ORIGINAL RESEARCH article

Front. Plant Sci., 02 June 2023

Sec. Plant Biophysics and Modeling

Volume 14 - 2023 | https://doi.org/10.3389/fpls.2023.1131778

Prediction of spatial distribution characteristics of ecosystem functions based on a minimum data set of functional traits of desert plants

  • 1. College of Ecology and Environment, Xinjiang University, Urumqi, China

  • 2. Key Laboratory of Oasis Ecology of Education Ministry, Xinjiang University, Urumqi, China

  • 3. Xinjiang Jinghe Observation and Research Station of Temperate Desert Ecosystem, Ministry of Education, Jinghe, China

Abstract

The relationship between plant functional traits and ecosystem function is a hot topic in current ecological research, and community-level traits based on individual plant functional traits play important roles in ecosystem function. In temperate desert ecosystems, which functional trait to use to predict ecosystem function is an important scientific question. In this study, the minimum data sets of functional traits of woody (wMDS) and herbaceous (hMDS) plants were constructed and used to predict the spatial distribution of C, N, and P cycling in ecosystems. The results showed that the wMDS included plant height, specific leaf area, leaf dry weight, leaf water content, diameter at breast height (DBH), leaf width, and leaf thickness, and the hMDS included plant height, specific leaf area, leaf fresh weight, leaf length, and leaf width. The linear regression results based on the cross-validations (FTEIW - L, FTEIA - L, FTEIW - NL, and FTEIA - NL) for the MDS and TDS (total data set) showed that the R2 (coefficients of determination) for wMDS were 0.29, 0.34, 0.75, and 0.57, respectively, and those for hMDS were 0.82, 0.75, 0.76, and 0.68, respectively, proving that the MDSs can replace the TDS in predicting ecosystem function. Then, the MDSs were used to predict the C, N, and P cycling in the ecosystem. The results showed that non-linear models RF and BPNN were able to predict the spatial distributions of C, N and P cycling, and the distributions showed inconsistent patterns between different life forms under moisture restrictions. The C, N, and P cycling showed strong spatial autocorrelation and were mainly influenced by structural factors. Based on the non-linear models, the MDSs can be used to accurately predict the C, N, and P cycling, and the predicted values of woody plant functional traits visualized by regression kriging were closer to the kriging results based on raw values. This study provides a new perspective for exploring the relationship between biodiversity and ecosystem function.

1 Introduction

Plant functional traits are morphological, physiological, and life history traits that indirectly affect plant fitness (Violle et al., 2007). In natural ecosystems, plants adapt to external changes by the changes of traits such as height, leaf area, leaf mass, leaf longevity, seed size, and seed dispersal mode, which may also lead to changes in ecosystem functions (Li et al., 2008; Messier et al., 2010; ; Kraft et al., 2015). However, plant species are diverse in nature, and plant functional traits are influenced by factors such as climate change and human disturbance (He et al., 2018). Therefore, using plant functional traits to reflect and predict changes in plant community and ecosystem function is of great importance.

Many studies have explored the relationship between plant functional traits and ecosystem processes or functions. Most researchers believe that the relative biomass of dominant species in plant communities and their specific traits dominate the dynamics of ecosystem processes in time and space (Vile et al., 2006; ; ; Lohbeck et al., 2015). For example, Grime (1998) proposed the “mass ratio hypothesis”, arguing that plant functional traits can be used to predict ecosystem functions or processes (; Niu et al., 2010; ). Furthermore, some scholars held that the change of one plant functional trait may lead to changes in multiple ecosystem functions, and one ecosystem function may be simultaneously affected by multiple plant functional traits (Temmerman et al., 2005; Kearney and Fagherazzi, 2016; Sun et al., 2020).

At present, there are two main ways to study the functional traits of plant communities. One is to use community functional parameters based on plant functional traits, for example, the community weighted mean (CWM) of plant functional traits, which is calculated using the weighted average of functional traits and relative abundances of species (Kattge et al., 2011; Zhang et al., 2011b; Zhang et al., 2020a). The other is to use plant functional trait diversity, for example, the size, range, and distribution of plant functional trait values in a community, which is considered important for biodiversity (; Huang et al., 2019b; Lin et al., 2022). Studies have shown that community functional parameters based on plant functional traits and plant functional trait diversity can influence plant community structure and ecosystem functions or processes (Flynn et al., 2011; Mei et al., 2017). However, the parameters are numerous, redundant, and cumbersome. Therefore, the selection of representative parameters that play an important role in ecosystem functioning has become the key to current research.

With the in-depth study of ecosystem functions, researchers gradually realize that ecosystems provide multiple ecosystem functions simultaneously, i.e. ecosystem multifunctionality (Pasari et al., 2013; Garland et al., 2020). Most previous studies have focused on the effects of a single trait on the functions of a single ecosystem (Fortunel et al., 2009; Zuo et al., 2016; Pakeman and Fielding, 2020) and the quantification of plant functional trait diversity and ecosystem functions (Petchey and Gaston, 2002; Laliberte and Legendre, 2010; Isbell et al., 2011). In recent years, quantitative analysis of the relationship between multiple plant functional traits and multiple functions of single ecosystems has been sought after (Wei et al., 2016a; Wei et al., 2016b; Liu et al., 2017; Liang et al., 2019; Li et al., 2020b). Spatial heterogeneity of single functions of single ecosystems has been proposed by ecologists and botanists at landscape and regional levels (Lin et al., 2010; Giese et al., 2013; Liu et al., 2018). Some scholars reported that morphological variation and spatial distribution of plant functional traits are the results of environmental filtering and biological interactions, reflecting plant adaptations to their habitats (; Ma et al., 2019; Zhang et al., 2020b). Therefore, by analyzing the spatial distribution of functional traits and their relationship to environment, it is possible to how plants respond to environmental changes and how the responses affect the functions of single ecosystems.

Arid regions account for about 41% of the world’s total land, and about 38% of the population lives in arid regions (Reynolds et al., 2007). Due to the influences of climate change and anthropogenic disturbances, the aridification of terrestrial ecosystems is exacerbating (Gao and Giorgi, 2008; ). In Xinjiang, China, the desert area (65.46 × 104 km2) accounts for 39% of the total area of Xinjiang, and has increased significantly (Ni and OuYang, 2006). Previous studies have shown that the proportions of C, N, and P in the total elemental content are relatively stable in desert ecosystems (; ). However, plants with different life forms affect C, N, and P cycling to a certain extent, which could impact the spatial distribution of the functions of C, N, and P cycling in ecosystems (Wang and Yu, 2008; Wang et al., 2019).

Therefore, based on the spatial heterogeneity of ecosystem functions, the spatial distributions of the functions of C, N, and P cycling were predicted using the MDS of the dominant plant functional traits in a temperate desert region by regression kriging (RK), a method that combines regression modeling with kriging (Sarmadian et al., 2014; Pham et al., 2019). The objectives were to: (1) Select the functional traits of woody and herbaceous plants that play a dominant role in temperate desert ecosystems to construct the wMDS and hMDS, (2) determine the spatial distribution characteristics of C, N, and P cycling in temperate desert ecosystems using geostatistical methods, and (3) predict the spatial distribution characteristics of C, N and P cycling using linear and non-linear models based on the wMDS and hMDS. This study will advance our understanding of the relationship between plant functional traits and ecosystem function.

2 Materials and methods

2.1 Study site, sampling and experiment design

The study area is located in the Xinjiang Ebinur Lake Wetland National Nature Reserve on the southwestern edge of the Junggar Basin (44°30’ - 45°09’N, 82°36’ - 83°50’E). Surrounded by mountains on three sides, it is the lowest depression and water and salt enrichment centre (Wang et al., 2021a). In the Reserve, swamps, rivers, salt lakes, riparian forests, and deserts are the main landscapes. Aeolian sandy soil, grey brown desert soil, and grey desert soil are the zonal soils, and saline soil is the intrazonal soil. Central Asian and Mongolian flora is the main part of vegetation (He et al., 2014; Yang et al., 2014).

Three plots (100 m × 100 m) perpendicular to the Aqikesu River were set up from southwest to northeast (A-C) in the riparian forest-desert transition zone in the north of the Aqikesu River, and the distance between the plots was about 1.5 km (Figure 1). From July to August 2018, Herbaceous and woody plants were surveyed separately and the soil under the canopy of their collections was collected. Each plot was divided into 100 subplots (10 m × 10 m) (300 subplots totally), and the multiplicity, plant height (H), crown width (CW), leaf thickness (LT), leaf length (LL), leaf width (LW), DBH/basal diameter, leaf area (LA), and leaf fresh weight (LFW) of plants were recorded.

Figure 1

2.2 Plant and soil physicochemical experiments

Three plants of each species in each subplot were selected for the following determinations. The CW of trees was measured using a laser rangefinder (Dimetix-DAE-10-050, Dimetix, Switzerland), and that of shrubs and herbs was measured using a steel tape measure. For all trees in the subplots, the DBH was measured using a tape measure at a height of 1.3 m. For shrubs and herbs, the basal diameter was measured at 2.54 cm from the ground. Three to five leaves at each plant position (upper, middle, and bottom) were collected to determine the LT, LL, and LW using digital vernier caliper. To measure the LA, a 1 cm scale was marked on the lower right corner of a paper, then 10-20 leaves were laid flatly on the paper, followed by the photographing using a camera parallel to the paper. The pictures were processed using Photoshop software (2020CC, Adobe, USA) to obtain LA. After LA measurement, the leaves were transferred in sealed bags, weighed immediately (fresh weight), and dried for the measurements of dry weight and C, N, and P concentrations according to the methods of (Table S1).

Within each subplot, five sampling points were selected along the diagonal, and the 0-30 cm soil layer was sampled at each sampling point for the determinations of soil C, N, and P concentrations () (Table S1).

2.3 Statistical analyses

2.3.1 CWM

The CWM was calculated as a sample of the functional trait values within each subplot, based on the species diversity in each subplot and the measured functional trait values.

where is the species abundance in a subplot, and is the functional trait value of a species in a subplot (Ren, 2021).

2.3.2 Construction of minimal data sets of plant functional traits

Sixteen plant functional traits were selected, including leaf carbon (LC), leaf nitrogen (LN), leaf phosphorus (LP), specific leaf area (SLA), specific leaf weight (SLW), LA, leaf dry matter content (LDMC), leaf water content (LWC), LFW, leaf dry weight (LDW), diameter at breast height (DBH) (stem base diameter (SBD)), H, LT, LL, LW, and width to length ratio (WLR) to perform Kaiser Mayer-Olkin (KMO) test and Bartlett’s test based on partial correlation. If tests were passed, factor analysis was performed on the selected traits. All above analyses were performed by using the psych and vegan package in R software (R Development Core Team, 2021).

The Norm value is an important reference for trait selection and MDS construction. The Norm value indicates the combined loading of a trait on multiple principal components with PC ≥ 1. So the larger the Norm value of a trait, the higher its combined loading value, and the more information on the principal components with PC ≥ 1 it has (Wu et al., 2019). Norm values were calculated as follows:

where is the combined loadings of the ith variable on k principal components with eigenvalue ≥ 1(Norm value), is the loading of the ith variable on the kth principal component, and is the eigenvalue of the kth principal component (Wu et al., 2019).

Based on the results of the factor analysis, principal component variables with eigenvalues greater than 1 (PC > 1) were screened out. Among the variables, the traits with factor loadings |PC| ≥ 0. 5 were screened out and grouped by different principal component variables. If a trait had a factor loading value |PC| ≥ 0.5 in two main variables, it was classified into the group with the lower correlation. Norm values of each group of traits were compared. Traits with Norm values in the top 10% in each group were retained, and the rest was discarded. If there were multiple traits in the top 10%, the correlations between the trait with the highest Norm value in each group and the traits in the top 10% were checked. If the correlation coefficient |r| ≥ 0.5, then the trait with higher Norm value and coefficient of variation was put into the MDS. If |r|< 0.5, then both were put into the MDS (Pulido et al., 2017).

2.3.3 Minimum data set test model

Linear and non-linear scores of each trait was used to assign scores to the functional traits in the MDS:

where SL is the score derived from the linear score, ranging from 0 to 1, X is the value of a functional trait in the MDS, and Xmax and Xmin are the maximum and minimum value of each functional trait, respectively.

Equation (3) was applied to positive functional traits (the higher the value, the better the plant growth), while equation (4) was applied to negative functional traits (the lower the value, the better the plant growth) (; Li et al., 2020a). The formula for the non-linear function:

where SNL is the score derived from the non-linear score, ranging from 0 to 1, a is the maximum value that can be obtained for a functional trait, which is defined as 1 in this study, X is the value of a functional trait in the MDS, Xm is the average value of a functional trait, and b is the slope, which is set to -2.5 for positive functional traits and +2.5 for negative functional traits (Zhang et al., 2011a; Raiesi, 2017).

After all functional traits in the MDS were assigned a score, the scores of all functional traits were summed using the following formula. The evaluation index FTEIA is the average of the scores of the traits in the MDS, while FTEIW is the sum of the scores of the functional traits multiplied by the corresponding weights ().

where FTEIA and FTEIW are the functional trait evaluation indices calculated without and with weights, respectively, Si is the score of the functional trait i, n is the number of functional traits in the MDS, and Wi is the weight of the functional trait i. The weight was determined by the ratio of the characteristic roots of the principal component of functional trait i in the PCA analysis to the sum of all characteristic roots in the MDS (Guo et al., 2017; Jin et al., 2018).

2.3.4 Ecosystem functions

The three ecosystem functions were the C, N, and P cycling in this study. The C cycling indicators included plant organic carbon and soil organic carbon, the N cycling indicators included soil nitrate nitrogen, ammonium nitrogen, and total nitrogen, and plant leaf total nitrogen, the P cycling indicators included soil available phosphorus and total phosphorus and plant leaf total phosphorus. All above were calculated by the average method (Maestre et al., 2012; ).

where EF is a single ecosystem function, fi is the measured value of function i, ri is the mathematical function that converts fi to a positive value, g is the normalization of all measured values, and n represents the number of functions measured.

2.3.5 Linear and non-linear predictive models

In this study, linear and non-linear models were used to predict the functions of single ecosystems based on the MDSs of plant functional traits. Models selected 70% sample size of C, N and P cycle indices of each first-level plot (100 m × 100 m) for training, and the remaining 30% sample size for model verification. The linear model was constructed by using partial least squares regression (PLS), which was based on covariance regression. PLS models can effectively catch the unique contributions of each independent variable to overcome multicollinearity. PLS models were constructed by the pls package in R software (Ge et al., 2018; Tan et al., 2020).

The non-linear model was constructed by using the Random Forest (RF) and BP neural network (BPNN). RF algorithm was based on the statistical learning theory of decision trees, which can effectively process high-dimensional data and overcome the overfitting. In this study, the RF model set the number of trees to 100, the GBM used the default setting, the maximum number of iterations was set to 500, the linear output unit was used, and the grid was set with the option to optimize the hyperparameters. The model constructions using the RF were completed by using the random Forest package (; Poggio et al., 2019; Vilchez-Mendoza et al., 2022). BPNN model is a multi-layer feed-forward neural network that continuously approximates the desired output based on the backward transmission of errors to obtain a prediction. In this study, the BPNN model built four hidden layers with five nodes in each layer and used backprop algorithm for calculation. The model constructions using the BPNN were completed by using the neural net package (Huang et al., 2019a; Morais et al., 2021).

Root mean square error (RMSE) and mean absolute error (MAE) were used to test the accuracy of the predictive models (Qiu et al., 2010; Guo et al., 2013; Knadel et al., 2021).

where is the predicted value of a subplot, is the measured value of a subplot, and N is the number of subplots. The smaller the RMSE and MAE, the higher the prediction accuracy.

2.3.6 Geostatistical analysis

2.3.6.1 Regression kriging

Regression kriging (RK) is a spatial interpolation technique that performs kriging interpolation on the prediction residuals by combining the regression of the dependent variable on the predictor variables (such as environmental variables) (Hengl et al., 2004). That is to say, RK is a hybrid method that combines a simple or multiple linear regression model with ordinary kriging for predicting residuals. RK allows the auxiliary variables to interpolate the dependent variables at unsampled locations (Hengl et al., 2007). In this study, Partial least squares kriging (PLSK) (Guo et al., 2021), Random forest kriging (RFK) (; ), and Back propagation neural network (BPNNK) models were constructed (Li et al., 2017; ). Regression Kriging (RK)was constructed using the predicted values from the PLS, RF and BPNN models in combination with the Ordinary Kriging (OK) method. Aim of this method was to establish linear and non-linear mapping relationships between the MDSs and a single ecosystem function by PLS, RF, and BPNN. Relative coordinates were established in each subplot, and then the residual terms were spatially interpolated using the OK method to obtain the final prediction results. This was completed by using the automap and gstat packages (Meng and Liu, 2013; Mukherjee et al., 2015).

where is the interpolation result at the predicted subplot , is the deterministic part of the fitting by regression, is the interpolation result on the regression residuals by OK method, k is the position number in the fitting by regression, p is the sample size of the regression model based on the predicted values, is the coefficient of the regression model, is the intercept when k=0, i is the position number at the regression residual interpolation, n is the sample size for kriging interpolation of the residual value based on the predicted values, is the value of the auxiliary variable at the predicted position , is the OK interpolation weight determined by the spatial correlation structure of the regression residual, and is the residual at position .

2.3.6.2 Semi-variable functions

The traits were tested for normal distribution before geostatistical analysis. If the traits did not follow normal distribution, they would be transformed with Box-Cox (Wang et al., 2021a). The spatial variability of ecosystem functional diversity was analyzed using geostatistical software (GS+, version 9.0, Gamma Design Software. LLC, USA), and a semi-covariance function model was fitted. By analyzing the nugget (C0), structural variance (C), sill (C0+ C), variation range (Range), and nugget-sill ratio (C0/(C0+ C)), the proportion of nugget variance in total spatial heterogeneity variance was determined, i.e. the proportion of structural variance in the total variance. It is often used to describe the degree of variation in the spatial heterogeneity of study objects. If C0/(C0+ C)< 25%, the variable has strong spatial autocorrelation; if C0/(C0+ C) is between 25% and 75%, the variable has moderate spatial autocorrelation among the variables; if C0/(C0+ C) > 75%, the variable has weak spatial autocorrelation (Robertson et al., 1993; Zartman, 2005).

where γ(h) is the semi-variance of the interval class h, N(h) is the number of samples separated by the lag distance, and Ai(xi) and Ai(xi+h) are the measurement variables for spatial locations i and i+h, respectively. There are four types of models: linear, spherical, exponential, and Gaussian. The coefficient of determination (R2) and the residual sum of squares (RSS) were used to select the best-fitting model. The larger the R2, the smaller the RSS, the better the model fitted (; Wang et al., 2021b).

3 Results

3.1 Indicators of minimum data set of functional traits

In the three plots (A-C), where were the species and frequencies of plant in Table S2, the wMDS and hMDS were constructed based on the functional traits of woody and herbaceous plants in the subplots (Table 1). The results showed that the SBD, H, LT, LL, LW, and WLR of herbaceous plants in plot C were different (P< 0.05) from those of herbaceous plants in plots A and B. The SBD and H in plot B was different (P< 0.05) from those in plot A. Therefore, the functional traits of woody and herbaceous plants in the three plots were combined to construct wMDS and hMDS.

Table 1

Woody plantHerbaceous plant
Trait variablesPlot APlot BPlot CTrait variablesPlot APlot BPlot C
LC0.992 a0.990 a0.989 aLC0.992 a0.978 a0.983 a
LN0.991 a0.991 a0.987 aLN0.990 a0.987 a0.981 a
LP0.991 a0.988 a0.981 aLP0.990 a0.982 a0.969 a
SLA0.985 a0.974 a0.966 aSLA0.975 a0.972 a0.980 a
SLW0.984 a0.971 a0.963 aSLW0.966 a0.962 a0.978 a
LA0.893 a0.944 a0.924 aLA0.938 a0.836 a0.924 a
LDMC0.988 a0.983 a0.985 aLDMC0.993 a0.948 a0.962 a
LWC0.993 a0.992 a0.991 aLWC0.994 a0.983 a0.991 a
LFW0.896 a0.950 a0.926 aLFW0.955 a0.890 a0.926 a
LDW0.898 a0.948 a0.927 aLDW0.950 a0.840 a0.917 a
DBH0.658 a0.763 a0.797 aSBD0.593 a0.879 b0.629 b
H0.928 a0.855 a0.885 aH0.982 a0.847 ab0.921 b
LT0.827 a0.905 a0.956 aLT0.983 a0.846 a0.965 b
LL0.844 a0.833 a0.747 aLL0.966 a0.803 a0.964 b
LW0.274 a0.845 a0.878 aLW0.485 a0.856 a0.933 b
WLR0.247 a0.913 a0.858 aWLR0.538 a0.887 a0.937 b

Analysis of variance for functional traits of woody and herbaceous plants in plots A, B and C.

The KMO test based on partial correlation on the CWMs of the 16 plant functional traits showed that the KMO value of woody plants was 0.77 (P< 0.001), with Bartlett’s test P< 1.7e-16, and that of herbaceous plants was 0.74 (P< 0.001), with Bartlett’s test P< 2.2e-16. This indicates that there is a correlation between the woody and herbaceous functional traits, and it is suitable for factor analysis.

Based on the results of the factor analysis (Table 2), the CWMs of the 16 woody plant functional traits were initially divided into seven groups. For group 1, because the ratios of the Norm values of LC, LN, LP, and LL to that of H were less than 90%, H was included in the wMDS. For group 2, the ratio of the Norm value of SLW to that of SLA was higher than 90%, and the Pearson correlation analysis (Figure 2A) showed that the correlation coefficient was greater than 0.5, thus SLA was included in the wMDS. With the same screening method as group 2, the LDW, LWC, and LW were included into the wMDS for groups 3, 4, and 7, respectively. Group 5 and 6 only had DBH and LT, respectively, which were included into the wMDS. Therefore, seven traits (H, SLA, LDW, LWC, LW, DBH, and LT) were ultimately included in the wMDS.

Table 2

Trait variablesPrincipal component (PC)GroupsNorm valueMDS
PC1PC2PC3PC4PC5PC6
LC0.062-0.055-0.0320.094-0.063-0.05110.218―
LN0.0030.0270.0740.009-0.0100.00910.119―
LP0.0650.0050.0810.063-0.0620.00710.201―
SLA-0.070-0.9570.1330.053-0.031-0.11021.620Enter
SLW0.0510.945-0.091-0.1070.0370.16321.602―
LA0.937-0.280-0.0640.044-0.045-0.02231.738―
LDMC-0.0040.083-0.966-0.0350.036-0.10041.439―
LWC-0.005-0.1270.9660.023-0.0550.08641.446Enter
LFW0.9510.1810.175-0.011-0.0130.08131.744―
LDW0.9600.207-0.108-0.005-0.0040.02331.753Enter
DBH-0.0540.053-0.072-0.0330.9490.02051.186Enter
H0.0730.072-0.243-0.0740.343-0.02310.590Enter
LT0.0640.2460.177-0.0590.0190.94661.220Enter
LL0.1020.101-0.023-0.0190.1280.03110.299―
LW0.013-0.0560.0430.982-0.004-0.01271.294Enter
WLR0.009-0.0900.0130.973-0.038-0.05271.288―
Characteristic value3.1742.7802.1771.7221.5251.377―――
Variance contribution
rate of the PC (%)		17.14412.99212.80612.2056.5866.044―――
Cumulative contribution
rate of the PC(%)		17.14430.13642.94155.14661.73267.776―――

Factor loading matrices, groupings and Norm values for functional traits in woody plant communities.

Figure 2

The CWMs of herbaceous plant functional traits were divided into six groups (Table 3). For group 1, the ratio of the Norm values of LC, LN, LP, SBD, and LT to that of H was lower than 90%, then H was included in the hMDS. For group 2, the ratio of the Norm value of SLW to that of SLA was higher than 90% and the Pearson correlation analysis (Figure 2B) showed that the correlation coefficient was greater than 0.5, then SLA was included in the hMDS. With the same screening method as group 2, LFW, LWC, and LW were included in the hMDS for groups 3, 4, and 6, respectively. Group 5 had LL only. Therefore, six traits (H, SLA, LFW, LWC, LW, and LL) were ultimately included in the hMDS

Table 3

Trait variablesPrincipal component (PC)GroupsNorm valueMDS
PC1PC2PC3PC4PC5
LC0.0860.3900.165-0.0100.09710.638―
LN0.0770.0550.0200.0050.05910.213―
LP-0.074-0.015-0.048-0.054-0.01010.201―
SLA-0.103-0.131-0.943-0.027-0.03721.341Enter
SLW0.0220.1100.9530.0470.02621.330―
LA0.8990.265-0.1940.0050.12532.189―
LDMC0.1700.8820.149-0.0340.17341.335―
LWC-0.210-0.884-0.155-0.006-0.15341.367Enter
LFW0.965-0.0610.1570.0250.08732.307Enter
LDW0.9110.2760.162-0.0070.11632.214―
SBD0.1320.1420.044-0.0030.06210.384―
H0.2890.4000.096-0.0040.26910.951Enter
LT-0.136-0.4610.014-0.016-0.14110.743―
LL0.2780.2970.0590.0470.86751.284Enter
LW-0.094-0.130.0150.944-0.18561.249Enter
WLR0.1300.1240.0650.9050.27461.246―
Characteristic value5.6381.9711.9131.6041.370―――
Variance contribution
rate of the PC (%)		18.09815.05612.38710.7576.634―――
Cumulative contribution
rate of the PC (%)		18.09833.15445.54156.29862.932―――

Factor loading matrix, groupings and Norm values for functional traits in herbaceous communities.

3.2 Minimum data set test

Four cross-validation methods including FTEIW - L (linear weight evaluation method), FTEIA - L (linear average evaluation method), FTEIW - NL (non-linear weight evaluation method), and FTEIA - NL (non-linear average evaluation method) were used to test the correlation between MDS and TDS in this study. The R2 of the linear regression of FTEIW - L, FTEIA - L, FTEIW - NL, and FTEIA - NL of MDS and TDS of woody plants were 0.29, 0.34, 0.75, and 0.57, respectively, and those of herbaceous plants were 0.82, 0.75, 0.76, and 0.68, respectively. Overall, the test results by FTEIW - L, FTEIA - L, FTEIW - NL, and FTEIA - NL all showed significant correlation (Figure 3).

Figure 3

In this study, the SL method outperformed the SNL method when the same evaluation system was used (Figure 3). By comparing different traits, it was found that the correlation coefficients calculated by FTEIA were higher than those calculated by FTEIW under the same assignment function method. Furthermore, MDS and TDS showed a positive correlation (P< 0.001) in all the resultant models tested by cross-validation. Therefore, the constructed wMDS and hMDS can replace TDS.

3.3 Spatial distribution characteristics of ecosystem functions

The optimal semi-variance function model for predicting C cycling based on the raw values, wMDS, and hMDS was the exponential model. The analysis results of C0/(C0+ C) showed that the structure of spatial variability had strong spatial autocorrelation (Table 4). This suggests that the spatial distribution of C cycling is mainly influenced by structural factors. The optimal semi-variance models based on raw and predicted values of N cycling were exponential and Gaussian models (Table 5). Except for the moderate spatial autocorrelation of the PLS model-predicted values based on the wMDS in plot B, the C0/(C0+ C) of the remaining raw and predicted values also showed strong spatial autocorrelation. This indicates that there is an error in PLS prediction accuracy and the spatial heterogeneity is dominated by structural factors. The optimal semi-variance models based on raw and predicted values of P cycling were exponential and Gaussian models (Table 6). The C0/(C0+ C) of raw values and predicted values based on wMDS for plot B showed moderate spatial autocorrelation, while that of the remaining values showed strong spatial autocorrelation, with structural factors dominating spatial heterogeneity. In summary, the C, N, and P cycling showed strong spatial autocorrelation, and the prediction accuracy of the RF and BPNN models were better than that of PLS model. By comparing the C, N, and P cycling predicted results of the three models based on the wMDS, it was found that the prediction accuracy of the RF model was higher than that of the BPNN and PLS models (Figures 4A–C). Furthermore, the same results were obtained by comparing the predicted results of the three models based on the hMDS (Figures 4D–F).

Table 4

(Woody)
Plots/Kriging		ModelsC0C0+CC0/(C0+C)
(%)		Range
(m)		R2RSS
ARealExp0.00440.07835.65.60.6143.70E-05
PLSExp0.00560.09326.03.40.0112.88E-04
RFExp0.00090.02843.34.50.3135.75E-06
BPExp0.00160.03364.85.80.6676.31E-06
BRealExp0.00140.06202.310.70.9132.25E-05
PLSExp0.17000.495034.362.70.9782.39E-04
RFExp0.01200.039530.440.40.9881.30E-06
BPExp0.00160.04923.311.20.8632.40E-05
CRealExp0.00030.07600.43.30.0091.97E-04
PLSExp0.00700.13005.46.80.7499.45E-05
RFExp0.00140.02825.03.60.0133.04E-05
BPExp0.00060.02132.74.40.0661.26E-05
(Herbaceous) Plots/KrigingModelsC0C0+CC0/(C0+C)
(%)		Range
(m)		R2RSS
ARealExp0.00670.050013.410.20.9041.21E-05
PLSExp0.00400.04159.65.50.6018.37E-06
RFExp0.00230.016913.79.10.8951.19E-06
BPExp0.00350.025114.17.10.8362.25E-06
BRealExp0.00790.10087.86.50.9891.90E-06
PLSExp0.08600.591014.616.60.9971.10E-04
RFExp0.00930.072312.99.40.9853.46E-06
BPExp0.00530.039713.410.70.9771.99E-06
CRealExp0.00040.06810.64.70.2697.42E-05
PLSExp0.02400.44605.44.40.4271.02E-03
RFExp0.00140.03793.75.00.3861.88E-05
BPExp0.00000.01470.14.30.1674.56E-06

Statistical parameters of the ecosystem C cycle function.

Table 5

(Woody)
Plots/Kriging		ModelsC0C0+CC0/(C0+C)
(%)		Range
(m)		R2RSS
ARealExp0.00170.016210.45.70.7597.11E-07
PLSExp0.00260.11122.38.70.5053.59E-04
RFExp0.00060.00599.35.60.6591.37E-07
BPExp0.00400.035311.55.80.7982.96E-06
BRealGau0.00070.004416.310.10.9364.41E-08
PLSGau0.08510.225237.841.70.9981.71E-05
RFGau0.00030.001916.112.70.9631.16E-08
BPGau0.00400.029413.610.10.9681.01E-06
CRealExp0.00100.008711.66.40.8072.24E-07
PLSExp0.00280.12362.30.30.0011.23E-04
RFExp0.00050.003314.09.00.8894.45E-08
BPExp0.00580.019611.76.50.8007.89E-06
(Herbaceous) Plots/KrigingModelsC0C0+CC0/(C0+C)
(%)		Range
(m)		R2RSS
ARealExp0.00160.015710.25.90.9082.63E-07
PLSExp0.00970.12247.92.50.0928.77E-06
RFExp0.00040.00597.06.00.8011.03E-07
BPExp0.00380.03869.86.30.9022.04E-06
BRealGau0.00050.004711.46.50.1181.30E-07
PLSGau0.00740.073710.09.00.7523.28E-05
RFGau0.00020.001510.57.40.2981.78E-08
BPGau0.00330.03549.35.70.0304.72E-06
CRealGau0.00080.007011.66.30.0992.03E-07
PLSGau0.00450.042510.67.20.4705.70E-06
RFGau0.00030.002211.07.20.5739.15E-09
BPGau0.00310.03638.51.80.0015.45E-06

Statistical parameters of the ecosystem N cycle function.

Table 6

(Woody)
Plots/Kriging		ModelsC0C0+CC0/(C0+C)
(%)		Range
(m)		R2RSS
ARealGau0.00060.00806.81.80.0012.38E-07
PLSGau0.01590.15989.91.90.0011.49E-04
RFGau0.00010.00284.81.90.0012.42E-08
BPGau0.00230.03107.51.80.0015.63E-06
BRealGau0.01070.026340.530.30.9981.98E-07
PLSGau0.12850.347037.045.80.9881.76E-04
RFGau0.00410.013031.229.50.9991.01E-08
BPGau0.02590.060642.729.20.9988.83E-07
CRealGau0.00090.02213.91.90.0013.89E-06
PLSGau0.00710.10626.77.80.2721.77E-04
RFGau0.00040.00814.38.30.3621.31E-06
BPGau0.00150.02925.21.90.0015.31E-06
(Herbaceous) Plots/KrigingModelsC0C0+CC0/(C0+C)
(%)		Range
(m)		R2RSS
ARealGau0.00060.00648.95.50.0021.09E-06
PLSGau0.01070.0713159.80.8372.80E-05
RFGau0.00020.002210.46.10.0131.20E-07
BPGau0.00150.02555.81.90.0011.77E-05
BRealExp0.00530.026719.716.40.9562.66E-06
PLSExp0.01940.124815.516.00.9921.53E-05
RFExp0.00070.00858.712.90.9364.41E-07
BPExp0.00620.042714.511.00.8941.25E-05
CRealGau0.00280.025111.210.30.8235.93E-06
PLSGau0.01020.12947.97.90.4861.17E-04
RFGau0.00110.008312.810.20.8026.61E-07
BPGau0.00750.055513.510.70.9231.26E-05

Statistical parameters of the ecosystem P cycle function.

Figure 4

The OK results showed that the ecosystem function C, N and P cycles from river bank to desert margin (Plot A-C). The C cycle of woody plants firstly decreased and then increased with gradients (Figure 5), while the carbon cycle of herbaceous plants was the opposite. The N cycles in both woody and herbaceous plants were weakened along the gradients (Figure 6). The P cycle of herbaceous plants was continuously decreased while the woody plants were continuously increased (Figure 7).

Figure 5

Figure 6

Figure 7

3.4 Regression kriging prediction

The RK visualization results based on wMDS are shown in (Figures 5, 6, 7A). The visualization results based on raw and predicted values all showed that the prediction accuracy of the BPNNK model was the highest, followed by RFK and PLSK models. The OK and RK visualization results of P cycling in plot B all showed “bull’s eye phenomenon”. The visualization results by hMDS (Figures 5, 6, 7B) based on raw and predicted values all showed that the prediction accuracy of the BPNNK model was similar to that of the RFK model, and the prediction accuracy of the PLSK model was the lowest. Furthermore, local extremum or over smoothing (lack of details) were shown in the P cycling in plot A, the C cycling in plot B, and the N cycling in plot C. Therefore, RF and BPNN are better than PLS in RK prediction. The accuracy test results also showed that RF and BPNN were better than PLS (Figure 4). Moreover, the RK visualization results based on wMDS were better than those based on hMDS (Figures 5–7).

4 Discussion

4.1 Minimum data set of functional traits

In this study, plant functional traits were used as predictor variables, and hMDS and wMDS were constructed by selecting important plant functional traits to predict ecosystem functions. According to the study results, morphological traits including SLA, DBH, PH, LW, and LT, and physiological traits including LWC and LDW were included in the wMDS. Previous studies have shown that plant leaf morphological traits could determine plant photosynthetic capacity (; Yu et al., 2014; Liu and Liang, 2016). Some scholars also reported that H and SBD (DBH) could reflect the adaptability of plants to environmental changes and their ability to acquire resources, and the length and thickness of the stems of woody plants were more sensitive to environmental filtering (; Liu et al., 2015). In this study, in desert ecosystems, water is the most important environmental factor affecting plant distribution and growth. LWC can reflect the water status of plant tissues (Li et al., 2013) and the resistance and adaptation of plants to drought stress (Joly et al., 2017). With the increase of drought degree, soil water content and LWC of plants decreased greatly (Maréchaux et al., 2015; Zhou et al., 2021). Furthermore, LFW can reflect the dehydration resistance of plants. The higher the LWC, the higher the LFW, and the stronger the drought resistance (Zou et al., 2019). Therefore, through the construction of hMDS and wTDS, the representative functional traits can be screened out. The results of this study showed that desert plants adapt to the arid environment mainly by the changes of morphological and physiological traits. Furthermore, by comparing the components in hMDS with those in the wMDS, it was found that woody plants invest more on stem.

4.2 Prediction of spatial distribution characteristics of ecosystem functions by functional trait MDS

Spatial heterogeneity is jointly affected by random and structural factors. In geostatistics, a high nugget-sill ratio (C0/(C0+ C) > 50%) indicates a high degree of spatial heterogeneity caused by the random factors. If the ratio is close to 100%, it means that the object variables have constant variation, i.e., the spatial heterogeneity comes from random factors (; Shi et al., 2020). According to the results of this study, the nugget-sill ratios of the raw and predicted values of C, N, and P cycling for the three plots were less than 25%, the functions in each plot had strong spatial autocorrelation, and the spatial variation was mainly influenced by structural factors. This indicates that natural factors (structural factors) such as topography, parent material, and vegetation play important roles in the spatial variation. The study area is in a national nature reserve, where human interference is minimal. Within each plot, soil type, relief, light radiation, temperature, and other conditions are nearly uniform, so structural factors may come from vegetation type and functional traits of plants. Furthermore, a small proportion of random factors may be caused by experimental errors such as sampling.

Plant functional traits are biological regulators of C, N, and P cycling (Wright et al., 2004; Xu et al., 2017; ). They can regulate hydrothermal and material redistribution (; Wasternack, 2017) to influence the extent and intensity of C, N, and P cycling in the ecosystem (McCormack et al., 2015). The OK model results showed that the relative contributions of woody and herbaceous plants to the ecosystem functioning C cycle were mainly determined by the plant biomass. In Plot B, the biomass of herbaceous plants dominated for more C stock, while in Plot A, woody plants were the dominant life forms, and in Plot C, woody plants were more deeply rooted and drought tolerant than herbaceous plants, so woody plants participated in carbon cycles with higher carbon stocks in Plots A and C. The contribution of both woody and herbaceous plants to the ecosystem functional N cycle decreased with gradients, potentially due to moisture limitation (). concluded that nitrogen accumulation and transport were influenced by soil moisture, and in natural communities with no anthropogenic nitrogen addition, nitrification and denitrification of plant residues would be more intense in wet areas than in arid areas, with wetter areas receiving more nitrogen accumulation (Liu et al., 2020; Zhao et al., 2021). The contribution of woody plants to the ecosystem functional C cycle and the contribution of herbaceous plants to the ecosystem functional C cycle showed overall contrary results in areas ranging from riparian forests to desert margins. Some studies suggested that in grassland ecosystems, phosphorus was slowing down grassland degradation. When herbaceous plants are moisture-limited, phosphorus stocks are also reduced (Liu et al., 2018). Woody, being perennial, retains more nutrients and reduces the phosphorus metabolic loss in extreme conditions (Zhang et al., 2012). At the small scale of desert ecosystems, ecosystem cycles were largely influenced by vegetation type and may be disturbed by animals and microorganisms in the soil (Zhao et al., 2018). Whereas at large scales (a few square kilometers or hundreds of square kilometers), it has been suggested that ecosystem functioning was affected by climate, topography or anthropogenic emissions (Marklein and Houlton, 2012).

The RK model results show that the wMDS and hMDS can improve the prediction accuracy of the spatial distribution of C, N and P cycling. It has been argued that plant community characteristics influence soil C, N, and P cycling and control the decomposition process in ecosystems. Furthermore, the abundance and composition of species or functional groups within a community influence the input and output of soil C (Oelmann et al., 2011; Zhang, 2020; Shen et al., 2021). According to the mass ratio hypothesis, ecosystem function is primarily determined by the traits of the biomass-dominant species in the community (Smith et al., 2001; Wright et al., 2003; Prentice et al., 2014). Therefore, the relative abundance and biomass of plants and their traits may be the main determinants of C, N, and P cycling in the ecosystem. It has also been shown that plant functional traits can influence C cycling in wetland ecosystems (Wang et al., 2010), plant trait combinations influence the diversity of soil decomposers through the diversity of habitat conditions they create. In turn, the diversity of decomposers may significantly affects soil C cycling (). Li and Wang (2021) argued that C and N cycling were related to plant morphological traits, and ecosystem function was strongly influenced not only by dominant species or functional group traits, but also by dominant functional traits. Meng et al. (2017) showed that C cycling was mainly influenced by vegetation type, which could explain 66.10% of the total variation. showed that the spatial distribution of C cycling was progressively enhanced by vegetation structure as farmlands were returned to forestland. Gong et al. (2017) also reported a significant positive correlation between plant leaves and soil organic matter content in their study on the C cycling. Therefore, it is reasonable that plant functional traits can predict the function of C cycling in ecosystems. Plant functional traits are the structural factors that have the strongest impact on ecosystem functions.

The results of the RK analysis also indicated that the N and P cycles were also predicted more accurately. Phenotypic traits can reflect changes in ecosystem functions, as well as changes in the spatial distribution of ecosystem processes and functions (McIntyre et al., 2009), such as SLW and SLA (). C, N, and P cycling in ecosystems are often affected by traits of multiple plant organs. For example, the C, N, and P concentrations of plant leaves have a stable positive relationship (Rawat et al., 2020). Furthermore, the maximum diameter of stems (Jucker et al., 2016) and the maximum plant height also have a stable positive relationship with the C, N, and P concentrations of plant leaves (). The accumulation and transformation of N and P between plants and soil is a complex process, which is affected by many environmental factors. Under the same climate and habitat conditions, the dynamic changes of vegetation factors and soil factors regulate the input and output of soil N and P, and then affect the accumulation of N and P in the soil (). Therefore, the functional traits of plants can directly reflect the N and P cycling in the ecosystem.

4.3 Comparison of regression kriging models

By comparing the prediction results of the PLS, RF, and BPNN models based on the Kriging method, it was found that the RF model can significantly improve the prediction accuracy on ecosystem functions based on the MDS, and it can accurately predict the spatial distribution of C, N, and P cycling in the desert ecosystem. This may be a non-linear relationship between RF analysis of multiple source auxiliary variables (MDS) and C, N, and P cycling through a classification algorithm to obtain a globally optimal solution, which can overcome the defect of the local minimum solution of the BPNN method (Tyralis et al., 2019). Song et al. (2017) compared the accuracy of Support Vector Machine (SVR), BPNN, and RF models in predicting SOM, and showed that the RF model had higher coefficient of determination and prediction accuracy. This is consistent with the results of this study. Furthermore, it was found that the RF model based on the wMDS had a better performance in predicting C, N, and P cycling than the RF model based on the hMDS, enabling global and point specific predictions. Zeng et al. (2014) also showed that the RF model could better reflect the “pure information” changes in the samples and obviously improve the prediction accuracy of the model.

The BPNN model with non-transparency of data operation has strong fault tolerance, but traditional BPNN models are also prone to over-fitting and local optimality (). The results of this study confirmed that the BPNN model was very unstable. PLS, on the other hand, had a low prediction accuracy. It may be that the algorithm has difficulty explaining the loading of independent latent variables. It is based on a cross product with the response variables, rather than on correlations between independent variables in conventional factor analysis (Lednev et al., 2018; Wang et al., 2022). Consequently, visualisation results from the non-linear RF and BPNN models combined with kriging demonstrate that the MDS of plant functional traits could be used to predict C, N, and P cycling in ecosystems.

5 Conclusion

In this study, the MDSs (hMDS and wMDS) of plant functional traits was constructed. The spatial distribution of C, N, and P cycling in the desert ecosystem in the Xinjiang Ebinur Lake Basin were accurately predicted based on the hMDS and wMDS using linear and non-linear models combined with regression kriging and geostatistical analysis. The wMDS included H, SLA, LDW, LWC, DBH, LW, and LT, and the hMDS included H, SLA, LFW, LWC, LL, and LW. The cross-validation performed in this study showed that the MDS can replace TDS in predicting ecosystem functions, and the constructed MDSs could accurately predict the spatial distribution of C, N, and P cycling in the ecosystem. Furthermore, C, N, and P cycles are strongly spatially autocorrelated due to structural factors, and C, N and P cycles in desert ecosystems do not behave uniformly between different life forms of plants, subject to water limitation. The RK predicted result was highly consistent with the distribution of the raw values.

Statements

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions

Conceptualization, YC and JW. Methodology, YC and JW. Software, YC and JW. Validation, YC and JW. Formal Analysis, GL. Investigation and Project Administration, LJ, HL, HW. Resources. All authors contributed to the article and approved the submitted version.

Funding

This study was financially supported by National Natural Science Foundation of China (No.42171026), Xinjiang Uygur Autonomous Region innovation environment Construction special project & Science and technology innovation base construction project (PT2107), and Excellent doctoral research and innovation Project of Xinjiang University (XJU2022BS060).

Acknowledgments

We greatly thank Xuemin He, Wenjing Li, Zhoukang Li, Kunduz Sattar and Shiyun Wang et al. for their strong help with field and laboratory work.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2023.1131778/full#supplementary-material

Abbreviations

LC, Leaf carbon; LN, Leaf nitrogen; LP, Leaf phosphorus; SLA, Specific leaf area; SLW, Specific leaf weight; LA, leaf area; LDMC, Leaf dry matter content; LWC, Leaf water content; LFW, Leaf fresh weight; LDW, Leaf dry weight; DBH, diameter at breast height; SBD, Stem base diameter; H, Height; LT, Leaf thickness; LL, Leaf length; LW, Leaf width; WLR, Width to length ratio; SOC, Soil organic carbon content; SAN, Soil ammonium nitrogen; SNN, Soil nitrate nitrogen; STN, Total nitrogen content; STP, Soil total phosphorus; SAP, Soil available phosphorus.

References

  • 1

    AlbertC. H.de BelloF.BoulangeatI.PelletG.LavorelS.ThuillerW. (2012). On the importance of intraspecific variability for the quantification of functional diversity. Oikos121 (1), 116–126. doi: 10.1111/j.1600-0706.2011.19672.x

  • 2

    AskariM. S.HoldenN. M. (2014). Indices for quantitative evaluation of soil quality under grassland management. Geoderma230, 131–142. doi: 10.1016/j.geoderma.2014.04.019

  • 3

    AskariM. S.HoldenN. M. (2015). Quantitative soil quality indexing of temperate arable management systems. Soil Tillage Res.150, 57–67. doi: 10.1016/j.still.2015.01.010

  • 4

    BaoS. D. (2000). Soil agrochemical analysis (Beijing: China Agricultural Press).

  • 5

    BehnamianA.MillardK.BanksS. N.WhiteL.RichardsonM.PasherJ. (2017). A systematic approach for variable selection with random forests: achieving stable variable importance values. IEEE Geosci. Remote Sens. Lett.14 (11), 1988–1992. doi: 10.1109/LGRS.2017.2745049

  • 6

    BoernerR.BrinkmanJ. A.SmithA. (2015). Seasonal variations in enzyme activity and organic carbon in soil of a burned and unburned hardwood forest. Soil Biol. Biochem.37 (8), 1419–1426. doi: 10.1016/j.soilbio.2004.12.012

  • 7

    BowkerM. A.MaestreF. T.MauR. L. (2013). Diversity and patch-size distributions of biological soil crusts regulate dryland ecosystem multifunctionality. Ecosystems16 (6), 923–933. doi: 10.1007/s10021-013-9644-5

  • 8

    BreimanL. (2001). Random forests. Mach. Learn.45 (1), 5–32. doi: 10.1023/A:1010933404324

  • 9

    ButlerE. E.DattaA.Flores-MorenoH.ChenM.ReichP. B. (2018). Mapping local and global variability in plant trait distributions. PNAS114 (51), E10937–E10946. doi: 10.1073/pnas.1708984114

  • 10

    CambardellaC. A.MoormanT. B.NovakJ. M.ParkinT. B.KonopkaA. E. (1994). Field-scale variability of soil properties in central Iowa soils. Soil Sci. Soc. America J.58 (5), 1501–1511. doi: 10.2136/sssaj1994.03615995005800050033x

  • 11

    CamposM. L.YoshidaY.MajorI. T.de Oliveira FerreiraD.WeraduwageS. M.FroehlichJ. E.et al. (2016). Rewiring of jasmonate and phytochrome b signalling uncouples plant growth-defense tradeoffs. Nat. Commun.7, 12570. doi: 10.1038/ncomms12570

  • 12

    CatorciA.CesarettiS.MalatestaL.TardellaF. M. (2014). Effects of grazing vs mowing on the functional diversity of sub-Mediterranean productive grasslands. Appl. Vegetation Sci.17 (4), 658–669. doi: 10.1111/avsc.12103

  • 13

    CavanaughK. C.GosnellJ. S.DavisS. L.AhumadaJ.BoundjaP.ClarkD. B.et al. (2014). Carbon storage in tropical forests correlates with taxonomic diversity and functional dominance on a global scale. Global Ecol. Biogeogr23 (5), 563–573. doi: 10.1111/geb.12143

  • 14

    ChenS. Y.GaoJ. S.FuH. G.ZhaoR. (2021). A selective ensemble modeling method based on accuracy and difference of game theory and its application. J. Beijing Univ. Technol.47 (1), 32–39. doi: 10.11936/bjutxb2019070012

  • 15

    ChenW.RanH.CaoX.WangJ.TengD.ChenJ.et al. (2020). Estimating PM2.5 with high-resolution 1-km AOD data and an improved machine learning model over shenzhen, China. Sci. Total Environ.746, 141093. doi: 10.1016/j.scitotenv.2020.141093

  • 16

    ChenX.SearleE.ChenC.ReichP. (2021). Negative to positive shifts in diversity effects on soil nitrogen over time. Nat. Sustainability4 (3), 225–232. doi: 10.1038/s41893-020-00641-y

  • 17

    CornelissenJ. H. C.LavorelS.GarnierE.DiazS.BuchmannN.GurvichD. E.et al. (2003). A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Aust. J. Bot.51 (4), 335–380. doi: 10.1071/Bt02124

  • 18

    DiazS.KattgeJ.CornelissenJ. H.WrightI. J.LavorelS.DrayS.et al. (2016). The global spectrum of plant form and function. Nature529 (7585), 167–171. doi: 10.1038/nature16489

  • 19

    DiazS.LavorelS.McIntyreS.FalczukV.CasanovesF.MilchunasD. G.et al. (2007). Plant trait responses to grazing - a global synthesis. Global Change Biol.13 (2), 313–341. doi: 10.1111/j.1365-2486.2006.01288.x

  • 20

    DijkstraF. A.PendallE.MorganJ. A.BlumenthalD. M.CarrilloY.LeCainD. R.et al. (2012). Climate change alters stoichiometry of phosphorus and nitrogen in a semiarid grassland. New Phytol.196 (3), 807–815. doi: 10.1111/j.1469-8137.2012.04349.x

  • 21

    DuanY.ChenL.LiY. M.WangQ. Y.ZhangC. Z.MaD. H.et al. (2021). N, p and straw return influence the accrual of organic carbon fractions and microbial traits in a mollisol. Geoderma403, 115373. doi: 10.1016/j.geoderma.2021.115373

  • 22

    DuanY. F.WangK. L.FengD.WuM.ZhangW.ChenH. S. (2018). Spatial pattern changes of soil organic carbon and total nitrogen and their responses to the conversion of cropland to forest and grassland in a typical karst watershed. Acta Ecologica Sin.38 (5), 1560–1568. doi: 10.5846/stxb201701220184

  • 23

    DuranS. M.MartinR. E.DiazS.MaitnerB. S.MalhiY.SalinasN.et al. (2019). Informing trait-based ecology by assessing remotely sensed functional diversity across a broad tropical temperature gradient. Sci. Adv.5 (12), eaaw8114. doi: 10.1126/sciadv.aaw8114

  • 24

    EdmondsonJ.TerribileE.CarrollJ. A.PriceE. A. C.CapornS. J. M. (2013). The legacy of nitrogen pollution in heather moorlands: ecosystem response to simulated decline in nitrogen deposition over seven years. Sci. Total Environ.444, 138–144. doi: 10.1016/j.scitotenv.2012.11.074

  • 25

    ElserJ. J.FaganW. F.KerkhoffA. J.SwensonN. G.EnquistB. J. (2010). Biological stoichiometry of plant production: metabolism, scaling and ecological response to global change. New Phytol.186 (3), 593–608. doi: 10.1111/j.1469-8137.2010.03214.x

  • 26

    ElserJ. J.SternerR. W.GorokhovaE.FaganW. F.MarkowT. A.CotnerJ. B.et al. (2000). Biological stoichiometry from genes to ecosystems. Ecol. Lett.3 (6), 540–550. doi: 10.1046/j.1461-0248.2000.00185.x

  • 27

    FalsterD. S.BrannstromA.DieckmannU.WestobyM. (2011). Influence of four major plant traits on average height, leaf-area cover, net primary productivity, and biomass density in single-species forests: a theoretical investigation. J. Ecol.99 (1), 148–164. doi: 10.1111/j.1365-2745.2010.01735.x

  • 28

    FengS.FuQ. (2013). Expansion of global drylands under a warming climate. Atmospheric Chem. Phys.13 (19), 10081–10094. doi: 10.5194/acp-13-10081-2013

  • 29

    FernándezP. L.PablosF.Martín.M. J.GonzálezA. G. (2002). Multi-element analysis of tea beverages by inductively coupled plasma atomic emission spectrometry. Food Chem.76 (4), 483–489. doi: 10.1016/S0308-8146(01)00312-0

  • 30

    FineganB.PeM.De OliveiraA.AscarrunzN.Bret-HarteM. S.CasanovesF.et al. (2015). Does functional trait diversity predict above-ground biomass and productivity of tropical forests? Testing three Altern. hypotheses. J. Ecol.103 (1), 191–201. doi: 10.1111/1365-2745.12346

  • 31

    FlynnD. F. B.MirotchnickN.JainM.PalmerM. I.NaeemS. (2011). Functional and phylogenetic diversity as predictors of biodiversity-ecosystem-function relationships. Ecology92 (8), 1573–1581. doi: 10.1890/10-1245.1

  • 32

    FortunelC.GarnierE.JoffreR.KazakouE.QuestedH.GrigulisK.et al. (2009). Leaf traits capture the effects of land use changes and climate on litter decomposability of grasslands across Europe. Ecology90 (3), 598–611. doi: 10.1890/08-0418.1

  • 33

    GaoX. J.GiorgiF. (2008). Increased aridity in the Mediterranean region under greenhouse gas forcing estimated from high resolution simulations with a regional climate model. Global Planetary Change62 (3-4), 195–209. doi: 10.1016/j.gloplacha.2008.02.002

  • 34

    GarlandG.BanerjeeS.EdlingerA.Miranda OliveiraE.HerzogC.WittwerR.et al. (2020). A closer look at the functions behind ecosystem multifunctionality: a review. J. Ecol.109 (2), 600–613. doi: 10.1111/1365-2745.13511

  • 35

    GeX.DingJ.WangJ.FeiW.SunH. (2018). Estimation of soil moisture content based on competitive adaptive reweighted sampling algorithm coupled with machine learning. Acta Optica Sin.38 (10), 1030001. doi: 10.3788/aos201838.1030001

  • 36

    GieseM.BrueckH.GaoY. Z.LinS.SteffensM.Kogel-KnabnerI.et al. (2013). N balance and cycling of inner Mongolia typical steppe: a comprehensive case study of grazing effects. Ecol. Monogr.83 (2), 195–219. doi: 10.1890/12-0114.1

  • 37

    GongZ. N.LIH.ChengQ. W. (2017). Spatial distribution of soil organic matter in water-level-fluctuation zone of guanting reservoir. Acta Ecologica Sin.37 (24), 8336–8347. doi: 10.5846/stxb201611182344

  • 38

    GrimeJ. P. (1998). Benefits of plant diversity to ecosystems: immediate, filter and founder effects. J. Ecol.86 (6), 902–910. doi: 10.1046/j.1365-2745.1998.00306.x

  • 39

    GuoL.SunX.FuP.ShiT.DangL.ChenY.et al. (2021). Mapping soil organic carbon stock by hyperspectral and time-series multispectral remote sensing images in low-relief agricultural areas. Geoderma398, 115118. doi: 10.1016/j.geoderma.2021.115118

  • 40

    GuoL. L.SunZ. G.OuyangZ.HanD. R.LiF. D. (2017). A comparison of soil quality evaluation methods for fluvisol along the lower yellow river. Catena152, 135–143. doi: 10.1016/j.catena.2017.01.015

  • 41

    GuoP. T.WuW.ShengQ. K.LiM. F.LiuH. B.WangZ. Y. (2013). Prediction of soil organic matter using artificial neural network and topographic indicators in hilly areas. Nutr. Cycl Agroecosys95 (3), 333–344. doi: 10.1007/s10705-013-9566-9

  • 42

    HeM.DijkstraF. A.ZhangK.LiX.TanH.GaoY.et al. (2014). Leaf nitrogen and phosphorus of temperate desert plants in response to climate and soil nutrient availability. Sci. Rep.4, 6932. doi: 10.1038/srep06932

  • 43

    HeN. P.LiC. C.ZhangJ. H.XuL.YuG. R. (2018). Opportunities and challenges in plant trait research: from organs to communities. Acta Ecologica Sin.38 (19), 6787–6796. doi: 10.5846/stxb201710241900

  • 44

    HenglT.HeuvelinkG. B. M.RossiterD. G. (2007). About regression-kriging: from equations to case studies. Comput. Geosciences33 (10), 1301–1315. doi: 10.1016/j.cageo.2007.05.001

  • 45

    HenglT.HeuvelinkG. B. M.SteinA. (2004). A generic framework for spatial prediction of soil variables based on regression-kriging. Geoderma120 (1-2), 75–93. doi: 10.1016/j.geoderma.2003.08.018

  • 46

    HuangT.LiaoX. Y.ZhaoD.GongX. G.CassidyD. P. (2019a). Delineation of soil contaminant plumes at a co-contaminated site using BP neural networks and geostatistics. Geoderma354, 113878. doi: 10.1016/j.geoderma.2019.07.036

  • 47

    HuangX. Q.ZhaoJ.ZhouX.ZhangJ. B.CaiZ. C. (2019b). Differential responses of soil bacterial community and functional diversity to reductive soil disinfestation and chemical soil disinfestation. Geoderma348, 124–134. doi: 10.1016/j.geoderma.2019.04.027

  • 48

    IsbellF.CalcagnoV.HectorA.ConnollyJ.HarpoleW. S.ReichP. B.et al. (2011). High plant diversity is needed to maintain ecosystem services. Nature477 (7363), 199–U196. doi: 10.1038/nature10282

  • 49

    JinH. F.ShiD. M.ChenZ. F.LiuY. J.YangX. (2018). Evaluation indicators of cultivated layer soil quality for red soil slope farmland based on cluster and PCA analysis. Trans. Chin. Soc. Agric. Eng.34 (7), 155–164. doi: 10.11975/j.issn.1002-6819.2018.07.020

  • 50

    JolyF. X.KurupasK. L.ThroopH. L. (2017). Pulse frequency and soil-litter mixing alter the control of cumulative precipitation over litter decomposition. Ecology98 (9), 2255–2260. doi: 10.1002/ecy.1931

  • 51

    JuckerT.SanchezA. C.LindsellJ. A.AllenH. D.AmableG. S.CoomesD. A. (2016). Drivers of aboveground wood production in a lowland tropical forest of West Africa: teasing apart the roles of tree density, tree diversity, soil phosphorus, and historical logging. Ecol. Evol.6 (12), 4004–4017. doi: 10.1002/ece3.2175

  • 52

    KattgeJ.DiazS.LavorelS.PrenticeC.LeadleyP.BonischG.et al. (2011). TRY - a global database of plant traits. Global Change Biol.17 (9), 2905–2935. doi: 10.1111/j.1365-2486.2011.02451.x

  • 53

    KearneyW. S.FagherazziS. (2016). Salt marsh vegetation promotes efficient tidal channel networks. Nat. Commun.7, 12287. doi: 10.1038/ncomms12287

  • 54

    KnadelM.RehmanH. U.PouladiN.de JongeL. W.MoldrupP.ArthurE. (2021). Estimating atterberg limits of soils from reflectance spectroscopy and pedotransfer functions. Geoderma402 (1-2), 115300. doi: 10.1016/j.geoderma.2021.115300

  • 55

    KraftN. J.GodoyO.LevineJ. M. (2015). Plant functional traits and the multidimensional nature of species coexistence. P Natl. Acad. Sci. U.S.A.112 (3), 797–802. doi: 10.1073/pnas.1413650112

  • 56

    LaliberteE.LegendreP. (2010). A distance-based framework for measuring functional diversity from multiple traits. Ecology91 (1), 299–305. doi: 10.1890/08-2244.1

  • 57

    LednevV.TretyakovR.SdvizhenskiiP.GrishinM.AsyutinR.PershinS. (2018). Laser induced breakdown spectroscopy for in-situ multielemental analysis during additive manufacturing process. J. Physics: Conf. Ser.1109, 12050. doi: 10.1088/1742-6596/1109/1/012050

  • 58

    LiS. F.HuangX. B.LangX. D.ShenJ. Y.XuF. D.SuJ. R. (2020b). Cumulative effects of multiple biodiversity attributes and abiotic factors on ecosystem multifunctionality in the jinsha river valley of southwestern China. For. Ecol. Manage.472, 118281. doi: 10.1016/j.foreco.2020.118281

  • 59

    LiT.ShenH.ZengC.YuanQ.ZhangL. (2017). Point-surface fusion of station measurements and satellite observations for mapping PM2.5 distribution in China: methods and assessment. Atmospheric Environ.152, 477–489. doi: 10.1016/j.atmosenv.2017.01.004

  • 60

    LiJ. S.SuP. X.ZhangH. N.ZhouZ. J.XieT. T. (2013). Leaf water and functional traits of desert plants and their relationship. Plant Physiol. J.49 (2), 153–160. doi: 10.13592/j.cnki.ppj.2013.02.011

  • 61

    LiY. Q.WangZ. H. (2021). Ecological function, geographic distribution and genesis of plant leaf morphology. Chin. J. Plant Ecol.45 (10), 1154–1172. doi: 10.17521/cjpe.2020.0405

  • 62

    LiP.WuM. C.KangG. D.ZhuB. J.LiH. X.HuF.et al. (2020a). Soil quality response to organic amendments on dryland red soil in subtropical China. Geoderma373, 114416. doi: 10.1016/j.geoderma.2020.114416

  • 63

    LiG. Y.YangD. M.SunS. C. (2008). Allometric relationships between lamina area, lamina mass and petiole mass of 93 temperate woody species vary with leaf habit, leaf form and altitude. Funct. Ecol.22 (4), 557–564. doi: 10.1111/j.1365-2435.2008.01407.x

  • 64

    LiangM.LiuX.ParkerI. M.JohnsonD.ZhengY.LuoS.et al. (2019). Soil microbes drive phylogenetic diversity-productivity relationships in a subtropical forest. Sci. Adv.5 (10), eaax5088. doi: 10.1126/sciadv.aax5088

  • 65

    LinY.HongM.HanG. D.ZhaoM. L.BaiY. F.ChangS. X. (2010). Grazing intensity affected spatial patterns of vegetation and soil fertility in a desert steppe. Agr Ecosyst. Environ.138 (3-4), 282–292. doi: 10.1016/j.agee.2010.05.013

  • 66

    LinJ. N.LinD. M.ZhuG. Y.WangH. J.QianS. H.ZhaoL.et al. (2022). Earthworms exert long lasting afterlife effects on soil microbial communities. Geoderma420, 115906. doi: 10.1016/j.geoderma.2022.115906

  • 67

    LiuY. R.Delgado-BaquerizoM.TrivediP.HeJ. Z.WangJ. T.SinghB. K. (2017). Identity of biocrust species and microbial communities drive the response of soil multifunctionality to simulated global change. Soil Biol. Biochem.107, 208–217. doi: 10.1016/j.soilbio.2016.12.003

  • 68

    LiuZ. M.GuoS. J.QinT. T.SunX. B. (2015). Growth response of 1a castanea mollissima seedlings to compost of forestry and agricultural residues. J. Cent. South Univ. Forestry Technol.35 (10), 62–68. doi: 10.14067/j.cnki.1673-923x.2015.10.011

  • 69

    LiuM. X.LiangG. L. (2016). Advances in studies on plant specific leaf quality. Chin. J. Plant Ecol.40 (8), 847–860. doi: 10.17521/cjpe.2015.0428

  • 70

    LiuY.TenzintarchenGengX.WeiD.DaiD.Xu-Ri (2020). Grazing exclusion enhanced net ecosystem carbon uptake but decreased plant nutrient content in an alpine steppe. Catena195, 104799. doi: 10.1016/j.catena.2020.104799

  • 71

    LiuS. B.ZamanianK.SchleussP. M.ZarebanadkoukiM.KuzyakovY. (2018). Degradation of Tibetan grasslands: consequences for carbon and nutrient cycles. Agr Ecosyst. Environ.252, 93–104. doi: 10.1016/j.agee.2017.10.011

  • 72

    LohbeckM.PoorterL.Martinez-RamosM.BongersF. (2015). Biomass is the main driver of changes in ecosystem process rates during tropical forest succession. Ecology96 (5), 1242–1252. doi: 10.1890/14-0472.1

  • 73

    MaX. L.MahechaM. D.MigliavaccaM.van der PlasF.BenavidesR.RatcliffeS.et al. (2019). Inferring plant functional diversity from space: the potential of sentinel-2. Remote Sens. Environ.233, 111368. doi: 10.1016/j.rse.2019.111368

  • 74

    MaestreF. T.QueroJ. L.GotelliN. J.EscuderoA.OchoaV.Delgado-BaquerizoM.et al. (2012). Plant species richness and ecosystem multifunctionality in global drylands. Science335 (6065), 214–218. doi: 10.1126/science.1215442

  • 75

    MaréchauxI.BartlettM. K.SackL.BaralotoC.ChaveJ. (2015). Drought tolerance as predicted by leaf water potential at turgor loss point varies strongly across species within an Amazonian forest. Funct. Ecol.29 (10), 1268–1277. doi: 10.1111/1365-2435.12452

  • 76

    MarkleinA. R.HoultonB. Z. (2012). Nitrogen inputs accelerate phosphorus cycling rates across a wide variety of terrestrial ecosystems. New Phytol.193, 696–704. doi: 10.1111/j.1469-8137.2011.03967.x

  • 77

    McCormackM. L.DickieI. A.EissenstatD. M.FaheyT. J.FernandezC. W.GuoD.et al. (2015). Redefining fine roots improves understanding of below-ground contributions to terrestrial biosphere processes. New Phytol.207 (3), 505–518. doi: 10.1111/nph.13363

  • 78

    McIntyreS.LavorelS.LandsbergJ.ForbesT. D. A. (2009). Disturbance response in vegetation - towards a global perspective on functional traits. J. Vegetation Sci.10 (5), 621–630. doi: 10.2307/3237077

  • 79

    MeiS.TianK.YunZ.HangW.GuanD. X.YueH. T. (2017). Research on leaf functional traits and their environmental adaptation. Plant Sci. J.35 (6), 940–949. doi: 10.11913/PSJ.2095-0837.2017.60940

  • 80

    MengQ. M.LiuZ. J. (2013). Assessment of regression kriging for spatial interpolation – comparisons of seven GIS interpolation methods. Cartography Geographic Inf. Sci.40 (1), 28–39. doi: 10.1080/15230406.2013.762138

  • 81

    MengG. X.ZhaT. G.ZhangX. X.ZhangZ. Q.ZhuY. S.ZhouY.et al. (2017). Effects of vegetation types and topography on vertical distribution of soil organic carbon in reclaimed farmland in loess region. Chin. J. Ecol.36 (9), 2447–2454. doi: 10.13292/j.1000-4890.201709.018

  • 82

    MessierJ.McGillB. J.LechowiczM. J. (2010). How do traits vary across ecological scales? A Case trait-based ecology. Ecol. Lett.13 (7), 838–848. doi: 10.1111/j.1461-0248.2010.01476.x

  • 83

    MoraisT. G.TufikC.RatoA. E.RodriguesN. R.GamaI.JongenM.et al. (2021). Estimating soil organic carbon of sown biodiverse permanent pastures in Portugal using near infrared spectral data and artificial neural networks. Geoderma404 (8), 115387. doi: 10.1016/j.geoderma.2021.115387

  • 84

    MukherjeeS.JoshiP. K.GargR. D. (2015). Regression-kriging technique to downscale satellite-derived land surface temperature in heterogeneous agricultural landscape. IEEE J. Selected Topics Appl. Earth Observations Remote Sens.8 (3), 1245–1250. doi: 10.1109/JSTARS.2015.2396032

  • 85

    NiY. M.OuYangZ. Y. (2006). Distribution characteristics and succession trend of desert ecosystem in xinjiang. J. Arid Land Resour. Environ.20 (2), 7–10. doi: 10.3969/j.issn.1003-7578.2006.02.002

  • 86

    NiuK. C.ZhangS. T.ZhaoB. B.DuG. Z. (2010). Linking grazing response of species abundance to functional traits in the Tibetan alpine meadow. Plant Soil330 (1-2), 215–223. doi: 10.1007/s11104-009-0194-8

  • 87

    OelmannY.RichterA. K.RoscherC.RosenkranzS.TempertonV. M.WeisserW. W.et al. (2011). Does plant diversity influence phosphorus cycling in experimental grasslands? Geoderma167-168, 178–187. doi: 10.1016/j.geoderma.2011.09.012

  • 88

    PakemanR. J.FieldingD. A. (2020). A functional assessment of the impact of changing grazing management of upland grassland mosaics. Appl. Vegetation Sci.23 (4), 539–550. doi: 10.1111/avsc.12504

  • 89

    PasariJ. R.LeviT.ZavaletaE. S.TilmanD. (2013). Correction for pasari et al., several scales of biodiversity affect ecosystem multifunctionality. Proc. Natl. Acad. Sci.110 (37), 15163–15163. doi: 10.1073/pnas.1314558110

  • 90

    PetcheyO. L.GastonK. J. (2002). Functional diversity (FD), species richness and community composition. Ecol. Lett.5 (3), 402–411. doi: 10.1046/j.1461-0248.2002.00339.x

  • 91

    PhamT. G.KappasM.Van HuynhC.NguyenL. H. K. (2019). Application of ordinary kriging and regression kriging method for soil properties mapping in hilly region of central Vietnam. Isprs Int. J. Geo-Information8 (3), 147–165. doi: 10.3390/ijgi8030147

  • 92

    PoggioL.LassauceA.GimonaA. (2019). Modelling the extent of northern peat soil and its uncertainty with sentinel: Scotland as example of highly cloudy region. Geoderma346, 63–74. doi: 10.1016/j.geoderma.2019.03.017

  • 93

    PrenticeI. C.DongN.GleasonS. M.MaireV.WrightI. J. (2014). Balancing the costs of carbon gain and water transport: testing a new theoretical framework for plant functional ecology. Ecol. Lett.17 (1), 82–91. doi: 10.1111/ele.12211

  • 94

    PulidoM.SchnabelS.ContadorJ. F. L.Lozano-ParraJ.Gomez-GutierrezA. (2017). Selecting indicators for assessing soil quality and degradation in rangelands of extremadura (SW Spain). Ecol. Indic74 (mar.), 49–61. doi: 10.1016/j.ecolind.2016.11.016

  • 95

    QiuY.FuB.WangJ.ChenL.MengQ.ZhangY. (2010). Spatial prediction of soil moisture content using multiple-linear regressions in a gully catchment of the loess plateau, China. J. Arid Environments74 (2), 208–220. doi: 10.1016/j.jaridenv.2009.08.003

  • 96

    RaiesiF. (2017). A minimum data set and soil quality index to quantify the effect of land use conversion on soil quality and degradation in native rangelands of upland arid and semiarid regions. Ecol. Indic75, 307–320. doi: 10.1016/j.ecolind.2016.12.049

  • 97

    RawatM.ArunachalamK.ArunachalamA.AlataloJ. M.KumarU.SimonB.et al. (2020). Relative contribution of plant traits and soil properties to the functioning of a temperate forest ecosystem in the Indian Himalayas. Catena194, 104671. doi: 10.1016/j.catena.2020.104671

  • 98

    R Development Core Team (2021). R: a language and environment for statistical computing (Vienna, Austria: R Foundation for Statistical Computing.).

  • 99

    RenS. Y. (2021). Responses of plant diversity-productivity relationship to canopy gap disturbance in tiantong evergreen broad-leaved forest (Shanghai: East China Normal University).

  • 100

    ReynoldsJ. F.SmithD. M.LambinE. F.TurnerB. L.2ndMortimoreM.BatterburyS. P.et al. (2007). Global desertification: building a science for dryland development. Science316 (5826), 847–851. doi: 10.1126/science.1131634

  • 101

    RobertsonG. P.CrumJ. R.EllisB. G. (1993). The spatial variability of soil resources following long-term disturbance. Oecologia96 (4), 451–456. doi: 10.1007/BF00320501

  • 102

    SarmadianF.KeshavarziA.RooienA.IqbalM.JavadikiaH. (2014). Digital mapping of soil phosphorus using multivariate geostatistics and topographic information. Aust. J. Crop Sci.8 (8), 1216–1223.

  • 103

    ShenJ.TaoQ.DongQ.LuoY. L.LuoJ. P.HeY. T.et al. (2021). Long-term conversion from rice-wheat to rice-vegetable rotations drives variation in soil microbial communities and shifts in nitrogen-cycling through soil profiles. Geoderma404, 115299. doi: 10.1016/j.geoderma.2021.115299

  • 104

    ShiB. K.HuG.HenryH. A. L.StoverH. J.SunW.XuW. L.et al. (2020). Temporal changes in the spatial variability of soil respiration in a meadow steppe: the role of abiotic and biotic factors. Agr For. Meteorol287, 107958. doi: 10.1016/j.agrformet.2020.107958

  • 105

    SmithB.PrenticeI. C.SykesM. T. (2001). Representation of vegetation dynamics in the modelling of terrestrial ecosystems: comparing two contrasting approaches within European climate space. Global Ecol. Biogeogr10 (6), 621–637. doi: 10.1046/j.1466-822X.2001.00256.x

  • 106

    SongY. Q.YangL. A.LiB.HuY. M.WangA. L.ZhouW.et al. (2017). Spatial prediction of soil organic matter using a hybrid geostatistical model of an extreme learning machine and ordinary kriging. Sustainability9 (5), 754. doi: 10.3390/su9050754

  • 107

    SunJ.ZhouT. C.LiuM.ChenY. C.LiuG. H.XuM.et al. (2020). Water and heat availability are drivers of the aboveground plant carbon accumulation rate in alpine grasslands on the Tibetan plateau. Global Ecol. Biogeogr29 (1), 50–64. doi: 10.1111/geb.13006

  • 108

    TanC. X.ZhangZ. T.XuZ. H.MaY.YaoZ. H.WeiG. F.et al. (2020). Unmanned aerial vehicle (UAV) multi-spectral remote sensing retrieval of soil moisture content in maize root region at various growth stages. Trans. Chin. Soc. Agric. Eng.36 (10), 63–74. doi: 10.11975/j.issn.1002-6819.2020.10.008

  • 109

    TemmermanS.BoumaT. J.GoversG.WangZ. B.De VriesM. B.HermanP. M. J. (2005). Impact of vegetation on flow routing and sedimentation patterns: three-dimensional modeling for a tidal marsh. J. Geophysical Research-Earth Surface110 (F4), 2169–9003. doi: 10.1029/2005jf000301

  • 110

    TyralisH.PapacharalampousG.LangousisA. (2019). A brief review of random forests for water scientists and practitioners and their recent history in water resources. Water-Sui11 (5), 910. doi: 10.3390/w11050910

  • 111

    Vilchez-MendozaS.Romero-GurdiánA.AvelinoJ.DeClerckF.BommelP.BetbederJ.et al. (2022). Assessing the joint effects of landscape, farm features and crop management practices on berry damage in coffee plantations. Agriculture Ecosyst. Environ.330, 107903. doi: 10.1016/j.agee.2022.107903

  • 112

    VileD.ShipleyB.GarnierE. (2006). Ecosystem productivity can be predicted from potential relative growth rate and species abundance. Ecol. Lett.9 (9), 1061–1067. doi: 10.1111/j.1461-0248.2006.00958.x

  • 113

    ViolleC.NavasM. L.VileD.KazakouE.FortunelC.HummelI.et al. (2007). Let the concept of trait be functional! Oikos116 (5), 882–892. doi: 10.1111/j.0030-1299.2007.15559.x

  • 114

    WangX. Y.LiY. G.GongX. W.NiuY. Y.ChenY. P.ShiX. P.et al. (2019). Storage, pattern and driving factors of soil organic carbon in an ecologically fragile zone of northern China. Geoderma343, 155–165. doi: 10.1016/j.geoderma.2019.02.030

  • 115

    WangM.LiuH.LennartzB. (2021b). Small-scale spatial variability of hydro-physical properties of natural and degraded peat soils. Geoderma399, 115123. doi: 10.1016/j.geoderma.2021.115123

  • 116

    WangP.ShengL. S.YanH.ZhouD. Y.SongY. T. (2010). Plant functional traits and soil carbon sequestration in wetland ecosystem. Acta Ecologica Sin.30 (24), 6990–7000.

  • 117

    WangJ.TengD.HeX.QinL.YangX.LvG. (2021a). Spatial non-stationarity effects of driving factors on soil respiration in an arid desert region. Catena207, 105617. doi: 10.1016/j.catena.2021.105617

  • 118

    WangK. Y.YangS.GuoC. Y.BianX. H. (2022). Research on spectral variable selection method based on LASSO algorithm. J. Analytical Testing41 (3), 398–402+408. doi: 10.19969/j.fxcsxb.21070503

  • 119

    WangS. Q.YuG. R. (2008). Ecological stoichiometric characteristics of carbon, nitrogen and phosphorus in an ecosystem. Acta Ecologica Sin.28 (8), 3937–3947. doi: 10.3321/j.issn:1000-0933.2008.08.054

  • 120

    WasternackC. (2017). A plant’s balance of growth and defense - revisited. New Phytol.215 (4), 1291–1294. doi: 10.1111/nph.14720

  • 121

    WeiX.MaZ.XinJ.HeJ. S.EcologyD. O. (2016a). Biodiversity and ecosystem multifunctionality: advances and perspectives. Biodiversity Sci.24 (1), 55. doi: 10.17520/biods.2015091

  • 122

    WeiX.XinJ.MaZ.HeJ. S. (2016b). A review on the measurement of ecosystem multifunctionality. Biodiversity Sci.24 (1), 72–84. doi: 10.17520/biods.2015170

  • 123

    WrightI.ReichP.WestobyM. (2003). Least-cost input mixtures of water and nitrogen for photosynthesis. Am. Nat.161, 98–111. doi: 10.1086/344920

  • 124

    WrightI. J.ReichP. B.WestobyM.AckerlyD. D.BaruchZ.BongersF.et al. (2004). The worldwide leaf economics spectrum. Nature428 (6985), 821–827. doi: 10.1038/nature02403

  • 125

    WuC. S.LiuG. H.HuangC.LiuQ. S. (2019). Soil quality assessment in yellow river delta: establishing a minimum data set and fuzzy logic model. Geoderma334, 82–89. doi: 10.1016/j.geoderma.2018.07.045

  • 126

    XuS.FuW.Pingq.HeX. Y.ChenW.WuX.et al. (2017). Research progress on effects of climate change on tree litter decomposition. Chin. J. Ecol.36 (11), 3266–3272. doi: 10.13292/j.1000-4890.201711.029

  • 127

    YangJ.ZhangG. C.CiX. Q.SwensonN. G.CaoM.ShaL. Q.et al. (2014). Functional and phylogenetic assembly in a Chinese tropical tree community across size classes, spatial scales and habitats. Funct. Ecol.28 (2), 520–529. doi: 10.1111/1365-2435.12176

  • 128

    YuH. Y.ChenY. T.XuZ. Z.ZhouG. S. (2014). Relationship and economic spectrum analysis of leaf functional traits in desert steppe of inner Mongolia. Chin. J. Plant Ecol.38 (10), 1029–1040. doi: 10.3724/SP.J.1258.2014.00097

  • 129

    ZartmanR. E. (2005). Spatial and temporal statistics: sampling field soils and their vegetation. Vadose Zone J.4 (4), 983–983. doi: 10.2136/vzj2005.0004br

  • 130

    ZengY.LuY. Z.ChangW. D.JianM. Z. (2014). Infrared photoacoustic spectroscopy and support vector machine model were used to determine soil organic matter content. Acta Pedologica Sin.51 (6), 1262–1269. doi: 10.11766/trxb201311110526

  • 131

    ZhangM. (2020). Linking soil nutrient cycling and microbial community with vegetation cover in riparian zone. Geoderma384, 114801. doi: 10.1016/j.geoderma.2020.114801

  • 132

    ZhangC. B.KeS. S.WangJ. A.GeY.ChangS. X.ZhuS. X.et al. (2011b). Responses of microbial activity and community metabolic profiles to plant functional group diversity in a full-scale constructed wetland. Geoderma160 (3-4), 503–508. doi: 10.1016/j.geoderma.2010.10.020

  • 133

    ZhangM. N.LiG. Y.LiuB.LiuJ. S.WangL.WangD. L. (2020b). Effects of herbivore assemblage on the spatial heterogeneity of soil nitrogen in eastern Eurasian steppe. J. Appl. Ecol.57 (8), 1551–1560. doi: 10.1111/1365-2664.13655

  • 134

    ZhangC. B.LiuW. L.LuoB.GuanM.WangJ.GeY.et al. (2020a). Spartina alterniflora invasion impacts denitrifying community diversity and functioning in marsh soils. Geoderma375, 114456. doi: 10.1016/j.geoderma.2020.114456

  • 135

    ZhangC.XueS.LiuG. B.SongZ. L. (2011a). A comparison of soil qualities of different revegetation types in the loess plateau, China. Plant Soil347 (1-2), 163–178. doi: 10.1007/s11104-011-0836-5

  • 136

    ZhangS.-B.ZhangJ.-L.FerryJ.CaoK.-F. (2012). Leaf element concentrations of terrestrial plants across China are influenced by taxonomy and the environment. Global Ecol. Biogeography21(8), 809–818. doi: 10.1111/j.1466-8238.2011.00729.x

  • 137

    ZhaoS.WangQ.ZhouJ.YuanD. (2018). Linking abundance and community of microbial N2O-producers and N2O-reducers with enzymatic N2O production potential in a riparian zone. Sci. Total Environ.642, 1090–1099. doi: 10.1016/j.scitotenv.2018.06.110

  • 138

    ZhaoS.ZhangB.SunX.YangL. (2021). Hot spots and hot moments of nitrogen removal from hyporheic and riparian zones: a review. Sci. Total Environ.762, 144168. doi: 10.1016/j.scitotenv.2020.144168

  • 139

    ZhouH. L.ZhouG. S.HeQ. J.ZhouL.JiY. H.LvX. M. (2021). Capability of leaf water content and its threshold values in reflection of soil-plant water status in maize during prolonged drought. Ecol. Indic124, 1470–1160. doi: 10.1016/j.ecolind.2021.107395

  • 140

    ZouQ. Y.ChenH. S.MaX. Y.NieY. P. (2019). Analysis of plant water source in karst exposed bedrock habitat based on water control experiment. Chin. J. Appl. Ecol.30 (3), 759–767. doi: 10.13287/j.1001-9332.01903.012

  • 141

    ZuoX.ZhouX.LvP.ZhaoX.ZhangJ.WangS.et al. (2016). Testing associations of plant functional diversity with carbon and nitrogen storage along a restoration gradient of sandy grassland. Front. Plant Sci.7 (e17476). doi: 10.3389/fpls.2016.00189

Summary

Keywords

arid regions, random forest, regression kriging, semi-variable functions, spatial variation

Citation

Chen Y, Wang J, Jiang L, Li H, Wang H, Lv G and Li X (2023) Prediction of spatial distribution characteristics of ecosystem functions based on a minimum data set of functional traits of desert plants. Front. Plant Sci. 14:1131778. doi: 10.3389/fpls.2023.1131778

Received

26 December 2022

Accepted

10 May 2023

Published

02 June 2023

Volume

14 - 2023

Edited by

Zhong-Hua Chen, Western Sydney University, Australia

Reviewed by

Lingcheng Li, Pacific Northwest National Laboratory (DOE), United States; Terry Lin, Western Sydney University, Australia

Updates

Copyright

© 2023 Chen, Wang, Jiang, Li, Wang, Lv and Li.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Guanghui Lv,

†These authors have contributed equally to this work and share first authorship

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

Outline

Figures

Cite article

Copy to clipboard


Export citation file


Share article

Share on WeChat

Scan with WeChat to share this article

Article metrics