Results
The analysis on spider communities included 3931 spider individuals separated into 16 taxa (Fig. 2). The variation in community composition was explained by a region-by-inland/coast interaction (Wald statistics=8.3, p<0.001) and not by the presence or absence of a thick wrack bed (Wald statistics=4.7, p>0.1). The region-by-inland/coast interaction arose because of a larger difference between southern and northern coastal sites compared with southern and northern inland sites (Fig. 2). When comparing abundances at the species level (Table 1), four species (Pardosa agrestis , P. agricola, Arctosa leopardus and Alopecosa cuneata ) were found almost exclusively at southern coastal sites and three taxa (Pardosa prativaga, P. amentata and Pirata spp. [mainlyP. piraticus ]) almost never occurred in these sites but were abundant elsewhere (Fig. 2). In addition, one species (P. monticola ) was mainly coastal whereas another species (Pardosa palustris ) occurred mainly inland, irrespective of region.
The variation in the prey community was explained by a region-by-wrack interaction (Wald statistics=11.4, p<0.001) and by season (Wald statistics=11.6, p<0.002) (Fig. 3). The region-by-wrack interaction occurred because Coleoptera (Deviance=14.1, p<0.02) and Sciaridae (Deviance=20.3, p<0.003) were positively affected by wrack availability only in northern sites, whereas Dolichopodidae (Deviance=11.5, p<0.05) was negatively affected by wrack availability only in southern sites (Fig. 3). The seasonal effect occurred because Empididae (Deviance=12.4, p<0.04) and Homoptera (Deviance=12.8, p<0.03) were more abundant during early season in June, whereas Chironomidae (Deviance=14.7, p<0.03) and Trichoptera (Deviance=12.5, p<0.04) were more abundant during August.
The number of prey species encountered in the gut of spider individuals varied between one and 15, with an average of 3.9 prey species. The dominant order in the guts was Diptera, both Brachycera (60%) and Nematocera (18%), with minor amounts of other groups; Homoptera (10%, mainly Cicadellidae and Delphacidae), Collembola (4%), other flying prey (3%, Hymenoptera and Lepidoptera), Formicidae (2%), Acari (2%) and Heteroptera (1%) (Fig. 4, SI Table 2). The diet contents varied considerably among sites and were mainly explained by wrack (Lawley-Hotelling trace statistics=6.1, p<0.001) and season (Lawley-Hotelling trace statistics=2.6, p<0.003), with an almost significant effect from a region-by-wrack interaction (Lawley-Hotelling trace statistics=1.7, p<0.07) (Fig. 4). However, there was no effect of spider species either when including this variable alone or in combination with other variables, or when run separately for region. Because of the almost significant region-by-wrack interaction on gut contents, we repeated the analysis for sites with or without wrack separately. In this analysis, region was significant for sites without wrack (p<0.03) but not for sites with wrack (p>0.2). The prey groups explaining the region difference for sites without wrack were Sphaeroceridae (F1,6=141, p<0.003) and Enchytraeidae (F1,6=111, p<0.05), that both had a higher frequency in spider guts from southern sites (Fig. 4). Finally, the comparison between sites with or without wrack suggested that mainly Sphaeroceridae (F1,11=17.2, p<0.06) and Heteroptera (F1,11=19.9, p<0.05) were more abundant in spider guts from site with wrack.
The species accumulation curves indicated that prey diversity was higher in southern sites and in sites with no wrack compared with northern sites and wrack sites (Fig. 5). When comparing diet consistency, we found that individual spiders had, on average, a Jaccard similarity of diets = 0.056 (sharing approximately 5.6% of the prey species consumed by two individuals). Diet consistency between any pair of individual spiders varied with region (β =0.025, p<0.001), wrack (β=0.080, p<0.001), and their interaction (β=-0.042, p<0.001). The interaction arose because spider from northern wrack sites had higher diet consistency than spiders from southern wrack sites, whereas spiders from northern non-wrack sites had lower consistency than those from southern non-wrack sites (Fig. 6). These diet similarities were larger when performed for pairs of the same spider species (Fig. 6) but patterns were otherwise similar (region: β=0.062, p<0.001; wrack: β=0.108, p<0.001; interaction: β=-0.099, p<0.001) and when diets were estimated at the prey family level both for all spider individuals (region: β=0.038, p<0.001; wrack: β=0.094, p<0.001; interaction: β=-0.059, p<0.001) and for pairs of the same spider species (region: β=0.070, p<0.001; wrack: β=0.136, p<0.001; interaction: β=-0.124, p<0.001).