Results
A combined total of 3542 arthropod individuals were collected in June and August, with an approximately equal number of spiders (1703) and beetles (1839). The spider assemblage included 18 families and 114 species, with Linyphiidae, Lycosidae and Salticidae being the most abundant families. The beetle assemblage included 22 families and 136 species, with Staphylinidae, Chrysomelidae and Carabidae being the most abundant families.
When comparing spider and beetle responses based on the design characteristics, we found that only total beetle density varied between categories (Χ 5,59=18.8, P<0.005), which occurred because sites classified as dry-dry had a higher beetle density (Fig. 4). Spider density did not vary between the different flooding categories (P>0.1), and neither did rarefied species richness (P>0.2) of beetles and spiders. When comparing explanatory power of the proportion of dry pixels between spatial scales, we found that the variance explained (R2) for both spider and beetle densities was highest when the proportion of dry pixels was estimated at the scale of 10x10 m2 (Fig. 5a). At this scale, both spider (LR=5.9, df = 56, p<0.02) and beetle (LR = 11.8, df = 56, p<0.0006) densities increased with the proportion of surrounding dry pixels (Fig. 5b), and there was no effect of the local (2x2 m2) wetness-dryness, as expressed by the local inundation frequency. Rarefied species richness was not significant at any scale for either spiders or beetles, and was therefore not further analysed. When we included environmental variables and the heterogeneity index, we found that none of these variables explained variation in either the density or species richness of either spiders or beetles and these variables are not further discussed.
The multivariate analyses through manyGLM similarly showed that the proportion of dry pixels affected the community structure of both spiders (Wald=4.6, p<0.007) and beetles (Wald=5.2, p<0.001). These differences occurred because the densities of Staphylinidae (Wald=3.6, p<0.005) and Chrysomelidae (Wald=2.9, p<0.03) increased, with a similar non-significant trend Linyphiidae (Wald=2.5, p<0.08), whereas Salticidae (Wald=2.7, p<0.04) decreased with the proportion of dry pixels at the respective optimal spatial scale (Fig. 6, adjusted p-values). Other included groups (Lycosidae, Thomisidae, Tetragnathidae, Carabidae, Curculionidae and Coccinellidae) did not vary with the proportion of dry pixels. To further examine families with significant responses, we repeated the manyGLM at a genus level for Linyphiidae, Salticidae and Staphylinidae for genera with at least 20 specimens. For Linyphiidae, we could include five genera (Bathyphantes, Oedothorax, Porrhomma, Savignia, Silometopus ) and for these only Savignia (Wald=2.7, p<0.03) increased with the proportion of dry pixels. For Salticidae, we could only include two genera (Attulus andMarpissa ) and only Attulus (Wald=2.6, p<0.03) decreased with the proportion of dry pixels. For Staphylinidae, we found somewhat similar responses for all five included genera (Amischa ,Erichsonius , Paederus , Rybaxis , Stenus ), but only Amischa (Wald=2.8, p<0.04) and Rybaxis (Wald=3.3, p<0.01) significantly increased with the proportion of dry pixels. The other genera had 0.05<p<0.1.
When comparing habitat preferences, we found a relationship between the proportion of dry pixels and the density of beetles (Wald=4.6, p<0.002) but not of spiders (Wald=2.1, p>0.2). The relationship for beetles was caused by a positive relationship between density and the proportion of dry pixels for habitat generalists (Wald=4.4, p<0.002). When comparing traits, we found positive relationships between density and the proportion of dry pixels for both beetles (Wald=4.6, p<0.001) and spiders (Wald=4.4, p<0.02). For beetles, all three trophic groups (carnivores, herbivores and other consumers) responded similarly (Waldcarnivores=3.2, p<0.007; Waldherbivores=2.7, p<0.009; Waldother=3.9, p<0.002). For spiders, only sheet web weaver densities (Wald=3.3, p<0.008) showed a relationship to the proportion of dry pixels, matching the result for Linyphiidae that includes the sheet web weavers.