Characterization of the spatial distribution of plateau pika burrows along an alpine grassland degradation gradient on the Qinghai–Tibet Plateau

Abstract Plateau pika burrows are common feature of degraded grassland in the Qinghai–Tibet Plateau (QTP) and serve as an important indicator of pika activity and grassland degradation. However, the current understanding of the spatial pattern changes of pika burrows and their critical thresholds across a degradation gradient in alpine grassland is deficient. In this study, we investigated and quantified changes in the spatial pattern of plateau pika burrows under typical degraded alpine shrub meadows in the northeastern QTP using an unmanned aerial vehicle and landscape pattern metrics. The degradation of the alpine shrub meadow leads to a change in landscape pattern from a two‐layered structure of alpine shrub and alpine meadow to a mosaic of alpine meadow and bare soil, with plateau pika burrows scattered throughout. Moderate degradation is the tipping point for changes in surface landscape patterns, followed by the disappearance of alpine shrub, the retreat of alpine meadows and the encroachment of bare soil, and the increasing density and size of pika burrows. The area characteristics of alpine meadows have influenced changes in the spatial pattern of pika burrow, and maintaining its proportional area is a vital measure to control the threat of pika burrows to pastures. The results of this paper provide a methodological reference and guidance for the sustainable utilization of grassland on the QTP.

The plateau pika (Ochotona curzoniae) has long been considered to be responsible for the degradation of alpine grassland due to their foraging, which can reduce vegetation height and alter plant community structure and biomass, and due to their burrowing behavior, which can destroy surface vegetation and increase the risk of soil erosion (Yi et al., 2016;Yu et al., 2017;Zhang et al., 2020). However, as research into the mechanisms behind alpine grassland degradation has increased, there is a growing consensus that the plateau pika is more of a consequence than a cause of grassland degradation. Plateau pikas provide many benefits for grassland ecosystems; for example, their burrows can provide shelter for other animals and their behavior serves to increase vegetation species richness, and they are also the prime prey for predator species on QTP (Dobson et al., 2019;Smith et al., 2019).
The burrows are important places for plateau pika to shelter, rest, and breed. Pika burrows can have detrimental or beneficial effects on grassland ecosystems (Sun et al., 2015); they not only can reduce plant productivity, soil organic carbon, and total nitrogen and can increase greenhouse gas emissions (Qin, Yi, et al., 2020;Zhao et al., 2019), but they can also provide homes for other animals and increase soil water infiltration efficiency (Delibes-Mateos et al., 2011;Wilson & Smith, 2015). In addition, the presence and development of pika burrows can strongly alter the spatial pattern of grassland . The fragmentation of intact grassland patches through the direct destruction and occupation of pika burrows, combined with the continued degradation of grassland and trampling by livestock, leads to the further expansion of bare patches (Cao et al., 2019;Chen et al., 2017;. Therefore, pika burrows are considered by pastoralists as a sign of grassland degradation and an obstacle to the sustainable use of rangelands. This is the main reason for the long-standing conflict between herders and plateau pikas, thus prompting herders to frequently use trapping and poisoning to exterminate the pikas . The spatial pattern of pika burrows can provide us with information on the population size and activity patterns of plateau pikas, as well as serve as a basis for determining the degree of grassland degradation (Wei et al., , 2020Zhao et al., 2019). The traditional survey method is manual field measurements of the burrows, but it still has the shortcomings of being time-consuming and costly.
The rapid development of unmanned aerial vehicle (UAV) in recent years has opened new methods for ecological research. UAVs have a fast revisit time and can provide data at high spatial resolution and low cost, which overcome the shortcomings of traditional surveys (Anderson & Gaston, 2013;Manfreda et al., 2018). A variety of studies have been conducted using UAV technology to investigate the spatial patterns of plateau pika burrows and their effects on surrounding grassland cover and bare patches Zhang et al., 2021), as well as on carbon emissions, ecosystem respiration, vegetation biomass, and soil organic carbon (Qin et al., 2015(Qin et al., , 2019Qin, Yi, et al., 2020). However, these studies have only focused on the spatial pattern of plateau pika burrows under a single level of degradation, and lack an understanding of the spatial pattern changes of plateau pika burrows and other types under the complete degradation gradient of alpine grasslands, so the key thresholds for the changes of pika burrow patterns are also unclear.
We investigated an alpine grassland pasture with a complete degradation sequence in the northeastern Qinghai-Tibet Plateau using an unmanned aerial vehicle. The objective of this paper was to quantitatively assess the changes in spatial pattern characteristics of plateau pika burrows and other surface types under different degrees of degradation in alpine grasslands, as well as the quantitative relationships between them, and thus obtain information on the key thresholds of pika burrow patterns corresponding to different degrees of grassland degradation. The results of this paper will improve our understanding of the changing pattern of pika burrows in degraded grassland on the QTP and provide a valuable reference for grassland management and wildlife conservation on the QTP.

| Study site
The study site is located on the northeastern edge of the Qinghai-Tibet Plateau, in an area of premountain pasture formed by alluvial fans on the southern slope of the Qilian Mountains, and is part of Menyuan County, Qinghai Province, China ( Figure 1). The region has a typical continental plateau climate with only two seasons a year: a cold and dry winter, and a humid and rainy summer. The average annual temperature is −1.7°C, with the lowest temperature occurring in January, when it can reach −14.8°C, and the highest temperature occurring in July, when it can reach 9.8°C. The total annual precipitation is 580 mm and is concentrated between May and September, which accounts for 80% of the total annual precipitation (Cao et al., 2007).
The vegetation type is alpine shrub meadow, with a two-layer structure that comprises an upper shrub layer (Potentilla fruticosa as the dominant species) and a lower herbaceous layer (dominated by Kobresia humilis). The area is used by local herders as summer pasture, it is grazed from June to September by Tibetan sheep and yaks. The short grazing period and the mixture of different grazing management regimes (family, joint family, and communal pastures) lead to more intensive grazing. Such grazing management has been practiced for at least 30 years, this has resulted in the degradation of previously widespread alpine shrub meadows from north to south, as is evidenced by a decline in the canopy, height, and proportion of shrub, and changes in the structure of plant communities and landscape patterns (Dai et al., 2020).

| Field sampling
Along the degradation gradient, we used the coverage of potentilla fruticosa shrub, the number of plant species, and the aboveground biomass to classify the degree of degradation into light degradation (LD), moderate degradation (MD), heavy degradation (HD), and severe degradation (SD), and to choose a grazing exclusion (GE) site as nondegraded alpine grassland (enclosed for more than 4 years) (Table 1) (Dai et al., 2020).

| Data processing
2.2.1 | UAV data processing and surface type classification The UAV imagery in each quadrat was processed using Agisoft PhotoScan to produce the remote-sensing images. We classified the surface types into alpine shrub, alpine meadow, bare soil, and plateau pika burrow using an object-oriented classification technique with the eCognition 9.0 software, consisting of multiresolution segmentation and decision tree classification. The multiresolution segmentation had a segmentation scale of 15, and a shape and compactness parameter of 0.5, which were obtained by trial and error. A simple decision classification tree was then constructed (Figure 3), and the images were classified by combining key indicators of the image objects, that is, the spectral, spatial, and textural information that was representative of the object ( Table 2). The classification results were then validated to assess the accuracy, and the decision tree parameters were continuously modified until the accuracy requirements were met, at which point the results were exported as vector files.
Four hundred and eighty ground truth samples of different landscape types (200 alpine meadow, 160 alpine shrub, 80 bare soil, and F I G U R E 1 Study site location. Note: Figure a, b, and c is the small watershed in which the study site is located, the location of the study site on the Qinghai-Tibet Plateau, and UAV imagery of the surface landscape at different levels of degradation, respectively. Abbreviations: GE, grazing exclusion; HD, heavy degradation; LD, light degradation; MD, moderate degradation; SD, severe degradation

| Selection and calculation of landscape pattern metrics
We used the characteristics of the pika burrows and other surface types to select six landscape pattern metrics (LPMs) at the class F I G U R E 2 Field sampling and UAV aerial photography. Note: a, b, and c is the spatial layout of the sample plots, the aerial photography process of UAV, and the interpretation results of aerial images. Abbreviations: GE, grazing exclusion; HD, heavy degradation; LD, light degradation; MD, moderate degradation; SD, severe degradation

Spectral information Mean brightness
To classify alpine shrub and alpine meadow, bare soil, and pika burrow

Excess green (EXG) Normalized green-red difference index (NGRDI) Red green-blue vegetation index (RGBVI)
To classify vegetation and nonvegetation Geometry information Shape index To classify bare soil and rodent burrow Texture information GLCM contrast To classify alpine shrub and alpine meadow level, representing attributes of the landscape such as area, fragmentation, stability, and connectivity (Table 3). We used the R package landscapemetrics (Hesselbarth et al., 2019) to calculate the LPMs for each quadrat.

| Statistical analysis
The characteristics of landscapes at different stages of degradation are presented as the average ± standard deviation. One-way analysis of variance (ANOVA) and a multicomparison of the least significant difference (LSD) test were performed using the SPSS 17.0 statistical software package (SPSS Inc.) to determine differences at a p = .05 level. Spearman's rank correlation was used to study the relationship between the pattern of pika burrows and other surface types.

| Changes in area and spatial pattern of different surface types along the degradation sequence
Alpine shrub meadows have similar area structures in GE and LD.
Both alpine meadows and alpine shrubs are absolutely dominant  which in turn decreased rapidly in HD and increased slightly in the SD stage (Table 5). This further illustrates the fact that alpine meadows tend to fragment as they degrade.
LPI indicates the proportion of the area of the largest patch in the landscape, which can reflect the stability and resilience potential of a certain landscape type. LPI in alpine meadows did not change much in LD, but increased significantly in MD (p < .05), indicating that the stability of the meadow landscape increases significantly at this stage, while it decreased continuously in HD and SD, indicating that the stability and resilience potential of alpine meadows continue to decline at these stages (Table 5). The ED reflects the extent to which the landscape is disturbed by the outside world, with the ED of alpine meadows decreased continuously in LD and MD and increased significantly in HD (p < .05), followed by a slight decrease in SD, indicating that alpine meadows are most strongly disturbed by other types in the HD stage (Table 5).
NP in alpine shrub continued to increase in LD and MD, while PD also increased significantly in LD (p < .05) and decreased slightly in MD. Area_MN and LPI continued to decrease and reached a minimum in MD. This indicates that the intact shrub is constantly being broken up into small isolated patches. The ED of the alpine shrub also showed a continuous decline, indicating that its boundaries tend to be simpler (Table 5).
Bare soil is one of the main indicators of alpine grassland degradation, and its landscape pattern generally changes significantly during the HD stage. The NP and PD of bare soil underwent a decrease followed by an increase at both the LD and MD stages, but only increased significantly and dramatically (p < .05) at HD, before decreasing slightly at SD. Area_MN, on the contrary, increased significantly in SD, and LPI continued to increase in HD and SD, indicating that bare soil expanded rapidly in both stages, occupying the original habitat of the alpine meadow. The ED of bare soil increased significantly in HD (p < .05) and then decreased slightly in SD, indicating that the bare soil boundary was complex and unstable under this stage and that degradation may still be developing (Table 5).

| Changes in the spatial pattern of plateau pika burrow along the degradation gradient
The pika burrows did not occur at the land surface at the GE and LD stages, but as grassland degradation increased, they gradually spread from being isolated and occasional occurrences to being between burrows shortened rapidly at the HD compared with MD level, from 3.51 ± 0.51 m to 0.60 ± 0.31 m. The distance between burrows reduced slightly further between the HD and SD (0.37 ± 0.09 m), but not significantly (Figure 4).

| The relationship between the spatial pattern of plateau pika burrow and other land surface types
There is a significant correlation between the spatial pattern of alpine meadow and pika burrows, except for Area_MN. LPI and PLAND for alpine meadow were more closely related to the spatial pattern of pika burrow, with LPI having a significant negative correlation with PD for pika burrows (Figure 5), and a significant positive correlation with ENN for pika burrows ( Figure 5). PLAND for alpine meadow had similar characteristics, but was more closely related to ENN for pika burrow, while its correlation coefficient with PD for pika burrows was less than that for LPI. Both ED and AREA_MN for alpine meadow were significantly negatively and significantly positively correlated with ENN_MN for pika burrows. The decrease in the proportional area of alpine meadow, and the shrinkage of its dominant patches, may have led to an increase in the number and density of pika burrow, as well as a decrease in their connectivity (Figure 5a).
The relationship between most spatial pattern features of bare soil and pika burrows is less clear than for alpine meadows and pika burrows. Only PLAND and AREA_MN correlate with all pattern indices for pika burrows at significant levels. There were a significant negative correlation between PLAND for bare soil and ENN_MN for pika burrows, and a significant positive correlation with other indices. The ED of bare soil was significantly and positively correlated with the PD and NP for pika burrows (Figure 5b).
As the consistently dominant landscape type, alpine meadow had a significantly greater effect on the spatial pattern of pika burrows than bare soil did. The decrease in PLAND and LPI for alpine meadow caused an increase in the number and density of pika burrows, while the increase in PLAND for bare soil caused the same effect.
We used PLAND and ED, which are closely related to the spatial pattern of pika burrows, to represent the area and edge characteristics of alpine meadow, respectively, and used PD and ENN, which represent the density and connectivity of pika burrows, to quantitatively analyze the relationship between spatial pattern of alpine meadow and pika burrows. There was a significant linear relationship between PLAND for alpine meadow and PD and ENN for pika burrow. For every 10% decrease in PLAND of alpine meadow, the PD of pika burrows increased by 0.15 number/m 2 , and ENN for pika burrows increased by 0.59 m. There was also a significant linear relationship between ED for alpine meadow and PD and ENN for pika burrows. For every 10 m/m 2 increase in ED for alpine meadow, PD for pika burrows increased by 0.38 number/m 2 , while ENN shrank by 1.56 m (Figure 6).

| D ISCUSS I ON
The two-layered structure of shrub and meadow of a landscape in pristine condition degraded to a landscape pattern dominated by meadows and bare soil, with pika burrows throughout as the level of landscape degradation increased. Moderate degradation was the key stage of change in the surface landscape pattern of alpine shrub meadows, as the area of shrub declined sharply, meadow became the dominant landscape type, and pika burrows began to appear. As degradation increased, meadows tended to fragment, while bare soil expanded and encroached on the meadow landscape, and pika burrows became a pervasive landscape feature.
The density, average size, and number of pika burrows increased significantly with the degradation of alpine shrub meadow, while the linear distance between burrows decreased significantly. Changes in the size of pika burrows, and the distance between them are related to the migration, development, and reproduction of pika 28.58 ± 7.21 a 0.10 ± 0.08 b SD -60.60 ± 11.62 bc 39.12 ± 11.64 a 0.31 ± 0.11 a Abbreviations: GE, grazing exclusion; HD, heavy degradation; LD, light degradation; MD, moderate degradation; SD, severe degradation.
Lower case letters represent significant differences (P < .05) in area proportion for the same surface type at different levels of degradation.  Abbreviations: GE, grazing exclusion; HD, heavy degradation; LD, light degradation; MD, moderate degradation; SD, severe degradation.

TA B L E 4 Changes in area proportion
Lower case letters represent significant differences (P < .05) in landscape patterns for the same surface type at different levels of degradation.
populations The results of this study showed a density of 5500 pika burrows/ha in degraded alpine shrub meadow, whereas surveys conducted in the western part of the southern Qilian Mountains have shown that the average number of pika burrows in normal grassland is 300-800 burrows/ha Qin et al., 2019;Yi et al., 2016). In contrast, studies of degraded alpine grassland in the hinterland of QTP found the average number of pika burrows to be 2700-4775 burrows/ha (Liu et al., 2017;Tian et al., 2019;Wang et al., 2020;Wei et al., 2013), and in the degraded grasslands of the eastern QTP, the density of pika burrows can even reach 7066 burrows ha −1 (Zhang et al., 2021), which is close to our results. The density of pika burrows has been proven to be significantly correlated with the coverage of vegetation and bare soil Zhang et al., 2021), and it is also a valid indicator to estimate the population size of pika in the plateau. There is evidence that increases in the density of plateau pika populations lead to increased reductions in grassland aboveground biomass, soil carbon, and nitrogen (Liu et al., 2017;Qin, Yi, et al., 2020;Yi et al., 2016). This suggests that the increase in the population of plateau pika does have a negative impact on grassland ecosystems, but that grassland degradation is still a complex process involving multiple factors, and that pika population may only exacerbate grassland degradation when a certain threshold is exceeded.
We found that the average area of a pika burrow opening was 0.003 m 2 , which is smaller compared with other studies (Ma et al., 2011;Pech et al., 2007). We also found that the average linear distance between pika burrows across all the stages of degradation was 0.49 m. However, there are few examples of studies for comparison, due to the limited means of investigation. Tang et al. (2019) investigated the spatial connectivity of pika burrows using the COHESION index, but as this index is a relative measure with an upper limit of 100, it is difficult to compare it with our study.
In this paper, we found that the landscape pattern of alpine meadows was closely related to that of pika burrows. Higher proportions of meadow area, large meadow patches, and larger average meadow patch size were all able to inhibit the development of pika burrows. Tang et al. (2019) also found that a similar relationship exists between vegetation cover and the number of pika holes. This is because dense vegetation is not conducive to pika escaping or avoiding predators, and they therefore prefer habitats that provide an open view and lower grass height, such as degraded grassland (Liu et al., 2003). Therefore, it is now generally accepted that pika does not cause grassland degradation. However, the activities of pikas do accelerate the rate of degradation in alpine grassland by reducing productivity and altering the structure of plant communities through their feeding and clipping effects. Furthermore, their burrow excavations of burrows can directly occupy grassland growth space and therefore, together with the alteration of soil properties, lead to an increase in the area of bare soil and a decrease in the proportion of grassland area Yu et al., 2017).
A further linear relationship was constructed to show the key threshold for changes in pika burrow patterns; that is, for every 10% increase in the proportion of meadow area, the density of pika F I G U R E 6 Relationship between PLAND and ED for alpine meadow and PD and ENN for pika burrows. Abbreviations: ED, edge density; ENN_MN, mean of Euclidean nearest-neighbor distance; PD, patch density; PLAND, percentage of landscape

| CON CLUS ION
Using high-resolution UAV images, this study provides a quantitative assessment of changes in the spatial pattern of plateau pika burrows along the alpine shrub meadow degradation gradient. The degradation of the alpine shrub meadow leads to a dramatic change in the surface landscape pattern, with the two-layered structure of shrub and meadow changing to a landscape pattern where meadow and bare soil coexist and pika burrows are scattered throughout.
Moderate degradation is the starting point for the dramatic changes in pika burrows and the loss of shrub. The increasing number and size of pika burrows naturally lead to an increase in their density and a reduction in their distance. Maintaining a certain proportion of meadow area is a key measure to prevent uncontrolled pika burrow increase. Therefore, improving grassland management is the key to preventing grassland degradation and pika activity from threatening the sustainable use of alpine grassland. UAV technology can provide a new observation tool for grassland degradation studies on the Qinghai-Tibet Plateau, and has great advantages and potential for plateau pika surveys.

CO N FLI C T O F I NTE R E S T
No conflict of interest exists in the submission of this manuscript, and the manuscript is approved by all authors for publication.