Decoupling species richness variation and spatial turnover in beta diversity across a fragmented landscape

Study site

The study was conducted in a land-bridge island system in a man-made reservoir, the TIL in East China. The climate is subtropical monsoon with an average annual temperature of 17 °C and precipitation of 1,430 mm (Editorial committee of development of Xin’an River, Xin’an River Development Corporation, 2009). The inundation following the Xin’an River Dam construction for power generation in 1959, resulted in 1,078 fragmented islands (>0.25 ha) when the water reached its highest level (108 m a.s.l.). The areas of most islands are less than one ha. Currently, most islands are covered with forests (∼90%) dominated by Masson pine (Pinus massoniana) in the canopy and broad-leaved plants in the sub-canopy and understory. Forests were formed by secondary succession through regeneration after dam construction (Wilson et al., 2016). The majority of TIL has been protected by government and the vegetation has experienced few human disturbances over the past several decades. The TIL system may be considered as an “experimental” fragmented landscape with homogeneous matrix and synchronous successional history (Hu et al., 2011; Wilson et al., 2016; Yu et al., 2012).

Data collection

Field work for plant communities

We established long-term forest plots on 29 islands (Fig. 1) with little human disturbance across different gradients of area and isolation in 2009–2010 (Hu, Feeley & Yu, 2016). To assess whether the 29 studied islands are representative of the larger fragmented landscape, Wilson (2018) compared the distributions of island area, distance to mainland (DM) and distance to nearest neighbor island of the 29 studied islands and 510 islands in the fragmented landscape using the Kolmogorov–Smirnov test. Results showed there are no significant differences between the distributions of the 29 islands and all islands in the landscape for island area (D = 0.20, P = 0.22). Plot establishment and data collection followed the protocol for plot censuses established by the Center for Tropical Forest Science—Forest Global Earth Observatory network (Condit, 1998; Condit et al., 1999). The census of plots took one of three shapes: (1) cover the entire island (island area ≤1 ha); (2) 0.5 ha sampling plot (island area one to five ha) and (3) one ha plot (island area >5 ha) on these study islands. All woody individuals with diameter at breast height (DBH) >1 cm were measured in height and DBH , and were identified to species. Rarefaction analyses indicated that sampling protocols were sufficient to capture most woody plant species on the islands (Yu et al., 2012).

In total, 186,781 individuals of 76 woody plant species were recorded on the 29 islands.

Data analysis

Pattern of α-diversity

Species richness variation between island pairs was significantly influenced by island area for all plants and individual PFTs (Table 1). More species were recorded in islands in the large island group. DNI and DM slightly affected richness variation of different PFTs.

Pattern of β-diversity

β_RC value of all plants, PFT1 and PFT2 were negative (Fig. 2C), and had significant correlation with island areas (Table 1). PFT3 and PFT4 had positive β_RC values. β_RC of PFT3 only correlated with area, while β_RC of PFT4 was affected by area, DM and the area-isolation interaction.

β_J of all plants, PFT1, PFT2, and PFT3 increased with the increasing island area and DM. But the area-DM interaction had negative effects (Table 1). DNI had no significant effects on β_J.

Contributions of richness variation and spatial turnover to β-diversity

Using variation partitioning, richness variation independently explained 51% of the β-diversity (β_J) for all islands, while spatial turnover only explained 1% (Fig. 3A). In the small island group, richness variation was also the main factor explaining the β-diversity. But in the large island group, the independent explanation of spatial turnover increased to 9%, while richness variation only explained 6% (Fig. 3B). Isolation has no effect on the relative contribution of richness variation and spatial turnover in β-diversity (Fig. 3C and 3D).

Plant functional types had different performances in the variation partitioning analysis. β-diversity (β_J) of PFT1, PFT2, and PFT4 were only explained by richness variation (Figs. 4A, 4B and 4D). But spatial turnover explained 17% variation in β-diversity for PFT3, regardless richness variation (Fig. 4C).

Effects of island area on α- and β-diversity

In the TIL decreased patch size due to habitat loss has a strong negative impact on species richness, whereas increased isolation has no significant impact on species richness (Wang et al., 2010; Yu et al., 2012). In our previous studies, we found the decline of habitat heterogeneity caused by area loss was the main driver of this decreased α-diversity for multiple taxa, such as plants, birds, and lizards (Wang et al., 2010; Yu et al., 2012).

Island area also influenced the variation in community structure, involving the α-independent spatial turnover (e.g., β_RC) and α-dependent dissimilarity (e.g., β_J) across islands. Variation partitioning showed that the variation in species richness explained nearly 50% of the variation in β-diversity (Fig 3A). The variance explained by each factor in GLM analysis also supports this finding, with island area explaining more of the deviation in GLMs predicting β-diversity than other parameters (Table 1). Communities on the island pairs with greater area variation were more dissimilar, suggesting an area effect working on β-diversity across islands. This relationship was also found in small mammals in Atlantic fragmented forests, which was explained by the higher vulnerability of these species and increasing environmental heterogeneity across small fragments (Pardini et al., 2005). We speculate area effect on plant β-diversity was caused by habitat divergence and following environmental filtering (Myers et al., 2015) on the different islands. This speculation is informed from our previous studies, which showed environmental filtering was the main driver of community assembly in seedling-sapling transition (Hu, Feeley & Yu, 2016).

However, we also found evidence that environmental filtering drove the species assembly at different scales. The large island group had relatively high α-diversity and high variation explained by spatial turnover. We also found that GLM coefficients associated with area were positive when predicting β_RC (e.g., as the difference is area between island pairs becomes larger the island communities become more dissimilar independent of richness variation). Together this suggests that higher habitat heterogeneity within large islands could be driving community assembly on these islands. In contrast, habitat diversity is lower within, but higher across the small islands (Liu et al., 2018b). As mentioned above, our GLMs showed that the difference is island area is positively related to community dissimilarity at the island scale once richness variation was eliminated. Combined with relatively low α-diversity and low variation explained by spatial turnover this suggests that environmental filters shared across small islands play an important role in driving community structure. Our previous studies have also found steeper species-area relationship in the more fragmented landscapes with smaller islands (Hu et al., 2012), which illustrated the severe richness variation occurred in small island groups, and drove the β-diversity pattern.

Effects of island isolation on α- and β-diversity

Isolation has no or slight effects on species richness in this system, as demonstrated by several studies in TIL (Hu et al., 2011; Hu, Feeley & Yu, 2016; Yu et al., 2012). However, isolation, especially the DM, significantly affected the β-diversity including β_J and β_RC, albeit with relatively low amounts of deviance explained. That species similarity and spatial turnover were both affected by DM illustrates that most species on islands were strongly linked with the species pool on the surrounding mainland. Selective immigration of pollinators or seed dispersers could be one explanation for this β-diversity pattern (Wang, Chen & Ding, 2011). However, different from oceanic islands, islands far from the mainland in TIL are located in the center of the lake. It caused the difference of habitat quality between near islands (close to lakeshore) and far islands (distant to lakeshore). The worse habitat quality on near islands was influenced by the disturbance from surrounding lakeshore. Higher density of beetles founded in the far islands (Zhou et al., 2017) also support this explanation. In TIL region, we speculated that habitat quality is more important than immigration, which caused DM positively affect the β-diversity.

Responses of PFTs

There was little difference in richness variation and β-diversity between PFTs. The only significant differences in the richness-independent spatial turnover were between common species (PFT1, 2) and rare species (PFT3, 4), illustrating different community assembly processes for each PFT. Lower spatial turnover of common species and high spatial turnover of rare species (Fig. 2C) suggests that deterministic competition exclusion or shared environmental filters across fragments drove the assembly process (Chase et al., 2011) of common species, while specific habitat demands or low dispersal among fragments (dispersal limitation) drove the assembly of rare species.

Generalized linear models analysis illustrated that PFT4 was impacted by the largest number of fragmentation altered parameters. This suggests that plants with high SLA leaves and low tolerance to disturbance could suffer severe impacts from habitat fragmentation. Positive β_RC values of PFT3, suggests that the specific habitat requirements regulated the occurrence of these species. PFT3 also showed a unique relationship between α-diversity and β-diversity relative to other PFTs, with spatial turnover of PFT3 primarily driving β-diversity (Fig. 4C). Specific habitat requirements and dispersal limitation might enhance the spatial turnover of PFT3 among islands. That greater habitat dissimilarity was found to cause higher β-diversity among smaller islands in TIL (Liu et al., 2018b) also supports this point. Similar to the performance of the whole communities, negative β_RC values of suggests that spatial turnover of PFT1 and PFT2 was determined by the shared environmental filtering (Chase et al., 2011) across islands. Thus PFTs differ significantly in which drivers of β-diversity control their response to habitat fragmentation in this landscape.

Our finding that β-diversity for plant communities in the TIL system is α-dependent could be viewed as support for previous work that has shown communities in the TIL to be nested (Hu et al., 2011; Wang et al., 2010). However, similar decompositions of β-diversity in the TIL system for bird, lizard, and spider communities have shown that species-turnover is the dominant component of β-diversity for these taxa (Si, Baselga & Ding, 2015; Wu et al., 2017). Future work should aim to understand why the variation in community structure varies so widely across taxonomic groups in this system.