Results
Effects of artificial ecological islands on macroinvertebrates
In total, 12956 macroinvertebrate specimens were collected in the park; 85 taxa belonging to 13 orders and 6 classes were identified, most of which were arthropods (62 species, accounting for 72.94%). Mollusks were second with 15 species (17.65%). There were only 8 species (9.41%) of annelids. SIMPER analysis showed that only 3 species were the primary contributors at all sites, Parafossarulus striatulus(35.72%), Palaemon modestus (34.84%) and Radi plicatula(13.49%), with a cumulative contribution of approximately 84% and a total similarity of 38.17 (the list was truncated when 80% was reached).
The species and abundance of macroinvertebrate fauna in the park increased continuously in 3 years. The abundance of Chironomuslarvae decreased, but those of Odonata and Rhynchobdellida Blanchard increased continuously (Figure 3).
The species abundance and diversity were higher in the island communities than in the open-water area. Regarding the two water levels on the islands, the richness in the shallow-water-level area was higher but not as evenly distributed as that in the deep-water-level area (Figure 4).
Two-way crossed ANOSIM showed that there were significant differences among the open-water area and island two levels regarding the macrobenthos community composition in 3 years (Table 1). ANOVA showed that in the 12 orders used as classifiers, the mean abundances of 8 orders in the open-water area and on the islands were statistically different (Table 2). Additionally, there were three orders of classifiers that were significantly different between the two island levels (Table 2). The hierarchical cluster dendrogram of the 20 macrobenthos communities assessed with SPSS software was basically consistent with the MDS. There were two groups: islands and open-water area (Figures 5 & 6). Due to the large changes in substrate, hydrophytes and depth in open-water area samples, the order was not relatively uniform.
Effects of substrate and construction time of the artificial ecological islands on macroinvertebrates
Following MDS, we placed island B into the SSI group, and the community composition differed significantly between the two substrates (Table 3, Figures 5 and 6). Seven taxa between the two were significantly different (Table 2). The PSI groups had more species and were more evenly distributed than the SSI. In the MDS results (Figure 7), the island group was divided into two groups (dimension 2), and the division was related to the different substrates. In the classification, PSI B was classified as a SSI. The possible reason is that island B has a large area, and a few stones were laid around the island. The outlying part of the island is covered with a few pebbles; only the part near the centre of the island is covered with more pebbles. It is also covered with cattails, reeds and other vegetation, making it similar to the soil islands. To explore the change trend in macroinvertebrate diversity with the extension of island construction time, we also compared the community compositions of the islands with two construction ages. The results showed that there were significant differences (Table 3), and 3 of the 12 orders were significantly different (Table 2). The species and diversity were higher on the islands that were built relatively later (Figure 7). Compared with the vegetation biodiversity and abundance on these islands, the PSIs were higher in plant species and biodiversity than SSIs although plant abundance was lower. In addition, the vegetation biodiversity on the PSI built first was higher than that on the PSI built later, while the vegetation biodiversity on SSI built first was lower than that on SSI built later. (Figure 8).
Effects of artificial ecological islands on waterfowl
In total, 5987 individuals were recorded, and 33 species identified in the park belonged to 8 families and 7 orders. SIMPER analysis showed that nine species contributed the most in three years: Fulica atra (15.49%), Podiceps cristatus (10.28%), Chlidonias hybrida (9.66%), Egretta alba (8.59%), Aythya ferina(8.47%),Anas poecilorhyncha (7.63%), Podiceps ruficollis (7.61%)Chlidonias leucoptera (6.83%) and Anas platyrhynchos(4.85%), with a cumulative contribution of approximately 80% and a total similarity of 65.42.
Changes in the species composition of waterfowl in the shallow water areas that consisted of islands and deep water areas without islands are shown in Figure 9. There were more wading bird species than swimming birds in the park, but the wading birds were less abundant. Comparing the distribution of waterfowl at the two water levels, the number and abundance of species in the shallow water area were higher than those in the deep water area, Charadriidae, Scolopacidae and Ardeidae, for example. In contrast, the shallow water areas had more Podicipedidae and Rallidae.
Pearson correlation analysis showed that there was a negative correlation between the diversities (H’) of waterfowl and macroinvertebrates (r = -0.997, P = 0.05). Therefore, the waterfowl and macroinvertebrate data were compared. The macroinvertebrate diversity index had an upward trend, but the waterfowl diversity index decreased (Figure 10). The possible reason for the effects on the composition of the waterfowl community is that summer is the peak tourist season in the park, and the number of visitors gradually increased during those years. The number of visitors was negatively correlated with the diversity (\(r_{H^{\prime}}\) = -1.000) and richness (\(r_{d}\) = -0.999) of waterbirds. Human activities interfere with the foraging behaviours of birds and reduce their activity space.