A Survey of the Wild Giant Panda Population and Habitat Reflects an Urgent In Situ Conservation Need: A Case of Meigu Dafengding National Nature Reserve

Abstract

An important tool for conserving species and their habitats and achieving sustainable development is the establishment of national parks. The population ecology of the species to be conserved, including their population size and suitable habitat, needs to be assessed prior to integrating protected areas into national parks. The Liangshan Mountains are currently the southernmost habitat of giant pandas and lie outside the newly established Giant Panda National Park (GPNP). The population is threatened with extinction due to severe isolation and human disturbance. However, there has been a ten-year gap in the census of giant pandas in this critical area. This means that conservation managers are unable to keep up to date with population and habitat dynamics. The Meigu Dafengding National Nature Reserve is the core area of the giant panda population in Liangshan and the link for gene exchange. The focus of this study is to assess the spatial distribution pattern of suitable habitat, habitat fragmentation and the habitat selection characteristics of giant pandas in the Meigu Dafengding Nature Reserve in Liangshan in order to lay the foundation for the future inclusion of Liangshan into the national park. A total of 151 giant panda fecal samples were collected in 2023. Using the distance–bamboo stem fragments method, this indicated that nearly 28 to 29 giant pandas exist within the reserve. Based on MaxEnt and FRAGSTATS, the giant panda population is concentrated in Wahei–Yiziyakou. They tend to choose gentle slopes of 10–30°, at altitudes between 2500 m and 3500 m, with average annual temperatures between 8.5 °C and 10 °C. They also select the areas close to roads and settlements for their frequent activities. Unsuitable habitat is widespread and surrounds low-, medium- and high-suitability habitats with poor habitat connectivity. In situ conservation measures are urgently needed due to the restricted distribution of populations and poor habitat suitability. This lays the foundation for the future inclusion of Liangshan into the national park, providing greater protection for the giant panda and other species in the region, and for other national parks to integrate all habitats into a single management unit. This will address conservation gaps and overlapping management, and promote the conservation of rare or endangered species.

1. Introduction

With the development and progress of science and technology, human activities have led to the further degradation of wildlife habitats, resulting in disturbance, population decline and even extinction [1], loss of biodiversity and increased human–wildlife conflict [2,3]. One of the most important strategies for halting the rapid loss of global biodiversity is the establishment of protected areas [4,5]. Protected areas, as the core carriers of biodiversity conservation and the nature conservation system, with national parks as the main body, lay the foundation for the protection of habitats and wildlife, the improvement of ecological environmental quality and the maintenance of national ecological security [6,7]. National parks for flagship species have been widely established in countries around the world: Rajaji and Corbett National Parks protect Indian elephant populations that have suffered from habitat fragmentation due to hydrological development [8]; and Pakistan’s Deosai National Park incorporates community participation to reduce the poaching of brown bears [9]. As a result of long-term multidisciplinary efforts [10], they are increasingly playing a key role in the conservation of biodiversity [11,12,13].

The giant panda (Ailuropoda melanoleuca) is a flagship species [14,15] that is endemic to China, and it is regarded as a “national treasure” and a “living fossil”. Due to human activities, global climate change and increased urbanization, the living space of giant pandas has been drastically reduced and divided into 33 local populations [16,17], restricted to six isolated alpine valleys in Qinling, Minshan, Qionglai, Daxiangling, Xiaoxiangling and Liangshan [18,19]. Conservation efforts for giant pandas in China have been ongoing for decades. To protect the fragmented metapopulation of giant pandas and integrate the management of fragmented habitats, the Chinese government established the GPNP, one of the country’s first national parks [20,21], with the aim of setting a good example for a global biodiversity conservation hotspot by establishing regulations to ensure connectivity between each habitat [22,23]. However, studies have shown that there are gaps in the coverage of the GPNP [24,25], and that the protection of marginal populations and habitats remains inadequate. In order to create source–sink dynamics, local populations outside of the parks should be integrated around the two core populations [26].

The Liangshan Mountains are the southernmost range of wild giant pandas. They are located in a global biodiversity hotspot and are of great importance for China’s biodiversity conservation. The giant panda population in the Liangshan Mountains has faced serious challenges of habitat loss and fragmentation since the 1970s and 1980s, and the population has declined [27,28]. However, despite these problems, the Liangshan Mountains are not currently included in the management of the GPNP as a whole, making the conservation of giant pandas in the Liangshan Mountains a very serious situation. At the same time, there has been a ten-year gap in the giant panda censuses in this critical area, and conservation managers are unable to track recent population and habitat dynamics.

Meigu Dafengding National Nature Reserve is rich in biodiversity and is the core range of the giant panda population in the Liangshan Mountains. It is also the link between the major giant panda reserves in the Liangshan Mountains and the key corridor linking the two local populations of giant pandas in the mountains. It plays an irreplaceable role in the reproduction, gene exchange and long-term survival of the giant panda population [29]. If the habitat of the giant pandas in this area is lost or destroyed, the Liangshan giant panda population is likely to be segregated into three populations, which will inevitably block the genetic exchange of the entire Liangshan giant panda population. Therefore, this reserve is the core foundation for the rejuvenation of the Liangshan giant panda population and the function of habitat expansion and radiation to ensure the stability and growth of the Liangshan giant panda population, including habitat quality, integrity and safety [30].

The study was conducted in the Meigu Dafengding National Nature Reserve with the following objectives: (1) to collect giant panda fecal samples in the main distribution area of the giant pandas in the Meigu Dafengding National Nature Reserve, and to calculate the number of giant pandas using the distance–bamboo stem fragments method; (2) to calculate the environmental capacity and nutrient capacity based on the existing data of the reserve, and to evaluate the effectiveness of protection compared with the existing number of giant pandas in the reserve; (3) to evaluate the spatial distribution pattern of suitable habitat for giant pandas and the habitat selection characteristics of giant pandas in the Meigu Dafengding Nature Reserve, based on the quantitative analysis of the relationship between giant panda tracks and environmental variables of the field survey; and (4) to conduct an assessment of habitat fragmentation. The results of the assessment will not only contribute to a deeper understanding of the relationship between giant pandas and their natural habitats, and provide a scientific basis for the in situ conservation of giant pandas in the reserve, but will also provide a reference for the future inclusion of giant panda populations in this mountainous system under national park management, and facilitate the expansion and improvement of the protection radius of the GPNP.

2. Overview of the Study Area

The Meigu Dafengding National Nature Reserve is located in the northeastern part of Meigu County, Liangshan Yi Autonomous Prefecture, Sichuan Province (longitude: 102°52′–103°20′ E, latitude: 28°30′–28°50′ N) (Figure 1A). It is situated in the middle of the Hengduan Mountain Range on the southeastern edge of the Tibetan Plateau. The altitude of this area is 1240 m to 3835 m, with an annual precipitation of 1089 mm, and a relative humidity of 80%. The frost-free period ranges from 230 d to 280 d [31]. The terrain is tilted from southwest to northeast, presenting a deep cut in the mountainous terrain, which is characterized by a subtropical monsoon humid climate. The Meigu Dafengding Nature Reserve has abundant natural resources, with a remarkable vertical structure of the soil. The soil is mainly represented by yellow, purple, yellow–brown, brown and other soil types, which facilitate the growth of bamboo, the giant panda’s staple food. Six bamboo species, including Bashania fangiana, Yushania brevipaniculata and Yushania ailuropodina, are widely distributed in the area with good growing conditions [32].

3. Materials and Methods

3.1. Field Investigation

Between February 2023 and June 2023, taking the survey area as a whole using 1:50,000 paper or electronic maps as the basis, the reserve was divided into several 2 km² survey grids after fully considering the topography, vegetation and ecological habits of giant pandas in the adjacent survey area. The “Z”-shaped route was taken as the investigation area, and the “U”-shaped or ring route was used as the investigation interval. The shortest distance through the giant panda habitat was guaranteed, and the total length of each survey plot was no less than 0.75 km [28] (Figure 1B).

If giant panda feces was found in a given habitat, polyethylene gloves were used to collect 1 to 3 fecal samples consisting of approximately 100 bamboo nodes. The degree of freshness of the feces (1~3 days, 4~15 days, >15 days) was determined. GPS was used to record the latitude, longitude, elevation and collection date of the fecal tracking site.

3.2. Distance–Bamboo Stem Fragments Analysis

First, a distance analysis was carried out. Based on previous investigations and studies, such as radiolocation, the thresholds used in the distance analysis were: 1~3 days for a distance of less than or equal to 1.5 km; 4~15 days for a distance of less than or equal to 2.5 km; and more than 15 days for a distances of less than or equal to 3.5 km [28]. Since the precise location of each giant panda tracepoint was determined using GPS during the field survey, the distance between each tracepoint was calculated by using ArcGIS 10.2 software for proximity analysis, and the distance was compared with the distinction threshold. If the distance was greater than the distinction threshold, it was counted as two distinct individuals. Trace data less than the distinction threshold were analyzed for the bamboo stem fragments analysis.

Second, the bamboo stem fragments analysis method was used to determine the composition of giant panda feces (stem, stem and leaf, stem and shoot, leaf and shoot). Only leaf or shoot were used when the average bamboo stem fragments were recorded as −1. The remaining bamboo stem fragment lengths were measured with electronic Vernier calipers to an accuracy of 0.01 mm. The mean bamboo stem fragment length was calculated and recorded as the average bamboo stem fragment length after retaining 2 valid digits. Taking 2.0 mm as the threshold for distinguishing bamboo stem fragments, when the difference between the mean values of the bamboo stem fragments was greater than 2.0 mm, it was determined as two individuals; otherwise, it was considered as the same individual [28].

3.3. Environmental Capacity

The environmental capacity of a population is an important concept in wildlife management. Determining the environmental capacity of a given animal involves the ecological characteristics and habitat characteristics of a given species, and requires knowledge of the animal’s standing, traveling, social interaction, feeding and other characteristics [33]. Currently, the methods commonly used to estimate environmental capacity mainly involve the measurement of nutrient and spatial capacities. The spatial capacity is estimated based on the overall habitat pattern and the minimum spatial requirements for wildlife to survive and thrive. Therefore, the environmental capacity (K) was estimated using the available habitat area in the Meigu Dafengding National Nature Reserve, based on the following formula:

$$\mathrm{K}=\frac{\mathrm{Area} \mathrm{of} \mathrm{available} \mathrm{habitat}\left(1+\mathrm{Percentage} \mathrm{of} \mathrm{area} \mathrm{of} \mathrm{non}-\mathrm{utilized} \mathrm{habitat} \mathrm{for} \mathrm{giant} \mathrm{pandas}\right)}{\mathrm{Minimum} \mathrm{space} \mathrm{requirements} \mathrm{of} \mathrm{individual} \mathrm{giant} \mathrm{pandas}}$$

According to the results of the Fourth National Survey on Giant Pandas, the total area of the Meigu Dafengding Protected Area is 506.55 km²; the area of giant panda habitat is 358.56 km² and the area of potential habitat is 6.90 km² [28]. Therefore, the available area for giant pandas in the protected area is 365.46 km². The minimum space requirement of giant pandas in Meigu Dafengding has yet to be reported. The minimum home area measured in Hu Jinchu in Wolong (3.9 km²) was used as the basis for the calculation [34].

Based on the growth parameters and distribution area of bamboo as the staple food in the reserve, the aboveground biomass of bamboo was estimated. Combined with the individual food requirements of giant pandas reported in the literature, the nutrient capacity of giant pandas in the reserve was comprehensively deduced. The following data were first calculated based on previous measurements of bamboo samples:

$$\mathrm{Annual} \mathrm{bamboo} \mathrm{biomass} \mathrm{per} \mathrm{unit} \mathrm{area} = \mathrm{average} \mathrm{number} \mathrm{of} \mathrm{annual} \mathrm{bamboo} \mathrm{per} \mathrm{unit} \mathrm{area} \ast \mathrm{average} \mathrm{weight} \mathrm{of} \mathrm{annual} \mathrm{bamboo}$$

$$\mathrm{Annual} \mathrm{bamboo} \mathrm{biomass} = \mathrm{annual} \mathrm{bamboo} \mathrm{biomass} \mathrm{per} \mathrm{unit} \mathrm{area} \ast \mathrm{bamboo} \mathrm{forest} \mathrm{area}$$

$$\mathrm{Net} \mathrm{weight} \mathrm{gain} \mathrm{per} \mathrm{unit} \mathrm{area} =\left(\mathrm{average} \mathrm{number} \mathrm{of} \mathrm{annual} \mathrm{bamboos} - \mathrm{average} \mathrm{number} \mathrm{of} \mathrm{dead} \mathrm{bamboos}\right) \ast \mathrm{annual} \mathrm{bamboo} \mathrm{biomass}$$

Total net gain = net weight gain per unit area * bamboo forest area

Previous studies have reported that the distribution area of bamboo forests in this reserve was 44,086 hm², accounting for 87.2% of the total area of the reserve. Six species of bamboo were distributed in the area: Yushania ailuropodina > Bashania fangiana > Yushania brevipaniculata > Yushania dafengdingensis > Yushania mabianensis > Chimonobambusa szechuanensis. Among them, the 2018 giant panda field survey and infrared monitoring data of giant panda activity and fecal tracepoints were compared based on the bamboo distribution map. The areas of giant panda activity and traces were mainly localized to Zhuluyida, Yiziyakou, Yiziheba, Amiduer, Dueryida, 514 forest farms and other areas. In these areas, the total distribution area of Yushania ailuropodina, Bashania fangiana and Yushania dafengdingensis was 41,558 hm², accounting for 94.27% of the distribution area of bamboo forests. Other bamboo species were scattered in small areas [30]. Therefore, this study selected the three bamboo species with the widest distribution area in the reserve to explore the nutrient capacity of giant pandas in the reserve. According to the daily intake of each giant panda, the fresh weight of bamboo was approximately 12.5 kg, and the discard rate was 60% [27]. This was used to determine the maximum nutrient capacity of giant pandas in the conservation area using the following formula:

$$\mathrm{Maximum} \mathrm{nutrient} \mathrm{capacity} = \mathrm{Total} \mathrm{net} \mathrm{gain}/\left\{\left[\mathrm{daily} \mathrm{intake}/\left(1- \mathrm{discard} \mathrm{rate}\right)\right]\times 365\right\}$$

However, in reality, giant pandas do not feed on all of the bamboo in the area, and instead feed selectively. The staple bamboo species consumed by giant pandas in the protected areas mainly consist of Yushania ailuropodina, Bashania fangiana and Yushania dafengdingensis. In the Liangshan Mountains, the feeding rates of giant pandas on the three types of bamboo were 14.15%, 9.29% and 12.91%, respectively [5]. Therefore, the nutrient capacity under natural conditions can only be obtained by combining the feeding rate of giant pandas using the following formula:

$$\mathrm{Actual} \mathrm{food} \mathrm{capacity}=\mathrm{Total} \mathrm{net} \mathrm{gain} \times \mathrm{intake} \mathrm{rate}/\left\{\right[\mathrm{daily} \mathrm{intake}/1-\mathrm{discard} \mathrm{rate}] \times 365\}$$

3.4. Habitat Suitability

3.4.1. Species Distribution

During the survey, 151 valid traces of giant pandas were obtained, and their latitude, longitude and altitude information were determined and recorded repeatedly. Within ArcGIS, the minimum home range of giant pandas was used as a differentiation criterion to avoid spatial autocorrelation. Points with a distance of 1125 m or less were excluded using proximity analysis, and ultimately, only 41 valid points were retained (Figure 2A).

3.4.2. Environmental Variables

During long-term natural selection, giant pandas have gradually adapted to the forest environment of deep subtropical mountains and valleys. Although they have survived in this environment, their distribution is attributed to the comprehensive effects of various environmental factors such as topography and landform. The habitat requirements are extremely stringent, with a preference for gentle slopes and high-density bamboo forests, associated with movement and energy conservation [35]. The spatial distribution is also influenced by factors such as climate and land use type [36,37]. In order to analyze the habitat suitability of giant pandas in the Meigu Dafengding Nature Reserve, four factors, namely, topography, climate, disturbance and resources, were selected. After deleting factors with a high covariance (|r| ≥ 0.8) and avoiding spatial autocorrelation, the remaining factors were used as comprehensive assessment indices.

Topographic factors including slope and slope aspect were extracted via spatial analysis of digital elevation model (DEM) data (30 m × 30 m) downloaded from the geographic space cloud (http://www.gscloud.cn/ accessed on 21 April 2023). The U.S. Geological Survey and NASA Cooperative Satellite Data (https://earthexplorer.usgs.gov/ accessed on 21 April 2023) were used to obtain vegetation information, and identify and accurately classify remote sensing images. The visual interpretation was performed according to the bamboo location, and the vector map of the bamboo cover was determined via manual interpretation. The average temperature and average precipitation data were extracted from the ERA5 dataset. Vegetation type, settlements, roads and river data were obtained from the Resource and Environment Science Data Center (www.resdc.cn accessed on 20 April 2023) and processed via the Spatial Analysis Distance tool in ArcGIS 10.8 to determine the Euclidean distance grid [38]. Vegetation type, based on the distribution of vegetation in the Meigu Dafengding Nature Reserve and field surveys, was categorized into five major groups: coniferous forest, mixed coniferous broad-leaved forest, broad-leaved forest, scrub and meadow. The global 30 m resolution land cover dataset (www.globeland30.Org acccessed on 19 April 2023) was acquired, extracted and masked by the Meigu Dafengding Nature Reserve layer, and the land use types covered only cropland, woodland, shrubs and grassland (

Table 1. Environmental factors affecting habitat of giant pandas in Meigu Dafengding National Nature Reserve.

Types	Factor Codes	Description of Factors	Unit	Date Sources
Climate	Temperature	Annual average temperature	Centigrade (°C)	Google Earth Engine platform (earthengine.Google.com acccessed on 19 April 2023)
	Precipitation	Annual average precipitation	Millimeter (mm)
Topography	Altitude	Altitude	Meter (m)	Geographic space cloud website (http://www.gscloud.cn/ acccessed on 21 April 2023)
	Aspect	Aspect	Degree (°)
	Slope	Slope	Degree (°)
Interference	Road	Distance to roads	Meter (m)	Resource and Environment Science Data Center (http://www.resdc.cn/ acccessed on 20 April 2023)
	Settlement	Distance to settlements	Meter (m)
Resources	River	Distance to rivers	Meter (m)	Resource and Environment Science Data Center (http://www.resdc.cn/ acccessed on 20 April 2023) U.S. Geological Survey and NASA Cooperative Satellite Data (https://earthexplorer.usgs.gov/ acccessed on 21 April 2023)
	Bamboo	Remote sensing of bamboo coverage
	Vegetation	Categorical variables divided into five categories: coniferous forest, mixed coniferous broad-leaved forest, broad-leaved forest, scrub and meadow	/
	Land use	Cropland, woodland, shrubs and grassland	/	www.globeland30.Org acccessed on 19 April 2023

).

In order to improve the accuracy and reliability of the model, the environmental variables were uniformly analyzed by setting the data coordinates to the WGS-1984-UTM-Zone-48N projection, resampling the uniform grid size and converting the data to the ASCII format for MaxEnt model operation.

3.4.3. Predictive Analysis Based on MaxEntNT Model

The MaxEnt version 3.3.3k was utilized to construct the model. To access the importance and contribution of environmental variables in the distribution of giant pandas in the study area, the jackknife method was employed. In this approach, 25% of the distribution points of giant pandas were randomly selected as test data, while the remaining 75% of the distribution points were utilized as training data for the model [39].

The area under the curve (AUC) of the receiver operating characteristic curve (ROC) is not affected by the probability of occurrence of distribution points and judgment thresholds, and is suitable for evaluating the model effect [40,41]. The AUC value tends to be stable during multiple simulations after more than 10 repetitions, and the prediction accuracy does not change significantly with the increase in the number of repetitions. Therefore, the model was cross-validated 10 times to ensure the predictive stability of the model. The maximum AUC was always less than 1 [42], but higher AUC values indicate that the selected environmental variables were highly correlated with the geographic distribution of the predicted species, suggesting a higher prediction accuracy.

MaxEnt was used to simulate the habitat suitability of the Meigu Dafengding National Nature Reserve, and the outputs of the simulation results were obtained in the ASCII format. These were then reclassified after superimposing the boundary of the Meigu Dafengding National Nature Reserve. Maximizing the sum of sensitivity and specificity as the classification criteria, the giant panda habitat was classified into four levels [43]—unsuitable habitat, poorly suitable habitat, moderately suitable habitat and highly suitable habitat—and the proportion and corresponding area occupied by each region in the study area was calculated.

3.5. Habitat Landscape Pattern Index

The landscape pattern index quantifies spatial information of the landscape, reflecting its structural composition and spatial distribution [44,45]. The landscape pattern index was calculated using FRAGSTATS 4.2, and five landscape pattern indices were selected to represent the connectivity and fragmentation of suitable habitats. The number of patches (NP) refers to the number of patches in a given type of landscape; the mean patch size (MPS) refers to the average size of habitat patches; and the largest patch index (LPI) measures the proportion of the entire landscape area occupied by the largest patches. The aggregation index (AI) is a metric that quantifies how aggregated or dispersed a particular patch type is in the landscape. The percentage of landscape (PLADJ) indicates the proportion of patches having neighbors of the same patch type.

$$\mathrm{MPS}=\frac{\mathrm{TA}}{\mathrm{N}}(\frac{1}{\mathrm{10,000}}) (\mathrm{MPS}>0)$$

(1)

$$\mathrm{LPI}=\frac{\max \left({\mathrm{a}}_{\mathrm{ji}}\right)}{\mathrm{A}} (0<\mathrm{LPI}\le 100)$$

(2)

$$\mathrm{AI}=\left[\frac{{\mathrm{h}}_{\mathrm{ik}}}{\max \to {\mathrm{h}}_{\mathrm{ik}}}\right](100) (0<\mathrm{AI}\le 100)$$

(3)

$$\mathrm{PLADJ}=\left[\frac{{\mathrm{g}}_{\mathrm{ii}}}{{{\displaystyle \sum }}_{\mathrm{k}=1}^{\mathrm{m}}{\mathrm{g}}_{\mathrm{ik}}}\right](100) (0 < \mathrm{PLADJ} \le 100)$$

(4)

In these equations, A is the total area; a_ij indicates the area of a specific patch labeled as ij; h_ik refers to the count of adjacent pixels within the same patch type i, calculated using the single-count method i based on the single-count method; max → h_ik represents the maximum number of similar adjacencies between pixels of patch type (class) i (see below) based on the single-count method; g_ii denotes the number of similar adjacencies between pixels of patch type (class) i based on the double-count method; and g_ik indicates the number of adjacencies between pixels of patch types (classes) i and k based on the double-count method.

4. Results

4.1. Population Size Survey

A total of 167 giant panda fecal bamboo stem fragment tables were completed in the field survey. Seven records were excluded with less than 30 bamboo stem fragments, six records were excluded due to missing specific collection dates, and three records were excluded due to missing information regarding the freshness of the feces. The final 151 bamboo stem fragment records were used to calculate the average value of each bite independently. The longitude and latitude of fecal collection, the mean value of bamboo stem fragments, the date of collection, the latest date, the furthest date, the degree of chewing and the presence or absence of small fecal data were imported into the wild giant panda population measurement system. The results show that a total of 28–29 giant panda individuals (95% confidence interval) were distributed in the Meigu Dafengding Nature Reserve. In addition, the population density of giant pandas in the Meigu Dafengding National Nature Reserve was analyzed. The giant pandas in the study area were mainly concentrated in Wahei–Yiziyakou. The highest distribution density was at Hongxi Station (0.682 pcs km²), followed by Lanlong Station (0.132 pcs km²). Traces of giant pandas were rarely detected at Shuwo, Longwo and Weiluohe Stations (0.005 pcs km²), with the entire reserve having a distribution density of only 0.272 pcs km² (Figure 2B).

4.2. Environmental Capacity

Using the area of available habitat in the reserve to estimate the environmental capacity, the reserve can provide a stable and healthy habitat for up to 120 giant pandas. In terms of food resources, the reserve is equipped with sufficient food resources to accommodate at least 121 giant pandas (

Table 2. Bamboo biomass and nutrient capacity of Meigu Dafengding National Nature Reserve.

Variables	Bashania fangiana	Yushania brevipaniculata	Yushania ailuropodina
Net gain (plant/m2)	0.12	0.14	0.97
Bamboo biomass per plant (kg)	0.078	0.150	0.150
Net increase in biomass per unit area (kg/km2)	9360	21,000	145,500
Bamboo forest area (km2)	103.04	161.70	44.17
Annual net increase in bamboo biomass (kg)	964,454.4	3,395,700	6,426,735
Feeding rate (%)	9.29%	14.15%	12.91%
Nutrient capacity (pcs)	7.86	42.13	72.74

).

In addition to the above three kinds of bamboo used as a nutritional staple by giant pandas, the reserve also contains other bamboo species, which suggests an adequate food supply for giant pandas in the reserve.

4.3. Analysis of Factors Affecting Habitat Selection by Giant Pandas

The results of the jackknife test show that the average precipitation ranked the highest in terms of importance, at 50.8%, followed by the degree of bamboo coverage, vegetation type and road distance, and settlement distance. The model showed that average precipitation, road distance and slope made the highest contributions, with a cumulative contribution of 78.5%. The contributions of bamboo cover, vegetation type and mean temperature decreased sequentially (Figure 3).

The response curves of each environmental variable and the distribution probability show that giant pandas in the Meigu Dafengding Nature Reserve tend to prefer areas close to roads and settlements for their frequent activities; the farther the distance from roads and settlements, the lower the distribution probability. In the range of 3000 m, the distribution probability increases with the increase in altitude. Beyond 3000 m, the distribution probability is negatively correlated with altitude. The altitude range for appropriate habitat selection is 2500 m to 3500 m. The probability of the distribution of gentle slopes with a slope orientation of 10–30° to the south is more than 50%. The distribution probability of slopes exceeding 60° tends to be stable and these are rarely detected. The mean annual temperatures ranged from 8.5 °C to 10 °C in appropriate habitats. At temperatures below 8.5 °C, the distribution probability increased with increasing temperature. The vertical zones are divided according to the topography, climate and vegetation types of the area. Subtropical and tropical montane coniferous forests and subtropical sclerophyllous evergreen broad-leaved forests are the main habitats for giant pandas in the Meigu Dafengding Nature Reserve (Figure 4).

4.4. Evaluation of Habitat Suitability

Based on the evaluation of the model using ROC analysis, the AUC values for the test data and training data were 0.97 and 0.96, respectively. These results suggest that the model’s predictive accuracy is exceptionally high, demonstrating a strong level of performance and the selection of highly representative environmental variables. According to the maximum sensitivity and specificity of the modeling data, the distribution threshold of the suitable areas was determined. The areas were reclassified to obtain a map of habitat suitable for giant pandas in the Meigu Dafengding National Nature Reserve. The results show that highly suitable habitat (>0.5) in the whole protected area comprised only 54.16 km², accounting for 10.68% of the total protected area; 23.58 km² was distributed in the core area, accounting for 7.37% of the total core area. The smallest proportion in the buffer zone was 4.58%. The experimental area accounted for the highest proportion of 9.71%. Poorly suitable (0.1–0.33) and unsuitable (<0.1) habitats were widely distributed, with areas of 82.57 km² and 341.52 km², accounting for 16.3% and 67.42% of the total area of the reserve, respectively. The unsuitable habitat in the core area was 238.30 km², accounting for 74% of the core distribution area. The proportion of unsuitable and low-suitability habitat was 88.27%, covering most of the core area. The area of moderately suitable habitat (0.33–0.50) was 28.3 km², which was the most widely distributed area in the experimental area, constituting 46.5% of the experimental area. This surrounded the highly suitable habitat in the form of fragmented distribution (Figure 5).

4.5. Habitat Fragmentation

MPS, NP and LPI reflect the degree of habitat fragmentation in the Meigu Dafengding National Nature Reserve, while AI and PLADJ reflect the degree of aggregation of habitats. The results show that the moderately suitable habitat existed in the form of multiple patches, with the smallest MPS and LPI values, suggesting a high degree of fragmentation. The AI was the lowest at 55.82%, reflecting a high degree of dispersion. The number of highly suitable habitat patches was 85; the number of patches was the lowest among the habitat types, and the AI and PLADJ were second only to unsuitable habitat. The LPI value of unsuitable habitat was up to 42.59%, and the AI and PLADJ were 94.02% and 93.51%, respectively. These existed together in the eastern part of the study area involving the primary forest area in a coherent manner, and surrounded the highly, moderately, and poorly suitable habitats in the western portion of the study area in the form of a large area boundary. The degree of aggregation was in the following order: unsuitable habitat > highly suitable habitat > poorly suitable habitat > moderately suitable habitat (

Table 3. Habitat landscape pattern index of Meigu Dafengding National Nature Reserve.

Type	Highly Suitable Habitat	Moderately Suitable Habitat	Poorly Suitable Habitat	Unsuitable Habitat
Mean patch size/km2	0.64	0.12	0.32	2.01
Number of patches	85	236	258	170
Largest patch index/%	2.86	0.643	9.15	42.59
Aggregation index/%	80.99	55.82	71.51	94.02
Percentage of landscape/%	79.53	54.73	70.69	93.51

).

5. Discussion

Population size characteristics are important quantitative features of wildlife populations, and an accurate population census is particularly complex and important for giant pandas, who have a significant tendency to avoid anthropogenic activities. It is one of the indicators for assessing regional biodiversity and conservation management effectiveness [46], which provides a basis for population dynamics. In the third and fourth national population surveys of giant pandas, the distance–bamboo stem fragments method was mainly used, combined with molecular biological methods [27,28]. The molecular method involves a mature technology, producing reliable results and higher accuracy. However, it has yet to be used as the primary method for giant panda population surveys because it is difficult to collect high-quality feces in practice routinely.

The distance–bamboo stem fragments method is based on the spatial use of wild giant pandas [47], and their feeding behaviors [48] and habitat selection. Giant pandas have now evolved almost exclusively to feed on bamboo [35,49], with more than 99% of their diet consisting of subalpine bamboo [50]. Bamboo stem fragments are chewed and swallowed. They pass through the digestive tract and are excreted, still maintaining their original length and shape. These bamboo stem fragments (BSF) in giant panda feces are thought to represent the characteristics of individuals, and can therefore be used to estimate the age and population structure of giant pandas in the wild [51]. Nonetheless, giant pandas are gregarious, and usually only one individual of the same age is found in a group. As giant pandas age, the spacing between teeth increases, the anal sphincter relaxes, and the probability of average fecal BSF in two individuals being extremely similar is very small [50]. Thus, it is possible to differentiate between individual giant pandas based on the distance–bamboo stem fragments method.

The analysis based on the distance–bamboo stem fragments method showed that 28 to 29 giant pandas lived in the Meigu Dafengding National Nature Reserve with a 95% confidence interval. The results of the Fourth National Giant Panda Survey, which was initiated in October 2011, based on the distance–bamboo stem fragments method combined with molecular genetics, and the 2017 Giant Panda Population Survey in the Meigu Nature Reserve showed that the number of giant pandas in the Meigu Dafengding Nature Reserve was 23 [28,29]. Excluding the effects of systematic and random errors, our results show a steady increase in the number of pandas in the study area, within the spatial accommodation of the habitat. Meanwhile, previous studies have shown that the minimum habitat area for wild giant panda populations is 114.7 km² [52]. The habitat area of Meigu Dafengding National Nature Reserve is 506.55 km², and the availability of habitat space in the reserve can ensure the abundance of wild giant pandas in the area and increase the probability of survival. However, it should be noted that using the biomass and net growth of bamboo alone to estimate the capacity of giant pandas is not entirely consistent with reality, because giant pandas like to wander in moderately sparse bamboo forests. Therefore, it should be combined with estimates based on the local micro-environmental estimation, in order to accurately judge the environmental capacity. Nonetheless, the assessment of suitable habitats should not be neglected, as wildlife responds differently to environmental changes due to its unique habitat and scaling needs [53]. In general, habitat fragmentation creates very small patches, making it difficult for animals to find appropriate habitats. Moreover, the negative edge effect further reduces the habitat quality, leading to the extinction of local species and a significant reduction in the population [54]. Therefore, the estimation of the environmental carrying capacity needs to be accompanied by the identification of suitable habitats for the target species.

The spatial utilization characteristics of wildlife not only determine the distribution pattern of the species, but also reflect the characteristics of suitable habitats [47]. Among the environmental factors in the Meigu Dafengding National Nature Reserve, average precipitation, distance from the road and slope, with a cumulative contribution rate of more than 70%, are the key contributing factors. They are also the most critical factors influencing the selection of habitat for giant pandas. Precipitation is the predominant factor as the survey of the giant panda trace sites in this study coincided with the rainy season. When the temperature rebounded, the rainfall gradually increased, and new bamboo shoots began to sprout from Yushania brevipaniculata and Yushania ailuropodina at low altitudes. Precipitation had a significant effect on the growth of spring shoots, which provided relatively high crude proteins and crude fats for giant pandas with carnivorous intestinal tracts [31].

Accelerated regional economic development and the construction of roads and other infrastructure have altered the land use pattern due to the lack of rational and scientific planning. This has led to increased pressure on the ecosystem, resulting in a changing habitat structure [55]. Distance from the road was a secondary factor in habitat selection, as the distance from the road was inversely related to the probability of giant panda occurrence, which is contrary to the results of other protected area studies [39]. Theoretically, the road is embedded in the giant panda habitat’s points, lines and surfaces. Vehicular traffic and pedestrian noise significantly interfered with the habitat and reproduction of giant pandas. The larger the radius of the spread of the interference, the greater the interference with the giant panda habitat and activities [56]. Analysis of the unique characteristics of this study area revealed that the range of frequent activities of giant pandas was restricted to the roadside and surrounding areas in Wahei–Yiziyakou, which has 103 provincial highway crossings. However, the avoidance of human activities led to the outward spread of the giant pandas along the longitudinal disturbance of the road, suggesting that the distance from the road is one of the limiting factors. Overpasses and bridges, as well as underpasses, are a solution to mitigate the road-induced fragmentation of wildlife populations during dispersal [57,58]. The aim is to connect critical habitats, mitigate the road barrier effects and reduce road mortality, as well as to allow the safe movement of animals on busy roads [59,60]. A direction for further research is therefore the installation of highly visible and well-lit crossing structures on Provincial Highway 103.

The slopes on which giant pandas carry out their activities in various mountain areas have two common characteristics, namely, giant pandas mostly travel in areas with slopes below 40 degrees; with increasing slope, the frequency of giant panda activities gradually increases and then slowly decreases. The selection of slope in the study area was similar to that reported in related studies, indicating that giant pandas tended to prefer sunny slopes or ravines with gentle slopes in different areas with different habitat characteristics [35]. The gentler the slope, the larger the tree diameter at breast height and the better the bamboo development farther away from the fallen trees. At the same time, giant pandas like to move in bamboo forests with gentle slopes [35,61], where they can rest on their backs or sit upright, releasing their forelimbs to grasp the bamboo and bamboo shoots for foraging [62]. Thus, they conserve energy by adapting to this low-quality food and low-energy intake [29], which is the basis for their life up to the present day.

The quality of habitat is critical for species continuation and biodiversity conservation [63]. Fragmentation and habitat loss are the main threats to biodiversity and can lead to species extinction [64,65]. Based on the MaxEnt results, the spatial distribution of suitable habitats for giant pandas in the Meigu Dafengding Nature Reserve is relatively concentrated, with a few scattered patches in the southwestern part of the reserve. Furthermore, the moderately and poorly suitable habitats show a wrapping trend outside the highly suitable habitats, and appear to be expanding in the southwestern portion of the study area. Poorly suitable habitats are partially penetrating and emerge on the periphery of moderately and highly suitable habitats, with a wide range of distributions; unsuitable habitats are integrated and continuously distributed over a large area, resulting in some suitable habitats occurring as “isolated islands”.

Under the existing functional planning of the reserve, the highly suitable habitat for giant pandas has been divided into two areas: the surrounding area of Wahei–Yiziyakou in the central part of the reserve and the northern part of the study area. Local populations are concentrated in these two areas, resulting in the highest kernel density of giant pandas in the vicinity of the Hongxi Station. No giant pandas were found at the Longwo and Shuwo stations, despite their proximity to these areas. This was due to their distribution in the towns and the severe anthropogenic disturbance. However, the segmentation between the two patches in the central and northern parts of the study area is mainly attributed to the spacing of unsuitable habitats and small areas of conservation vacancies in the form of habitat fragmentation, resulting in the restricted migration of giant pandas. The small vacant area originates from the national agenda of vigorously developing the return of farmland to forests and vegetation restoration, alongside other giant panda conservation measures. As a result, the habitat patches are interconnected to form a large area of suitable habitat for the long-term survival and reproduction of giant pandas. Giant pandas actively avoided areas in the experimental area within these patches where anthropogenic disturbances were not eliminated during migration, resulting in gaps in moderately and highly suitable habitats. However, corridors can increase the degree of connectivity between habitat patches, increase species migration rates between patches, and facilitate species and gene exchange [66,67]. Establishing corridors in areas of high anthropogenic disturbance is also a high priority to reduce conservation vacancies.

The northern part of the Meigu Dafengding National Nature Reserve is bounded by the county border between Meigu County and Ebian County, and is adjacent to the Heizhugou National Nature Reserve. The surrounding areas are centered on central Wahei–Yiziyakou and the northern part of the study area, which represent the two areas of concentrated distribution of giant panda populations. Normal reproduction and gene exchange with the Heizhugou giant panda population will be promoted through the autonomous dispersal of giant pandas. The neighboring Shen Guozhuang Provincial Nature Reserve, located to the west, sees less genetic exchange of giant panda populations between the two reserves due to the unsuitable distribution of large areas of habitat in the southwestern part. The Meigu Dafengding National Nature Reserve, bounded by the Dafengding Ridge, has experienced a significant reduction in gene exchange due to the isolation of primary forests in unsuitable habitats and the southward migration of the Mabian giant panda population to avoid unsuitable habitats in the western and northern parts of the country [30]. Geographically, the Meigu Reserve links the three reserves of Heizhugou, Mabian and Shenguozhuang. Effective conservation measures in Meigu are essential to strengthen the habitat connectivity of the protected areas, and to promote the normal breeding and genetic exchange of giant pandas.

Landscape pattern is defined as the spatial distribution and configuration of landscape elements of different sizes, shapes and attributes, which is manifested by the different types of landscape patches in a random, uniform or clustered pattern in space [68]. Landscape pattern analysis is one of the most important methods to study the response of habitat fragmentation to biodiversity [69]. It can effectively reflect the habitat quality, dynamic changes in habitat suitability and population responses to habitat fragmentation. At the landscape level, the fragmented distribution of suitable habitat for giant pandas in the Meigu Dafengding National Nature Reserve, in the experimental area, shows severe fragmentation, and weak and decentralized patch aggregation. With social development, population increase and the fragmentation of native vegetation due to anthropogenic disturbances, moderately suitable habitats were further lost and cannot be linked to highly suitable habitats. The highly suitable habitats AI and PLADJ were second only to those of the unsuitable habitat, originating from Wahei–Yiziyakou in the central part of the study area. It improved the connectivity of the landscape and thereby contributed to the migration and dispersal of species. It also increased the likelihood of a continued increase in the population. The results complement previous studies demonstrating that ecological corridors are highly connected, link different patches and increase species movement between patches [70,71]. This suggests that conservationists should focus on connectivity between patches so that biodiversity continues to function [72].

Previous studies have shown that forest age is positively correlated with the probability of the presence of giant pandas [50,73]. As the largest primary forest in the reserve, the Dafengding area covers 159.90 km², and bamboos are widely distributed in the lower layers of the continuous primary forest, with nutrient-rich areas and growing conditions [40]. In addition, the primary forest has a sufficient number of caves suitable for Sichuan giant pandas to procreate and raise their young [41]. The primary forest provides space for avoidance and supplies sufficient food resources [74]. However, the activity trajectories of giant pandas were not determined, probably because environmental pollution and grazing disturbances due to mining in the 1980s and 1990s severely limited the distribution of giant pandas in the primary forests of the Meigu Dafengding area [75]. Grazing has become an important anthropogenic disturbance in the protected area [76]. A few minor disturbances can lead to landscape fragmentation, while large-scale and intense disturbances may lead to landscape homogenization rather than fragmentation. Overgrazing in the Dafengding area and large numbers of livestock (e.g., free-range cattle, sheep and horses) in the giant panda habitat pose a major threat to the survival of giant pandas, resulting in avoidance behaviors of giant pandas towards grazing sites and the formation of contiguous, large, unsuitable areas, resulting in unsuitable habitat occupying the vast majority of the study area [67]. In addition, large-scale grazing destroyed the habitats for giant pandas by damaging the growth and regeneration of bamboo, forcing pandas to use alternative habitats [77].

6. Recommendations of In Situ Conservation

Our results show the existence of small vacant areas in the Meigu Dafengding National Nature Reserve following the fragmentation of suitable habitats. The highest proportion of highly suitable habitats was distributed in the experimental area, which was susceptible to anthropogenic activities, resulting in an inability to ensure the long-term survival of the giant panda population. At the same time, some of the giant panda trace points in the study area were beyond the boundaries of the protected area. Due to the lack of effective management and care, the current conservation zoning does not cover all of the suitable habitats, thereby limiting the efficiency of giant panda conservation. To achieve comprehensive protection and systematic restoration of the giant panda’s habitat, evaluate the efficiency of functional areas, reassess and define the functional zoning of reserves, establish a new pattern of ecological protection, and provide a reference for the later stage of the GPNP, it is recommended to include the reserve and its surrounding suitable habitat into the management scope to expand its coverage and consolidate protection and management.

Second, there are frequent signs of giant panda activity in the vicinity of Provincial Highway 103, and there is an increased risk of mixing animal and human tracks, resulting in the limited migration of giant pandas. Therefore, it is imperative to build corridors to improve connectivity between highly suitable habitats. Funnel-shaped barriers of native vegetation and edible bamboo species for giant pandas established around the corridor can create a suitable environment. The real-time monitoring of pandas using the corridor, timely access to information to analyze and explore any challenges along the corridor, and timely policy measures are necessary to remedy the situation. For example, the monitoring equipment in areas frequently visited by pandas should be increased, and standing signs should be set up to highlight the presence of wild giant pandas to remind people to slow down and to prohibit the sounding of horns.

Ultimately, we hope to inform the expansion of the GPNP and other national parks, as well as assist in the conservation of small populations of the species.

7. Study Limitations

The results of this study are specific to giant pandas. However, the freshness of the feces in this study limited the availability of comprehensive genomic data, resulting in the individual counting method being used only for the distance–bamboo stem fragments method. When sufficient genomic data are available to be able to identify individuals from feces, the study methodology will have a wider range and provide a reference for assessing population size and habitat quality in other species.