Detecting Nestedness in City Parks for Urban Biodiversity Conservation

Urbanization has been a global phenomenon producing great environmental and ecological challenges including species composition shifts in urban areas. To evaluate the interaction networks of the inhabitant metacommunities in urban parks, we applied the nestedness analysis using ecological survey data of birds, reptiles, frogs, and butterflies among 16 parks in Taipei City of Taiwan and identified critical environmental factors for biodiversity conservation. Results found significant nestedness of the metacommunities among these parks, which indicated strong interaction networks and the importance of urban parks for sustaining these metacommunities. In addition, various types of land use had different effects on the species incidence of birds, reptiles, frogs, and butterflies. It demonstrates that landscape design should consider the taxonomic differences in the development of conservation guidelines. Results also showed that the area of parks and the area of woodlot in the park were critical elements for the nested structure. Nonetheless, abundant bird generalists were found to be highly related to the area of buildings within parks, distance to forests, rivers, and the neighbouring parks. Moreover, the exotic species were found to affect the interaction networks of native species. However, based on the current information, we cannot confirm the actual effects of the exotic species on the native biotasassemblages. Reptiles and frogs appeared to be more easily threatened by land development and habitat fragmentation due to their lower vagility. In this regard, we suggest that conservation action should focus on preserving green space with large areas and constructing green corridors, creating crucial elements, and increasing the vertical complexity in the parks for species with various vagilities. We believe these strategies can help reduce the impact of urbanization to retain healthy interactions and connections of metacommunities among habitats for effective biodiversity conservation in urban areas.


Introduction
The rapid economic developments have motivated and attracted nearly 4 billion people to live in cities worldwide, and it is predicted that nearly 70% of the world population will reside in urban areas by 2050 (United Nations 2018). As urbanization has known to result in a series of economic, social, and environmental changes (Berry 2008), this trend is still undergoing and has led to various ecological impacts. In particular, serious habitat deterioration and fragmentation occur, along with its edge effects that further reduce landscape connectivity (Danielle et al. 2011) and change the inhabitant species composition and distribution (Habel et al. 2019), affecting species abundance, compositional variation, and biodiversity (Alexandre et al. 2008; Danielle et al. 2011). Although some species may have adapted to the niches provided by these urban environments, species with habitat speci city or sensitive to environmental changes, suffer serious threats of survival (Dennis and Eales 1997) and the integrity of metacommunity network. Under the constraints of limited land with increasing human population, land developments are spread out to suburban areas around many cities worldwide. Nowadays, there is general awareness of the importance of green space for species survival and biodiversity conservation (Lepczyk et al. 2017), so that the idea of "wildlife-inclusive urban design" has been recently getting promoted and incorporated into urban planning that give much value of the green space like parks and yards within the urban areas (Apfelbeck et al. 2020).
To explore the importance of green space towards the state of ecological communities, the concept of nestedness proposed by Darlington (1957)  biogeography, the richness of species is distributed in association with the migration and extinction probability determined by the size and the isolation level of the correspondent habitat conditions (MacArthur and Wilson 2001). This mechanism drives the biogeographical pattern, which can be depicted by the nestedness, showing that the species presented in species-poorer islands are often a subset of the species composition in species-richer islands (Atmar and Patterson 1993). As a result, the nestedness is an effective way to quantify and provide information of ecological interactions among the community networks.
Quanti cation of nestedness has been developed from "Temperature index" (Atmar and Patterson 1993) to "Nestedness metric based on Overlap and Decreasing Fill" (NODF) that applied advanced programming and techniques to detect the nested structure and standardize the bias resulted from the size differences of the composition matrix (Ulrich and Gotelli 2007). However, the NODF requires information of species presence and absence (Almeida-Neto et al. 2008), which often causes problem with the existence of rare species. To reduce the in uence from rare species, "Weighted Nestedness metric based on Overlap and Decreasing Fill" (WNODF) improved the quanti cation process by using the weighted method within the abundance matrix (Almeida-Neto and Ulrich 2011) and further considered the inequality and overlap of matrix as a combination of marginal sums of gradients to form the "Weighted nestedness based on overlap and decreasing abundance" (WNODA) (Pinheiro et al. 2019). Under the assumption of the nestedness of WNODA, the internal species composition information can be better predicted within a higher gradient of the marginal sums of composition matrix, con rming stronger habitat-species interactions rather than randomness. Compared with the nestedness index calculated from the presence-absence matrices of binary system, WNODA resolves the rare species in uence by more conservatively presenting the nestedness using the abundance matrix to mimic the real interactions among the ecological systems of metacommunities and their habitats (Pinheiro et al. 2019).
Suitable for detecting the metacommunity interactions in isolated habitats (Atmar and Patterson 1993), the nestedness may be applied to explore the ecological relationship and status of green space in urban areas as they are often situated apart as ecological islands, although for some generalized species, they are likely to quickly occupy new niches even in the highly urbanized areas (Čeplová et al. 2017;Apfelbeck et al. 2020). To explore the importance of urban green space in the biodiversity conservation, and to investigate how different environmental factors affect the species composition and connectivity between metacommunities in urban areas, in this study, we apply the WNODA to examine the nestedness of metacommunities interaction networks among the selected 16 parks in the highly urbanized Taipei city of Taiwan. Followed by the nestedness analysis, we examine the relationship between gradients of environmental factors and species incidences from nestedness by the Spearman's correlation test (Schouten et al. 2007;Ulrich 2009). At the end, we list trade-offs between land development and biodiversity conservation, and provide considerations and suggestions for park design and urban planning.

Study area & data description
In this study, we assembled species observation data in the 16 parks in the Taipei city of Taiwan ( Fig. 1 The species composition data were organized into two versions. One included the exotic and native species, and the other included only the native species. We arranged the species-habitats matrix with vertical axis representing the species composition of each habitat, and horizontal axis the species incidences in all the habitats (Atmar and Patterson 1993; Almeida-Neto and Ulrich 2011) using the software of R 4.0.3. After that, we adjusted the abundance of each species by the frequency of surveys to produce the weighted-abundance.

The nestedness analysis
To evaluate the role of parks and their surrounding con guration for sustaining the biodiversity, we applied WNODA that considers the inequality and overlap of matrix as a combination of gradient presented by marginal sums to examine the nestedness (Pinheiro et al. 2019). The nestedness, therefore, represents the interaction networks of the inhabitant metacommunities of bird reptile, frog, and butter y among the 16 parks in Taipei and the associated gradient change (Pinheiro et al. 2019). Before calculating WNODA, we lled in the weighted-abundance into each cell to construct the species-habitats matrix. Based on the assumption of the hierarchical relationship shown in the nestedness, in the rows and columns, the matrix was organized in an order by the magnitude of the marginal sums of the weighted-abundance. As such, parks with higher richness were ordered at the upper rows, and species with higher incidences were ordered to the columns at left. After the arrangement, the more general species will be organized into the upper left corner in the matrix

Identi cation of critical environmental factors
As the nestedness is driven by the interaction among heterogeneous environmental conditions, it is associated with different probabilities of selective immigration and extinction of species (Atmar and Patterson 1993; Vázquez et al. 2007), which leads to species compositional variation and unequal abundance. Based on the rationale of island biogeography and the calculated nestedness structure, we allocated potential environmental factors that may affect the survival of species and the interaction networks of the metacommunities (Rico-Silva et al. 2021), including area of different land use types or landscape components within the park, habitat heterogeneity, identi cation of green space within each park, area or boundary ratio of parks, road or building density around parks, and distance to usable habitats. The landsat-8 satellite image from the USGS Earth Explorer was used to digitalizing the chosen environmental factors. Then we created a correlation matrix, and excluded some of the factors having a correlation coe cient higher than 0.7 and at a signi cance level of P < 0.05 (Dormann et al. 2013). We examined the in uence of the surrounding environmental con guration on the interaction networks of the metacommunities using a Spearman's correlation test that taking account the relationship between ecological gradients and species incidences from nestedness using the following equation (Ulrich 2009): D. Area or boundary ratio of parks.
To represent the possible edge effects, boundary ratio was calculated as the perimeter of the park divided by the area.
E. Road or building density around parks.
Building or road density around parks were determined separately as the proportion of area occupied by buildings or roads within a buffer of 100 m around each park (Rico-Silva et al. 2021).
F. Distance to usable habitats.
To examine the ecological connection between usable habitats, like green spaces and aquatic environments, we calculated the shortest distance from each park to the nearest habitats, including forests, rivers, and the neighbouring parks. In this study, forests were de ned as an area occupied by trees larger than 5 hectares. The composition data of the metacommunity of frog were found to include 2 exotic and 13 native species. The exotic frog species were American Bullfrog (Lithobates catesbeianus) and Brown Tree Frog (Polypedates megacephalus); while the native species were mainly composed of Asian Common Toads (Duttaphrynus melanostictus). In terms of the metacommunity of butter y, all of the 71 recorded species were native species, in which Pieris canidia was accounted for the majority.

The nestedness structure among metacommunities and parks
Excluding the exotic species in the composition data, results showed that all the metacommunities of native species had signi cant nestedness among these 16 parks in Taipei (Table 1 and Fig. 2). Reptiles appeared the strongest nestedness (WNODA = 60.93, P < 0.001), followed by frogs (WNODA = 51.12, P = 0.017) and birds (WNODA = 48.17, P < 0.001), and butter ies the weakest (WNODA = 28.91, P < 0.001). However, when including the exotic species into the analysis, the nestedness of all the biological groups slightly declined (Table 1 and Fig. 2).

Identi cation of critical Environmental by Spearman's correlation test
According to the correlation matrix, we excluded some of the highly correlated environmental factors, including area of grass, roads, and buildings, boundary ratio of parks, and road density around parks (Fig.  3). Analysis results from the Spearman's correlation test found that in general, the land use types within parks showed different level of in uence, and the habitat heterogeneity showed lower and non-signi cant correlations on the nestedness of the metacommunities ( Table 2). In particular, we found signi cant positive correlation between certain land use types with the weighted abundance of native species of birds, reptiles, and butter ies. For example, the correlation coe cients of the area of trees to native birds, reptiles, and butter ies were 0.52, 0.78, and 0.70, respectively. Yet, the correlation coe cient of frogs to the area of trees was at a lower value (0.27), and not signi cant. In addition, we also found shrubs a critical land use type for birds and butter ies, with a higher positive correlation coe cient of 0.67 for native birds at a signi cance level of P < 0.01, and for butter ies with a correlation coe cient of 0.58 at P < 0.05. Furthermore, results found signi cant correlations of birds and butter ies to the area of parks with a correlation coe cient of 0.54 and 0.47, respectively. Besides, the metacommunity of bird was found to be mostly affected by distance to rivers with a positive correlation coe cient of 0.77 at a signi cance level of P < 0.01. On the contrast, the distance to the neighbouring parks appeared a negative correlation of -0.56 at a signi cance level of P < 0.05 to the weighted abundance of birds (Table 2).
More importantly, results found random effects of exotic species on the interaction networks of native species. For example, when including the exotic species into the analysis, the signi cance level of the correlation of birds to "area of trees" turned from signi cant to non-signi cant. An interesting result was found that with the exotic species, the correlation of building density around parks on birds decreased from -0.41 to -0.34, and the signi cance level dropped from P < 0.1 to insigni cance. Similarly, the area of parks was found to be a signi cant factor whether we included the exotic reptile species. However, "average NDVI within green space" was the factor that changed the correlation from 0.63 (P < 0.01) to 0.41 (P < 0.1) when the exotic species were included ( Table 2).

Discussion
To examine the interaction networks of metacommunities of bird, reptile, frog, and butter y among the parks in the urban areas, we applied the nestedness analysis of WNODA to measure and observe the order/disorder of the community ecological patterns in habitats facing fragmentation or isolation conditions (Atmar and Patterson 1993). Results showed signi cant nestedness of all the metacommunities among these parks, implying the importance of parks in the urban ecosystems as "island-like" habitats governing the assemblage networks and ecological community interactions of bird, reptile, frog, and butter y. Given the selective colonization and extinction pressure caused by the respective environmental properties of each habitat, the assemblage networks are prone to develop nested structure with unequal biogeographical patterns of the species distribution and abundance (Atmar and Patterson 1993;MacArthur and Wilson 2001). As a result, when the system of these ecological islands become stable, the ubiquity of nestedness shows (Thebault and Fontaine 2010).
Through the analysis of WNODA, the size of the species-habitat matrix representing the nested structure can be standardized allowing unbiased comparison of different communities (Pinheiro et al. 2019). Based on the nestedness calculated from WNODA with native species only, reptiles and frogs showed stronger nested structure than birds and butter ies. We thought that this result may indicate greater in uence of habitat fragmentation on the biological groups with lower mobility being more restricted for their movements (Habel et al. 2019). This necessitates the conservation strategy of green spaces during land development for those being threatened or struggling to survive, such as the lower mobility biological groups, in the highly fragmented urban environments. However, with consideration of both the native plus the exotic species in the WNODA calculation, results appeared a declined nestedness of all the four selected metacommunities showing a more disordered interaction networks of the associated assemblages among these parks (Čeplová et al. 2017). We argue that the calculated WNODA with exotic species included may be a result of randomness, rather than a nested structure in the parks of Taipei. This randomness implies that the niches of the exotic species are more general than those inhabitant native species in terms of the adaptation to the urban environments (Atmar and Patterson 1993; Čeplová et al. 2017). As such, for effective biodiversity conservation in urban areas, it is recommended to focus on restricting the spread out of the exotic species and regulating the transaction or trade. If an exotic species has been found to form proli c population in urban areas, it should be considered to lower the risk to the survival of native species by conducting effective control or removal on the exotic species to prevent their territory development or spread out (Gaertner et al. 2016).
Given the importance of parks in sustaining the interaction networks among different metacommunities from the nestedness analysis, we cooperated the Spearman's correlation test to explore the critical environmental factors affecting the metacommunities to provide science-based suggestions of crucial elements and proper designs for landscape planning and improvement ). Results found that various types of land use had different effects on the species incidence of birds, reptiles, frogs, and butter ies. In particular, areas of trees and shrubs within parks were found to be positively correlated with the species incidence of native species of birds, reptiles, and butter ies. This reveals that the design of multi-layered environment could potentially bene t biodiversity conservation for providing diverse habitats for different species, enhancing the robustness of the ecological interaction networks (Oliveira and Scheffers 2019).
In addition, area of parks exhibited a positive correlation with the species incidence of birds and butter ies signi cantly. According to the correlation matrix, we observed a negative correlation of "area of parks" to the boundary ratio of parks, and road or building density around parks (Fig 3). It demonstrates that a larger park with lower boundary and isolation may help reduce the edge effects and maintain diverse natural habitats, which is bene cial to the richness and abundance of birds and butter ies. It also strengthens the knowledge ground for preservation of large green spaces to produce various niches to accommodate different species for biodiversity conservation. Surprisingly, habitat heterogeneity, as a transformed index, did not show signi cant correlations to birds, reptiles, frogs, and butter ies. It might be due to the scale of the parks in Taipei being relatively small, so that the associated correlation coe cients of the habitat heterogeneity were relatively minor in comparison with other environmental factors. This implies the more direct impacts of the habitat features on shaping the distribution and survival of species (Ulrich 2009), yet the environmental factors interacted with the habitat heterogeneity can together determine the nestedness of communities (Schouten et al. 2007;Rico-Silva et al. 2021). This warns the irreplaceable role of the measurements or observations to be more directly linked to the reality or the in-situ conditions, as well as the use of the transformed index to be taken as a supplementary and/or a more comprehensive addition.
Moreover, our results found that birds were negatively correlated with distance to the neighbouring parks, and positively correlated to distance to the nearest forests and rivers, indicating the importance of the connection between the parks and natural areas as habitats to the bird metacommunities. Studies found that loss of connection could potentially impact the networks and resilience of metacommunities, increasing the probability of extinction (Alexandre et al. 2008). Interestingly, the incidence of birds, either native or exotic, were found to remain a positive correlation with distance to the nearest forests and rivers signi cantly. This situation may be caused by the habits of the generalist species accounted for the majority of total weighted abundance, such as Eurasian Tree Sparrow (Passer montanus) accounted for 33.7% to prejudice the arrangements of the abundance. Nonetheless, since the ecological requirements for the generalists of birds, often the omnivorous, granivorous, or cavity nesting birds, are more easily ful lled in urban areas (Simberloff and Cox 1987), it may represent that indirect impact of urbanization on the richness of native birds with narrow niches (Chace and Walsh 2006). As a result, improvement on the connection between habitats, such like the urgent advocate of green corridor developments (Simberloff and Cox 1987), increasing the natural designs, planting native nectars and fruits between parks and green space (Akif et al. 2020), and preservation of existing green space, can help stabilize the interaction networks and increase the re-colonization rates of native species. These strategies can uphold the decreasing biodiversity (Alexandre et al. 2008), in the highly urbanized environment, especially for native birds.
When including the exotic species, we observed that the signi cance level of some correlations shifted. In terms of land use types within each park, the factor of "area of trees" turned into non-signi cant, while "area of shrubs" remains its positive and signi cant correlation. Besides, with the inclusion of the exotic bird species, "density of buildings around the parks" became less negative and not signi cantly related, re ecting the species-speci c adaptability of the exotic birds in the urban areas and their utilization of the roads and buildings. If the thrived population of exotic bird species outcompete the native bird species, it may cause the loss of biodiversity (Chace and Walsh 2006). Similarly, the species incidence of native reptiles was found to be signi cantly correlated to "area of trees" and "average NDVI within each park", expressing the need of native reptiles, such as Swinhoe's Japalure and Hekou's Gecko (Gekko hokouensis), for higher quality of green vegetation and large arbors. However, once the exotic reptile species were included, the correlation coe cient between species incidence and "area of parks" became higher and the correlation to "average NDVI within each park" decreased. These changes may be caused by the dominant exotic species of red-eared sliders, an aquatic species that have often been abandoned in large parks, and have been found to establish their population there (Zhang et al. 2020). In this regard, these results reinforce the need to control the introduction of the exotic species for native species protection. Also, it is vital to preserve and plant native vegetation to conserve the connection or provide niche for native species as another strategy for biodiversity conservation.
In conclusion, metacommunities of bird, reptile, frog, and butter y showed signi cant nestedness among the 16 parks in Taipei, indicating strong interaction networks of the metacommunities among these habitats (Atmar and Patterson 1993). As a result, the existence of parks plays a vital role for sustaining these metacommunities (Almeida and Ulrich 2011). In addition, land development may affect the biological groups with lower mobility more seriously, such that in our analysis reptiles and frogs were found to be more easily threatened by habitat fragmentation (Habel et al. 2019). To reduce the impact of urbanization, and help retain healthy interactions and connections of metacommunities among the habitats, we suggest to conserve green space with large areas and construct the green corridors (Alexandre et al. 2008). In the parks, it is recommended to increase the area of shrubs and trees, and create crucial elements such as increase the vertical complexity with large and multi-layered trees for various mobility level species (Oliveira and Scheffers 2019). As for the exotic species, control the introduction is important preventive work for native biodiversity conservation in urban areas. We believe that these strategies are essential to construct wildlife-inclusive cities and conserve the community ecology in urban areas. Figure 1 A schematic map showing the land-use types in the Taipei city of Taiwan and the location of the selected parks with their names and an example of the land-use types within park #16 Figure 2 Nestedness plot of the 16 parks with only the native species of (a) bird, (b) butter y, (c) reptile, and (d) frog Figure 3 Correlation matrix of (a) correlation coe cients and (b) P-value of original environmental factors. Factors to be excluded were highlighted in red boxes for their high association with other factors (correlation coe cient > 0.7) at a signi cance level of P < 0.05