Introduction

Plant functional traits are considered as reflecting the adaptations to environmental change and trade-offs among different functions within a plant, thus they can better link environmental drivers and plant responses to ecosystem structure and functioning (Díaz et al. 1998; Díaz et al. 2007; Lavorel et al. 2007; Wright et al. 2004). In recent years, more attempts to predict plant responses to grazing have focused on using plant functional traits, especially for some ‘soft’ traits, which can be easily and quickly measured, such as plant height, leaf size, leaf toughness, leaf dry matter content, and specific leaf area (SLA) (Adler et al. 2005; Díaz et al. 2007; Díaz et al. 2001; De Bello et al. 2005; Golodets et al. 2009; Vesk et al. 2004). Generally, these ‘soft’ traits are closely correlated with the ‘harder’ traits, such as photosynthetic capacity, relative growth rate, leaf nutrient concentrations, and leaf turnover rate, which further reflect the fundamental trade-offs between fast growth and anti-herbivore defence (or nutrient storage) in plant functioning (Grime 2001; Reich et al. 1997; Reich et al. 2007; Westoby et al. 2002).

Many studies showed that plant height and leaf size decreased under grazing pressure (Díaz et al. 2001; Klimesova et al. 2008; Landsberg et al. 1999; Rusch et al. 2009). Because tall species receive most grazing pressure and short species are protected from grazing by tall species. Also, larger leaves provide better bites for grazer, and smaller leaves require more bites at a given leaf area (and mass) if taken in the same number of bites (Vesk et al. 2004). SLA, the ratio of leaf area to leaf mass, as an important leaf trait that integrates plant investment into growth and storage, has been widely used to predict plant strategy. Westoby (1999) reported that SLA might be related to grazing responses in a different manner at low and high grazing intensities. The high SLA species decreased faster than low SLA species at low grazing intensity. However, under high and non-selective grazing intensity, all species are grazed and high SLA species may be advantaged by faster regrowth. Therefore, plant resistance to grazing is associated with both avoidance traits (e.g., small height and leaf size), and tolerance traits (e.g., high SLA and leaf N content) (Cingolani et al. 2005; Díaz et al. 2001). Díaz et al. (2001) proposed plant species responses to grazing could be predicted from simple traits, such as plant height, leaf size, life history, and SLA in subhumid grasslands. Yet, Vesk et al. (2004) found little evidence for predictability of grazing effect on plant growth with simple traits in the semi-arid and arid shrublands and woodlands. Recently, several studies demonstrated plant traits response to grazing are largely mediated by resource availability (e.g., precipitation and soil nutrients) (Adler et al. 2004; Osem et al. 2004; Pakeman 2004; Vesk et al. 2004) and grazing intensity or evolutionary history of grazing (Adler et al. 2004; Díaz et al. 2007; Graff et al. 2007). For example, Adler (2004) found that aridity favored grazing resistance traits, while grazing impact was intermediated in wetter sites. Thus, there are still large uncertainties associated with the effects of grazing on plant functional traits and the causal linkage between shifts in species traits and change in ecosystem functioning under different environmental conditions.

As typical semi-arid grasslands, the Xilin River Basin is characterized by sparse and irregular rainfall and soil nutrient deficiency (Chen 1988; Wang and Cai 1988). The traditional land use type in this area is grazing, which changed from semi-nomadic farming systems to intensified settled livestock farming in the last 20 years. Consequently, species diversity and aboveground net primary productivity (ANPP) decreased markedly in recent decades, and grassland degradation has become a severe problem in Xilin River Basin (Tong et al. 2004). Leymus chinensis (tall perennial C3 rhizome grass) and Cleistogenes squarrosa (short perennial C4 bunchgrass) are two dominant species distributed widely across a broad range of grassland ecosystems on the Mongolian plateau (Chinese Academy of Sciences Integrative Expedition Team to Inner Mongolia and Ningxia 1985; Hilbig 1995). However, both species have distinctly different habitats: L. chinensis predominates in relatively wet and fertile habitats with high vegetation cover, while C. squarrosa dominates in dry and infertile habitats with low vegetation cover, particularly in the overgrazed communities. In the lightly and moderately grazed communities, L. chinensis is located at the upper canopy layer of the community while C. squarrosa is at the lower canopy layer. Previous studies demonstrated that both species have distinct individual traits in terms of leaf morphology (Wan et al. 2008), photosynthetic physiology (Chen et al. 2002; Chen et al. 2005; Cui et al. 2001), root distribution (Chen et al. 2001), seed size and seed mass (Zhong et al. 2001). Yet, little is known about how functional traits of the two contrasting species respond to grazing and their differences in avoidance and tolerance strategies under different stocking rates. Several previous studies have predicted that overgrazing will lead to a shift in dominance from L. chinensis to C. squarrosa and consequently cause a decrease in biodiversity and ecosystem functioning (Auerswald et al. 2009; Bai et al. 2000; Bai et al. 2007; Wang et al. 2003), although mechanisms underpinning this shift are not very clear.

In this study, both morphological and physiological traits of L. chinensis and C. squarrosa grown in a traditional grazing system with different stocking rates were examined in 2008 (a wetter year) and 2009 (a dry year). Specifically, we address the following three research questions: First, how do functional traits of the two dominant species vary with grazing intensity and precipitation? Second, are there any fundamental trades-offs linked to trait responses to grazing and variation in precipitation? Third, what are the avoidance and tolerance strategies employed by two species to improve their grazing resistance?

Materials and methods

Study area and experimental design

This study was conducted at the Inner Mongolia Grassland Ecosystem Research Station (IMGERS), Chinese Academy of Sciences, which is located at the middle reach of the Xilin River (Bai et al. 2004). Based on the long-term meteorological data (1970–2007), the mean annual temperature in the study area was 0.4°C with mean monthly temperature ranging from −21.4°C in January to 19.0°C in July, and annual precipitation was 336.9 mm, with 51–89% occurring in the growing season (May–August). The meteorological data collected within the experimental site showed that the patterns of monthly mean temperature were similar between 2008 and 2009, but the patterns of monthly precipitation were quite different between the 2 years (Fig. 1), with much higher total annual precipitation occurred in 2008 (297 mm) than that in 2009 (226 mm). The growing season precipitation (May–August) was 263 mm in 2008 but only 157 mm in 2009.

Fig. 1
figure 1

Monthly precipitation (bars) and monthly mean air temperature (line) of the Xilin River Basin in 2008 and 2009. Meteorological data were collected from the eddy tower adjacent (approximately 100 m) to the experimental plots

Full size image

The topography at our experimental site is consisted of two landscape units, i.e., slope and flat ground, with elevations ranging from 1220 to 1280 m. The soil is dark chestnut soil (Calcic Chernozem according to ISSS Working Group RB, 1998), with a loamy sand texture (Bai et al. 2010). Topsoil organic carbon, total nitrogen, and total phosphorus contents (0–10 cm) are 20.0, 1.90 and 0.32 g.kg−1, respectively. Topsoil pH ranges from 6.61 to 7.98 and the topsoil cation exchange capacity (CEC) is 15.7 cmol.kg−1. The vegetation at the study area is dominated by L. chinensis and Stipa grandis (C3 perennial bunchgrass), which is a widely distributed community type in the Eurasia steppe (Bai et al. 2008; Coupland 1993).

Field sampling and measurements were carried out at the Sino-German grazing experimental site, which was established in June 2005 and occupied a total area of 160 ha including 56 experimental plots (Schonbach et al. 2009). The experimental treatment was composed of two management systems (e.g., traditional versus mixed system) and 7 levels of stocking rate (SR = 0, 1.5, 3.0, 4.5, 6.0, 7.5, and 9.0 sheep/ha). The standard plot size was 2 ha for SR = 0, 3.0, 4.5, 6.0, 7.5, and 9.0 sheep/ha, while it was 4 ha for SR = 1.5 sheep/ha in order to achieve a minimum of 6 sheep per plot. All plots have been fenced against sheep migration between plots since the beginning of the experiment. The traditional system was managed with grazing and haymaking on two different areas. In the grazing area, sheep was allowed to graze continuously during the vegetation period from June to September. The mixed system was managed with a yearly shift between grazing and haymaking (Schonbach et al. 2009). In this study, the traditional grazing system including 5 levels of stocking rates (i.e., 0, 3.0, 4.5, 7.5 and 9.0 sheep/ha), representing a gradient of increasing grazing intensity, was selected as our field sampling and measurement plots.

Vegetation and soil properties

Within each plot, three 2 m × 2 m grazing exclosures were established randomly in the early June to determine the aboveground biomass before sheep grazing started, and the exclosures were monthly moved to new locations to measure the productivity of grazed plots until the end of September each year. Two sampling quadrats (1 m × 1 m) were randomly located inside and outside each exclosure and paired into block each month. The difference in total biomass between each pair of quadrats was used for estimating monthly biomass increment. On the 8–10th of each month, all living biomass within each quadrat was harvested by clipping to the soil surface, separated to species, oven dried at 70°C for 24 h to constant mass, and weighed. Litter biomass within each quadrat was collected. The total aboveground biomass available for grazing sheep (Bg) and the ungrazed aboveground biomass (Bu) were collected outside and within each grazing exclosure during June, July, August, and September, thus the monthly biomass increment could be estimated. Annual biomass accumulation (Ba) was calculated by the biomass at the initial date of grazing (early June) summed to the monthly biomass growth using the following equation as per suggested by Schonbach et al. (2009).

$$ {{\hbox{B}}_{\rm{a}}} = {{\hbox{B}}_{\rm{g1}}} + \left( {{{\hbox{B}}_{\rm{u2}}} - {{\hbox{B}}_{\rm{g1}}}} \right) + \left( {{{\hbox{B}}_{\rm{u3}}} - {{\hbox{B}}_{\rm{g2}}}} \right) + \left( {{{\hbox{B}}_{\rm{u4}}} - {{\hbox{B}}_{\rm{g3}}}} \right) $$

where Bgi and Bui are the aboveground biomass collected outside and inside the exclosure at sampling time i (i=1, 2, 3, 4, i.e., beginning of June, July, August, and September, respectively).

Since the standing aboveground biomass of this area usually reaches its annual peak in August, thus aboveground net primary productivity (ANPP) for each treatment approximated Ba (Bai et al. 2004). For SR = 0, the peak aboveground biomass sampled in August was equated with Ba.

Five soil samples were collected at the surface layer (0–10 cm) for each plot, oven-dried at 105°C for 48 h, and weighed to determine soil moisture (%) in mid July, 2008 and 2009.

Plant traits

On July 25–26, 2008 and 2009, we randomly collected 30 individuals (ramet) of L. chinensis (a clonal plant) and the same number of bunches for C. squarrosa at each plot. The height was determined by the distance from the basal stem to the natural crown of each plant individual. After the height measurement, the aboveground part of each individual was collected and taken back to the laboratory for separating stem and leaf biomass. All stem and leaf samples were oven-dried at 70°C for 24 h to constant mass and weighted. Thus, stem biomass, leaf biomass, stem: leaf ratio (S/L), and plant individual biomass were calculated.

For each species, 30–50 individuals with fully expanded leaves were chosen for measuring leaf morphological traits. According to leaf size, 5–20 leaves were picked as a sample and totally 30 replicates for each species. The projected area of each sample was measured with a planimeter (Li-3100, Li-Cor, Lincoln, NE, USA), and then oven-dried at 70°C for 24 h to a constant mass. The specific leaf area (SLA, cm2 g−1) was calculated as the ratio of leaf area (LA) to leaf dry mass (LM). After SLA measurement, leaf sample was grounded to 80-mesh and leaf N content was analyzed using a Kjeltec analyzer (Kjeltec 2300 Analyzer Unit, Foss, Sweden). The leaf N content was expressed on both mass basis (Nmass, %) and area basis (Narea, g·m−2), which was calculated as Nmass divided by SLA.

Gas exchange

The photosynthetic gas exchange parameters of two species were measured with a Li-6400 Portable Photosynthetic System (Li-6400, Li-Cor, Lincoln, NE, USA) on 25–26 July, 2008 and 2009. Net photosynthetic rate (Pn), transpiration rate (Tr), stomatal conductance (Gs), and intercellular CO2 concentration (Ci) were measured at a CO2 concentration of 400 µmol mol−1 (using the built-in Li-Cor 6400 CO2 controller) and a saturating irradiance of 1500 µmol m−2 s−1 provided by a built-in red LED light source. Instantaneous water use efficiency (WUE) was determined by Pn/Tr. The value of stomatal limitation (Ls) was calculated by the following formula:

$$ {{\hbox{L}}_{\rm{s}}} = 1 - {{\hbox{C}}_{\rm{i}}}/{{\hbox{C}}_{\rm{a}}} $$

where Ca is atmosphere CO2 concentration, which is controlled at 400 µmol mol−1 in this study.

For each treatment, the photosynthetic gas exchange parameters were measured on three fully expanded leaves of the plant, and a total number of 10 individuals for each species were measured.

Statistical analysis

Statistical analyses were performed with the software SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) and Multi-Variate Statistical Package (MVSP, Kovach Computing Services, Anglesey, UK). The effects of stocking rate, sampling year, and their interaction on plant functional traits were tested by two-way ANOVAs. One-way ANOVAs followed by the LSD multiple range tests were used for multiple comparisons of plant traits among treatments. The differential responses of plant trait to grazing between L. chinensis and C. squarrosa were tested using Independent-Samples T test at P < 0.05. The canonical correlation analyses were used to test the relationships between whole-plant traits (height, plant biomass, stem biomass, leaf biomass, stem: leaf ratio) and leaf traits (LA, LM, SLA, Pn, Tr, Gs, Ci, WUE, Ls, Nmass, and Narea), and between leaf morphological (LA, LM, and SLA) and physiological traits (Nmass, Narea, Pn, Tr, Gs, Ci, WUE, and Ls). For both species, trait data were pooled into three groups, i.e., no grazing (control), low grazing intensity (SR = 3.0 and 4.5 sheep/ha), and high grazing intensity (SR = 7.5 and 9.0 sheep/ha) in both years. For each group, canonical redundancy analysis was conducted to explore the variance explained by the canonical variates. The relationships between plant functional traits were tested by Pearson correlation analysis. The principal components analysis (PCA) was conducted to identify fundamental trade-offs between traits and associated trait syndromes between the two species at no grazing, low grazing intensity, and high grazing intensity. Data were standardized before the analysis. The differences in loading score on PCA 1 for traits between 2 years or two species were tested using Independent-Samples T test at P < 0.05. Using regression analysis, we also analyzed the relationships between the responses of ecosystem properties (soil moisture, ANPP) and plant functional traits (Pn, and plant biomass) to grazing. Before regression analysis, the response ratios of treatment to control at different stocking rates were transformed with the natural logarithm to improve the normality.

Results

Response of whole-plant traits

The whole-plant traits including height, total individual aboveground biomass, stem biomass, and leaf biomass for both species, and stem: leaf ratio (S/L ratio) for C. squarrosa were all significantly affected by grazing (P < 0.001). The effect of sampling year was only significant for C. squarrosa (P < 0.05), except for S/L ratio. However, the interactive effects of stocking rate × year on whole-plant traits were all highly significant for both species (P < 0.01).

For L. chinensis, height, total individual biomass, stem and leaf biomass decreased significantly at the highest stocking rate (9.0 sheep/ha) in the wet year (2008), and at high stocking rates (7.5 and 9.0 sheep/ha) in the dry year (2009); although no significant changes in total individual biomass, stem and leaf biomass were observed at low and moderate stocking rates (3.0 and 4.5 sheep/ha) compared to control (no grazing) in the dry year (Fig. 2). Compared with control, total individual biomass of L. chinensis was decreased by 72% in 2008 and by 70% in 2009 at the highest stocking rate (9.0 sheep/ha). The stem: leaf (S/L) ratio generally increased at low, moderate, and high stocking rates in the wet year, but declined in the dry year (Fig. 2).

Fig. 2
figure 2

Effects of grazing on whole-plant traits of L. chinensis and C. squarrosa. Vertical bars indicate standard errors of the mean (n = 30). Different letters represent significant differences among treatments (LSD test, P < 0.05), with lowercases for 2008 and capital letters for 2009

Full size image

For C. squarrosa, height, total individual biomass, stem and leaf biomass diminished significantly with increasing stocking rate in both years, with a greater reduction occurred in the wet year than that of the dry year (Fig. 2). At the highest stocking rate, plant biomass of C. squarrosa was declined by 92% in 2008 and by 81% in 2009 compared to control. Under moderate to high stocking rates, grazing also reduced the S/L ratio of C. squarrosa in the wet year, but enhanced it in the dry year (Fig. 2).

In the control plots, total individual biomass, stem and leaf biomass of L. chinensis were significantly higher in the wet year than in the dry year (P < 0.05), while height and stem:leaf ratio showed a reverse pattern (P < 0.0001). For C. squarrosa, however, there was no significant difference in each trait between the wet and dry years.

Response of leaf morphological traits

For both L. chinensis and C. squarrosa, the effects of stocking rate, year, stocking rate × year on leaf morphological traits, i.e., leaf area, leaf dry mass, and specific leaf area, were highly significant (P < 0.01), except for the interactive effect of stocking rate × year on leaf dry mass of C. squarrosa.

The leaf area of L. chinensis did not show a significant change at low and moderate stocking rates but declined significantly at high stocking rates in the wet year. In the dry year, however, the leaf area of L. chinensis was significantly reduced by grazing at moderate and high stocking rates (Fig. 3). The leaf dry mass of L. chinensis also showed a slight but significant decrease at low, moderate, and high stocking rates and a marked reduction at the highest stocking rate in the wet year, whereas it was diminished sharply by grazing at moderate and high stocking rates in the dry year (Fig. 3). Grazing generally increased the specific leaf area of L. chinensis in the wet year. In the dry year, in contrast, the response of specific leaf area of L. chinensis was much complex, declined first at low and moderate stocking rates and enhanced then at high stocking rates (Fig. 3).

Fig. 3
figure 3

Effects of grazing on leaf morphological traits of L. chinensis and C. squarrosa. Vertical bars indicate standard errors of the mean (n = 30). Different letters represent significant differences among treatments (LSD test, P < 0.05), with lowercases for 2008 and capital letters for 2009

Full size image

Leaf area and leaf dry mass of C. squarrosa were diminished by grazing for almost all treatment plots and in both wet and dry years (Fig. 3). In the wet year, grazing enhanced specific leaf area of C. squarrosa at low and moderate stocking rates. In the dry year, however, grazing increased SLA of C. squarrosa only at low stocking rate and had no effect on SLA at moderate and high stocking rates (Fig. 3). For all treatment and control plots, the SLA of C. squarrosa increased significantly in the dry year compared to the wet year (P < 0.0001).

In the control plots, leaf area and leaf dry mass of L. chinensis and leaf dry mass of C. squarrosa were significantly higher in the wet year than in the dry year (P < 0.05), however, the SLA of both species were significantly higher in the wet year than in the dry year (P < 0.0001). Also, L. chinensis had much higher leaf area and leaf dry mass but significantly lower SLA than C. squarrosa across all treatment and control plots (P < 0.0001).

Response of leaf physiological traits

All leaf physiological traits of the two species were significantly affected by grazing (P < 0.001), and the effects of year and stocking rate × year were mostly significant (P < 0.05). In the wet year, Pn, Tr, Gs, and Ci of L. chinensis first decreased at low stocking rate, then increased at moderate and high stocking rates, and declined again at the highest stocking rate; whereas WUE and Ls of L. chinensis first increased with stocking rate and then decreased at high stocking rates but still significantly higher than those of control (Fig. 4). In the dry year, Tr, Gs, and Ci of L. chinensis also declined at low stocking rate, then increased at moderate and high stocking rates, and elevated to the level of control at the highest stocking rate, with Pn declining first at low stocking rate and then increasing slightly at moderate and high stocking rates (Fig. 4). A significant reduction in WUE and Ls of L. chinensis was observed only at the highest stocking rate, while they were relatively unchanged at low, moderate, and high stocking rates (Fig. 4). On the whole, grazing had more negative effects on photosynthetic capacity of L. chinensis in the dry year than that in the wet year.

Fig. 4
figure 4

Effects of grazing on leaf photosynthetic traits of L. chinensis and C. squarrosa. Vertical bars indicate standard errors of the mean (n = 10). Different letters represent significant differences among treatments (LSD test, P < 0.05), with lowercases for 2008 and capital letters for 2009

Full size image

In most cases, grazing significantly diminished Pn, Tr, Gs, and WUE of C. squarrosa in the wet year (Fig. 4). On average, Pn, Tr, Gs, and WUE of C. squarrosa decreased by 72%, 17%, 58%, and 62% in 2008, respectively, compared to control. In the dry year, however, grazing generally had no effect on Pn Tr, Gs, and WUE (Fig. 4). In both wet and dry years, grazing elevated Ci but reduced Ls of C. squarrosa in most cases. Compared to control, the average value of Ci in the grazed plots increased by 31% in 2008 and by 55% in 2009; whereas Ls decreased by 22% in 2008 and by 30% in 2009. Across all grazed plots, the values of WUE of C. squarrosa were much higher in 2009 than those in 2008 (Fig. 4). Moreover, compared with L. chinensis, C. squarrosa showed much lower Pn, Tr, Gs, Ci, Nmass, and Narea, but higher Ls (Figs. 4, 5).

Fig. 5
figure 5

Effects of grazing on mass (Nmass) and area (Narea) based leaf nitrogen contents of L. chinensis and C. squarrosa. Vertical bars indicate standard errors of the mean (n = 5). Different letters represent significant differences among treatments (LSD test, P < 0.05), with lowercases for 2008 and capital letters for 2009

Full size image

For both L. chinensis and C. squarrosa, values of Nmass and Narea were consistently higher in the wet year than those in the dry year (Fig. 5). Grazing increased Nmass and Narea of L. chinensis in the dry year, whereas decreased Narea of L. chinensis and generally had no effect on Nmass in the wet year (Fig. 5). For C. squarrosa, however, grazing enhanced Nmass and Narea in both wet and dry years.

Relationships between plant functional traits

The canonical correlation analysis showed that whole-plant traits were significantly correlated with leaf morphological and physiological traits when all data of the two species were pooled together. The correlation coefficient (r) for the first group of canonical variates was 0.98 (P < 0.001) at no grazing (control), 0.95 (P < 0.001) at low grazing intensity (3.0 and 4.5 sheep/ha), and 0.93 (P < 0.001) at high grazing intensity (7.5 and 9.0 sheep/ha). Canonical redundancy analysis further showed that the canonical variates were Pn, Gs, Tr, Ci, Ls, Ci, LA, LM, SLA, and Narea for leaf traits at control plots, explaining 66% of total variation; for whole-plant trait, height was the canonical variate, accounting for 24% of the total variation. For both species, the height was positively correlated with Pn, Gs, Tr, Ci, LA, LM, and Narea, but negatively correlated with SLA, WUE, and Ls (Appendix 1). Plant height had a predominant effect on leaf physiological traits at the control plots, explaining 63% of total variance. Canonical redundancy analysis also showed that the canonical variates for leaf trait were Pn, Gs, Tr, LM, LA, SLA, and Narea, which explained 32% of the total variation at the low grazing intensity and 36% at the high grazing intensity. The canonical variates for whole-plant trait were total individual biomass, stem biomass, leaf biomass, and height, which accounted for 69% and 67% of total variation at low and high grazing intensities, respectively. All whole-plant traits were positively correlated with Pn, Gs, Tr, LM, LA and Narea, but negatively correlated with SLA (Appendix 1). Leaf physiological traits had predominant effects on plant biomass at low grazing intensity, explaining 63% of total variance. However, leaf morphological traits had leading influence on plant biomass at high grazing intensity, accounting for 57%.

Results from canonical correlation analysis also demonstrated that leaf physiological traits were positively correlated with leaf morphological traits under no grazing disturbance (r = 0.997), at low (r = 0.989) and high (r = 0.987) grazing intensities. Pn, Gs, Tr, and Narea were positively correlated with LA and LM, but negatively correlated with SLA (Appendix 1). The canonical variate for leaf morphological traits was LM and LA under no grazing disturbance (accounting for 83%) and at low grazing intensity (58%), and was SLA at high grazing intensity (83%). The canonical variate for leaf physiological traits was Narea under no grazing disturbance (accounting for 64%), at low (23%) and high grazing intensities (32%). The canonical redundancy analysis further illustrated that Narea explained separately 83%, 57%, and 81% of variation in leaf morphological traits under no grazing disturbance, at low and high grazing intensities.

Differential response between two species

The principal components analysis (PCA) showed that the first PCA axis (PCA 1) was positively correlated with traits related to plant height, leaf size, and photosynthetic capacity under no grazing disturbance, explaining 52% of the total variance, while the second axis (PCA 2) was related to plant biomass, accounting for only 22% (Fig. 6). We identified two fundamental trade-offs linked to consumptive and conservative plant growth strategies at low and high grazing intensities. The PCA 1 was correlated with traits related to leaf size, photosynthetic capacity, and plant size, which increased, while SLA decreased along the PCA 1 (Fig. 6). The PCA 1 explained separately 46% and 52% of the total variance at low and high grazing intensities. The PCA 2 was positively correlated with WUE and Ls, but negatively correlated with Ci, reflecting the conservative water utilization for plant growth. The PCA 2 explained only 19% of total variance at both low and high grazing intensities.

Fig. 6
figure 6

PCA biplot diagram for first two axes showing the functional traits of L. chinensis and C. squarrosa at no grazing (a) and low (b) and high (c) grazing intensities. The absolute values of each point on x- and y-coordinate represent loading scores of each trait for PCA 1 and 2. Trade-off is characterized by negative correlation between two traits when they have high loading on the same axis but in an opposite direction. For each species, whiskers denote standard error of the mean of PCA weighted loading scores. Abbreviations: PH, plant height; PB, plant biomass; SB, stem biomass; LB, leaf biomass; S/L, stem: leaf ratio; LA, leaf area; LM, leaf dry mass; SLA, specific leaf area; Pn, net photosynthetic rate; Tr, transpiration rate; Gs, stomatal conductance; Ci, intercellular CO2 concentration; WUE, instantaneous water use efficiency; Ls, stomatal limitation; Nmass, mass based nitrogen content; Narea, area based nitrogen content; Lc, L. chinensis; Cs, C. squarrosa

Full size image

The PCA 1 distinctly separated L. chinensis with higher height, plant size, leaf size, and photosynthetic capacity from C. squarrosa with higher SLA under no grazing disturbance, at both low and high grazing intensities (Fig. 6). For C. squarrosa, there was a significant difference in loading score on PCA 1 between the wet and dry years under no grazing disturbance (P = 0.025), at low (P < 0.001) and high (P = 0.043) grazing intensities. For L. chinensis, however, there was no significant difference in loading score on PCA 1 between the 2 years at low grazing intensity, while the difference was significant under no grazing disturbance (P < 0.001) and at high grazing intensity (P = 0.031). Thus, high grazing intensity decreased inter-variation of two species, and grazing had more influence on L. chinensis than C. squarrosa. For a given grazing intensity or year, the loading score on PCA 1 was also significantly different between the two species (P < 0.01). Also, the difference between the two species gradually diminished from no grazing disturbance, low grazing intensity to high grazing intensity, especially in the dry year (Fig. 6).

Relationship between responses of species and community

Under grazing treatment, the Pn response of C. squarrosa was negatively correlated with soil moisture response, whereas the Pn response of L. chinensis was little affected by change in soil moisture (Fig. 7). For both species, the response of total individual biomass to grazing was also negatively correlated with soil moisture response (Fig. 7). The plant biomass response of C. squarrosa was positively correlated with its Pn response, while no significant relationship was found between the plant biomass response and Pn response of L. chinensis (Fig. 7). The ANPP response was positively correlated with plant biomass responses of two species, especially with L. chinensis (Fig. 7).

Fig. 7
figure 7

Relationships between responses of net photosynthetic rate (Pn, a) and total individual biomass (b) and soil moisture, between plant biomass response and Pn (c) and ANPP responses (d). Data were log-transformed (base e) before analysis to improve normality. Lc, L. chinensis; Cs, C. squarrosa

Full size image

Discussion

Grazing impact on plant functional traits of two species

In this study, we demonstrated that the whole-plant traits of the two species, including plant height, total individual biomass, leaf biomass, and stem biomass, decreased with increasing stocking rate, which are consistent with previous studies (Díaz et al. 2001; Klimesova et al. 2008; Landsberg et al. 1999; Rusch et al. 2009). The magnitude of change for each species, however, was generally greater in the wet year than that in the dry year. The decreased total individual biomass of both species was attributed mainly by the reduction in both leaf biomass and leaf area that diminished assimilation production of the plant at high stocking rates. This is because leaf Pn is closely related to plant photosynthetic capacity and relative growth rate (Field and Mooney 1986; Shipley 2002, 2006), and plant stature and leaf size are directly associated with light interception for photosynthesis (Milla and Reich 2007; Poorter and Rozendaal 2008). For both species, the less negative impact of grazing on whole-plant traits was likely caused by the increased stomatal conductance and intercellular CO2 concentration in the dry year, which made the net photosynthetic rate unchanged or even enhanced, particularly for C. squarrosa.

The canonical correlation and redundancy analyses further suggest that grazing at low intensity decreased plant stature and consequently had a negative impact on leaf photosynthetic capacity, while grazing at high intensity diminished leaf size and resulted in the decrease of plant biomass. Our findings indicate that species mainly suffers from light competition associated with plant stature at no grazing or low grazing intensity, while heavy grazing generally leads to miniaturization of plant individual (Wang et al. 2000). In addition, our results have shown that the response of community ANPP to grazing was positively correlated with the responses of plant biomass of the two species, especially with that of L. chinensis. This suggests that heavy grazing, which decreased plant growth of dominant species, greatly influenced the ecosystem functioning.

Compared with C. squarrosa, total individual biomass of L. chinensis was relatively less affected by grazing at low and moderate stocking rates in both wet and dry years. This is due mainly to the relatively low impact of grazing on both leaf biomass and leaf area of L. chinensis. At high stocking rates, however, the plant individual biomass of both species was significantly reduced by grazing. The high plasticity of L. chinensis at low and moderate stocking rates partly support the hypothesis that the effect of grazing on plant traits could be mediated by grazing intensity (Adler et al. 2004). For C. squarrosa, however, plant individual biomass, leaf biomass, and leaf area were all decreased sharply by grazing even at low and moderate stocking rates.

Trade-offs among plant functional traits in dry and wet years

Water availability is the key limiting factor controlling plant growth and biomass production in arid and semiarid grassland ecosystems (Bai et al. 2004; Bai et al. 2008). In this study, the differential responses of plant trait to variation in precipitation suggest that some fundamental trade-offs among traits may exist in wet versus dry years. For L. chinensis, both mass and area based leaf N contents and net photosynthetic rate were significantly higher in the wet year than those in the dry year under grazed conditions. These trade-offs indicate that L. chinensis may adopt the high regrowth rate to compensate for the biomass loss by frequent grazing disturbance in the wet year. In the dry year, however, it may adopt the low growth rate for more conservative resource use to survive under water stress conditions and grazing disturbance. Our findings are also corroborated by previous studies that trade-offs among plant functional traits within a species may promote its survivorship under the fluctuating environments (Grime 2001; Reich et al. 1997; Reich et al. 2007; Westoby et al. 2002).

For C. squarrosa, the trade-offs of high SLA and net photosynthetic rate in the dry year versus low SLA and Pn in the wet year suggest that it is likely to adopt high growth rate to deplete the available resources and improve its competitive ability when the growth of L. chinensis was inhibited by water stress. This was further supported by the results that plant individual biomass and stem and leaf biomass were much higher in the dry year than those in the wet year under grazed conditions. It has proposed that C4 species generally exhibit high photosynthetic capacity due to their particular mechanisms of CO2 fixation and high resource use efficiency in dry habitats with high irradiation (Sage et al. 1999). Our results also revealed that C. squarrosa was more resistant to droughts as indicated by the negative relationships between Pn and biomass responses and soil moisture response under the grazed conditions.

In the present study, the impacts of grazing on leaf area, leaf dry mass, and net photosynthetic rate of L. chinensis were less negative in the wet year than that in the dry year. However, grazing enhanced the net photosynthetic rate of C. squarrosa in the dry year but decreased it in the wet year. This implies that high annual precipitation may modulate the negative grazing effect on functional traits of L. chinensis, whereas low annual rainfall is likely to promote the growth of C. squarrosa. Thus, the effect of grazing on the functional traits was also mediated by water availability (Adler et al. 2004; Osem et al. 2004; Pakeman 2004; Vesk et al. 2004).

Plant avoidance and tolerance strategies to grazing

Plant species may adopt some avoidance (escape from grazers) and tolerance (regrowth capacity after defoliation) strategies to improve their grazing resistance (Cingolani et al. 2005; Díaz et al. 2001; Klimesova et al. 2008; Landsberg et al. 1999; Rusch et al. 2009). In this study, we found that the two species exhibited both similar and different strategies in terms of their avoidance and tolerance to grazing pressure.

First, plant height and leaf size of both species decreased with increasing stocking rate, which reduced foraging selectivity by herbivores (Vesk et al. 2004), with C. squarrosa being much shorter than that of L. chinensis. Second, the SLA, leaf Nmass and Narea (except L. chinensis in the dry year) of both species increased or relatively unchanged under high grazing pressure, which might be favorable for leaf turnover and shoot regrowth (Cingolani et al. 2005; Evju et al. 2009; Pérez-Harguindeguy et al. 2003; Westoby 1999). C. squarrosa had much higher SLA but lower area based N content compared with L. chinensis. Third, both L. chinensis and C. squarrosa employed more conservative water use strategy to maintain or improve their photosynthetic capacity in the dry year. This is because Gs and Ci of both species increased, Ls decreased at high grazing intensity, whereas WUE was relatively unchanged. Compared with L. chinensis, C. squarrosa had the higher WUE in the dry year. Collectively, compared with L. chinensis, C. squarrosa had shorter plant height and lower total individual biomass, stem and leaf biomass, LA, LM, Pn, and Narea but higher SLA. This implies that C. squarrosa is more resistant to grazing in terms of its avoidance and tolerance traits than that of L. chinensis, particularly under heavy grazing pressure and in the dry year.

Our findings have important implications for understanding species coexistence and for the management of semiarid ecosystems in the face of climate change. First, future climate change is likely to produce more frequent extreme precipitation and drought events (Easterling et al. 2000), which may have greater impacts on the functioning of L. chinensis and C. squarrosa dominated steppe ecosystems. The dominance of L. chinensis is likely to increase in the wet yeas because it has the competitive advantage both in light and soil moisture competition. In the dry years, however, when L. chinensis is suppressed, C. squarrosa may take advantage of available resources because of its drought-resistance traits. Second, we expect that the increased droughts together with heavy grazing may accelerate the process of shift in dominance from L. chinensis to C. squarrosa in the typical steppe, and consequently decrease the ecosystem functioning, such as primary production and carbon sequestration (Li 1989; Wittmer et al. 2010; Xiong et al. 2003). Third, it seems clear from our study that the currently high stocking rate in the Inner Mongolia grassland should be reduced in order to avoid drastic alterations of ecosystem structure and functioning in a long run.