Effects of Dynamic Changes of Soil Moisture and Salinity on Plant Community in the Bosten Lake Basin

Abstract

To estimate the potential risks of plant diversity reduction and soil salinization in the Bosten Lake Basin, the dynamic changes in the plant community and species diversity affected by soil moisture and salinity were analyzed from 2000 to 2020 based on remote sensing technology and field experiments. A model for simulating soil moisture, salinity, and the productivity of the plant communities was proposed. The results demonstrated that: (1) The soil moisture index (SMI) increased but the soil salinity index (SSI) decreased from 2000 to 2020 in the study areas. Accordingly, the plant community productivity indices, including the vegetation index (NDVI), enhanced vegetation index (EVI), and ratio vegetation index (RVI), exhibited an increasing trend. It was found that the Alpine meadow, Alpine steppe, and temperate steppe desert were the main types of plant communities in the study areas, accounting for 69% of its total area. (2) With increasing SMI or decreasing SSI, the vegetation productivity such as NDVI, RVI, and EVI all exhibited an increasing trend. With the increment of SMI, the species diversity indices of the Simpson, Shannon–Wiener, and Margalef exhibited a distinctly increasing trend. However, the indices of the Simpson, Shannon–Wiener, and Alatalo increased with the decreasing SSI. (3) The study discovered from the SVM model that the species diversity index was optimal when the soil salinity was 0–15 g/kg and the soil moisture was 12–30% in the study areas. It was found that soil moisture, not soil salinity, controls the plant species diversity change in the study areas. (4) A multiple linear regression model was established for simulating the effect of soil water-salinity on the vegetation productivity index at the watershed scale. The model indicated that higher salinity would reduce vegetation productivity and higher soil moisture would promote vegetation growth (except for RVI). The SSI had a higher impact on NDVI and EVI than the SMI in the study areas. This study would support decision-making on grassland ecosystem restoration and management in the other arid areas.

1. Introduction

The species diversity and productivity of plant communities play important roles in maintaining ecosystem stability and biodiversity [1,2]. Especially in the arid areas, hydrological processes, topography, soil moisture, salinity, and other environmental factors affect the ecological dynamics of plant communities [3]. This affects the direction and degree of biological processes in the ecosystem, while its instability strongly restricts the developing direction of the ecosystem [4]. The strong spatial heterogeneity of soil moisture and salinity, which is controlled by many environmental factors, including topography, hydrological processes, and land use, determines the spatial distribution of the vegetation [5]. The key components of hydrological processes include precipitation [6,7], surface water [8], and groundwater [9]. They all change the soil moisture and salinity of a habitat and then significantly affect the composition and productivity of plant communities [10]. Thus, studies on the relationships between plant communities and soil water or salt gradients would help to understand the current habitat conditions and can help identify future changes of surface ecosystems [11,12].

Several studies have been conducted on the effect of soil salinity on plant species diversity and community structure [13,14,15]. High soil salinity can reduce the primary production of some natural plant communities and can affect plant community structures [16,17]. In turn, the magnitude of plant community responses to modifications in soil salinity depends on their structures and compositions [18]. Several studies indicated that the soil moisture and salinity both affected the distribution and ecological response of plant communities [15,19]. The soil moisture and salinity variation are the main environmental factors that affect the composition, result, function, and succession of plant communities [20]. However, few studies have considered the temporal dynamics of species composition and structure characteristics caused by plant community growth [12,21] and few have focused on qualifying the contribution of soil moisture or salinity to the plant community at a watershed scale.

In the past 50 years, natural ecological processes in the Bosten Lake Basin have been changed significantly due to intense human activities with the development and utilization of water resources [22]. As a result, the natural vegetation has been replaced mostly by farmland, leading to soil salinization. The lake salinity increased, the balance of the Bosten Lake wetland ecosystem was damaged, plant biodiversity was reduced, and sustainable development was affected adversely in the basin [23]. The influence of soil moisture and salinity on the succession of plant communities must be examined to provide information that a decision-maker can use to protect and facilitate the integrated management of the Bosten Lake Basin ecosystems.

Based on the remote sensing technology and field monitoring, this paper analyzed the effect of the soil moisture and salinity change on the productivity, distribution, and composition of the natural plant community from 2000 to 2020 in the Bosten Lake Basin. The specific objectives of this paper are to: (1) characterize and compare spatial-temporal characteristics of the soil moisture, soil salinity, and natural plant community in the Bosten Lake Basin; (2) discuss the effects of soil moisture/salinity and their interaction on natural plant community change; (3) propose a model for soil moisture, salinityand plant community change in the study area. The results of this study will contribute for controlling soil moisture or salt and protecting the terrestrial ecosystem and animal husbandry development in the Bosten Lake Basin.

2. Materials and Methods

2.1. Study Area

The Bosten Lake Basin is located in Xinjiang, northwest China and mainly includes Heshuo County, Hejing County, Yanqihui Autonomous County, and part of Bagrax County. The spatial boundaries of the study area are as follows: 41°21′19″–43°21′34.8″ N. and 82°54′10″–88°21′06″ E. Its total area is 43,930 km² with altitudes of 1008–4801 m. In terms of topography, the region has high elevations in its northwest region and low elevations in the southeast. The study area consists mainly of the Kaidu River Basin (including the Greater and Lesser Yudus basins), Huangshuigou River Basin, Qingshuihe River Basin, Wushtala River Basin, Yanqi Basin, and more than 20 smaller river basins (Figure 1).

The Bosten Lake Basin is characterized by a temperate continental arid climate. The annual average precipitation is only 60 mm, with more than 80% occurring in summer and while annual average evaporation is 2368 mm [24]. The water supply of the rivers in the basin depends mainly on melting ice and snow from the mountains and rainfall. There are different land covers in the Bosten Lake Basin including a glacier snow belt, meadow steppe belt, oasis plain, desert steppe and desert belt, and the Bosten Lake area with the decrease in the elevation [25].

2.2. Remote Sensing Data

The remote sensing data used were the Tier 1 data from the Landsat TM/ETM+/Oli sensors. The Landsat data were processed using the QA band generated by the CFMASK algorithm, a cloud, and a cloud shadow mask to retrieve nearly cloudless images. The band assimilation coefficient of the Landsat series data proposed by Roy was applied to assimilate the data from the three sensors [26]. The Landsat time series data from the Bosten Lake Basin included 305 Landsat-5 data images; 1114 Landsat-7 data images; and 528 Landsat-8 data images from 2000 to 2020. In total, 1947 images were processed using the remote sensing big data platform.

2.3. Investigation of Plant Community Types and Diversity

Field surveys were conducted in the lake and plain area and the Piedmont alluvial fan of the Hoe San Estate in the Bosten Lake Basin from May to July 2020. A total of 43 sample plots of typical vegetation types were selected. Ecological indices such as plant species, quantity, canopy width, coverage, growth tendency, and community types were measured. In order to meet the sample statistical requirements, three samples were randomly selected from each typical plant community sample plot. The tree sample size was 10 m × 10 m, the shrub sample size was 5 m × 5 m, and the herb sample size was 1 m × 1 m. Soil samples were extracted from each sampling area according to the three-point sampling method. The soil sample depths were 0–10 cm, 10–20 cm, and 20–30 cm. The soil samples were analyzed in the laboratory in terms of their soil moisture and soil salinity. The soil moisture content is 15–24%, the soil salinity is 10–20 g/kg in the study areas. The plant communities in the study area were classified based on the field survey data and the results of classifications of vegetation in the Xinjiang grasslands (

Table 1. Classification System of plant communities in the Bosten Lake Basin.

Number	Plant Communities/Land Cover
1	Halophyte meadow community
2	Marshy meadow community
3	Low swamp community
4	Alpine steppe community
5	Alpine meadow community
6	High swamp community
7	Sub-class of sandy temperate desert community
8	Temperate steppe desert community
9	Temperate desert community
10	Gravelly temperate steppe desert sub-class community
11	Mountain meadow community
12	Montane temperate steppe community
13	Alpine temperate desert-steppe community
14	Temperate desert community in saline soil
15	Cultivated land
16	Water area
17	Forest
18	Glaciers and snow
19	Construction Land

).

2.4. Data Processing and Calculation Methods

In this study, based on the series data from the Landsat, the normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), ratio vegetation index (RVI), soil salinity index (SSI), and soil moisture index (SMI) were utilized to quantify the spatial and temporal change of plant communities. The calculation formulas are presented in

Table 2. The calculation formulas for the main vegetation, soil moisture, and salinity indices.

Index Name	Formula *	Reference
NDVI	$\frac{NIR - Red}{NIR + Red}$	[27]
EVI	$\frac{2.5 \times \left(NIR - Red\right)}{NIR + 6 \times Red - 7.5 \times Blue + 1}$	[28]
RVI	$\frac{NIR}{Red}$	[29]
SSI	$\frac{Red - NIR}{Red+NIR}$	[30]
SMI	$SMI\left(i\right)o=\sqrt{{(e_p\left(max\right) - e_p \left(i\right))}^2 + MPDI{\left(i\right)}^2}$	[31]

* Blue, Red, and NIR is B-2, B-4, and B-8 bands, respectively. e_p(i) and MPDI(i) stand for the corresponding emissivity and microwave polarization difference index at day i, respectively, and e_p(max) represents multiyear maximum value of e_p(i) for a specific satellite pixel.

.

In order to remove noise and canopy structure information from Landsat data, this paper uses the spectral index of soil moisture or salt salinity as following:

$$R{I}_{\left(i,j\right)}=\frac{R_i}{R_j}$$

(1)

$$ND{I}_{\left(i,j\right)}=\frac{(R_i-R_j)}{(R_i+R_j)}$$

(2)

$$D{I}_{\left(i,j\right)}=R_i-R_{j }$$

(3)

where $R{I}_{\left(i,j\right)}$ is the ratio spectral index, $ND{I}_{\left(i,j\right)}$ is the normalized spectral index, $D{I}_{\left(i,j\right)}$ is the difference spectral index, and i and j are any two-band data from the Landsat OLI data band.

The Support Vector Regression (SVR) is accurate for the challenge of the non-linearity due to the high generalization characteristics of the Support Vector Machine (SVM) models [32,33]. Thus, the SVM model is adopted to construct the regression model among the spectral index of soil moisture, salt salinity, and plant diversity. Based on the 43 pixels data of RI, NDI, and DI from Landsat and the 43 sampling data of the soil moisture and salt salinity from field investigation, 30 groups of data were selected for modeling randomly and 13 groups were used to validate the results.

The plant species diversity indices are analyzed using the Margalef, Simpson, Shannon–Wiener, and Alatalo indices [34], as shown in Equations (4)–(7), respectively:

$$Margalef index=\frac{\left(S-1\right)}{Ln N}$$

(4)

$$Simpson index=1-Ln\left[ {{\displaystyle \sum }}_{i=1}^s{\left(\frac{N_i}{N}\right)}^2 \right]$$

(5)

$$Shannon-Wiener index=-{\displaystyle \sum }{P}_i\mathit{ln}{P}_{i }$$

(6)

$$Alatalo index=\left[\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$({\displaystyle \sum }P{i}^2)$}\right.-1\right]/\left[\exp \left(-P_{iln}{P}_i\right)-1\right];Pi=N_i/N$$

(7)

where S is the total number of species in a sample, N is the total number of individuals in a sample, N_i is the number of individuals of species i, and P_i is the proportion of the number of individuals of species i to the total number.

3. Results

3.1. Dynamic Characteristics of Soil Moisture, Salinity, and Plant Communities in the Bosten Lake Basin

The SMI and SSI were analyzed as shown in Figure 2. The SMI values exhibited a gradually increasing trend and the increased rate is 0.003/10a from 2000 to 2020. Firstly, the SMI values decreased from 2000 to 2004. Then, they exhibited some variability from 2005 to 2010. Finally, they increased gradually from 2011. The SSI exhibited a decreasing trend and the decreased rate was 0.005/10a. It increased from 2000 to 2010, was up to the minima value in 2017, and began to increase since 2017.

Figure 3 showed that the vegetation productivity, such as the NDVI, EVI, and RVI, exhibited a similar increasing trend from 2000 to 2020 in the study areas. The value of NDVI fluctuated between 0.22 and 0.28 and was 0.25 averagely over the 21-year period. The value of EVI was 0.56 averagely and varied between 0.50 and 0.64. The RVI varied between 1.78 and 2.22, was 2.00 averagely, and showed a generally increasing trend.

There were significant differences among the area percentage of the different plant community types in the study areas (Figure 4). Alpine meadow, Alpine steppe, and temperate steppe desert were found to be the main community types in the basin, accounting for 69% of the total area. Temperate steppe and lowland halophytic meadow communities were predominant, accounting for 9% of the total communities. The plant community with the smallest area was the mountain meadow community (0.04%). The proportions of the areas of different plant communities in the study areas can be classified in the following order: Alpine meadow community > Alpine steppe community > Temperate steppe desert community > Low swamp community > Mountain temperate steppe community > Temperate desert community > High swamp community > Sub-class of sandy temperate desert community > Mountain meadow community.

3.2. Relationship between Soil Characteristics and Plant Productivity in the Bosten Lake Basin

The effects of soil moisture and salinity on the productivity of the vegetation communities were analyzed in Figure 5 and Figure 6. The value of the NDVI, RVI, and EVI all exhibited decreasing trends with the increasing SSI. The value of the NDVI, RVI, and EVI all decreased with the increase in the SMI.

The investigation data from Figure 7 showed the effects of soil moisture and salinity on plant species diversity in the Bosten Lake Basin. The indices of the Margalef, Simpson, and Shannon–Wiener except for the Alatalo exhibited distinctly increasing trends with the increasing soil water content.

Figure 8 shows the relationship between soil salinity and plant species diversity. It was found that increasing the soil salinity resulted in decreasing trends of the Simpson, Shannon–Wiener, and Alatalo except for the Margalef index.

3.3. Effects of Soil Characteristics on the Coverage Area of Plant Communities in the Bosten Lake Basin

The dynamic changes in the plant community area percentage, soil moisture, and salinity in the Bosten Lake Basin were analyzed (Figure 9 and Figure 10). Figure 9 demonstrated that soil salinity changes somewhat affected the plant community area percentage in the study region. As the SSI increased, the area percentage of six plant communities (high swamp, alpine temperate desert steppe, montane temperate steppe, halophyte meadow, temperate steppe desert, and alpine steppe) exhibited a similar increasing trend. However, the area percentage of the temperate desert positive community and alpine meadow community decreased with the increasing SSI.

Figure 10 showed that the percentage of high swamp, alpine steppe, and alpine meadow community areas all decreased with gradually increasing the SMI. Other communities, including alpine temperate desert steppe, montane temperate steppe, halophyte meadow, temperate desert positive, and temperate steppe desert gradually increased with the increment of SMI.

3.4. Effects of Soil Moisture-Salinity Interaction on the Plant Diversity and Productivity of Plant Communities in the Bosten Lake Basin

Based on the SVM model and series data of the four plant diversity indices (Margalef index, Simpson index, Shannon–Wiener index, Alatalo index) from 2000 to 2020, the effect of the soil moistureand salinity interaction on plant diversity change was analyzed. This paper took 3% of soil water content (SWC) and 3 g/kg of soil salinity content (SSC) as a step and then looked for a certain adapted threshold for plant diversity (Figure 11). When the soil salinity content is between 0 and 15 g/kg and the soil water content is between 12 and 30%, the plant diversity index is the highest in the study areas. Therefore, when the soil salinity content is 0–15 g/kg and the soil water content is 12–30%, the plant diversity index is optimal in the Bosten Lake Basin. It was also found that soil moisture, not soil salinity, controls the plant diversity change in the study area.

In order to qualify the effect of the soil moisture-salinity interaction on the productivity of plant communities in the study areas, the 21-year data of soil moisture and salinity were utilized to construct a multiple linear regression model of the vegetation productivity index(Y), SSI (X₁) and SMI (X₂). The three regression equations passed the significance test (p < 0.001), as shown in

Table 3. Regression model of soil moisture and salinity and vegetation index in the Bosten Lake Basin.

	Linear Regression Model	Equation Test Significance
		R2	F	Sig.
1	YNDVI = −0.879 X1 + 0.026 X2 + 0.007	0.918	572.734	0.000
2	YRVI = −5.190 X1 − 1.139 X2 + 1.026	0.838	264.243	0.000
3	YEVI = −2.160 X1 + 0.545 X2 − 0.227	0.810	217.453	0.000

.

The regression equation for NDVI, SSI, and SMI demonstrated that NDVI was negatively affected by SSI and positively affected by SMI. Moreover, SSI had a higher impact on NDVI than SMI. The regression equation for EVI, SSI, and SMI indicated that the affection of SSI and SMI on the EVI was similar to the affection of the NDVI. This study indicates that high soil salinity would reduce the vegetation productivity while higher soil water content would promote vegetation growth. The regression equation for RVI, SSI, and SMI demonstrated that the RVI index was negatively affected by SSI and SMI. That the RVI exhibited a decreasing trend as both SSI and SMI increased.

4. Discussion

Climatic factors are the most important factors affecting vegetation change in arid inland regions [35,36]. The annual temperature and precipitation gradually increased from 2000 to 2020 in the Bosten Lake Basin [23]. The climate-driven SMI also increased while the SSI decreased. The NDVI, EVI, and RVI were at a high level and showed an increasing trend from 2000 to 2020. Moreover, the precipitation or temperature change were mostly consistent with the change of the NDVI, RVI, and EVI. The vegetation growth improved, driven by an increase in large-scale regional precipitation [37]. Thus, sufficient heat and water conditions can lead to high vegetation coverage and productivity in the study areas. Due to the increasing precipitation in the study area, the soil salinity decreased continuously from 2000 to 2020. Reductions in salinity are generally beneficial to the growth and survival of plant communities [38]. This study has verified that plant community productivity has generally improved and that changes in the microclimate affect the distribution of plant communities and reflect the overall improvement of the regional environment [39,40].

Soil moisture and salinity both strongly affect the structure, composition, and dynamic change of plant communities in arid regions [41]. Some studies have demonstrated that soil salinity affects plant niches and species diversity by changing the physicochemical properties of the soil [10]. When habitat conditions change from good to poor in arid areas, the water gradients had more significant and more directed effects than salinity gradients on plant species and communities [42]. This study firstly quantified the soil water and salinity content when the species diversity index was optimal in the Bosten Lake Basin. The result was more beneficial to quantitatively identify the plant community response range to changes in soil moisture and salinity.

Soil moisture and salinity strongly affected the productivity and diversity of plant communities but had only a minor effect on the areas of some plant types. The proportion of alpine meadows, temperate steppe deserts, high marsh, mountainous temperate desert steppe, mountainous temperate steppe, and lowland halophytic meadows were not significantly correlated with SSI and SMI in the Bosten Lake Basin. This indicates that the distribution of areas of different plant communities in the basin was more affected by other environmental factors.

Although the regression relationships between vegetation productivity and soil moisture and salinity factors have been quantified, other environmental factors including topography, soil texture, and soil nutrients should be considered [12]. Theoretically, they can also affect the characteristics and distribution of plant communities. Of course, the uncertainty of remote sensing data may affect the research results [36,43]. Our results were also affected by the limitation or uncertainty of Landsat-5 data images. The field survey data may be limited by the number of plant community investigation sampling points in our current study. Thus, the results from simulating the SVM model need to verify and optimize model parameters in the future.

5. Conclusions

This study provides an in-depth analysis of the dynamic change of soil water and salt and their effect on the plant community based on remote sensing data for 21 consecutive years. It can provide a method for improving the species diversity controlled by soil water and salt. It is very important for the local government to estimate the potential risks of plant community function reduction and soil salinization and restore destroyed grassland ecosystem, especially in the arid areas.

It can be concluded that there has been an increase in soil moisture and decrease in soil salt in the last 21 years; the indices of the Simpson and Shannon–Wiener exhibited distinctly increasing trends. The plant species diversity index is optimal when the soil salinity was 0–15 g/kg and the soil moisture was 12–30% when simulating the SVM model. It is indicated that soil moisture, not soil salinity, mainly controls the plant diversity change in the study area.

The productivity of the plant community is also affected by both the soil moisture and salinity. However, the SSI had a negative effect and the SMI had a positive effect on the plant productivity. The SSI had a higher impact on NDVI and EVI than SMI. The research provides findings that can provide a foundation for supporting decision-making related to managing the hydrology, regulating the salinity, and restoring the vegetation in the study areas. Moreover, this study provides a model of the soil water-salinity-plant community and was proposed to reflect the succession of plant communities in the arid areas. Therefore, this research is considered to delicately qualify plant growth and succession at the watershed scale.