial scale of individual transects is not large compared to the high mobility of many of the species considered. The differences observed between the locations placed 100 km apart are due to changes in the relative density of species rather than differences in species composition. Between-location variability in the univarite community descriptors was not significant. However, within-location (intermediate scale) differences were significant for density but not for biomass. This is related to the large number of small individuals found in one Formentera site. These results are consistent with the hypothesis that P. oceanica meadows from different locations in the western Mediterranean might display a similar carrying capacity although large-scale hydrodynamic conditions and meadow structure at the between-location level might lead to differently shaped fish communities (for example, the larger meadow complexity in Formentera might favour smaller sizes, since small species and/or individuals are known to find more shelter and food there), while differences at an intermediate spatial scale remain irrelevant.

Introduction

The Posidonia oceanica beds are among the most important Mediterranean ecosystems, and their conservation is a high national and international priority. Posidonia oceanica beds have a multifunctional role within coastal systems that is comparable to that of other seagrasses in temperate and tropical areas. They offer substrate for settlement, food availability and shelter, recruitment and nursery areas, as well as participating in key biogeochemical and geological processes (Orth et al., 1984, Orth, 1992, Proccacini et al., 2003, Nakaoka, 2005).

Fish assemblages associated with Posidonia oceanica seagrass meadows have been extensively studied in the Mediterranean. For instance, community structure and diel variations (Bell and Harmelin-Vivien, 1982, Harmelin-Vivien, 1984), spatial and temporal fluctuations (Harmelin-Vivien, 1982, Francour, 1997), feeding habits (Bell and Harmelin-Vivien, 1983, Khoury, 1984, Harmelin-Vivien et al., 1989), comparisons with other inshore habitats (Guidetti, 2000), effect of bottom trawling (Sánchez-Jerez and Ramos Esplá, 1996), effect of protection (Francour, 1994, Francour, 2000, Macpherson et al., 2002) and methodological bias in sampling methods (Harmelin-Vivien and Francour, 1992) should be considered well known in this area.

Nevertheless, little attention has been given to the variability in the fish assemblages associated with Posidonia oceanica meadows at different spatial scales. Studying spatial patterns is of ecological importance in order to understand the causes of abundance of organisms, and it also provides valuable insights for management and conservation. The topic of variability at several spatial scales, from meters up to hundreds or thousands of kilometres, has been addressed for fish communities other than the one considered here. Specifically, there are numerous examples concerning reef fishes (Choat and Ayling, 1987, Sale, 1998, Chesson, 1998, García-Charton and Perez-Ruzafa, 2001, Guidetti et al., 2002, García-Charton et al., 2004, Chittaro, 2004, Anderson and Millar, 2004, Nuñez-Lara et al., 2005). This is a relevant topic because the patterns that can be observed (and the processes that lead to them) depend on the extent to which the system is examined (Sale, 1998). For example, analyzing multivariate variability in reef fish assemblages at different spatial scales revealed that the greatest variation occurred at the smallest spatial scale, between individual transects (separated by a few meters), and in contrast, variability from site to site (separated by hundreds to thousands of meters) and from location to location (separated by hundreds of kilometres) were comparable (Anderson and Millar, 2004).

Hierarchical multi-scale approaches are conceived for understanding how information is transferred across scales but they are rarely integrated into survey methodologies and analysis of the results. This is probably due to the huge effort that simultaneous sampling at multiple spatial scales with the appropriate number of replicates at each spatial level implies. The marine environment has particular practical, logistical and financial challenges that make it more difficult to obtain this type of dataset (Raffaelli et al., 2005). The analysis of anthropogenic and geomorphologic effects on reef fish communities along 400 km of the coast of the Yucatan Peninsula (Nuñez-Lara et al., 2005) and the analysis of the effect of habitat structure on reef fishes at a large scale of hundreds of kilometres along the north-eastern coast of New Zealand (Anderson and Millar, 2004) are two notable exceptions. Similarly, this specific matter has been studied in relation to Mediterranean rocky reef fish assemblages covering more than 500 km, but at the price of extending the sampling program from June to October (García-Charton et al., 2004). According to this contribution, the variability observed at the largest spatial scale in the structure of the fish assemblages studied seems to be largely determined by differences in local carrying capacity and hydro-climatic conditions, but at small-to-intermediate spatial scales the habitat structure is more likely to be responsible for a large part of the observed differences. The importance of habitat structure has also been studied in the case of Australian seagrasses (Bell and Westoby, 1986a, Bell and Westoby, 1986b, Bell and Westoby, 1986c). Specifically, the effect of physical complexity seems to affect the abundance and distribution of fishes differently depending on the spatial scale (i.e. local scale versus an entire bay).

Therefore, a multi-scale analysis should at least consider the potentially confounding effects of both seasonality and habitat structure (at the intermediate scale, i.e. from more than 1 km to less than 10 km), but a fully factorial experimental design that takes all these factors into account is usually unavoidable. Consequently, a combination of data coming from different sampling programs is proposed, after testing and insuring that the specificities of each program cannot be confounded with spatial variability per se. The two preliminary experimental designs adopted here to allow the datasets to be combined were: (1) the differences between two consecutive years were tested by sampling the same meadow (Formentera) in June 2001 and June 2002; and (2) the differences attributable to the meadow structure (at the within-location scale; in terms of cover and shoot density) were evaluated by sampling two meadows (Mallorca 2000 and Formentera 2001). The absence of any significant correlation pattern related to these factors allows us to combine the data from the three locations (Formentera 2001, Mallorca 2000 and Alacant 2000) in order to carry out a single analysis aimed at determining how much variability is explained by the three spatial scales considered (i.e. <1 km, <10 km and >100 km).

In addition, in terms of multi-scale analyses, this contribution represents the first attempt in the Mediterranean to analyze the correlation patterns between the structure of the Posidonia oceanica meadows and the fish assemblages that inhabit them. Moreover, in the geographical context of the Balearic Islands and the Iberian coast, detailed descriptions of the characteristics of the fish assemblages associated with P. oceanica meadows are scarce (Massutí, 1962, Reñones et al., 1995, Jiménez et al., 1997) despite the fact that these meadows are among the most important habitats in the area since they cover nearly 3100 km² (Más et al., 1993).