Testing relationship between plant productivity and diversity in a desertified steppe in Northwest China

Study site

This research started in the desertified steppe of northwestern China, Ningxia Hui Autonomous Region of China (106°30′–107°41′E, 37°04′–38°10′N), and was approval by The Grassland Committee of Ningxia Hui Autonomous Region (#15032). The study site had a continental, temperate, monsoon, semiarid climate. The mean annual potential evaporation is approximately 1223.8–2087.6 mm, and the average elevation is 1450 m. Further, the mean annual precipitation is approximately 298 mm over the past 12 years. In July, it has the highest monthly mean temperature (22.4 °C), and in January it has the lowest monthly mean temperature (8.7 °C), respectively. The soil type is mainly dominant by the desertification sierozem with the lower soil fertility, and plants are solitary and show drought-resistant characteristics, including some small xerophytic shrubs and annual weeds. The leaves of plants generally have clear xerophytic morphological characteristics, mainly including members of Asteraceae, Gramineae, Zygophyllaceae, Liliaceae, Cruciferae and Leguminosae (Li, 2001; Pei, Fu & Wan, 2008; Liu et al., 2011).

Sampling methods

Six plant communities were sampled, which have not been grazed for 20 years, i.e., Agropyron mongolicum, Stipa bungeana, Cynanchum komarovii, Glycyrrhiza uralensis, Sophora alopecuroides, Artemisia ordosica, which are dominant plant species in the desertified steppe in northwestern China. AGB and BGB were sampled at their peak times in July and August in 2015. Each plant community consisted of five sampling sites (100 ×100 m), and thirty sampling sites were selected from within the six plant communities (Fig. 1). In order to make this study representative, each sampling site was spaced approximately 1 km apart with the same natural conditions. Data were collected by previous study in Yang et al. (2018). Specially, sampled 15 quadrats (1 × 1 m²) for herbaceous plants and 15 quadrats (3 × 3 m²) for shrubs around with 200 m², and AGB and BGB of all plant species in each quadrat was harvested. Meanwhile, we measured plant height (the maximum, plant species abundance, plant cover, and then dug up the whole root (measured BGB) in each quadrat. The collected AGB and BGB were dried (dry mass) at 85 °C for more than 72 h until a constant weight, and then weighed by electronic scales. In addition, the altitude, latitude, longitude were recorded by a GPS receiver. We investigated the growth characteristics of these plant species and morphological characteristics, which partly reflected the general characteristics of plant communities.

Each soil sample was collected in triplicate from 0 cm to 15 cm of soil profile, and separated into two parts: (1) One part of the fresh soil sample was oven-dried by aluminium container to measure SW (soil water content, %); (2) The other part of the fresh soil sample was used to measure soil physicochemical properties and nutrients by passing through a 2 mm sieve.

Soil sample analysis

Soil pH and EC (soil electrical conductivity, μscm⁻²) were measured in 1:1 (v/v) and 1:5 (v/v) soil water solution, respectively. BD (soil bulk density, gcm⁻³) was measured by oven-dried (for 72 h until the constant weight). TC (soil total carbon, gkg⁻¹) and TN (soil total nitrogen, gkg⁻¹) were measured by an elemental analyzer (elementar vario MACRO cube, Germany). AN (soil available nitrogen, mgkg⁻¹) was measured by NaOH-H₃BO ₃. AP (available phosphorus, mgkg⁻¹) and TP (total phosphorus, gkg⁻¹) were measured by molybdenum-antimony using model 722-Spectrometer. Finally, we used fumigation-extraction method to measure and calculate SMB-C (soil microbial biomass C, mgkg⁻¹) and SMB-N (soil microbial biomass N, mgkg⁻¹) (Liu et al., 2011).

Statistical analysis

In each sample, the following plant diversity indices were calculated (Zhang, Bai & Han, 2004), including Patrick index, Pielou index, Shannon–Wiener index, and Simpson index: (1)${\displaystyle Pa=S}$ (2)${\displaystyle D=1-\sum {\left(Pi\right)}^2}$ (3)${\displaystyle H=-\sum Pi\cdot \mathrm{Ln}Pi}$ (4)${\displaystyle JP=H∕\mathrm{Ln}S=-\sum Pi\cdot \mathrm{Ln}Pi∕\mathrm{Ln}S}$ (5)${\displaystyle Pi=\left(\text{relative abundance+relative plant cover+relative height}\right)∕3}$

Pa, Patrick index; Pi, Plant species dominance; D, Simpson index; H, Shannon–Wiener index; JP, Pielou index (Hurlbert, 1971; Van der Heijden, Bardgett & Van Straalen, 2008).

Fi, Sample species i/The total number of samples; S means the number of species in samples (Bedford, Walbridge & Aldous, 1999; Van der Heijden, Bardgett & Van Straalen, 2008).

All the statistical analyses were conducted by using SPSS 18.0 software (SPSS Inc., Chicago, IL, USA). The significant difference among different plant diversity and productivity were tested using one-way ANOVA (Analysis of variance). Before one-way ANOVA, we tested the normality and homogeneity of variance assumptions of these plant diversity and productivity which were normally distributed, and thus one-way ANOVA followed by Student’s t-test was carried out to test the significant difference in plant diversity and productivity at p < 0.05 level. Further, the dominant driving factors for plant diversity and productivity were performed by CANOCO 5.0 software by RDA (redundancy analysis). Specially, the length of the arrows was determined, and the direction of the arrows for individual driving factors indicated the correlation coefficient among these variables. Plotting was done in Origin 9.2 software. Meanwhile, Pearson correlation, normal regression were used to explore the effect of soil factors on plant diversity and productivity by using R 3.6.0 package (R Core Team, 2019). Finally, GAMs (generalized additive models) were used to obtain the explained variation for plant diversity and productivity, which presented in a Venn diagram.

Occurrence frequency of all plant species and diversity

The occurrence frequency (Fi) values of all plant species were calculated (Table S1). In detail, plant species with the highest frequencies in Stipa bungeana communities mainly included Euphorbia esula (0.47), Artemisia vestita (0.33) and green bristlegrass (0.33); Potentilla bifurca, Artemisia scoparia, Ixeris chinensis, and Salsola ruthenica were auxiliary species with a lower occurrence frequency and therefore had a lower effect on the community structure. Plant species occurring at higher frequencies in Agropyron mongolicum communities mainly included Ixeris chinensis (0.67), Pennisetum flaccidum (1.00), Polygala tenuifolia (0.67), Lespedeza bicolor (0.67), and Heteropappus altaicus (1.00), and the other plants were auxiliary species with a lower occurrence frequency. Plant species with the highest frequencies in G. uralensis communities were Leymus secalinus (0.67), Peganum harmala (0.47), Euphorbia humifusa (0.47), and C. komarovii (0.47). In S. alopecuroides communities, the occurrence frequencies of plant species were relatively similar except for that of Pennisetum flaccidum, and the other plant species were auxiliary species with a lower occurrence frequency. Plant species with the highest frequencies in Artemisia ordosica communities were Agropyron mongolicum (0.67), Agropyron mongolicum (1.00) and Corispermum hyssopifolium (0.67), whereas C. komarovii, green bristlegrass and Ixeris chinensis were auxiliary species with a lower occurrence frequency. The C. komarovii communities had high plant species richness, C. komarovii and Oxytropis racemosa were the dominant species, and species occurring at high frequency were Polygala tenuifolia, Agropyron mongolicum, and Euphorbia esula.

In Fig. 2, C. komarovii and G. uralensis communities had the highest number of plant species, followed by Artemisia ordosica, Stipa bungeana, and Sophora alopecuroides, which all had the same number of plant species. This result indicates that the structural complexity of C. komarovii and G. uralensis communities was higher than that of other plant communities. C. komarovii communities had a higher Patrick index, whereas Stipa bungeana communities had a lower Patrick index, and the Patrick index of C. komarovii communities was nearly two times that of Stipa bungeana communities; there was no significant difference in Patrick index (p > 0.05). Simpson’s index for the C. komarovii and Sophora alopecuroides communities was higher than that for the other plant communities, with no significant difference (p > 0.05). Stipa bungeana communities had a lower Simpson’s index value compared with the other plant communities (p < 0.05), thus indicating that Stipa bungeana communities were relatively homogenous. We found no significant difference in Simpson’s index for the G. uralensis, Agropyron mongolicum, and Artemisia ordosica communities (p > 0.05). Pielou’s index for the C. komarovii and Sophora alopecuroides communities was higher than that for the other plant communities, whereas Pielou’s index for the Artemisia ordosica and Stipa bungeana communities was the lowest for these plant communities (p < 0.05), suggesting that plant species distributions of C. komarovii and Sophora alopecuroides communities were uniform.

Above-ground biomass, Below-ground biomass and R:S ratios

AGB, BGB, R:S ratios of different plant communities showed a higher variation (Table 1). BGB was nearly double than AGB (p < 0.01). The proportion of AGB of Agropyron mongolicum, Stipa bungeana, C. komarovii, G. uralensis, Sophora alopecuroides, Artemisia ordosica were 40.12%, 38.41%, 18.11%, 17.16%, 30.57%, and 31.26%, respectively. AGB ranged from 43.9 to 300.6 gm⁻², with a mean value of 132.2 gm⁻², and BGB ranged from 80.7 to 682.6 gm⁻², with a mean value of 290.7 gm⁻². AGB showed the following order: Artemisia ordosica > Stipa bungeana > Agropyron mongolicum > Sophora alopecuroides = Cynanchum komarovii > G. uralensis with significant differences.

Based on AGB and BGB, we calculated R:S ratios (BGB/AGB) and plotted their frequency distribution. As shown in Fig. 3, R:S ratios had a strong heterogeneity, and ranged between 0.4 and 7.3. Specially, R:S ratios of the Gramineae plants (Stipa bungeana and Agropyron mongolicum) had no significant difference with Leguminosae plants (G. uralensis and Sophora alopecuroides) (p > 0.05). Totally, R:S ratios of most plant communities (more than 40%) ranged from 1 to 2. Besides, we found that there was a lower AGB, BGB and R:S ratios compared with the other grasslands around the world (Table 2).

Relationships among litter, biomass, and the Shannon–Wiener index

To study the contributions of AGB, BGB, and litter to these plant communities and their growth patterns in the desertified steppe, we evaluated the relationships among litter, biomass, and Shannon–Wiener index using regression models (Fig. 4). There was a positive linear relationship between AGB and BGB, indicating that plant communities presented equivalent growth patterns (R² = 0.9025, p < 0.001). Likewise, there was a positive linear relationship between litter and AGB (R² = 0.8485, p < 0.001), BGB (R² = 0.8361, p < 0.001), suggesting that BGB can be accurately estimated by AGB.

Further, we found that the curves between Shannon–Wiener index and AGB were unimodal (R2 = 0.4572, p < 0.05), whereas no significant relationship between Shannon–Wiener index and BGB and litter (p > 0.05). These findings indicated that the degree of resource sharing and the interactions between species of different plant communities changed in the same manner, thus indicating that plant diversity inhibited AGB but did not inhibit BGB. In most grasslands, large amount of studies have confirmed the positive relationship between plant diversity and productivity. In this case, our results also supported this viewpoint. In other words, plant productivity increased with plant diversity because of sufficient natural resources that promoted the reproduction of the plant population at a low level of plant diversity. While at a high level of plant diversity, it was limited by the natural resource, so inter-specific competition intensified, and plant productivity tended to a decrease as a consequence.

Relationships between soil factors and productivity and diversity

Firstly, we found that soil factors had a large effect on plant diversity and productivity by correlation analysis (Table S2). Specially, plant diversity and soil nutrients presented the positive correlations, whereas plant diversity and soil pH, BD presented the negative correlations, and significant positive correlations were observed among AGB, BGB, litter, and TC (p < 0.05).

Secondly, we conducted ordination analysis of plant diversity and productivity; all of plant plots were ordered in NMDS. NMDS showed a higher explanation for plant diversity and productivity on these two axes (first axis/dimension = 52.47, R² = 0.579, p < 0.05; second axis/dimension = 32.09, R² = 0.507, p < 0.05), and plant productivity was strong related to plant diversity (Fig. 5). Further, RDA was used to determine the driving factors affecting plant diversity and productivity (Fig. 6A). In RDA, axes 1 (p < 0.01) and 2 explained 62.19% and 19.58% of the data, respectively, suggesting that soil factors (BD, SW, EC, pH, TC, TN, TP, AP, AN, SMB-N, SMB-C) can adequately account for plant diversity and productivity. In addition, the canonical coefficients of soil factors on each axis indicated that soil factors had an obvious effect on plant diversity and productivity (Table S3). Specifically, there was a similar direction for soil factors arrows and AGB and diversity, which indicated a significant correlation (the longer the arrows, the stronger the correlation) of soil factors and plant diversity and productivity. As for AGB and plant diversity, the first axis was correlated with BD, pH, SW, TC, TN, SMB-C, SMB-N, and the second axis was correlated with BD, pH, SMB-C, SMB-N. Totally, these two axes were negatively related to BD and pH. While soil factors had no significant correlation with BGB, these findings were in agreement with the results from Pearson correlation analysis, indicating that soil factors had a large influence on plant diversity and productivity. Additionally, TC, TN, SMB-C, and SMB-N can be regarded as the key factors driving plant diversity and productivity in desertified steppe.

Finally, we used generalized additive models (GAMs) to explore the relationship among plant productivity (AGB and BGB), plant diversity, and soil factors (Fig. 6B). GAMs showed that soil factors explained more variation in plant diversity and productivity (78.24%). In detail, soil factors had a greater effect on plant productivity (27.01%) than plant diversity (25.41%).

The dynamics of plant community traits in the desertified steppe

Based on the vegetation and soil investigation, we calculated plant evenness, the dominance index, plant productivity (AGB, BGB), and plant diversity. The same soil texture and characteristics existed among these plant communities, and we found that plant Richness index, Evenness index, Simpson index, Shannon–Wiener index had a large difference among these plant communities (Fig. 2), similar to previous studies (Sala & Austin, 2000; Zhang, Bai & Han, 2004). Generally, plant community structure was determined by plant species numbers and their composition (Rejmánek & Richardson, 1996; Zobel, 1997; Bruelheide et al., 2018; Liu et al., 2018). In our study, the Richness index was similar to Simpson’s index. Specifically, the Richness index, Evenness index, Shannon–Wiener index of the Stipa bungeana community were lower than the other plant communities, reflecting the plate-block distribution of the Stipa bungeana community. However, G. uralensis and C. komarovii communities had a large plant species number and higher Shannon–Wiener index compared with the Stipa bungeana community, supporting the fierce inter-specific competition for natural resources.

AGB, BGB, and R:S ratios of different plant communities showed a higher variation (Table 1), and BGB was higher than AGB. According to optimal allocation, plants in drought ecosystems tend to accumulate more BGB for more nutrient absorption to adapt to the fragile environment (extreme drought). Thus, our findings are agreement with previous research from grasslands in arid region around the world (Hedlund, Santa Regina & Van der Putten, 2003; Tilman, Reich & Knops, 2006; Cardinale, Wright & Cadotte, 2007; Chen et al., 2018). However, R:S ratios of most plant communities (more than 40%) ranged between 1 and 2, which were lower than the other grasslands around the world (Table 2). Generally, the lower R:S ratios indicated that these plant communities presented high plasticity for their own survival and reproduction. In order to adapt to the fragile conditions, plant vegetative organs need to more biomass in their growth, which is called to the survival strategy in the desertified steppe (Whittaker & Heegaard, 2003; Cardinale, Wright & Cadotte, 2007). Thus, R:S ratios were lower compared with the other grasslands around the world. Besides, there were some reasons resulting in these discrepancies: (1) There are some special differences in extremely environmental conditions (e.g., climate and soil factors), and also the plant community composition exhibited large differences. For example, the size of plant species in this region was small due to the arid environment (wind erosion, water loss and soil erosion, extreme drought, and so on), leading to the lower AGB and BGB. However, in order to adapt to the extremely environmental conditions, these plant species must enhance the growth of their roots to absorb soil nutrients, thus resulting in the higher BGB compared with AGB. (2) Over the history of our study area, long-term grazing, the fragile environment, the low levels of precipitation, and the interference of human activities have caused the lower AGB and BGB compared with the global level (Tang et al., 2018).

The linkage and driving factors of plant diversity and productivity in the desertified steppe

To explore the contributions of AGB, BGB and litter to plant diversity and productivity, we tested the relationships between AGB, BGB, litter, and plant diversity by using a regression model (Fig. 4). We found a positive linear relationship between AGB and BGB, which presented an equivalent growth pattern (R² = 0.9025, p < 0.001), and this result agrees with most studies from the other grasslands in the world (Hedlund, Santa Regina & Van der Putten, 2003; Ma & Fang, 2006; Cardinale, Wright & Cadotte, 2007; Bai, Wu & Clark, 2012). Further, the positive linear relationships between litter and AGB (R² = 0.8485), BGB (R² = 0.8361) indicated that AGB and BGB were mainly dependent on litters. In recent decades, the dominant view supported the hump-shaped or unimodal pattern between plant diversity and productivity around the world (Fang, Liu & Xu, 1996; Mittelbach, Steiner & Scheiner, 2001; Tilman, Reich & Knops, 2006; Yang et al., 2018). For example, Tilman et al. build a specific model to predict the number of plant species that coexisted on a limited resource base, and showed that plant diversity first increased and then decreased when the supply of natural resources was limited (Liu et al., 2018; Tang et al., 2018). In our study, the curves between plant diversity and AGB were unimodal (R² = 0.4572, p < 0.05), supporting the previous viewpoint (Mittelbach, Steiner & Scheiner, 2001; Tilman, Reich & Knops, 2006; Bai, Wu & Clark, 2012). The hump-shaped form indicated that plant productivity increased at a low level of diversity but decreased at high level of diversity (Bai, Wu & Clark, 2012; Tang et al., 2018). At a low level of plant diversity, plant productivity increased with diversity because the sufficient natural resource promoted the reproduction of the plant population (Mittelbach, Steiner & Scheiner, 2001; Hedlund, Santa Regina & Van der Putten, 2003; Ma & Fang, 2006). At a high level of diversity, it was limited by the natural resource, so inter-specific competition was intensified, and plant productivity tended to decrease as a result (Mittelbach, Steiner & Scheiner, 2001). However, there was no significant relationship between plant diversity and BGB. The reason was that AGB was strong related to plant diversity; when AGB got to the peak, plant diversity decreased because of the fierce competition among plant species, but the fierce competition did not affect BGB since BGB was mainly dependent on soil nutrients below-ground, thus plant diversity had no effect on BGB (p > 0.05), so the findings are agreement with those results reported from Leibold (Leibold, Holyoak & Mouquet, 2004) and Tilman (Tilman, Reich & Knops, 2006).

Additionally, we tested the relationships between plant diversity, productivity and soil factors in the desertified steppe using Pearson correlation coefficients (Table S2). We found that soil factors strongly affected plant diversity and productivity, which were similar to the results from other studies (Liu, Zhao & Zhao, 2012; Bai, Wu & Clark, 2012). Actually, plant diversity and soil nutrients (TC, TN, TP, AN, AP, SMB-C, and SMB-N) presented the positive correlations, whereas plant diversity and soil pH, BD presented the negative correlations, suggesting that soil nutrients positively contributed to plant diversity, whereas soil pH and BD had a negative contribution to plant diversity. In addition, GAMs showed that soil factors explained more variation in plant diversity and productivity (78.24%). In detail, soil factors had a greater effect on plant productivity (27.01%) than plant diversity (25.41%). RDA results demonstrated that soil factors accounted for plant diversity and productivity. Specifically, TC, TN, SMB-C, and SMB-N can be regarded as the key factors driving plant diversity and productivity in desertified steppes in northwestern China.