Introduction

Both large and small mammalian herbivores are keystone species in grasslands worldwide, exerting profound impacts on ecosystem structure and functions (1-6). They have co-evolved with grasslands and with each other over millions of years, but how these mammals co-exist remains poorly understood (6-10). By sharing food and other resources, guilds of different herbivore species can participate in varied interactions ranging from negative (competition) to positive (facilitation) for one or more of the species (11-14). However, such interactions often are assumed to be highly asymmetric: large herbivores have been demonstrated to affect the abundance, diversity, and demography of small mammals (15-20), whereas the potential for reverse effects has received little attention (10,21-23).

Small mammals, including pikas, voles, prairie dogs, pocket gophers, and other small herbivores, are not only primary consumers but also key ecosystem engineers in many grasslands (6,9,24-26). Through their feeding, clipping, and burrowing activities they profoundly alter vegetation and soil properties, with potential cascading effects on co-occurring large herbivores (6,10,21-23). In rangelands and other less productive ecosystems, small herbivorous mammals are often considered to be pests because their population outbreaks can lead to competition with livestock for food (27-29). Despite this, the stress gradient hypothesis predicts that the direction of interspecific interactions should shift as environmental stress changes (30). Specifically, a competition-facilitation balance between species may transition with population density: facilitation should occur when species populations are at low density, while competition should occur when densities are high; such density-transitions may promote community stability and productivity (31,32). Despite such expectations, it remains unclear whether the competitive effects of outbreaks of small herbivorous mammals on larger herbivores diminish or even shift to facilitation when the smaller species are at lower, non-outbreak levels. Livestock grazing currently uses ∼77% of global agricultural land, and sustains billions of people worldwide (33). A critical assessment of small mammals’ impacts on livestock production is urgently needed to guide management for both production and biodiversity conservation.

The Qinghai-Tibetan Plateau supports approximately 14 million yaks (Bos grunniens), forming one of the world’s most extensive pastoral systems (34). These herds are vital to pastoral livelihoods and key ecosystem functions across the region’s 2.5 million km² expanse (35). The dominant small mammal species―plateau pika (Ochotona curzoniae)―is an iconic keystone species that commonly coexists with yaks, and is the focus of long debate on whether its impact of yaks is positive or negative (9,36-38). In regions with high population densities (e.g., over 500 active burrows/ha), plateau pikas can suppress livestock production by heavily consuming nearly all plant species foraged by yaks, leading to extensive poisoning campaigns targeted at eradicating these small mammals (37-39). At low and moderate densities of pikas (e.g., below 200 active burrows/ha), however, competition for food is mitigated and dietary partitioning can occur: yaks graze selectively on monocotyledonous plants such as grasses and sedges (40,41), whereas pikas prefer to clip and feed on the leaves of dicotyledonous plants (42,43). Thus, whether plateau pikas and yaks compete for pasture largely depends on their diets and population densities.

In line with global trends in weed invasions in rangelands (44), intensive livestock grazing and other human disturbances have facilitated grassland degradation on the Qinghai-Tibet Plateau, creating opportunities for the proliferation of toxic weeds (45). Among these, the wolf poison Stellera chamaejasmehas become a predominant species, now covering approximately 20 million hectares, as it outcompetes palatable grasses and sedges while thriving under grazing pressure (46). The expansion of poisonous plants on the Qinghai-Tibetan Plateau poses a dual threat to livestock, either by causing direct toxicity (47) or by reducing forage availability through competition with palatable species for light and soil nutrients (45,46). However, plateau pikas may indirectly benefit yak foraging by selectively clipping large plants (48,49), especially those poisonous forbs such as S. chamaejasme (37) (Fig. 1)—a behavior that reduces predation risk (50,51) but also suppresses these competitors, potentially enhancing the quality and quantity of desirable grasses and sedges. Despite these plausible interactions, empirical studies directly testing such interactions remain scarce.

Figure 1.

Here, we experimentally assess how a moderate population density of pikas (∼200 active burrows ha⁻¹) affects yak performance in an alpine ecosystem of the Qinghai–Tibetan Plateau. We hypothesized that, at moderate densities, plateau pikas facilitate yaks by suppressing tall poisonous forbs, thereby increasing the abundance and nutritional quality of palatable grasses and sedges during the growing season (June–August). Specifically, we predicted that (i) pika clipping reduces poisonous plant cover; (ii) this enhances the availability and quality of palatable forage (i.e., grasses and sedges) for yaks; (iii) yaks increase their foraging efficiency in the presence of pikas; and (iv) these cascading effects ultimately improve yak weight gain (Fig. 1). To test these predictions, we first conducted two field surveys to examine diet partitioning between pikas and yaks, and the associations among pika density, Stellera abundance, and yak grazing activity. We then performed an in-situ field experiment using 150 × 150 m fenced enclosures to test the interactive effects of pikas and Stellera on yak body growth and the underlying mechanisms driving these effects.

Results

In the field surveys, we found that pikas and yaks have distinct diets: pikas very frequently clipped but did not eat large, poisonous Stellera plants and fed mostly on other forbs, whereas yaks strongly preferred grasses and sedges (Fig. 2A,B, table S1,2), supporting findings of previous studies (40-43). We also found that the cover abundance of Stellera was associated negatively with the density of active pika burrows (R² = 0.39, P <0.001; Fig. 2C) and with yak foraging activity, as indicated by dung density (R² = 0.43, P < 0.001; Fig. 2D).

Figure 2.

In the manipulative field experiment, pika removal led to a 90% reduction in the number of active burrows per hectare, with 182.9 (SE ± 24.5) burrows/ha in pika-present plots compared to 19.0 (SE ± 1.6) burrows/ha in the pika-removed plots (Fig. 3A, table S3). In the presence of Stellera, yaks co-occurring with pikas showed daily weight gains 67% greater than those in the no-pika treatment, but such effects disappeared once Stellera was removed (Fig. 3B, table S3,4). Notably, yak weight gain showed a hump-shaped relationship with pika density, with the highest growth at ca. 220 active burrows/ha, after which, yak growth decreased linearly as pika density increased (R² = 0.75, P < 0.001; Fig. 3C).

Figure 3.

In the presence of Stellera, pikas suppressed the cover abundance of this forb by two-thirds (Fig. 3D, table S3,4), which led to increases of 118% and 18% in cover abundances of co-occurring grasses and sedges, respectively (Fig. 3E,F, table S3,4). These shifts in vegetation composition also improved the nutritional value of the total available forage for yaks, with increases of 15% in crude protein (CP) content and 6% in acid detergent fibre (ADF) content in the pika + Stellera treatment compared with the Stellera only treatment (Fig. 3G,H, table S5,6). Pikas and Stellera had no interactive effects on abundance of sedges, forbs, and neutral detergent fibre (NDF) of total forage for yaks (Fig. 3F,I, fig. S1, table S3,5). These results suggest that pikas facilitate yak growth by suppressing poisonous plants, increasing both the availability and quality of food for the livestock.

Pika-yak facilitation was also linked to improved foraging efficiency in yaks when grazing alongside pikas. Bite rate (i.e., the number of bites taken on plants per hour) and bites/step ratio (i.e., the number of bites taken on plants per step) are two key indicators of foraging efficiency in large herbivores (52). In the presence of Stellera, sedge bite rate and sedge bites per step of yaks significantly increased by 48% and 89%, respectively (Fig. 4A,C, table S7,8), and grass bite rate and grass bites per step increased similarly by 41% and 80% (Fig. 4B,D, table S7,8), when yaks occurred with compared without pikas. These enhancements in yak foraging efficiency can be attributed to the decline in cover of the poisonous Stellera (Fig. 3D), which removed a grazing deterrent and improved access by yaks to palatable food items. Pikas and Stellera had no interactive effects on yaks’ foraging efficiency on forbs (fig. S2, table S7).

Figure 4.

Discussion

By combining field surveys and manipulative experiments, we have demonstrated that pikas—when occurring at moderate density―can benefit yak body growth by suppressing large poisonous plants and increasing the availability of palatable forage in the Qinghai-Tibetan Plateau. Our findings thus quantify the beneficial impacts of small mammals that, during outbreak years, are often considered as pests and competitors with livestock for food (1,37-39). These observations support predictions of a density-dependent transition in competition-facilitation interactions by Zhang (2003) (31) and provide new insights into how small and large herbivores co-exist in nature.

Ecosystem engineering is a key mechanism driving interspecific facilitation (24,53,54). Engineering activities of one herbivore species can indirectly benefit another by either increasing access to food resources or ameliorating abiotic conditions in particular habitats (11,12,40,55). In our system, the selective clipping activities of pikas greatly suppressed the abundance of large poisonous forbs (Fig. 3D), and increased both the quantity and quality of yaks’ palatable food plants (Fig. 3E-H). These improvements in food availability and nutrition for yaks can be attributed to the release of grasses and sedges from competition with the forbs for limiting above- and below-ground resources, including light, soil moisture and nutrients (45,46). Pika-induced increases in food resources, along with fewer foraging barriers for yaks after poisonous forb removal by pikas, together increased foraging efficiency (Fig. 4) and thus facilitated weight gain by yaks (Fig. 3B). In addition to increasing food resources, the removal of poisonous plants by pikas may also benefit yaks by reducing the risk of incidental consumption (45,46), which can cause biochemical or physiological stresses in yaks that lead to sickness or even death (47).

Our results add to a growing list of studies that highlight the importance of small mammal impacts on large herbivores. In grasslands of the North American Great Plains, for example, black-tailed prairie dogs (Cynomys ludovicianus) can alter growing-season forage quality and quantity both on and off their colonies, exerting either competitive or facilitatory impacts on daily forage intake rates and mass gain of cattle (10,22,23,56). By contrast, in drought years on the Mongolian mountain steppes, overlap in use of forage grasses (Stipa krylovii and Agropyron cristatum) can lead to competition for food between Mongolian pika (Ochotona pallasi) and goats, sheep, and cattle, with potential negative impacts on livestock production (57). Given the ubiquity of small mammals, and their ability to strongly modify plant and soil properties via herbivory and ecosystem engineering, the impacts of small mammals on co-occurring large herbivores may be complex, resulting in either positive or negative outcomes which would benefit from further investigations.

The classic stress gradient hypothesis predicts an increase in the intensity of facilitation as environmental conditions become increasingly stressful (30), and this has been widely documented in plant communities (58,59). In contrast to this prediction, however, we found that facilitation between herbivores tends to occur in less (i.e., low- or moderate-density of pikas), rather than in more stressful environments (during pika population outbreaks). In an examination of wild ungulate-cattle interactions in African savannas, it has also been found that herbivore facilitation is greater in wet than in dry seasons (12). Similar facilitation was also reported between migratory grazers (14), between prairie dogs and cattle (23), and between snails and caterpillars (60) in habitats with moderate, but not limited resource levels. Our results, along with these prior studies, therefore suggest a different competition-facilitation balance in herbivore communities compared to those exhibited in plant communities (58,59). These observations also highlight the importance of population density in regulating competition-facilitation transitions in species interactions (31). More studies should performed to understand how herbivore interactions may change in strength and direction along stress gradients, which is key to predicting how herbivore communities assemble under global change.

Our results reveal the importance of wild small herbivores in counteracting livestock grazing-induced vegetation imbalance in rangelands. Small herbivores have comparatively high metabolic demands and small gut capacities and often prefer dicotyledonous plants (e.g., forbs) with high concentration of nutrients (61-62). In contrast, larger herbivores such as livestock prefer monocotyledonous plants (e.g., grasses) because they can tolerate low plant nutrient contents but require a greater quantity of food (61-62). The coexistence of the small and large herbivores is therefore can lead to an “compensatory effect” on grass and forb biomass that helps to maintain a balance and diverse plant community (62), with importance consequences for ecosystem functioning and services. In our system, the widespread of Stellera across the Qinghai-Tibetan Plateau is itself a product of long-term grazing pressure, which suppresses palatable grasses and sedges while releasing unpalatable, stress-tolerant forbs (45,46). The pika’s selective clipping of forbs helps promoting graminoid recovery, benefiting not only livestock production but may also other key rangeland functions (e.g., carbon storage and nutrient cycling) (41). Such modification effects of small herbivores, especially those colonial and hyper-abundant ones (e.g., rodents and insects), on large grazers’ ecological functions may be more common and importance than previously perceived (7,55,63,64).

Collectively, our experiments provide empirical evidence that small mammals can facilitate large herbivores through altering vegetation properties. By suppressing tall poisonous forbs, plateau pikas improved forage composition and quality, increased yak foraging efficiency, and ultimately boosting yak weight gain. These findings overturn the dominant perception that small mammals act purely as rangeland pests (27-29,37-39) and reveal that small-bodied herbivores can contribute positively to pastoral systems when maintained at moderate densities. Our study examined pika–yak interactions only during the summer period, when food resources are most abundant. Whether such facilitative effects weaken or even shift toward competition under more stressful conditions—for example, when forage becomes limited during autumn or winter—remains to be tested.

Crucially, these insights carry direct policy relevance. Current rangeland management practices often rely heavily on rodenticides and other toxic compounds, leading to widespread small-mammal eradication (27-29,37-39). Our findings show that coexistence when small herbivorous mammals at low to moderate densities can enhance both livestock production and biodiversity. Recognizing and managing such facilitatory interactions between herbivore guilds supports international goals to integrate biodiversity conservation with food security and climate adaptation.

Rangeland policy should therefore move beyond pest eradication toward ecologically based management (28) that regulates herbivore populations through ecosystem processes and habitat managements. Such an approach embraces the full spectrum of herbivores—from rodents to megafauna to livestock—as contributors to multifunctional landscapes in the Anthropocene (67,68).

Materials and Methods

All pika and yak manipulations were carried out in accordance with the Law of the People’s Republic of China on the Protection of Wildlife (1988).

Study system and background

We conducted the study at a grazing grassland in Menyuan County, Qinghai Province, China (37o48’ N, 101o56’ E, 3200 m a.s.l.). The site is located in the northeast Qinghai-Tibetan Plateau and has continental cold/humid climate conditions, with a summer rainy season and a winter dry season. The mean annual temperature is 4.2°C, and rainfall was 750 mm. The vegetation is typical of alpine meadows. The grassland is dominated by sedges such as Kobresia spp. (e.g., K. humilis and K. graminifolia), subdominant species include grasses such as Elymus spp. (i.e., E. nutans), Festuca spp. (e.g., F. ovina), and Stipa spp. (i.e., S. aliena), and companion species of forbs such as Potentilla spp. (e.g., P. anserina and P. multifida) and Medicago ruthenica. In recent decades, the poisonous Stellera chamaejasme forb, commonly named wolf poison, which is toxic to livestock, has encroached and has become a dominant plant owing to human disturbances and climate changes, competing with forage plants for shared resources such as light, soil water, and nutrients (45,46).

The plateau pika (Ochotona curzoniae) is a small (body length ca. 120-190 mm, and body weight ca. 110-170 g) lagomorph endemic to and dominant in the alpine meadows of the Qinghai-Tibetan Plateau (69). When populations outbreak, pika population densities can reach over 500-1000 active burrows per hectare (37-39). However, our study grassland hosts a moderate density (ca. 100-300 active burrows per hectare) of plateau pikas during the forage growing seasons (June to August), providing us with an ideal opportunity to investigate the effects of non-outbreak pika density on yak performance. Pikas live in social groups, breed during the warm summer and have lifespans about 120-250 days (69). The plateau pika is a keystone species due to the feeding, forb-clipping, and burrowing activities of individuals that exert profound effects on soils, plants, and other animals in the meadows (70). By selectively feeding on and clipping large dicotyledonous plants such as forbs (42,43), pikas occupy a different plant-food utilization niche compared with domestic livestock such as yaks and sheep, which often prefer monocotyledonous plants such as grasses and sedges (40,41). Besides pikas, other small mammals such as rabbits and zokors occur rarely in the area.

Tibetan yak (Bos grunniens) is the major ruminant species on the Tibetan rangelands due to the species’ excellent adaptability and production performance (34). It is estimated that about 14 million domestic yaks live on the Tibetan Plateau, providing local herdsmen with daily necessities like meat, milk, wool, skins, fuel and economic benefit (35). In our study site, the grassland has been managed by pastoralism of domesticated yaks for years, and the grazing intensity is mainly controlled by pastoral practice.

Field survey #1: Diet preferences of pikas and yaks

In July (peak growing season) 2021, we investigated the diet/clipping selection of pikas and yaks in the study site. For pika feeding/clipping preferences, we randomly selected 10 2 m × 2 m plots separated by at least 100 m from each other in the study site. A large cage enclosure 1.5 m high and 2 m × 2 m bottom surface area, covered with a 5 × 5 mm plastic mesh window screen, was assigned to each plot. We then captured pikas nearby and placed one adult into each enclosure, allowing each animal to freely feed/clip plants within the cage for 20 min before it was released. We then laid out a 2 m linear transect in each cage plot consisting of 10 0.2 × 0.2 m quadrats. If vegetation had been consumed (i.e., plant tissues were removed and digested) or clipped (i.e., plant tissues were cut down and lay on the ground surface), we assigned that quadrat a value of one for that vegetation group (i.e., Stellera, sedges, grasses, and forbs); if there was no sign of consumption, a value of zero was assigned. Values assigned for each vegetation group were summed for the transect and divided by 10 to obtain a frequency of feeding/clipping use ranging from 0% to 100% (64). To document yak diet preferences, we placed 10 1 × 1 m quadrats on the ground at approximately 20 m intervals along 10 250 m transects that were randomly located on fresh grazing paths of yaks. We recorded and calculated how frequently different plant groups were grazed by yaks using the same methods as above.

Field survey #2: Associations among pikas, poisonous plants, and yak activities

In July 2021, we investigated the potential ecological interactions among pikas, poisonous plants, and yaks in the study site. We firstly selected 30 10 m × 10 m plots separated by at least 200 m from each other in the study site. The site was grazed by yaks at low to moderate intensity (i.e., 0.5-1.5 animal units/ha), with varying abundance of pikas and wolf poison Stellera forbs. Within each plot, we then assessed the abundance of pikas and Stellera, and foraging activities of yaks. We visually counted the number of active burrows (hole entrances characterized by clear openings, fresh soil or pika feces) to indicate pika abundance. For poisonous plant abundance, we visually estimated the percentage of the ground surface covered by Stellera in four 1 m × 1 m quadrats within each plot. For yaks, we recorded the number of dungs present in each plot, which is regarded a good measure for assaying grazing pressure in grasslands (71).

Field manipulative experiment: Interactive effects of pika and poisonous plants on yak performance

Experimental design.―In May 2022, we established four replicate blocks of experimental plots, for a total of 16 plots in a large area with similar plant community composition and pika densities (table S9). The site had not been disturbed by human activities (e.g., grazing or mowing) for two years prior to the initiation of the study. Each block had the following 2 × 2 factorial design: presence of pikas and poisonous plants, pikas only, poisonous plants only, and where neither pikas nor poisonous plants was present. Plot treatments were randomly assigned within each block. Minimum distances between the four replicate blocks of plots were 200 – 300 m. Each of the four plots in a replicate block was separated by 50 m, and each plot was 150 × 150 m. To avoid edge effects, we sampled plant, pika, and yak variables within the 100 × 100 m area at the center of each plot.

At the start of experimental treatments each year, we obtained 32 yak steers aged 2 years and weighing 115 kg ± 7.8 (SD) from adjacent and/or nearby pastures, and randomly grouped them into 16 herds of two yaks each. We then randomly allocated these yak herds to the 16 experimental plots (one herd/plot), creating a light to moderate grazing intensity recommended by local government guidelines, which often allow yak to maximize their growth rates (72). Electric fencing was used to confine yaks within the experimental plots. To align with their local grazing habits, yaks were allowed to graze daily in the experimental plots between 08:00 and 18:00, after which the animals were removed from the plots and housed in shelters overnight without feeding. All yaks had free access to fresh water and a mineral-lick block (Cangzhou Leysin Biotechnology Co., Ltd, Cangzhou, China) during the experimental period. Yak grazing under this regime continued for two growing seasons (June to August) in 2022 and 2023.

Pika treatments.―We installed exclosures to control pika populations in the plots. Initially, pikas were present on all plots, and the treatments with pika absence were implemented by removing pikas and preventing recolonization by fencing using an iron sheet around the perimeter of the plots. The iron sheet extended 0.60 m aboveground to prevent pikas jumping in or out, and was buried 1.5 m below the soil surface to deter animals from burrowing underneath.

In May 2022, after the establishment of the iron sheet fences, the pika exclusion treatments were initiated by removing pikas from the allocated experimental plots. Pikas were trapped using live traps and relocated elsewhere in the study site. Pika activities (e.g., number of active burrows) in the experimental plots were monitored monthly (June to August) thereafter, and pikas entering the exclosure plots were removed as necessary to maintain the exclusion treatments.

Poisonous plant treatments.―For the experimental plots without poisonous plants, we clipped and removed Stellera forbs using garden clippers. To simulate the clipping activities of pikas, we clipped only those large Stellera forbs with a height exceeding 20 cm, as these are often preferred by pikas and can exert significant impacts on the plant community and on yak grazing behaviors (Z.Z., field observations). We removed these poisonous plants once a month from June to August. For the experimental plots with poisonous plants, Stellera forbs were left intact.

Yak body growth and grazing activity.―During the growing seasons (June to August) in 2022 and 2023, we recorded the initial and final body weights (Weighbridge, Shanghai Jiujin Electronics Apparatus Co. Shanghai, China) of yaks each month to calculate their average daily weight gain. Each month, we also observed the grazing activities of each yak for six 2-hour focal periods. During these observations, we recorded the number of bites that the yaks took on different forage plant groups (i.e., sedges, grasses, and forbs) and the number of steps they took. The foraging efficiency of yaks on each plant group (bites/step) was calculated as the number of bites on specific plant group/the total number of steps during the observations (52).

Forage quantity and quality.―From June to August, along with the measurements of yak behavior and attributes as described above, we assessed how the food resources of yaks changed in the experimental plots each month. To assess forage quantity, we randomly assigned ten 1 × 1 m quadrats spaced at least 10 m from each other, and recorded the percentage of the ground surface covered by each plant group (i.e., Stellera, sedges, grasses, and forbs). To assess forage quality, forage samples were collected to quantify their nutritive values. To obtain a sample of forage apparently consumed by yaks, we tracked and observed yak foraging activities along their grazing paths, and used the hand-plucking technique to collect the corresponding plants species and tissues that were consumed by animals (73). Forage sub-samples from different plant species and tissues were pooled into a single forage sample within each plot, which was then dried at 60 ° C for 48 h in a forced-air drying oven, milled using a 1-mm mesh, and stored in plastic bags for subsequent analyses.

We analysed forage samples for crude protein (CP), acid detergent fibre (ADF), and neutral detergent fibre (NDF). Total nitrogen content was determined by the Kjeldahl method (2300; Foss Tecator AB, Hoganas, Sweden) using selenium as the catalyst, and CP was calculated as 6.25 × nitrogen. ADF and NDF were analysed with an automatic fibertec apparatus (M6, Foss Tecator AB, Hoganas, Sweden) by the method of Van Soest et al. (1991) (74).

Statistical analyses

All data were analyzed using linear models, generalized linear mixed models, or generalized additive models, with the choice of model and statistical family guided by the structure and distribution of the data. Post-hoc comparisons were conducted only when the pika × Stellera interaction term was significant. For the 2021 field surveys, we fitted generalized linear mixed models with plot and month as random effects. We then used generalized additive mixed models for the cover abundance of Stellera and active pika burrow density, with plot as a random effect, and linear regression models for dung density and Stellera cover. For the field manipulation experiments in 2022 and 2023, we constructed generalized linear mixed models with the dependent variables regressed against the interactive effect of pika and Stellera treatments, while including block, year, and month as random effects to capture the hierarchical structure of the data. Models assumed Gaussian, beta (for proportions), or Tweedie (for non-normal data) distributions, selected based on data type and model fit. A significance threshold of P = 0.05 was applied, with Tukey-HSD or Sidak post-hoc tests used where appropriate. All data management, modeling, and visualization approaches were carried out in R, with dependencies managed using renv. The main modeling packages were glmmTMB and mgcv, with DHARMa used for model diagnostics. Data management relied on the tidyverse suite of packages. A complete record of package versions is available in the renv.lock file in the repository Zenodo (75): https://doi.org/10.5281/zenodo.18290921.

Supplementary Materials

Figure S1.

Figure S2.

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Table S6.

Table S7.

Table S8.

Table S9.

Data availability

A complete record of package versions is available in the renv.lock file in the repository Zenodo (75): https://doi.org/10.5281/zenodo.18290921.

Additional information

Funding

National Natural Science Foundation of China (32371587)

• Zhiwei Zhong