Our study locations were situated along the coast of Granada (Mediterranean coast of southern Spain). This area is characterized by the abundance of cliffs composed of carbonate rocks, which allow the presence of many marine caves. Six caves were studied: Cueva de Calahonda (36°42’46”N, 3°22’18”W), Cueva de la Punta del Vapor (36°43’22”N, 3°42’35”W), Raja de la Mona (36°43’10”N, 3°44’6”W), Cueva de los Treinta Metros (36°43’12”N, 3°44’9”W), Cueva de Cantarriján (36°44’16”N, 3°46’41”W) and Cueva de las Gorgonias (36°44’17”N, 46°46’42”W) (see also Navarro-Barranco et al. 2013). All of them show a similar topographic profile, with a single submerged entrance followed by a rectilinear blind-ending tunnel. These locations comprise a wide range of depth (from 6 to 32 m deep) and a heterogeneous granulometry (from coarse sand to predominantly muddy sediments).

During July and August 2011, two stations were established and sampled at each location, one inside the cave and one in the exterior area (each one approximately 10 m from the cave entrance). Seven replicate samples were taken at each station using a hand-held rectangular core of 0.025 m² to a depth of 10 cm by SCUBA diving. Three of these samples were immediately frozen for physicochemical characterization of each station. Granulometric parameters were determined following the methodology proposed by Guitián and Carballas (1976). Organic matter and major, minor and trace element concentrations in sediments were determined following the methodology presented in Navarro-Barranco et al. (2012). Four replicate samples were used for the biological study of the soft-bottom community. Samples were sieved through a 0.5-mm mesh sieve and preserved in 70% ethanol stained with Rose Bengal. In the laboratory, all amphipod species were sorted, counted and identified to species level when possible using binocular microscopes.

Mean values ± standard deviation (SD) of number of species, total abundance, Shannon-Wiener diversity index (Shannon-Weaver 1963) and abundance of each amphipod species were calculated for each station. Differences between external and internal stations for these indices were tested using an analysis of variance (ANOVA) with two factors: Position (Po) and Site (Si). Po was a fixed factor with two levels (internal and external stations). Si was a fixed factor, orthogonal with Po, with six levels (each marine cave). Four samples were considered in each position (n=4). Prior to analysis, homogeneity of variances was checked using Cochran’s test and transformations were applied when necessary (Underwood 1997). A preliminary correlation analysis was done to determine which chemical variables were highly correlated with each other (p<0.05) and these were eliminated from later analyses to reduce the number of variables considered. Correlation analyses were also done to evaluate the relationship between environmental (depth, granulometry and chemical data) and biological variables (abundance, species richness and Shannon diversity). A permutational analysis of multivariate dispersions (PERMDISP) was carried out to test differences in the variation of the amphipod community between internal and external habitats. Finally, the relationships between environmental variables and the amphipod species were explored by a canonical correspondence analysis (CCA). The species with very low abundances (those with five or fewer individuals) were not included in the analysis.

Univariate analyses were performed using the GMAV5 program (Underwood et al. 2002) for ANOVA and the SPSS.15 program for the correlations. CCA was carried out using the PC-ORD program (McCune and Mefford 1997), and the PRIMER v.6 + PERMANOVA package (Clarke and Gorley 2001) was used for PERMDISP.

Species were classified in ecological groups in function of their sensitivity to contamination, following the criteria of the Azti Marine Biotic Index (AMBI) and the BENTIX index and classification lists (Borja et al. 2000, Simboura and Zenetos, 2002). Both indices are among the most widely used and require classification to species level. AMBI is the most applied index and classifies the amphipods into three groups: sensitive, indifferent and tolerant species. On the other hand, BENTIX is a biotic index specifically designed for the Mediterranean Sea and only discriminates between sensitive and tolerant species, including the indifferent species (those always present in low densities and with no significant variations in time) in the sensitive group. The percentage of species belonging to each group was calculated for internal and external stations.

Species were also assigned to different biogeographic regions on the basis of information and classifications in Ruffo (1982-1998), Conradi and López-González (1999), Guerra-García et al. (2009) and the World Register of Marine Species (WoRMS) (www.marinespecies.org). The following groups were established:

Mediterranean endemic
Mediterranean and Atlantic Ocean
Mediterranean, Atlantic and Indo-Pacific Ocean
Mediterranean, Atlantic and Arctic Ocean
Cosmopolitan

Traditionally, the diversity impoverishment of cave faunas has been a widely accepted pattern. This fact was explained by the reduced trophic supply inside submarine caves (Zabala et al. 1989, Fichez 1990). However, in recent years quantitative studies conducted simultaneously in different caves have highlighted the difficulty of finding general and consistent patterns in marine cave communities (Bussotti et al. 2006, Navarro-Barranco et al. 2013). In the present study, mean values of abundance, species richness and Shannon diversity were higher at external stations. However, although this trend has been observed in most studies of marine caves, it cannot be considered as a constant pattern. The finer sediments inside caves are often more stable and show a higher concentration of organic matter, often allowing the existence of complex and rich soft-bottom communities. For example, internal stations of the Cantarrijan cave showed abundances of amphipods more than ten times higher than external stations because of an unusually high density of Harpinia pectinata. We found another exception to the general pattern in Punta del Vapor, where samples showed lower diversity values at external stations, mainly because of the overwhelming numerical dominance of Siphonoecetes sabatieri. This lack of general patterns is not surprising. Despite their apparent environmental homogeneity, marine caves often show high variability because many complex mechanisms, both physical and biological, structure the soft-bottom communities (e.g. Morrisey et al. 1992, Snelgrove 1998, Norén and Lindegarth 2005). Moreover, each marine cave has particular environmental conditions determined by features such as its topography, depth and orientation. As a result, the behaviour of the soft-bottom communities inside marine caves is difficult to predict.

Most of the species found inside the caves were also present at external stations. Among the species only found at external stations, some of them seemed to show a special avoidance of cave habitats. Siphonoecetes sabatieri, Urothoe elegans (Bate, 1857) and Pontocrates arenarius (Bate, 1858) are some of the species with a wider distribution in the study, being present in most of the sampling areas but with no single individual inside the caves. On the other hand, some species, such as Microdeutopus algicola Della Valle, 1893, Liljeborgia sp. and Gammarella fucicola (Leach, 1814), were only present at internal stations. Anyway, none of the species found in the present study are cave-exclusive, having all been previously cited in open habitats (Ruffo 1982-1998). This is a common feature in marine cave fauna, since most of the species only use the caverns for shelter and can also be found outside the caves in similar habitats (Scipione et al. 1981, Gerovasileiou and Voultsiadou 2012). The genus Harpinia, for example, was found predominantly inside marine caves, but was also present at some external stations, thus facilitating intercommunication among cave assemblages. Therefore, amphipod cave communities are not as isolated as could be expected, and the stepping stones model can probably be used to understand their population dynamics.

It was difficult to compare the diversity results obtained in the present study with those found in shallow soft-bottoms habitats of other localities of the Iberian Peninsula, since there is a great variation among studies in the sampling methodology (number of samples, degree of spatial and temporal replication, etc). Studies carried out in the Atlantic zone of the Iberian Peninsula found between 34 and 79 different amphipod species in the northwestern area (Cunha et al. 1999, Lourido et al. 2008, Moreira et al. 2008, Cacabelos et al. 2010) and 93 species on the southern Portuguese coast (Carvalho et al. 2012). On the other hand, in the Mediterranean area, an inventory of soft-bottom amphipods from the east coast of the Iberian Peninsula revealed the presence of 55 different species (de-la-Ossa-Carretero et al. 2010). This value is slightly higher than that found in the present work. However, taking into account the huge differences in the sampling effort between these two studies (40 stations along 250 km of coast were sampled during five consecutive years in the east coast study), the total amphipod species richness in shallow sediments could be expected to be higher in the Alboran Sea area. The relevance and high diversity of the Alboran Sea in comparison with other Mediterranean areas, mainly owing to the overlap of Atlantic and Mediterranean faunas, has already been reported (Guerra-García et al. 2009). In this sense, only in the sediments of Ceuta Harbour and Algeciras Bay (both of them in the Strait of Gibraltar), 44 and 55 amphipod species were collected, respectively, by Conradi and López-Gónzalez (1999) and Guerra-García and García-Gómez (2004). Moreover, the existence of several upwelling areas in the littoral of southern Spain may also contribute to the high biological richness of this area (Delgado 1990).

Most of the species found have a wide range of distribution, present in more than one biogeographical region. However, only one species, Ampelisca brevicornis (Costa, 1853), can be considered cosmopolitan. The percentage of Mediterranean endemism (20.6%) was similar to that obtained in previous amphipod studies in the Strait of Gibraltar area (Conradi and López-Gónzalez 1999, Guerra-García et al. 2009), but surprisingly higher than that obtained on the east coast of the Iberian Peninsula (12.7%) (de-la-Ossa-Carretero et al. 2010).

All the variables included in the CCA were significant but sediment granulometry, organic matter and N concentration were the main parameters explaining the distribution of amphipods; this finding is a common feature in soft-bottom ecology studies (Pearson and Rosenberg 1978, Carvalho et al. 2012).

Regarding the ecological quality of the communities studied, the high abundance of sensitive species indicated the good state of the assemblages. The chemical data obtained agree with that conclusion, since almost all the values at all stations were far below the highest reference values tolerated in marine sediment quality guidelines (Grimwood and Dixon 1997, Long et al. 1998). Thus, because of the lack of polluted stations, it is not possible for us to evaluate the suitability of the indices for detecting the anthropogenic impacts in marine caves. Although axis 1 in the CCA correlated significantly with the concentration of organic matter, N, P, Pb and Cu, this first axis does not represent a pollution gradient, since the concentration of heavy metals and the organic enrichment levels was not very high at any station. The rise in these concentrations is associated with the increase (mainly at internal stations) in the percentage of silt and clay, which have a higher capacity to retain elements. The a priori tolerant species, such as Gammarella fucicola, Pariambus typicus and Leptocheirus hirsutimanus, appear separated in the CCA. However, another way to evaluate the suitability of the indices is to analyse the level of agreement between results obtained by different indices (Borja and Dauer 2008). In this case, both indices showed similar results for the percentage of sensitive species and BENTIX showed slightly higher results for tolerant species. These small differences were due to some discrepancies in the assignment of ecological groups for some species. The caprellid Phtisica marina Slabber, 1769, for example, is classified as tolerant by BENTIX and sensitive by AMBI. The analysis of the ecology of soft-bottom caprellids in Southern Spain carried out by Guerra-García and García-Gómez (2001) and Guerra-García et al. (2012) supports the BENTIX classification, since it is able to tolerate conditions of high organic enrichment and low oxygen concentrations. The gammarid Perioculodes longimanus is considered a tolerant species by AMBI and a sensitive one by BENTIX. P. longimanus was present at most of the stations, regardless of the granulometry, depth or position (external or internal). Its abundance at these stations with high ecological quality was often very high, suggesting that P. longimanus should probably be considered an indifferent species in the AMBI’s list of species. Other discrepancies between indices occur for the gammarids Pontocrates arenarius, Leptocheirus pectinatus (Norman, 1869) and Microdeutopus algicola. However, since all our stations showed a good state of conservation and some of these species were in very low densities, we were unable to discriminate the tolerance of these species. The study area is subjected to a low anthropogenic influence and some of the caves are even inside a marine protected area (Rodriguez et al. 2003). Future studies involving marine caves with a higher gradient of pollution stress would be very useful to properly evaluate the tolerances of different taxa and the suitability of biotic indices in marine caves.