The effect of grazing on winter survival of midday gerbil (Meriones meridianus) of different genders

Abstract The objective of this study was to investigate the effects of grazing on midday gerbil (Meriones meridianus) population characteristics and survival of animals of different genders. The experiment used a randomized complete block design and was conducted in Alxa Left Banner, Inner Mongolia, China, in 2002 (The agricultural reclamation plots set up in 1994). From April 2006 to October 2010, midday gerbils were live‐trapped in 3 light grazing plots, 3 overgrazed plots, and 3 grazing exclusion plots. The quantity of vegetation was investigated in the two different grazing intensity areas and grazing exclusion area to determine the relationship between gerbils and plant food availability. The results suggested that there was higher gerbil density, individual body mass, and daily body mass growth rate in the grazing exclusion sites than the other sites across the whole year. Females had higher survival in grazing exclusion areas than in other treatments, but the males’ survival showed the opposite pattern. Our results indicated that grazing negatively influenced the midday gerbil population by reducing food availability. Grazing influenced the survival rates of male midday gerbils positively, but had negative effects on females. The reason for gendered differences in survival rates of midday gerbils requires further investigation.


| INTRODUC TI ON
Livestock grazing is the most common land use worldwide (Kemp et al., 2013). Grazing induced changes in plant communities and soil properties affect the dynamics of small-mammal populations and communities (Fuhlendorf et al., 2010;Keesing, 1998a;Li et al., 2016;Saetnan et al., 2012;Steen et al., 2005;Wang et al., 2019). For example, wild and domestic ungulate grazing may affect small mammals through changing grassland ecosystems and food availability (Keesing, 1998b;Milchunas et al., 1998). Food availability, for example, plays a key role in the fitness, home range size, predation risk, physiology, behavior, life history, and overwinter survival of small mammals (Brown et al., 2004;Liu et al., 2011;Schoepf et al., 2015).
The food availability hypothesis posits that grazing changes food plant availability and food plant quality for small herbivorous mammals and thus may affect their population size, survival, body mass, and reproduction (Jon et al., 1993;Keesing, 1998b;Li et al., 2016;Schmidt et al., 2005). However, in arid ecosystems, grazing may produce also positive effects on rodents depending on their ecology, and in a relatively dry environment, grazing will drive the rodent community to develop into a community with fewer dominant species (Jones et al., 2003;Tabeni & Ojeda, 2005). So the grazing affect the life of rodent may influenced by many external condition and different object. Most studies have assessed the responses of rodent density or population growth rates to grazing (Cockburn & Lidicker, 1983;Klemolaet al., 2000;Rosi et al., 2009), while few studies have investigated the effects of grazing on demographic rates such as survival probabilities of small mammals with different genders (Keesing, 1998b;Korslund & Steen, 2006;Yuan et al., 2018).
Because food is scarce, the environment is harsh, and winter is a difficult period for mammalian nonhibernators in high northern latitudes (Coltrane & Barboza, 2010;Solonen, 2006). Mammals cope with this predictable period of food scarcity by caching food in the late fall and/ or hibernating (Geiser & Ruf, 1995;Turbill et al., 2011). Food caching is a widespread behavioral adaptation used primarily by nonmigratory and nonhibernating species to store food for future use during periods of low resource availability or uncertain environmental dynamics (Sutton et al., 2016). A caching specie relies on stored food for survival during periods of limited food availability and, in some cases, for reproduction, and food quality could have major downstream effects on fitness and population dynamics (Sutton et al., 2016). Caching species can generally be divided into two classes based on the cache duration: "short-term hoarders" and "long-term hoarders" (Vander Wall, 1990).
The midday gerbil (Meriones meridianus) is an example of the latter class. This caching species engages in intense periods of caching, usually in the late summer or fall (Degange et al., 1989;Jansson et al., 1981;Vander Wall, 1990). Both autumn and winter food availability may play crucial roles in the overwinter survival of midday gerbils.
Midday gerbil spans the major arid and semiarid biogeographic regions in northern of China and northwest China as far west as the Caspian Sea and occur in open desert habitats on stabilized, semistabilized, or nonstabilized sands (Luo et al., 2000;Rogovin, 2007;Shenbrot et al., 1999). The range of the midday gerbil covers a large region of the Eurasian arid zone. This superspecies is characterized by remarkably high geographical variation (Gromov & Erbajeva, 1995;Heptner, 1968). The type specimen was produced in the lower reaches of the Urals in Kazakhstan (Pallas, 1773). Midday gerbil is a small-sized desert dewelling rodent. The average weight of males was 52.26 ± 1.00 g and females was 50.51 ± 0.98 g (Zhang et al., 2017). The duration of pregnancy in uncontrolled laboratory conditions was found to last 24-27 days, and new-born gerbils weigh about 2.45 g (Özkurt et al., 2001). It has a mixed diet of seeds, insects, and green plant parts, but seeds dominate in the diet (Shenbrot et al., 1999;Rogovin, 2007). The reproduction period of midday gerbil, which occurs two or three times per year, is from late March to early November in our study area, and breeding activity reaches its maximum in July (Jin et al., 2009;Yang et al., 2017;Zhou et al., 1999). In general, a lower population density occurs in late autumn, and is higher in spring (Zhou et al., 2012). Thus, the breeding period may start in a season without new food resources, and both the food stored and vegetation situation may play a key role in its first breeding activity every year. Midday gerbils mainly feed on seeds, leaves, stems and insects, and food composition may vary depending on conditions in different microhabitats. Midday gerbils engage in food caching behavior in late autumn, but since this is insufficient to provide food for the whole winter, they go out of their borrows in winter (Song & Liu, 1984). Reproductive activity of midday gerbils starts in early spring when the environment could not provide any new foods, so the amount of stored food in winter will affect reproduction. In some species, females will reduce reproductive costs in harsh external conditions to ensure survival over the winter (Doonan & Slade, 1995). This specie is typically nocturnally active, and we never caught any specimens during the day when we live traps were set at day and night.
The purpose of our study was to test how livestock grazing affects the density and body mass of midday gerbils in early spring and late autumn, and the overwinter survival of gerbils. In particular, we were interested in whether there are gendered differences in these effects. We proposed three hypotheses. First, compared to overgrazed and lightly grazed areas, grazing exclusion should induce higher male and female gerbil population density both in early spring and late autumn. Second, compared to overgrazed and lightly grazed areas, grazing exclusion should induce higher body mass and body mass daily growth rate in both male and female gerbils. These effects could occur because grazing can affect plant productivity and ecosystem functions (Wang et al., 2019). Because females show a positive correlation between survival probability and body mass while males do not (Korslund & Steen, 2006), our third hypothesis was that, compared to overgrazing and lightly grazed areas, grazing exclusion would induce higher winter survival rates of gerbils.  (Yuan et al., 2018). Midday gerbil, northern three-toed jerboa (Dipus sagitta), and Mongolian five-toed jerboa Corsac fox (Vulpes corsac) are also present in the system (Wu & Fu, 2005;Xu, 2013).

| Study area
This sample areas had been established at the beginning of this century. The experiment adopted a randomized block design with 3 blocks and 4 treatments (i.e., light grazing, overgrazing, grazed exclusion, and land reclamation) to study the dynamics of desert small mammals community under different grazing intensity. Each block is 240 ha, and each treatment unit is 60 ha. The agricultural reclamation plots previously had plant species similar to the ungrazed plots but were reclaimed in 1994 by planting saxaul (Haloxylon ammodendron), sunflowers, and maize. Before the study areas were set up (prior to 2002), every treatment blocks experienced the same grazing events. The treatments assessed in this study were overgrazing, light grazing, and grazing exclusion only, which were established with standard sheep fencing (110 cm high). In the overgrazing sites, sheep grazing intensity was close to prevailing grazing intensity in the local areas and was controlled within the range of 3.75 to 4.23 sheep per ha. In the light grazing sites, sheep grazing intensity was in line with Inner Mongolia government standards and was controlled within the range of 0.83 to 1.00 sheep per ha. A 7 × 8 gerbil trapping grid (0.96 ha) at a 15-m intertrap distance was established at the center of each plot (60 ha). One wire-mesh live trap (42 cm × 17 cm × 13 cm) was placed at each trap station (Yuan et al., 2018). A wooden protective box (15 cm × 7 cm × 10 cm) is placed inside each live cage to protect rodents from natural enemies, precipitation, or low temperatures.

| Trapping of gerbils
We live-traped midday gerbils from April to October, October 2006 to October 2010, on 4 consecutive days at 4-week intervals. We did not trap during winter (from November to March) due to the low temperature. Traps were baited with fresh pignuts and checked twice (morning and afternoon) each day. Each captured individual was sexed and marked with a passive integrated transponder (PIT) tag with a unique identification number (ID) injected under the pelage. The sex, capture station, body mass, and reproductive condition of each captured individual were recorded. Males were considered in reproductive condition if they had scrotal testes. Females were considered reproductive if they possessed enlarged nipples surrounded with white mammary tissue, or a bulging abdomen. We classified gerbils as juveniles if they weighed <26 g and adults (subadult) if they weighed ≥26 g (Zhang et al., 2017), and we only captured 5 juvenile midday gerbils during the entire experiment. Therefore, in the data analysis of this study, we ignore the difference between larvae and adults. In order to minimize the impact of cage capture on midday gerbils, a wooden thermal insulation box was placed in the live cage to avoid the danger of midday gerbils from natural enemies, low temperature, and precipitation. The time to check the cage in the morning is generally before 6 o'clock in the morning.
Other experiments (infrared camera monitoring) in the same study sites found that the gerbils continue to be active after 6 o'clock, so it will reduce the impact of the experiment on the gerbils. Snap trapping is a traditional survey method for long-term studies of rodent populations and does not affect population dynamics (Christensen & Hörnfeldt, 2003;Hörnfeldt, 2004).
In all experiments, we carried out animal care and treatment in accordance with the guidelines issued by the Ethical Committee of Inner Mongolia Agricultural University. The committee requires that all researchers and students related to wildlife and experimental animals are certified in accordance with the requirements of the Institutional Animal Care and Use Committee of the Institute of Zoology, Chinese Academy of Sciences (IOZ11012).

| Vegetation sampling
Vegetation sampling was carried out monthly from April to October during 2006 to 2010. We randomly placed 3 100-m 2 square sampling plots in each treatment unit to sample shrubs and randomly placed 3 1-m 2 quadrats in each 100-m 2 square plot to sample grasses and forbs. We estimated aboveground standing biomass of shrubs, grasses, and forbs by species (Yuan et al., 2018). An additional experiment on feeding behavioral observations of midday gerbils was conducted in 2017 to determine the food resource from all plants (Table S1). Preferred foods and potential food resources were divided according to the preference index (PI) (Batzli & Pitelka, 1983).
Plant species from food resources were chosen to calculate total food biomass (TFB) and preferred food biomass (PFB) ( Table S1).

| Statistical analyses
The Cormack-Jolly-Seber (CJS) models were used as implemented in the program MARK 6.0 to estimate monthly apparent survival probabilities of the midday gerbils (White & Burnham, 1999).  (Table S2). Thus, we used quasi-Akaike's information criterion corrected for small sample size (QAICc) to select the most parsimonious and competing models of midday gerbil monthly survival (Burnham & Anderson, 2002). The most parsimonious model is the model with the lowest QAICc among all candidate models considered. A model with ∆QAICc < 2 was considered as a competing model (Burnham & Anderson, 2002). In a preliminary MARK analysis of the capturerecapture data by treatments, we found that the candidate models had gender effects (Table S3). Therefore, we assumed that the local survival probabilities of individuals would differ between the sexes. We built 15 candidate models including all possible models for the effects of treatment, time, and their interaction on survival probabilities and all possible models of the effects of time, treatment, and time-treatment interaction on recapture probabilities.
If a model including treatment effects on survival probabilities was the most parsimonious model or a competing model, we concluded that sheep grazing significantly affected survival of the midday gerbil. A ninety-five percent confidence interval (CI) was provided for survival probability. Nonoverlapping CIs were considered to be significantly different .
We indexed the late autumn (October) and early spring (April) population density of midday gerbils from 2006 to 2010 using the minimum number known alive (MNKA) (Krebs, 1966;Hilborn, Redfield, & Krebs, 1976). All of the male and female adults in both grazing and nongrazing treatments were compared by body mass in late autumn. Daily proportional body mass growth rate (DPBMGR, i.e., growth rate) of a midday gerbil in late autumn was calculated as the difference in body mass between two successive trapping occasions divided by the initial body mass at the first trapping occasion and the number of days between the two trapping occasions (Agrell et al., 1992). All the data involved in this paper have been Shapiro-Wilk tested, in which body weight (p > .05, n = 30), growth rate (p > .05, n = 28), and population density (p > .05, n = 30) all conform to the normal distribution. In addition, the sex ratio, total food biomass (TFB), and preferred food biomass (PFB) have been normalized (log(n + 1)) prior to the analysis.   Table 1). The main reason for this result is that the population of midday gerbils is significant higher in 2008 than in other years. However, we did not detect a sex difference in gerbil abundance both in late autumn and early spring (p > .05, Table 1). Grazing exclusion significantly enhanced the gerbil population compared to the light and overgrazing treatments in both late autumn (F 2,27 = 5.77, p = .011) and early spring (F 2,27 = 5.40, p = .008; Figure 2). The light grazing treatment was not significantly different from the other two areas in late autumn.

| Body mass and daily proportional body mass growth rate
Grazing treatments influenced late autumn body mass and daily proportional body mass growth rate of adult midday gerbils (F 2,72 = 14.00, p = .001; F 2,21 = 1.602, p = .011; Table 1). Midday gerbils in grazing exclusion had a significantly higher body mass than in light grazing sites and overgrazing sites (F 2,72 = 52.80, p = .023; Figure 3a). Midday gerbils in the grazing exclusion sites had higher body mass growth rates than in overgrazing sites before overwintering (F 2,21 = 3.39, p = .026; Figure 3b). There was no difference in body mass or body mass growth rate (p > .05, Table 1) between the two genders. There is no interyear difference between individual body mass (p > .05, Table 1). When analyzing the daily proportional body mass growth rate, most years have data missing (Especially in the overgrazing plot), so the impact of the year on body weight growth is not analyzed.

| Survival probability
There were gender effects in the competing models in all three sites (Table S2), so we analyzed the effect of grazing on survival by gender. Almost all competing models included the effects of grazing on both males and females (Table 2). In all of the grazing treatments, male survival probability had a time effect, and survival probability in late autumn was lower than at other times. Males had a greater survival probability in the overgrazing and lightly grazed sites than in the grazing exclusion sites (Figure 4), but the opposite was found for females ( Figure 5). Because the survival model is built in a long-term continuous marking process, the model may include a different age stage of a midday gerbil, so the model cannot analyze the survival status at a certain point in time. At the same time, there is no significant difference in age structure between them.

| Sex ratio (females/males)
The sex ratio of midday gerbils captured showed a significant difference between the light grazing or grazing exclusion areas and the overgrazing areas in the early spring (F 2,6 = 41, p < .05) and late autumn (F 2,6 = 9, p < .05). There was no significant difference between the light grazing and grazing exclusion areas ( Figure 6).

| Food resources
In early spring, grazing exclusion and light grazing areas significantly enhanced TFB (F 2,19.57 = 3.85, p < .05) and PFB (F 2,19.55 = 4.2 p < .05) compared to the overgrazing areas (Figure 7).  g p = .331) in late autumn. There was a significant impact of TFB and PFB on female density and gerbil population density. TFB affected male density, but PFB has no effect on male population density ( Figure 8).

| D ISCUSS I ON
Winter is a bottleneck period for mammalian nonhibernators due to food shortage and harsh environmental conditions at high northern latitudes (Coltrane & Barboza, 2010;Solonen, 2006). The overwinter survival of gerbils is affected by many mortality risks. The mortality risk faced by wild animal populations depends on their intrinsic conditions (e.g., age-related experience, body condition, immune system status, and population structure), but the intrinsic conditions are always influenced by many external conditions (e.g., meteorological conditions, predation risk, food availability, and disturbance) (Chantepie et al., 2015;Forslund & Pärt, 1995;Théoret-Gosselin et al., 2015). In our study, all treatments were applied under the same meteorological conditions, so meteorological conditions do not play an important role in our study. A previous study has shown that the midday gerbil is a typical arid-area rodent, which prefers habitats with high vegetation coverage in extremely arid desert area (Fu et al., 2004), which was also confirmed by our finding of greater TA B L E 2 Models of local survival probabilities (φ) in male and female midday gerbils using Cormack-Jolly-Seber (CJS) models respective  Grazing exclusion Light Grazing Over Grazing density in grazing exclusion and light grazing areas than in overgrazing areas. Other studies had shown that low grazing can have negative effect on rodents in range lands. However, the vegetation and climate of the abovementioned study area are different from the situation in this study area (Tabeni & Ojeda, 2005;Tchabovsky et al., 2019). The environment in above study area will change greatly, and grassland types will have succession phenomena when long-term reduced grazing intensity, and different species show different coping strategies (Jones et al., 2003). Our result is consistent with former studies suggesting that grazing exclusion may increase the number of rodents (Heske & Campbell, 1991;Keesing, 1998b;Rosi et al., 2009). Our models show a significant effect of grazing on overwinter survival. The body mass, growth rate, and food availability of gerbils differed between the overgrazing and grazing exclusion areas. Therefore, there are three main clusters of factors influencing the population characteristics of gerbils.
The first is food availability. The results of this study indicate that food availability is highest in the grazing exclusion areas, and significantly higher than in light grazing and overgrazing areas.
Adequate food availability results in faster weight accumulation and larger body weight of gerbils in grazing exclusion areas. Our result is consistent with previous studies which observed heavier individuals in grazing exclusion plots than in grazing plots (Bueno et al., 2012;Keesing, 1998b). Previous studies suggested that food addition in the growing or breeding season can increase population densities of rodents through the enhancement of recruitment and reproduction, but does not improve survival, mainly due to the trade-off between reproduction and winter survival in harsh environments (Andrzejewski, 1975;Doonan & Slade, 1995;Taitt & Krebs, 1981).
Winter food availability would limit winter survival of voles (Johnsen et al., 2017) and densities of breeding mouse the following spring.
Thus, increased food availability during winter would result in higher breeding bird densities the following spring. A study on Mongolian gerbils (Meriones unguiculatus) found that group sizes were larger in food supplemented chambers than in unsupplemented chambers (Liu et al., 2011). This is consistent with our results. Grazing exclusion areas, similar to food addition, held higher TFB and PFB than in the overgrazing areas across the whole year. Therefore, compared to the light grazing and overgrazing plots, body mass, body mass growth rates, and the density of gerbils were higher in the grazing exclusion sites. Other study have predicted that in long-term perspective, intense grazing as well as low grazing can cause qualitative changes in vegetation and habitat structure, which may have long-term positive or negative effects on rodents depending on their ecological requirements. In particular, negative long-term effects of low grazing on population abundance and demographic parameters (including survival) were shown for gerbils (Tchabovsky et al., 2016(Tchabovsky et al., , 2019. In above cases, the gerbils demographic parameters keep a relatively negative relationship with the light grazing. The main reason is that the type of grassland has changed from desert to steppe and the gerbils cannot adapt to this living environment. The second cluster of factors affects reproduction. Previous research predicted that sufficient fat and body mass would cause a higher overwinter survival both in male and female hibernating rodents (Schorr et al., 2009). However, female Microtus oeconomus showed a positive correlation between probability of survival and body mass, while no such effect was observed in males (Korslund, 2006), and females showed lower survival rates after breeding (Aars & Ims, 2002).
A previous study indicated that both male and female meridian gerbils can form stable mate preferences, with behavioral characteristics of monogamous species . The female gerbil had enough energy for mating and reproduction due to their higher body mass and higher PFB in grazing exclusion areas. Therefore, the female gerbils had higher overwinter survival in grazing exclusion sites. This study indicated a lower sex ratio (females/males) in overgrazing areas than in grazing exclusion areas. Male gerbils thus allocated less energy and time for mating behavior, and they need less food to support their daily activities due to higher survival in the overgrazing sites. The third aspect is predation risk. In our study, male gerbils had higher overwinter survival in overgrazing areas than in lightly grazing and grazing exclusion areas, but female gerbils did not, showing higher survival properties in grazing exclusion areas.
The removal of livestock could increase plant cover, thereby reducing the exposure of small mammals to their avian predators F I G U R E 6 Sex ratio (females/males, x ± SE) of midday gerbils captured from overgrazing, light grazing, and grazing exclusion sites in early spring and late autumn, 2006 to 2010 in Alxa Left Banner, Inner Mongolia, China. Different letters (A, B, and C) indicate significant differences (p < .05) between overgrazing, light grazing, and grazing exclusion sites, according to Tukey's test.

Early Spring Late Autumn
F I G U R E 7 Total food biomass (TFB, x ± SE) and preferred food biomass (PFB, x ± SE) in overgrazing (OG), light grazing (LG), and grazing exclusion (EG) sites in April (early spring(a)) and October (late autumn (b)). Different letters (A, B, C or a, b, c) indicate significant differences (p < .05) (Peles & Barrett, 1996). Rodents may be at risk of avian predation in exposed microhabitats, but were more susceptible to predation by snakes in sheltered microhabitats (Bouskila, 1995;Kotler et al., 1991Kotler et al., , 1993. Gerbils are nocturnal rodents, so they are more likely to be hunted by night-time predators such as foxes, snakes, and weasels, rather than birds. In overgrazing plots, male gerbils spend less time out of the nest due to their lower reproductive investment, thereby facing lower predation risk. Therefore, they would face higher predation risk in the grazing exclusion sites.
Thus, gerbils showed higher survival ability in overgrazing than in grazing exclusion sites.
In our study, a lower population density occurs in late autumn and a higher in spring in our experiment. There are two main reasons for this result. The first reason is that the autumn midday gerbils are in a period of concentrated storage of food, so they face a higher predation risk, and the population will decline. Second, because autumn is a peak period for midday gerbils' reproduction, their pups grow into adults in the spring of the second year, increasing the population of midday gerbils, so the number of midday gerbils in spring is generally higher than the previous year's late autumn. This result also shows that winter has a great influence on the population of small rodents.
We concluded that grazing influenced midday gerbil population density and body mass negatively due to reduced food availability in late autumn and early spring. For female gerbils, overwinter survival was higher in grazing exclusion sites than in light grazing sites and overgrazing sites, while male gerbils showed higher survival in overgrazing than in lightly grazed sites and grazing exclusion sites.
Further research is required on the factors affecting survival rates, with a focus on other factors that differentially affect male and female survival rates. These factors may include adjustments within their populations or other abiotic factors, such as climate.

ACK N OWLED G M ENTS
This experiment was funded by the National Natural Science Wilkes for their help in writing this paper.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.