Ecological Indicators 128 (2021) 107830

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Determining the effects of biotic and abiotic factors on the ecosystem

multifunctionality in a desert-oasis ecotone
Hengfang Wang a, Guanghui Lv a, *, Yan Cai a, Xueni Zhang a, Lamei Jiang a, Xiaodong Yang b
a
College of Resources and Environment Science, Xinjiang University, Urumqi 830046, China
b
Institut Conjoint des Universities de Ningbo et d’Angers, Ningbo University, Ningbo 315211, China

A R T I C L E I N F O A B S T R A C T

Keywords: Questions: Differences in the vertical structures of communities, nutrient cycling, multiple diversity attributes,
Multifunctionality and environmental factors are important forces driving ecosystem multifunctionality. However, the mechanisms
Taxonomic diversity underlying these processes remain unclear.
Phylogenetic diversity
Location: The study took place at the Ebinur Lake Wetland Nature Reserve of the Xinjiang Uygur Autonomous
Functional diversity
Abiotic
Region, China.
Biotic Methods: This study integrated taxonomic diversity, functional diversity, phylogenetic diversity, and environ­
mental factors to evaluate ecosystem multifunctionality and the factors influencing nutrient cycling within 66
dryland communities with different vertical structures.
Results: Both unweighted and weighted diversity had significant impacts on ecosystem multifunctionality and the
cycling of C, N, and P. However, only weighted diversity had a significant impact on the woody and herb layers.
The main factors influencing ecosystem multifunctionality at the community level were soil moisture and
functional diversity, whereas those influencing the woody layer were soil moisture and plant functional traits,
and those influencing the herb layer were phylogenetic diversity and taxonomic diversity. The multifunctionality
of the woody layer and community showed a positive relationship with changes in soil moisture and salinity.
Conclusions: The results of the study showed the existence of both mass ratio effects and richness effects of
ecosystem multifunctionality at the community level, whereas the woody and herb layers were mainly affected
by the complementary effects. Biotic and abiotic factors explained the multifunctionality and nutrient cycling of
the ecosystem at the community level to a greater extent than those in the woody and herb layers separately. In
addition, biotic and abiotic factors explain ecosystem multifunctionality more than nutrient cycling, and
ecosystem multifunctionality was found to explain more than a single nutrient cycle. The multifunctionality of
the ecosystem and the ability to restore specific nutrient cycles can be maximized through the hierarchical
assessment of community diversity to prevent desertification in drylands.

1. Introduction multifunctionality is the ability of an ecosystem to simultaneously

maintain multiple ecological functions or services (Gamfeldt et al.,
Changes to the environment and loss of diversity are major threats to 2008; Hector and Bagchi, 2007; Zavaleta et al., 2010). Indicators related
natural ecosystems (Millennium Ecosystem Assessment, 2005). Global to soil nutrient cycling are generally used to evaluate ecosystem multi­
declines in diversity and altered community composition may have functionality (Berdugo et al., 2017; Maestre et al., 2012a; Le Bagousse-
adverse effects on ecosystem functions (Durán et al., 2018), while Pinguet et al., 2019), primarily because soil-vegetation systems are
environmental changes will also strongly affect ecosystem functions complex and highly interdependent systems that exhibit considerable
(van der Plas et al., 2016). Thus, understanding the relationships be­ feedback. Soils are critical compartments within ecosystems because of
tween species loss and environmental changes, in addition to how these their roles in nutrient and seed storage, and can largely reflect multiple
processes affect ecosystem functioning, is important for predicting ecosystem functions, such as carbon (C), nitrogen (N), and phosphorus
future ecosystem functions (Cardinale et al., 2011). Ecosystem (P) storage, soil and water conservation, and the capacity to support

* Corresponding author.
E-mail address: [email protected] (G. Lv).

https://doi.org/10.1016/j.ecolind.2021.107830
Received 6 March 2021; Received in revised form 2 May 2021; Accepted 16 May 2021
Available online 27 May 2021
1470-160X/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
H. Wang et al. Ecological Indicators 128 (2021) 107830

animal and plant populations. conditions, in addition to how individual nutrient cycles are balanced.
Early studies on biodiversity-ecosystem multifunctionality relation­ The desert-oasis ecotone is the connecting zone between the desert
ships have mostly focused on species richness (Hector & Bagchi, 2007; ecosystem and the oasis ecosystem. Its material and energy exchanges
Maestre et al., 2012a; Soliveres et al., 2014). With further understanding are frequent, and the ecological environment is fragile. It can either
of the attributes of diversity, studies have increasingly begun to evaluate evolve into an oasis or degenerate into a desert, and the difficulty of
multiple attributes of biodiversity, such as functional and phylogenetic evolution is much higher than that of degradation (Tang et al., 2002).
diversity, which can affect the multifunctionality of ecosystems (Cadotte Studying the ecosystem functions of the desert-oasis ecotone plays an
et al., 2012; Craven et al., 2018; Flynn et al., 2011; Naeem et al., 2016; important role in the stability of the desert-oasis ecotone ecosystem, and
Roger et al., 2016). Zirbel et al. (2019) evaluated the restoration of is one of the key contents in the research on biodiversity conservation in
grassland ecosystem multifunctionality and demonstrated that func­ the oasis region (Li et al., 2016).
tional diversity is positively related to multifunctionality. Furthermore, The present study investigated the vegetation and soil characteristics
in a study of 123 arid ecosystems on six continents, Le Bagousse-Pinguet of 66 plots in a desert-oasis ecotone of the Ebinur Lake Wetland Nature
et al. (2019) concluded that phylogenetic diversity and functional di­ Reserve of the Xinjiang Uygur Autonomous Region, China, to obtain an
versity are the key forces driving improvements to the multi­ improved understanding of the impact of varying diversity and envi­
functionality of ecosystems. Indeed, the assessment of how multiple ronmental factors on ecosystem multifunctionality. More specifically,
attributes of biodiversity (e.g., taxonomic, functional, and phylogenetic the effects of various plant diversity attributes on the multi-functionality
diversity) affect multifunctionality can provide a more comprehensive and nutrient cycling of the ecosystem in different vertical structures of
evaluation of ecosystem multifunctionality. In addition, we investigated the community was evaluated. The present study analyzed three soil
whether the ecosystem multifunctionality is affected by the richness indicators related to the multifunctionality of the ecosystem: (1) the C
effect or the mass ratio effect. The mass ratio effect means that the effect cycle, (2) the N cycle, and (3) the P cycle. The present study also pro­
of plant species on ecosystem functions is proportional to the biomass vides a comprehensive discussion of the relationship between biotic and
(mass ratio effect). Relatively insensitive to species richness, it is mainly abiotic factors and ecosystem multifunctionality. We hypothesized that:
affected by individual species or dominant species (Carroll et al., 2011). (1) ecosystem multifunctionality and nutrient cycling of the different
The richness effect refers to the correlation with species richness (Tilman vertical structures are affected by the mass ratio effects or richness ef­
et al., 1997). The richness effect is represented by diversity index based fects; (2) the impact of biotic and abiotic factors on ecosystem multi­
on weighted (Tilman et al., 1997), and the mass ratio effect is repre­ functionality depends on different vertical structures; and (3) different
sented by diversity index based on unweighted (Carroll et al., 2011). vertical community structures (woody layer, herb layer, and community
Exploring the mass ratio effect and richness effect can provide a more level) and different nutrient cycles (C, N, and P cycles) have different
comprehensive understanding of the relationship between biodiversity tradeoff strategies.
and ecosystem multifunctionality.
Dryland ecosystems are characterized by sparse vegetation that is 2. Materials and methods
mainly composed of an herb ground layer and an overstory woody layer,
which are affected by differing structuring mechanisms and ecological 2.1. Study area
processes. There are different relationships between diversity and
ecosystem multifunctionality among the layers, as they might be The study area is located in the National Nature Reserve of Ebinur
assembled through different mechanisms (Zhang et al., 2017a). The Lake Wetland in Xinjiang, China (44◦ 30′ –45◦ 09′ 5N, 82◦ 36′ –83◦ 50′ 3E).
different vertical structures of the community are manifested in the The research area has a typical northern temperate continental arid
appearance and characteristics of plants that are differentially affected climate, with an average annual precipitation and annual evaporation of
by environmental factors. On the one hand, vegetation with different 107 mm and > 1,600 mm, respectively. The area also falls within the
vertical structures show different responses to environmental factors path of the Alashankou winds, with an average of 164 windy days each
(Meng et al., 2015). A study by Qin et al. (2020) on the plant diversity of year, with wind speeds exceeding 17 m s− 1 (Wang et al., 2019). The
above-ground forest communities found that environmental filtering diverse soil types in the area support enriched plant communities,
plays a leading role in the diversity of woody plants, while that of herb especially xerophytic desert species. The primary plant species include
plants is subject to the combined effects of environmental filtering and Populus euphratica, Haloxylon ammodendron, Halimodendron haloden­
diffusion restrictions. However, the different vertical structures of dron, Halostachys belangeriana, Kalidium foliatum, Apocynum venetum,
communities have different effects on ecosystem functions. Studies such and Aeluropus pilosus.
as that by Luo et al. (2019) have shown that species richness and
phylogenetic diversity are the most important predictors of herb 2.2. Community survey
biomass, whereas functional traits are the main predictors of woody
plant biomass. Sampling was conducted from July to August 2017 over a large area
Ecosystem multifunctionality and nutrient cycling, as well as the extending from the Aqikesu River to the Mutter Desert, with a width and
tradeoffs between woody plants, herb plants, and the overall community length of 1,200 m and 3,300 m, respectively. The overall sample area
ecosystem occur simultaneously; thus, an understanding of how these was divided into 66 grid cells, each with a size of 200 × 300 m. A 30 m ×
functions promote each other or otherwise develop trade-offs among 30 m plot was established in each grid center for vegetation surveys and
each other is important (Pasari et al., 2013; Zirbel et al., 2019). The for the collection of soil and plant samples. The species names,
global dryland analysis by Maestre et al. (2012b) found that ecosystems maximum heights of the plants, and plant abundances were recorded
in colder and sandier regions have higher multifunctionality. They also during the survey.
found that multiple functions related to P cycling decreased as average
dryland temperatures increased, whereas species richness was critical 2.3. Measurement of plant and soil functional traits
for maintaining ecosystem functions related to C and N cycles. Drivers of
ecosystem multifunctionality may comprise the simultaneous operation Soil heterogeneity from the roots to the crown edge was reduced by
of all functions, or the operation of only a few functions that strongly randomly selecting 5–10 sampling points using a 10 cm × 10 cm square
influence ecosystem multifunctionality or environmental conditions box under each crown of species in the 66 sample plots. Three sampling
(Byrnes et al., 2014). In addition to the above considerations, it is also points were also collected from bare areas. All soil samples were
important to understand how ecosystem multifunctionality at different collected from depths of 0–10 cm.
community levels is associated with biodiversity and environmental Several functional traits were measured, including maximum height

2
H. Wang et al. Ecological Indicators 128 (2021) 107830

(H, m), leaf length (LL, mm), leaf thickness (LT, mm), specific leaf area averaged according to the ratio of the bare land area to the area covered
(SLA, m2/kg), leaf carbon concentration (LC, g/kg), leaf nitrogen con­ by vegetation, and then transformed and standardized (Maestre et al,
centration (LN, g/kg), and leaf phosphorus concentration (LP, g/kg). 2012b). The formula for calculating the multifunctionality index MFa is:
These physiological characteristics affect the growth, survival, and
reproduction of individual plants, as well as the function of the com­ 1 ∑F
1∑ 9
MF a = g(ri (fi ) ) = g(ri (fi ) )
munity. Plant traits were measured during the period of plant growth F i=1 9 i=1
(July). The current study used nine soil function indicators related to
ecosystem processes. C, N, and P cycling dynamics were selected to where F represents the number of measured functions, fi represents the
comprehensively evaluate ecosystem multifunctionality (Maestre et al., measured value of function i, ri is a mathematical function that converts
2012a). Among them, the selected indicators associated with C cycling fi to a positive value, and g represents normalization of measured values.
were organic C and β-glucosidase activity, whereas those associated We useed MinMax’s method for standardization.
with N cycling were total N, nitrate N, ammonium N, and available N, Ecosystem multifunctionality and the C, N, and P cycles were
and those associated with P cycling were total P, phosphatase activity, calculated at the community level, woody layer, and herb layer. The C,
and available P. In addition, the soil moisture (SM) content, soil salinity N, and P cycles were calculated separately, and the method was
content (SSC), and soil pH of each plot of bare land were determined. consistent with the method of ecosystem multifunctionality. The cor­
Appendix S1 outlines the methods used to determine plant and soil relations between ecosystem multifunctionality and the C, N, and P
traits. cycles at the community level, woody layer, and herb layer were then
calculated (see Appendix S2).
2.4. Multivariate diversity metrics
2.6. Data analysis
The current study calculated several biodiversity indices, including
taxonomic diversity, phylogenetic diversity, and functional diversity. The factors influencing ecosystem multifunctionality among the 66
Taxonomic diversity is the most basic, intuitive and commonly used plots could be divided into abiotic factors, such as soil moisture, soil salt
metric for biodiversity, which is mainly determined by the number of content, and pH, and biotic factors, such as multi-faceted biodiversity
species. The indices indicating taxonomic diversity we used included metrics of different vertical community structures, such as the woody
species richness (SR), Shannon index (Shannon, 2001), and Simpson layer, herb layer, and community level. The present study clarified the
index (Simpson, 1949). relationship between taxonomic, phylogenetic, and functional diversity,
Phylogeny diversity is a measure related to the evolution of species, as well as ecosystem multifunctionality, in dryland ecosystems through
and phylogeny requires understanding of the gene structure of the selection of three species diversity indices (the SR, Shannon-Wiener,
commensal species in the community. The indices of phylogenetic di­ and Simpson indices), four functional diversity indices (Fdis, Fdiv, Fric,
versity we used included the mean nearest taxon distance (MNTD) and RaoQ), three phylogenetic diversity indices (PD, MNTD, and MPD),
(Webb et al., 2002), mean pairwise distances (MPD) (Webb et al., 2002), and CWM of functional traits.
and Faith’s PD (PD) (Faith, 1992). Functional diversity is a method to One variable was eliminated when a pair of candidate variables with
measure the variation of species traits in a community, which is mainly a correlation coefficient exceeding 0.6 was observed (Appendix S3-
characterized by various characteristic values measured by species (such Appendix S5) to avoid high autocorrelations between variables. Linear
as related traits of leaves, stems or roots). mixed models were then used to investigate the impact of each diversity
The calculation process of community phylogenetic diversity is metric on ecosystem multifunctionality. The standardized coefficients
based on the following steps: sort out the Latin names of all species, (slope: b) and determination coefficients (R2) were then calculated.
input them into APGIII to search for family and genus names, then use Random forest modeling analysis was first used to determine which
Phylomatic software to generate a phylogenetic tree, and then import variables had the greatest explanatory power and were the best pre­
the phylogenetic tree into Phylocom software (Webb et al., 2008), dictors of variation in ecosystem multifunctionality.
through the bladj module Estimate the differentiation time of the species The Boruta package of the random forest algorithm within the R
on the phylogenetic tree. Subsequent calculations are calculated by language was used to select the importance of the biotic and abiotic
importing the generated phylogenetic tree through the R (3.6.2) factors that affect ecosystem multifunctionality and the C, N, and P
language. cycles in different community vertical structures. These factors were
The functional diversity we used included the functional richness then ranked, and the optimal factors were identified based on the
index (Fric) (Villéger et al., 2008), the functional divergence index importance of the predictor variables (see Appendix S6-Appendix S8).
(Fdiv) (Villéger et al., 2008), the functional separation (Fdis), the RaoQ The Boruta package was preferred in the current analysis because of its
quadratic entropy index (Ricotta, 2005), and the community weighted computational efficiency and lower variable selection error rate (Speiser
mean (CWM) of functional traits (Garnier et al., 2004). In the current et al., 2019). The Boruta algorithm is based on a permutation test using a
study, metric indicators of functional diversity and phylogenetic di­ holdout approach for importance measures. This algorithm assesses
versity were weighted (e.g., MNTD_W, Fdiv_W) and non-weighted (e.g., whether the score of each biotic or abiotic factor exceeds that of a
MNTD, Fdiv) by species abundance to distinguish between the mass random feature during the iterative process, with the factor confirmed
ratio effects and richness effects of plant species on ecosystem multi­ or rejected if the score of the factor exceeds or is below that of a random
functionality. In addition, these diversity indicators were calculated at feature, respectively (Kursa & Rudnicki, 2010). The linear mixed effects
the community level, woody layer, and herb layer. (lme) model was used to calculate the effects of all variables on
ecosystem multifunctionality, during which all variables were first input
2.5. Measurement of ecosystem multifunctionality into the model to test predictors and their interactions. All variables
were scaled and normalized during the calculation (a mean of 0 and
We use the average method to calculate the ecosystem multi­ standard deviation of 1).
functionality. It is relatively simple and intuitive to use the average The “MuMIn” package was then used to calculate a complete set of
method to calculate the multifunctionality. First, the ecosystem function models based on maximum likelihood (ML) estimations, and the models
values in the plot are transformed, then the data is standardized, and were ranked using the AICc (Akaike Information Standard) score (Luo
finally the average value of each function value is calculated as the et al., 2019), with the model with the smallest AIC difference selected as
multifunctionality index. In our sample plot, the vegetation coverage the final model. For each response variable, a marginal R2m (variance
rate is low, and the measured function values need to be weighted and explained by fixed factors only) was calculated using the selected model

3
H. Wang et al. Ecological Indicators 128 (2021) 107830

(Nakagawa & Schielzeth, 2013). Linear mixing models were calculated ratio and richness effects, whereas the ecosystem multifunctionality
using the nlme package. Finally, triangle diagrams were used to inves­ values of the woody and herb layers were mainly influenced by the
tigate the balance of C, N, and P cycle functioning, in addition to the richness effects.
tradeoffs of community woody and herb plants. Hierarchical segmen­
tation was used to determine the contribution of the redundancy anal­ 3.2. Effects of biotic and abiotic factors on ecosystem multifunctionality
ysis (RDA) single explanatory variable. Hierarchical partitioning was and the C, N and P cycles
then used to decompose the explanatory rate of each explanatory vari­
able. Rates were quantified using the “rdaenvpart” package. All of the Random forest analysis showed that biotic and abiotic factors had an
above packages were used in the R computing environment (ver. 3.5.1 important impact on ecosystem multifunctionality and the community-
Development Core Team, https://www.r-project.org/). level C, N, and P cycles (Appendix S6-Appendix S8). Abiotic factors and
functional and phylogenetic diversity played important roles in
3. Results ecosystem multifunctionality and the C, N, and P cycles in the woody
layer. The results suggested that soil moisture, salinity, and community-
3.1. The bivariate relationships between diversity attributes and ecosystem weighted means of functional traits and SR were important for
multifunctionality ecosystem multifunctionality and the C, N, and P cycles of the herb
layer.
As shown in Fig. 1, multiple diversity indicators were closely related Based on the importance of random forest screening, the current
to ecosystem multifunctionality and nutrient cycling. The unweighted study used a mixed linear model (Table 1) to analyze and demonstrate
Fdis had a significant impact on ecosystem multifunctionality and the C, that ecosystem multifunctionality and the community-level C, N, and P
N, and P cycles. The unweighted taxonomic diversity index had a sig­ cycles were mainly affected by SM and Fdis. The ecosystem multi­
nificant impact on the N and P cycles. The degree of functional diver­ functionality of the woody layer was mainly affected by SM, Fdiv, the
gence (Fdiv_W) weighted by abundance also had a significant impact on community-weighted mean of the crown area, and leaf phosphorus
ecosystem multifunctionality and the C, N, and P cycles. Weighted content. Ecosystem multifunctionality and the C, N, and P cycles of the
phylogenetic diversity (MPD_W) had a significant impact on ecosystem herb layer were mainly affected by SR and MPD. In addition, the
multifunctionality and the C and N cycles. Therefore, mass ratio effects interpretation of the community-level R2m within the model was higher
and richness effects occurred at the community-level multifunctionality than that of the woody and herb layers.
and within the C, N, and P cycles.
Ecosystem multifunctionality and the C, N, and P cycles of the woody 3.3. Tradeoffs between ecosystem multifunctionality
layer (Fig. 2) were mainly affected by the community-weighted average
of plant functional traits (such as CWM.LN and CWM.LP), whereas Fdis 3.3.1. Tradeoffs between the C, N, and P cycles among different vertical
and average phylogeny distance (MPD) also had a significant impact. structures
Ecosystem multifunctionality and the C, N, and P cycles of the woody The contributions of the C and N cycles to the multifunctionality of
layer were less affected by weighted plant diversity; that is, the richness the community were stronger than those of the P cycle (Fig. 4A). The C,
effects had a greater impact on the woody layer. Fdiv had a significant N, and P cycles of the woody layer were more balanced (Fig. 4B),
negative effect on ecosystem multifunctionality and C, N, and P cycles. whereas the contribution of the N cycle to the function was higher
The biodiversity indicators that had a significant impact on within the herb layer (Fig. 4C). Although different factors affected the
ecosystem multifunctionality and the C, N, and P cycles of the herb layer community and woody and herb layers (Fig. 4D–F), as shown in Fig. 1,
included species richness (SR), the phylogenetic diversity index (MPD), the C, N, and P cycles of each group were relatively balanced.
and the community-weighted mean of plant leaf phosphorus content and As shown in Fig. 4 A, B, and C, although the C, N, and P cycles
leaf thickness (Fig. 3). Among them, the community-weighted mean of maintained a relatively stable tradeoff relationship, they were affected
leaf thickness had a significant negative effect on ecosystem multi­ by different biotic and abiotic factors. Among them, the most important
functionality and the C, N, and P cycles of the herb layer. factor affecting the community-level C and N cycles was Fdis, followed
Therefore, the ecosystem multifunctionality and the community- by SM, and the most important factor affecting the P cycle was SSC,
level C, N, and P cycles were mainly affected by the combined mass followed by SM (Fig. 4D). The most important factor affecting the P

Fig. 1. The relative effects of diversity metrics on

ecosystem multifunctionality and functions related to
carbon (C), nitrogen (N), and phosphorus (P) cycling
within 66 plots in the Ebinur Lake Wetland Nature
Reserve of the Xinjiang Uygur Autonomous Region,
China. Effect sizes were standardized coefficients of
linear mixed models individually evaluated for each
predictor variable. Dark green symbols represent
functional diversity metrics, light green represents the
community-weighted means of functional diversity
metrics, red represents taxonomic diversity metrics,
and blue represents phylogenetic diversity metrics.
The lines indicate 95% confidence intervals. Signifi­
cances are * P < 0.05 and ** P < 0.01. MFa represents
ecosystem multifunctionality at the community level;
C_MFa represents the C cycle at the community level;
N_MFa represents the N cycle at the community level;
P_MFa represents the P cycle at the community level.
(For interpretation of the references to color in this
figure legend, the reader is referred to the web version
of this article.)

4
H. Wang et al. Ecological Indicators 128 (2021) 107830

Fig. 2. The relative effect of diversity metrics on

ecosystem multifunctionality and functions related to
C, N, and P cycling for woody plants within 66 plots in
the Ebinur Lake Wetland Nature Reserve of the Xin­
jiang Uygur Autonomous Region, China. Effect sizes
were standardized coefficients from linear mixed
models that were individually estimated for each
predictor variable. Dark green symbols represent
functional diversity metrics, light green symbols
represent the community-weighted means of func­
tional diversity metrics, red symbols represent taxo­
nomic diversity metrics, and the blue symbols
represent phylogenetic diversity metrics. The lines
indicate 95% confidence intervals. Significances are *
P < 0.05 and ** P < 0.01. Woody-MFa represents
ecosystem multifunctionality in the woody layer;
Woody-C_MFa represents the C cycle in the woody
layer; Woody-N_MFa represents the N cycle in the
woody layer; Woody-P_MFa represents the P cycle in
the woody layer. (For interpretation of the references
to color in this figure legend, the reader is referred to
the web version of this article.)

Fig. 3. The relative effect of diversity metrics on

ecosystem multifunctionality and functions related to
C, N, and P cycling for herb plants within 66 plots in
the Ebinur Lake Wetland Nature Reserve of the Xin­
jiang Uygur Autonomous Region, China. Effect sizes
were standardized coefficients from linear mixed
models that were individually estimated for each
predictor variable. Dark green symbols represent
functional diversity metrics, light green symbols
represent community-weighted means of functional
diversity metrics, red symbols represent taxonomic
diversity metrics, and blue symbols represents phylo­
genetic diversity metrics. The lines indicate 95%
confidence intervals. Significances are * P < 0.05 and
** P < 0.01. Herb-MFa represents ecosystem multi­
functionality in the herb layer; Herb-C_MFa represents
the C cycle in the herb layer; Herb-N_MFa represents
the N cycle in the herb layer; Herb-P_MFa represents
the P cycle in the herb layer. (For interpretation of the
references to color in this figure legend, the reader is
referred to the web version of this article.)

cycle was SSC, followed by SM (Fig. 4D). Similarly, the main factors woody plants, with variable contributions from the remaining factors. In
affecting the balance of the C, N, and P cycles of the woody layer were other words, the factors influencing ecosystem multifunctionality and
SM and the community-weighted means of leaf phosphorus (Fig. 4E). the C, N, and P cycles differed among the community levels and the
The main factors affecting the herb layer were SSC and MPD (Fig. 4F). woody and herb layers.

3.3.2. Tradeoffs in multifunctionality at the community, woody layer, and 3.3.3. Relationship between the community and herb and woody layers
herb layer with soil moisture and soil salinity
Unlike the tradeoff relationships between C, N, and P shown in Fig. 4, Fig. 7 shows the correlation analysis between abiotic factors and
there were uneven tradeoffs between the community and woody and ecosystem multifunctionality and the C, N, and P cycles. Community-
herb layers (Fig. 5 A–D). Among these trade-offs, those among level and woody layer ecosystem multifunctionality and the C, N, and
ecosystem multifunctionality and the C, N, and P cycles showed that the P cycles showed the same trend with increasing soil moisture and
contribution at the community level was significantly higher than that at salinity, and community-level and woody layer multifunctionality were
the woody and herb layers, with the first two centers of gravity in the C significantly correlated with SM and salinity (Appendix S9).
and N cycles (Fig. 5 C, D), with one biased towards the community level Ecosystem multifunctionality and the C, N, and P cycles of the herb
and the other biased towards the herb layer. Therefore, the herb layer layer showed a downward trend with increasing SM and soil salt content
also played an important role in the C and N cycles at certain times. (Fig. 7, Appendix S9), whereas the woody layer exhibited the opposite
As shown in Fig. 6, the factors influencing ecosystem multi­ trend. The herb layer was more multifunctional than the woody layer
functionality at the community, woody layer, and herb layer levels were under SM < 5% and soil salt content > 3.5%. In other words, the herb
quite different; SM was a common factor influencing the community and layer plays a leading role under these abiotic conditions; otherwise, the

5
H. Wang et al. Ecological Indicators 128 (2021) 107830

Table 1 the results provide a basis for the mass ratio effects and identification of
Model selection results for biotic and abiotic sources of variation in ecosystem richness effects of ecosystems across different vertical structures. Abiotic
multifunctionality and sets of functions related to C, N, and P cycling for 66 plots factors were further combined, the relative importance of biotic and
in the Ebinur Lake Wetland Nature Reserve of the Xinjiang Uygur Autonomous abiotic factors to ecosystem multifunctionality was discussed, and the
Region, China. tradeoffs between different vertical structures and different cycles were
Type Group Predictor logLik AICc Wt R2m further analyzed. These analyses provide a basis for studying ecosystem
variables multifunctionality in different vertical structures of desert ecosystems.
Community MFa SM + SSC + Fdis 51.75 − 91.50 0.03 0.521 The results of the present study also provide a reference for the mech­
C_MFa SM + Fdis 39.44 − 68.88 0.04 0.482 anisms that influence ecosystem multifunctionality.
N_MFa SM + Fdis + SSC 44.01 − 76.03 0.04 0.467
P_MFa SM + Fdis + SSC 42.25 − 72.50 0.03 0.382
Woody MFa SM + CWM.S + 31.75 − 49.50 0.05 0.463
Fdiv + CWM.LP 4.1. The mass ratio effects and identification of the richness effects of
C_MFa SM + CWM.S + 32.15 − 52.30 0.04 0.413 ecosystem multifunctionality
CWM.LP
N_MFa SM + CWM.LP + 31.72 − 51.45 0.05 0.455
An understanding of how the biodiversity attributes of different
Fdiv
P_MFa CWM.S + CWM. 23.33 − 34.66 0.05 0.390 vertical structures affect ecosystem multifunctionality and nutrient
LP + SM cycling in natural communities is essential for restoring and managing
Herb MFa SR + MPD + SSC 38.74 − 48.07 0.09 0.334 desert ecosystems (Balvanera et al., 2014; Brum et al., 2017). At the
C_MFa SR + MPD + SSC 34.27 − 51.56 0.11 0.335 same time, we can gain an understanding of the impact of diversity on
N_MFa MPD + SR 39.59 − 71.18 0.13 0.226
P_MFa MPD + SR 38.90 − 68.54 0.04 0.229
ecosystem multifunctionality by considering the mass ratio effects and
richness effects of different vertical structures on ecosystem
Note: The modeling was conducted for three different groups of the desert multifunctionality.
ecosystem at the community, woody plant, and herb plant levels. Only models
The results of the present study showed that the weighted functional
exhibiting significant differences are presented. Marginal R2m was calculated as
diversity, phylogenetic diversity, unweighted functional diversity, and
the variance explained only by fixed effects. MFa represents ecosystem multi­
functionality, C_MFa represents the C cycle, N_MFa represents the N cycle, and abundance indices all had significant effects on ecosystem multi­
P_MFa represents the P cycle. functionality and community-level C, N, and P cycles. Therefore, both
mass ratio effects and richness effects exist in ecosystem multi­
functionality and in community-level C, N, and P cycles (Cadotte, 2017;
woody layer plays a leading role.
Luo et al., 2019). The woody layer was mainly affected by functional
diversity and phylogenetic diversity based on non-abundance weight­
4. Discussion
ing. In other words, the richness effects of these two factors played a
major role in the woody layer, indicating that symbiotic species
The present study evaluated the factors influencing ecosystem mul­
collaborate within the overall use of resources, with niche comple­
tifunctionality and the C, N, and P cycles across different vertical
mentarity playing the primary role (Naeem et al., 1994). The herb layer
structures within a dryland ecosystem. The effects of biodiversity in­
was also mainly affected by the MPD between the richness and phylo­
dicators and ecosystem functions, both weighted and non-abundance
genetic diversity indices. Therefore, the richness effects constituted the
weighted, on ecosystem multifunctionality were first analyzed, and
main adaptive mechanism of the herb layer, and the ecosystem

Fig. 4. Ternary diagrams of community-level C, N, and P cycling for the populations of the herb and woody layers (A, B, and C) along with an explanation of the
contribution of biotic and abiotic factors to ecosystem multifunctionality (D, E, and F) for 66 plots in the Ebinur Lake Wetland Nature Reserve of the Xinjiang Uygur
Autonomous Region, China. The shading of color represents the distance from the center, and the values in parentheses show the average for panels A, B, and C. The
histograms in D, E, and F show the contribution of each factor from lowest to highest. MFa represents ecosystem multifunctionality; C_MFa represents the C cycle;
N_MFa represents the N cycle; P_MFa represents the P cycle.

6
H. Wang et al. Ecological Indicators 128 (2021) 107830

Fig. 5. Ternary diagram showing ecosystem multifunctionality in addition to the contributions of the C, N, and P cycles at the community, herb layer, and woody
layer levels (A, B, C, and D) for 66 plots in the Ebinur Lake Wetland Nature Reserve of the Xinjiang Uygur Autonomous Region, China. The values in parentheses show
the mean values for that axis. The color shading represents the distance from the center. MFa represents ecosystem multifunctionality; C_MFa represents the C cycle;
N_MFa represents the N cycle; P_MFa represents the P cycle.

multifunctionality of the woody and herb layers was not mainly driven average phylogenetic distance between individuals, excluding pairs of
by dominant species (Tilman et al., 1996). The relationships between identical species (Webb et al., 2002). The significant correlation be­
biodiversity and ecosystem multifunctionality at the woody layer, herb tween MPD_W and ecosystem multifunctionality in the present study
layer, and community levels showed that the impact of dryland biodi­ also showed that weighted phylogenetic diversity had a significant effect
versity on community-level multifunctionality was not merely a simple on ecosystem multifunctionality. Phylogenetic diversity may effectively
superposition of the woody and herb layers. This result may be attrib­ reflect hidden traits related to the ecosystem multifunctionality of dry­
uted to the complexity of the underlying process, which may be related lands, including pathogen protection, pollination rate, mycorrhizal as­
to processes other than those related to the effect of the woody and herb sociation (Gilbert and Webb, 2007), and plant population strategy
layers, which requires further exploration in future research. (Salguero-Gómez et al., 2015). Phylogenic diversity is also an important
force driving ecosystem multifunctionality. MPD_W, based on weighted
4.2. Impact of biotic and abiotic factors on ecosystem multifunctionality phylogenetic diversity, had a significant impact on ecosystem
multifunctionality.
The results showed that both biotic and abiotic factors are important The degree of functional divergence (Fdiv_W) based on abundance
drivers of ecosystem multifunctionality (Table 1 and Fig. 6). The main weighting was negatively correlated with the ecosystem multi­
biotic factors affecting community-level ecosystem multifunctionality functionality of the woody and herb layers. The negative effects of
were unweighted functional dispersion, weighted functional divergence, Fdiv_W observed in the current study were consistent with the previ­
and weighted average phylogenetic distance. The unweighted func­ ously observed negative effects of trait dispersion on dryland ecosystem
tional diversity index (Fdis) refers to the degree of functional separation, multifunctionality (Gross et al., 2017) and on the production and sta­
representing the average distance from each community to the center bility of biomass in several ecosystems (Craven et al., 2018; Flynn et al.,
point in the representative shape space (Anderson, 2006; Laliberté and 2011; García-Palacios et al., 2018). At the same time, the community-
Legendre, 2010). Functional dispersion (Fdiv_W) represents the distance weighted means of leaf phosphorus content and crown area had a sig­
between the most abundant taxa and the community center in the trait nificant impact on the woody layer (Table 1), whereas SR and MPD of
space (Villéger et al., 2008). Therefore, ecosystem multifunctionality is the phylogenetic diversity index had significant effects on the herb layer.
strongly correlated with the distance from the trait to the community It is possible that different symbiotic species behaved differently in most
center, that is, the occupation of the trait space. MPD_W represents the spaces. These results support the hypothesis that niche complementarity

7
H. Wang et al. Ecological Indicators 128 (2021) 107830

Fig. 6. The explanation of contribution of biotic and

abiotic factors on ecosystem multifunctionality and C,
N, and P cycling at the community, woody layer, and
herb layer levels (A, B, C, and D) for 66 plots in the
Ebinur Lake Wetland Nature Reserve of the Xinjiang
Uygur Autonomous Region, China. The histogram
shows the contribution of each factor from lowest to
highest. MFa represents ecosystem multifunctionality;
C_MFa represents the C cycle; N_MFa represents the N
cycle; P_MFa represents the P cycle.

Fig. 7. Relationships between soil moisture (SM) content and ecosystem multifunctionality for 66 plots in the Ebinur Lake Wetland Nature Reserve of the Xinjiang
Uygur Autonomous Region, China. The grey shaded area represents the 95% confidence intervals, and the lines of the same color indicate ecosystem multi­
functionality, C cycling, N cycling, and P cycling. The unit of SM is percentage (%), whereas the unit of soil salt content (SSC) is g kg− 1.

explains the multifunctionality of ecosystems (Cadotte, 2017). The the perspective of a single community to the perspective of different
contrasting patterns between ecosystem multifunctionality and biodi­ vertical structures of the community.
versity across different vertical structures of a community also provides Soil moisture and salinity are the main abiotic factors explaining
a new perspective on the importance of species evolution to the multi­ ecosystem multifunctionality and the community-level C, N, and P cy­
functionality of ecosystems. The contrasting operational modes of cles and the woody layer (Table 1, Figs. 4, 6, and 7). These abiotic factors
phylogenetic diversity and ecosystem multifunctionality across different are important in drylands because of low annual precipitation, low
vertical structures also provide a new perspective on the multi­ water availability, and serious soil desertification and salinization (Gong
functionality of ecosystems. Therefore, a shift from a single taxonomic et al., 2019; Wang et al., 1999). Community-level ecosystem multi­
perspective to a perspective that includes multiple attributes of biodi­ functionality showed the same trend as that of the woody layer with
versity is required to better understand the impact of ecosystem multi­ respect to changes in SM and salinity (Fig. 7). In contrast, the herb layer
functionality (Le Bagousse-Pinguet et al., 2019), along with a shift from played an important role in the natural ecosystem under high SM and

8
H. Wang et al. Ecological Indicators 128 (2021) 107830

salt content, which is consistent with the vegetation observed during the traits, and the herb layer was mainly affected by phylogenetic diversity
investigation. The dominant species along the riverbank to the edge of and species diversity. There were also differences in ecosystem multi­
the desert changed from woody plants to herb plants, and plant abun­ functionality among the different vertical structures and within the
dance decreased with decreasing soil moisture. These observations can balance of the C, N, and P cycles, illustrating that the community-level
provide guidance for the restoration of desert ecosystems and demon­ ecosystem multifunctionality exceeded those of the woody and herb
strate that different vertical structures of communities may have layers. The present study provides a new perspective on the study of
different adaptive mechanisms. Therefore, the relationships between ecosystem multifunctionality and can provide a theoretical basis for the
biotic and abiotic factors and different levels of multifunctionality may protection of biodiversity in the desert ecosystems of Xinjiang. The re­
also differ across different vegetation structures (Zhang et al., 2017b), sults of the study are also beneficial for the protection, restoration, and
and the relationship between different aspects of ecosystem functions management of ecosystem diversity according to species attributes. The
may also differ within the same vertical structure. In addition, biotic and current study also has important practical significance for the restora­
abiotic factors explain community-level ecosystem multifunctionality to tion and reconstruction of degraded ecosystems.
a greater degree than the woody and herb layers (Table 1), and the
explanation of ecosystem multifunctionality is greater than that of a CRediT authorship contribution statement
single nutrient cycle.
Hengfang Wang: Data curation, Investigation, Methodology, Visu­
4.3. Dependencies and tradeoffs across different community vertical alization, Writing - original draft, Writing - review & editing. Guanghui
structures Lv: Conceptualization, Funding acquisition, Supervision, Writing - re­
view & editing. Yan Cai: Investigation, Writing - review & editing.
With the concurrent consideration of multiple biodiversity attri­ Xueni Zhang: Writing - review & editing. Lamei Jiang: Writing - review
butes, biodiversity generally shows a positive but weak impact on & editing. Xiaodong Yang: Writing - review & editing.
multifunctionality. This relationship is mainly due to the differences
among the factors influencing ecosystem multifunctionality according to
functions related to different biogeochemical cycles, in addition to the Declaration of Competing Interest
influence of different community vertical structures. The contribution of
the C and N cycles exceeds that of the P cycle at the community level and The authors declare that they have no known competing financial
in the woody layer (Fig. 4). The contributions of the C and P cycles interests or personal relationships that could have appeared to influence
exceeded those of the N cycle within the herb layer. the work reported in this paper.
The factors influencing the multifunctionality of different commu­
nity ecosystems depend not only on the combination of functions Acknowledgments
examined in the current study and the characteristics of the assessed
biodiversity attributes but also on the vertical structures of different This work was financially supported by the Natural Science Foun­
communities (Fig. 5). The degree to which the community level dation of Xinjiang Province (XJEDU2020I002, XJ2019G021), and the
explained the vertical structure exceeded that of the woody and herb National Natural Science Foundation of China (U1903216, 51964043,
layers (Fig. 5), which was consistent with the influence of the biomass of 41571034, 31700354).
different strata examined in the current study (Luo et al., 2019). Func­
tional traits differ among the different community structures; the con­ Appendix A. Supplementary data
straints of environmental factors are also greater (Yan et al., 2019), and
woody plants are more tolerant to extreme drought (Meng et al., 2015). Supplementary data to this article can be found online at https://doi.
Qin et al. (2020) found that environmental filtering plays a leading role org/10.1016/j.ecolind.2021.107830.
in the diversity of woody plants, whereas that of herb plants was found
to be affected by the combined effects of environmental filtering and References
diffusion restrictions through the study of community construction
mechanisms. Ali, A., Yan, E.-R., 2017. The forest strata-dependent relationship between biodiversity
and aboveground biomass within a subtropical forest. For. Ecol. Manage. 401,
In summary, ecosystem multifunctionality is affected by both di­
125–134. https://doi.org/10.1016/j.foreco.2017.06.056.
versity and environmental factors. On the one hand, it is important to Anderson, M.J., 2006. Distance-based tests for homogeneity of multivariate dispersions.
consider the effects of environmental factors on ecosystem multi­ Biometrics 62, 245–253. https://doi.org/10.1111/j.1541-0420.2005.00440.x.
functionality when analyzing the mechanism of ecosystem multi­ Balvanera, P., Siddique, I., Dee, L., et al., 2014. Linking biodiversity and ecosystem
services: current uncertainties and the necessary next steps. Bioscience 64, 49–57.
functionality (Gong et al., 2019), and soil moisture and salinity have a https://doi.org/10.1093/biosci/bit003.
significant impact on ecosystem functions. On the other hand, the Berdugo, M., Kéfi, S., Soliveres, S., et al., 2017. Plant spatial patterns identify alternative
various attributes reflected by biodiversity indicators should not be ecosystem multifunctionality states in global drylands. Nat. Ecol. Evol. 1, 0003.
https://doi.org/10.1038/s41559-016-0003.
ignored, including taxonomic diversity, phylogenetic, and functional Brum, F.T., Graham, C.H., Costa, G.C., et al., 2017. Global priorities for conservation
diversity indicators (Ali and Yan, 2017; Mori, 2018; Zhang et al., 2017b) across multiple dimensions of mammalian diversity. Proc. Natl. Acad. Sci. 114,
and the influences of different vertical structures of communities. 7641–7646. https://doi.org/10.1073/pnas.1706461114.
Byrnes, J., Lefcheck, J.S., Gamfeldt, L., et al., 2014. Multifunctionality does not imply
Maintenance of community composition is crucial for maintaining that all functions are positively correlated. Proc. Natl. Acad. Sci. 111, E5490.
ecosystem multifunctionality (Hautier et al., 2018). https://doi.org/10.1073/pnas.1419515112.
Cadotte, M.W., 2017. Functional traits explain ecosystem function through opposing
mechanisms. Ecol. Lett. 20, 989–996. https://doi.org/10.1111/ele.12796.
5. Conclusions Cadotte, M.W., Dinnage, R., Tilman, D., 2012. Phylogenetic diversity promotes
ecosystem stability. Ecology 93, S223–S233. https://doi.org/10.1890/11-0426.1.
Dryland ecosystem multifunctionality and the community-level C, N, Cardinale, B.J., Matulich, K.L., Hooper, D.U., et al., 2011. The functional role of producer
diversity in ecosystems. Am. J. Bot. 98, 572–592. https://doi.org/10.3732/
and P cycles were jointly affected by mass ratio and richness effects,
ajb.1000364.
whereas the woody and herb layers were mainly affected by the richness Carroll, I.T., Cardinale, B.J., Nisbet, R.M., 2011. Niche and fitness differences relate the
effects. The factors influencing the different vertical structures were maintenance of diversity to ecosystem function. Ecology 92, 1157–1165. https://doi.
quite different. Community-level ecosystem multifunctionality was org/10.2307/23213777.
Craven, D., Eisenhauer, N., Pearse, W.D., et al., 2018. Multiple facets of biodiversity
mainly affected by soil moisture and functional diversity, whereas the drive the diversity–stability relationship. Nat. Ecol. Evol. 2, 1579–1587. https://doi.
woody layer was mainly affected by soil moisture and leaf functional org/10.1038/s41559-018-0647-7.

9
H. Wang et al. Ecological Indicators 128 (2021) 107830

Durán, J., Delgado-Baquerizo, M., Dougill, A.J., et al., 2018. Temperature and aridity Pasari, J.R., Levi, T., Zavaleta, E.S., et al., 2013. Several scales of biodiversity affect
regulate spatial variability of soil multifunctionality in drylands across the globe. ecosystem multifunctionality. Proc. Natl. Acad. Sci. 110, 10219–10222. https://doi.
Ecology 99 (5), 1184–1193. https://doi.org/10.1002/ecy.2199. org/10.1073/pnas.1220333110.
Faith, D.P., 1992. Conservation evaluation and phylogenetic diversity. Biol. Conserv. 61, Qin, H., Zhang, Y.B., Dong, G., et al., 2020. Altitudinal patterns of taxonomic,
1–10. https://doi.org/10.1016/0006-3207(92)91201-3. phylogenetic and functional diversity of forest communities in Mount Guandi,
Flynn, D.F., Mirotchnick, N., Jain, M., et al., 2011. Functional and phylogenetic diversity Shanxi, China. Chinese J. Plant Ecol. 43 (9), 762–773. http://doi.org/10.17521
as predictors of biodiversity–ecosystem-function relationships. Ecology 92, /cjpe.2018.0088.
1573–1581. https://doi.org/10.1890/10-1245.1. Ricotta, C., 2005. A note on functional diversity measures. Basic Appl. Ecol. 6, 479–486.
Gamfeldt, L., Hillebrand, H., Jonsson, P.R., 2008. Multiple functions increase the https://doi.org/10.1016/j.baae.2005.02.008.
importance of biodiversity for overall ecosystem functioning. Ecology 89, Roger, F., Bertilsson, S., Langenheder, S., et al., 2016. Effects of multiple dimensions of
1223–1231. https://doi.org/10.1890/06-2091.1. bacterial diversity on functioning, stability and multifunctionality. Ecology 97,
García-Palacios, P., Gross, N., Gaitán, J., et al., 2018. Climate mediates the 2716–2728. https://doi.org/10.1002/ecy.1518.
biodiversity–ecosystem stability relationship globally. Proc. Natl. Acad. Sci. 115, Salguero-Gómez, R., Jones, O.R., Archer, C.R., et al., 2015. The COMPADRE plant matrix
8400–8405. https://doi.org/10.1073/pnas.1800425115. database: an open online repository for plant demography. J. Ecol. 103, 202–218.
Garnier, E., Cortez, J., Billès, G., et al., 2004. Plant functional markers capture ecosystem https://doi.org/10.1111/1365-2745.12334.
properties during secondary succession. Ecology 85, 2630–2637. https://doi.org/ Shannon, C.E., 2001. A mathematical theory of communication. ACM SIGMOBILE
10.1890/03-0799. Mobile Comput. Commun. Rev. 5, 3–55. https://doi.org/10.1002/j.1538-
Gilbert, G.S., Webb, C.O., 2007. Phylogenetic signal in plant pathogen–host range. Proc. 7305.1948.tb01338.x.
Natl. Acad. Sci. 104, 4979–4983. https://doi.org/10.1073/pnas.0607968104. Simpson, E.H., 1949. Measurement of diversity. Nature 163, 688. https://doi.org/
Gong, Y., Ling, H., Lv, G., et al., 2019. Disentangling the influence of aridity and salinity 10.1038/163688a0.
on community functional and phylogenetic diversity in local dryland vegetation. Sci. Soliveres, S., Maestre, F.T., Eldridge, D.J., et al., 2014. Plant diversity and ecosystem
Total Environ. 653, 409–422. https://doi.org/10.1016/j.scitotenv.2018.10.358. multifunctionality peak at intermediate levels of woody cover in global drylands.
Gross, N., Le Bagousse-Pinguet, Y., Liancourt, P., et al., 2017. Functional trait diversity Glob. Ecol. Biogeogr. 23, 1408–1416. https://doi.org/10.1111/geb.12215.
maximizes ecosystem multifunctionality. Nat. Ecol. Evol. 1, 1–9. https://doi.org/ Speiser, J.L., Miller, M.E., Tooze, J., et al., 2019. A comparison of random forest variable
10.1038/s41559-017-0132. selection methods for classification prediction modeling. Expert Syst. Appl. 134,
Hautier, Y., Isbell, F., Borer, E.T., et al., 2018. Local loss and spatial homogenization of 93–101. https://doi.org/10.1016/j.eswa.2019.05.028.
plant diversity reduce ecosystem multifunctionality. Nat. Ecol. Evol. 2, 50–56. Tang, L.S., Zhang, J.B., Cheng, X.J., 2002. Land Degradation and Ecological
https://doi.org/10.1038/s41559-017-0395-0. Regeneration in the Ecotones between the Oases and Deserts in the Arid Areas. Arid
Hector, A., Bagchi, R., 2007. Biodiversity and ecosystem multifunctionality. Nature 448, Zone Res. 19 (3), 43–48. http://doi.org/10.13866/j.azr.2002.03.009.
188–190. https://doi.org/10.1038/nature05947. Tilman, D., Lehman, C.L., Thomson, K.T., 1997. Plant diversity and ecosystem
Li, X., Yang, K., Zhou, Y., 2016. Progress in the study of oasis-desert interactions. Agric. productivity: theoretical considerations. Proc. Natl. Acad. Sci. 94, 1857–1861.
For. Meteorol. https://doi.org/10.1016/j.agrformet.2016.08.022. https://doi.org/10.1073/pnas.94.5.1857.
Kursa, M.B., Rudnicki, W.R., 2010. Feature selection with the Boruta package. J. Stat. Tilman, D., Wedin, D., Knops, J., 1996. Productivity and sustainability influenced by
Softw. 36, 1–13. http://doi.org/10.18637/jss.v036.i11. biodiversity in grassland ecosystems. Nature 379, 718–720. https://doi.org/
Laliberté, E., Legendre, P., 2010. A distance-based framework for measuring functional 10.1038/379718a0.
diversity from multiple traits. Ecology 91 (1), 299–305. https://doi.org/10.1890/08- Van der Plas, F., Manning, P., Soliveres, S., et al., 2016. Biotic homogenization can
2244.1. decrease landscape-scale forest multifunctionality. Proc. Natl. Acad. Sci. 113,
Le Bagousse-Pinguet, Y., Soliveres, S., Gross, N., et al., 2019. Phylogenetic, functional, 3557–3562. https://doi.org/10.1073/pnas.1517903113.
and taxonomic richness have both positive and negative effects on ecosystem Villéger, S., Mason, N.W., Mouillot, D., 2008. New multidimensional functional diversity
multifunctionality. Proc. Natl. Acad. Sci. 116, 8419–8424. https://doi.org/10.1073/ indices for a multifaceted framework in functional ecology. Ecology 89, 2290–2301.
pnas.1815727116. https://doi.org/10.1890/07-1206.1.
Luo, Y.H., Cadotte, M.W., Burgess, K.S., et al., 2019. Greater than the sum of the parts: Wang, G.X., Cheng, G.D., Xu, Z.M., 1999. The utilization of water resource and its
how the species composition in different forest strata influence ecosystem function. influence on eco-environment in the northwest arid area of china. J. Nat. Resourc. 14
Ecol. Lett. 22, 1449–1461. https://doi.org/10.1111/ele.13330. (2), 109–116.
Maestre, F.T., Castillo-Monroy, A.P., Bowker, M.A., et al., 2012a. Species richness effects Wang, H., Cai, Y., Yang, Q., et al., 2019. Factors that alter the relative importance of
on ecosystem multifunctionality depend on evenness, composition and spatial abiotic and biotic drivers on the fertile island in a desert-oasis ecotone. Sci. Total
pattern. J. Ecol. 100, 317–330. https://doi.org/10.1111/j.1365-2745.2011.01918.x. Environ. 697, 134096 https://doi.org/10.1016/j.scitotenv.2019.134096.
Maestre, F.T., Quero, J.L., Gotelli, N.J., et al., 2012b. Plant species richness and Webb, C.O., Ackerly, D.D., Kembel, S.W., 2008. Phylocom: software for the analysis of
ecosystem multifunctionality in global drylands. Science 335, 214–218. https://doi. phylogenetic community structure and trait evolution. Bioinformatics 24 (18),
org/10.1126/science.1215442. 2098–2100. https://doi.org/10.1093/bioinformatics/btn358.
Meng, T.T., Wang, H., Harrison, S.P., et al., 2015. Responses of leaf traits to climatic Webb, C.O., Ackerly, D.D., McPeek, M.A., et al., 2002. Phylogenies and community
gradients: adaptive variation versus compositional shifts. Biogeoences 12 (9), 1–14. ecology. Annu. Rev. Ecol. Syst. 33, 475–505. https://doi.org/10.2307/3069271.
https://doi.org/10.5194/bg-12-5339-2015. Yan, R.Y., Lou, A.R., Xie, S.Y., et al., 2019. The composition and distribution pattern of
Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-Being: Synthesis. plant life forms in Alxa Plateau. J. Beijing Normal Univ. (Natural Science) 55 (3),
Island Press, Washington DC. 349–355. https://doi.org/10.16360/j.cnki.jbnuns.2019.03.009.
Mori, A.S., 2018. Environmental controls on the causes and functional consequences of Zavaleta, E.S., Pasari, J.R., Hulvey, K.B., et al., 2010. Sustaining multiple ecosystem
tree species diversity. J. Ecol. 106, 113–125. https://doi.org/10.1111/1365- functions in grassland communities requires higher biodiversity. Proc. Natl. Acad.
2745.12851. Sci. 107, 1443–1446. https://doi.org/10.2307/40536347.
Naeem, S., Prager, C., Weeks, B., et al., 2016. Biodiversity as a multidimensional Zhang, Q., Li, J., Xu, W., et al., 2017a. Estimation of biomass allocation and carbon
construct: a review, framework and case study of herbivory’s impact on plant density of Rhododendron semi shrub land in the subtropical mountainous areas of
biodiversity. Proc. Royal Soc. B Biol. Sci. 283, 20153005. https://doi.org/10.1098/ China. Chinese J. Plant Ecol. 41, 43–52. https://doi.org/10.17521/cjpe.2016.0174.
rspb.2015.3005. Zhang, Y., Chen, H.Y., Taylor, A.R., 2017b. Positive species diversity and above-ground
Naeem, S., Thompson, L.J., Lawler, S.P., et al., 1994. Declining biodiversity can alter the biomass relationships are ubiquitous across forest strata despite interference from
performance of ecosystems. Nature 368, 734–737. https://doi.org/10.1038/ overstorey trees. Funct. Ecol. 31, 419–426. https://doi.org/10.1111/1365-
368734a0. 2435.12699.
Nakagawa, S., Schielzeth, H., 2013. A general and simple method for obtaining R2 from Zirbel, C.R., Grman, E., Bassett, T., et al., 2019. Landscape context explains ecosystem
generalized linear mixed-effects models. Methods Ecol. Evol. 4, 133–142. https:// multifunctionality in restored grasslands better than plant diversity. Ecology 100,
doi.org/10.1111/j.2041-210x.2012.00261.x. e02634. https://doi.org/10.1002/ecy.2634.

10