Drought and grazing drive the retrogressive succession by changing the plant–plant interaction of the main species in Inner Mongolia Steppe

Abstract Plant–plant interactions play a key role in the function and structure of communities. The combined effect of drought stress and grazing disturbance on shaping plant–plant interactions is still poorly understood, while this combination is common in semiarid ecosystems. Four species including Stipa grandis, which is dominant in the typical steppe, and Stipa krylovii, Artemisia frigida, and Cleistogenes squarrosa, which are dominant species in the S. grandis degraded communities, were selected as study targets. We conducted a competition experiment (uniformly dense monoculture or mixture, respectively) under controlled conditions, including both drought stress and mowing disturbance, and calculated the relative interaction index (RII) of tiller number and RII of biomass for each species under each condition. (a) Under the same condition, the RII of tiller number and that of biomass for the same species usually showed reverse trends. (b) Mowing disturbance rather than drought stress played a negative role in influencing S. grandis’ or S. krylovii’s RII of tiller number and played a positive role in influencing A. frigida's RII of biomass. (c) Drought stress rather than mowing disturbance played a positive role in influencing C. squarrosa’s RII of tiller number. (d) Neighbor species significantly influenced S. grandis’ RII of tiller number, S. krylovii’s RII of tiller number, A. frigida's RII of tiller number and biomass, and C. squarrosa’s RII of biomass. These results could provide an explanation for why S. krylovii, A. frigida, and C. squarrosa can replace S. grandis and become the dominant species when S. grandis communities undergo a process of degradation due to overgrazing or climatic drought in natural communities. The present study provided powerful evidences for species replacement in the typical steppe of Inner Mongolia and elucidated the driving mechanisms of S. grandis communities’ retrogressive succession.


| INTRODUC TI ON
Understanding the process of retrogressive succession of plant community is very important in the restoration and protection ecology as well as in applied ecology (He & Bertness, 2014). Plant-plant in-teraction is thought to be the fundamental for determining community dynamic and one of the most fundamental issues in community ecology, especially how the balance between positive interaction (competition) and negative interaction (facilitation) responds to variation in environmental stress and disturbance (Bertness & Callaway, 1994;Brooker et al., 2008;Bruno, Stachowicz, & Bertness, 2003;Grime, 1974). The negative change of plant-plant interactions of dominant species could lead to being replaced by the prosperity of others (Soliveres & Maestre, 2014). Therefore, understanding how plant-plant interactions change with the environmental condition is the core to explain the community succession and to predict the community dynamics in the face of environmental changes.
The shift of competitive superiority from one species to another species would happen in stress or disturbance condition, which could cause the retrogressive successive. More recently, researchers have focused on understanding how species characteristics (ecotype, tolerance) and stress types affect competitive capacity in themselves.
Although the tolerance or resistance of the species in the same community are thought to be similar, it shows difference when these species are under the stress or disturbance condition. On the other hand, many documents have reported that neighbor species could play an important role in affecting the competitive hierarchy of the target species (Ulrichet et al., 2016). That is to say, the plant-plant interaction could be positive if the target species grows with one species or negative if it grows with another species. As a result, the target species could keep its superiority with one species but lose its superiority with another species (Olsen, Topper, Skarpaas, Vandvik, & Klanderud, 2016). Even though the relationships between the characteristics (or stress types) of the target species and the result of the plant-plant interaction are complicated and may be species specific, it goes to the core of retrogressive succession. Therefore, fully understanding such relationships of the main species in a particular community is very important for ecologist.
The semiarid Inner Mongolian Steppe of China is an important part of the terrestrial ecosystem of the Eurasian steppe, and plant communities are often dominated by the tussock species Stipa grandis (Chen, Zhao, Zhang, & Gao, 2013;Lu & Wu, 1996). However, due to drought caused by climate change and overgrazing by cattle or/ and sheep, a large area of grassland has become degraded. During the process of retrogressive succession, the dominant species S. grandis is gradually replaced by Stipa krylovii in the light degraded communities, by Artemisia frigida and Cleistogenes squarrosa in the heavy-degraded communities. Accompanying with the replacement of dominant species is the loss of diversity and the steep decline in aboveground net primary productivity (Pan et al., 2016;Wan et al., 2015). Previous studies have focused more on water utilization and the life history to explore the species replacement during the degradation of S. grandis communities (Zhang, Zhu, Liu, & Pan, 2014).
Researchers have tried to examine how drought or disturbance influenced the plant-plant interactions of the main species to provide some explanations for the species replacement in some other communities (Michalet et al., 2014;Smit, Rietkerk, & Wassen, 2009).
Moreover, although some hypotheses, such as stress gradient hypothesis (SGH) or niche complementarity, could help us to predict how stress or disturbance affects the results of plant-plant interactions, they could not help us to efficiently predict how a combination of stress and disturbance shapes plant interactions (Barner, Hacker, Menge, & Nielsen, 2016;He, Bertness, & Altieri, 2013). Therefore, understanding how the combination of grazing disturbance and drought stress influences plant-plant interactions of the main species in Inner Mongolia Steppe of China is of critical importance for explaining the community retrogressive succession in this region, and guiding sustainable management practices of semiarid grassland ecosystems of China.
Therefore, in the present study, we selected the four species, S. grandis, S. krylovii, A. frigida, and C. squarrosa, as study subjects, and conducted a two-factor (drought and mowing) isodensity displacement experiment to assess the variation of the relative interaction index (RII) between pairwise species under different conditions. Specifically, we intended to address the following questions.

| Experimental design
We conducted a two-factor (drought stress and mowing disturbance) isodensity displacement experiments, and each factor had two levels (present (+) and absent (−)). Soil moisture in the drought treatments (D+) was maintained at 8% ± 3%, which was equivalent to that of a semiarid grassland, while that in the nondrought treatments (D−) was kept at 15% ± 3% (Zhang et al., 2014). A soil compaction moisture tester (SCMT) was used to monitor the soil moisture every three days. Mowing was carried out to simulate moderate-to-heavy grazing disturbance, and the aboveground plant biomass was removed to a stubble height of 10 cm twice during the experiment.
We selected seedlings of equal size of the four species from seedling trays two weeks after germination and transplanted them into pots (10 cm in diameter and 20 cm in height) containing vermiculite (150 g) on 1st March 2014. According to the principle of replacement, there were ten types of planting, including 6 twospecies mixtures (S. grandis-S. krylovii, S. grandis-A. frigida, S. grandis-C. squarrosa, S. krylovii-A. frigida, S. krylovii-C. squarrosa, and A. frigida-C. squarrosa) and four monocultures. There were seven replicates per treatment, and we got 280 pots (10 × 2×2 × 7) in total. In mixtures, the one is the target species and the other is its neighbor species, therefore, each species had three different neighbor species per treatment. Mowing was performed on 1st May (2 months after plantation) and 1st July (4 months after plantation) 2014.
The study site was established in the experimental field at Nankai University in Tianjin (39°10N, 117°10E), China, where the mean annual air temperature is 12.6° and the mean annual precipitation over the last 40 years is 523.2 mm. The pots were placed under a rain-proof shed to avoid natural precipitation. The shed was at a height of 5 m and could penetrate 95% natural light. The locations of the pots were randomly changed every month.
On 1st October 2014, we measured tiller number and harvested the plants by individual. The harvested individuals (including shoot and root) were dried at 70°C to a constant weight.

| Data analysis
The relative interaction index (RII) of each target plant species was estimated by using the following formulas (Armas, Ordiales, & Pugnaire, 2004): RII = P +N −P −N P −N +P +N . P +N represents the target species' performance of tiller number (biomass) in mixture with a neighbor species under one treatment; P -N represents the target species' average performance of tiller number (biomass) of an individual in monocultures under the same treatment. The RII of tiller number (biomass) was calculated according to three different neighbor species within each drought treatment (2) and disturbance treatment (2) cross-treatment, totaling 12 RIIs of tiller number (biomass) for each target species. The RII varies from −1 to 1, with a negative value indicating a negative interaction, a positive value indicating positive interaction, and nonsignificant difference with 0 indicating a neutral interaction.
In order to estimate the effects of drought stress, mowing disturbance, and neighbor species on each target species' RII of tiller number (biomass), general linear model (GLM) was used with RII as dependent variables, drought stress, mowing treatment and neighbor species as fixed factors. If there were significant effects of the main factor(s) and no significant interaction between main factors, independent-sample test or one-way ANOVA was used to measure the effect of main factor(s) on the variables. In addition, one-sample t test was applied to assess the difference between each species' RII of tiller number (biomass) under each treatment and 0. Analysis was carried out in R (3.1.1, www.r-project.org), and the figures were created with Sigmaplot 12.5.

| S. grandis' RII of tiller number and biomass
The overall effects of drought (D) and mowing (M) significantly F I G U R E 1 One-way ANOVA of drought (a) and mowing (b) on Stipa grandis' RII of tiller number. The different lowercase letters represent significant difference between treatments at 0.05 level. The asterisk represents significant difference between the average value of RII and 0 at 0.05 level

| S. krylovii's RII of tiller number and biomass
The overall effects of N, M, D × M, and M × N were significant (p < 0. 05) on S. krylovii's RII of tiller number (Table 2) Greek letters within the same treatment represent nonsignificant difference among different neighbor species. The asterisk represents significant difference between the average value of RII and 0 The p values that are lower than 0.05 are in bold type.

| A. frigida's RII of tiller number and biomass
The

| C. squarrosa's RII of tiller number and biomass
The overall effects of N and D were significant (p < 0. 05) on C. squarrosa's RII of tiller number. The result of main effect analysis showed that C. squarrosa's RII of tiller number under drought treatment was higher than 0 and higher than that under nondrought treatment Greek letters within the same treatment represent nonsignificant difference among different neighbor species. The asterisk represents significant difference between the average value of RII and 0 at 0.05 level The p values that are lower than 0.05 are in bold type.

| D ISCUSS I ON
Our study revealed the reverse trends between the RIIs of tiller number and the RIIs of biomass for the same species under the same condition. That is to say, under the same condition, if target species' RII of tiller number was positive, its RII of biomass usually was negative (Figures 1 and 5). Therefore, in the present study, we used RII of tiller number to show the competitive capacity of the three Gramineae species, and RII of biomass to show the competitive capacity of the small subshrub A. frigida because A. frigida expands by generating adventitious roots from its tillers.
Our study showed that the mowing disturbance but not drought stress was the main driver of reducing competitive capacity of S. grandis, which would cause the retrogressive succession of S. grandis communities. First, disturbance of mowing played more important role in affecting S. grandis' RII of tiller number than drought stress ( Figure 1). As the dominant species in the typical steppe of Inner Mongolia Steppe, S. grandis exhibits superior growth relative to the other species in the community, but disturbance by mowing would decrease its competitive advantage. High competitive ability and low tolerance to disturbance enable a species to minimize the costs of negative interactions and to maximize the benefits of habitat amelioration provided by neighbors (Liancourt, Callaway, & Michalet, 2005).
For example, all the neighbor species showed significantly positive effects on S. grandis' RII of tiller number under un-mowing treatment, but no significant effect was found under mowing treatment (Tables 1 and 2; Figures 1 and 2). While, comparing with the RIIs under nondrought treatment, the RIIs of tiller number were higher under drought treatment. Mowed target species may suffer more under drought condition than under nondrought condition and thus may be more demanding of stress mitigation by neighbors, as shown in physically disturbed communities (Maalouf, Bagousse-Pinguet, Marchand, Touzard, & Michalet, 2012). Second, drought increased S. grandis's RII of tiller number but did not affect the counterpart of S. krylovii (Figures 2 and 3), indicating that drought stress could not be the main driver for the replacement of S. grandis by S. krylovii.  Table 2) in affecting it, which could cause the replacement of S. krylovii by other species, such as A. frigida (Zhang, Li, Jiang, Lin, & Du, 2005).

Source of variation
Furthermore, the researches about changes in competition hierarchies caused by physical environment or/and neighbor specie are one of the most attractive issues in modern ecology (Jiménez & Rossi, 2012). A few studies have manifested that neighbor species could change species competition hierarchies (Chen et al., 2013;Michalet et al., 2014). Under D−M+ treatment, S. grandis' RII of tiller number was negative when S. grandis was grown with A. frigida and neutral when S. grandis was grown with the other two species (Figure 2b) (Yang, Bai, Li, & Han, 2009).
Comparing with the other three species, the stronger tolerance to drought played an important role in helping C. squarrosa become the dominant specie in the heavy-degraded communities. C. squarrosa's RII of tiller number was significantly affected by drought, mowing, and neighbor species, especially drought ( Figure 5a, Table 4).
Drought significantly increased C. squarrosa's RII of tiller number, especially under drought and mowing (D+M+) treatment (Figure 6c, 6d), and C. squarrosa's RIIs of tiller number were positive under drought treatment and were not influenced by neighbor species (Figure 5b).
As a C 4 plant, C. squarrosa has a higher photosynthetic capacity and water-use efficiency than the C 3 plants and a relatively higher adaptive capacity to drought (Wang, 2004;Wang, Wang, & Chen, 2003).
In heavy-degraded steppe, similar results that C. squarrosa could take advantage of increasing tiller and become dominant species have been reported by researchers (Qin et al., 2017;You & Yang, 2016).
These results supported that the changes in growing abilities and drought-induced habitat amelioration co-determine the dominance of the species in the community (Hacker & Gaines, 1997).

| CON CLUS IONS
The

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R S CO NTR I B UTI O N
Shaobo Gao wrote the manuscript and accomplished the experiment, Zhirong Zheng analyzed the data, Yukun Wang and Lei Liu took part in the experiment and collected data, Nianxi Zhao designed the experiment and revised the manuscript, and Yubao Gao revised the manuscript.

DATA ACCE SS I B I LIT Y
Data underlying this article can be accessed on Dryad. We agree to deposit our data to a public repository.