Plant and Animal Ecology

The Effects of Plateau Pika (Ochotona curzoniae) Presence and Population Control on the Structure of an Alpine Grassland Bird Community

  • Joseph P. LAMBERT , 1 ,
  • Johanna V. HARTMANN 2, 3 ,
  • SHI Kun , 1, 4, * ,
  • Philip RIORDAN 2, 3
  • 1. Wildlife Institute, School of Nature Conservation, Beijing Forestry University, Beijing 100083, China
  • 2. School of Biological Sciences, University of Southampton, Southampton SO171BJ, UK
  • 3. Marwell Wildlife, Winchester, Hampshire SO211JH, UK
  • 4. Eco-Bridge Continental, Beijing 100085, China
*SHI Kun, E-mail:

Joseph P. LAMBERT, E-mail:

Received date: 2021-05-20

  Accepted date: 2022-03-20

  Online published: 2023-01-31

Supported by

The National Natural Science Foundation of China(31470567)


China’s Qinghai-Tibetan Plateau (QTP) is an important area for bird conservation, with many endemic and Threatened species. Colonial burrowing mammals play an important role in structuring bird communities in arid grasslands around the world. On the QTP, the plateau pika Ocho tona curzoniae builds colonies which provide a dense source of resources for many bird species. However, pikas are regarded as a pest by local pastoralists, and they are the target of a population reduction campaign which could have a significant impact on the bird communities. We surveyed bird communities at Gansu Yanchiwan National Nature Reserve to investigate the differences in community structure between sites with pika colonies (on colony) and sites without them (off colony), and between pika colonies which had been poisoned and those which had not. Using non-metric multidimensional scaling (NMDS) combined with permutational multivariate analysis of variance (PERMANOVA) and Wilcoxon rank-sum tests, we found that there was no significant difference in bird community composition or abundance between the poisoned and untreated colonies. However, there was a very large and statistically significant difference in bird community structures between on- and off-colony sites. Only horned lark Eremophila alpestris was consistently observed at sites without pika colonies, while ten bird species were observed on colonies. Six species were significantly more abundant on colony than off. While we could not claim that the poisoning campaign at Yanchiwan is altering bird communities, the presence of pika colonies seems to be an indispensable resource for the resident birds.

Cite this article

Joseph P. LAMBERT , Johanna V. HARTMANN , SHI Kun , Philip RIORDAN . The Effects of Plateau Pika (Ochotona curzoniae) Presence and Population Control on the Structure of an Alpine Grassland Bird Community[J]. Journal of Resources and Ecology, 2023 , 14(1) : 46 -56 . DOI: 10.5814/j.issn.1674-764x.2023.01.005

1 Introduction

Alpine grassland environments, with homogeneous vegetation and a harsh climate, can pose challenges for birds. China’s Qinghai-Tibetan Plateau (QTP) is one such habitat. The QTP is a centre of high local endemism among birds (Li et al., 2018) that is also subject to rapid economic development and the accompanying land-use shifts and habitat degradation (Li et al., 2017), making it an important focal point for bird conservation in China.
Allogenic ecosystem engineers, those species whose actions modify the structure of the surrounding habitat (Jones et al., 1994, 1997), are especially relevant for conservation in such semi-arid and grassland environments. Their presence can increase habitat heterogeneity, species diversity, and biomass (Root-Bernstein and Ebensperger, 2013) in a variety of taxa (Kinlaw, 1999). Small burrowing mammals are a well-studied group of ecosystem engineers whose activities have been shown to significantly alter the structure of bird communities around the world (e.g., Delibes-Mateos et al., 2008). They can function as a dense source of prey for predatory birds, provide breeding sites for cavity-nesting species, and can alter the suitability of the surrounding habitat for foraging (Root-Bernstein et al., 2013) or breeding (Li et al., 2015). From a community ecology perspective, these interactions function by the modification of resources available to bird species for exploitation, increasing the amount of habitat that falls within the fundamental niche space (Bruno et al., 2003). They can also be thought of as reducing stress, i.e., improving fitness, in either one or both partners. This process is known as facilitation, and it is an important driver of biodiversity (Stachowicz, 2001).
As the dominant small burrowing herbivore on the QTP, the plateau pika Ochotona curzoniae is considered an important ecosystem engineer and keystone species (Smith and Foggin, 1999; Hogan, 2010). Its presence has been shown to significantly alter the community structure of mammals (Badingqiuying et al., 2016), plants (Bagchi et al., 2006; Smith et al., 2019), and birds (Lai and Smith, 2003; Badingqiuying et al., 2016). Interactions with bird communities are perhaps the most frequently studied of the pika’s interspecific interactions. In total, 18 species of birds have been reported in the literature to interact strongly with plateau pikas (see Supplementary material). However, pikas are commonly regarded as a pest in many grassland areas of China (Badingqiuying, 2016) and have been extensively poisoned since 1958, sometimes to the point of local extirpation (Smith et al., 2019). This has significantly altered the structure of the local bird communities (Lai and Smith, 2003; Badingqiuying et al., 2016), and has provided a natural experimental setup in which to investigate the pika’s impact on other animal species.
Pikas can alter bird community structure via several distinct pathways: by providing breeding sites for cavity-nesting birds in the form of burrows, by functioning as prey, and by altering nest-site suitability for some ground-nesting birds. For example, Lai and Smith (2003) found an increase in species richness and abundance at sites which had not been poisoned compared to poisoned sites. Specifically, burrow-nesting species of snowfinch (Tibetan snowfinch Montifringilla adamsi, white-winged snowfinch M. nivalis, plain-backed snowfinch Pyrgilauda blanfordi, small snowfinch P. davidiana, rufous-necked snowfinch P. ruficollis, and white-rumped snowfinch Onychostruthus taczanowskii) were significantly more abundant at sites where pika had not been poisoned. Three of these species were never seen on poisoned sites. These six species have all been found to nest in pika burrows (Lu et al., 2009). Elsewhere, the abundance of P. ruficollis and O. taczanowskii has been shown to increase with pika burrow density (Arthur et al., 2008). Isabelline wheatear Oenanthe isabellina preferentially nests in the burrows of both Daurian pika Ochotona dauurica and plateau pika O. curzoniae (Lai and Smith, 2003; Li and Peng, 2014). Hume’s ground-pecker, Pseudopodoces humilis, was found to be completely absent at sites where pika had been poisoned (Lai and Smith, 2003).
In addition to burrow-nesting passerines, birds of prey have also been shown to occur in greater numbers in areas with high pika densities. Lai and Smith (2003) reported significantly more black-eared kite Milvus nigrans in areas where pika numbers had not been controlled. Pellets of upland buzzard Buteo hemilasius and saker falcon Falco cherrug were found to consist of over 90% pika material (Schaller, 1998), and these species have all been observed in significantly reduced numbers at sites where pika had been eradicated (Badingqiuying et al., 2016).
Pikas may also impact breeding success in species which build open nests on the ground. Open-nesting birds are at increased risk of predation if they nest in areas with high densities of shared prey. For example, oriental skylark Alauda gulgula has been shown to select nest sites with a lower density of pika burrows than randomly selected sites (Li et al., 2015). Nest failure occurred most often at nests with a greater density of pika burrows nearby, due to increased rates of predation by mountain weasel Mustela altaica.
Previous studies on the relationships between pika control and bird communities have focussed on comparing sites where pika had been eradicated, or at least greatly reduced (e.g., Lai and Smith, 2003; Badingqiuying et al., 2016). However, the rapid population growth potential of plateau pikas means that the control of pika numbers by poisoning on the QTP often does not result in complete eradication (e.g., Pech et al., 2007). Less intense poisoning campaigns, and other methods of control such as sterilisation, are becoming popular with managers in China as they reduce pika populations without leading to the collapsed burrows that prevent some burrow-dwelling birds from establishing nests (Liu, 2019). However, a reduction in pika numbers could still lead to declines in either birds of prey or cavity-nesting birds which prefer occupied burrows such as O. taczanowskii (Zeng and Lu, 2009), or to increasing competition between birds for a smaller pool of breeding sites (Arthur et al., 2008).
Previous studies on the ecosystem engineering effects of plateau pikas have focused on the Kobresia-dominated alpine meadows which comprise the majority of its range. The northernmost part of the pika’s range contains more varied habitat types, including arid montane steppes, deserts, and shrubland (Liu et al., 2010). Engineering effects vary between habitats (Root-Bernstein et al., 2013), and how pikas might alter bird communities in different landscapes is currently unknown.
In this study, we aim to understand the effects on birds of the pika control programme in China’s Gansu Yanchiwan National Nature Reserve (NNR) by comparing the bird communities at sites in which yearly pika poisoning occurs with those at sites where pika numbers are not controlled. We also compare the bird communities at sites with and without pika colonies, to estimate the potential effect of a more complete poisoning programme and to understand the engineering effects of pikas more broadly.

2 Methods

2.1 Study area

Yanchiwan NNR is located in China’s Gansu Province, on the north-eastern edge of the QTP (see Fig. 1). It is characterised by a continental and semi-arid climate. The Kobresia steppe, which is dominant over much of the Qinghai-Tibetan Plateau, is absent in Yanchiwan. Instead, the reserve is dominated by alpine desert and montane steppe dominated by Stipa purpurea (Liu et al., 2010). The reserve is listed as part of an Important Bird Area (Bird Life International, 2009) and is home to a breeding population of Endangered black-necked cranes (Grus nigricollis) as well as many other species endemic to China and the Qinghai-Tibetan plateau (Liu et al., 2010).
Fig. 1 Location of Gansu Yanchiwan NNR within China’s Gansu Province
The poisoning of pika in Yanchiwan began in 1976, and has continued on a yearly basis, with poison being applied by reserve staff at the beginning of the pika’s breeding season (Dabuxilite, personal communication). As indicated by staff, the poison is most likely botulinum toxin type-C, the most frequently used poison against pikas (Liu, 2019), and is applied across the entire targeted colony. Details on the exact dosage of poison were unavailable. The poisoning is part of a comprehensive system of mammal control that also includes artificial perches designed to attract birds of prey to areas where small mammal densities are high (Liu et al., 2010). According to reserve staff, poisoning in Yanchiwan sub-reserve is confined to the centre of the park. However, our observations indicate that local people may engage in poisoning, without the reserve’s knowledge or guidance, in other parts of the park.

2.2 Sites

During 2017, surveys of bird communities were carried out in the vicinity of six pika colonies that had been subjected to recent poisoning operations, and six colonies that had not been subjected to poisoning. The so-called “poisoned sites” were those subjected to poisoning operations during the summer of 2016, and were identified based on location data obtained from the local nature reserve staff (see Fig. 2). An additional 20 sites were surveyed during summer of 2018: Ten located on pika colonies, and ten located in areas with no evidence of pika presence (see Fig. 3).
Fig. 2 Map of poisoned and untreated (i.e., not poisoned) survey sites in Yanchiwan NNR for 2017
Fig. 3 Map of survey sites on and off pika colonies surveyed in 2018 in Yanchiwan NNR
Each site consisted of a circular plot with a radius of 200 m and was over 250 m away from its nearest neighbour to maximise the likelihood of independence. All sites were more than 250 m away from wetlands or running water to prevent birds unique to these habitats from being included in the count. The centre of each on-colony site was marked by existing artificial perches which were placed by reserve teams to attract birds of prey as a form of ecological pest control. The presence of these perches suggested that pika colonies were present at these sites. On-colony sites surveyed in both years were randomly selected from a subset of all perches in the park that were located during a brief survey in 2017. Off-colony sites were randomly selected from a set of sites that were comparable with on-colony sites in terms of vegetation structure and soil type, but with no pika colonies present. All sites were thoroughly checked for pika colony presence after random selection.
In both years, all sites were located in either arid montane steppe, dominated by the sedge Stipa purpurea and the forbs Ajania fruticulosa and Krascheninnikovia compacta, or in alpine desert shrubland dominated by Potentilla fruticosa. Many of the sites showed evidence of desertification and grassland degradation to varying degrees.

2.3 Bird point survey

Surveys were conducted between June 5th and July 20th in 2017, and between July 11th and August 5th in 2018. Bird point counts were carried out twice at each site, once in the morning and once in the evening (as in Sutherland et al., 2004). Two observers were positioned at the centre of the site facing north and south, respectively. A five-minute pause after arriving at the site allowed any birds that were disturbed to habituate to our presence. For 15 minutes, all birds observed within the 200 m radius of the site were recorded and visually identified to the species level whenever possible. Only birds landing in the site were recorded; birds flying directly overhead were not recorded unless they subsequently landed within the site during the survey. Surveys were carried out in the morning between 8:00 and 10:00, and in the evening between 17:00 and 19:00 when birds were likely to be more active. The total number of birds observed was summed for each site. All surveyors were equally experienced at observing birds in this habitat, and the flat, open and uniform nature of the habitats at each site provided confidence that few birds were missed and that the detection rates were relatively high and constant.

2.4 Habitat variables

To control for possible variations in bird communities caused by changes in habitat unrelated to pika presence or poisoning, we collected data on several variables related to the vegetation structure at each site. Vegetation data were collected by walking four 200 m-long transects, beginning at the centre of each site and moving outwards along the four major points of the compass, and placing a quadrat at 20 m intervals (See Fig. 4). In each quadrat, we estimated plant species richness, percentage cover of vegetation (further subdivided into grasses and forbs/shrubs), rocks, and livestock dung (as a proxy measurement of recent grazing pressure). We also measured the height of the tallest plant in each quadrat. The vegetation variables were then averaged across the whole site. The GPS coordinates and elevation above sea-level for the centre of each site were also recorded.
The number of pika burrow entrances was counted at each of the poisoned and non-poisoned on-colony sites during 2017. Each site was divided into eight segments, according to the lines of the compass emanating from the centre, and every other segment was surveyed by walking along the outer edge of the site and back again, gradually moving towards the centre in a zig-zag pattern, with every burrow entrance recorded (see Fig. 4).
Fig. 4 Schematic diagram of the site and transects

3 Data analysis

3.1 Impact of poisoning on burrow density

Due to the small sample size and non-normal distribution of data, the non-parametric Wilcoxon rank sum test was used to test for differences in mean burrow density between poisoned and untreated sites.

3.2 Exploring community composition

A non-metric multidimensional scaling (NMDS) approach was used to visualise patterns of bird community structure at poisoned and untreated sites in 2017, and at off-colony and on-colony sites in 2018. NMDS is an ordination technique that uses rank-order of a given community dissimilarity metric to ordinate sites so that the sites with more similar communities appear closer together in the ordination space. Bray-Curtis dissimilarity was used as the distance metric, as it is robust to sampling error (Schroeder and Jenkins, 2018) and maintains a strong monotonic relationship with original ecological differences even at large distances (Faith et al., 1987). NMDS was performed using the function metaNMDS in the R package ‘vegan’ (Oksanen et al., 2019). Bird abundance data were square-root transformed prior to ordination to decrease the effect of abundant species as outliers (Van Nimwegen et al., 2008), and a very small constant (0.0001) was added to each species count to enable the inclusion of sites with no species present. The ‘envfit’ function in vegan was used to superimpose vectors representing habitat variables and centroids representing the categorical treatment variables onto the ordinated data. Correlations between these vectors and centroids and the distance in ordination space is represented by an R2 value, and P-values are calculated based on bootstrapping with 999 permutations. Fitted vectors (or centroids) with an R2 value >0.05 were plotted onto the ordinations. We used the non-parametric Wilcoxon rank-sum test to test for differences in abundance of individual bird species between the different treatments.

3.3 Permutational multivariate analysis of variance (PERMANOVA)

To test for significant differences in community composition between a) poisoned and untreated sites and b) on-colony and off-colony sites, a permutational multivariate analysis of variance (PERMANOVA; Anderson, 2001) was used. PERMANOVA tests the null hypothesis that there is no significant difference in the locations (i.e., centroids) of two groups (Anderson and Walsh, 2013) by calculating a pseudo-F-ratio between the within-group sum of squares and between-group sum of squares (Anderson, 2001). The null distribution of the F-statistic is derived using permutation, and the associated P-values can be calculated (Anderson, 2001). PERMANOVA can become overly conservative in the face of heterogeneous variances between groups in unbalanced designs when the larger group has greater variance (Anderson and Walsh, 2013), and so we set alpha to 0.01.

3.4 Constrained ordination: Partial distance-based redundancy analysis

We used constrained ordination to test for differences in bird community structure between poisoned and untreated sites in 2017, and between off-colony and on-colony sites in 2018. A partial distance-based redundancy analysis (db-RDA) was used that allows the patterns in the multivariate species data to be expressed as a linear function of environmental or spatial covariates (Legendre and Legendre, 2012), and then partitions the observed variation into that caused by environmental factors and that caused by spatial factors. Redundancy analysis is essentially an extension of multiple linear regression for use on multivariate response data (Legendre and Legendre, 2012). db-RDA is a variant of RDA used on distance metrics between sites, rather than on raw community data (Legendre and Anderson, 1999). In this instance, an RDA was carried out on the outputs from a principal coordinate analysis (PCoA) using the Bray-Curtis dissimilarity metric.
A partial form of db-RDA was used to partition the variation in bird community structure into four distinct components: variation caused by spatial structure, variation caused by non-spatial environmental variables, variation caused by spatially structured environmental variables, and variation unexplained by either component (Borcard et al., 1992). Spatial data were incorporated into the analysis by first generating Moran’s eigenvector maps (MEMs) (Dray et al., 2006). MEMs are generated by eigenanalysis of a spatial weighting matrix (SMW). This SMW can be thought of as the product of two component matrices: A binary connection matrix which describes whether or not a site is connected to another one, and a weighting matrix which describes the weight given to each distance. The SMW was selected from a series of candidate matrices using the procedure described in Bauman et al. (2018), in the R package ‘adespatial’ (Dray et al., 2020). Details of the procedure can be found in the Supplementary materials. The significance of the db-RDA, along with individual axes and environmental variables, was tested using an ANOVA with 5000 permutations in the ‘vegan’ package.

4 Results

4.1 Poisoning and burrow density

Poisoned sites contained 1051 burrow entrances on average (Range: 189 to 1844), while untreated sites contained an average of 1282 burrow entrances (Range: 94-2554). There was no statistically significant difference between these two means (W = 15, P > 0.1).
Fig. 5 Box plot showing burrow densities at poisoned and untreated sites

4.2 Bird survey

A total of eight bird species were observed during our survey in 2017 (see Fig. 6), while a total of 11 species were observed across all sites in 2018 (see Fig. 7). Of these, the saker falcon (Falco cherrug) is listed as Endangered by the IUCN (Bird Life International, 2017). Bird communities at on-colony sites were distinct from those at off-colony sites. By far the most commonly observed bird at both on-colony and off-colony sites in 2018 was horned lark (Eremophila alpestris, n=80, 25, respectively). Apart from a lone observation of snow pigeon (Columba leuconota), E. alpestris was the only species observed at off-colony sites.
Fig. 6 Bar chart showing total observations of each bird species at poisoned and untreated sites in 2017

Note: None of the differences were significant according to Wilcoxon rank sum tests (P >0.05).

Fig. 7 Graph of total observations of bird species at sites off and on pika colonies in 2018

Note: Differences were significant for P. humilis (W = 72, P = 0.02), U. epops (W = 72, P = 0.02), S. paradoxus, (W = 72, P = 0.02), P. ruficollis (W = 90, P = <0.001), O. taczanowskii (W = 72, P = 0.02) and B. hemilasius (W = 76.5, P = 0.01).

4.3 Community composition

4.3.1 Correlations with pika poisoning

The NMDS of bird communities in 2017 converged on a two-dimensional solution with stress of 0.05 and a non- metric fit R2 = 0.997, indicating that a great deal of the multidimensional variation in the community data was preserved in the two-dimensional ordination. Four non-poisoned sites appeared close together in the ordination, though envfit vectors indicated that these four sites were also similar in terms of elevation (3722 ± 20 m a.s.l). Elevation in turn was strongly positively correlated with total vegetation cover (Pearson’s R2= 0.7), grass cover (R2 = 0.7) and burrow density (R2 = 0.7). Poisoned sites were all situated within a small range of lower elevations (3373 ± 28 m a.s.l), making it difficult to partition the variation caused by these variables. Wilcoxon rank-sum tests showed no significant differences in the variables between poisoned and untreated sites. P. humilis, O. taczanowskii and B. hemilasius were associated with sites at higher elevation, with a greater percentage of vegetation cover, and with a higher density of pika burrows. Other species showed no clear associations with environmental vectors or poisoning. Of the six habitat variables, only elevation correlated with the variation in community structure to a degree statistically different from zero (R2 = 0.7, P = 0.01). Of the eight species observed, three were observed in great enough numbers to perform Wilcoxon rank-sum tests, namely P. humilis, P. ruficollis, and O. taczanowskii. While all of these species occurred in greater numbers at non-poisoned sites than poisoned sites (see Fig. 6), these differences were not significant according to Wilcoxon rank-sum test (P > 0.05). PERMANOVA was unable to reject the null hypothesis of significant differences in centroid location (F = 1.2, P = 0.3) between poisoned and untreated sites.
Fig. 8 NMDS of bird communities at poisoned and untreated sites during 2017

Note: Envfit vector displayed for elevation (R2 = 0.7, P = 0.01).

4.3.2 Correlations with colonies

The NMDS of bird communities in 2018 converged on a two-dimensional solution after 20 random starts with stress = 0.1 and a non-metric fit R2 = 0.989. The most pronounced distinction between groups was between sites on colonies and sites off colonies (see Fig. 9). Community composition was significantly correlated with colony type according to envfit (R2 = 0.5, P = 0.001). PERMANOVA tests showed a significant difference in centroid locations between these two groups, even with our conservative alpha (F = 7.32, P < 0.001). Community composition at the sites was also significantly correlated with elevation (R2 = 0.5, P = 0.001).
Fig. 9 NMDS of bird communities at sites on and off pika colonies in 2018

Note: Crosses represent significantly different centroid locations for categorical variables according to PERMANOVA (F = 7.32, P < 0.001). Envfit vectors are plotted for variables where the correlation was significantly different from zero (i.e. elevation, R2 = 0.5, P = 0.001) and grass cover (R2 = 0.2, P = 0.05).

The iterative nature of NMDS and the fact that the starting positions of sites are random means that multiple runs of the algorithm do not always produce the same results. In some runs, community composition was found to correlate significantly with percentage grass cover at a site (R2 = 0.2, P < 0.05), while in others it did not. Eremophila alpestris was the only species associated with both on-colony and off-colony sites. Columba leuconota was associated with off-colony sites, though it was only observed once. Syrrhaptes paradoxus was found in greater numbers at on-colony sites, but was also strongly associated with sites at higher elevation. Oenanthe deserti was also found only at on-colony sites, and was associated with sites with a greater percentage cover of grasses. Six species were significantly more abundant at on-colony sites than off-colony sites, according to Wilcoxon rank-sum tests: P. humilis (W = 72, P = 0.02), U. epops (W = 72, P = 0.02), S. paradoxus (W = 72, P = 0.02), P. ruficollis (W = 90, P < 0.001), O. taczanowskii (W = 72, P = 0.02) and B. hemilasius (W = 76.5, P = 0.01).
The SWM chosen for use in the partial db-RDA was a combination of a minimum spanning tree and a linear function (Šidák-corrected P = 0.001, corrected α = 0.009). Three Moran’s eigenvectors were selected from this matrix, all representing relatively broad-scaled patterns (the 1st, 2nd and 8th eigenvectors were considered significant). This was unsurprising, as the study site can easily be thought of as consisting of two halves, one with colony and one without (see Fig. 3). The overall partial db-RDA was statistically significant (P = 0.009), though only the first of the two axes was statistically significant (P = 0.03). That axis broadly separates the sites into two groups: on-colony and off-colony. Elevation (P = 0.047), plant species richness (P = 0.019) and colony (P = 0.004) were statistically significant. Plant richness appears to correlate with colony to some degree, but no significant difference was found in plant richness between on- and off-colony sites (W = 42.5, P = 0.6).

5 Discussion

5.1 Effect of poisoning

There was no significant difference in the average number of burrows at poisoned and untreated sites (see Fig. 5). However, burrow density showed a greater standard error at untreated sites. This could indicate that poisoning has been effective at reducing the number of sites with very high densities of pika, although this is unlikely given that pikas take only several months to return to pre-control densities after poisoning (Pech et al., 2007). We visited Yanchiwan during May 2017, and while systematic pika poisoning typically occurs there in the summer months, the pikas would likely have returned to their pre-control densities by the time of our survey. Though not the most accurate proxy measure, total burrow density does have a significant linear relationship with pika density (Wei et al., 2020b). The fact that there were untreated sites with fewer burrows than poisoned sites is likely because the poison is, according to policy, only applied when pikas reach a high density. Detailed descriptions of the timing and intensity of the poisoning programme were unavailable to us, and we were unable to make any observations as to whether local guidelines regarding poisoning were strictly followed. Despite this, given the continued presence of pika colonies in Yanchiwan for the duration of our work there over four years, we can conclude that the poisoning programme was ineffective at removing pikas in the long-term at Yanchiwan.
Fig. 10 Plot of the partial distance-based redundancy analysis (partial db-RDA) of bird communities at on-colony and off-colony sites in 2018

Note: Statistically significant vectors plotted for elevation (P = 0.047) and plant richness (P = 0.019).

Poisoning of pikas did not appear to affect the bird community structure in Yanchiwan. Poisoning was not found to be significantly correlated with differences in bird community structure, and there was no significant difference in bird abundance between poisoned and untreated sites. However, we must stress that our small sample size means that any more subtle effects on bird communities, such as through secondary poisoning or effects on the abundance of individual species, would go unnoticed. We cannot distinguish between a lack of significance in either the Wilcoxon rank-sum tests or envfit correlations caused by limited sampling and that caused by lack of a true effect. Other artefacts of the small sample size include the disparity in bird community composition between years, and the large difference in relative abundance of E. alpestris.

5.2 Effect of pika colonies

While poisoning of pikas in Yanchiwan did not significantly alter bird communities, sites with no pika colonies supported a much more impoverished bird community than those with pika colonies. Except for a single observation of snow pigeon Columba leuconota, sites without pikas were occupied only by E. alpestris. This could be a result of both pikas and birds selecting for similar habitat. Sites without pika were often rockier and more of them had less vegetation than sites with pika colonies, although Wilcoxon rank sum tests showed no significant difference in the distributions of any habitat variables between on-colony and off-colony sites. A principal component analysis (PCA) of continuous habitat variables revealed a gradient ranging from low altitude sites with high vegetation cover to higher altitude sites with low vegetation cover. Off-colony sites were all found clustered around the mid-point of this gradient. This gradient is present in the only habitat variables significantly correlated with community structure in the NMDS (grass cover and elevation, Fig. 9), and off-colony sites are also clustered together at the midpoint. The lack of variation in vegetation cover and elevation could be responsible for the similarity in bird communities observed off colony. However, the partial-dbRDA indicated that presence of pika colonies, plant species richness, and elevation all had significant effects on community structure despite the spatial clustering of sites.

5.2.1 Passerines

Pyrgilauda ruficollis and O. taczanowskii were both observed in significantly greater numbers at sites on colony than those off colony, which reflects previous results from other authors (Arthur et al., 2008; Lai and Smith, 2003). These birds are cavity-nesters and build their nests entirely in burrows constructed by small mammals, most often pikas but also voles (Lu et al., 2009). They were also observed in greater numbers at sites that had not been poisoned, although this difference was not significant according to the Wilcoxon rank sum test and could not be separated from the variation caused by other factors by the NMDS. The common anticoagulants bromadiolone and brodifacoum have been implicated in the deaths of passerines elsewhere (e.g., Zahler et al., 2004), although there is not yet any evidence of this occurring in China (Liu, 2019), or in ways that could result in population declines outside of total eradication of the pikas.

5.2.2 Birds of prey

Buteo hemilasius and F. cherrug were only observed at sites with pika colonies. B. hemilasius was observed significantly more frequently at sites with pikas. These findings are unsurprising as pikas form the primary part of the diets of these two species on the QTP (Li et al., 2004; Dixon et al., 2015). Previous work has shown a significant decrease in the abundance of several species of birds of prey at sites where pikas have been poisoned, including F. cherrug and B. hemilasius (Lai and Smith, 2003; Badingqiuying et al., 2016). The QTP is home to a sizeable breeding population of Endangered saker falcons and it is also an important wintering area for individuals migrating from further north in Russia and Mongolia (Dixon et al., 2015). Research has shown that saker falcons preferentially choose habitat with a high degree of grassland cover, habitat which also supports denser pika colonies (Dixon et al., 2017). While the effects of complete pika eradication on birds of prey are known to some extent (Lai and Smith, 2003; Badingqiuying et al., 2016), the indirect effects of poisoning with common anticoagulants or botulinum toxin C in China are unknown. However, there are anecdotal reports of raptor deaths occurring after poisoning of pikas with botulinum C (Badingqiuying, 2016; Liu, 2019).

5.2.3 Pseudopodoces humilis

Pseudopodoces humilis was observed only at on-colony sites in 2018 and was more abundant at untreated sites with a greater number of pika burrows in 2017. This is similar to the results from Lai and Smith (2003), who observed no P. humilis at sites where pikas had been removed. These authors attributed this to the fact that P. humilis nests in pika burrows, although this is incorrect because P. humilis excavates its own burrows and rarely uses those of other species (Ke and Lu, 2009; Lu et al., 2011). Interestingly, the pattern observed in this study and by Lai and Smith (2003) has been reported by other authors, indicating an overlap in the habitat used by the two species (Londei, 1998). The precise reason for this overlap is currently unclear, although Li et al. (2018) report an increased abundance of P. humilis in sites where grasslands were degraded due to overgrazing, a trend also reported in plateau pikas (Harris et al., 2015). Unlike in Lai and Smith’s (2003) study, the abundance of P. humilis we observed was not significantly lower at poisoned sites according to the Wilcoxon rank-sum test, although it was significantly lower at off-colony sites than on-colony sites. It was also noted that the sites where P. humilis was observed in 2017 were all at a similar elevation and had similar vegetation structure in terms of grass cover. These sites were spatially closer together, indicating a potential for spatial autocorrelation. For this reason, we cannot claim that the poisoning of pika reduces numbers of P. humilis at Yanchiwan, although we confirm observations of an overlap in habitat use by the two species.

5.2.4 Upupa epops

Hoopoe, Upupa epops, were observed significantly more frequently at sites on colony than off colony. They are not normally thought to associate with burrowing mammals. Hoopoes forage for invertebrates in the ground (Krištín and Kirwan, 2020) and are likely attracted to areas with softer soils where burrowing is also easier for pikas. While some burrowing animals increase the abundance and diversity of arthropods in arid environments (Davidson and Lightfoot, 2008; Buyandelger et al., 2021), as of yet there have been no studies on this effect in pika colonies.

5.2.5 Syrrhaptes paradoxus

Pallas’s sandgrouse, S. paradoxus, has not been associated with burrowing mammals in previous work. Here we found a significantly higher abundance at sites on pika colonies than off. However, we suspect that this is due to the flocking behaviour of S. paradoxus, which leads to large observations (e.g., 18 individuals at one site) that do not accurately reflect the abundance caused by any of the environmental factors measured in this study. Sandgrouse were associated with high elevation sites with little grass cover, indicating they could be responding to environmental factors other than the presence of pikas. Sandgrouse are frequently found in arid semi-steppe with sparse vegetation (de Juana and Boesman, 2020), and the sites where they were observed in this study all matched this description.

5.3 Limitations and future research

NMDS is not a statistical tool used for understanding causation. It is more useful as an exploratory tool for revealing patterns that can be investigated later. In this study, we found interesting patterns, but caution against interpreting them as strong evidence for pikas being solely responsible for bird community structure. For example, it is possible that birds and pikas are selecting for similar habitat e.g., areas where burrowing and foraging are easier. Future studies could take an experimental approach involving pika removal and measurements of more habitat variables to untangle these complex patterns. Future research should also investigate the three levels of treatment together: poisoned colony, untreated colony, and off colony, to make more useful comparisons.
Many of the actual processes of niche expansion which lead to the facilitation of bird communities by plateau pikas remain unknown and were beyond the scope of this study. For example, while we know that they provide a source of prey for raptors and breeding sites for cavity nesters, we do not know how they alter food resources for insectivorous birds, or how their modifications of vegetation structure and community composition could impact bird communities. The spatially clustered nature of pika colonies, especially in areas of lower density (Wei et al., 2020a), could also lead to areas of increased competition between bird species.
Pika colonies provide a useful opportunity to study the processes driving community composition. So far, as with most other putative ecosystem engineers (e.g., Coggan et al., 2018), most studies have focussed on simple mensurative studies comparing community composition between sites with and without pikas. Advances in both applied conservation problems and theoretical questions about community composition could be made by examining the interactions between both pikas and co-occurring species, and among the different co-occurring species themselves.

6 Conclusions

This is the first study to investigate the engineering effects of pikas outside the Kobresia steppe. We found no significant difference in bird community structure between poisoned and untreated colony sites, and no significant difference in the relative abundance of birds. However, the small sample size means we are unable to claim conclusively that this is evidence of no effect. We found significant differences in bird community structure and abundance between on-colony and off-colony sites, with only one species (E. alpestris) being observed with any frequency off colonies. This suggests that pikas play a vital role in structuring bird communities in habitats where resources are scarce, such as the desert shrublands and arid montane steppe found in Yanchiwan. However, more research is needed to provide mechanistic explanations for these changes.


The poisoning programme in Yanchiwan is not producing large enough effects on bird community structure that they could be ascertained by our small sample size. However, if poisoning efforts were to increase to a degree that pikas were eradicated at sites, this would have a drastic effect on the bird communities.
The authors wish to thank all the staff at Gansu Yanchiwan NNR, especially Wuliji, Dabuxilite, and Dou Zhigang. Ma Bing, Sun Yakuan, Gao Jian, Peng Xiaoxu, Wang Jiahui, Li Deng, Wang Jun, and Sydney Greenfield all provided valuable assistance collecting data and in the field. Robyn Geldard gave helpful comments on several early drafts of this manuscript.


The authors wish to thank all the staff at Gansu Yanchiwan NNR, especially Wuliji, Dabuxilite, and Dou Zhigang. Ma Bing, Sun Yakuan, Gao Jian, Peng Xiaoxu, Wang Jiahui, Li Deng, Wang Jun, and Sydney Greenfield all provided valuable assistance collecting data and in the field. Robyn Geldard gave helpful comments on several early drafts of this manuscript.
Anderson M J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology, 26(1): 32-46.

Anderson M J, Walsh D C I. 2013. PERMANOVA, ANOSIM, and the Mantel test in the face of heterogeneous dispersions: What null hypothesis are you testing? Ecological Monographs, 83(4): 557-574.


Arthur A D, Pech R P, Davey C, et al. 2008. Livestock grazing, plateau pikas and the conservation of avian biodiversity on the Tibetan Plateau. Biological Conservation, 141(8): 1972-1981.


Badingqiuying. 2016. Pikas, grasslands, and pastoralists: Understanding the roles of plateau pikas in a coupled social-ecological system. Diss., Phoenix, USA: Arizona State University.

Badingqiuying, Smith A T, Senko J, et al. 2016. Plateau pika Ochotona curzoniae poisoning campaign reduces carnivore abundance in southern Qinghai, China. Mammal Study, 8: 1-8.

Bagchi S, Namgail T, Ritchie M E. 2006. Small mammalian herbivores as mediators of plant community dynamics in the high-altitude arid rangelands of Trans-Himalaya. Biological Conservation, 127(4): 438-442.


Bauman D, Drouet T, Dray S, et al. 2018a. Disentangling good from bad practices in the selection of spatial or phylogenetic eigenvectors. Ecography, 41(10): 1638-1649.


Bauman D, Drouet T, Fortin M J, et al. 2018b. Optimizing the choice of a spatial weighting matrix in eigenvector-based methods. Ecology, 99(10): 2159-2166.


Bird Life International. 2009. Directory of important bird areas in China (Mainland): Key sites for conservation. Cambridge, UK: Bird Life International.

Bird Life International. 2017. The IUCN red list of threatened species 2017. DOI: 10.2305/IUCN.UK.2017-1.RLTS.T22696495A110525916.en.


Blanchet F G, Legendre P, Borcard D. 2008. Forward selection of explanatory variables. Ecology, 89(9): 2623-2632.


Borcard D, Legendre P, Drapeau P. 1992. Partialling out the spatial component of ecological variation. Ecology, 73(3): 1045-1055.


Bruno J F, Stachowicz J J, Bertness M D. 2003. Inclusion of facilitation into ecological theory. Trends in Ecology and Evolution, 18(3): 119-125.


Buyandelger S, Enkhbayar T, Otgonbayar B, et al. 2021. Ecosystem engineering effects of Mongolian Marmots (Marmota sibirica) on terrestrial arthropod communities. Mongolian Journal of Biological Sciences, 19(2): 17-30.


Coggan N V, Hayward M W, Gibb H. 2018. A global database and “state of the field” review of research into ecosystem engineering by land animals. Journal of Animal Ecology, 87(4): 974-994.

Davidson A D, Lightfoot D C. 2008. Burrowing rodents increase landscape heterogeneity in a desert grassland. Journal of Arid Environments, 72(7): 1133-1145.


de Juana E, Boesman P D F. 2020. Pallas’s Sandgrouse (Syrrhaptes paradoxus). In: del Hoyo J, Elliott A, Sargatal J, et al(Ithaca, eds.). Birds of the world. USA: Cornell Lab of Ornithology.

Delibes-Mateos M, Delibes M, Ferreras P, et al. 2008. Key role of European rabbits in the conservation of the western Mediterranean Basin hotspot. Conservation Biology, 22(5): 1106-1117.


Dixon A, Li X, Rahman L, et al. 2017. Characteristics of home range areas used by Saker Falcons (Falco cherrug) wintering on the Qinghai-Tibetan Plateau. Bird Conservation International, 27(4): 525-536.


Dixon A, Ming M, Batbayar N. 2015. Importance of the Qinghai-Tibetan Plateau for the endangered saker falcon falco cherrug. Forktail, 31: 37-42.

Dray S, Bauman D, Jombart T, et al. 2020. Package “adespatial”.

Dray S, Legendre P, Peres-Neto P R. 2006. Spatial modelling: A comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecological Modelling, 196(3-4): 483-493.


Faith D P, Minchin P R, Belbin L. 1987. Compositional dissimilarity as a robust measure of ecological distance. Vegetatio, 69(1-3): 57-68.


Harris R B, Wang W, Badingqiuying, et al. 2015. Herbivory and competition of Tibetan steppe vegetation in winter pasture: Effects of livestock exclosure and Plateau Pika reduction. Plos One, 10(7): e0132897. DOI: 10.1371/journal.pone.0132897.


Hogan B W. 2010. The Plateau Pika: A keystone engineer on the Tibetan Plateau. Diss., Phoenix, USA: Arizona State University.

Jones C G, Lawton J H, Shachak M. 1994. Organisms as ecosystem engineers. Oikos, 69(3): 373-386.


Jones C G, Lawton J H, Shachak M. 1997. Positive and negative effects of organisms as physical ecosystem engineers. Ecology, 78(7): 1946. DOI: 10.1890/0012-9658(1997)078[1946:PANEOO]2.0.CO;2.


Ke D, Lu X. 2009. Burrow use by Tibetan Ground Tits Pseudopodoces humilis: Coping with life at high altitudes. Ibis, 151(2): 321-331.


Kinlaw A. 1999. A review of burrowing by semi-fossorial vertebrates in arid environments. Journal of Arid Environments, 41(2): 127-145.


Krištín A, Kirwan G M. 2020. Eurasian Hoopoe (Upupa epops). In: del Hoyo J, Elliott A, Sargatal J, et al(eds.). Birds of the World. version 1. Ithaca, USA: Cornell Lab of Ornithology.

Lai C H, Smith A T. 2003. Keystone status of plateau pikas (Ochotona curzoniae): Effect of control on biodiversity of native birds. Biodiversity and Conservation, 12(9): 1901-1912.


Legendre P, Anderson M J. 1999. Distance-based redundancy analysis: Testing multispecies responses in multifactorial ecological experiments. Ecological Monographs, 69(1): 1-24.


Legendre P, Legendre L. 2012. Numerical ecology:Developments in environmental modeling. Oxford, UK: Elsevier.

Li L, Fassnacht F E, Storch I, et al. 2017. Land-use regime shift triggered the recent degradation of alpine pastures in Nyanpo Yutse of the eastern Qinghai-Tibetan Plateau. Landscape Ecology, 32(11): 2187-2203.


Li L, Tietze D T, Fritz A, et al. 2018. Rewilding cultural landscape potentially puts both avian diversity and endemism at risk: A Tibetan Plateau case study. Biological Conservation, 224: 75-86.


Li L, Yi X, Li M, et al. 2004. Diet composition of upland buzzard: Analysis on stomach content and food pellet. Zoological Research, 25(2): 162-165.

Li S, Lu X. 2012. Reproductive ecology of Isabelline wheatears at the extreme of their altitude distribution. Ardeola, 59(2): 301-307.


Li S, Peng W. 2014. Nest-site selection by Isabelline wheatears Oenanthe isabellina on the Tibet Plateau. Forktail, 30: 132-133.

Li S, Peng W, Guo C. 2015. Factors affecting nest success of the oriental skylark on the Tibetan Plateau. Ornithological Science, 14(1): 47-52.


Liu N, Zhang H, Dou Z. 2010. Synthetical scientific investigation on Yanchiwan National Reserve in Gansu Provence. Lanzhou, China: Lanzhou University Press.

Liu X. 2019. Rodent biology and management: Current status, opinion and challenges in China. Journal of Integrative Agriculture, 18(4): 830-839.


Londei T. 1998. Observations on Hume’s groundpecker Pseudopodoces humilis. Forktail, 14: 74-75.

Lu X, Huo R, Li Y, et al. 2011. Breeding ecology of ground tits in northeastern Tibetan Plateau, with special reference to cooperative breeding system. Current Zoology, 57(6): 751-757.


Lu X, Ke D, Zeng Z, et al. 2009. Reproductive ecology of two sympatric Tibetan snowfinch species at the edge of their altitudinal range: Response to more stressful environments. Journal of Arid Environments, 73(12): 1103-1108.


Oksanen J, Blanchet F, Friendly M,. 2019. Vegan: Community ecology package—R package version 2.4-2.

Pech R P, Jiebu, Arthur A, et al. 2007. Population dynamics and responses to management of plateau pikas Ochotona curzoniae. Journal of Applied Ecology, 44(3): 615-624.


Root-Bernstein M, Ebensperger L A. 2013. Meta-analysis of the effects of small mammal disturbances on species diversity, richness and plant biomass. Austral Ecology, 38(3): 289-299.


Root-Bernstein M, Fierro A, Armesto J, et al. 2013. Avian ecosystem functions are influenced by small mammal ecosystem engineering. BMC Research Notes, 6: 549.


Schaller G B. 1998. Wildlife of the Tibetan steppe. Chicago, USA: University of Chicago Press.

Schroeder P J, Jenkins D G. 2018. How robust are popular beta diversity indices to sampling error? Ecosphere, 9(2): e02100. DOI: 10.1002/ecs2.2100.


Smith A, Wang X. 1991. Social relationships of adult black-lipped pikas (Ochotona curzoniae). Mammalogy, 72(2): 231-247.

Smith A T, Badingqiuying, Wilson M, et al. 2019. Functional-trait ecology of the Plateau Pika Ochotona curzoniae (Hodgson, 1858) in the Qinghai-Xizang ecosystem. Integrative Zoology, 14: 87-103.


Smith A T, Foggin J M. 1999. The Plateau Pika (Ochotona curzoniae) is a keystone species for biodiversity on the Tibetan Plateau. Animal Conservation, 2(4): 235-240.


Stachowicz J J. 2001. Mutualism, facilitation, and the structure of ecological communities. BioScience, 51(3): 235-246.


Sutherland W J, Newton I, Green R E. 2004. Bird ecology and conservation:A handbook of techniques. Oxford, UK: Oxford University Press.

Van Nimwegen R E, Kretzer J, Cully J F. 2008. Ecosystem engineering by a colonial mammal: How prairie dogs structure rodent communities. Ecology, 89(12): 3298-3305.


Wei W, He J, Zheng Q. 2020a. Plateau pikas (Ochotona curzoniae) at low densities have no destructive effect on winter pasture in alpine meadows. Rangeland Journal, 42(1): 27-43.


Wei W, He J, Zheng Q, et al. 2020b. A comparison of density estimation methods in Plateau Pika populations in an alpine meadow ecosystem. Journal of Mammalogy, 101(4): 1091-1096.


Zahler P, Dolgormaa L, Hinds L, et al. 2004. The management of Brandt’s Vole in Mongolia: Toward an ecologically based means of control. Mongolian Journal of Biological Sciences, 2(2): 51-55.


Zeng X, Lu X. 2009. Breeding ecology of a burrow-nesting passerine, the white-rumped snowfinch. Ardeola, 56(2): 173-187.

Zhao X, Zhao L, Xu T, et al. 2020. The Plateau Pika has multiple benefits for alpine grassland ecosystem in Qinghai-Tibet Plateau. Ecosystem Health and Sustainability, 6(1): 1-5