EN中文

Journal of Geographical Sciences >

2023 , Vol. 33 >Issue 8: 1660 - 1680

DOI: https://doi.org/10.1007/s11442-023-2147-1

Special Issue: Human-environment interactions and Ecosystems

Understanding the spatial heterogeneity of grazing pressure in the Three-River-Source Region on the Tibetan Plateau

  • GU Changjun , 1, 2, 3 ,
  • LIU Linshan 1 ,
  • ZHANG Yili , 1, 2, 3, * ,
  • WEI Bo 1, 2 ,
  • CUI Bohao 1, 2 ,
  • GONG Dianqing 1, 2
Expand
  • 1. Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • 2. College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China
  • 3. National Disaster Reduction Center of China, Ministry of Emergency Management, Beijing 100124, China
* Zhang Yili (1962-), Professor, E-mail:

Gu Changjun (1992-), specialized in grassland change, climate change, and grazing management. E-mail:

Received date: 2022-07-05

  Accepted date: 2023-02-23

  Online published: 2023-08-29

Supported by

The Second Tibetan Plateau Scientific Expedition and Research(2019QZKK0603)

Strategic Priority Research Program of the Chinese Academy of Sciences(XDA20040201)

National Natural Science Foundation of China(41671104)

Fold

Abstract

Elucidating the distribution of the grazing pressure requires an understanding of the grazing activities. In this study, we analyzed the grazing behavior of yaks in Three-River- Source Region (TRSR) and identified the main factors influencing the distribution of grazing intensity (GI) using trajectory data and remote sensing datasets. Our results revealed that a semi-resident transhumance strategy is employed in this region. The average grazing time (GT) of four GPS collars over the year was 11.84 h/day (N6), 11.01 h/day (N11), 9.25 h/day (N18), and 11.61 h/day (N24). GT was generally higher in warm seasons (summer and autumn) than in cold seasons (spring and winter). The average daily moving speed was found to be closely related to the pasture size of different herders and the seasons. Geodetector analysis identified the distance to camp (DOC) as the most important single factor influencing the distribution of GI, explaining up to 52% of the GI variations. However, relying solely on this factor may not accurately depict the actual GI distribution. When pairwise factors interacted, the explanatory power of the model increased, ranging from 34.55% to 63.26%. Our study highlights the importance of considering multiple factors when predicting grazing intensity, as grazing activities tend to cluster near settlements, but other factors may also be influential.

Key words： grazing intensity; GPS-tracking; spatial heterogeneity; yak; Tibetan Plateau

Cite this article

GU Changjun , LIU Linshan , ZHANG Yili , WEI Bo , CUI Bohao , GONG Dianqing . Understanding the spatial heterogeneity of grazing pressure in the Three-River-Source Region on the Tibetan Plateau[J]. Journal of Geographical Sciences, 2023 , 33(8) : 1660 -1680 . DOI: 10.1007/s11442-023-2147-1

1 Introduction

Grasslands, which cover about a quarter of the global land surface, harbor roughly 10% of terrestrial biomass and contribute 20%-30% to the global pool of soil organic carbon (SOC) (Scurlock et al., 1998; Conant et al., 2001). Grasslands also provide 30%-50% of the global livestock product (Chen et al., 2000) and account for about 70% of the agricultural area (Conant, 2012). As such, grasslands play a crucial role in the global ecosystem and social economy system, providing multiple ecosystem services (Zhang et al., 2014; Obermeier et al., 2019). However, grassland degradation has become a global ecological problem over the past few decades due to the dual effects of climate change and intensifying human activities (Melillo et al., 1993; Gao et al., 2014; Seddon et al., 2016; Li et al., 2018a). Grassland degradation poses a significant threat to ecosystem functions (Lehnert et al., 2014), exacerbates soil degradation (Li et al., 2014), and jeopardizes the livelihood of local people (Harris, 2010). While the causes of grassland degradation are still debated, overgrazing is widely recognized as one of the primary drivers (Abdalla et al., 2018; Li et al., 2018b).
Previous research has shown that grazing can significantly impact various aspects of grassland, including phenology (Li et al., 2019), grassland productivity (Venter et al., 2019), grass species composition (Cingolani et al., 2005), and soil physicochemical properties (Du et al., 2019). Both overgrazing and undergrazing may cause unwanted environmental problems (Follett et al., 2000). Undergrazing, for instance, can increase dry biomass and raise the risk of wildfires (Van Auken, 2000; Dubinin et al., 2011). Therefore, obtaining the spatial explicit distribution of grazing pressure is crucial for understanding the impacts of grazing on grassland and for making informed management decisions.
Grazing activity is very complicated, and a massive effort has been initiated in the quantification of grazing pressure, including field surveys, remote sensing skills (Feng et al., 2011; Yang et al., 2012), and land use modeling (Robinson et al., 2014; Hankerson et al., 2019). However, these methods may not accurately capture the spatial distribution of grazing animals (Dara et al., 2020; Gu et al., 2023). Statistical and remote-sensing-based techniques often assume that grazing activities are evenly distributed based on features such as the distance from rivers or settlements (Coppolillo, 2001; Homewood et al., 2004). This assumption indicates that grazing pressure is uniformly distributed over a certain distance from a river or settlement (Liao et al., 2017). Nevertheless, the actual situation is much more complex. Research has shown that though the distance from campsites was the best proxy for grazing intensity, it can only explain the 30% of its variation (Dorji et al., 2013). Fortunately, advances in GPS-tracking technology and spatial analysis tools have enabled researchers to better understand the actual distribution and spatial preferences of grazing activities (Anderson et al., 2012).
GPS devices have been widely to monitor grazing animals and to quantify grazing behavior and locations in previous research (Ungar et al., 2011; Augustine et al., 2013). Rutter et al. (1997) tracked domestic sheep by GPS to identify their home ranges. Kawamura et al. (2005) combined GPS, GIS, and remote sensing techniques to quantify grazing intensity in the Xilingol steppe region of Inner Mongolia, China, by rasterizing the study area and counting GPS points in a single pixel after a short grazing period. Brennan et al. (2021) classified season-long livestock grazing behavior using a low-cost GPS and accelerometer. Liao et al. (2017) used GPS tracking to reveal the complexity of spatial utilization of rangelands in the Horn of Africa, providing valuable information on livestock movement and selection, grassland use, and changes in forage preference over time. Overall, GPS tracking has great potential for obtaining grazing locations, grazing selection, and understanding changes in forage preference over time.
The Three-River-Source Region (TRSR), in the hinterland of the Tibetan Plateau (TP), is the birthplace of the Yangtze, the Yellow, and the Lancang rivers (Han et al., 2018; Zhang et al., 2019). The region’s dominant land cover type is alpine grassland, which covers more than 80% of the TRSR area and provides essential ecosystem services to both the local community and downstream regions (Qian et al., 2010). However, this ecosystem is highly vulnerable to external disturbances due to the harsh environmental conditions prevailing in the region (Sun et al., 2016; Yang et al., 2018; Bai et al., 2020). Grassland degradation caused by overgrazing is a pressing issue in the TRSR (Zhang et al., 2006; Liu et al., 2008). Although previous research has explored the relationship between grazing pressure and alpine grassland ecosystems, the lack of spatially explicit grazing pressure data limits our understanding of this relationship, particularly at a large scale. Therefore, identifying the primary factors that determine the spatial heterogeneity of grazing pressure in the TRSR is crucial.
This research aims to investigate the behavioral preferences of yaks, which are the primary grazing livestock in the TP, and to identify the main factors that determine their spatial utilization patterns. To achieve this, portable GPS collars were used to record the daily grazing activities of yaks in the TRSR. The specific objectives of this research are: 1) to quantify the grazing behavior characteristics of yaks, including their grazing time, moving speed, grazing ranges, and grazing preferences; 2) to identify and analyze the main factors that influence the livestock behavior patterns in this region.

2 Methods

2.1 Study area

The Three-River-Source Region (TRSR) (31°39′-36°12′N, 89°45′-102°23′E) is located in the center of the Tibetan Plateau (TP) (Figure 1a), covering about 15% of the TP and with an average elevation of 4200 m (Wang et al., 2016). As the name suggests, TRSR is the origin of the Yangtze, Yellow, and Lancang rivers, and is commonly referred to as the “Chinese Water Tower” (Ma et al., 2020). The region experiences a typical continental plateau climate with intense radiation, low temperatures, and significant seasonal temperature variation (Wang et al., 2016). The average annual temperature in the region ranges from -5.6 to 3.8℃, and the mean annual rainfall ranges from 262.2 to 772.8 mm (Ma et al., 2020). Generally, the temperature and precipitation decrease gradually from southeast to northwest in TRSR. The grassland in the region is primarily composed of alpine grassland and alpine steppe (Figure 1b), which are essential components of the alpine ecosystem (Xu et al., 2018).
Figure 1 Three-River-Source Region (TRSR) is located in central Tibetan Plateau (TP) (Zhang et al., 2021) and has a first national park in China and also the biggest natural reserve in China (a). As a typical alpine ecosystem, TRSR is characterized by alpine grassland and alpine meadow (b). The Chinese government has implemented ecological conservation and restoration project in this region. Since the beginning of the 21st century, the greenness of vegetation has maintained positive improvement generally but decreased locally (c). Four GPS tracking devices were set in Maqin and Dari county (d).
The Chinese government has taken measures to protect the ecological integrity of the TRSR, including initiating first-stage and second-stage ecological conservation and restoration projects in 2005 and 2014, respectively, and establishing the nature reserve and national park (Shao et al., 2016). Even though the ecosystem in this region has been improving, grassland degradation situations were not fundamentally reversed (Shao et al., 2016) (Figure 1c). Grazing is the main use of the grassland and a crucial source of livelihood for local communities, who practice a semi-nomadic grazing model utilizing sheep, yaks, and horses as their primary livestock (Fan et al., 2010). Due to the limited productivity of the alpine grassland, most herders drive their animals seasonally between summer and winter pastures (Fan et al., 2010). Winter pastures, which account for about 40% of the grazed area of TRSR, are located in warmer, sheltered, and sunnier locations closer to the settled residences of herders and are subjected to higher grazing pressure as they are used for a more extended period than summer pastures. Overgrazing was regarded as the important driver causing the grassland degradation and thus this research set four GPS tracking devices in different vegetation greenness change types (Figure 1d).

2.2 GPS tracking and experimental design

The challenge we often face with GPS tracking is the trade-off between the battery life of GPS devices and the extensive movement behaviors of grazing livestock. In addition, obtaining records of grazing activities for a year or more typically requires significant effort from researchers. To address these issues, we carefully selected GPS devices that are lightweight and have a long battery life. Specifically, we used Shenzhou-VI GPS Trackers (Geman E-Commerce co., Ltd, Shenzhen, China) (Figure 2a), which have large capacity lithium batteries (12000 milliamperes) and can operate for over two weeks without charging when set to log at 10-minute intervals. Furthermore, these devices are relatively affordable, costing only 498 yuan (≈77 USD) and weighing less than 0.1% of the body mass of a mature yak in TRSR (around 300 g) (Figure 2b). To capture more detailed information on the grazing activities of yaks, we set the GPS logger to collect a fix (latitude/longitude) at 1-minute intervals. Over the course of our study (August 2020 to September 2021), we collected a total of 782,527 GPS tracking records of grazing yaks, including coordinates, timestamp, GPS location, moving speed, and direction (Table 1).
Figure 2 Diagram of GPS collar used in this study (a): ① A Nylon collar; ② A screwdriver; ③ Waterproof box; ④⑤ USB cable; ⑥ Instructions; ⑦ GPS tracker; ⑧ Packing box; Selected yak with GPS collar in Dari county, located in central TRSR (b)
Table 1 List of GPS location records of grazing yaks during the past year (August 2020 to September 2021)
Month N6 N11 N18 N24
1 16724 14802 6535 17694
2 20826 16521 15066 16944
3 16097 16814 16008 17919
4 14956 13081 16736 14977
5 24786 15124 24575 19966
6 19404 14783 20521 21365
7 18598 17251 12824 18627
8 13130 22483 17355 21545
9 10021 20469 13910 17042
10 7909 8894 17366 18087
11 2945 11916 17280 15958
12 17458 14679 18456 16100
Total 182854 186817 196632 216224

2.3 Preprocess of the original data

GPS signals can be lost due to weather or terrain conditions, resulting in the absence of location information. In such cases, the recorded location may be hundreds of kilometers away from the actual location when the GPS receiver receives the signal again. Therefore, the abnormal value needs to be removed. Additionally, there may be a deviation between the recorded and real sites due to the different coordinate systems used in the GPS collar. To rectify this deviation, a python package was applied in this research (Figure S1).
It is observed that grazing activities always begin and end at fences. Therefore, we defined a grazing time (GT) term as the time difference between the start and end of grazing activities in one day. Locations within the fences were dropped from the dataset using information recorded by one of the four herders throughout the year (Figure S2 and Table 2). Significant differences were observed in GT in different seasons. During summer, grazing activities started at an average of 8:08 a.m. and ended at an average of 8:48 p.m., resulting in the longest GT. However, during winter, yaks come out quite late and come back early due to limited grassland resources and harsh climate conditions. Therefore, two threshold values of 6:00 a.m. and 9:00 p.m. were set, respectively. After preprocessing, 42% of the original data was dropped.
$GT=Grazin{{g}_{end}}-Grazin{{g}_{start}}$
where GT denotes the grazing time throughout a day, Grazingend refers to the time at which the grazing activities end, and Grazingstart denotes the time at which the yaks are let out of the fence to graze.
Table 2 A detail time table of grazing activities across the four seasons
Spring Summer Autumn Winter
Avg Min Max Avg Min Max Avg Min Max Avg Min Max
Start 10:16 6:38 17:12 8:08 5:58 14:27 9:40 7:43 12:26 11:26 10:29 13:57
End 19:52 11:21 22:39 20:48 19:18 23:05 19:55 18:02 21:19 18:33 11:00 20:13
GT 9:36 12:40 10:15 7:07

2.4 Identifying the behavior characteristics of grazing yak

Animal behavior is critical for understanding their foraging strategy and resource utilization within their habitat (Brennan et al., 2021). Various parameters such as moving speed, direction, distance, duration of stay, and home range are commonly used to study animal behavior. In this research, we employed three indices to describe the behavior of yaks, namely, moving speed, moving distance, and home range. The GPS collar’s velocity transducer collected the moving speed information, and we used MovingPandas (available at: https://anitagraser.github.io/movingpandas/) (Graser, 2019), a python package for movement data analysis, to obtain the daily moving distance and home range information.
This study employed a method similar to the one described by (Kawamura et al., 2005) to quantify the grazing intensity (GI). A 30 m × 30 m grid was created using ArcGIS (Ver10.8) based on the boundary of GPS location data. The grid layer was overlaid with the location data to count the number of points in each grid, as shown in Figure S3. To investigate the main factors influencing the distribution of GI, we collected several environmental factors (Table 3) based on previous studies (Rivero et al., 2021). All the environmental factors had a resolution of 30 m × 30 m. To account for seasonal differences in vegetation greenness, we used monthly Landsat NDVI maximum value composite instead of an annual maximum value composite. To explore the factors that explain the distribution of GI and the interactions between pairwise environment factors, we used the Geodetector model (GDM) (available at: http://www.geodetector.cn/). GDM is a statistical method that can detect stratified spatial heterogeneity and reveal the driving factors behind it (Wang et al., 2017). The q-statistic of the factor detector is used to measure the contribution of each independent variable, assuming that the high-influence independent variable should share a similar spatial distribution pattern with the dependent variable (Wang et al., 2017). The formula is as follows:
$q=1-\frac{1}{N{{\sigma }^{2}}}\underset{h=1}{\overset{m}{\mathop \sum }}\,{{N}_{h}}\sigma _{h}^{2}$
where h(1, …, m) denotes the number of factors; Nh indicates the number of samples in subregion k; N represents the total number of spatial units across the whole study area; σ2 and σh2 represent the global variance in the study area, and variance in the samples in subregion k, respectively. A greater q-statistic value (range from 0 to 1) indicates higher explanatory power of the selected independent variable. Furthermore, GDM provides another module to detect whether the explanatory powers of two independent variables are enhanced, weakened, or separated (Wang et al., 2017).
Table 3 Potential environmental factors influencing the distribution of grazing intensity
Data source Resolution Year
Elevation NASA SRTM Digital Elevation 30 m, https://cmr.earthdata.nasa.gov/search 30 m×30 m 2000
Slope Obtained from DEM 30 m×30 m 2000
Aspects Obtained from DEM 30 m×30 m 2000
NDVI Landsat 8 Collection 1 Tier 1 32-Day NDVI Composite, https://earthengine.google.com/ 30 m×30 m 2016-2021
(Monthly average)
Distance to camp (DOC) Digitization of google earth images ——
Distance to road (DOD) Digitization of google earth images ——
Distance to river (DOR) Digitization of google earth images ——
Month —— ——

3 Results

3.1 Moving patterns of grazing yaks

Despite the short distances between the grazing camps, there are distinct differences in the grassland use patterns among the four households (Figure 3). The most noticeable feature is that the points collected from GPS collars indicate non-overlapping grazing areas, which suggests that rotational grazing is a common practice. However, specific rotation grazing strategies vary among the four households. For example, in collar N6, the herdsman migrated three times in one year. From January to February, the yaks grazed in the first camp, which was located close to the main road. The home range was relatively small and close to the campsite (around 3 km). In March (2021/03/14), the herdsman would migrate to the second camp for about two months and then return to the first campsite for another four months. On September 19, the herdsman would continue to migrate to the third camp for about two months, and then return to the first camp at the end of November. In contrast, the grazing patterns in N11 differed, as the herdsmen did not move to another location during the year. However, the grazing direction varied across the different months. By the end of April, the herdsmen were grazing the yaks to the north of the road, while in May, most of the grazing activities occurred on the south side of the road. Collars N18 and N24 had the same grazing strategy.
Figure 3 Spatial patterns of GPS tracks during the whole year (August 2020 to September 2021) in four locations
Significant differences in Grazing Time (GT) between different seasons and months can be observed (Figure 4). The average GT values of four GPS collars are 11.84 h/day (N6), 11.01 h/day (N11), 9.25 h/day (N18), and 11.61 h/day (N24), respectively. From the view of different seasons, changes in GT show a similar single peak curve from spring to winter. GT in spring (February, March, and April) and winter (November, December, and January) are lower than that in summer (Mayday, Jun, and July) and autumn (August, September, and October). Take collar N24 as an example, the average GT values in summer and autumn are 13.20 h/day and 12.11 h/day, respectively. While in spring and winter, these values are 11.08 h/day and 10.08 h/day, respectively. The maximum average GT of all GPS collars occurs in summer except for collar N6 (in autumn). From a monthly perspective, changes in GT show a similar trend as seasonal changes. The minimum average GT of four GPS collars occurs in November (N6, 5.8 h/day), December (N11, 9.07 h/day), January (N18, 4.99 h/day), and December (N24, 9.94 h/day), while the maximum average GT of four GPS collars occurs in July (N6, 13.70 h/day), July (N11, 11.67 h/day), July (N18, 13.03 h/day), and June (N24, 13.60 h/day).
Figure 4 Grazing time (GT) in different seasons and months of four GPS collars during the past whole year (August 2020 to September 2021)
The data from the four GPS collars showed significant differences in average moving speed across different seasons and months (Figure 5). GPS collar N6 had the highest average moving speed of 2.04 km/h, while GPS collar N24 had the lowest average moving speed of 0.86 km/h. GPS collars N11 and N18 had average moving speeds of 1.46 km/h and 1.48 km/h, respectively. Generally, yaks moved faster in autumn and winter than in summer and spring. As depicted in Figure 5, the highest average moving speed was observed in autumn (2.46 km/h in N6 and 1.08 km/h in N24) and winter (1.70 km/h in N11 and 1.75 km/h in N18). On the other hand, the lowest average moving speed occurred in summer (1.19 km/h in N11 and 1.08 km/h in N18) and spring (1.80 km/h in N6), except for N24. A similar trend was found in GPS collars N6 and N24, while GPS collars N11 and N18 had a similar trend. The lowest average moving speed in GPS N6 and GPS N24 was 1.33 km/h in February and 0.65 km/h in January. The lowest average moving speeds in GPS N11 and GPS N18 were 1.01 km/h and 0.99 km/h, respectively, in May.
Figure 5 Moving speed of yaks in four GPS collars in different seasons and months
From the view of the moving distance and home range of yaks (Figure 6), the average daily moving distance and home range of four GPS collars are 7.01 km and 168.84 ha in GPS N6, 10.59 km and 634.21 ha in GPS N11, 10.87 km and 566.84 ha in GPS N18, and 7.82 km and 241.89 ha in GPS N24, respectively. In GPS N6 and N11, the smallest average daily home range was observed in summer, while the most extensive average daily home range occurred in winter. Meanwhile, the longest average moving distance occurred in summer (N6) and autumn (N11). On the other hand, the largest average daily home range of N6 and N11 was found in winter, while the shortest average daily moving distances of N6 and N11 were found in spring. In GPS N18 and N24, the average daily home range and moving distance were lower than in GPS N6 and N11. The most extensive average daily home range and the longest average daily moving distance of GPS N6 found in spring. Meanwhile, the smallest average daily home range of GPS N6 was found in autumn, and the shortest average daily moving distance of GPS N6 was in winter. While in GPS N24, the largest average daily home range and the longest average daily moving distance were observed in autumn and summer, respectively. Meanwhile, the smallest average daily home range and the shortest average daily moving distance of GPS N24 were observed in winter.
Figure 6 Average daily moving distance and home range of yaks in different seasons and months

3.2 Main factors determining the distribution of grazing intensity

The distribution of grazing intensity (GI) among the four GPS collars exhibited a notable difference (Figure 7). In the case of GPS N6, the high GI values were concentrated primarily near the first and second camps but less so in the third camp. This indicates that grazing activities in the first two camps lasted longer than in the third camp. On the other hand, the other three GPS collars exhibited similar distribution patterns, with grazing activities revolving around the only camp and being limited by the boundary of the pasture. Using Geodetector, this study analyzed the primary factors that influence the distribution of GI.
Figure 7 Distribution of grazing intensity of four GPS collars
The factor detector analysis revealed the main factors that influence the GI distribution, as ranked by their q-statistics values in Table 4. In GPS N6, the elevation factor had the greatest impact on the GI distribution with a q-value of 0.16, followed by factor DOD with a q-value of 0.15. Other factors with q-values greater than 0.1 included DOC and month. Conversely, in the other three GPS collars, DOC was the most significant factor contributing to GI distribution, with q-values of 0.19 (N11), 0.20 (N18), and 0.52 (N24). However, one single factor can explain less GI distribution variation. Even if the DOC in N24 is the best proxy for GI distribution, it can only describe the 52% variation in GI distribution. These findings suggest that multiple factors work together to control GI distribution.
Table 4 The q-statistics value of seven main influencing factors of grazing intensity
q-statistics Elevation Slope Aspect NDVI Distance to camp (DOC) Distance to road (DOD) Distance to river (DOR) Month
N6 0.16 0.02 0.02 0.09 0.12 0.15 0.01 0.34
N11 0.10 0.04 0.01 0.03 0.19 0.07 0.04 0.05
N18 0.02 0.01 0 0.02 0.20 0.08 0.03 0.05
N24 0.03 0.04 0.02 0.09 0.52 0.13 0.04 0.25
The interaction detector was further utilized to identify the interactive effects of all pairwise factors on GI distribution. As depicted in Figure 8, the interactive q-statistic of the pairwise factors were higher than those of single factor, indicating that the two factors are mutually enhanced. Compared to a single factor, the explanatory power of GI distribution can be improved when interacting with the other influencing factors. In GPS N6, the maximum q-statistic (0.32) occurred when elevation interacted with DOC. Notably, the q-statistic of DOC-DOD interaction was also greater than 0.3. While in GPS N11, the interaction between DOC and DOR generated the maximum q-value, with a q-value of 0.35. However, the q-statistic of the other pairwise factors interactions was lower than 0.3, except for the month-DOC interaction. In GPS N18, the maximum q-value appeared in the interaction between DOR and DOC, with a q-value of 0.34. The interactions of DOC-month and DOC-slope were 0.32 and 0.30, respectively. GPS 24 showed similar interaction features with GPS N11, and the interaction between DOC and DOR generated the maximum q-value (0.63). In GPS N11, N18, and N24, the interactions between DOC and NDVI, slope, elevation, and month were generally stronger than other pairwise factors.
Figure 8 Interactions (measured by q value) between pairs of factors on grazing intensity distribution. DOC, DOR, and DOD represent distance to camp, distance to river, and distance to road, respectively.

3.3 Significant differences between pairwise factors and the optimal range

According to the ecological detector, this study further explored whether there were significant differences in the pairwise factors’ impacts on GI distribution. As depicted in Figure 9, significant differences existed in pairwise factors influencing GI distribution, which indicated a significant difference in effects on GI distribution among the factors selected in this research.
Figure 9 Statistical significance of the pairwise factors influences on grazing intensity distribution (95% confidence level). DOC, DOR, and DOD represent distance to camp, distance to river, and distance to road, respectively. Y indicates that there are significant differences existed in the effects of the pairwise factors on grazing intensity distribution.
Based on the risk detector, the optimal range of eight factors corresponding to the maximum GI was calculated in Table 5. According to the results of the single-factor detector, GI varied across different months. For GPS N6, the maximum GI occurred in January, while the minimum GI was observed in December. Similarly, GPS N18 and N24 show the highest GI in January and the lowest in December, while GPS N11 achieves the maximum GI in June and the minimum in December. DOC was also the critical factor influencing the GI distribution. We found that GI decreased as the distance from camp increased. In GPS N6, the maximum GI was observed from 12.3 m to 270 m, while the minimum was observed when the DOC was higher than 1.29 km. While in GPS N11, the GI reached its maximum at the DOC range of 40.8 m to 87.2 m. Meanwhile, the minimum GI occurred in the DOC range of 1.75 km to 3.71 km. In GPS N18, the maximum GI occurred in the DOC range of 28.9 m to 65.9 m, while the minimum GI was observed in the DOC range of 748 m to 1.68 km. Finally, in GPS N24, the maximum GI was not found within the nearest DOC range but within the second nearest DOC range (49.8 m to 112 m), while the minimum GI was observed within the DOC range of 1.23 km to 2.73 km.
Table 5 Corresponding range of eight factors to maximum grazing intensity
Elevation (m) Slope (°) Aspect NDVI DOC (m) DOR (m) DOD (m) Month
N6 4000-4030 3.68-7.81 20-73.7 0.16-0.24 12.3-270 653-816 42.7-138 1
N11 4080-4130 1.85-2.9 178-237 0.41-0.55 40.8-87.2 0.04-44.9 0.07-321 6
N18 4258-4261 3.57-4.48 182-243 0.08-0.17 28.9-65.9 252-336 0.07-3.58 1
N24 4033-4041 4.98-9.25 211-236 0.16-0.26 49.8-112 112-194 40.4-107 1

Note: DOC, DOR, and DOD represent the distance to camp, the river, and the road, respectively.

4 Discussion

4.1 Reliability of using GPS collars to monitor the grazing activities in the Three- River-Source Region (TRSR)

Since the early 1990s, numerous studies have explored the usability of wearable GPS devices to monitor the grazing system (Rodgers, 2001; Rivero et al., 2021). Exponential improvement in tracking technology allows us to get even closer to the real but complicated world (Kays et al., 2015). Discarding the traditional method based on the manual record helped to lower the cost. Rutter et al. (1997) developed a GPS-based animal behavior and tracking system to monitor the grazing areas of hill sheep on a hill sheep farm in upper Eskdale, West Cumbria, UK. Their results demonstrated that GPS could be used to track domestic sheep. A further study carried out by Manning et al. (2017b) showed that the presence of GNSS collars did not modify any cow behaviors. This study analyzed the grazing activities in Three-River-Source Region (TRSR), including grazing time, grazing speed, and grazing home range based on low-cost GPS collars. Previous research has argued that the grazing period’s duration essentially determines forage availability and quality (Fetzel et al., 2017). Furthermore, grazing time variations significantly influence Tan sheep’s behavior on desert steppe (Chen et al., 2013). They would adjust the time distribution among intake, bite rate, grazing velocity, resting, and meditating time to improve grazing efficiency. In terms of grazing time (GT), the GT is generally higher in summer and autumn but shorter in spring and winter (Figure 3). As the saying goes—yaks are “alive in summer, strong in autumn, thin in winter, and tired in spring” (Dong et al., 2003). A previous study stated that yaks increase their body mass in summer and decline in body mass in winter (Liu et al., 2019). For this purpose, herders would increase the GT properly in summer because of the better quality of grassland and climate conditions, a widely adopted strategy for herders to achieve maximum grassland utilization (Long et al., 2008).
Moving speed is another crucial animal behavior that affects all behaviors and underlies activity intensity (Wilson et al., 2015). According to the optimal foraging theory (OFT) (Brennan et al., 2021), a critical theory of the behavioral ecology model, animals adopt a foraging strategy that provides the most benefit (energy) for the lowest cost. Fast-moving behavior is typically associated with higher energy consumption (Hoyt et al., 1981) and is only adopted by yaks when necessary according to OFT. As depicted in Figure 5, the moving speed in GPS N6 and N24 are generally higher in summer and autumn but lower in spring and winter. A more extensive home range can explain this in summer and autumn (Figures 3 and 6). In summer and autumn, the moving distance of yaks in GPS N6 and N24 is longer than in spring and winter. However, yaks in GPS N11 and N18 shared a similar moving speed distribution pattern. Still, they conversed with the mentioned two (Figures 3 and 6). OFT cannot be applied to explain this kind of difference because of the uncertainty of whether animals adopt the current speed or external environmental constraints (Wilson et al., 2015). Limited by ownership of grassland, grazing activities here are not free-range grazing, the speed-choice is a complex trait driven by internal and external factors. The home range of yaks in different seasons can be used to outline the pasture size of four herders in this study. As depicted in Figure 6, a herder in GPS N11 holds the largest pasture size while the N24 occupies the smallest one.

4.2 Which factor can be the most explanatory proxy for grazing intensity?

The emergence of GPS tracking provided new opportunities for understanding how behavioral preferences relate to spatially constrained environmental factors (Swain et al., 2011). Nevertheless, most studies have focused on describing livestock distribution rather than exploring the driving forces behind it (Brennan et al., 2021). Due to the lack of animal preference details, traditional conceptual models assumed that GI is evenly distributed within a defined distance from the grazing camp, water location, or other grazing-related geographical features (Western, 1975; Homewood et al., 1991; Spencer, 2012). Previous research has identified water access as a crucial factor influencing animal preference for specific landscapes, particularly in arid regions (Putfarken et al., 2008; Tomkins et al., 2009; González et al., 2014). However, animal preferences for water sources can vary depending on the season and vegetation type. For instance, cattle tend to spend more time near water sources during the dry-cold season but not during the wet-warm season (Von Müller et al., 2017). This study found that the distance to the river (DOC) only explains about 4% of the GI variations, which may be because the alpine vegetation has a high-water content, resulting in yaks spending relatively less time drinking water.
Terrain conditions are essential environmental factors influencing grassland utilization and can be taken as an important index to predict livestock distribution (Ganskopp et al., 2009; Homburger et al., 2015). In this research, elevation was the second most important factor influencing the GI distribution in GPS N6 (16%) and N11 (10%), which can be explained by the rotation grazing strategy adopted by these two herders. In the warm season (summer and autumn), herder would pasture yaks to higher elevations, a little far from the permanent settlements. While in the cold season (winter and spring), yaks would spend more time in low winter pastures close to accommodations (Wiener, 2013). Nevertheless, elevation in GPS N18 and N24 can only explain 2% and 3% of the GI distribution variations, respectively. Presumably, grazing activities occurred in the region of smaller relief. Previous studies have demonstrated that livestock prefers areas with gentler terrain (Pittarello et al., 2016; Henkin et al., 2012). However, this study did not observe obvious preferences in slope (Table 4). The maximum q value is 0.04 in GPS N11, indicating the relatively weak explanation power. However, a significant trend can be observed that GI decrease with the increase of slope. Though this trend was not found in GPS N18 and N24, the minimum GI was also distributed in the maximum slope range. Furthermore, we found that the aspects play a small part in GI distribution, which can be explained by the grassland ownership, which indicates a resident or semi-resident transhumance instead of an entirely nomadic existence (Long et al., 2008).
Livestock is always clustered around settlements; thus, distance to settlement is a critical indicator for predicting the GI distribution (Coughenour et al., 2008; Hankerson et al., 2019). Distance to settlement was also applied to divided grazing gradients, the closer and the stronger (Li et al., 2008; Wang et al., 2017). This study also found that distance to camp (DOC) is the most important factor in explaining the distribution of grazing intensity, accounting for up to 52% of the variations observed. Akasbi et al. (2012) assessed goat grazing patterns and intensity in Southern Morocco and found that the highest grazing intensities were distributed in the 250 m nearest the settlement. Li et al. (2008) set the maximum sheep GI at a DOC of 0.6 km while the lowest GI at a DOC of 4.6 km. Hankerson et al. (2019) set the maximum search radius, representing the relative immobility of this livestock, as 2 km to spatialize the GI in Kazakhstan. in this study, the maximum grazing intensity was observed at a DOC ranging from 65.9 m (N18) to 270 m (N6) (Table 5), with grazing activities’ sphere of influence reaching a DOC range from 1.68 km (N18) to 3.71 km (N11). Thus, it is challenging to present a universal DOC for grazing activities.
Vegetation conditions varied across the different seasons and further influenced the choices made by herders and yaks. NDVI has been widely used for herbage biomass and nutritional quality (Zengeya et al., 2013; Manning et al., 2017a). According to the optimal foraging theory (OFT), also proved by existing studies, animals tend to spend more time at high forage allowance conditions (Homburger et al., 2015). Nevertheless, NDVI can only explain the largest 9% of the GI distribution (GPS N6 and N24). Moreover, with the increased NDVI, GI does not always increase simultaneously (Figures S4-S7). This may be due to the monthly variations in NDVI that we accounted for using monthly maximum value composite NDVI instead of the yearly maximum value composite NDVI. Interestingly, this study found that the largest GI occurred in January, except for GPS N11 (June), which is consistent with previous studies that grazing pressure is higher in winter pastures than in summer in the TRSR (Fan et al., 2010).

4.3 Interactions between pairwise factors

Our findings suggest that the distribution of grazing intensity (GI) is influenced by multiple factors, and a single factor alone cannot fully explain the actual distribution. While the distance to camp (DOC) is the best proxy of GI, it can explain at most 52% of variations of GI distribution, higher than the research of (Dorji et al., 2013), which reported 30%. In Section 3.3, significant differences existed in the influence of selected factors on GI distribution. Geodetector was further applied to explore the main factors influencing the GI distribution. We found that though the distance to camp (DOC) can only explain a part of the variation of GI distribution, this kind of explanation power can be enhanced when this factor interacts with the other factor, including month, distance to the river (DOR), and distance to road (DOD) (Figure 8). For instance, the interaction between DOR and DOC can explain up to 63% of the GI distribution, indicating that multifactorial interaction may better explain the distribution of GI.

4.4 Uncertainty and limitations of this study

Though we adopted a cooperative approach with herders to solve the battery capacity of GPS collars, a primary limiting factor (Augustine et al., 2013), we cannot fully overcome this problem. On the one hand, it is pretty hard to find ideal herders willing to cooperate and help us constantly charge the GPS collars, even though we paid them. Additionally, bad weather, terrain factors, or failure to charge the GPS collars on time can result in incomplete data, which can introduce uncertainty.
Our research showed that the grazing pattern in the study area is a semi-resident transhumance, rather than a completely nomadic lifestyle. This means that there are established corridors, as observed in our fieldwork (Figure 10). During this period, the yaks’ moving choice is no longer solely determined by them (Von Müller et al., 2017). Nevertheless, we keep the data records during this period. We believe that regular livestock traffic and trampling are also a critical part of GI, as proved by previous studies (Tefera et al., 2007; Dunne et al., 2011; Egeru et al., 2015). In this case, the GI of these areas would appear at a relatively high value and further influence the explanatory power of selected factors.
Figure 10 Routine corridors of grazing activities

5 Conclusion

This study provides clear evidence that grazing behavior varies significantly across different pastures, seasons, and months. The use of GPS tracking technology in conjunction with the cooperation of local herders proved to be an effective and cost-efficient method for obtaining continuous monitoring data. Despite longer grazing times during the warmer months, our findings show that the highest grazing intensity occurred in January across almost all GPS collars. Our analysis also revealed that a single factor, distance to camp (DOC), can explain up to 52% of the variation in grazing intensity, and this explanatory power can be further enhanced by considering interactions with other factors such as month, distance to the river (DOR), and distance to road (DOD). These results suggest that multifactorial interactions must be considered when spatializing grazing pressure, and a clear understanding of the contribution rate of each element is essential.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Abdalla M, Hastings A, Chadwick D R et al., 2018. Critical review of the impacts of grazing intensity on soil organic carbon storage and other soil quality indicators in extensively managed grasslands. Agriculture, Ecosystems & Environment, 253: 62-81.

[2]
Akasbi Z, Oldeland J, Dengler J et al., 2012. Analysis of GPS trajectories to assess goat grazing pattern and intensity in Southern Morocco. The Rangeland Journal, 34(4): 415-427.

DOI

[3]
Anderson D M, Winters C, Estell R E et al., 2012. Characterising the spatial and temporal activities of free-ranging cows from GPS data. The Rangeland Journal, 34(2): 149-161.

DOI

[4]
Augustine D J, Derner J D, 2013. Assessing herbivore foraging behavior with GPS collars in a semiarid grassland. Sensors, 13(3): 3711-3723.

DOI PMID

[5]
Bai Y F, Guo C C, Degen A A et al., 2020. Climate warming benefits alpine vegetation growth in Three-River Headwater Region, China. Science of the Total Environment, 742: 140574.

DOI

[6]
Brennan J, Johnson P, Olson K, 2021. Classifying season long livestock grazing behavior with the use of a low-cost GPS and accelerometer. Computers and Electronics in Agriculture, 181: 105957.

DOI

[7]
Chen Y, Luo H L, Liu X L et al., 2013. Effect of restricted grazing time on the foraging behavior and movement of tan sheep grazed on desert steppe. Asian-Australasian Journal of Animal Sciences, 26(5): 711.

DOI PMID

[8]
Chen Z Z, Wang S P, 2000. Typical Grassland Ecosystem in China. Beijing: Science Press. (in Chinese)

[9]
Cingolani A M, Noy-Meir I, Díaz S, 2005. Grazing effects on rangeland diversity: A synthesis of contemporary models. Ecological Applications, 15(2): 757-773.

DOI

[10]
Conant R T, 2012. Grassland Soil Organic Carbon Stocks: Status, Opportunities, Vulnerability. Dordrecht: Springer Press.

[11]
Conant R T, Paustian K, Elliott E T, 2001. Grassland management and conversion into grassland: Effects on soil carbon. Ecological Applications, 11(2): 343-355.

DOI

[12]
Coppolillo P B, 2001. Central-place analysis and modeling of landscape-scale resource use in an East African agropastoral system. Landscape Ecology, 16: 205-219.

DOI

[13]
Coughenour M, Behnke R, Lomas J et al., 2008. Forage Distributions, Range Condition, and the Importance of Pastoral Movement in Central Asia:A Remote Sensing Study. Dordrecht: Springer Press.

[14]
Dara A, Baumann M, Freitag M et al., 2020. Annual Landsat time series reveal post-Soviet changes in grazing pressure. Remote Sensing of Environment, 239: 111667.

DOI

[15]
Dong S K, Long R J, Kang M Y et al., 2003. Effect of urea multinutritional molasses block supplementation on liveweight change of yak calves and productive and reproductive performances of yak cows. Canadian Journal of Animal Science, 83(1): 141-145.

DOI

[16]
Dorji T, Totland Ø, Moe S R, 2013. Are droppings, distance from pastoralist camps, and pika burrows good proxies for local grazing pressure? Rangeland Ecology & Management, 66(1): 26-33.

DOI

[17]
Du Y G, Ke X, Guo X W et al., 2019. Soil and plant community characteristics under long-term continuous grazing of different intensities in an alpine meadow on the Tibetan Plateau. Biochemical Systematics and Ecology, 85: 72-75.

DOI

[18]
Dubinin M, Luschekina A, Radeloff V C, 2011. Climate, livestock, and vegetation: What drives fire increase in the arid ecosystems of southern Russia? Ecosystems, 14: 547-562.

DOI

[19]
Dunne T, Western D, Dietrich W E, 2011. Effects of cattle trampling on vegetation, infiltration, and erosion in a tropical rangeland. Journal of Arid Environments, 75(1): 58-69.

DOI

[20]
Egeru A, Wasonga O, MacOpiyo L et al., 2015. Piospheric influence on forage species composition and abundance in semi-arid Karamoja sub-region, Uganda. Pastoralism, 5(1): 1-17.

DOI

[21]
Fan J W, Shao Q Q, Liu J Y et al., 2010. Assessment of effects of climate change and grazing activity on grassland yield in the Three Rivers Headwaters Region of Qinghai-Tibet Plateau, China. Environmental Monitoring and Assessment, 170: 571-584.

DOI

[22]
Feng X M, Zhao Y S, 2011. Grazing intensity monitoring in northern China steppe: Integrating CENTURY model and MODIS data. Ecological Indicators, 11(1): 175-182.

DOI

[23]
Fetzel T, Havlik P, Herrero M et al., 2017. Seasonality constraints to livestock grazing intensity. Global Change Biology, 23(4): 1636-1647.

DOI PMID

[24]
Follett R F, Kimble J M, Lal R, 2001. The Potential of US Grazing Lands to Sequester Carbon and Mitigate the Greenhouse Effect. Chelsea: CRC Press.

[25]
Ganskopp D C, Bohnert D W, 2009. Landscape nutritional patterns and cattle distribution in rangeland pastures. Applied Animal Behaviour Science, 116(2-4): 110-119.

DOI

[26]
Gao J G, Zhang Y L, Liu L S et al., 2014. Climate change as the major driver of alpine grasslands expansion and contraction: A case study in the Mt. Qomolangma (Everest) National Nature Preserve, southern Tibetan Plateau. Quaternary International, 336: 108-116.

DOI

[27]
González L A, Bishop-Hurley G, Henry D et al., 2014. Wireless sensor networks to study, monitor and manage cattle in grazing systems. Animal Production Science, 54(10): 1687-1693.

DOI

[28]
Graser A, 2019. Movingpandas: Efficient structures for movement data in python. GIForum, 1: 54-68.

[29]
Gu C J, Zhang Y L, Liu L S et al., 2023. Differentiation of grazing pressure expressions and its applicable scenarios. Chinese Journal of Applied Ecology, 34(2): 433-441. (in Chinese)

DOI

[30]
Han Z, Song W, Deng X Z et al., 2018. Grassland ecosystem responses to climate change and human activities within the Three-River Headwaters Region of China. Scientific Reports, 8(1): 1-13.

[31]
Hankerson B R, Schierhorn F, Prishchepov A V et al., 2019. Modeling the spatial distribution of grazing intensity in Kazakhstan. Plos One, 14(1): e210051.

[32]
Harris R B, 2010. Rangeland degradation on the Qinghai-Tibetan Plateau: A review of the evidence of its magnitude and causes. Journal of Arid Environments, 74(1): 1-12.

DOI

[33]
Henkin Z, Ungar E D, Dolev A, 2012. Foraging behaviour of beef cattle in the hilly terrain of a Mediterranean grassland. The Rangeland Journal, 34(2): 163-172.

DOI

[34]
Homburger H, Lüscher A, Scherer-Lorenzen M et al., 2015. Patterns of livestock activity on heterogeneous subalpine pastures reveal distinct responses to spatial autocorrelation, environment and management. Movement Ecology, 3(1): 1-15.

DOI

[35]
Homewood K M, Rodgers W A, 2004. Maasailand ecology: Pastoralist development and wildlife conservation in Ngorongoro, Tanzania. Cambridge: Cambridge University Press.

[36]
Hoyt D F, Taylor C R, 1981. Gait and the energetics of locomotion in horses. Nature, 292(5820): 239-240.

DOI

[37]
Kawamura K, Akiyama T, Yokota H et al., 2005. Quantifying grazing intensities using geographic information systems and satellite remote sensing in the Xilingol steppe region, Inner Mongolia, China. Agriculture, Ecosystems & Environment, 107(1): 83-93.

DOI

[38]
Kays R, Crofoot M C, Jetz W et al., 2015. Terrestrial animal tracking as an eye on life and planet. Science, 348(6240): aaa2478.

DOI

[39]
Lehnert L W, Meyer H, Meyer N et al., 2014. A hyperspectral indicator system for rangeland degradation on the Tibetan Plateau: A case study towards spaceborne monitoring. Ecological Indicators, 39: 54-64.

DOI

[40]
Li G Y, Jiang C H, Cheng T et al., 2019. Grazing alters the phenology of alpine steppe by changing the surface physical environment on the northeast Qinghai-Tibet Plateau, China. Journal of Environmental Management, 248: 109257.

DOI

[41]
Li L H, Zhang Y L, Liu L S et al., 2018a. Spatiotemporal patterns of vegetation greenness change and associated climatic and anthropogenic drivers on the Tibetan Plateau during 2000-2015. Remote Sensing, 10(10): 1525.

DOI

[42]
Li L H, Zhang Y L, Liu L S et al., 2018b. Current challenges in distinguishing climatic and anthropogenic contributions to alpine grassland variation on the Tibetan Plateau. Ecology & Evolution, 8(11): 5949-5963.

[43]
Li Y H, Wang W, Liu Z L et al., 2008. Grazing gradient versus restoration succession of Leymus chinensis (Trin.) Tzvel. grassland in Inner Mongolia. Restoration Ecology, 16(4): 572-583.

DOI

[44]
Li Y Y, Dong S K, Wen L et al., 2014. Soil carbon and nitrogen pools and their relationship to plant and soil dynamics of degraded and artificially restored grasslands of the Qinghai-Tibetan Plateau. Geoderma, 213: 178-184.

DOI

[45]
Liao C, Clark P E, Degloria S D et al., 2017. Complexity in the spatial utilization of rangelands: Pastoral mobility in the Horn of Africa. Applied Geography, 86: 208-219.

DOI

[46]
Liu J Y, Xu X L, Shao Q Q, 2008. Grassland degradation in the “Three-River Headwaters” region, Qinghai province. Journal of Geographical Sciences, 18(3): 259-273.

DOI

[47]
Liu P P, Ding L M, Zhou Y Q et al., 2019. Behavioural characteristics of yaks grazing summer and winter pastures on the Qinghai-Tibetan Plateau. Applied Animal Behaviour Science, 218: 104826.

DOI

[48]
Long R J, Ding L M, Shang Z H et al., 2008. The yak grazing system on the Qinghai-Tibetan Plateau and its status. The Rangeland Journal, 30(2): 241.

DOI

[49]
Ma B R, Xie Y X, Zhang T Z et al., 2020. Identification of conflict between wildlife living spaces and human activity spaces and adjustments in/around protected areas under climate change: A case study in the Three-River Source Region. Journal of Environmental Management, 262: 110322.

DOI

[50]
Manning J, Cronin G, González L et al., 2017a. The behavioural responses of beef cattle (Bos taurus) to declining pasture availability and the use of GNSS technology to determine grazing preference. Agriculture, 7(5): 45.

DOI

[51]
Manning J K, Cronin G M, González L A et al., 2017b. The effects of global navigation satellite system (GNSS) collars on cattle (Bos taurus) behaviour. Applied Animal Behaviour Science, 187: 54-59.

DOI

[52]
Melillo J M, Mcguire A D, Kicklighter D W et al., 1993. Global climate change and terrestrial net primary production. Nature, 363(6426): 234-240.

DOI

[53]
Obermeier W A, Lehnert L W, Pohl M J et al., 2019. Grassland ecosystem services in a changing environment: The potential of hyperspectral monitoring. Remote Sensing of Environment, 232: 111273.

DOI

[54]
Pittarello M, Probo M, Lonati M et al., 2016. Effects of traditional salt placement and strategically placed mineral mix supplements on cattle distribution in the western Italian Alps. Grass and Forage Science, 71(4): 529-539.

DOI

[55]
Putfarken D, Dengler J, Lehmann S et al., 2008. Site use of grazing cattle and sheep in a large-scale pasture landscape: A GPS/GIS assessment. Applied Animal Behaviour Science, 111(1/2): 54-67.

DOI

[56]
Qian S, Fu Y, Pan F F, 2010. Climate change tendency and grassland vegetation response during the growth season in Three-River Source Region. Science China Earth Sciences, 53(10): 1506-1512.

DOI

[57]
Rivero M J, Grau-Campanario P, Mullan S et al., 2021. Factors affecting site use preference of grazing cattle studied from 2000 to 2020 through GPS tracking: A review. Sensors, 21(8): 2696.

DOI

[58]
Robinson T P, Wint G W, Conchedda G et al., 2014. Mapping the global distribution of livestock. Plos One, 9(5): e96084.

DOI

[59]
Rodgers A R, 2001. Tracking animals with GPS: The first 10 years. Tracking Animals with GPS, 1-10.

[60]
Rutter S M, Beresford N A, Roberts G, 1997. Use of GPS to identify the grazing areas of hill sheep. Computers and Electronics in Agriculture, 17(2): 177-188.

DOI

[61]
Scurlock J, Hall D O, 1998. The global carbon sink: A grassland perspective. Global Change Biology, 4(2): 229-233.

DOI

[62]
Seddon A W, Macias-Fauria M, Long P R et al., 2016. Sensitivity of global terrestrial ecosystems to climate variability. Nature, 531(7593): 229-232.

DOI

[63]
Shao Q Q, Fan J W, Liu J Y et al., 2016. Assessment on the effects of the first-stage ecological conservation and restoration project in Sanjiangyuan region. Acta Geographica Sinica, 71(1): 3-20. (in Chinese)

DOI

[64]
Spencer P, 2012. Nomads in alliance:Symbiosis and growth among the Rendille and Samburu of Kenya. Cambridge: Oxford University Press.

[65]
Sun J, Wang H M, 2016. Soil nitrogen and carbon determine the trade-off of the above- and below-ground biomass across alpine grasslands, Tibetan Plateau. Ecological Indicators, 60: 1070-1076.

DOI

[66]
Swain D L, Friend M A, Bishop-Hurley G J et al., 2011. Tracking livestock using global positioning systems: Are we still lost? Animal Production Science, 51(3): 167-175.

DOI

[67]
Tefera S, Snyman H A, Smit G N, 2007. Rangeland dynamics in southern Ethiopia: (1) Botanical composition of grasses and soil characteristics in relation to land-use and distance from water in semi-arid Borana rangelands. Journal of Environmental Management, 85(2): 429-442.

PMID

[68]
Tomkins N W, O Reagain P J, Swain D et al., 2009. Determining the effect of stocking rate on the spatial distribution of cattle for the subtropical savannas. The Rangeland Journal, 31(3): 267-276.

DOI

[69]
Ungar E D, Schoenbaum I, Henkin Z et al., 2011. Inference of the activity timeline of cattle foraging on a Mediterranean woodland using GPS and pedometry. Sensors, 11(1): 362-383.

DOI PMID

[70]
Van Auken O W, 2000. Shrub invasions of north American semiarid grasslands. Annual Review of Ecology and Systematics, 31(1): 197-215.

DOI

[71]
Venter Z S, Cramer M D, Hawkins H, 2019. Rotational grazing management has little effect on remotely-sensed vegetation characteristics across farm fence-line contrasts. Agriculture, Ecosystems & Environment, 282: 40-48.

DOI

[72]
Von Müller A R, Renison D, Cingolani A M, 2017. Cattle landscape selectivity is influenced by ecological and management factors in a heterogeneous mountain rangeland. The Rangeland Journal, 39(1): 1-14.

DOI

[73]
Wang D L, Du J, Zhang B T et al., 2017. Grazing intensity and phenotypic plasticity in the clonal grass Leymus chinensis. Rangeland Ecology & Management, 70(6): 740-747.

[74]
Wang J F, Xu C D, 2017. Geodetector: Principle and prospective. Acta Geographica Sinica, 72(1): 116-134. (in Chinese)

DOI

[75]
Wang Q, Wang J, Zhao Y et al., 2016. Reference evapotranspiration trends from 1980 to 2012 and their attribution to meteorological drivers in the Three-River Source Region, China. International Journal of Climatology, 36(11): 3759-3769.

DOI

[76]
Wang X Y, Yi S H, Wu Q B et al., 2016. The role of permafrost and soil water in distribution of alpine grassland and its NDVI dynamics on the Qinghai-Tibetan Plateau. Global and Planetary Change, 147: 40-53.

DOI

[77]
Western D, 1975. Water availability and its influence on the structure and dynamics of a savannah large mammal community. African Journal of Ecology, 13(3/4): 265-286.

DOI

[78]
Wiener G, 2013. The yak, an essential element of the high altitude regions of Central Asia. Études Mongoles et Sibériennes, Centrasiatiques et Tibétaines, 43/44.

[79]
Wilson R S, Husak J F, Halsey L G et al., 2015. Predicting the movement speeds of animals in natural environments. Integrative and Comparative Biology, 55(6): 1125-1141.

DOI PMID

[80]
Xu M, Kang S C, Chen X L et al., 2018. Detection of hydrological variations and their impacts on vegetation from multiple satellite observations in the Three-River Source Region of the Tibetan Plateau. Science of the Total Environment, 639: 1220-1232.

DOI

[81]
Yang S X, Feng Q S, Liang T G et al., 2018. Modeling grassland above-ground biomass based on artificial neural network and remote sensing in the Three-River Headwaters Region. Remote Sensing of Environment, 204: 448-455.

DOI

[82]
Yang X H, Guo X L, Fitzsimmons M, 2012. Assessing light to moderate grazing effects on grassland production using satellite imagery. International Journal of Remote Sensing, 33(16): 5087-5104.

DOI

[83]
Zengeya F M, Mutanga O, Murwira A, 2013. Linking remotely sensed forage quality estimates from WorldView-2 multispectral data with cattle distribution in a savanna landscape. International Journal of Applied Earth Observation and Geoinformation, 21: 513-524.

DOI

[84]
Zhang Y L, Li B Y, Liu L S et al., 2021. Redetermine the region and boundaries of Tibetan Plateau. Geographical Research, 40(6): 1543-1553. (in Chinese)

DOI

[85]
Zhang Y L, Liu L S, Bai W Q et al., 2006. Grassland degradation in the source region of the Yellow River. Acta Geographica Sinica, 61(1): 3-14. (in Chinese)

DOI

[86]
Zhang Y L, Liu L S, Wang Z F et al., 2019. Spatial and temporal characteristics of land use and cover changes in the Tibetan Plateau. Chinese Science Bulletin, 64: 2865-2875. (in Chinese)

[87]
Zhang Y L, Qi W, Zhou C P et al.., 2014. Spatial and temporal variability in the net primary production of alpine grassland on the Tibetan Plateau since 1982. Journal of Geographical Sciences, 24(2): 269-287.

DOI

Options
Outlines

/

〈 〉