Introduction

Native grasses of the Chinese semi-arid steppe, defined as a rhizome and bunchgrass steppe with many or few forbs, are well adapted to herbivory and have supported grazing as the dominant land use for thousands of years (Zhu 1993). However, in the past four decades, overgrazing has resulted in substantial degradation of the Inner Mongolian steppe (Xie and Wittig 2004; Zhong et al. 2005) and drastically reduced community productivity (Liu et al. 1998; Katoh et al. 1998), litter accumulation (Snyman 2005), and site fertility, because of loss of fertile topsoil (Zhong et al. 2005). These losses, in combination with altered soil physical properties and water availability have a strong impact on species composition and long-term succession direction in this ecosystem (Banyikwa 1988; Liu et al. 1998; Li and Li 2002). It has been demonstrated that, when heavy grazing persists on degraded grasslands, ruderals, typically unpalatable herbaceous and woody perennials, become dominant (Milton et al. 1994). In the study area, the abundance of Leymus chinensis and Stipa grandis is substantially reduced under high grazing pressure (Li 1989; Liu et al. 1998; Xie and Wittig 2003) whereas dicotyledonous species, for example Artemisia frigida, Potentilla acaulis, the C4 grass Cleistogenes squarrosa, and the sedge Carex duriuscula are typical grazing increasers (Liu et al. 1998; Li and Li 2002; Gao et al. 2007).

In addition to grazing disturbance, drought can significantly affect species composition and functional components, because of species differences in drought resistance or tolerance (Biondini et al. 1998; Thurow and Taylor 1999; Bai et al. 2001; Briggs and Knapp 2001; O’Connor et al. 2001; Nippert et al. 2006). Generally, drought-tolerant species or functional groups become more abundant and/or make a greater contribution to total community productivity in relatively dry years (Bai et al. 2001, 2004; O’Connor et al. 2001). Briggs and Knapp (2001) found peak biomass of grasses correlated strongly with soil water availability whereas biomass of forbs did not. However, drought and grazing always interactively affect ecosystem processes in grazed grasslands (Milchunas et al. 1989; Hartvigsen 2000). Drought stress on heavily grazed sites is likely to be manifested more frequently and more severely than on ungrazed sites because of a higher portion of unproductive water loss by evaporation from bare soil. Consequently, plants from heavily grazed sites may have less water available for growth compared with ungrazed sites (Fahnestock and Detling 2000). Therefore, our hypothesis is that drought combined with heavy grazing can drastically accelerate the shift speed of species or functional groups because of species differences in drought resistance and grazing tolerance.

The relationship between species richness and grassland productivity is fundamental to the management and preservation of biodiversity, but remains controversial when different scales (spatial and temporal), disturbances, stressors (abiotic and biotic), and ecosystems are considered (Mittelbach et al. 2001). The relationship between biodiversity and productivity is strongly affected by the mode and severity of disturbance in semiarid steppe ecosystems, and it has been shown that heavy grazing can lead to both lower productivity and lower species diversity (Zhou et al. 2006; Bai et al. 2007). However, few studies have investigated potential interactions of grazing disturbance and drought on species diversity and productivity (Cingolani et al. 2005).

In the work reported here, we investigated:

  1. 1

    whether different species or functional groups responded differently to grazing over three years with variable amounts of rainfall; and

  2. 2

    whether the relationship between species richness, diversity, and productivity was similar on differently managed sites facing extreme weather conditions.

Materials and methods

Experimental sites

The principal study area is the Xilin River watershed (3,800 km2) in Inner Mongolia. A short vegetation period of approximately 150 days (Xiao et al. 1995) and precipitation, mainly of convective nature in combination with high evapo-transpiration, are the factors most limiting plant growth of the typical steppe which covers more than 60% of the Xilin river catchment. Main soil types in the low hills and plains are chestnut soils and chernozems with high soil organic carbon and total N contents. Main vegetation types are L. chinensis and S. grandis communities with ground cover of 30–70%, depending on precipitation between the years and a vegetation height of 0.3 m in the long-term exclosures. This steppe is mainly used for pastoralism, with only a few fertile sites of artificial grassland used for hay production. Compared with the local long-term average precipitation from May to September (294 mm), rainfall was 288, 131, and 212 mm in 2004, 2005, and 2006, respectively, indicating that 2005 and 2006 were relatively dry years.

The grazing experiment was conducted at three sites within this area. One site of 24 ha, established and maintained by the Inner Mongolia Grassland Ecosystem Research Station (IMGERS), Institute of Botany, the Chinese Academy of Sciences, has been protected from grazing since 1979. Within this site (ungrazed since 1979, hereafter denoted UG79), a representative plot of 150 m × 150 m was used for sampling. The second site of 40 ha has been used for winter grazing since 2000 (winter grazing, hereafter denoted WG). Before 2000, this area was continually grazed at a low to moderate intensity (two sheep per hectare). Within this site, a representative plot of 200 m × 200 m was used for sampling. The sampling plots of UG79 and WG were 600 m apart. The third site of approximate 100 ha has been heavily grazed in the past 30 years during the vegetation period. The average grazing intensity during the last 10 years has been four sheep per hectare. Within this heavily grazed site (hereafter denoted HG), a representative area of 200 m × 200 m was selected for sampling. Site HG is 3.3 km from sites UG79 and WG. Soil types of these three sites are similar, although site HG has a slightly higher coarse sand fraction in the very topsoil. For further soil characterization, see Krümmelbein et al. (2006).

Biomass harvest

The aboveground net primary production (ANPP) was determined by measuring the peak biomass, because our study sites contained mostly herbaceous plants with above ground tissues that die back annually. The aboveground biomass of individual species was harvested by clipping biomass within ten 1 m × 1 m quadrats at the three sites down to the soil surface in the middle of August of 2004–2006. The tillers or stolons of each species were counted in each quadrat and the height was measured with a meter rule. Shrubs and semishrubs (for example Caragana microphylla and A. frigida) were clip-harvested in late April and mid-August and biomass differences were considered to represent ANPP. Dry mass weight was measured after oven-drying at 75°C to constant weight. Species were grouped into categories of total community and plant functional group according to Bai et al. (2004) on the basis of life form: annuals and biennials (AB), perennial rhizome grass (PR), perennial bunchgrass (PB), perennial forbs (PF), perennial grass (PG), and shrubs and semishrubs (SS).

Data calculation

The position and/or competitive ability of each species in the community is partially reflected by its relative biomass (Connolly and Wayne 2005). Therefore, the relative biomass was expressed as the ratio of the contribution of single-species biomass to total biomass. Average height of the plant community was calculated from those species for which relative biomass was more than 1%. The Simpson diversity index (D) was used to describe species diversity

$$ D = 1\left/\sum\limits_{1}^{S} {({{N_{i} } \mathord{\left/ {\vphantom {{N_{i} } N}} \right. \kern-\nulldelimiterspace} N})^{2} } \right.$$

where S is the number of species per quadrat, N i is the biomass of species i, and N is the total biomass per quadrat.

Statistical analyses

Because of the field-history of the long-term established grassland sites, treatment comparison was unavoidably based on pseudo-replicated data collection. Statistical analyses were performed using SAS version 8.0. Normality distribution of residuals was tested with Proc Univariate. Transformation of variables was not required in any case. ANPP, community height, plant density, species diversity, and different functional components were subjected to analysis of variances by Proc Mixed including grazing treatments (fixed effect) and years (random effect). In the case of non-significant interaction between grazing treatments and years, comparison of main factor means was based on Tukey’s HSD test with an experimental wise error of 0.05. Because of the significant interaction between grazing treatments and years, multiple comparisons of ANPP, plant density, and different functional components least square means were based on the Bonferoni–Holm procedure with nine comparisons. The significance of species differences between years was analysed with one-way analysis of variance, and comparison of means was based on Tukey’s HSD test with an experimental wise error of 0.05.

Results

ANPP, community height, and total plant density

ANPP was significantly higher in the year with average rainfall (2004) compared with the dry years 2005 and 2006 (Tables 1, 2). In all years ANPP at site UG79 was significantly higher than that at site HG. ANPP at site UG79 was significantly higher than that at site WG in 2004, and ANPP at both sites was similar in 2005 and 2006. ANPP of site WG was not significantly different from that of site HG in 2004; it was higher in 2005 and 2006. Average community height was significantly higher at sites UG79 and WG than at site HG; it was similar at sites UG79 and WG. The community height was significantly higher in 2004 than in the dry years 2005 and 2006 (Tables 1, 2). In all three years, total plant density was highest at site HG and lowest at site UG79 and was higher in 2004 than in 2005 and 2006 (Tables 1, 2). Total plant density decreased from 2005 to 2006 (in the second dry year) at sites WG and HG, but not at site UG79.

Table 1 The probability (Pr > F) of two-factors ANOVA on aboveground net primary productivity, community height, plant density, different functional groups, and species diversity
Full size table
Table 2 Aboveground net primary productivity (ANPP), average height of community, and total plant densities at differently managed sites in 2004, 2005, and 2006 (X ± SE)
Full size table

Relative biomass of single species

Leymus chinensis, S. grandis, Carex spp., Agropyron michnoi, and C. squarrosa were found at all three sites (Table 3). Caragana microphylla was only found at site UG79 and its relative biomass ranged from 4 to 7%. Anemarrhena asphodeloides was absent at site HG, and quite rare at site WG, while its relative biomass varied between 5 and 7% at site UG79. Achnatherum sibiricum had high relative biomass at sites UG79 and WG, while it was completely absent at site HG, and relative biomass of C. squarrosa (C4, bunchgrass) at site HG was substantially higher than at sites UG79 and WG.

Table 3 The relative biomass contribution of single species (%) to total biomass at three differently managed sites (see footnote of Table 2) in 2004, 2005, and 2006
Full size table

Sites and precipitation had strong interactive effects on some species (Table 3). Chenopodium aristatum and Chenopodium glaucum usually are opportunity species at all three sites, but heavy grazing in combination with drought significantly increased the relative biomass of these two species. The relative biomass of L. chinensis was reduced under drought, and heavy grazing with prolonged drought made this species almost absent at site HG. Carex spp. did not respond consistently at the three grassland sites in the dry years 2005 and 2006. Relative biomass was reduced at site UG79 and increased at site WG; there was no clear trend at site HG. However, this species made a large contribution to total biomass at all three sites after the two dry years. Relative biomass of S. grandis increased in dry years and this trend was most evident at site WG in the extremely dry year 2005. Similarly, Achnatherum sibiricum did not respond to drought at site UG79 whereas its relative biomass decreased significantly after drought at site WG. P. acaulis, which is regarded as an important indicator of heavy grazing, was a rare species at sites UG79 and WG, and was dominant at site HG in 2004 and 2005. However, it was almost absent in the second dry year. A. frigida, a semi-dwarf shrub, which was not found at site WG and was a rare species at site UG79, became the dominant species at site HG after the two dry years. Total numbers of species involved in the contribution to total biomass are quite different among these sites (Table 3). Approximately 80% of total biomass came from more than 15 species at site UG79 whereas more than 90% of total biomass was from only five species at site WG and six or seven species at site HG.

Functional groups

Variation of precipitation and grazing interactively affected plant functional group composition (Table 1; Fig. 1). Usually, AB only made very small contribution to total biomass at all sites, however, it became important at site HG after two continuous dry years and contributed 22% to total biomass (Fig. 1a). SS exhibited a more or less stable contribution to total biomass over time at site UG79 (Fig. 1b) and was not found at site WG, while SS significantly increased its contribution to total biomass with drought at site HG (Fig. 1b). PR, PB, and PF were the most important components of this perennial grassland. At site UG79, PG accounted for 60 and 69% of total biomass in 2004 and 2005, whereas its contribution to total biomass significantly decreased in 2006 (Fig. 1f), when PF made a greater contribution to total biomass (Fig. 1e). The reduction of PG was due to a decrease of PR (Fig. 1c), while the relative biomass of PB remained more or less stable (Fig. 1d). At site WG, PG was, irrespective of years, by far the most dominant plant functional group and accounted for more than 97% of total biomass (Fig. 1f). The relative biomass of PR and PB compensated each other from year to year (Fig. 1c, d). PF made very small contribution to total biomass at this site (Fig. 1e). PF and PG varied greatly from year to year at site HG, contribution of PF and PG to total biomass ranging from 5 to 44 and 49 to 75%, respectively (Fig. 1e, f).

Fig. 1
figure 1

The relative biomasses of different plant functional groups in 2004, 2005, and 2006 at three grassland sites of Inner Mongolia. Site UG79 has been protected from grazing since 1979, site WG is used for winter-grazing, and site HG is heavily grazed. AB, annuals and biennials; SS, shrubs and semishrubs; PR, perennial rhizome grass; PB, perennial bunchgrass; PF, perennial forbs; PG, perennial grass. A significant interaction occurred between sites and years for all different functional groups; differences among years within a site are indicated by letters (a–c) and differences among sites within a year by letters (x–z)

Full size image

Species richness and species diversity

In total, 55 species were found at site UG79, and 25 and 26 species at sites WG and HG, respectively, in 2004, a relatively wet year (Table 4). The total number of species decreased at all three sites after two drought years, and variation of species richness between years was mainly from functional groups AB and PF, whereas perennial grasses (PR + PB) and SS remained more stable during extreme drought events (Tables 3, 4). The Simpson diversity index was not significantly (P = 0.126) different between years but tended to decrease with drought (Tables 1, 5). Grazing had a strong effect on species diversity index (P < 0.0001), and the index was significantly higher at site UG79 than at sites WG and HG, while no significant difference between the other two sites was found during the study period.

Table 4 Species richness of different functional groups on the differently managed sites in 2004, 2005, and 2006
Full size table
Table 5 Simpson diversity index at differently managed sites in 2004, 2005, and 2006
Full size table

Discussion

ANPP, community height, and total plant density

From our data it is obvious that community height was reduced more by heavy grazing than by drought (Table 2), and that drought and heavy grazing had opposite effects on total plant density. Drought tended to reduce it and heavy grazing significantly increased total plant density (Table 2). Therefore, drought and grazing had interactive effects on plant height and density, which resulted in different ANPP under fluctuating levels of rainfall (Table 1). It must be mentioned that although plants produced more small and thin tillers or stolons, this increase could not compensate for the sharp decrease of the height, which led to lower ANPP at site HG.

Relative biomass of species

Grazing-sensitive species, for example Achnatherum sibiricum and Anemarrhena asphodeloides, disappeared at the heavily grazed site whereas P. acculis, A. frigida, and C. duriuscula became dominant under heavy grazing conditions. These species are adapted to grazing because of their strong gap colonization ability and prostrate growth habit, which are effective grazing avoidance mechanisms (Milchunas and Noy-Meir 2002; Bullock et al. 2001). Grazing tolerance is another important adaptation mechanism to grazing. C. squarrosa, despite being preferentially eaten by sheep, has a high overcompensatory ability because of strong tillering and high photosynthetic capacity (Liang et al. 2002; Wang and Wang 2001; Gao et al. 2007), which makes it an important species on the heavily grazed site. The relative biomass of L. chinensis was significantly reduced after drought at sites UG79 and WG. Because of sheep preference for this species much more in drought years than in years of normal rainfall (Wang 2000), it was almost completely absent at site HG after two dry years (Table 3).

Carex spp. characterised by strong colonization ability and drought tolerance (Liu et al. 1998) can produce many thin and small tillers or stoloniferous branches, so it occupies huge horizontal space. This strategy can partially explain why Carex spp. became an important species at all three sites after drought (Table 3). It should be noted that Carex has two main species in this area, more than 98% of Carex spp. is Carex korshinskyi at sites UG79 and WG, and more than 98% of Carex spp. is C. duriuscula at site HG (personal communication with Bai). C. duriuscula has stronger colonization ability and drought tolerance than Carex korshinskyi.

Annual and biennial usually make a very small contribution to total community productivity in terms of relative biomass and species richness in this perennial steppe grassland (Fig. 1a; Table 4). The increases in these species were only temporary and highly affected by rainfall distribution and abundance. In the first dry year, 2005, almost no annual species, for example Chenopodium glaucum, Chenopodium aristatum L., and Salsola collina, were found. However, these opportunity species have the ability to quickly occupy space during periods of water availability, resulting in a biomass contribution of 22% in 2006 at site HG where the amount of bare soil was large after the first dry year, 2005.

Carry-over effects of drought on species were reflected by the substantial increase of biomass contribution of Agropyron michnoi and P. acculis in 2005 compared to 2004. Because of the favourable growth conditions in 2004, these two rhizome species probably accumulated enough root carbohydrates for the second year’s growth, but failed to do so in 2005, resulting in a substantial decrease of relative biomass in 2006.

Competition and facilitation for resources are two of the most decisive factors determining vegetation pattern (Michalet 2007) and the relative importance of facilitation and competition varies inversely across gradients of abiotic stress, with facilitation dominant under highly stressful conditions (Brooker and Callaghan 1998). However, it seems that interspecific competition, instead of facilitation, plays a vital role in determining community structure and species composition under harsh environmental conditions (Zhou et al. 2006). Despite its strong colonization ability and low palatability (Wang 2000), the abundance of drought-susceptible P. acculis was substantially reduced on site HG after two dry years, and drought-tolerant A. frigida (Gao et al. 2007) and opportunity species of AB significantly increased because of release from competition by P. acculis. It has been argued that high abundance of P. acaulis is indicative of severe degradation of the grassland (Liu et al. 1998; Li and Li 2002) and it is regarded as the last perennial species in this native grassland, which in cases of disappearance indicates a high danger of extinction (Li and Li 2002). However, as suggested by our data, prolonged drought periods may reduce the predictability of degradation by means of this typical indicator species, because the decrease of P. acculis is partially compensated by A. frigida.

Species compensatory effects

A growing number of experiments have demonstrated that species compensatory effects can ensure greater and more stable production (Tilman and Downing 1994; Bai et al. 2001, 2004). Types of compensatory effect are very different in different land-use types. PR and PF compensated each other, while PB and SS remained more or less stable in the long-term exclosure (Fig. 1); the sharp decrease of PF was mainly compensated by functional groups AB and SS after two dry years on the heavily grazed site; compensatory effects between species were more evident on the winter grazing area, where the perennial grasses were overridingly dominant (Fig. 1f). The relative decrease of L. chinensis was partially compensated by the relative increase of Carex spp. and Achnatherum sibiricum was mainly compensated by S. grandis in the dry year (Table 4). However, species and/or functional groups compensatory effects did not prevent low total productivity in 2005 and 2006 at all three sites (Table 1).

Species diversity

A review (Mittelbach et al. 2001) of 171 published studies revealed the relationship between species richness and productivity is hump-shaped and positive. To stress this relationship in specific grazing ecosystems, Cingolani et al. (2005) argued that grazing can substantially change species composition, instead of species richness or diversity, based on the assumption that the numbers of less grazing-resistant species are similar to the numbers of grazing-resistant species. In contrast with argument, our data illustrated that heavy grazing significantly reduced species richness and biodiversity. In addition, the positive relationship between productivity and species diversity was only found in the relatively normal year 2004, and no relationship was observed between species diversity and productivity in the dry years 2005 and 2006. This discrepancy indicates that regulatory processes at the community level in response to varying soil nutrient and water availability are quite complex in grassland ecosystems in general, and in this semi-arid grassland in particular (Baer et al. 2003; Cingolani et al. 2005). When one resource, for example soil water availability, becomes limiting, all ecosystem processes are slowed down. In addition, the responses of productivity to drought are quicker than species richness, which leads to the divergence between productivity and species diversity. It should be noted that this ecosystem seems to have greater resilience, as indicated by the fact that none of the important perennial species disappeared after extreme drought, despite high fluctuation of species contributions to community production in different years.

In conclusion, drought stress and grazing disturbance interactively control species composition and community structure in this ecosystem. Nonetheless, we emphasized that severe drought combined with heavy grazing has greatly changed species composition and functional components, which led to unstable production and degradation of this perennial grassland.