Distribution of Grazing Paths and Their Influence on Mountain Vegetation in the Traditional Grazing Area of the Tien-Shan Mountains

Abstract

In the Tien-Shan Mountains, Ili Prefecture, Xinjiang, China, the livestock industry has experienced rapid growth in recent decades. However, this expansion has led to increased overgrazing behavior, resulting in the proliferation of grazing paths and a decline in vegetation cover. These factors are considered the main causes of vegetation degradation in the region. To investigate this issue, we conducted a study utilizing unmanned aerial vehicle imagery in the Zollersay Mountains of Ili to examine the distribution of grazing paths and their effects on mountain vegetation, including grassland and Malus sieversii. The results of our study revealed that grazing paths in the area exhibited various formations, including parallel, oblique intersection, and grid. On the hilltop, the grazing paths were not only shorter but also wider, whereas on the hillside, they were denser, indicating a higher concentration of livestock trampling events. It was found that grazing path density played a pivotal role in grassland degradation, with a negative correlation observed between grazing path density and indicators such as the grassland quality index and grass vegetation coverage. As grazing path density increased, the damage inflicted on Malus sieversii by livestock also intensified. However, as the trees grow older, their height surpasses the feeding range of livestock, resulting in reduced grazing impact. The findings of our study carry significant implications for developing scientifically informed livestock policies and promoting the conservation of wild fruit forests.

1. Introduction

Landscapes consisting of small paths and patches formed by livestock trampling are widely distributed in arid and semi-arid regions. The academic community has provided different definitions for these landscapes, as follows: Higgins [1] referred to this landscape type as grazing steps, which represent a broad staircase of narrow pathways formed as a result of grazing behavior on natural slopes. Jin et al. [2] defined it as livestock tracks, specifically referring to paths created by animals such as goats. In this study, we define it as a grazing path, a micro-land formation resulting from the repeated trampling of livestock as they move to and from grazing grassland while foraging along the optimal path. This unique landscape is characterized by an interlacing step pattern, appearing as parallel or perpendicular lines in mountainous areas.

Grazing paths are commonly observed in undulating mountains or hills as well as on gentle slopes, especially in places where livestock gather in the mountains or where water resources are abundant [3]. The slopes where grazing paths are located are typically steep. In terms of trampling intensity, grazing paths tend to form rapidly and easily on sloping hillsides compared to relatively flat areas. These paths can persist for extended periods, potentially exacerbating soil erosion [4].

The types of livestock associated with grazing path formation include domesticated hoofed herbivores such as cattle [4,5,6], sheep [7,8,9,10,11], horses [12], and yaks [3]. The length of grazing paths is closely related to plant abundance [2], and the width correlates strongly with livestock body length, with significant differences between cattle and sheep [3,5,13]. Grazing paths are diverse, with primarily parallel, oblique, and grid patterns [14]. In contrast, in areas with high slopes, animals move at a lower rate, so the density and tortuosity of grazing paths are greater [15]. Zhang [16] found that grazing has an influential effect on the structure of plant communities, with foraging and trampling by goats reducing plant density as well as cover. The trampling of livestock increased the soil capacity, resulting in substantial spatial heterogeneity of grazing paths with individual differences and distribution characteristics [2]. Knowledge of the distribution characteristics and heterogeneity of grazing paths can help to understand the foraging strategies of livestock, which in turn can provide better management measures and policies for ecosystem sustainability.

The impact of livestock on montane vegetation is mainly caused by trampling, foraging, and excretion. Of these, the effects of trampling on grassland community structure are not only complex but also controversial [17,18], as they may be influenced by a combination of environmental conditions [19], grazing intensity [20], study scale [21], and grazing history [22]. Classical ecological theory suggests that appropriate grazing has positive effects on grassland production as well as compensatory growth effects [23] and the highest species diversity under moderate disturbance [21]. However, several studies have found negative effects of grazing on plant diversity, yield, and height [24,25,26], especially in degraded, malnourished, or low-productivity grassland ecosystems with low grazing resilience [27]. The different formations of grazing paths reflect the different selections of plants on both sides of the path by livestock, increasing the number of species with high palatability. Furthermore, it is potentially causing more rampant growth of toxic grasses, seriously endangering the lives of livestock [28], with cover rebounding after the adoption of grazing exclusion measures [29]. Grazing behavior not only has an effect on herbaceous vegetation but also affects the growth of young trees [30]. Trees are gnawed by hoofed animals when they are young, which affects the early radial growth of trees, but trees experience wider radial growth once they are above the range of animals gnawing [31]. The study of grazing path influences on mountain vegetation is important for enriching grazing ecology and conserving vegetation.

Zollersay Mountains in Gongliu County, Ili Prefecture, Tien-Shan Mountains, Xinjiang, China, is a mountain ecosystem with rich species diversity, where the oldest variety of apple tree (Malus sieversii) in the world grows [32]. In recent years, driven by population growth and economic interests, the grazing behavior and density of grazing paths have increased, which has caused a decrease in plant cover and the deterioration of the growing environment of Malus sieversii [33], affecting the safety of the entire mountain ecosystem. However, the spatial distribution of grazing paths and their potential effects on montane vegetation have not been determined.

Therefore, based on grazing paths extracted from unmanned aerial vehicle (UAV) imagery in the traditional grazing area of the Zollersay Mountains in the Tien-Shan Mountains of Xinjiang, China, our study aims to investigate the impact of grazing behavior on mountain vegetation, including grassland and Malus sieversii, providing valuable insights for making informed decisions regarding grazing policies and implementing effective conservation measures for mountain vegetation. To achieve these goals, our study focuses on the following three objectives: quantifying the amount, configuration, and spatial distribution characteristics of grazing paths, assessing the effect of grazing paths on the degradation of mountain grasslands, investigating the role of grazing on the development of Malus sieversii.

2. Materials and Methods

2.1. Study Area

Our study area is located in the villages of Ahele, Kuldening Town, Gongliu County, Ili Prefecture, Xinjiang Uygur Autonomous Region, China (42°9′–42°15′N, 82°42′–82°47′E), which is contiguous with the Mohur river to the east, the Kuldening township to the west, and the Big Gilgalang river to the north. It is part of the Zollersay Mountain in the western part of the Tien-Shan Mountains and has an area of about 50 km² (Figure 1). It is a typical alpine valley landscape with high terrain in the southeast, low terrain in the northwest, and a mountain elevation of 1100–3780 m [34]. Zollersay Mountain has a warm winter and a cool summer, an average annual temperature of 5–6 °C, and an annual rainfall of 300–700 mm. The climate is suitable for the growth of vegetation, with a large number of Malus sieversii and a few Prunus armeniaca L. growing, while the hills have lush vegetation resources with herbs, including Stipa capillata, Festuca sulcata, and Carex, and shrubs, including Spiraea hypericifolia, Caragana sinica, and Rhamnus davurica Pall [35].

As a combined agricultural and livestock village, the village of Azeil on the Zollersay Mountain, where the study area is located, has about 280 regular inhabitants and a total population of 980 people, of whom 80% are Kazakh. Most of the herders live in the flat arches of the hills, while farmers grow Triticum aestivum, Zea mays, and cash crops such as Fritillaria and Helianthus annuus around the hills. Relying on the rich mountainous vegetation resources, the grazing industry in this village has rapidly grown. The number of livestock animals has grown from approximately 2000 in 2004 to more than 3600 in 2022. This is derived from the records of the village committee, with large animals such as cattle and horses being the main types. They are kept mainly in a semi-grazing or semi-captive manner, relying mostly on pasture grazing in summer and autumn and captive breeding with hay in winter and spring. Because of its isolation and low accessibility, no tourism has been developed, and there is little human activity outside of the agro-pastoral industry, making it an ideal area for the study of grazing paths.

2.2. Data and Processing

2.2.1. Image Data

For the grazing path survey of Zollersay Mountain, a PHANTOM 4PRO quadcopter unmanned aerial vehicle (UAV) (

Table 1. PHANTOM 4PRO device parameters.

Indicator	Detailed Parameter
Number of rotors	4
Satellite positioning module	GPS/GLONASS dual-mode
Image sensor	1 inch CMOS, effective pixels: 20 million
Lens	FOV 84°, 8.8 mm/24 mm (35 mm format equivalence), f/2.8–f/11 with autofocus (focusing distance: 1 m–infinity)

) was used to collect high-resolution charge-coupled device (CCD) imagery. Missions were conducted in cloud-free conditions at noon in July 2022 (Figure 2a). The aerial photography mode used single-lens tilt-shift photography for ground-varying flight. A high-resolution digital elevation model (DEM) obtained using UAV photogrammetry in July 2022 was used as the reference elevation to obtain imagery with uniform spatial resolution and minimize feature deformation in consideration of terrain effects. The flight parameters were set as follows: relative altitude of 150 m, spatial resolution (ground sample distance, GSD) of 0.08 m, aircraft flight speed of 6.5 m/s, heading overlap of 70%, and lateral overlap of 60%.

To obtain the coordinates of the control points, 20 ground control points (GCPs) evenly distributed over the flight area were selected; each had easily recognizable targets placed, and the GCPs were measured using the Trimble real-time kinematic (RTK). As a result, measurements with an accuracy of approximately 8 mm were obtained.

We combined Structure from Motion (SfM) with Agisoft PhotoScan software to preprocess the CCD high-resolution imagery, including importing and aligning photographs, ortho-mosaicing, and processing photos by generating dense point clouds, meshes, and textures. The final output resulted in a digital surface model (DSM) and digital orthophoto map.

Of all the GCPs, 10 were selected for spatial adjustment, while the other GCPs were used as pre-defined check points to assess the accuracy of the DSM and digital orthophoto map. After quality checks, the vertical positioning accuracy of the DSM was within 10 cm, and the horizontal accuracy of the digital orthophoto map was within 5 cm.

2.2.2. Sample Data

In July 2022, during the field survey, sixty samples (Figure 1) were set up in areas with typical grazing path distribution based on the distribution characteristics of grazing behavior and other factors, including accessibility and terrain (Figure 1), which were evenly distributed on the hilltop (No. 1–20), hillside (No. 21–40), and hillbottom (No. 41–60). The length of each sample was 50 m. We employed a comprehensive approach. Firstly, we conducted a field survey in which we recorded the number of grazing paths in every sample. Additionally, we carefully selected grazing paths from the field survey data and performed detailed measurements to obtain their length, width, amplitude, curvature, and number of planar bends. These field-measured parameters were then used as reference data to assess the accuracy of the grazing paths extracted through remote sensing. By comparing the extracted trails with the ground-truth measurements, we were able to quantitatively evaluate the accuracy of our remote sensing method.

Five 1 m × 1 m herbaceous samples were set up in each sample. The number of toxic, poorly palatable, fair, and good grasses in each herbaceous sample square was recorded during the field survey (Figure 2e) and used to analyze grassland quality. We extracted the grass vegetation cover area for each sample square on the digital orthophoto map for statistical purposes to determine the extent of herbaceous vegetation cover.

Sample cores of 60 young Malus sieversii trees were collected from the distribution area of the grazing path sample zone and used to analyze the effect of livestock on tree growth. Tree growth cones with an internal diameter of 5.15 mm were used to collect the cores (Figure 2g) separately at the tree’s breast diameter position along two different directions. The collected cores were stored in paper tubes with assigned numbers. The sample cores were cross-dated after natural drying, fixing, and sanding until the surface was smooth, the annual rings were quite visible, and the cell walls were well defined. Then the width of the tree rings was measured using the tree ring analysis instrument LINTAB from Heidelberg, Germany, which has a measurement accuracy of 0.01 µm. The measured tree ring sample sequences were quality checked and controlled using the accuracy checking computer program COFECHA to eliminate some errors in the dating and measurement processes.

2.3. Study Method

2.3.1. Extraction of the Grazing Path

We used the example-based feature extraction workflow of ENVI 5.3 software to classify the digital orthophoto map of the initially extracted grazing paths. However, due to the interference of herbaceous vegetation, most of the grazing paths were composed of multiple line segments, which did not form a complete grazing path (Figure 2i). Therefore, based on this, the grazing paths within each sample were further refined with the help of ArcGIS 10.8 (Figure 2j–m) for the study of the characteristics of the grazing paths (see Supplementary Table S1).

2.3.2. Parameter Extraction of the Grazing Path

The quantitative characteristics of the grazing path, including number, curved length, and average width, are key metrics used to characterize its quantitative properties, as presented in

Table 2. The distribution of the number of grazing paths.

Location	Metrics	Maximum Value	Minimum Value	Mean Value	Variance	Standard Deviation	Standard Error	95% Confidence Interval
Hilltop	Number	30.00	14.00	21.05	18.37	4.18	0.96	21.05 ± 2.01
	Width (m)	1.64	0.32	0.62	0.07	0.27	0.06	0.62 ± 0.13
	Length (m)	48.45	14.36	32.75	142.61	11.64	2.67	32.75 ± 5.59
Hillside	Number	33.00	18.00	24.05	22.58	4.63	1.06	24.05 ± 2.22
	Width (m)	0.95	0.37	0.51	0.02	0.13	0.03	0.51 ± 0.06
	Length (m)	50.61	15.01	27.25	120.01	10.68	2.45	27.25 ± 5.13
Hillbottom	Number	29.00	10.00	20.60	28.15	5.17	1.19	20.60 ± 2.48
	Width (m)	0.83	0.28	0.44	0.01	0.11	0.02	0.44 ± 0.05
	Length (m)	53.63	19.40	35.47	93.51	9.42	2.16	35.47 ± 4.53

. On the other hand, metrics such as amplitude, tortuosity, and number of plane bends are used to characterize the morphological characteristics of the grazing path (

Table 3. Morphological parameters of grazing paths among samples.

Location	Morphological Characteristics	Maximum Value	Minimum Value	Mean Value	Variance	Standard Deviation	Standard Error	95% Confidence Interval
Hilltop	Amplitude (m)	31.00	11.00	24.00	25.53	4.92	1.13	24.00 ± 2.36
	Tortuosity	1.26	1.02	1.09	0.00	0.07	0.02	1.09 ± 0.03
	Number of plane bends	7.00	3.00	4.85	0.77	0.85	0.20	4.85 ± 0.41
Hillside	Amplitude (m)	35.00	15.00	27.00	28.96	5.24	1.20	27.00 ± 2.52
	Tortuosity	1.74	0.89	1.13	0.03	0.18	0.04	1.13 ± 0.09
	Number of plane bends	8.00	2.00	4.25	1.36	1.13	0.26	4.25 ± 0.54
Hillbottom	Amplitude (m)	34.00	15.00	24.00	24.66	4.84	1.11	24.00 ± 2.32
	Tortuosity	1.18	1.01	1.08	0.00	0.04	0.01	1.08 ± 0.02
	Number of plane bends	8.00	1.00	4.50	2.30	1.48	0.34	4.50 ± 0.71

). Density characterizes the distribution of the grazing path.

Amplitude of the Grazing Path

The change in amplitude indicates the physical quantity that reaches its maximum range. The amplitude is used to express the complexity of the topography of the grazing path distribution area in the vertical direction, i.e., the distance from the lowest point to the highest point when the livestock is foraging and trampling, using the following equation:

$$A_i=H_{ih}-H_{il}$$

(1)

where A_i is the elevation drop through the ith sample grazing path; H_ih and H_il are the highest and lowest elevations in the distribution area of the ith sample grazing path, respectively.

We used ArcGIS 10.8 to calculate the amplitude of each sample, then obtained the maximum, minimum, mean, and standard deviation for the hilltop, hillside, and hillbottom.

Tortuosity of the Grazing Path

The tortuosity [36] was used to express the grazing path formation using the following equation:

$$S_{ij}=\frac{L_{ij}}{l_{ij}}$$

(2)

where S_ij, L_ij, and l_ij are the curvature, straight line length, and curve length of the jth grazing path in the ith sample, respectively.

The curvature of each grazing path was calculated using ArcGIS 10.8. We obtained the maximum, minimum, mean, standard deviation, and variance for the hilltop, hillside, and hillbottom.

Number of Plane Bends

The number of plane bends was used to represent the horizontal complexity of the terrain in the grazing path distribution area.

We counted the number of plane bends for each grazing path, then obtained the maximum, minimum, mean, and standard deviation for the top, side, and bottom of the hill.

Density of the Grazing Path

The DW mean was used to calculate the grazing path density using the following equation:

$$D{W}_i=\frac{{\displaystyle \sum _{j=1}^n{W}_{ij}}}{C_i}\times 100\%$$

(3)

where DW_i denotes grazing path density in the ith sample, ${\displaystyle \sum _{j=1}^n{W}_{ij}}$ represents the sum of all grazing path widths in the ith sample, i.e., the width of all grazing paths measured once on the central axis at 25 m (Figure 2d,j), and C_i denotes the length of the ith sample (i.e., 50 m).

Accuracy Evaluation

We used R² and root mean squared error (RMSE) to evaluate the grazing path parameters extracted by remote sensing methods and compared the parameters with those from field surveys.

$$\mathrm{RMSE}=\sqrt{\frac{1}{n}{\displaystyle \sum _{i=1}^n{(Y_i-Y_i^{\prime })}^2}}$$

(4)

where Y_i represents the extracted grazing path, Y_i’ is the measured grazing path, Y is the measured average grazing path, and n is the number of samples.

Figure 3 demonstrates the R² values obtained for our extracted grazing path parameters, which were found to be consistently higher than 0.7. These results serve as an indication of the satisfactory accuracy achieved by our method and meet the requirements of the study.

2.3.3. Grassland Quality Characteristics

Grassland Quality Index (IGQ)

The IGQ [37] was used to indicate the degree of grassland degradation, which was divided into four levels, namely, normal (75 ≤ IGQ ≤ 100), mild (51 ≤ IGQ < 75), moderate (25 ≤ IGQ < 50), and severe (0 ≤ IGQ < 25) degradation.

The equation for calculating the IGQ is as follows:

$$\mathrm{IGQ}=\frac{1}{3}{\displaystyle \sum _{i=-1}^{3}i}\times S_i$$

(5)

where i = −1 represents poisonous grass, i = 0 represents grass with poor palatability, i = 1 represents grass with general palatability, i = 2 represents grass with good palatability, and i = 3 represents grass with excellent palatability. S_i is the percentage of the number of corresponding grasses.

Using the number of grasses surveyed in the field, we calculated the IGQ for each herbaceous sample, which in turn led to the mean IGQ within each grazing path sample.

Grass Vegetation Coverage (GVC)

The GVC was used to express the degree of herbaceous cover using the following equation:

$$GV{C}_i=\frac{A_{ig}}{A_{is}}$$

(6)

where IGQ_i is the GVC of the ith sample square, A_ig is the area of herbaceous vegetation cover of the ith sample square, and A_is is the area of the ith sample square.

Using the herbaceous vegetation cover information on the digital orthophoto map, we calculated the GVC of each herbaceous sample square, which in turn led to the mean GVC value within each grazing path sample square.

2.3.4. Correlation of Grazing Paths with Montane Vegetation

The Impact of the Grazing Path on Grassland

To understand the potential influence of grazing behavior on grass quality, we established linear equations for the number, length, width, amplitude, tortuosity, number of planar bends, and density of grazing paths in the sample with IGQ, respectively. To further verify the relationship between grazing path density and grass quality, we established linear equations for density and GVC.

The Effect of the Grazing Path on the Malus sieversii Tree

To investigate the relationship between the grazing path and the radial growth of the trees, we established a linear equation between the radial growth of a sample of 60 Malus sieversii trees and the density of the grazing path in samples where the trees were located.

3. Results

3.1. Characteristics of the Grazing Path

3.1.1. Quantitative Characteristics of the Grazing Path

As shown in Table 2, the maximum number of grazing paths among all samples was 33, the minimum number was 10, and the average number was 22. The mean number of grazing paths on the hillside was higher than on the hilltop and hillbottom. There was a notable difference between the grazing path widths. The maximum width was 1.32 m, the minimum width was 0.14 m, the average width was 0.52 m, and the standard error was 0.03. The mean width of the grazing path on a hilltop was higher than on a hillside or hillbottom. The maximum length of grazing paths intercepted by the samples was 53.63 m, the minimum value was 14.36 m, the average length was 31.82 m, and the standard error was 1.45. The largest mean length of grazing paths was found on the hillbottom, while the smallest was found on the hillside. Slope was an important factor influencing the choice of foraging areas for livestock. As a result of the high human activity at the hillbottom and the decreasing availability of forage, livestock gradually moved to the slopes, but the steeper slopes made it difficult for livestock to stay for long periods of time, so the width of the grazing path was smaller. As the foraging area shifts from the bottom to the sides and top of the slopes, an essential area to pass through, the number of grazing paths increases. The shorter and wider grazing paths at the hilltop indicated that foraging was relatively easier for livestock there.

3.1.2. Morphological Characteristics of the Grazing Path

As shown in Table 3, the maximum amplitude of grazing paths among all samples was 35, the minimum value was 11, the mean value was 25, and the standard error was 0.68. The hillside has the greatest amplitude. Livestock foraging and trampling are generally done from the bottom of the hill, inclining their bodies up the hill, with the undulating terrain causing the amplitude of the hillside to be higher than the hilltop and hillbottom. The average tortuosity of the grazing path was 1.10, with a maximum value of 1.74, a minimum value of 0.89, and a standard error of 0.01. The mean tortuosity was lower at the hilltop and hillbottom. A field survey showed that these areas have not only relatively simple vegetation types but also relatively gentle topography. In these areas, the trampling and foraging paths were almost straight. However, in hillside areas, the terrain affects the need for livestock to shift their stance back to balance their bodies when foraging. Furthermore, because of the sparse distribution and slow renewal of vegetation due to grazing behavior, the more complex the grazing path, the greater the tortuosity [38]. The maximum number of plane bends in samples for a single grazing path was 8, the minimum value was 1, and the average was 4. The number of plane bends varied with the topography. The less variation in mountainous terrain, the fewer the number of plane bends and the shorter the foraging path for livestock. According to the optimal path theory, livestock will choose hills with few bends to trample when foraging in order to save energy [39].

The morphology of grazing paths on Zollersay Mountain mainly included parallel, oblique, and grid shapes, each of which had a different grazing intensity [8]. Parallel grazing paths (Figure 4a) were mainly distributed in areas with gentler slopes, a relatively large width between foraging and grazing paths, less mellow soil on the surface, and moderate vegetation cover. This type accounted for 5% of the grazing path combination formation. Oblique grazing paths (Figure 4b) were often located in areas with large undulations or at the upper neck of the mountain, where livestock foraging and trampling were heavier and bare ground was formed after the surface mellow soil was peeled off. This type accounted for 5% of the grazing path combination formation. Grid grazing paths (Figure 4c) were distributed on relatively steep slopes where livestock trampled and foraged heavily, producing bare ground or crumbling areas, accounting for 90% of the grazing path combination formation. The grazing path combination formation in the study area was dominated by the grid shape, which was relatively stable with frequent trampling by livestock.

3.1.3. Distribution Characteristics of the Grazing Path

The average density of grazing paths among all samples was 21.18%, the maximum was 36.12%, and the minimum was 11.05% (Figure 5). Compared to the hilltop and hillbottom, the hillside had not only a higher density of grazing paths but also a greater concentration of livestock trampling events. This may be due to the high accessibility of the hillbottom, the fact that there are more roads in the hillbottom, the fact that livestock seldom need to make paths when foraging, and the fact that palatable grasses were consumed early. In comparison to the top and bottom, the hillside was less vegetated as well as steeper, making foraging more dispersed and trampling more frequent. In addition, trampling was more frequent on the hillside because those paths were necessary to reach the bottom.

3.2. Influence of the Grazing Path on Grassland Degradation

3.2.1. Effect of the Grazing Path on IGQ

Our field investigation revealed that grazing has been a long-standing practice in all samples. The quality index of grassland in these samples ranged from 28 to 65, with 67% of the grassland being moderately degraded and 33% being mildly degraded (Figure 6). The IGQ on the hilltop was significantly higher than that on the hillside and hillbottom, indicating that the grassland degradation on the hillside was the most severe.

To gain insights into the potential impact of grazing behavior on grassland, linear equations were established for the number, length, width, amplitude, tortuosity, number of plane bends, and density of grazing paths in all samples to assess their correlation with grass quality (

Table 4. Linear equation for grazing path characteristics and IGQ.

Characteristics of the Grazing Path	Linear Equations in One Element	R2	p
number	y = −0.0334x + 46.913	0.0003	0.89
length	y = −0.1522x + 49.94	0.0082	0.49
width	y = −2.228x + 48.626	0.0008	0.83
amplitude	y = 0.7311x + 42.929	0.0103	0.44
tortuosity	y = −0.0334x + 46.913	0.0003	0.89
number of plane bends	y = −0.2632x + 54.557	0.1119	<0.01
density	y = 0.5546x + 45.892	0.0002	0.92

).

The density of the grazing path had the highest correlation with IGQ (R² = 0.6069), emphasizing its importance as a factor in grassland quality. The more frequent the trampling and foraging activities of livestock, the higher the density, which leads to a lower survival rate for palatable herbs. This, in turn, increases the chances of toxic grass growth and decreases the overall quality of the grassland.

3.2.2. Effect of the Grazing Path on GVC

In order to understand the effect of grazing paths on grassland degradation, a linear equation (Figure 7) was established between grazing path density and GVC for all samples.

The GVC among all samples exhibited a wide range, varying from 23% to 71%, and the vegetation growth was highly variable. There was a strong linear relationship (R² = 0.65) between grazing path density and GVC. The higher the grazing path density, the more barren the land and the smaller the GVC. This finding further supports the notion that grazing paths play an important role in affecting vegetation cover and contributing to grassland degradation.

3.3. Effect of the Grazing Path on Malus sieversii

3.3.1. Changes in Radial Growth of Malus sieversii

Livestock affect the annual growth of trees, mainly when the trees are young. When trees grow to a certain height, livestock are unable to nibble on young leaves and branches, so the trees grow further at diameter breast height (DBH) [40]. To analyze the changes in radial growth with tree age, radial growth and relative tree age were correlated for the Malus sieversii sample trees (Figure 8).

The results showed that the pattern of variation in DBH with age tended to be consistent among the sixty Malus sieversii sample trees (Figure 8). The DBH was about 1.79 mm below four years old, with a clear trend of increasing radial growth from four to six years old (Figure 8, red rectangular area), and about 2.47 mm after six years old. During the initial four years, Malus sieversii primarily focuses on height growth to move out of the livestock’s grazing range as quickly as possible. Once the height surpasses 2 m, it becomes challenging for livestock to reach the leaves or branches, as livestock in the Zollersay Mountains area, mainly cattle and sheep, have standard heights of 1.44 m and 0.9 m, respectively. Thus, the likelihood of Malus sieversii being nibbled by livestock is closely tied to changes in radial growth.

3.3.2. Relationship between the Grazing Path and Radial Growth of Malus sieversii

Building upon the observed changes in radial growth discussed in Section 3.3.1, a linear regression relationship was established between the radial growth of sixty Malus sieversii and the grazing path density of the samples where the trees were located (Figure 9).

The radial growth of Malus sieversii showed a linear relationship with the grazing path density (R² = 0.70). Greater grazing path density corresponded to smaller radial growth in Malus sieversii. Grazing path density partly reflects the frequency of foraging and trampling by livestock. In areas with a high livestock population and limited forage availability, livestock tend to feed more on the young leaves or branches of young wild Malus sieversii, thereby impacting its radial growth.

4. Discussion

4.1. Evolution of Grazing Path Characteristics

The length, width, formation, and density of the grazing path could visually reflect the trampling route of livestock during foraging. Field observations revealed that the livestock are dominated by large animals, such as cattle and horses, in the Zollersay Mountain area. The width of the grazing path is higher than that in the upper reaches of the Minjiang River as well as the Loess Plateau in China, where sheep are the main breeding animals [12,41], but also in the California region of the USA, where cattle and horses are the main breeding animals (40 cm) [13]. The tortuosity of the grazing path ranged from 0.89 to 1.74, which was lower than in the scrub [12], mainly on account of the fact that livestock did not have to avoid the obstruction of their feeding by the scrub on the grassland. The shape of the grazing path developed from a parallel to an oblique intersection and finally evolved into a grid, caused by the gradual increase in grazing intensity. Factors such as long-term feeding and trampling, rainfall, and surface runoff contribute to the strengthening of grazing paths, enabling their reasonable use by livestock during the feeding process. Therefore, through interactions with the surrounding environment, livestock play a role in the formation of grazing path networks during foraging. The process of forming grazing path morphological features and distribution patterns is complex [11], and tracing the process of grazing path morphological development to capture the grazing path of livestock movement is a worthwhile direction of study.

4.2. Mechanisms of Grazing Path Action on Montane Vegetation

Our results showed that the IGQ was meaningfully correlated with the density of grazing paths and that this can be solved with the help of livestock optimal foraging theory [42]. Grazing path density influenced the distribution of vegetation, favoring short, trample-tolerant plants with a high capacity for compensatory growth, as they are more likely to survive and reproduce under high-intensity foraging and trampling. A higher grazing path density leads to more pronounced differences in herbaceous density and increased herbaceous heterogeneity. Livestock’s trampling behavior influences the formation of grazing path combinations, often resulting in a grid-shaped pattern and a decrease in plant cover. These findings are consistent with the perspectives of many scholars [24,25,26,27,28]. Further investigation of grazing path dynamics over time could provide more reliable evidence to support this conclusion.

Grazing paths not only contribute to soil erosion and grassland degradation but also directly damage trees [43]. Trees with a height lower than two meters have an injury rate of approximately 90%. Livestock directly damage the trees through foraging, trampling, and horn play. Young trees are more susceptible to livestock damage and can experience limitations in their early radial growth. However, once grazing ceases, they can continue to grow [30]. On the one hand, as livestock, particularly cattle, forage on Malus sieversii and excrete cow dung containing encased seeds, this process increases germination and promotes healthy seedling growth in the Zollesay Mountains. On the other hand, there are few surviving seedlings, for example, live seedlings and sprouting tiller seedlings of Malus sieversii [35], because many young seedlings are stunted by livestock feeding and trampling, affecting their formation, radial growth, and height development [44]. The severity of livestock damage to young Malus sieversii seedlings is higher in areas with high grazing path density. While grazing behavior can help expand the growth range of Malus sieversii trees, overgrazing remains a primary factor contributing to the reduction in young Malus sieversii saplings [34]. Further in-depth studies are needed to explore how to scientifically regulate grazing intensity and maximize the positive effects of grazing on vegetation turnover.

4.3. Remodeling Effect of Grazing Paths on Mountainous Terrain

It is worth noting that grazing paths influence the composition and spatial distribution of mountain vegetation. When the vegetation cover in mountain ecosystems decreases, the exposed surface becomes susceptible to erosion and transforms into bare ground, resulting in reduced water retention capacity [45]. Under high-intensity rainfall, the combined effect of livestock trampling and rainfall scouring increases the likelihood of geological hazards such as landslides [46]. During the field survey conducted on Zollersay Mountain, 21 obvious landslides were observed (Figure 10). The size of the collapse area, the level of livestock foraging, and the density of grazing paths showed a positive relationship. Additionally, grazing paths serve as important channels for surface runoff in the Zollersay Mountains. On rainy days, livestock foraging and trampling churn the soil into mud, reducing percolation rates, accelerating surface erosion, increasing surface runoff, intensifying soil erosion, and disrupting the mountain ecosystem [47].

Although we employed an object-oriented remote sensing classification algorithm in our experiments to partially extract grazing paths using UAV imagery, the extracted grazing paths were mostly discontinuous. Therefore, we further refined the grazing paths using ArcGIS, a method that achieved an accuracy rate exceeding 90%. It is important to note that the UAV imagery was acquired in July, during the summer months when herbaceous plants were growing relatively abundantly and tended to mask some grazing paths (Figure 11). In addition, we acknowledge that the relatively short time span of drone collection and ground surveys in our study area may not fully capture the long-term distribution pattern of grazing paths.

Thus, our future work will primarily focus on continuously monitoring the dynamics of grazing paths and vegetation. We aim to quantitatively assess the correlation between grazing paths and landslides, acquire UAV imagery from different time phases, and employ remote sensing algorithms and other possible methods for automated extraction of grazing paths. These tasks will significantly contribute to our understanding of grazing path dynamics and their implications.

5. Conclusions

In this study, we utilized high-precision UAV imagery and conducted field surveys in the traditional grazing area of the Tien-Shan Mountains in Xinjiang, China. Our aim was to investigate the number, formation, and distribution characteristics of grazing paths and assess their effects on mountain vegetation, including grassland and Malus sieversii. The findings of our study revealed several key insights. Firstly, we observed that grazing paths at the hilltop were shorter and narrower, while those on the hillside exhibited the largest number and the greatest amplitude of grazing paths. This indicates that the hillside serves as a hotspot for grazing path distribution. The formation of the grazing path tended to evolve towards a grid shape on account of the frequent trampling behavior of livestock. Moreover, grazing path density emerged as an important factor influencing IGQ, with the highest density observed on hillsides, where grassland degradation was most severe. Furthermore, grazing behavior led to a greater degree of damage to young Malus sieversii, with the radial growth of trees under four years old being the most affected by grazing paths. In the future, it is recommended to explore the great potential of remote sensing classification techniques based on UAVs for monitoring grazing paths. Additionally, it is crucial to regulate grazing intensity in a reasonable manner and strengthen monitoring of geological hazards to mitigate the negative impact of overgrazing on mountain vegetation and geology. By undertaking these measures, we can contribute not only to the sustainable management of the livestock industry but also to the conservation of the wild fruit forests in the Tien-Shan Mountains in China.