Land‐use changes conservation network of an endangered primate (Rhinopithecus bieti) in the past 30 years in China

This study aims to propose a conservation network that contains suitable habitat and connectivity corridors for mitigation due to habitat transformation and fragmentation of Yunnan snub‐nosed monkeys (Rhinopithecus bieti). Further, we also aim to understand the effects of land‐use changes on the conservation network of R. bieti from 1990 to 2020.


| INTRODUC TI ON
Land-use change is a major threat to terrestrial ecosystems, including substantial declines in suitable habitats of wildlife (Díaz et al., 2019;Jung et al., 2020;Powers & Jetz, 2019). Due to the impact of landuse change alone, by 2070, approximately 1700 species of terrestrial vertebrates globally are predicted to be endangered, including species of high conservation value and functional importance (Penjor et al., 2021;Powers & Jetz, 2019). Recent projections of worldwide forest degradation or deforestation suggest a potentially substantial increase in the extinction risk of forest-associated vertebrate species (Betts et al., 2017). For millennia (7000-9000 years ago), China experienced significant changes in patterns of land use and food production associated with animal and plant domestication, sedentarization and a major expansion of its human population . In moving from a largely agrarian economy to a highly industrial economy in recent decades, great changes have taken place in land-use patterns in China (Liu, 2010). China has transformed its natural environment in ways that have negatively impacted biodiversity and species survival Ren et al., 2015). Today, due to its large human population (1.42 billion, the seventh China National Census) (National Bureau of Statistics, 2021), long history of agricultural production, deforestation, and recent expansion of cities, China is facing significant environmental challenges in attempting to balance social development and biodiversity conservation . Dozens of government forestry bureaus were established for purposes of logging and the expansion of the timber industry from the 1950s to 1980s that led to the loss of 1.5 × 10 4 km 2 of forest in addition to losses incurred by the activities of private companies and illegal loggers Sun et al., 2004). China's timber resources are distributed principally in three major regions: northeast, southwest and south China, which coincide with zones of high primate diversity (Liu, 2014). Deforestation and degradation lead to habitat transformation and fragmentation of wildlife. Habitat transformation and fragmentation are primary causes for ongoing biodiversity loss, such as negatively affecting species' distributions, population structures, genetic diversities and likelihoods of survival . Habitat fragmentation transforms large tracts of continuous habitat into smaller and spatially distinct patches immersed within a dissimilar matrix (Wilson et al., 2016;. It also may hinder gene exchange among groups, generating isolated populations that must increase feeding effort to consume more energy to support greater movements (Li et al., 2020).
Identifying species and locations most at risk from habitat loss and fragmentation is key for prioritizing in management of biodiversity (Powers & Jetz, 2019). Three Parallel Rivers Region (TPRR) is a biodiversity hotspot in south Asia and a critical part of Asia's water source and glacier ice repository of the 'Third Pole' in the Himalayas (Penjor et al., 2021). The extreme altitudinal variation and topographical complexity influence weather patterns and create a unique microclimate that harbours a wide assemblage of wildlife such as the endangered Yunnan snub-nosed monkey (Rhinopithecus bieti) . Biodiversity in the TRRR is under siege from human-induced changes to the landscape Mahmoud et al., 2020). Habitat transformation and fragmentation, important forms of habitat degradation, are a consequence of forest degradation or deforestation. R. bieti is an endangered flagship and umbrella colobine species of the TPRR and leaf-eating primate endemic to China (Li et al., , 2020. The range of R. bieti includes coniferous forests (such as Pinus yunnanensis) and alpine mixed broad-leaved coniferous forests (Grueter et al., 2013;Li et al., 2020). There are approximately 3500 individuals and 24 wild groups of R. bieti that inhabit the highest elevation (ranging up to 4500 m) of any nonhuman primate; the species' population appears to have increased relative to IUCN estimates (Li et al., 2020;Long et al., 1994;Su et al., 2019). The primary factor contributing to the increase in population size of R. bieti is a decrease in anthropogenic disturbances including hunting and deforestation. For example, beginning in the late 1990s, China implemented large reforestation initiatives called 'The Natural Forest Protection Program (NFPP)' and 'National Nature Reserve Program' Zhao et al., 2018). Despite the large amount of human and financial capital invested by China to increase forest cover, unless policies are specifically designed to expand the natural ecosystems required by R. bieti, the effectiveness of these policies for conservation is limited . Most of these programmes are not aimed to regenerate native habitats that are crucial for R. bieti survival (Viña et al., 2016). Suitable habitat of R. bieti may decrease by 8.0%-22.4% by the year 2050 due to increased climate change and anthropogenic disturbances, such as forest products collection and the conversion of natural forest to pasture land for livestock grazing (Huang et al., 2017;Li et al., 2020;. Designing a conservation network will help to maintain population and wild groups of R. bieti. A conservation network contains suitable habitat and connectivity corridors among different habitats for mitigating habitat transformation and fragmentation of R. bieti. Establishing a conservation network is important for making robust conservation management decisions. The species distribution model, also commonly referred to as the ecological niche model, is a main tool used to derive spatially explicit predictions of environmental suitability for species (Guisan et al., 2013). Habitat suitability mapping is the first step in building a conservation network and often necessitates the integration of limited data from multiple sources, especially for R. bieti in remote mountainous regions of limited accessibility (Zhang et al., 2020).
Information on occurrence sites of wildlife can be provided by predictions of species occurrences derived from environmental suitability models that combine biological records with spatial environmental data (Guisan et al., 2013). The second step to build a conservation network is the construction of connectivity corridors between the largest and the closest habitat fragments, which can provide one of the highest returns on investment for primate species (Newmark et al., 2017). In the short-term, human-made bridges or other constructed zones of safe passage and strips of non-native fast-growing tree species can be made as temporary connectivity corridors to achieve rapid recovery of R. bieti populations and wild groups . However, in the long-term, these corridors should be comprised of native forest communities with the goal of expanding suitable habitat and ultimately increasing the size and genetic variability of R. bieti populations and wild groups .
To understand the effects of land use on the conservation network of the endangered primate R. bieti from 1990 to 2020, we documented (1) shifts in habitat suitability, (2) construction of connectivity corridors among potential core habitat (PCH patches) and protected areas (PAs), and (3) proposes new conservation network and changes assessment within it.

| Study area
The study area, which is rich in biodiversity and mineral resources,

| Data collection
We acquired eight variables including elevation, slope, aspect, distance to water, settlements and roads, vegetation type and land use and cover change (LUCC) for calculating habitat suitability of R. bieti.
To survey forest ecosystem changes and to verify the accuracy of the LUCC surface as of 2020, we conducted a field investigation to collect 5342 GPS points with vegetation types along the route shown in the red line in Figure 1. LUCC data in 1990LUCC data in , 2000LUCC data in and 2010 were provided by the 'Environmental & Ecological Science Data Center for West China' (http://westdc.westg is.ac.cn). LUCC data in 2020 were downloaded from GLOBELLAND 30 (total accuracy was 85.72% and the Kappa coefficient was 0.82 tested by 5342 GPS points), which were provided by the Ministry of Natural Resources of the People's Republic of China (http://www.globa lland cover.com).
All LUCC data were raster with a resolution of 30 m × 30 m. We derived lakes from LUCC data in 1990, 2000, 2010 and 2020. We downloaded a Digital Elevation Model (DEM) data from the United States Geological Survey (USGS) (http://earth explo rer.usgs.gov/) at 30-m spatial resolution and derived the slope and aspect from those data. We obtained distribution data of settlements (locations of cities, counties and villages) and roads (national highways, provincial highways, county roads and village roads) in 1990, 2000, 2010 and 2020 and river data (vector) from 'National Tibetan Plateau Data Center' (http://westdc.westg is.ac.cn). Combining field survey results with occurrences from the literature, we identified 24 locations where R. bieti were known to occur to verify the accuracy of the GISbased niche model (Su et al., 2019;).

| Habitat variables and GIS-based niche model
The GIS-based niche model incorporated eight environmental and biological variables that were deemed as potentially the most important factors based on ecological requirements and natural history of R. bieti (Huang et al., 2017;Su et al., 2015Su et al., , 2019Venne & Currie, 2021;Zhu et al., 2016).
We tested for multicollinearity to avoid including highly correlated environmental variables. None of the variables were highly correlated (r ≥ 0.8), and so, all were included in the models (Behdarvand et al., 2014;Cerqueira et al., 2021). For vegetation preferences, researchers have identified the main plant species in the diets of R. bieti, through analysing the mean relative density of plant fragments in faeces Li et al., 2009Li et al., , 2011Li et al., , 2013Su et al., 2019). The mean per cent of fragments was a proportion of different vegetation types indicating the main food of the R. bieti (Li et al., 2009(Li et al., , 2011Su et al., 2019). Wild groups of R. bieti distributed in regions with elevation ranging from 2500 to 4700 m (Huang et al., 2017;Su et al., 2015Su et al., , 2019Zhu et al., 2016). There were no distributions of R.bieti wild groups with elevations above 4700 m. Because most of R. bieti wild groups distributed elevation ranged from 2800 to 3600 m, we used segmented assignment to identify suitability threshold values of elevation ranging from 2800 to 3600 m were one hundred (100). We assigned 75 with elevation ranging from 2701 to 2800 m and 3601 to 4300 m. From this, we normalized and assigned values from 0 to 100 to the eight variables with greater numbers indicating the preference by R. bieti, due to different values of eight variables. (Table 1) (Su et al., 2019). For example, if the value of slope was zero (0), it meant that R. bieti did not select this area as habitat, the same as vegetation, LUCC and so on. Firstly, we multiplied these eight normalized variables together to divide habitat suitability in the GIS-based niche model for yielding PCH patches for R. bieti. Second, the resulting maps from the GIS-based niche model were then reclassified with natural break into 4 kinds of potential habitats for R. bieti, in which 0 (zero) stood for unsuitable areas, 1-25 for lowly suitable habitats, 26-50 for moderately suitable habitats and 51-100 for highly suitable habitats (51-75 selected as PCH patches and 76-100 selected as PPHs) (Su et al., 2019). Finally, we used field survey data and references data to verify which suitability threshold values were more appropriate for R. bieti. Based on verification results, suitability threshold values (25, 50, 75 and 100) determined for dividing suitability levels can be used in this research. We defined all PAs as PPHs for R. bieti. Due to minimum requirements for habitat area of R. bieti, PCH patches with more than 30 km 2 were chosen to build connectivity corridors in the circuit model (Xiang et al., 2007(Xiang et al., , 2013.

| Analyses in the circuit model
Circuit models can be used as simple movement models to inform areas of ecology that deal with networks when data are limited or lacking (McRae et al., 2008 (Su et al., 2019). It is typically determined by cell characteristics, such as topography (e.g., elevation, aspect and slope) or human disturbances (e.g., settlement/urban area and road), combined with species-specific landscape resistance models (Roever et al., 2013). Based on existing researches, we derived the elevation, slope and aspect preferred by R. bieti to draw the resistance map (Huang et al., 2017;Xia et al., 2016). We used buffer analysis to obtain the spatial distribution of the effective range from settlement/urban area and roads and selected 1, 3.5 and 5 km from human activities as buffer distances in 1990, 2000, 2010 and 2020 (Xiang et al., 2013). The cell value of the resistance surfaces ranged from zero (minimum resistance value) to 100 (maximum resistance value).

| Design the NIVI index
We designed a normalized importance value index ( where IVI is the importance value index of the PCH patches or PAs in different years; A is the area of PCH patches or PAs; T L is the total length of all connectivity corridors that connect one

| Habitat suitability changes from 1990 to 2020
There were 13 wild groups of R. bieti located in highly suitable to 2020 (Table 2).

| Spatio-temporal distributions of PCH patches and PPHs
Areas of PCH patches and PPHs decreased 615.1 and 221.0 km 2 from 1990 to 2020 (Table 2). There were 20 PCH patches with a total area of 1663.2 km 2 in 1990, 14 PCH patches with a total area of 988.2 km 2 in 2000, 12 PCH patches with a total area of 959.6 km 2 in 2010 and 13 PCH patches with a total area of 1048.2 km 2 in 2020.
There were 2 PPHs with a total area of 424.2 km 2 in 1990, 3 PPHs with a total area of 328.2 km 2 in 2000, 7 PPHs with a total area of 541.1 km 2 in 2010 and 2 PPHs with a total area of 203.2 km 2 in 2020.
Most of PPHs were around Baimaxueshan NNR and Laojunshan scenic spots from 1990 to 2020 (Figure 3a-d). None of PCH patches and PPHs were located inside or around Yunlingtianchi NNR from 1990 to 2020.

| Spatio-temporal changes of conservation network
There

| Habitat transformation and fragmentation
Habitat transformation and fragmentation are widely regarded as among the greatest near-term threats to biodiversity survival, which causes reductions in population sizes of wild species and leads to greater likelihoods of extinctions and the collapse of ecosystems Haddad et al., 2017). Meanwhile, habitat transformation can affect populations of wildlife through well-known effects on connectivity (i.e., the degree to which landscapes alter movement among habitats) and habitat edge to cause extinction (Fletcher, Didham, et al., 2018). To address the extinction of wildlife species, biodiversity conservation often TA B L E 2 Areas of habitat, PCH patches and PPHs, and length of connectivity corridors among PCH patches, PPHs and protected areas Note: PCH patches-potential core habitats; PPHs-priority protection habitats. focuses on protecting species' populations, wild groups and habitats through a variety of legal and policy mechanisms such as the establishment of PAs (Evans & Malcom, 2021;UNEP-WCMC & IUCN, 2017). Understanding the roles of habitat transformation is highly relevant for decision-making regarding biodiversity conservation. Populations and wild groups of R. bieti are strictly protected by PAs and local governments that have overseen a stable rise to 3500 individuals and 24 groups during the past 30 years (Li et al., 2020). In the case study, land-use change has negative environmental impacts on the species' habitats that have become degraded and fragmented due to land-use changes, which result in an expansion of unsuitable areas and lowly suitable habitats and shrinkage of highly and moderately suitable habitats. As the main land-use change in the study area, forest degradation and deforestation are leading to a reduction in food availability, increasing habitat fragmentation and subpopulation isolation of R. bieti . It is the main anthropogenic challenge that popula- We suggest that all connectivity corridors should be protected by PAs and re-built through restoration of native forest and construction of artificial forests (such as Pinus yunnanensis), which can increase movements of R. bieti between habitats and across PAs.

F I G U R E 3 PCH patches and
Simultaneously, the restoration of native forest and construction of artificial forest communities also benefit the livelihoods of people in the local human communities, who collect and use a variety of forest products (Dosen et al., 2017;Li et al., 2018;Newmark et al., 2017).

| Strong protective effect of protected areas
Human activities have both positive and negative effects on biodiversity conservation. As a negative human activity, illegal hunting was one of the greatest threats to the survival of R. bieti, due to local villagers hunting it for meat and fur and as traditional medicine from the 1960s to 1990s .
Due to the enactment by the Chinese government in 1988 of the Wildlife Protection Act that listed R. bieti within the first class of key protected primates in China, it is protected strictly and illegal hunting has a limited effect on survival (Zhao et al., 2018;. Land-use change is impacting biodiversity across the earth, which directly causes habitat transformation and fragmentation of wildlife (Fletcher, Didham, et al., 2018). As a positive human activity, PA policies can mitigate this negative impact to a certain extent at a local scale. The establishment of PAs is essential to local biodiversity conservation (Estrada & Real, 2018). PAs are designed to safeguard and protect all environmental components that maintain biodiversity and ecosystem services, including populations, habitats of wild species and natural ecosystems (Frederico et al., 2018;Margules & Pressey, 2000). The establishment and improvement of PAs have protected subpopulations of R. bieti residing inside these PAs, and in some areas, this has resulted in an increase in the size of the remaining local populations of R. bieti (Su et al., 2019;. However, people and governments only focus on protecting the population of R. bieti and wild groups but ignore the strict protection of its habitats and connectivity corridors that are affected directly by land-use change, such as large areas of primary forest have been converted into pastures for cattle grazing (Xiao et al., 2003). In general, the wildlife inhabiting PAs enjoy higher protection and lower disturbance by human activities (land-use change) than populations living outside the areas (Estrada & Real, 2018). The evaluation of protective effect of PAs on their capacity to preserve species distributions and their habitats is a key topic in conservation biology (Estrada & Real, 2018).
In this research, the length and number of connectivity corridors among 5 PAs have changed little during the study period. Land-use change has a limited effect on the PA conservation network, which contains all PAs and connectivity corridors from 1990 to 2020. PAs have strong protective effect on R. bieti conservation due to limiting effects of land-use change.

| Framework adjustment and optimization of PAs
With an aim to optimize PA frameworks and enhance protective Therefore, this NP conservation framework should give full consideration to local human well-being and biodiversity conservation, which aim to protect biodiversity and maintain sustainable development for society. Another key issue of NP management is the effective implementation of laws and regulations that protect populations and habitats of wildlife from degradation caused by land-use change (Evans & Malcom, 2021). A significant limitation in biodiversity conservation has been the ineffective implementation of laws and regulations. It needs flexible, efficient and effective monitoring and enforcement methods to help conservation policies (laws and regulations) realize their full benefit (Evans & Malcom, 2021).

| CON CLUS IONS
In conclusion, using the GIS-based niche model and circuit model to re-build conservation network of R. bieti, we clarify that land-use change lead to habitat transformation and fragmentation of R. bieti   it has practical significance to re-adjust the system of PAs that conservation network of R. bieti should be as fundamental management unit for PA development.
Here, we propose that expectations to find parallel patterns between endangered species protection and natural resources utility, especially in the TPRR which is a biodiversity hotspot. It may be an alternative way to design a NP which should completely protect all populations and wild groups of R. bieti, its habitats, all PCH patches, PPHs and connectivity corridors. For full consideration of local human well-being and biodiversity conservation, the management framework of this NP should be a coupled ecological-social system linking human-ecosystem-wildlife in local.

ACK N OWLED G EM ENTS
We thank all people for helping us to complete the research. R.B.
Boone provided editorial comment. This study was supported by No. XDA20020402.

CO N FLI C T S O F I NTE R E S T
The authors declare there is no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in Dryad at https://doi.org/10.5061/dryad.fttdz 08tr.