The Impact of Groundwater Burial Depth on the Vegetation of the Dariyabui Oasis in the Central Desert

Abstract

Vegetation and groundwater are important components of the ecological environment of oases in desert hinterlands and their relationship is crucial to ecosystem stability. In this study, Sentinel-2 data for 2016–2022 and measured groundwater burial depths were analysed for the Dariyabui Oasis in the hinterland of the Taklamakan Desert. The spatial and temporal changes in vegetation and groundwater burial depth from 2019 to 2022 were analysed based on the image–element dichotomous model of the normalised difference vegetation index, utilising the inverse distance weight interpolation method, cubic curve regression, image–element difference, slope trend analysis, and the Markov transfer matrix for determining the temporal and spatial response law between the two. Finally, the threshold value of groundwater burial depth for different vegetation cover types was clarified. The fractional vegetation cover of the Dariyabui Oasis showed a slight increase from 2016 to 2022. Vegetation in the northwest and southeast of the oasis increased, whereas vegetation decreased in the mid-north and northeast regions; 5.14% of the total area experienced increased coverage, whereas 3.35% experienced decreased coverage. The depth of groundwater in the oasis showed a pattern of gradual increase from the entrance to the end of the oasis, that is, south to north. The depth of groundwater in the oasis from 2019 to 2022 was stable, with a 4-year average depth of 4.1069 m and a maximum fluctuation of 0.4560 m. The interannual changes in the groundwater level showed an increasing trend in January–April, while groundwater levels showed a decreasing trend in May–July and August–October and remained constant in June–July and October–December. Oasis vegetation cover showed a negative correlation with groundwater depth, with a depth interval for the highest low-cover vegetation distribution of 3–6 m, and an ultimate depth threshold of 7 m. The depth interval with the highest medium-cover vegetation distribution was 3–4 m, that with the highest high-cover distribution was 2–4 m, and the ultimate depth threshold was 6 m. The depth of the oasis ranged from 3 to 6 m and the ultimate depth threshold was 7 m.

1. Introduction

Vegetation is a key component of the biosphere, playing an irreplaceable role as the “lungs of the Earth” in the regulation of biogeochemical cycles, including water and carbon, and Earth’s energy balance. Vegetation provides key ecosystem services, such as habitat for wildlife and food [1], and contributes to human social and economic activities [2,3]. Vegetation changes are susceptible to various factors, including temperature, precipitation, light, soil, and groundwater [4,5,6,7], to varying degrees.

In arid and semi-arid zones, vegetation primarily relies on groundwater as its primary water source. To some extent, the depth of groundwater burial determines the distribution, development, and succession of desert plant communities [8,9]. Thus, studying the relationship between oasis vegetation and groundwater in arid and semi-arid zones is crucial for preserving biodiversity and maintaining ecosystem stability [10,11].

In recent years, scholars have carried out substantial research on the distribution of vegetation, groundwater depth, water level, and water quality, as well as the response of vegetation to spatial and temporal changes in groundwater, in arid and semi-arid ecosystems. For example, Bahir et al. [12] and Ouhamdouch et al. [13] used the water quality index and Rafik et al. [14] used isotope and GIS techniques to analyse the hydrochemical characteristics of the Essaouira Basin (Morocco) on the northwest coast of Africa and assessed the quality of local water for domestic use and irrigation. The results provided recommendations and theoretical support for the sustainable development of groundwater sources in the basin. The spatial heterogeneity of vegetation interacts in a complex manner with vegetation growth conditions and environmental changes, with climate change, groundwater levels, soil salinity, microtopography, and human activities influencing vegetation. In particular, the groundwater level is strongly correlated with vegetation. The over-exploitation of groundwater causes a decline in the water table, leading to an accumulation of soil salts and toxins, especially in drought years, in areas of high agricultural activity and increased water stress [15,16,17,18,19,20,21,22]. In China, the relationship between vegetation and groundwater burial depth, as well as the water table, water quality, and soil, has been studied in the Tarim River [23], Hei River [24], Shiyang River Basin [25,26], Sanguang River [27], Inner Mongolia Loop [28], Loess Plateau [29], Hulunbeier in Inner Mongolia [30], Ejina [31], and Gurbantunggut Desert [32,33]. Such studies have provided theoretical support for protecting the stability of vegetation in arid and semi-arid regions.

Located in the hinterland of the Taklamakan Desert in the extreme arid zone, the Dariyabui Oasis is the largest existing natural oasis in China. The oasis is subject to minimal human intervention and possesses a complete, original, and independent ecosystem, making it a one-of-a-kind “pristine oasis” [34]. The ecological environment of the Dariyabui Oasis is relatively fragile, having a single type of vegetation, that is, drought-tolerant vegetation, including the dominant poplar, tamarisk, and reeds, and less common camel’s thorn and bitter pea [35]. During the flood season, water from the Kriya River recharges the groundwater of the Dariyabui Oasis, and the groundwater becomes the only source of water to sustain the growth of the vegetation in the oasis [36]. No modern industry or agriculture is present within the oasis, so the surface water reaching the oasis from the Kriya River can also be said to have good ecological quality. In recent years, the villagers have used primitive earth dams in the oasis to divert surface water to irrigate the oasis vegetation in different areas and, thus, the growth of the oasis vegetation is affected by certain anthropogenic factors [37]. Overall, the vegetation in the oasis is relatively fragile. According to one field study, the depth of groundwater in the oasis is relatively low in the south and relatively high in the north. Reeds and tamarisks dominate the vegetation in locations with shallow groundwater, poplars and tamarisks dominate in areas with deep groundwater, and poplars and a few tamarisks dominate the northern part of the oasis, where groundwater is the deepest. Vegetation in the northern part of the oasis has been degraded to varying degrees, and studies have been conducted on plant communities affected by the groundwater depth [38], the spatial distribution pattern of poplar [39], and the estimation of vegetation cover in the oasis based on visible and multispectral indices from UAV remote sensing [40].

This study aimed to investigate the oasis vegetation–groundwater interaction in the Dariyabui Oasis and reveal the law of ecological evolution of the oasis in its natural state for the first time. Sentinel-2 remote sensing images were used to extract the fractional vegetation cover (FVC) of the study area based on the image–element dichotomous model of the normalised difference vegetation index (NDVI), and its accuracy was verified using UAV aerial images of vegetation samples. Data on groundwater burial depth from 2019 to 2022 were used to calculate a raster map of burial depth in the study area using the inverse distance weight (IDW) interpolation method, and the correlation between oasis vegetation and groundwater burial depth was analysed using cubic curve regression. The objectives of this study were to: (1) analyse the spatial and temporal distribution patterns of the oasis vegetation; (2) analyse the spatial and temporal distributions of the groundwater burial depth in the oasis; and (3) reveal the spatial and temporal response relationship between vegetation and groundwater in primitive oases in the hinterland of extreme arid deserts, as well as clarify the thresholds of groundwater depth for different vegetation cover. Our findings provide data and theoretical support for the maintenance of ecosystem stability in the hinterland of deserts.

2. Materials and Methods

2.1. Study Area

The Kriya River is the second-largest instream river at the southern margin of the Tarim Basin in Xinjiang, which originates at the northern margin of the Kunlun Mountains [41], flows northwards, and eventually disappears into the hinterland of the Taklamakan Desert. The Kriya River flows through the area, forming the Yutian Oasis in the middle reaches and the Dariyabui Oasis at the end of the river [37], which is the farthest oasis from the desert hinterland, as well as the largest natural oasis in China (Figure 1) [34]. The study area was located at 38°15′–38°50′ N, 81°43′–82°15′ E, covering a total area of 3248 km². The oasis vegetation area accounts for approximately 350 km². The altitude is 1110–1300 m, with a warm temperate desert climate [42], and the average annual precipitation is <10 mm [43]. The multi-mean temperature is 12.1 °C, potential annual evaporation is 2480 mm, and an extreme temperature difference exists between day and night. Reeds of the Gramineae family, tamarisk shrub, and poplar tree were the dominant species (Figure 2), and minor vegetation types included camel thorn and bitter pea [38,44].

2.2. Data Sources

2.2.1. Remote Sensing Data

Sentinel-2 remote sensing imagery data obtained from the ESA Copernicus Data Centre (http://scihub.copernicus.eu/dhus/#/home (accessed on 15 December 2023)) and USGS (https://Earthexplorer.usgs.gov/ (accessed on 15 December 2023)) were used for the study area. Sentinel-2 has two satellites, A and B, with a 5-day revisit period and 12 bands, of which B2 (blue), B3 (green), B4 (red), and B8 (NIR) have a spatial resolution of 10 m. The original remote sensing images of seven Sentinel-2L1C views (20160919, 20170919, 20180919, 20190919, 20200913, 20210918, and 20220923) were downloaded for summer when vegetation flourished and the sky was cloudless. The SNAP Toolbox (http://step.esa.int/main/download/snap-download/ (accessed on 15 December 2023)), a free and open-source Sentinel application platform developed by the ESA, was used for the science development of the Sentinel mission. The Sen2Cor algorithm [45,46] (https://step.esa.int/main/snap-supported-plugins/sen2cor/sen2cor-v2-11/ (accessed on 15 December 2023)) in the SNAP Toolbox version 2.11.0 was used for atmospheric corrections, radiometric calibrations, and to obtain the Sentinel-2L2A image data. The NDVI was calculated to determine the confidence interval and the raster calculator tool was used to calculate the FVC, which was cropped to obtain the vegetation cover of the study area.

2.2.2. Drone Data

From 23 to 25 August 2021, a DJI Phantom 4 multispectral quadcopter drone (DJI Innovation Technology Co., Ltd., Shenzhen, China) photographed the sample site. The Phantom 4 multispectral drone was equipped with a multispectral sensor with CMOS lenses for visible (RGB), NIR, and red-edge imaging. The sensor bands were the B band, with a wavelength of 450 nm ± 16 nm; G band, with a wavelength of 560 nm ± 16 nm; R band, with a wavelength of 650 nm ± 16 nm; red edge (RE) band, with a wavelength of 730 nm ± 16 nm; and via the NIR band, with a wavelength of 840 nm ± 26 nm. In this study, we used images of eight sample plots taken by UAV aerial photography at a size of 1000 m × 300 m, as shown in Figure 1, to calculate the vegetation cover using the image–element dichotomous model for the NDVI of Sentinel-2 in the same period and to verify the accuracy.

2.2.3. Groundwater Observation Data

The group arranged 19 wells in the Dariyabui Oasis in October 2018 (Figure 1) and placed HOBO (www.onsetcomp.com (accessed on 15 December 2023)) inductive water pressure level metres (Figure 2) for continuous observation of groundwater in the oasis, with a monitoring frequency of 4 h, six times a day. Two field trips were made to the study area, from 27 March to 6 April 2021 and from 10 to 18 May 2022, to collect groundwater data. Continuous groundwater burial depth data from 2019 to 2022 were used in this study. The groundwater burial depth in the monitoring wells was calculated as follows:

$$H=h-\frac{p_i-p_j}{\rho \mathrm{g}},$$

(1)

where H is the depth of groundwater burial (m), $h$ is the height of the water level gauge in the ground (m), $p_i$ is the groundwater water pressure data collected by the HOBO monitoring instrument (Pa), $p_j$ is the atmospheric pressure data collected by HOBO (Pa), $\rho$ is the density of water (Kg.m⁻³), and $\mathrm{g}$ is the acceleration of gravity, taken as 9.8 N.Kg⁻¹.

2.3. Research Methodology

2.3.1. Normalised Difference Vegetation Index

After pre-processing the Sentinel-2 remotely detected images, the NDVI was calculated using ArcGIS (Equation (2)):

$$NDVI=\frac{NIR-R}{NIR+R} ,$$

(2)

where $NIR$ is the reflectance value of the spectral information in the near-infrared band and $R$ is the reflectance value of the spectral information in the visible red-light band.

2.3.2. Calculation and Accuracy Verification of Vegetation Cover

The image–element dichotomous model is widely used in FVC studies and has high simulation accuracy [47,48]; it can effectively and objectively reflect the vegetation canopy. In this study, we applied the image–element dichotomous method using Equation (3):

$$FVC=\frac{NDVI-NDV{I}_{soil}}{NDV{I}_{veg}-NDV{I}_{soil}},$$

(3)

where $NDV{I}_{soil}$ represents the NDVI value of pure bare ground partial cover and $NDV{I}_{veg}$ represents the NDVI value of pure vegetation cover. Ideally, pure bare ground would have an $NDV{I}_{soil}$ value of 0 and pure vegetation would have an $NDV{I}_{veg}$ value of 100%. However, remote sensing images are affected by the atmosphere, surface complexity, and soil when obtaining spectral information, and scholars often investigate the study area in the field and determine confidence intervals by combining various factors. For example, confidence values differ, with cumulative likelihoods of NDVI purely bare ground taken as values of 0, 0.5%, 1%, 2%, and 5%, and cumulative likelihoods of NDVI purely vegetation taken as values of 100%, 99.5%, 99%, 98%, and 95% [2,49,50,51,52,53]. The study area was an arid desert hinterland with a single type of surface vegetation. Combined with the field survey and references [54], we determined the NDVI cumulative image value of 1% equal to the $NDV{I}_{soil}$ and the NDVI cumulative image value of 99% equal to $NDV{I}_{veg}$, which were used to calculate the FVC.

Verification of the accuracy of vegetation cover extracted by satellite remote sensing is traditionally conducted using field surveys to quantify vegetation cover by delineating sample plots, which is a time-consuming and labour-intensive method. Therefore, this study was based on the Sentinel Remote Sensing data and UAV aerial photographs of the sample plots. A linear regression model of vegetation cover was developed using the FVC extracted using Image J (Fiji) software (Version number: ij153-win-java8) as the measured values, and the image meta-dichotomous model was tested by the goodness-of-fit. Image J is based on the JAVA language development of image processing software (http://image.nih.gov/ij/ (accessed on 15 December 2023)); the visual interface is simple to operate and was originally mainly used in the field of medicine to calculate the area of cells and wounds. The study area mainly comprised poplar, reed, and tamarisk, with a single type of vegetation, and the orthophoto was taken with UAV, so the vegetation contour is highly similar to the contour of cells and wounds. The resolution of the UAV aerial photography ranges up to centimetre level, which can easily identify the vegetation and the bare ground, and the $FVC$ extracted by Image J was corrected using visual interpretation.

2.3.3. Image Difference Method

To better reflect the dynamic changes in vegetation cover in the study area, the image difference method was used to calculate changes in vegetation cover during different years. A difference of >0 indicates an increase in vegetation cover, a difference equal to 0 indicates stable vegetation cover, and a difference of <0 indicates a decrease in vegetation cover. The formula is shown in Equation (4):

$$\Delta FV{C}_g=FV{C}_{year2}-FV{C}_{year1},$$

(4)

where $FV{C}_{year1}$ and $FV{C}_{year2}$ are the vegetation cover before and after two different periods, respectively. Here, $\Delta FV{C}_g$= −3 (the number represents the magnitude of the difference) indicates severe degradation, $\Delta FV{C}_g$ = −2 indicates moderate degradation, $\Delta FV{C}_g$ = −1 indicates slight degradation, $\Delta FV{C}_g$ = 0 indicates stability, $\Delta FV{C}_g$ = 1 indicates slight improvement, $\Delta FV{C}_g$ = 2 indicates moderate improvement, and $\Delta FV{C}_g$ = 3 indicates extreme improvement [55]. To better visualise changes in vegetation cover from 2016 to 2022, a difference-in-difference analysis was conducted using vegetation cover in different years.

2.3.4. Slope Trend Analyses

$\mathrm{Slope}$ trend analysis is a method of predicting trends in quantities that change over time using linear regression. Linear regression was used to analyse the spatial trend of vegetation cover in the study area from 2016 to 2022 using Equation (5):

$$Slope=\frac{n{{\displaystyle \sum }}_{i=1}^n\left(i\times FV{C}_i\right)-{{\displaystyle \sum }}_{i=1}^n\times {{\displaystyle \sum }}_{i=1}^{n}FV{C}_i}{{{\displaystyle \sum }}_{n=1}^n{i}^2-{({{\displaystyle \sum }}_{n=1}^{n}i)}^2},$$

(5)

where $i$ is the number of years, $n$ is the duration of this study, $\mathrm{and} FV{C}_i$ is the value of each image element in years. The slope is the slope of the FVC trend for each image element. A $\mathrm{slope}$ of <0 represents a decreasing trend in the FVC value of the image element; thus, smaller slope values represent faster vegetation degradation. A slope of 0 represents a constant FVC value for the image element, that is, the vegetation is stable and unchanged. A slope of >0 represents the increasing trend of the FVC value of the image element; larger slope values represent rapid vegetation improvement.

Based on Dariyabui field research and reference to the relevant literature [51,56], the slope was categorised into five classes based on the size of the trend: significantly decreased (slope < −0.015), mildly decreased (−0.015 ≤ slope < −0.008), essentially stable (−0.008 ≤ slope < 0.008), mildly increased (0.008 ≤ slope < 0.015), and significantly increased (slope ≥ 0.015).

2.3.5. Inverse Distance Weight Interpolation

Several interpolation methods exist for the spatial distribution characteristics of groundwater burial depth, such as the spline function method, IDW interpolation, kriging interpolation, and the trend surface method [57,58], each of which has its own advantages and disadvantages. IDW is widely used and is optimal when the study is small and the sample size is large and evenly distributed [59]. Nineteen buried groundwater wells were present in this study area, and were relatively evenly distributed; therefore, the IDW method was used with the following equation:

$$P=\frac{{{\displaystyle \sum }}_{i=1}^n\frac{1}{{(D_i)}^2}{P}_i}{{{\displaystyle \sum }}_{i=1}^n\frac{1}{{(D_i)}^m}},$$

(6)

where $P$ is the interpolated value of the point of interest, $P_i$. The sample value is point i, and $D_i$ is the distance between the interpolated and sampled values. The index $m$ determines how the weight decreases with increasing distance [60].

Groundwater burial depth data collected in the field during September 2019–2022 were selected for correlation analysis with vegetation cover for the same period. Using ArcGIS software to create a vector file of groundwater depth range (Figure 1), the groundwater depth values at each point were interpolated via IDW to obtain the raster image of the groundwater depth, and the raster image of the groundwater depth of the Dariyabui Oasis was obtained through data masking and cropping.

3. Results

3.1. Test of the Degree of Fit of Vegetation Cover

In this study, eight sample plots were aerially photographed by UAVs to obtain orthophotos of Dariyabui Oasis, and RGB images were obtained by splicing and radiometrically correcting the sample plot photographs using DJI Zhitu software. Based on the Sentinel-2A remote sensing imagery with a spatial resolution of 10 m, a number of vegetation sample plots were laid out in the Phantom 4 Multispectral UAV aerial photography at 100 m × 100 m to ensure that the plots readily overlapped with the satellite image pixels. To avoid homogenisation of the images, the samples were selected by spacing them at 100 m intervals; the schematic layout of the samples is shown in Figure 3. The vegetation sample plots were processed in ArcGIS 10.8 with tools including creating fishing nets, creating element classes, and cropping to complete the deployment of sample plots, and a total of 50 vegetation sample plots were deployed in the study area. The vegetation coverage of the UAV image sample squares is shown in Figure 4.

As shown in Figure 5, the vegetation cover of the Sentinel-2 inversion and the FVC extracted by the UAV converged to y = 1.0465x + 3.5167 (R² = 0.83, p < 0.01). The correlation was good and the linear model fit was high. The results showed that the vegetation cover simulated by the image–element dichotomous model based on the NDVI of remote sensing data had high simulation accuracy and reliability in the Dariyabui Oasis.

3.2. Spatial and Temporal Changes in Vegetation Cover

Sentinel-2L1C data were pre-processed using Sen2Cor software, and the vegetation cover was calculated in ArcGIS using the image–element dichotomous model for seven years (2016–2022) in September. The FVC thresholds were delineated based on a field survey of the study area, regarding the relevant literature [55]: <0.15 is considered bare ground and very low vegetation cover, 0.15–0.3 is low vegetation cover, 0.3–0.5 is medium vegetation cover, and >0.5 is high vegetation cover. We obtained the FVC map after the 7-phase partition for 2016–2022. As shown in Figure 6, in terms of spatial distribution, the vegetation cover of the Dariyabui Oasis gradually decreased from southwest to northeast, with medium- and high-coverage vegetation mainly concentrated in the central part of the oasis; the northern regions represent lower vegetation cover. The field study found that the terrain south of the centre of the oasis is low and the depth of the groundwater is shallow.

As shown in

Table 1. Area of vegetation cover of different classes in Dariyabui.

Grade	Vegetation Cover (%)	Year
		Area (km2)
			2016	2017	2018	2019	2020	2021	2022
Ⅰ	0–15	Bare ground, very low vegetation cover	3164	3090	3119	3167	3016	3103	3127
Ⅱ	15–30	Low vegetation cover	166	199	182	176	247	181	194
Ⅲ	30–50	Medium vegetation cover	71	100	95	67	121	91	87
Ⅳ	≥50	High vegetation cover	27	39	32	18	44	53	20
Total area of vegetation cover			264	338	309	261	412	325	301

, the vegetation cover of the Dariyabui Oasis showed a fluctuating increasing trend from 2016 to 2022, with the smallest area covered by vegetation in 2019 and the largest area covered by vegetation in 2020. The areas of bare ground and very low cover show a decreasing trend, and the areas of low, medium, and high cover show an increasing trend from 2016 to 2022, in which a fluctuating increase in low cover is relatively distinct.

To further analyse the changes in the vegetation cover of the oasis, the absolute change in vegetation cover from 2016 to 2022 was calculated using the image difference method, where a negative difference indicates a decrease, a zero difference indicates no change, and a positive difference indicates an increase. The results are shown in Figure 7a–f, with relatively stable changes in the centre of the oasis and significant changes in the periphery. In terms of the time series, oasis vegetation cover mainly increased during 2016–2017, with a localised decrease in the southwestern region. From 2017 to 2018, the oasis cover increased in the southwest and southeast of the oasis and decreased in the central area. The decrease in the oasis cover predominantly occurred during 2018–2019, with a small, localised increase in cover in the central part of the oasis. A significant increase in oasis cover was present in the southwest and southeast of the oasis and a localised decrease in cover in the vegetation northwest during 2019–2020. In 2020–2021, the central and southwestern parts of the oasis were dominated by an increase in oasis cover, whereas the southeastern region experienced a decrease in oasis cover. The decrease in oasis cover from 2021 to 2022 was pronounced, with small sporadic increases in the central north.

Slope analysis was used to forecast the time-varying trend of the FVC in the Dariyabui Oasis from 2016 to 2022. As shown in Figure 7g, a significant decrease in FVC was present in the central part of Dariyabui green to the north and east and a mild decrease in FVC in the eastern fringe of the oasis. A mild increase in FVC was present in the southwest of the oasis, with significant increases in FVC in the southeast and northwest.

As shown in

Table 2. Area and proportion of vegetation cover change in the Dariyabui Oasis.

Trend Rating	Area (km2)	Proportion (%)
Significant reduction	47.17	1.38
Mild reduction	67.70	1.97
Basic stability	3136.97	91.51
Mild increase	83.88	2.45
Significant increase	92.28	2.69
Total area	3428.00	100

, the area changes in vegetation cover in the Dariyabui Oasis from 2016 to 2022 were stable, exhibiting the largest area and percentage. Based on the trend of area change, the percentage of the increasing area was greater than that of the decreasing area. The percentage of oases that were essentially stable was 91.51%, showing a decrease of 3.35% and an increase of 5.14%.

To further analyse the changes in Dariyabui Oasis cover, state transfer matrix operations were performed for different classes of vegetation cover in 2016–2022. This analysis was based on Sentinel-2 using the vegetation index inversion of vegetation cover, and ArcGIS software was used for reclassification. Raster to surface was then used to obtain vector data, the fusion tool was used to obtain the statistical area, and the intersection tool was used to obtain the change area. Finally, data were exported to Microsoft Excel to draw the pivot table (

Table 3. Vegetation cover class transfer matrix for the Dariyabui Oasis, 2016–2022.

Year	Type	2016
		Bare Ground, Very Low Cover		Low Coverage		Medium Coverage		High Coverage
		Area km2	%	Area km2	%	Area km2	%	Area km2	%
2022	Bare ground, very low cover	3078	97.29	44.73	26.94	3.36	4.71	0.57	2.13
	Low coverage	75.75	2.39	91.77	55.28	22.46	31.46	2.47	9.21
	Medium coverage	9.16	0.29	27.54	16.59	37.34	52.3	12.82	47.84
	High coverage	0.61	0.02	1.96	1.18	6.23	8.73	10.94	40.8
Amount of change		−36.80		26.40		17.47		−7.06

). The vegetation cover of the Dariyabui Oasis in 2016–2022 mainly shifted from bare ground and very low cover to low cover, from low cover to medium cover, and from high cover to medium cover. An area of 75.75 km² was converted from bare ground with very low cover to low cover, accounting for 2.39% of the total amount of bare ground with very low cover. An area of 27.54 km² was converted from low cover to medium cover, accounting for 16.59% of the total low-cover area. An area of 12.82 km² was converted from high cover to medium cover, accounting for 47.84% of the total high-cover area. In terms of absolute change, the amount of bare ground with very low-cover, low-cover, medium-cover, and high-cover area converted was −36.80 km², 26.40 km², 17.47 km², and −7.06 km², respectively.

3.3. Spatial and Temporal Groundwater Distribution Patterns

The measured groundwater burial depth from 19 wells in the Dariyabui Oasis was used to calculate the average monthly groundwater burial depths in September from 2019 to 2022. Furthermore, the IDW method was used in ArcGIS to obtain a contour map of groundwater burial depths in the study area for September (Figure 8).

The spatial variation in groundwater in the Dariyabui Oasis and the depth of groundwater in the oasis gradually increased from south to north, with the minimum depth at the inlet of the oasis in the southwestern part of the oasis and the maximum depth at the northeastern end of the oasis. The main cause of changes in the groundwater depth is groundwater recharge from the runoff of the Kriya River, which enters the oasis from the southwest during the flood season. In the runoff process to the north of the oasis, the amount of recharge gradually decreases, creating a spatial distribution pattern of groundwater depth in the oasis, which is shallow in the south and deep in the north. The distribution of the water systems in the oasis was analysed by dividing the oasis into three parts: east, central, and west. The depth of groundwater in the west was lower than in the central and eastern parts of the oasis; furthermore, according to the investigation of the study area, the villagers in the oasis divert water to the west using the retaining dams all year round to cultivate traditional Chinese medicinal herbs (Cistanche), which leads to recharge conditions in the western part of the oasis that are better than those in the central and eastern parts.

Figure 9 shows that the monthly average depth of the groundwater in the Dariyabui Oasis fluctuated. The trend in the average monthly depth of burial over the 4-year period (2019–2022) was consistent, with the minimum depth of groundwater in the oasis occurring in March–April and the maximum depth of groundwater occurring in October–December of each year. According to monitoring data from the groundwater burial depth observation wells along the Kriya River, the groundwater burial depth decreases in March–April every year. This decrease indicates that floods pass through the river channel during that period and that the Dariyabui Oasis receives surface water recharge regularly from March to April. After September, the Kriya River enters a dry period, with no water or only a little runoff present within the river, and the recharge of groundwater to the oasis is accordingly reduced. Furthermore, at that time, the vegetation is in a growth period in which is it continuously consuming water, so the depth of groundwater is increasing. Owing to the change in burial depth during the year, relatively small fluctuations in the groundwater depth were present in June–July and October–December. Thus, we concluded that the groundwater depth was maintained at a shallow depth and fluctuated less in June–July due to high recharge during the flood period. Furthermore, June–July was the peak water depletion period for vegetation growth, whereas the Kriya River was in a dry period from October to December, when the vegetation entered a low water consumption state during the nongrowing period, resulting in a higher and more stable groundwater depth. In terms of interannual changes in burial depth, the average annual burial depth of groundwater in 2019, 2020, 2021, and 2022 was 4.0187, 4.2220, 4.1092, and 4.0148 m, respectively, with a slight annual increase observed from 2020 to 2022. The interannual fluctuation in the groundwater depth in the oasis was relatively smooth.

3.4. Groundwater Dynamics in Relation to Changes in Vegetation Cover

Figure 10 shows the superimposed map of groundwater burial depth contours and vegetation cover distribution for September in 2019–2022. The depth of groundwater in the Dariyabui Oasis ranged from 0 to 8 m. In terms of spatial distribution, the vegetation cover in the oasis decreased with increasing depth to groundwater from southwest to northeast, with a threshold depth of <7 m for low vegetation, <6 m for medium vegetation, and between 2 and 6 m for high vegetation. Furthermore, in the northeastern area of the oasis, where the depth of groundwater is >7 m, the main vegetation is poplar with a small amount of tamarisk, but a large area of wilting and death also exists (Figure 2). This is close to the threshold value of groundwater burial depth affecting vegetation growth in the arid zone [25,61,62]. The results showed that the suitable threshold interval of groundwater burial depth for vegetation in the Dariyabui oasis is 0–6 m.

Groundwater data for September 2019–2021 and vegetation cover data for the corresponding years were selected for analysis. ArcGIS was used to generate a raster map of vegetation cover and groundwater burial depth with a resolution of 90 m, and raster turning points and values were extracted to exclude invalid data and outliers and obtain valid raster data: 25,096 pairs of data in 2019; 25,110 pairs of data in 2020; 25,131 pairs of data in 2021; and 25,261 pairs of data in 2022. Using Origin software, a scatter plot of vegetation cover versus groundwater burial depth was obtained, as shown in Figure 11a–d. The scatter plot reflects the correlation between groundwater burial depth and vegetation cover. A certain regularity was present between vegetation cover and groundwater burial depth; within a certain interval of groundwater burial depth, a negative correlation was present between groundwater burial depth and vegetation cover.

Taking the groundwater burial depth of 0.2 m as an interval, the mean value of vegetation cover within the interval and the threshold value of 95% as the larger value of vegetation cover were counted, respectively, to obtain the fold plot in Figure 11a1–d1. For the average and larger values of the oasis vegetation cover, the overall trend with an increase in the depth of the groundwater burial vegetation cover decreased, and the maximum value of the vegetation cover occurred at the depth of the groundwater burial 2–4 m interval.

The oasis groundwater burial depth and vegetation cover were further analysed using cubic curve regression to obtain the following fitted equations for 2019 to 2022:

f_e1 = 10.2735 + 0x − 66.6314x² + 28.3644x³ (1.0 ≤ x < 8) R² = 0.46

(7)

f_e2 = 9.8642 + 47.6233x − 936.8025x² + 0x³ (1.0 ≤ x < 8) R² = 0.81

(8)

f_f1 = −3.5434 +120.7277x − 358.7674x² + 287.3367x³ (1.8 ≤ x < 7.6) R² = 0.68

(9)

f_f2 = −3.5685 + 285.8194x − 2259.3774x² + 4862.0011x³ (1.8 ≤ x < 7.6) R² = 0.59

(10)

f_g1 = 4.7449 + 33.6876x − 95.6626x² + 45.2605x³ (2 ≤ x < 7.8) R²= 0.69

(11)

f_g2 = 6.8606 + 18.3331x + 0x² − 1860.8400x³ (2 ≤ x < 7.8) R² = 0.86

(12)

f_h1 = −3.4597 + 1.1667x − 0.0375x² + 0.0003x³ (1.2 ≤ x < 7.4) R² = 0.53

(13)

f_h2 = −26.9488 + 10.0060x − 0.9203x² + 0.0251x³ (1.2 ≤ x < 7.4) R² = 0.65

(14)

where f_e1, f_f1, f_g1, and f_h1 correspond to the maximum value of oasis vegetation cover in 2019, 2020, 2021, and 2022, respectively, and f_e2, f_f2, f_g2, and f_h2 correspond to the average value of oasis vegetation cover in 2019, 2020, 2021, and 2022, respectively. x is the burial depth of groundwater in m.

From the above equations, cubic curve regression of the larger value of $FVC$ with the burial depth of groundwater in the Dariyabui Oasis was performed. The fitted curve correlation was highest in 2021 (R² = 0.69) and lowest in 2019 (R² = 0.46). A curve regression of mean $FVC$ values against oasis groundwater burial depth showed the highest fit correlation in 2021 (R² = 0.86) and the lowest in 2022 (R² = 0.59). Overall, the relationship between the mean $FVC$ values and groundwater burial depth was better than the fit curve between the larger values and groundwater burial depth.

To analyse the relationship between vegetation and groundwater burial depth in the Dariyabui Oasis, the number of image elements of different types of vegetation was counted at groundwater burial depth intervals of 1 m. As shown in

Table 4. Depth of groundwater and vegetation type in the Dariyabui Oasis.

Type Number
Year	Depth of Burial (m)	Bare Ground, Very Low Cover	%	Low Coverage	%	Medium Coverage	%	High Coverage	%
2019	0–1	4	0.02	0	0	0	0	0	0
	1–2	357	1.79	91	2.52	24	1.92	8	3.35
	2–3	2801	14.01	592	16.38	233	18.63	55	23.01
	3–4	4693	23.47	907	25.10	303	24.22	49	20.50
	4–5	3366	16.84	593	16.41	209	16.71	36	15.06
	5–6	4010	20.06	715	19.78	253	20.22	40	16.74
	6–7	3559	17.80	574	15.88	189	15.11	38	15.90
	7–8	1202	6.01	142	3.93	40	3.20	13	5.44
2020	0–1	0	0	0	0	0	0	0	0
	1–2	20	0.12	11	0.24	1	0.04	2	0.27
	2–3	493	2.88	130	2.79	63	2.45	14	1.89
	3–4	4437	25.90	1419	30.41	833	32.40	243	32.75
	4–5	4016	23.44	1091	23.38	568	22.09	174	23.45
	5–6	4880	28.49	1376	29.48	770	29.95	216	29.11
	6–7	2756	16.09	553	11.85	302	11.75	82	11.05
	7–8	528	3.08	87	1.86	34	1.32	11	1.48
2021	0–1	0	0	0	0	0	0	0	0
	1–2	82	0.44	24	0.65	6	0.30	5	0.51
	2–3	1653	8.95	410	11.12	258	12.87	119	12.16
	3–4	5913	32.03	1401	38.00	769	38.37	363	37.08
	4–5	3222	17.45	590	16.00	326	16.27	159	16.24
	5–6	4511	24.44	931	25.25	462	23.05	253	25.84
	6–7	2368	12.83	271	7.35	157	7.83	67	6.84
	7–8	711	3.85	60	1.63	26	1.30	13	1.33
2022	0–1	0	0	0	0	0	0	0	0
	1–2	1108	5.71	260	7.24	148	8.06	34	7.73
	2–3	3679	18.97	827	23.04	456	24.82	101	22.95
	3–4	4167	21.49	756	21.06	391	21.28	93	21.14
	4–5	2944	15.18	474	13.20	210	11.43	56	12.73
	5–6	4798	24.74	894	24.90	466	25.37	113	25.68
	6–7	2432	12.54	339	9.44	157	8.55	39	8.86
	7–8	266	1.37	40	1.11	9	0.49	4	0.91

, the cumulative proportions of low-cover vegetation areas with groundwater depths of less than 7 m in 2019, 2020, 2021, and 2022 were 96.07%, 98.14%, 98.37%, and 98.74%, respectively. The groundwater depth where low cover vegetation occurs most was 3–4, 5–6, 3–4, and 5–6 m for 2019, 2020, 2021, and 2022, respectively. This indicates that the limiting value of the groundwater burial depth for low-cover vegetation is 7 m and that the groundwater burial depth at which low-cover vegetation occurs with the greatest distribution is 3–6 m. The maximum percentages of mid-cover vegetation occurrences were 24.22%, 32.4%, 38.37%, and 25.37% in 2019, 2020, 2021, and 2022, respectively; these values correspond to a depth of 3–4 m. The cumulative percentage of high-cover vegetation with groundwater depth of <6 m was 78.66%, 87.47%, 91.83%, and 90.23% in 2019, 2020, 2021, and 2022, respectively. The groundwater depth where the highest percentage of high vegetation cover occurred was 2–3, 3–4, 3–4, and 5–6 m for 2019, 2020, 2021, and 2022, respectively. Therefore, the limiting value of the groundwater burial depth for high vegetation cover was 6 m and the groundwater burial depth for the highest distribution of high vegetation cover was 2–4 m. A field survey of the Dariyabui Oasis found that the vegetation in the middle of the oasis grows readily with high coverage (Figure 2d). At the northern end of the oasis, the surface water of the Kriya River lies at depths of 6–8 m. The dominant species in the northern end, poplar, had a large withering area, with relatively low vegetation coverage (Figure 2c), which agrees with the results of the field survey and statistical analyses.

4. Discussion

4.1. Extraction of Vegetation Cover

The NDVI to extract vegetation cover has been used extensively [2,63,64] and results have been relatively reliable. Based on the actual situation in the study area, using other vegetation indices to invert the vegetation cover of the Dariyabui Oasis has also been reported considering the effects of soil, vegetation patches, water bodies, and other elements. For example, Wang et al. [40] used seven common vegetation indices, including the NDVI, SAV, and DVI, based on UAV and Sentinel-2 images to extract the Dariyabui vegetation cover and determine the most suitable vegetation indices for the extraction of desert vegetation cover. The SAV index was optimal for the extraction of natural vegetation cover in the extreme arid zone. In this study, the NDVI was selected to calculate the vegetation cover of the Dariyabui Oasis; however, in a follow-up study, the vegetation cover will be inverted with different vegetation indices based on the results from the study area, and its correlation with the depth of groundwater burial will be explored.

4.2. Influence of Groundwater Depth on Vegetation Distribution Patterns

Desert vegetation distribution patterns in arid and semi-arid regions are influenced by changes in groundwater dynamics [4]. In the arid desert hinterland of the Dariyabui Oasis, vegetation growth largely depends on groundwater and seasonal flooding of the Kriya River [65]. Numerous scholars have analysed the relationship between vegetation change and groundwater burial depth in Ejin and Minqin based on satellite remote sensing data and long time series of groundwater burial depth monitoring data [25,31]. The research has shown that vegetation is negatively correlated with the underwater burial depth and that vegetation is controlled by the groundwater burial depth in the long term. In the present study, we calculated the FVC of the Dariyabui Oasis based on Sentinel-2 remote sensing images and fitted it to the groundwater burial depth in cubic curve regression. Subsequently, we calculated the coefficient of determination of the mean value of the FVC and the depth of groundwater burial for 2019–2022 (R² = 0.59–0.86). The results of the current study showed that the correlation between vegetation and groundwater in the Dariyabui Oasis is a close match to the results reported for other arid and semi-arid regions. However, the Dariyabui Oasis is a natural and pristine oasis in the hinterland of a desert, which is virtually unaffected by anthropogenic influence. Annual precipitation is <10 mm [43] and groundwater is the only source of water for the growth of vegetation in the oasis [36], so the sensitivity (tolerance) of oasis vegetation to the depth of groundwater is higher than that in other arid-region oases.

The groundwater data and Sentinel-2 remote sensing imagers used in this study have a shorter time series than those used in similar studies in other regions and did not adequately reflect the dynamic changes in groundwater or vegetation cover over a longer time series in the oasis. Therefore, further research is required into the relationship between vegetation cover response and groundwater burial depth in the desert hinterland of the Dariyabui Oasis by selecting remote sensing data with a longer time series, as well as establishing a long-term groundwater simulation prediction model through continuous observations and groundwater simulation prediction models [66,67], in addition to methods.

4.3. Climatic and Anthropogenic Impacts on Vegetation Distribution Patterns

Changes in vegetation distribution patterns are influenced by factors such as temperature, precipitation, light, surface water, groundwater, soil, and human activities [25,31]. The Dariyabui Oasis is located in the hinterland of an extremely arid desert, with very little precipitation and extreme evaporation [65]. Floods and the infiltration of floodwater recharging the groundwater during the flood season are the only water sources sustaining the life of the oasis. Moreover, variation in the amount of water flowing into the oasis from the Kriya River is a determining factor in the growth of the local vegetation [36]. Therefore, the effect of precipitation on vegetation is minimal. In this study area, a medium-sized meteorological station has been established; however, it ceased to work in 2017 due to the harsh climatic conditions of the desert hinterland damaging the equipment, although no meteorological data were present in the same time series as the vegetation cover. Therefore, the impact on the distribution pattern of the vegetation cover was not analysed from a meteorological viewpoint. For subsequent studies, established meteorological stations should be repaired or new stations should be constructed to comprehensively analyse the influence of other factors on oasis vegetation distribution patterns.

Fieldwork showed that the population of the Dariyabui Oasis is relatively stable. In 2017 and 2019, to improve the living environment of the villagers and to protect the ecology of the natural oasis, the villagers were relocated to the middle reaches of the Kriya River. Now, the population of Dariyabui is mainly concentrated in two locations (Figure 12): the new village has a population of approximately 700 people (Figure 12b) and the old village has a population of approximately 100 people (Figure 12a). The villagers are mainly engaged in animal husbandry and herbal medicine (Cistanche) cultivation. Villagers often artificially control the floodwater of the Kriya River during the flood season using temporary earth dams, and the depth of groundwater in the western part of the oasis is shallower than that in the eastern part [37]. As shown in Figure 7g, the area with a considerable increase in vegetation cover was in the northwestern part of the oasis. The construction of the Giyin Reservoir in the upper reaches of the Krya River in 2017, which intercepts the renewal Krya River floods, regulates the ecological water coming from the Tail Coccyx Oasis to a certain extent. Thus, anthropogenic activities also affect the vegetation distribution pattern of the natural oasis.

5. Conclusions

This study shows the spatiotemporal distribution pattern of vegetation and groundwater burial depth in the Dariyabui Oasis for 2016–2022 and 2019–2022, respectively. Furthermore, the spatiotemporal response between the two factors was established in the desert hinterland to provide recommendations and theoretical support for the conservation of vegetation and groundwater ecosystems of the natural oasis. The spatial and temporal distribution patterns of vegetation cover according to the thresholds of water burial depth correspond to differences in vegetation cover in the oasis. From our results, we offer the following specific recommendations. Firstly, the growing period of the oasis vegetation in the upstream reservoirs should be considered to regulate the Keriya River water. This will allow the recharge of oasis groundwater so that the oasis vegetation in dry seasons has access to more groundwater and, thus, improved vegetation growth and health. Secondly, the groundwater in the northern part of the oasis is deep, and only floodwater from the Kriya River can reach the northern part of the oasis during the flood season. Furthermore, the oasis is located in the hinterland of the desert where evaporation is extremely strong. The construction of channels (especially culverts) can reduce evaporation and seepage along the course. This would allow surface water from the Keriya River to reach the northern part of the oasis for groundwater recharge in all seasons except the flood season. This would also reduce the depth of groundwater in the northern part of the oasis and improve vegetation health.

The following conclusions were drawn from this study. The overall trend in the FVC in the Dariyabui Oasis from 2016 to 2022 was slightly increasing, with fluctuating decreasing trends in bare ground and very low cover, and fluctuating increasing trends in low, medium, and high cover, where the fluctuating increase in low cover was relatively significant. The depth of groundwater in the oasis showed a pattern of gradual increase from the entrance to the end of the oasis, that is, south to north.

The depth of groundwater in the Dariyabui Oasis showed a distribution pattern of low in the south and high in the north, and the seasonal dynamics of the depth of groundwater in the oasis changed markedly, with the depth in the spring being low and the depth in the autumn being high.

Finally, vegetation cover decreased with increasing groundwater depth within a certain range of oasis depth. Medium and high vegetation cover in the Dariyabui Oasis occurred mainly in the central part of the oasis, where the groundwater was relatively shallow, and in the northern part of the oasis, where the groundwater was high and the vegetation cover was low.