Biomass Allocation and Competitive Ability of a Semiarid Perennial Grass and a Legume in Mixtures under Periodical Soil Water Decreasing Conditions

: Soil moisture is the key factor controlling plant growth in semiarid grasslands. Here, we sought to evaluate the effects of soil moisture decreasing at different growth stages on biomass accumulation, water use efficiency, and plant-plant interaction of a C 3 leguminous subshrub Lespedeza davurica (L) and a C 4 perennial grass Bothriochloa ischaemum (B) when sown singly and as a mix in ﬁve different ratios in a pot experiment. Results showed that soil water decrease signiﬁcantly reduced total biomass production of the mixtures by 3.7–53.8% compared with well-watered conditions, and plants at the heading and ﬂowering periods were more vulnerable to soil water decline than those at the late stage. The relative yield total (RYT) of the mixtures was mostly greater than those sown singly. Soil water decreasing increased root/shoot ratio and water use efﬁciency (WUE) of the mixtures, and such effects were mediated by mixture ratio and/or growth stage. In the mixtures, a strong intraspeciﬁc competition was observed in B. ischaemum , whereas interspeciﬁc competition in L. davurica . The highest overall biomass (86.47 g pot − 1 ) and WUE (6.33 g kg − 1 ) were observed when the mixture ratio was B:L = 10:2 regardless of soil moisture, and thus could be considered an optimal mixture ratio for establishing restored grassland using the two species. Our results suggest that sown seed mixtures of the two species with an appropriate ratio could sustain a relatively high total biomass production and improve WUE under soil water decreasing conditions in the semiarid Loess Plateau.


Introduction
The semiarid grassland consists of ca. 30-40% total land area on the Loess Plateau of China and provides key ecosystem functions such as biodiversity preservation, carbon storage, and soil and water conservation [1,2]. Water is the key environmental factor limiting grassland growth, and discrete rainfall events following periodic droughts are the common field condition in the region [3][4][5]. Although the annual precipitation seems to be unchanged over the whole Loess Plateau region [6], the precipitation regime may shift to more extreme rainfall events following frequent and prolonged droughts [7]. Limited soil water conditions could trigger plant reactions from cellular to the whole plant level and often results in adverse effects on plant growth [8]. Thus, understanding the physiological responses of plants under drought conditions is of great significance to predict the regional vegetation distribution and its dynamics under altered precipitation regimes in the context of climate change [9,10].
Soil water decreasing treatments (i.e., soil drying) were conducted according to the growth stage of B. ischaemum. When B. ischaemum was at the heading period ( Figure 1). After watering regimes were 'decreased' to the target regimes, they were kept consistent until the end of the growing season (10 October 2013). Pots under three soil water regimes (i.e., H, M, and L) without any soil water decreasing throughout the whole growing season were taken as references and were noted as 'no decreasing'. A total of 420 pots were used: 7 (mixture ratios) × 12 {3 (unchanged soil water regimes throughout the whole growth stage) + [3 (soil water decreasing treatments) × 3 (growth stages)]} × 5 (replicates). The mixture ratios and soil water decreasing treatments were assigned to pots following a completely random design, and pots were randomly placed on the benches under the rainout shelter. stage) + [3 (soil water decreasing treatments) × 3 (growth stages)]} × 5 (replicates). The mixture ratios and soil water decreasing treatments were assigned to pots following a completely random design, and pots were randomly placed on the benches under the rainout shelter.

Biomass Harvest
At the end of the growing season on 10 October 2013, the shoots and roots of species were harvested from randomly chosen three pots per treatment. Aboveground parts were carefully harvested. Roots were gently washed from the soil, and no root nodulation was observed in L. davurica. The above-and below-ground parts were oven-dried to constant weight at 75 °C and weighed separately. The root/shoot ratio (RSR) was obtained by dividing the weight of belowground and aboveground parts.

Water Use Efficiency
Each pot was weighed every day at 18:00 h to estimate daily water loss from the pot. Three additional pots with the identical setup but without growing plants per soil watering treatment were weighed daily at 18:00 h to estimate water loss via soil evaporation. Then, plant water use efficiency (WUE) over the season was calculated as total biomass accumulation divided by total plant water use (summing daily water loss minus daily soil evaporation).

Competitive Indices
The inter-and intra-specific competition of both species were compared by relative competition intensity (RCI) [29], and to assess the relative competitiveness of each species in mixtures, the competitive balance (CB) index was calculated [30]. The biological efficiency of the mixed cropping system was assessed by the relative yield total (RYT) [31]. These indices were calculated as follows:

Biomass Harvest
At the end of the growing season on 10 October 2013, the shoots and roots of species were harvested from randomly chosen three pots per treatment. Aboveground parts were carefully harvested. Roots were gently washed from the soil, and no root nodulation was observed in L. davurica. The above-and below-ground parts were oven-dried to constant weight at 75 • C and weighed separately. The root/shoot ratio (RSR) was obtained by dividing the weight of belowground and aboveground parts.

Water Use Efficiency
Each pot was weighed every day at 18:00 h to estimate daily water loss from the pot. Three additional pots with the identical setup but without growing plants per soil watering treatment were weighed daily at 18:00 h to estimate water loss via soil evaporation. Then, plant water use efficiency (WUE) over the season was calculated as total biomass accumulation divided by total plant water use (summing daily water loss minus daily soil evaporation).

Competitive Indices
The inter-and intra-specific competition of both species were compared by relative competition intensity (RCI) [29], and to assess the relative competitiveness of each species in mixtures, the competitive balance (CB) index was calculated [30]. The biological efficiency of the mixed cropping system was assessed by the relative yield total (RYT) [31]. These indices were calculated as follows: where Y BB and Y LL are dry weights (above/belowground biomass or total biomass depending on indices) of B. ischaemum and L. davurica in monoculture, respectively. Y BL and Y LB

Statistical Analysis
All statistical analyses were conducted using SPSS 17.0 (IBM, Chicago, IL, USA). One-way analysis of variance (ANOVA) with Tukey's test was used to detect differences in WUE, biomass production, and competitive indices among soil watering treatments. Three-way ANOVA was used to test the effects of growth stage, soil water decreasing, mixture ratio, and their interactions on biomass production, WUE, and competition indices of the two species in mixtures. Differences were considered significant at the p < 0.05 level.

Plant Biomass Production (BP)
The total BP and the respective BP of B. ischaemum or L. davurica in the mixtures were significantly affected by soil water decreasing, mixture ratio, growth stage, and their interactions (p < 0.05), except for the interaction of soil water decreasing and mixture ratio on BP of B. ischaemum, and the interaction of soil water decreasing and growth stage on BP of L. davurica (Tables 1 and 2). Table 1. Results of three-way ANOVA testing effects of soil water decreasing (SWD), growth stage (GS), mixture ratio (MR), and their interactions on total biomass production (BP) and water use efficiency (WUE) of the mixtures.  The total BP of the mixture increased with the proportion of B. ischaemum, and the highest BP was observed in the B10L2 mixture ratio and monoculture B. ischaemum ( Figure 2). Replacement series diagrams based on the biomass showed that the curves of B. ischaemum were concave while convex for L. davurica, and those curves were not intersected at any mixture ratio ( Figure 2). The total BP of the mixture increased with the proportion of B. ischaemum, and the highest BP was observed in the B10L2 mixture ratio and monoculture B. ischaemum (Figure 2). Replacement series diagrams based on the biomass showed that the curves of B. ischaemum were concave while convex for L. davurica, and those curves were not intersected at any mixture ratio ( Figure 2).

Root/Shoot Ratio (RSR)
The RSR of both species were significantly affected by soil water decreasing, growth stage, mixture ratio, and their interactions, except for the effect of mixture ratio on B. ischaemum (p = 0.056; Table 2). There were no notably changing trends for the RSR of the two species under each water treatment along with different mixture ratios, particularly in B. ischaemum ( Figure 3). RSR of B. ischaemum ranged from 0.39 to 0.68 in different mixture ratios under varied soil water decreasing and from 0.50 to 1.54 in L. davurica (Figure 3).
When there was no soil water decreasing, the RSR values of the two species were highest under L, moderate under M, and lowest under H (Figure 3 insets). The soil water decreasing treatment increased the RSR in general (Figure 3 insets), with the highest RSR observed under the M-L treatment in both species in all three growth stages.

Root/Shoot Ratio (RSR)
The RSR of both species were significantly affected by soil water decreasing, growth stage, mixture ratio, and their interactions, except for the effect of mixture ratio on B. ischaemum (p = 0.056; Table 2). There were no notably changing trends for the RSR of the two species under each water treatment along with different mixture ratios, particularly in B. ischaemum

Relative Yield Total (RYT)
There were no significant effects of soil water decreasing, growth stage, mixture ratio, and their interactions on the RYT, with the lone exception of a significant interaction of soil water decreasing and growth stage (Table 3). There was no evident changing trend for the RYT values across the different mixture ratios ( Table 4). The RYT at varied mixture ratios was always higher than 1.0, apart from under M treatment at the B2L10 mixture ratio and under M-L treatment at B2L10 and B4L8 during the mature period (Table 4). Table 3. Results of three-way ANOVA testing the effects of soil water decreasing (SWD), growth stage (GS), mixture ratio (MR), and their interactions on relative yield total (RYT) and relative competition intensity (RCI) calculated from biomass of both species, and competitive balance (CB) calculated from total biomass, shoot biomass and root biomass of B. ischaemum.

Relative Competition Intensity (RCI)
The interactive effects of soil water decreasing, growth stage, and mixture ratio, were significant on the RCI of both species (Table 3). The RCI values of B. ischaemum gradually increased (less negative) as its proportion increased in mixtures under each water treatment, but L. davurica showed the opposite (Figure 4). RCI values of B. ischaemum were all negative, which indicating that the competition was stronger in intra-than inter-specific; while RCI values of L. davurica were greater than zero, indicating stronger interspecific competition ( Figure 4). The soil water decreasing treatments increased the RCI of B. ischaemum (particularly under M-L), while decreased the RCI of L. davurica. interspecific competition (Figure 4). The soil water decreasing treatments increased t RCI of B. ischaemum (particularly under M-L), while decreased the RCI of L. davurica.

Competitive Balance (CB)
Same as for RCI, the CB values of B. ischaemum were significantly affected by t interactive effects of soil water decreasing, growth stage, and mixture ratio ( Table 3). values of B. ischaemum calculated from shoot, root, and total biomass were gradually creased with the proportion of B. ischaemum ( Figure 5).

Competitive Balance (CB)
Same as for RCI, the CB values of B. ischaemum were significantly affected by the interactive effects of soil water decreasing, growth stage, and mixture ratio (Table 3). CB values of B. ischaemum calculated from shoot, root, and total biomass were gradually increased with the proportion of B. ischaemum ( Figure 5).

Water Use Efficiency (WUE)
WUE of the mixtures was significantly affected by soil water decreasing, growth stage, mixture ratio, the interaction of soil water decreasing and mixture ratio (Table 1). Furthermore, the WUE increased with the proportion of B. ischaemum ( Figure 6). The WUE of B. ischaemum was significantly higher than L. davurica in monoculture. The B10L2 had the significantly highest WUE in most cases (p < 0.05; Figure 6). Soil water decreasing treatment generally increased the WUE of the mixtures, and this effect was attenuated with the increasing of L. davurica proportion and was greatest at heading and flowering periods ( Figure 6).

Water Use Efficiency (WUE)
WUE of the mixtures was significantly affected by soil water decreasing, growth stage, mixture ratio, the interaction of soil water decreasing and mixture ratio (Table 1). Furthermore, the WUE increased with the proportion of B. ischaemum ( Figure 6). The WUE of B. ischaemum was significantly higher than L. davurica in monoculture. The B10L2 had the significantly highest WUE in most cases (p < 0.05; Figure 6). Soil water decreasing treatment generally increased the WUE of the mixtures, and this effect was attenuated with the increasing of L. davurica proportion and was greatest at heading and flowering periods ( Figure 6).

Discussion
Limited soil moisture in frequently occurring drought events largely constrains vegetation growth in semiarid and arid grasslands [4,5]. In the present study, our results indicated that the limited soil water conditions (60% and 40% of field capacity) across the whole growing season substantially decreased the total biomass production of the C3 subshrub L. davurica and C4 grass B. ischaemum mixtures in a controlled pot experiment, particularly the grass biomass accumulation, which largely affects the total biomass production of the mixtures (Figure 2). The declined biomass under drought conditions in B. ischaemum may be ascribed to inhibited photosynthetic performance under soil water deficit [32]. Contrarily, there was only a limited decline in biomass production of L. davurica subshrub in mixtures, and almost no decline in L. davurica monoculture under limited soil water conditions (Figure 2). These suggested that L davurica was likely more drought-resistant than B. ischaemum no matter in a mixture or monoculture. The great drought resistance of L. davurica may be due to its various physiological and morphological traits. Yang et al. [33] reported that the adjustment of enzymes and osmotic compounds of L. davurica greatly contributed to its drought resistance. Our previous studies indicated that L. davurica had higher maximum photochemical efficiency than B. ischaemum under drought conditions [34], which may also contribute to its greater drought resistance. Furthermore, the root system/morphology determines plants' responses to drought in the regional semiarid grassland [35], the deep-rooted L. davurica could potentially be more accessible to deep soil water than the shallow-rooted B. ischaemum, and thus has greater drought resistance. However, this was unlikely to be the

Discussion
Limited soil moisture in frequently occurring drought events largely constrains vegetation growth in semiarid and arid grasslands [4,5]. In the present study, our results indicated that the limited soil water conditions (60% and 40% of field capacity) across the whole growing season substantially decreased the total biomass production of the C 3 subshrub L. davurica and C 4 grass B. ischaemum mixtures in a controlled pot experiment, particularly the grass biomass accumulation, which largely affects the total biomass production of the mixtures (Figure 2). The declined biomass under drought conditions in B. ischaemum may be ascribed to inhibited photosynthetic performance under soil water deficit [32]. Contrarily, there was only a limited decline in biomass production of L. davurica subshrub in mixtures, and almost no decline in L. davurica monoculture under limited soil water conditions (Figure 2). These suggested that L davurica was likely more drought-resistant than B. ischaemum no matter in a mixture or monoculture. The great drought resistance of L. davurica may be due to its various physiological and morphological traits. Yang et al. [33] reported that the adjustment of enzymes and osmotic compounds of L. davurica greatly contributed to its drought resistance. Our previous studies indicated that L. davurica had higher maximum photochemical efficiency than B. ischaemum under drought conditions [34], which may also contribute to its greater drought resistance. Furthermore, the root system/morphology determines plants' responses to drought in the regional semiarid grassland [35], the deep-rooted L. davurica could potentially be more accessible to deep soil water than the shallow-rooted B. ischaemum, and thus has greater drought resistance. However, this was unlikely to be the case in our pot experiment due to the limited pot size and short experimental duration for root development. Soil water decreasing treatments at varying growth stages revealed that biomass production of the mixtures was more affected by drought at the peak growth stages (heading and flowering periods; Figure 2). Similar results were reported on crops, e.g., Eck et al. [36] showed that the water stress caused the greatest decline in the yield of soybean during the pod development stage; Wagg et al. [37] suggested that the water deficit at the vegetative and tuberization stages affected the growth of potato the most. It is expected that plants commonly require abundant resources such as water/nutrients to sustain their rapid growth and development during these important phases of their life cycle. On the other hand, the drought events during the late growing period, e.g., at the mature stage in this study, may only have limited effects on plant growth.
RSR is often used to assess plant strategies of biomass allocation in response to drought perturbances [30,38]. Following the optimal partitioning theory [39], plants would allocate more photosynthates to roots than to shoots to sustain water uptake when soil moisture is limited, which leads to a great RSR. This has also been observed in both species in this study (Figure 3 insets), which corroborates with several other studies focusing on the two species [32,40]. Furthermore, the increase of RSR was mediated by the mixture ratio and growth stage (Table 2), with more evident increases in L. davurica during all three growth stages ( Figure 2). Different responses of biomass allocation to drought conditions between the two species may attribute to their distinct root morphology (tap roots vs. fibrous roots) since the root architecture largely controls root plasticity in responses to drought [41]. Xu et al. [42] studied the root morphology of L. davurica when intercropped with B. ischaemum under water deficit conditions, and their results indicated that L davurica had reduced root diameter and increased specific root area and specific root length when soil moisture was limited, which may all contribute to its improved root water uptake. However, a precipitation manipulating study at the inner Mongolian semiarid steppe showed that the effects of drought stress do not necessarily alter plant biomass allocation. The grassland plants can adjust their vertical distribution of roots to adapt to drought [43]. Future studies investigating the root vertical distribution of the two species in the mixtures are needed to fully explore plant responses to drought conditions, preferably in a field setting.
RYT for the species biomass production of the mixtures under different soil watering treatments were similar and were mostly greater than 1.0 (Table 4), indicating a net gain of biomass production by mixing the two species. The increased yield under drought conditions has been widely reported on many crop intercropping systems, e.g., maize and legume (common beans, cowpea, and groundnut) [44], sorghum and groundnut [45], soybean and pigeon pea [46], maize and grass pea [47]. Our previous study with the same setup but soil water increasing (i.e., soil rehydration) on the two species mixtures also showed increased RYT of their mixtures, indicating intercropping of the two could thus improve the RYT regardless of soil water regime [15]. Competition indices are commonly used to explore plant-plant interactions in mixtures [48]. The RCI was used to assess the intraand interspecific competition between the two species in the mixtures. When intraspecific competition is equal to interspecific, the RCI would equal zero, and an RCI greater than zero indicates stronger interspecific competition, and smaller indicates stronger intraspecific competition [29]. In the mixtures, our results showed that dominant intraspecific competition in the grass is contrary to dominant interspecific competition in the subshrub (Figure 4). Similar results were reported in our previous study testing the soil rehydration on the mixtures [15]. Together suggest that B. ischaemum grass was in a more dominant position in the mixtures. The competition balance index (CB) [30] was employed to evaluate the relative competitive ability in the mixture, and a value of zero indicates equal competitive ability between the two species, and a positive value indicates one outperformed another. Results demonstrated that the CB value was mostly positive in the mixtures, implying strong competition of B. ischaemum in the mixtures ( Figure 5), which corroborates the results of RCI. Overall, both competition indices employed here suggest that soil water decreasing altered species competition between the grass and the subshrub, and these effects were mediated by growth stage and mixture ratio together ( Table 3).
The increased WUE under drought conditions has been commonly reported in many dryland plants (see review by Van Duivenbooden et al. [49]). Our results showed a similar increased WUE by soil water decreasing across different growth stages ( Figure 6). This increased WUE under relatively drought conditions may be caused by multilevel biochemical and physiological adjustments of plants to limited soil water conditions and is considered an important mechanism for plants to adapt to dry environments [49,50]. Furthermore, the mixture of the two species, particularly under the B10:L2, had the highest WUE under soil water decreasing treatments ( Figure 6). The overall increased yield and WUE of the mixtures under drought conditions may involve various mechanisms such as niche complementarity, facilitation, and resource sharing [16]. For instance, the tap-rooted subshrub L. davurica and the shallow-fibrous rooted B. ischaemum may maximize the utilization of soil resources at different soil depths via niche complementarity [20].
Here, the mixtures of the two species were grown in pot conditions, which constrained the plant root development, particularly for the tap-rooted L. davurica subshrub. Furthermore, the study was only conducted for one growing season. In other words, it focused on the early development stage ('emergence period') of the mixtures. Future works under field conditions should be conducted over multiple years, especially on mature mixtures, to fully assess plant performance by mixing the two species. Additionally, the interaction between soil moisture and other important drivers, e.g., grazing and fertilization, should also be tested. Lastly, the L. davurica seeds were not priorly inoculated with Rhizobia in this study, and inoculation was not occurring naturally either since we did not detect any root nodulation when washing roots. However, when seed rhizobial inoculation (either artificially or naturally) occurs, the intercropped L. davurica with N fixation potential may reduce soil N competition and/or directly transfer N to the B. ischaemum grass [22,23], which would further improve the overall production of the mixture, and this needs to be explored in field trials.

Conclusions
Our controlled pot experiment showed that soil water decreasing significantly reduced the total biomass production of L. davurica and B. ischaemum mixtures, with a more notable decline during the peak growing periods (e.g., heading and flowering periods) than in the mature stage. On the other hand, the root-to-shoot ratio and overall WUE of the mixtures increased under soil water decreasing treatments, indicating that the two species intercropping could improve the drought resistance of the mixture system, which may be due to niche complementarity, species facilitation, and resource sharing. This increased WUE is of particular importance on the water-limited Loess Plateau. Plant-plant interaction was evident in mixtures, and dominant intraspecific competition was observed in B. ischaemum grass, whereas interspecific competition in the leguminous L. davurica subshrub hints that B. ischaemum was the superior dominant in their mixtures. Additionally, B. ischaemum:L. davurica mixture ratio of 10:2 showed the highest overall biomass and WUE under soil water deficit and without fertilization conditions, this ratio could be considered when sowing these two species to create renovated grasslands on the semi-arid Loess Plateau of China. Besides, the mixing of the legume could also increase grassland quality by N fixation and improving feeding value.