Potential impact of climate change on the distribution of Capricornis milneedwardsii, a vulnerable mammal in China

Abstract Climate change significantly impacted on the survival, development, distribution, and abundance of living organisms. The Chinese serow Capricornis milneedwardsii, known as the “four unlike,” is a Class II nationally protected species in China. In this study, we predicted the geographical suitability of C. milneedwardsii under current and future climatic conditions using MaxEnt. The model simulations resulted in area under the receiver operating characteristic curve (AUC) values above 0.9 for both current and future climate scenarios, indicating the excellent performance, high accuracy, and credibility of the MaxEnt model. The results also showed that annual precipitation (Bio12), slope, elevation, and mean temperature of wettest quarter (Bio8) were the key environmental variables affecting the distribution of C. milneedwardsii, with contributions of 31.2%, 26.4%, 11%, and 10.3%, respectively. The moderately and highly suitable habitats were mainly located in the moist area of China, with a total area of 34.56 × 104 and 16.61 × 104 km2, respectively. Under future climate change scenarios, the areas of suitability of C. milneedwardsii showed an increasing trend. The geometric center of the total suitable habitats of C. milneedwardsii would show the trend of northwest expansion and southeast contraction. These findings could provide a theoretical reference for the protection of C. milneedwardsii in the future.

reliable spatial information regarding current species distributions is therefore essential to effectively model future ranges under climate change scenarios (Qin et al., 2017).To overcome issues with data incompleteness and predict future species range shifts, species distribution models (SDMs) are commonly utilized (Guan et al., 2022).SDMs are statistical tools that quantify the relationship between observed species occurrences and associated environmental conditions (Carvalho et al., 2011).Various modeling approaches exist for species distribution modeling.Some methods, such as the generalized additive model (GAM; Kosicki, 2018) and classification and regression tree (CART; Zhang et al., 2014), use presence-absence data.Others, like the genetic algorithm for rule set prediction (GARP; Padalia et al., 2014) and maximum entropy model (MaxEnt; Yang et al., 2023), leverage presence-only records.Among these methods, MaxEnt has been widely adopted by researchers due to its ability to perform well with small sample sizes (Wisz et al., 2008), its use of presence-only occurrence data and relevant environmental variables (Elith et al., 2011), and its direct output of habitat suitability maps (Purohit & Rawat, 2022).
The Chinese serow Capricornis milneedwardsii is also known as "four unlike" because they look like deer but are not deer, have hooves like cattle but are not cows, have heads like sheep but are not sheep, and have tails like donkeys but are not donkeys (Figure 1).This species is native to China and Southeast Asia and has been listed as a Class II nationally protected species in China (Gong et al., 2016).Although it is widespread throughout its distributions, C. milneedwardsii is listed as a vulnerable species on the IUCN Red List (IUCN, 2011).This solitary and nocturnal species mainly inhabits coniferous and broad-leaved mixed forests and natural karst scrublands (Liu, Guan et al., 2021).Its elusive lifestyle and rugged habitat make in-depth study challenging (Liu, Lü et al., 2021).However, previous studies indicate that populations have become fragmented and are declining in many areas due to progressive habitat loss and overhunting (Francis, 2008;Thuc et al., 2014).Currently, habitat fragmentation and anthropogenic disturbances seriously threaten the survival of C. milneedwardsii (Liu et al., 2023).Urgent conservation interventions, such as strengthened habitat protections, are therefore needed.
In this study, we leveraged species location data compiled from literature searches and databases along with climate variables from WorldClim to map current and future potential distributions for C. milneedwardsii in China using MaxEnt modeling.We also aimed to determine key environmental factors restricting C. milneedwardsii's range.Specifically, we evaluated changes in the geographical extent of suitable habitat between current and future climate scenarios.

| Collecting occurrence data of C. milneedwardsii
Species distribution data were obtained from the Global Biodiversity Information Facility (GBIF) database, China National Knowledge Infrastructure (CNKI) database, Web of Science, and literature.After integrating the occurrence data from each source, the distribution information was depicted using R software (Chen, Guan et al., 2021).To ensure the accuracy, distribution point data were carefully screened.Data points that were repetitive, inaccurate, or significantly deviated from the species' known distribution range were eliminated following the previous research methods (Liu, Lü et al., 2021;Guan et al., 2022).To avoid overfitting the model, only one point was retained in each environmental factor grid with a spatial resolution of approximately 1 km × 1 km.
ENMTools was used to randomly delete multiple coordinates that fell in the same grid based on the climate variables layer (Zhao et al., 2023).Finally, a total of 73 distribution points were obtained (Figure 2).

| Environmental variables
To determine which environmental variables most influence the distribution of C. milneedwardsii, 19 bioclimatic variables and three topographic variables (elevation, slope and aspect) were selected., 2014).There are four RCP scenarios: RCP2.6,RCP 4.5/6.0 and RCP 8.5, which indicate low-, medium-, and high-concentration greenhouse gas emission scenarios, respectively (Remya et al., 2015).Ultimately, RCP 4.5 and RCP 8.5 were selected to assess the distribution of suitable habitat of C. milneedwardsii in the future.The topographic data (elevation, slope, and aspect) were derived from digital elevation model (DEM) (http:// www.gsclo ud.cn/ ) with a resolution of 1 km.
To reduce the risk of overfitting in the MaxEnt model due to environmental variables, pairwise Pearson's correlations among the variables were tested using ENMTools.Variables with |r| ≥ 0.8 were considered highly correlated, and the variables having the lower contribution value were retained.In the end, 13 environmental variables were selected for MaxEnt (Table 1).

| Species distribution modeling
Based on the selected distribution data and environmental variables, the MaxEnt model was established and run 10 times.In each run, 75% of the data were randomly selected for model training, while the remaining 25% were used for model testing (Guan et al., 2022;Phillips, 2008;Zhao et al., 2023).Jackknife tests were performed to identify variables that reduce model reliability when omitted (Chen et al., 2022;Yang et al., 2023;Zhao et al., 2023).The area under the receiver operator curve (AUC) was used to evaluate model performance.An AUC value less than or equal to 0.7 indicates poor model performance, a value between 0.7 and 0.9 indicates moderate performance, and a value between 0.9 and 1.0 indicates excellent performance (Zhao et al., 2023(Zhao et al., , 2024)).
For visualization and further analysis, the results of the Maxent models predicting the presence of C. milneedwardsii were imported into ArcGIS Pro (Reference ID: 602162530176).According to IPCC's explanation of the probability (p) of species' presence, potential habitats were divided into four categories: unsuitable habitat (p < .1),poorly suitable habitat (.1 ≤ p < .3),moderately suitable habitat (.3 ≤ p < .6), and highly suitable habitat (p ≥ .6).Following the methods of previous studies, the centroid of suitable areas under different climate change scenarios was calculated using the SDMtoolbox 2.4 in ArcGIS.Changes in centroid position under different scenarios were compared, and the migration distance of centroids was calculated (Chen et al., 2022;Guan et al., 2022;Zhao et al., 2023).4).Within these suitable habitat areas, C. milneedwardsii is predominantly found in the moist regions of China.
Under future RCP4.5 and RCP8.5 scenarios, the suitable habitat for C. milneedwardsii is projected to change (Figure 5).The areas of highly, moderately, and poorly suitable habitats are expected to increase significantly (Table 3).By the 2050s, the highly suitable habitat area is projected to increase to 23.46 × 10 4 km 2 under RCP4.5 and 22.91 × 10 4 km 2 under RCP8.5.The moderately suitable habitat area is expected to increase to 40.91 × 10 4 km 2 under RCP4.5 and 44.20 × 10 4 km 2 under RCP8.5.The poorly suitable habitat area is projected to increase to 73.67 × 10 4 km 2 under RCP4.5 and 83.82 × 10 4 km 2 under RCP8.5 (Table 3).The expansion of highly and moderately suitable habitats is mainly concentrated in southeastern Tibet, central and western Yunnan, central Sichuan, southwestern Zhejiang, central and northern Fujian, with a small increase in northwestern Tibet.In contrast, reductions in highly and moderately suitable habitats are primarily observed in eastern Yunnan and south-central Sichuan (Figure 5).By the 2070s, the highly suitable habitat area is projected to increase to 19.41 × 10 4 km 2 under RCP4.5 and 20.99 × 10 4 km 2 under RCP8.5.The moderately suitable habitat area is expected to increase to 37.48 × 10 4 km 2 under RCP4.5 and 41.15 × 10 4 km 2 under RCP8.5.The poorly suitable habitat area is projected to increase to 74.78 × 10 4 km 2 under RCP4.5 and 82.12 × 10 4 km 2 under RCP8.5 (Table 3).
Under current climate conditions, the geometric center of the potential suitable habitat for C. milneedwardsii is located in the southeast of Sichuan Province, China (Figure 6).By the 2050s, under  6).

| Performance of the MaxEnt model
The MaxEnt model has proven to be a robust tool for studying the spatial distribution of species (Elith et al., 2006).In this study, we em- Naemorhedus griseus in the Qinling Mountains (Liu et al., 2022).(Sheng & Lu, 1985), which provides abundant food resources and habitat conditions that affect the growth and development of C.

| Key environmental factors
milneedwardsii (Liu et al., 2023).We found a strong coupling between habitat suitability and annual precipitation (Bio12), slope, elevation, Furthermore, habitat suitability showed a significant correlation with slope, which may be attributed to the species' habit of basking on bare rock and using steep mountains for escape, as well as its extensive food habits (Liu et al., 2023;Wu & Hu, 2001).

| Habitat requirements and potential impacts of climate change
Forest type comprehensively reflects the characteristics of food composition, temperature, light, topography, and landform that animals require, and it meets the needs of animals for habitat selection to the greatest extent (Wu & Hu, 2001).Previous studies have shown that the main habitats of C. milneedwardsii are broad-leaved forests and coniferous broad-leaved mixed forests (Liu et al., 2023;Sun et al., 2007).The breeding season of C. milneedwardsii occurs in late October and November, with a single kid born after a 7-month gestation period.The kid may stay with its mother for almost a year (Thuc et al., 2014), and C. milneedwardsii reaches reproductive maturity after 2-3 years, around late May or early June (Francis, 2008).Moreover, this species tends to inhabit rugged, steep hills, and rocky terrain (Francis, 2008)  Climate is a key factor affecting the distribution of species (Castex et al., 2018).In the future, global warming may cause the subtropical monsoon climate boundary to move northward, potentially providing new suitable habitats for C. milneedwardsii and inducing upward and northward distributional shifts from lower to higher elevations and latitudes.Indeed, our modeling results showed that the highly suitable area for C. milneedwardsii will increase under both RCP 4.5 and RCP 8.5 scenarios compared with the current potential distribution.These results align with previous studies demonstrating how climate change could shift species distributions poleward and to higher elevations, as observed for Aspidoscelis costata costata (Güizado-Rodríguez et al., 2012), Inachis io (Ryrholm, 2003), and smallmouth bass (Chu et al., 2005).

| Conservation implications and recommendations
Our study underscores the significance of China's moist areas as the primary habitats for C. milneedwardsii and highlights the need to prioritize these regions, known for their high levels of wildlife diversity and activity (Liao et al., 2024), in conservation efforts.The MaxEnt model developed here identifies temperature, rainfall, elevation, and slope as the key environmental factors influencing the species' habitat suitability, providing valuable insights for targeted conservation strategies.

Following
the methods described byZhao et al. (2024), the climate data in this study were accessed through the Worldclim database (http:// www.world clim.org/ ) (accessed on 20 June, 2023) at 30 arcsecond (1 km × 1 km).The dataset included 19 bioclimatic variables that represent current temperature and precipitation conditions F I G U R E 1 Capricornis milneedwardsii.
from 1970 to 2000(Fick & Hijmans, 2017).To evaluate the impact of climate change on the species distribution, climate data for future predictions were also downloaded from the WorldClim database at 30 arc-second (1 km × 1 km).The future climate data represented long-term average climatic conditions in the 2050s (average for 2041-2060) and 2070s (average for 2061-2080) from the CCSM4 climatic system model released by CMIP6.The long-term climate change projections are based on new scenarios known as representative concentration pathway (RCPs, IPCC

F
I G U R E 2 Geographical distribution of Capricornis milneedwardsii occurrence points.TA B L E 1 Environmental variables used in MaxEnt model and their percentage contribution.
ployed the MaxEnt model to predict the current and future potential distributions of C. milneedwardsii using bioclimatic and topographic variables.Our model achieved a high AUC value of 0.975, surpassing those reported for small-scale studies of this species by Feng et al. (2022); AUC = 0.878, Liu et al. (2023); AUC = 0.944, and Meng et al. (2023); AUC = 0.852.The variable AUC accuracy across scales highlights the limitations of small-scale niche modeling and validates our large-scale approach.Comparable evaluations for other species also demonstrate a range in MaxEnt prediction success, with reported AUC values of 0.915 for Ailuropoda melanoleuca in the Minshan Mountains (Chen, Zhu et al., 2021), 0.970 for Chrysolophus pictus in central and western China (Ye et al., 2021), and 0.923 for and mean temperature of the wettest quarter (Bio8).The response curves showed that the presence probability of C. milneedwardsii increased when annual precipitation exceeded 720.93 mm and the mean temperature of the wettest quarter ranged from 9.99 to 20.23°C, indicating that the species' presence is strongly depends on rainfall and temperature.Additionally, C. milneedwardsii tends to prefer medium-high-elevation gradients, with a higher probability of occurrence at elevations ranging from 996.85 to 3953.99 m, which is consistent with previous studies byFeng et al. (2022) andLiu et al. (2023).The combination of different latitudes and elevations may form similar local climatic and hydrothermal conditions, thus creating suitable habitat conditions for C. milneedwardsii.
the Minshan, Qionglai, and other Sichuan mountain systems(Feng et al., 2022) identified differing environmental factors as key drivers of C. milneedwardsii distribution.For example, Liu et al. (2023) found that precipitation seasonality (Bio15), vegetation, isothermality (Bio3), elevation, and distance to rivers most influenced local C. milneedwardsii distributions.These discrepancies in results may be explained by several factors.Firstly, our study only considered bioclimatic and topographic data, while Feng et al. (2022) and Liu et al. (2023) integrated F I G U R E 3 Response curve of dominant environmental factors.F I G U R E 4 Distribution map of suitable habitat of Capricornis milneedwardsii under current climate.additional variables like vegetation indices and proximity to water and human activity into their models.Secondly, our study used large-scale occurrence data, whereas Feng et al. (2022) and Liu et al. (2023) based their analyses on localized presence records.To elucidate the most influential drivers of C. milneedwardsii distribution, future modeling efforts should incorporate more comprehensive environmental predictors and evaluate a multi-scale approach.

F
Changes in the distribution pattern of suitable habitat in highly and moderately of Capricornis milneedwardsii under different climatic conditions.
We recommend the following actions: (1) Conservation authorities should focus on monitoring and managing these specific environmental variables to ensure the persistence of suitable habitats for C. milneedwardsii.Long-term monitoring of vegetation status and climate factors in the species' habitats is crucial for understanding the impacts of climate change and other anthropogenic pressures, and for developing adaptive management strategies.(2) A combination of GPS collars, infrared camera monitoring techniques, metagenomics, and other macrocosm and microcosm approaches should be employed to comprehensively study the conservation ecology of C. milneedwardsii.(3) Different types of protected areas, such as national parks, nature reserves, and natural parks, should be established and strengthened as important forms of in situ conservation for C. milneedwardsii.(4) Based on a comprehensive database of species diversity and considering environmental factors such as climate, vegetation, terrain, and human activities, a variety of niche models should be used to construct spatially explicit maps of suitable habitat for the species.(5) Laws and regulations on wildlife protection should be improved and strictly enforced, with efforts made to raise public awareness and resolutely curb illegal hunting and trade in wildlife products.Our research contributes to the growing knowledge on the ecological requirements and potential vulnerabilities of C. milneedwardsii, providing a foundation for evidence-based conservation planning.We urge policymakers to utilize these findings to formulate and implement effective, science-driven strategies for the F I G U R E 6 Suitable location of geometric center for Capricornis milneedwardsii under different climate scenarios.protection of C. milneedwardsii and its critical habitats, ensuring the long-term viability of this iconic species in the face of ongoing environmental challenges.5 | CON CLUS ION This study used the MaxEnt model to predict the current and future potential geographical distribution of C. milneedwardsii, a vulnerable species native to China, under different climate change scenarios.Results suggest that the distribution of C. milneedwardsii is likely to expand in the future, with highly suitable areas increasing under both RCP 4.5 and RCP 8.5 scenarios.Annual precipitation, slope, elevation, and mean temperature of the wettest quarter were identified as the most important environmental variables, contributing 78.9% to the predictions.The habitat suitability maps generated in this study can inform conservation planning and management of C. milneedwardsii by prioritizing areas for monitoring, protection, and habitat restoration efforts.However, the model's limitations include the lack of consideration for other potentially important environmental variables such as interspecific competition, distance to water sources, vegetation types, and anthropogenic factors.Future research should integrate these additional variables to achieve more comprehensive predictions.Despite these limitations, our study provides valuable insights for the conservation and management of C. milneedwardsii.We recommend that conservation authorities and policymakers use this knowledge to develop evidence-based strategies for the protection of this vulnerable species and its critical habitats.As climate change continues to threaten biodiversity, studies like ours will become increasingly important for guiding proactive, science-based conservation strategies.
The MaxEnt model performed exceptionally well in predicting the distribution of C. milneedwardsii under current climate conditions, with an AUC value of 0.975 for the training data.The average training AUC values for the four future climate scenarios were also high, all exceeding 0.95, indicating the model's high accuracy and credibility (Table2).
The geographical distribution of C. milneedwardsii in China under current climate conditions, as predicted by MaxEnt, is shown in Figure 4.The suitable habitat for C. milneedwardsii is primarily located south of the Yellow River in mainland China and in the central region of Taiwan Province.The total suitable habitat area is approximately 123.21 × 10 4 km 2 , representing about 12.83% of China's land area (Table 3).The moderately and highly suitable habitats are mainly distributed in southern Shaanxi, southeastern Gansu, central and western Sichuan, central and western Hubei, central and western Hunan, southeastern Tibet, northwestern Yunnan, northern Guizhou, southwestern Zhejiang, central Taiwan, and Fujian, with total areas of 34.56 × 10 4 and 16.61 × 10 4 km 2 , respectively (Figure Area under the receiver operating characteristic curve (AUC) under current climate and four future scenarios.
Our results indicate that the suitable habitat of C. milneedwardsii is mainly distributed south of the Yellow River in mainland China and in the central area of Taiwan Province.These areas are located in TA B L E 2 the tropical and subtropical regions of Southeast Asia with a humid climate Potential habitat area (×10 4 km 2 ) of Capricornis milneedwardsii under current and future climate change scenarios.