SRUD: A simple non‐destructive method for accurate quantification of plant diversity dynamics

Predicting changes in plant diversity in response to human activities represents one of the major challenges facing ecologists and land managers striving for sustainable ecosystem management. Classical field studies have emphasized the importance of community primary productivity in regulating changes in plant species richness. However, experimental studies have yielded inconsistent empirical evidence, suggesting that primary productivity is not the sole determinant of plant diversity. Recent work has shown that more accurate predictions of changes in species diversity can be achieved by combining measures of species’ cover and height into an index of space resource utilization (SRU). While the SRU approach provides reliable predictions, it is time‐consuming and requires extensive taxonomic expertise. Ecosystem processes and plant community structure are likely driven primarily by dominant species (mass ratio effect). Within communities, it is likely that dominant and rare species have opposite contributions to overall biodiversity trends. We, therefore, suggest that better species richness predictions can be achieved by utilizing SRU assessments of only the dominant species (SRUD), as compared to SRU or biomass of the entire community. Here, we assess the ability of these measures to predict changes in plant diversity as driven by nutrient addition and herbivore exclusion. First, we tested our hypotheses by carrying out a detailed analysis in an alpine grassland that measured all species within the community. Next, we assessed the broader applicability of our approach by measuring the first three dominant species for five additional experimental grassland sites across a wide geographic and habitat range. We show that SRUD outperforms community biomass, as well as community SRU, in predicting biodiversity dynamics in response to nutrients and herbivores in an alpine grassland. Across our additional sites, SRUD yielded far better predictions of changes in species richness than community biomass, demonstrating the robustness and generalizable nature of this approach. Synthesis. The SRUD approach provides a simple, non‐destructive and more accurate means to monitor and predict the impact of global change drivers and management interventions on plant communities, thereby facilitating efforts to maintain and recover plant diversity.

Many studies have demonstrated that community-level changes in primary productivity following human disturbances regulate changes in grassland plant diversity, with specific disturbances having either positive or negative effects on plant diversity (Bakker, Ritchie, Olff, Milchunas, & Knops, 2006;Hillebrand et al., 2007;Olff & Ritchie, 1998;Proux & Mazumder, 1998;Worm, Lotze, Hillebrand, & Sommer, 2002). Studies to date have found that an increase in primary productivity, for example, in response to nutrient addition, usually reduces plant diversity while a decrease in standing biomass, for example, in response to herbivory, usually maintains more diversity, especially under productive conditions unless fertility is really too low and only a few stress tolerators are found (humpback curve) (Adams, 2009;Fraser et al., 2015;Oba, Vetaas, & Stenseth, 2001;Tredennick et al., 2016). However, recent studies using data from more than 40 grasslands world-wide within the Nutrient Network (NutNet) indicate that herbivores control grassland diversity primarily through their effects on ground-level light (Borer, Seabloom, et al., 2014b) and that plant diversity is reduced even at sites where productivity is not increased by nutrient addition . Moreover, a recent meta-analysis of 189 nutrient addition field experiments shows that loss of plant diversity is correlated with increased productivity, but with substantial variation (Soons et al., 2017). Thus, community-level productivity is not the sole determinant of plant community changes (Harpole et al., 2017).
Combining measures of cover and height for each plant species in the community and deriving a community-level volume-based indicator of plant competition for space and resources, that is, space resource utilization (SRU), may lead to better predictions of plant species richness than productivity (Zhang, Zhou, Li, Guo, & Du, 2015). Briefly, species-level SRU (SRU Si ) is the product of the percent cover and maximum height of each species in a given area ( Figure 1a). SRU represents the competitive ability of a species for space and resources in both the horizontal and vertical dimensions, and as such, may act as a surrogate measure combining the effects of multiple factors. Individual species-level SRU's can be added together to form the community-level SRU (SRU C ). This novel approach is attractive because it provides improved predictions of plant diversity dynamics in response to perturbations compared to productivity while using non-destructive measurements. In comparison, community productivity is estimated destructively by clipping at ground level and then drying and weighing all above-ground biomass of the community.
Biomass identification per species can provide detailed information on species dynamics, but is highly time-consuming and requires extensive taxonomic expertise for the identification of all the species in the community. The SRU approach suffers from similar drawbacks, as the percent cover and maximum height must be measured for each species separately . The ideal method would alpine grassland. Across our additional sites, SRU D yielded far better predictions of changes in species richness than community biomass, demonstrating the robustness and generalizable nature of this approach.
4. Synthesis. The SRU D approach provides a simple, non-destructive and more accurate means to monitor and predict the impact of global change drivers and management interventions on plant communities, thereby facilitating efforts to maintain and recover plant diversity.

K E Y W O R D S
biomass, dominant species, herbivory, maximum plant height, nutrient enrichment, percent cover, plant population and community dynamics, space resource utilization provide accurate predictions of plant diversity dynamics, while using an easy, rapid, non-destructive and broadly applicable tool. The central premise of this study is that the measurement of SRU for only a few dominant species (SRU D ) provides exactly such a tool.
The relative distribution of species abundance within a local community (hereafter 'abundance curve') is characterized by a minority of locally relatively more abundant species (hereafter 'dominant species') and a vast majority of locally relatively less abundant species (hereafter 'rare species') interspersed with species of locally relatively intermediate abundance (hereafter 'intermediate species') (Matthews & Whittaker, 2015;McGill et al., 2007). It follows that the greatest contribution to the cumulative percentage of abundance (measured as biomass or cover) is represented by a few dominant species, while the intermediate and rare species contribute much less ( Figure 1b). Thus, ecosystem processes and community structure and composition should be driven primarily by dominant species (those contributing most to abundance), which has been referred to as the 'mass ratio effect' (Grime, 1998) (Figure 1c). Moreover, dominant and rare species could have opposite contributions to community-level abundance. For example, the increase in community-level abundance in response to nutrient enrichment is usually the result of an increase in the abundance of some dominant species at the cost of a decrease in the abundance of some rare species (Harpole & Tilman, 2007;Zhang et al., 2015). Ultimately, the opposing effects We tested these hypotheses by quantifying the effect of nutrient addition and herbivore exclusion on species-and community-level plant biomass, height, cover, SRU and species richness using data from the Nutrient Network (Borer, Harpole, et al., 2014a). First, we carried out a detailed analysis in an alpine grassland that measured each species in the communities. We classified species into three groups according to their abundance rank: dominant, intermediate and rare, in addition to characterizing the communities based upon all species. We examined the extent to which species abundance impacted predictions of changes in plant diversity with the goal of providing a relatively simple, yet highly robust predictor of plant diversity dynamics. Next, we assessed the broader applicability of our approach by measuring the first three dominant species in each plot of five grassland sites with different habitat types across three continents. The performance of the SRU D approach was compared to that of conventional use of destructive biomass and total community SRU measures.

| Study site
The five study sites are part of the Nutrient Network, a cooperative globally distributed experiment (NutNet; http://nutnet.org/) (Borer, Harpole, et al., 2014a). The sites used in our study included a tallgrass prairie (cbgb.us), a shortgrass prairie (sgs.us) and a shrub steppe (shps.us) in North America, a pasture (frue.ch) in Europe and an alpine grassland (azi.cn) in Asia (Table S1). The sites are F I G U R E 1 (a) Space resource utilization of species i (SRU Si ) is the product of the maximum height (H i ) and percent cover (C i ) of that species in a plot and the plot area (A). (b) Conceptual abundance curve (based on cover or biomass) for all species in the community and highlighting four abundance groups: C = community, D = dominant, I = intermediate and R = rare species. (c) Conceptual diagram illustrating the relationship between changes in SRU or biomass and changes in plant species richness in response to human disturbance for each of the four abundance groups. Relative changes in SRU and species richness (log response ratio or LRR) are calculated as the natural logarithm of the ratio of the variable within a treatment plot to the control plot in the same block [Colour figure can be viewed at wileyonlinelibrary.com] dominated by herbaceous vegetation and referred to as 'grassland' here. Mean species richness in the untreated control plots among these sites varied from 8 to 32 species (cbgb.us: 8; sgs.us: 8; shps. us: 15; frue.ch: 13; azi.cn: 32), and mean richness of the local species pool from 21 to 65 species (cbgb.us: 46; sgs.us: 21; shps.us: 50; frue.ch: 27; azi.cn: 65). Cover, height, species richness and biomass were sampled after 3-5 years of treatment (cbgb.us: 3; sgs. us: 4; hps.us: 4; frue.ch: 3; azi.cn:5).

| Experimental design
Each site consists of a completely randomized block design of nutrient addition and herbivore exclusion with three blocks of ten 5 × 5 m plots per block (Borer, Harpole, et al., 2014a). Nutrient addition treatments consist of a factorial combination of phosphorus (P), nitrogen (N) and potassium (K +μ ; including a one-time addition of micronutrients) for a total of eight nutrient treatment combinations per block. Herbivore exclusion treatments consist of a fencing treatment crossed with the control and NPK treatments for a total of two treatments per block. N, P and K were applied annually, before the beginning of the growing season, using the following application rates and sources: 10 g N m −2 year −1 as time-release urea or ammonium nitrate (NH 4 NO 3 ), 10 g P m −2 year −1 as triple-super phosphate (Ca(H 2 PO 4 ) 2 ) and 10 g K m −2 year −1 as potassium sulphate (K 2 SO 4 ).

| Measurements of plant biomass, height, cover and species richness
Measurements were carried out at the seasonal peak in biomass in a fixed 0.5 × 0.5 m subplot randomly assigned within each plot for azi.cn site and in standard 1 × 1 m subplots for the other sites (cbgb.us, sgs.us, shps.us and frue.ch). For all sites, cover was estimated independently for each species in each plot (Table S2). Note that total summed cover can exceed 100% for multilayer canopies and include two-story vegetation types (e.g. shrublands and forests), where herbaceous species play a minor role. Above-ground live biomass was estimated destructively by clipping at ground level all above-ground biomass of individual plants rooted within two 0.1 m 2 (10 × 100 cm) strips immediately adjacent to the permanent 1 × 1 m plot, followed by drying to constant mass at 60°C and weighing to the nearest 0.01 g. Biomass was sorted to species for azi.cn and to functional group (i.e. grass, forb and legume) for the other sites. We used above-ground live biomass as a measure of primary productivity. Maximum height was estimated for one to five randomly selected individuals per species in each plot as the shortest distance between the upper boundary of a plant (flower stalk or leaf) and the ground level. Maximum height was estimated for each species for azi.cn and for the three most dominant species in each plot for the other sites.

| Calculations for biomass and SRU
To test our hypothesis that predictions of plant diversity dynamics based on changes in biomass or SRU depend on species abundance, we carried out a detailed analysis in an alpine grassland (azi.cn) that measured each species in the communities. SRU for each species (SRU Si ) in each plot was calculated as: where H i is average maximum height and C i the percent cover for species i in a plot and A is the plot area ( Figure 1a). We ranked all the species based on their abundance (biomass or cover) within each plot (Figure 1b) using the 'BiodiversityR' package (Kindt & Coe, 2005) and calculated biomass and SRU per plot using Equations 2 and 3 respectively. Biomass is the sum of the individual species biomass (Biomass Si ) per plot from species i to species j within a plot and is calculated as: and SRU is the sum of the individual species SRU (SRU Si ) per plot from species i to species j within a plot and is calculated as: where i and j are the species' ranks in each plot based on species percent biomass for biomass or percent cover for SRU. Note that indices i and j can take different values depending on the approach used for modelling (see statistical analyses section hereafter and Table 1). Note also that while total summed cover can exceed 100% Cumulative percentage of abundance, species ranks and sets of thresholds explored for each of four abundance groups determined across all plots in azi.cn. Species ranks determined across all plots were used to identify the species belonging to an abundance group within each plot

Abundance groups
Relative cumulative percentage of total abundance (biomass or cover) Note: i and j are the species' ranks in each plot based on species percent biomass for biomass or percent cover for SRU.

Sets of thresholds explored (i, j)
for multilayer canopies, this does not affect the calculation of SRU as this calculation is based on species' rank.
To assess the broader applicability of our approach, we examined data for the three most dominant species in each plot across a diverse range of five grassland sites (including azi.cn) (Table S1). We ranked species based on their cover within each plot and calculated SRU for the first (i = 1, j = 1), the first two (i = 1, j = 2) and the first three (i = 1, j = 3) dominant species in each plot using Equation 3.
We calculated biomass, SRU and species richness responses to treatments (log response ratio or LRR) as the natural logarithm of the ratio of the variable within a treatment plot to the control plot in the same block.

| Statistical analyses
We began by a detailed analysis of an alpine grassland (azi.cn). We modelled the relationships between changes in above-ground biomass or SRU and changes in plant species richness in response to human disturbance with linear mixed effects models using two approaches: the cumulative abundance approach and the abundance groups approach. Sites and blocks nested within sites were treated as random effect in all models.
For the cumulative abundance approach, we examined the im- Next, to assess the generality of our results across five disparate grassland sites, we used the cumulative abundance approach described above to examine the impact of adding plant species from the most dominant to the third most dominant species within each plot on the slope and predictive power of the relationships and compared our results based on SRU D to total community biomass.
We modelled the relationships between changes in above-ground biomass or SRU D and changes in plant species richness in response to human disturbance using linear mixed effects models with block nested within site as a random effect. We calculated conditional R 2 using the 'piecewiseSEM' package (Lefcheck, 2016). We allowed both the intercepts and slopes of regressions to vary between sites if supported by model selection approach based on minimization of BIC (Pinheiro & Bates, 2000).
For each regression, we extracted the slopes with 95% confidence intervals (CI) and extracted the percentage of variation explained by each of the relationships using R 2 values as an indicator of the predictive power for both approaches (higher R 2 values represent better predictive power). In the text, we present estimates of the slopes from the linear regression with their 95% confidence intervals (95% CI). Slopes were considered significant if the intervals did not overlap zero. All analyses were conducted in r 3.4.2 (R Development Core Team, 2014).

| Single study site-alpine grassland (azi.cn)
The abundance curves across all plots show that more than 60% of total abundance was accounted for by a small number of abundant species (hereafter 'dominant species'); while less than 10% of total abundance was represented by the vast majority of much less abundant species (hereafter 'rare species') ( Figure 2, We examined the extent to which changes in biomass or changes in SRU could explain changes in plant species richness in response to nutrient addition or herbivore exclusion (Figure 3). Our cumulative abundance approach revealed that a sizable portion of the total variance was explained by the biomass or SRU of only the single most dominant plant species (R 2 = 0.57 and 0.55, respectively, rank 1-1 in Figure 3a). The percentage of variance explained increased by the inclusion of the biomass or SRU data from other dominant species. This explanatory power soon reached a maximum when including the three dominant species for biomass and the five dominant species for SRU (R 2 = 0.62 and 0.70 respectively; highlighted in red in Figure 3a). Inclusion of additional species led to a decline in explanatory power, reaching rather stable levels of R 2 until all species were included (R 2 = 0.41 and 0.62 for biomass and SRU respectively; Figure 3a). Changes in the biomass or SRU measured for a few F I G U R E 2 Abundance curve for (a) above-ground biomass and (b) cover across all experimental plots of the alpine grassland in Asia (azi.cn). Abbreviations and colours are as in Figure 1b and c [Colour figure can be viewed at wileyonlinelibrary.com] F I G U R E 3 Results of the cumulative abundance approach for the alpine grassland in Asia (azi.cn). Percentage of variance explained (R 2 ) (a) and slopes with 95% CI (b) of the relationship between changes in species richness and changes in biomass or SRU in response to human disturbance, for different sets of increasing species cumulative abundance rank. Log response ratios (LRR) are calculated as in Figure 1c.

Examining the individual factors generating SRU D (Cover D and
Height D ) reveals that while changes in each factor partly and independently contribute to explaining changes in species richness ( Figure S1), their aggregation into SRU D leads to much better predictions of diversity dynamics as compared to each of Biomass D (Figures 3 and 4), Cover D and Height D ( Figure S1).
Because the choice of the thresholds used to classify species ranks into abundance groups is arbitrary to some extent, we explored the effects of selecting a range of different sets of thresholds (Table 1). We found that our results were independent of the thresholds used to classify species into different abundance groups ( Figure   S2; Figure S3). Taken together, these results indicate that the opposite direction of predictions derived from rare species (Figure 4g,h; F I G U R E 4 Results of the abundance groups approach for the alpine grassland in Asia (azi.cn). Relationships of changes in biomass (upper row) and changes in SRU (lower row) with changes in plant species richness in response to human disturbance for different species abundance groups (Table 1)

| Multiple study sites
The abundance curves for cover across all plots within each of the five study sites confirms that rare species are numerous while dominant species are few ( Figure S4) (McGill et al., 2007).
Similar to results of our alpine grassland, changes in plant species richness were negatively associated with changes in community biomass or SRU measured for the dominant species (black lines in Figure 5). Moreover, whether measured for the most (Figure 5b; SRU D1 = −0.09 (−0.14 to −0.05)), the two most  Figure 5).
These results clearly show, across multiple and disparate study sites, that SRU measured for the two most dominant species in each plot gives better predictions of changes in species richness than community biomass (compare Figure 5a,c). Predictions under either nutrient addition ( Figure S5) or herbivore exclusion ( Figure   S6) show similar trends as when analysed together, but with weaker predictive power due to reduction in sample size, especially under herbivore exclusion.

| D ISCUSS I ON
We investigate a new method for measuring plant diversity dynamics aimed at tracking plant community responses to environmental change including human disturbances. Specifically, we assessed how well changes in SRU predict changes in plant species richness and compared this approach to the use of conventional destructive biomass measures. We found that SRU provides stronger predictive power than biomass measurements, while also being non-destructive and easier to perform. We also found that species richness predictions are best achieved by utilizing biomass or SRU assessments of only the dominant species as compared to the entire community.
Based on a single-site level analysis of an alpine grassland, we found that changes in community-level SRU (SRU C ) and biomass (Biomass C ) are strong predictors of changes in plant species richness in response to human disturbances. Disturbances that increased SRU C or Biomass C relative to the undisturbed conditions led to a decrease in plant species richness. Several previous observational and experimental studies have reported such an inverse relationship between community plant productivity and species richness upon human disturbance (Borer, Seabloom, et al., 2014b;Clark et al., 2007;Harpole & Tilman, 2007;Hautier et al., 2009). Here, we extend these findings by showing that human disturbances that increase plant competition for space and resources lead to a reduction in plant species richness.
We hypothesized that changes in SRU in response to human disturbances would better predict changes in species richness than changes in biomass. Our results indeed show that changes in SRU have higher predictive power as compared to changes in biomass, regardless of the species abundance investigated (Figure 3). This result reinforces the earlier finding that SRU C is a better predictor of species richness than biomass  and extends it to predictions of changes in species richness across dominance groups and in response to multiple human disturbances (nutrient enrichment and herbivore exclusion).
We also hypothesized that dominant species would contribute the most to community-level predictions. Our results show that, although the reduction in species richness can be predicted from F I G U R E 5 Relationships of changes in biomass (a) and changes in SRU for the most (b), the two most (c) and the three most (d) dominant species with changes in plant species richness in response to human disturbance across five sites that are part of the international Nutrient Network. The sites include a tallgrass prairie (cbgb.us), a shortgrass prairie (sgs.us) and a shrub steppe (shps.us) in North America, a pasture (frue.ch) in Europe and an alpine grassland (azi.cn) in Asia (Table S1). Black lines are the fixed-effect linear regression slopes among sites from the mixed effects model with block nested within site as a random effect; coloured lines show patterns within sites. Conditional R 2 represents model variation explained by the combination of fixed and random effects. Log response ratios (LRR) are calculated as in Figure 1c [Colour figure can be viewed at wileyonlinelibrary.com] increases in community biomass or SRU, the best predictions were obtained when only a few dominant species were included. Thus, predictions based on changes in the dominant species (SRU D , Biomass D ) were much better than those based on the entire community (SRU C , Biomass C ). Our results support the mass ratio effect (Grime, 1998) and earlier studies reporting that only a few dominant species drive community structure, composition and functioning (Hoover, Knapp, & Smith, 2014;Smith & Knapp, 2003;Winfree, Fox, Williams, Reilly, & Cariveau, 2015). Most interestingly, we found that the best pre- the entire community data, as opposed to only the dominant species . In sum, our study shows that the aggregation of measures of species' cover and height of a few dominant species can provide a powerful, non-destructive and robust tool to assess competitive outcomes in response to human disturbance. For example, the total biomass or SRU of two communities may show little response to human disturbance, but the community with the greater changes in biomass or SRU of a few dominant species will experience a greater change in species richness and evenness.
Our in depth single-site analysis suggests that the best predictions are obtained when including the first three dominant species for biomass and the first five dominant species for SRU. However, it is interesting to note that the predictive power is already higher than the community-based approach when including only the most dominant species for biomass or the two most dominant species for SRU. Our results suggest that the higher predictive power of SRU comes from the aggregation of two key factors with their independent contributions, together acting as a collective wrapper for plant competitive responses in multiple dimensions. Particularly, cover and height represent the competitive ability for space and light in the horizontal dimension and for light in the vertical dimension (Damgaard, 2011). The finding that cover and height are only weakly correlated ( Figure S1) suggests that SRU is a better predictor than biomass because it combines two factors that, to a large extent, represent different species' resource accumulation strategies and independently contribute to asymmetric competition for light and community-level thinning (Hautier et al., 2009;Kaarlejarvi, Eskelinen, & Olofsson, 2017;Suding et al., 2005;Yang et al., 2015). As such, SRU represents a very useful and reliable wrapper of multiple factors related to plant competitive abilities, but future studies will be required to assess the exact suite of factors and underlying mechanisms that make SRU such a good predictor of plant species dynamics.
The SRU D approach is highly simplified; for example, while it precisely measures changes in trait plasticity (height) in response to human disturbance, it ignores the contribution to plant diversity dynamics of intraspecific trait plasticity within a community.
Moreover, while plants exhibit an enormous range of shape and volume (Ingram & Hudson, 1994), our approach simplifies plants' form into a volume representing a cylinder. Despite its simplicity, our model captures the variation in cover and the highly plastic response of height, a plasticity that is not available from trait databases and that needs to be measured in the field (Kattge et al., 2011), into a robust and generalizable predictor of competitive outcomes in response to multiple human disturbances across a wide range of habitats.
Not only does our approach allow for the accurate tracking of management success with respect to promoting species richness, it also stresses how management measures tailored to reducing the SRU of dominant species could represent successful interventions for enhancing biodiversity. Thus, our study suggests that selective harvesting of the dominant species, or introduction of natural enemies (e.g. herbivores, plant or soil pathogens and (hemi)parasites) acting in a density-dependent manner or having a greater effect on the dominant species in a community could promote coexistence and diversity. For example, specialist pathogens or negative biotic soil-effects can promote coexistence by limiting the abundance of the dominant plant species (Allan, van Ruijven, & Crawley, 2010;Creissen, Jorgensen, & Brown, 2016;Heinze, Bergmann, Rillig, & Joshi, 2015). If fast-growing species dominate the community, the introduction of (hemi)parasitic species likely to infect dominant species via abundance-based mechanisms (e.g., due to increased encounter rates) could help grassland restoration (Bardgett et al., 2006;Bullock & Pywell, 2005;DiGiovanni, Wysocki, Burke, Duvall, & Barber, 2017;Pywell et al., 2004). This is because the reduction in competitive dominance of the dominant species by selective harvesting or natural enemies impairs future resource uptake, competitive ability and future abundance of the target dominant species and helps other species, especially rare species, to establish and persist (Allan et al., 2010;Bullock & Pywell, 2005;Hautier, Hector, Vojtech, Purves, & Turnbull, 2010;Heinze et al., 2015). Thus, our results have implications for the development of restoration and management strategies as well as providing an accurate and tractable tool for monitoring subsequent changes in species richness.

ACK N OWLED G EM ENTS
This study was generated using data from five sites of the Nutrient

AUTH O R S' CO NTR I B UTI O N S
P.Z. and Y.H. conceived the study and analysed the data, with input from all authors; P.Z., Y.H., X.Z., X.Z., C.C., G.D., Z.G. and J.F. collected the data used in this study; P.Z. and Y.H. wrote the paper with input from all authors.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available from the Dryad Digital Repository: https ://doi. org/10.5061/dryad.1bm144m .