3 RESULTS
Overall, two amphibian species, eleven fish species, three plankton orders, and 48 benthic macroinvertebrate families were recorded in the three surveys. Assemblages of amphibians, fishes, macroinvertebrates, and planktons, respectively, all varied with connectivity gradient and season (Fig. 3, Table S2; two-way PERMANOVA, p <0.05). Consequently, the overall aquatic assemblage including all faunal groups was significantly influenced both by connectivity gradient and season, and their interaction was marginally significant (Fig. 4, Table S3). In spring, “No-flow” pond assemblages were significantly different from “Late” ponds and “Flowing” sites. In summer, significant differences between “No-flow” pond assemblages and “Late” ponds remained, and “Flowing” site assemblages were significantly different from all other pond sites. In autumn, differences between “No-flow” pond and “Late” pond assemblages became insignificant, and only the difference between “Early” pond and “Flowing” site assemblages were significant (Fig. 4). In autumn, mean square of variation was lower between categories and their residuals were higher.
Variation partitioning with generalized dissimilarity modelling showed the factors driving the β-diversity between sites change seasonally. The explained deviance of aquatic assemblage was 38% in spring, 57% in summer, and 53% in autumn (Fig. 5a). Within the explained variation, the proportion explained by connectivity (pure connectivity or combination with environment) was 46% in spring, 44% in summer, and then declined to 20% in autumn (Fig. 5b). The importance of each environmental variable changed seasonally (Table S4). In spring, connectivity at the downstream end and area of ponds significantly influenced the assemblage, explaining 23% and 22% of the total deviance, respectively. In summer, dissolved oxygen, water temperature, area of ponds, and connectivity at the downstream end significantly influenced on the assemblage, explaining 11%, 9%, 8% and 3% of the total deviance, respectively. In autumn, area of ponds, flow velocity, DO, temperature, and grain size significantly influenced on the assemblage, explaining 28%, 13%, 7%, 6% and 3% of the total deviance, respectively. Supplementary, we conducted variation partitioning with generalized dissimilarity modelling for benthic macroinvertebrates, fish, and planktons in the same manner (we could not apply the analysis to amphibian because of data limitations), and confirmed that the contribution of connectivity to the whole variation declined across seasons for all faunal groups (Fig. S1). CCA analysis of the amphibious organisms (benthic macroinvertebrates and amphibians) in autumn visually explained the effects of local environmental factors, and the permutation test showed significant influence of the local environmental factors on the assemblages (F 9,12=1.56, P = 0.038) (Fig. 5c).
SIMPER analysis showed seasonal changes in the taxa contributing to the assemblage difference along hydrological gradients (Table 1). Notably, the simper analysis shows that the three taxa, Hynobius retardatus , Rana pirica and Limnephilidae, which primarily contributed to the variation between “No flow” and “Early” as well as the variation between “Early” and “Late,” decreased by summer and then autumn, potentially causing loss in variation between those habitats in autumn. Furthermore, Leptophlebiidae and Nemouridae, which were the two primarily important taxa creating the differences between all ponds and flowing sites in summer, were both present in ponds and flowing sites in spring but disappeared only in ponds by summer. This led to the clear difference of the ponds and flowing site assemblages.