Three Decades of Oasis Transition and Its Driving Factors in Turpan–Hami Basin in Xinjiang, China: A Complex Network Approach

Abstract

As a predominant ecosystem-providing area and distinctive landscape in arid regions, an oasis plays an important role in maintaining land stability, human production, and daily activities. Studying the dynamics of oasis and its driving factors is vital to supporting arid regions’ sustainable development. As a typical mountain–desert–oasis landscape, the Turpan–Hami (Tuha) Basin, located in Xinjiang, China, includes complex interactions among different land types. For this study, we revealed the spatio-temporal patterns and transition processes of the oasis using a complex network method between 1990 and 2020 in the Tuha Basin. In the oasis transition network, the degree value, betweenness centrality, and average path length were calculated to express the transition relationship, key oasis type, and oasis structural stability, respectively. Six factors related to climate change and human actives were selected to investigate the driving forces behind oasis transitions, including the average temperature and precipitation in the growing season, the total power of agricultural machinery (TAMP), the production of raw coal (PRC), the total output value of the plantation industry (TPI), and the population (Pop). Our results show that the oasis area of the Tuha Basin, including the natural oasis and artificial oasis, all grew from 1990 to 2020, with the natural oasis expanding more than the artificial oasis. The transitions between oasis types became more frequent as the area of oasis land types increased throughout the study period. Grassland acted as the most important oasis type in the network, with the highest betweenness centrality, but its importance declined due to the increasing complexity of the oasis transition network from 1990 to 2020. The transitions between oasis types became simpler, and the oasis structural stability decreased. Through driving force analysis, the oasis changes showed a positive correlation with the temperature (p-value < 0.05, r = 0.88), and urbanization and industrialization factors prompted transitions to built-up areas and cropland from grassland and shrubland. In summary, our results suggest that to create a harmonious symbiotic relationship between the natural environment in dryland and human activities, preventing grassland degradation and excessive reclamation of land cover is an available way. Meanwhile, the protection of shrubland and water resources is also important. This study provided reference and theory support for promoting the sustainable development of oases.

1. Introduction

Oases are the most productive, dynamic, and vulnerable ecological systems in drylands and supply stable water resources, basic resources, and the conditions for human survival and living conditions [1,2,3,4]. Oases are also crucial ecological barriers in drylands, defeating desertification, salinization [5], and sandy weather [6]. Since the 1950s, artificial oases have expanded due to land reclamation and agriculture in northwest China [7], which has contributed to sufficient and stable food production and economic growth [8] and has caused a series of environmental problems. Because of the extreme drought weather and limited water supply, oases have experienced groundwater reduction, desertification expansion, and grass degradation, which have deeply influenced the stability of oasis systems. Therefore, it is necessary to quantify the oasis spatio-temporal patterns and reveal the driving factors behind oasis variation.

Previous studies have explored spatio-temporal oasis patterns and the interactions between oases and various factors (e.g., natural and human factors). Land-use dynamic models and landscape pattern indexes are common methods used to quantify and forecast oasis dynamic changes [9,10,11,12]. For example, Amuti and Luo [13] analyzed Hotan oasis temporal changes, using land-use/land-cover (LULC) data, and the main driving factors of oasis changes between 1990 and 2008 and found that intensified human activities reduced the desert–oasis ecotone. Tan et al. [14] analyzed the oasis changes in Zhangye city from 1980 to 2015, and they simulated future oasis changes in 2030 by applying a future-land-use simulation model (FLUS), LULC data, and geographic and socioeconomic data; the authors found that the main transfers of oasis transitions existed in cropland, built-up areas, water bodies, and desert from 2015 to 2030. These methods mainly describe oasis transitions from a specific land type to another in terms of type and quantity, ignoring the changing processes in the entire oasis transition system [15]. Moreover, in past studies, the identification of key land types has mostly relied on the extent of the changed area and the change rate, with little consideration of the influence of the land type alone on the entire process of oasis transitions. For oasis stability, many works have focused on the ecological perspective [4,16,17], but structural stability also needs further exploration.

Therefore, a complex network model was introduced in this paper to analyze the dynamic processes within oasis types. As a theoretical method, a complex network is composed of a large number of interconnected nodes and the edges connected to these nodes, performing substantial non-trivial topological features of the target system and quantifying the interactions among the sub-systems [18,19]. It has been demonstrated that the complex network is an effective tool in modeling and analyzing land-use change processes (e.g., urbanization, afforestation/deforestation) and identifying the key land types during those processes. Xu et al. [20] studied the complex relationships in the procession of LULC using a complex network, concluded that grassland is an essential land type for the transition process, and calculated the proper scale of each land type for maintaining artificial land stability. Zhang et al. [4] constructed a complex network from the landscape index and patch-type index, and evaluated the stability of the landscape and network, in the Pingshuo opencast mining area in China between 1986 and 2015. The authors found that the landscape heterogeneity tended to be unstable, and the network length became shorter, demonstrating more human activities in mining areas during the study period. So far, the applications of complex networks in investigating oasis transitions and stability are scarce, especially on a large, long-term scale.

Considering the driving force behind oasis variations, many studies have a similar conclusion that the oasis variation is mainly impacted by human and natural factors [15,20,21,22], and present studies have paid more attention to human factors. Cai et al. [21] found that irrigation, population growth, and the farmers’ income were important factors that affected crop expansion in Xinjiang. Wang et al. [23] believe that there exist associations among the proportion of the tertiary industry, rural capita net income, and the fraction of forest. Liu et al. [12] also found that human factors are the main driving force of landscape pattern changes in Jinghe County. However, human factors are multiple, complex, and adapt to local conditions [24]. Thus, this study considered the characteristics of the Tuha Basin and evaluated the driving force behind oasis variation.

In this study, we conducted oasis transition processes and structural stability investigations using the constructed oasis transition network and analyzed the driving factors in Tuha Basin between 1990 and 2020. Our research aimed to (1) identify the spatio-temporal patterns for different oasis types, (2) simulate the oasis transition processes and identify the key oasis types, (3) rate the oasis structural stability, and (4) discover the main driving factors behind natural and artificial oases.

2. Materials and Methods

2.1. Study Area

The Tuha Basin is the abbreviation of the Turpan and Hami Basin; the basin is located in eastern Xinjiang, China (86°44′55″–96°25′0″E, 40°49′13″–45°3′21″N, Figure 1), which comprises two districts (Gaochang and Yizhou) and four counties (Tuokexun, Shanshan, Balikun, and Yiwu). With an average elevation of 1173.36 m and an elevation drop that can reach 5538 m, the Tuha Basin comprises high mountains in the north and a low basin in the south.

The Tuha Basin has an uneven distribution of precipitation and temperature. Rainfall occurs primarily in the mountain areas between April and September annually. Annual precipitation ranges from 0 mm/a in the plain areas to 46.5 mm/a in the mountain areas, with an average of 6.12 mm/a, while evaporation can reach 3000 mm, causing a typical arid climate, with the water resources mostly relying on the snow meltwater from the mountains and groundwater. The highest temperature in the plain areas can reach 44.3 °C; in the same period, this is only 30.1 °C in the mountain areas. In permeant snow-covered areas, the average temperature is below −2.4 °C.

We categorized the oasis systems into natural and artificial oasis subsystems according to the climate differences, plant community function, and oasis function [25,26]. Natural oases are defined as areas that hold ecological stability and balance and provide basic protection for biodiversity and desertification control [27]. Artificial oases are defined as places that are suitable for human living and productive activities in arid/semi-arid areas [28]. In this study, oases are situated in the plains and plateaus of the Tuha Basin, with an elevation range from −124 to 2358 m. There is a small percentage of oases embedded in river basins formed by the meltwater from snowy mountains.

2.2. Datasets

Oasis data were extracted from GLC_FCS30 in Earth System Science Data [29], with a 30 m spatial resolution and a 5-year interval from 1990 to 2020. GLC_FCS30 data classified land types into cropland, forests, shrubland, grassland, wetlands, built-up areas, bare land, water bodies, and permanent ice and snow. Because of their tiny coverage in the Tuha Basin, wetlands were merged with water bodies. LUCC data accuracy, in 2015 and 2020, was validated through high-resolution images from Google Earth; the overall accuracy and kappa coefficient were selected as the accuracy metrics in this study. In total, 260 sample point coordinates (50 for cropland, 30 for each other land type) were randomly taken from Google Earth high-resolution images in 2015 and 2020, respectively. The validation results show that the overall accuracy reached 91.02% and 94.46% for the LUCC data in 2015 and 2020, and that of the kappa coefficient is 89.69% and 93.81%, respectively. These data were developed through continuous change detection globally, and then an optimization algorithm was applied to allow local improvement, so the other difficult-to-validate data (after 2015) used in this study were assumed to be of similar accuracy. These data were also validated in the Xinjiang range in other studies and were rated as having high accuracy [30].

Compared with the apparent vertical zonality of mountainous areas, oases have intrazonal characteristics [15,31]. Therefore, in this study, the areas with an elevation higher than 2500 m, or between 500 m and 2500 m with a slope greater than 25°, were regarded as mountainous areas and masked out [32,33,34].

Oasis dynamics are mainly affected by natural and human factors. Natural factors are characterized by wide ranges and a long duration, but they have minor short-term effects on oasis dynamics [8,15]. Two natural factors, precipitation and temperature, were collected from the National Tibetan Plateau Science Data Center (https://data.tpdc.ac.cn, accessed on 5 April 2023) with 1 km spatial resolution throughout the growing season (from May to July). Human factors had further impacts on oasis dynamics, but they were limited to local ranges [10,13]. Four human factors—total power of agricultural machinery, total output value of the plantation industry, production of raw coal, and population—were acquired from the Xinjiang statistic yearbook, the data of which were collected during the same periods as the LULC data.

2.3. Complex Network Construction

Complex networks are intricate structures composed of multiple interconnected sub-systems, with interactions occurring among them [19,20]. The complexity of these networks is affected by the number of sub-systems (nodes) involved and the nature of their interactions (edges) [35]. Among complex network theories and related studies, network node properties and topology attributes are of the most interest, and these properties are explored from degree, path, and clustering effect dimensions in general. Thus, many methods and indicators comprehensively or specifically concretize these properties from these dimensions, such as the degree distribution, path length, and clustering coefficient [36]. Within these dimensions, the clustering effect mainly occurs in the network with affiliation relationships or with the community structure [18,19]. However, for the LUCC transition relationship, each land type is considered as independent at the first classification level. Thus, the clustering dimension will not be discussed. In this study, all land types are denoted as nodes, and their interactions are represented as the edges. This model quantifies the relationship among nodes and facilitates the analysis of the entire network system [37].

The flowchart of the oasis transition network construction in this study is shown in Figure 2. The entire oasis transition process was viewed as a complex network, and land types played the roles of sub-systems. The transformation relationships between land types were regarded as the network edges. Depending on the edge direction and the weight variation, a complex network can be categorized into directional and non-directional networks, as well as weighted and non-weighted networks, based on the edge attributes. According to the land-use transition matrix in the Tuha Basin, six phases of directional and weighted oasis transition network were created according to the time period, i.e., 1990–1995, 1995–2000, 2000–2005, 2005–2010, 2010–2015, and 2015–2020. The oasis transition process, key land types, and oasis structural stability were then evaluated using three metrics: the degree value, betweenness centrality, and average path length.

2.4. Complex Network Analyses

2.4.1. Oasis Transition

In an oasis transition network, the degree value of one oasis type node refers to the number of edges connected to the node in the network, i.e., the number of land types that have existing transfer relationships with this land type, whether unidirectional or bi-directional. The higher the degree value, the more connections the node has with other nodes [38]. The direction and intensity of the node interactions are important indicators in a directional and weighted network. The intensity mentioned here was expressed through weight, and in this study, the weight was equal to the transferred area between two land types. Thus, considering the transfer direction and weight, the degree can be divided into the weighted in-degree and out-degree in further steps for the indicated node’s transferred area and direction simultaneously. The ratio of the weighted out-degree to the weighted in-degree (named ratio) is employed to determine whether a land type is a transfer-in or transfer-out type. If the ratio is greater than 1, then the land type is defined as a transfer-out type; otherwise, it is a transfer-in type. The in-degree and out-degree can be calculated as

$$K_{n\times n}=A_{n\times n}·B_{n\times n}^T$$

(1)

where $A$ is the adjacency matrix of the network. If there are any connections between the $i$th node and $j$th node, $A_{ij}=1$; otherwise, $A_{ij}=0$. Here, $i$, $j$ $\in$ N, n represents the number of nodes in a network, and when $i=j$, $A_{ij}=0$. $B$ is the transition matrix. $k_{ij}$ is the element in $K$; if $i>j$, $k_{ij}$ represents the weighted out-degree from the $i$th node to the $j$th node. Otherwise, $k_{ij}$ represents the weighted in-degree from the $i$th node to the $j$th node. The number of $k_{ij}\ne 0$, when $i>j$ is equal to the out-degree of the $i$th node; otherwise, it is equal to the in-degree of the $i$th node.

2.4.2. Identification of Key Land Types

Betweenness centrality is used to identify the key node in a complex network, which refers to the ratio of the number of shortest paths that pass through a node to the number of all the shortest paths in the network. The shortest path refers to the node sequence that can connect two nodes at a minimum cost, and the cost refers to the weighted sum [39,40]. The node with a higher betweenness centrality has a greater influence on information circulation. In the oasis transition network, a land type with a higher betweenness centrality has more control over the whole transition network, because more transition relationships will rely on this node. Thus, the land type in the transition network with a higher betweenness centrality will be considered as the key land type, which plays a vital role in affecting the interactions of land types and the oasis transition network vitality. The betweenness centrality can be calculated as

$$bc(i)={\displaystyle \sum _{j\ne k}{p}_{jk}(i)/p_{jk}}$$

(2)

where $bc(i)$ is the $i$th node betweenness centrality, $p_{jk}(i)$ is the number of the shortest path passing the $i$th node, and $p_{jk}$ is the number of the shortest path in the whole network.

2.4.3. Oasis Structural Stability

The average path length was used to rate a network’s structural stability through measuring the average length of all the shortest paths in a network. This indicator represents the possibility that there are interactions between an arbitrary combination of two nodes in a network [36]. In the oasis transition network, the shorter the average path length, the higher the transfer efficiency is among the oasis types, but it is a more unstable oasis structure. The average path length is calculated as

$$L=\frac{1}{N(N-1)}{\displaystyle \sum _{i\ne j}{d}_{ij}}$$

(3)

where $L$ is the average path length in the network. $N$ refers to the number of nodes in the network. $d_{ij}$ represents the length of the shortest path between nodes $i$ and $j$.

2.5. Driving Forces for Oasis Transition

Oasis transition driving forces were explored from natural and human perspectives. Hence, two natural indicators, including temperature and precipitation, and four human indicators, including total power of agricultural machinery, total output value of the plantation industry, production of raw coal, and population, were selected to explore the driving factors behind oasis transitions. Pearson’s correlation coefficient (r) was used to express the linear correlation between two variables [41]. In this study, we quantified the driving forces between driving factors and oasis changes using the r-value. The r-value can be expressed as

$$r=\frac{{\displaystyle \sum _{k=1}^n(X_k-\overline{X})}(Y_k-\overline{Y})}{\sqrt{{\displaystyle \sum _{k=1}^n{(X_k-\overline{X})}^2}}\sqrt{{\displaystyle \sum _{k=1}^n{(Y_k-\overline{Y})}^2}}}$$

(4)

where $X_k$ and $Y_k$ represent the oasis type and the area at the $k$th year. $n$ is the number of samples. $\overline{X}$ and $\overline{Y}$ represent the mean value vectors of $X$ and $Y$, respectively. $r$ is the correlation coefficient between $X$ and $Y$.

3. Results

3.1. Oasis Spatio-Temporal Change Patterns

Subjected to geographical, meteorological, and human factors, there is an obvious difference between natural oases and artificial oases in spatial distribution variation (Figure 3); natural oases mostly occur by Tianshan Mountain and show a sparse-to-dense trend from west to east. There is also a certain proportion of natural oases scattered around the periphery of artificial oases. Artificial oases are mainly distributed in the plain area of the Tuha Basin, which has a flatter topography, stable water resources, high population density, and is well developed.

The area and proportion of the oasis area in the Tuha Basin for different oasis types were counted from LULC data between 1990 and 2020 in the Tuha Basin. As shown in Figure 4, the oasis in the Tuha Basin expanded between 1990 and 2020, with the area increasing from 14,346.43 km² to 19,977.09 km².

Natural oases expanded the most, with the area increasing from 11,219.26 km² to 15,922.68 km², and increased by 29.54%. Within natural oases, shrubland contributed the most expansion, with the area increasing from 1121.93 km² to 4756.68 km² and from 2.57% to 29.87% in natural oasis areas in 1990 and 2020. The area of grassland accounted for the largest proportion of natural oases but showed insignificant variations in area (increasing from 10,714.67 km² in 1990 to 10,958.13 km² in 2020) and, on average, accounted for 85.35% of the area of natural oases.

Artificial oases also showed a slight rising trend, with the area increasing from 3127.17 km² to 4054.41 km², and increased by 22.87%. Specifically, cropland accounted for the greatest area of artificial oases, the area changing from 2914.63 km² to 3597.91 km² but decreasing from 93.17% to 88.74% with respect to the area of oases in 1990 and 2020. Artificial oasis expansion was mainly attributed to the increase in built-up areas; with the area increasing from 213.53 km² to 456.50 km², the proportion with respect to artificial oases increased from 6.83% to 11.26%.

According to the above results, artificial oases are distributed more intensively than natural oases in spatial patterns, but the growing speed has remained at a slow level over the last 30 years. Natural oases are growing more observable in both the spatial and temporal dimensions.

3.2. Oasis Transition Process Analysis

The oasis transition process is reflected in Figure 5 through nodes and edges in the oasis transition network. The width of the edges in the subgraphs refers to the converted areas from one node to another. The complexity of the network is decided by the number of nodes and edges. Generally, natural and artificial oases all show the transfer-in type in the oasis transition network. The ratio of natural oases was 0.87, 0.68, 0.88, 0.89, 0.97, and 0.65 in six sub-periods, respectively, indicating that the area of natural oases showed an increasing trend but the increasing rate was decreasing in general. In the 1995–2000 and 2015–2020 periods, the shrubland and forests were both transfer-in types and were the reason for the ratio of natural oases reaching 0.68 and 0.65 in these two periods. These two land types mainly converted from grassland and bare land. The ratio of artificial oases was 1.68, 0.42, 0.85, 0.90, 0.54, 0.85 in the six sub-periods, respectively, indicating the area of artificial oases decreased from 1990 to 1995 and then increased until 2010 obviously, but the increase rate decreased in the same period. Afterward, the area of artificial oases was increased again and finally held on to a stable level in 2020. In the 1990–1995 period, the artificial oases were the transfer-out type, mainly due to part of the proportion of cropland being converted to bare land and grassland. In 1995–2000 and 2010–2015, the area of artificial oases expanded significantly, owing to the areas of cropland and the built-up area increasing simultaneously, and the built-up area hardly converted to other land types. During 1990–2000, encouraged by the autonomous government, a series of policies relevant to individuals and collectives to reclaim land was enacted, and these policies stimulated the cropland to expand to a certain degree [42]. Part of the built-up area was converted from cropland, and according to the relevant policy starting from 1988, the state-owned land was allowed to be used for paid-for use, which made the cropland conversion to the built-up area possible [43]. Moreover, the industrialization and urbanization development requirement acerated the built-up area expansion speed during 2010–2013 [44]. The interactions among all the land types became more frequent from 1990 to 2020, with more edges connecting the nodes. Grassland had the most edges, proving to be the most active node in the oasis transition network. The interaction between cropland and shrubland was dramatic during the first three periods from 1990 to 2005 and remained steady in the last three periods. There was no transfer-out of built-up areas to other land types from 1990 to 2000, and the edge of the transfer-in type was substantially wider than the transfer-out one throughout the whole study period.

To reduce the redundancy of the network, the principal component of the network was extracted by calculating the average weight of edges and then selecting edges with an above-average weight and corresponding nodes (Figure 6). After reducing the redundancy of the network, the principal component network showed that the primary conversions occurred among grassland, cropland, shrubland, and bare land, indicating that these land types contributed more to the oasis change and transition. The interaction between grassland and bare land dominated the oasis transition network throughout the whole study period The transition between bare land and shrubland was obvious after 1995 and gradually became more observable. Combined with the area change in shrubland, natural oasis expansion can be attributed more to the shrubland being converted from bare land. The edge weight of the relevant cropland gradually decreased until it was below the average weight, at which point it was eliminated as a primary component in the oasis transition network in the period of 2015–2020, which means that the cropland area variation and transition all stayed at a stable level. Further, the dynamics of the artificial oasis decreased.

Table 1. The node ratio, out-degree, and in-degree in oasis transition network.

Land Type	1990–1995 (Ratio, Out-Degree, In-Degree)	1995–2000 (Ratio, Out-Degree, In-Degree)	2000–2005 (Ratio, Out-Degree, In-Degree)	2005–2010 (Ratio, Out-Degree, In-Degree)	2010–2015 (Ratio, Out-Degree, In-Degree)	2015–2020 (Ratio, Out-Degree, In-Degree)
Cropland	1.75, 5, 4	0.44, 6, 5	0.92, 7, 5	0.96, 6, 4	0.62, 7, 6	0.92, 7, 6
Built-up areas	0.00, 0, 3	0.00, 0, 4	0.00, 2, 5	0.01, 2, 4	0.00, 2, 5	0.02, 3, 6
Forests	1.35, 2, 1	0.57, 3, 2	1.65, 3, 2	0.74, 2, 2	1.64, 3, 3	0.23, 4, 5
Shrubland	0.32, 3, 3	0.17, 4, 3	0.32, 4, 3	0.68, 4, 4	1.10, 5, 3	0.19, 6, 4
Grassland	0.91, 7, 6	0.80, 7, 6	1.09, 7, 7	0.95, 7, 7	0.95, 7, 7	1.15, 7, 7
Water bodies	1.33, 2, 4	3.73, 3, 2	3.51, 4, 3	2.39, 3, 3	0.17, 3, 5	1.34, 5, 5
Permanent ice and snow	4.50, 2, 1	0.00, 1, 3	1.92, 2, 4	0.60, 4, 4	0.48, 4, 3	1.22, 6, 5
Bare land	0.97, 5, 4	1.91, 6, 5	1.19, 6, 6	1.17, 6, 6	1.16, 7, 6	1.65, 7, 7

shows the results of node ratios, out-degree and in-degree from the oasis transition network to quantify the conversions among various oasis types. A ratio less than 1 indicates a transfer-in oasis type in the sub-period. If the ratio declined between 1990 and 2020, the transfer-in area of the oasis type accelerated. A ratio greater than 1 showed a transfer-out oasis type. If the ratio grew between 1990 and 2020, then the transfer-out area of the oasis type sped up.

Natural oases and artificial oases were expressed as transfer-in types from 1990 to 2020, with ratios of 0.46 and 0.52, respectively. Throughout the study periods, grassland had the highest out-degree and in-degree values, and the ratios were steadily moving around 1, implying that it was the most active land type in the oasis transition network. Shrubland had a low ratio and an increasing tendency regarding out-degree and in-degree values throughout the study period. The ratio of shrubland dramatically dropped from 1.10 to 0.19, from 2015 to 2020, indicating the shift from transfer-out to transfer-in. The ratio of forests varied during the study period, and it peaked at 0.23 between 2015 and 2020. Generally, water bodies comprised a transfer-out type, and the ratio fluctuated between 1990 and 2020. The ratio of cropland was a transfer-in type, with the ratio smaller than 1 after 1995. The out-degree and in-degree values were maintained at a high level and changed slightly from 1990 to 2020. Built-up areas had the lowest ratio (0.00–0.02), suggesting that the built-up area was fixed, and it was difficult to convert it into other land types.

The oasis transition between 1990 and 2020 is shown in Figure 7, which illustrates the distribution features of various oasis types in the Tuha Basin. In total, the area of oasis converted to non-oasis was 3940.70 km², and 9570.38 km² of non-oasis was converted to oasis, which demonstrates the oasis expansion and is shown in Figure 4. Specifically, as an almost transfer-in type, the converted area from non-oasis to the artificial oasis was 1418.18 km², distributed in the southeast of Gaochang District, surrounding Yizhou District, and northern Balikun County. Of the natural oasis type, 3520.84 km² was converted to non-oasis, mainly occurring in northern and eastern Balikun. As an important transfer-in oasis, non-oasis converted 8152.20 km² into natural oasis, mainly concentrating in the northern part of Balikun County.

3.3. Key Land Types in Oases

The betweenness centrality was calculated for all nodes in the oasis transition network (Figure 8) to reveal the importance of nodes and identify the key land types in the oasis transition network. During the six sub-periods, grassland had the highest betweenness centrality, with values of 17.67, 11.33, 10.67, 12.17, 7.33, and 3.18, respectively, proving to be the most important type in the oasis transition process. Grassland was transferred in mostly from cropland and bare land and transferred out to shrubland. However, the betweenness centrality showed a descending tendency as the oasis expanded and a more complicated topological relationship among the land types. The second highest betweenness centrality was cropland, with values of 2.67, 3.83, 4.67, 3.33, 4.83, and 2.78, respectively, during the six sub-periods. The betweenness centrality of cropland was close to grassland in 2015–2020, becoming the key land type in the oasis transition network.

3.4. Oasis Structural Stability Analysis

The oasis structural stability was assessed using the average path length for all the sub-periods in the oasis transition network. If the average path length was longer, then the transition between the oasis types was more difficult, and the oasis had stronger structural stability. As shown

Table 2. Average path length of the oasis transition complex network.

Periods	1990–1995	1995–2000	2000–2005	2005–2010	2010–2015	2015–2020
Average path length	1.47	1.39	1.38	1.39	1.32	1.20

, the average path lengths of the oasis transition networks were all shorter than 1.5 in the seven sub-periods, implying the unstable structure of the oasis system. Additionally, the average path length of the oasis transition network descended from 1.47 to 1.20 from 1990 to 2020, which demonstrated the oasis system becoming increasingly unstable.

The average path length decreased by 5.44% between 1990–1995 and 1995–2000 and remained stable between 1995 and 2010. In this period, cropland areas expanded significantly, and many efforts like reclamation were offered. However, the average path length decreased to 1.20 between 2010 and 2020, which indicated that transitions among different land types were easier, and the oasis transition network was more complex.

3.5. Driving Factors for Oasis Changes

The inter-annual variations in driving factors between 1990 and 2020 are displayed in Figure 9. The average monthly temperature in the growing season ranged from 15.64 °C and 22 °C to 20.15 °C and 25.79 °C. The mean temperature in the growing season ranged from 19.22 °C to 23.07 °C, with a slight upward tendency. Precipitation decreased obviously from 1990 to 2020, and variations in the growing season ranged from 1.44 mm and 16.99 mm to that between 2.77 mm and 6.24 mm. In the growing season, precipitation expressed a significant decreased tendency. The four human driving factors—the population, the total power of agricultural machinery, the production of raw coal, and the total output value of the plantation industry—all showed an upward tendency over the study period. The production of raw coal and the total output value of the plantation industry increased dramatically after 2005.

Pearson’s coefficient and t-test were utilized to express the relevance between oasis transitions and driving factors, as shown in Figure 10. Among the natural driving factors, oasis transitions showed a positive correlation with the temperature (p-value < 0.05, r = 0.88) and an insignificant correlation with precipitation. The temperature was positively related to forests (p-value < 0.01, r = 0.94) and shrubland changes (p-value < 0.01, r = 0.95). Notably, precipitation had negative effects (p-value < 0.05) on shrubland changes. This indicated that the most natural oasis changes in the Tuha Basin were highly sensitive to temperature and scarcely affected by precipitation, due to the extremely low precipitation and its downward trend, between 1990 and 2020, over the Tuha Basin. Generally, natural factors mainly affected natural oases, with an average coefficient of 0.92 (p-value < 0.01), and had insignificant effects on artificial oases.

All the human driving factors showed positive correlations with the oasis changes, with coefficients of 0.83, 0.94, 0.90, and 0.91 (p-value < 0.05), respectively. Cropland, shrubland, and built-up area changes were mainly controlled by human activities, and built-up areas had the most positive response to the population, total agricultural machinery power, raw coal production, and agricultural output, with coefficients of 0.93, 0.94, 0.96, and 0.98 (p-value < 0.01), respectively. Shrubland changes were influenced significantly by the total power of agricultural machinery, the raw coal production, and the agricultural output value, with coefficients of 0.87, 0.89, and 0.87, respectively (p-value < 0.05). With the same coefficient of 0.89, human activities had obvious driving forces on both natural oases and artificial oases. Human activities mainly affected shrubland in natural oases. The four human factors showed significant positive correlations with artificial oases, with coefficients of 0.86, 0.86, 0.85, and 0.89, respectively (p-value < 0.05). The spatial allocation of water resources improved with the development of agriculture and industry in the Tuha Basin. Moreover, policies supporting shrub planting and growth were encouraged to combat desertification and reinforce sand.

4. Discussion

4.1. Oasis Variation Impacted by Policies

The interactions among land types are complicated but traceable. Complex network theories provide us with a workable plan for a deeper understanding of the conversion relation among land types. In this study, a network made of nodes and edges was constructed. The land types were viewed as nodes, and the conversion relation among them was abstracted as directional edges.

According to the network, the area of the cropland expressed a slightly increased trend in all periods, apart from 1990–1995, when its area decreased from 2914.63 km² to 2565.92 km² with a ratio of more than 1 (Table 1). In this period, cropland mainly converted to built-up areas. After 1988, the ‘system of paid-for use for state-owned land’ policy was enforced, which meant that collective-owned land (cropland dominated) could possibly be converted to built-up areas [43,45]. However, after 2000, regulations for the protection of basic farmland in the Xinjiang Uygur Autonomous Region were published (https://www.xinjiang.gov.cn/, accessed on 5 December 2023), which regulated the cropland types that should be listed as basic protected cropland, and the proportion of it should not be less than 80% of the area of all cropland. This stimulated the cropland area at a stable level that then increased continually, and the ratio of cropland was less than 1 again after 1995. Meanwhile, the built-up areas still increase but are mainly converted from shrubland, grassland, and bare land. Finally, subjected to the policy constraints, the active reclamation was prohibited gradually [46]; cropland relatively seldom interacted with other land types and was sifted out of the principal components of the network.

Between 2006 and 2020, the policies paid more attention to ecological restoration and environmental protection problems. Many effective ecological reservation methods, such as returning cropland to forests, afforestation, and combating or controlling desertification, were carried out [47,48]. According to statistics from the Xinjiang Autonomous Region Forestry Bureau, as of the end of 2019, Xinjiang has completed a cumulative conversion of farmland to forests and grassland, with an area of about 150 million acres. Among them, the forest area is about 80 million acres, and the grassland area is about 70 million acres [48]. From 2005 to 2020, the interaction between cropland and forests appears observably weak; more cropland converted to forests in 2005–2010 and then remained in a dynamic stable condition (Figure 5), and the forests, from the lowest area of 71.98 km² in 1995, increased to reach the highest area of 87.61 km² in 2020.

The betweenness centrality was used to identify the key land type, which explained the possibility of the land type being the transitional type. Grassland was the key oasis type in the oasis transition process, and its importance decreased between 1990 and 2020. During that period, grassland was not the most important transitional land type due to the efforts of desertification control operations [49,50].

Planting shrubland vegetation is an effective general method of combating desertification [51,52] and is often coordinated with forests, cropland, and water bodies to shape a system to prevent desertification [53]. From 1990 to 2020, the area of shrubland showed an obvious increased tendency and expanded spatially toward bare land. In the transition network, the area of shrubland increased significantly. However, the interaction among shrubland, grassland, and cropland was not obvious, which means that the expanded part of shrubland was not utilized quickly and effectively. And, this also can explain the reason behind the betweenness centrality of shrubland remaining under a low level, even though its area increased significantly.

The oasis structural stability was quantified using the average path length in the oasis transition network, which was a crucial index for expressing the oasis sustainability in dryland. So far, few studies had investigated the structural stability of the entire oasis system in dryland. In this study, we applied the average path length to reveal the difficulty of the interactions among the oasis types. With the results of more unstable oasis structural stability, we concluded that there have been easier transitions among different land types in recent years than there were before. Therefore, we suggest that more attention should be paid to key land types (e.g., grassland) in dryland sustainable development.

4.2. Oasis Variation Impacted by Climate, Population and Economic Factors

Even though the correlations between oases and precipitation were insignificant, water resources constituted one of the most important restraint factors in dryland development [16]. The surface water in the Tuha Basin has obviously been shrinking in recent years, partly because of the increasing temperature and intense evaporation [54,55,56] in the Tuha Basin. Additionally, water resources have been over-exploited, and the underground water level have continuously dropped for agricultural irrigation and mining purposes [57,58,59]. A more adaptive strategy for regulating cropland areas and water resource allocation is needed. Artificial driving factors mainly impacted the artificial oases area [60]. Like many other studies, population was shown to be the main factor for oasis changes in dryland [8,61], and the reason can be attributed to an increased requirement for food as the population increased. As such, a large number of bare land or grassland was converted to cropland [8,61,62,63], but an area of cropland reserved later for high-efficiency agriculture production was also needed. Thus, the total power of agricultural machinery and the total output value of the plantation industry also increased and can be strongly related to the area of cropland increasing. Moreover, the development of urbanization and industrialization accelerated the demand for built-up [64,65]. Thus, the production of raw coal is also highly relevant to the artificial oases area.

4.3. Comparison with Other Studies

Traditional transition analysis method descriptions of land type processes mainly rely on the area and its changing rate in different years [66,67]; these methods highlight the variation patterns in a single land type, but they omit the interactions with other land types, and few consider the land-use-change system [15,68]. For example, shrubland, in this study, owing to the difficulty in quickly converting it to other land types, has little importance, although it represented a large area in 2015–2020. However, cropland has always maintained the second biggest importance, even though the cropland area was smaller than shrubland in 2020. Moreover, the built-up areas always have the highest increasing rate but hardly convert to other land types, so built-up areas do not have a controlling ability and are not counted as a key land type. Moreover, Zhang et al. [15] also compared the performance of betweenness centrality and the land-use dynamic degree and concluded that the complex network method was more scientific than traditional methods in identifying key land types.

For the oasis transition, we found that the cropland mainly converted from bare land and expressed as the transfer-in type, which means the cropland expanded spatially. Based on spatial heterogeneity, the cropland increasing rate slowed down after 2010. Cai et al. [21] also achieved the same conclusion, but compared to this study, we considered more interactions among of many land types rather than a single one.

Regarding the identification of key land type, the betweenness centrality was used, and we viewed grassland as a key land type. Xu et al. [20] also used the same way to recognize the key land type in Chengdu, which proved the successful appliance of complex theory in LUCC.

Regarding the structural stability of the oasis, we evaluated it by quantifying the interactions of the oasis transition and obtained a satisfactory result. Also, many other studies have rated the oasis stability through an ecological perspective [23,69,70] and analyzed the ecological carrying potential systematically, but the shortcomings of these studies can be summarized as follows: (1) they introduced subjective willing in classifying the indicator layer or set index attribute [71]; (2) when subjected to the data dimension, more attention was paid to the natural oasis; and (3) the partial data used in these study was hard to acquire, and studying time and costs were high.

4.4. Limitation in This Study

There are also some limitations to this study. The human driving force was considered to be the predominant reason behind the oasis transition; this can be further divided into policy factors, economic factors, and social factors. However, the impact of policy is difficult to quantify, and some relevant impacts of economic and social factors are hard to explore, owing to a lack of consecutive data. Furthermore, the oasis transition stability was evaluated in a spatial dimension, but for oasis land use, ecological stability is also valuable. Thus, more land properties, like the fraction of vegetation, soil production, etc., and the interactions among these should be considered in future works that further discuss suitable indicators.

5. Conclusions

To allow us to further understand the oasis spatial–temporal variations in the Tuha Basin from a systematic perspective and provide some references for future oasis development management, the complex network theory was introduced in this study, and related methods were input to reveal the oasis variation and transition processes, key land types in this process, and the oasis structural stability. The driving factors (human factors and natural factors) behind oasis variation and transitions were explored based on Pearson’s coefficient. The following conclusions can be drawn through this study:

(1)

Areas of natural and artificial oases all showed an increasing tendency; natural oases increased from 11,219.26 km² to 15,922.68 km², and artificial oases increased from 3127.17 km² to 4054.41 km². Natural and artificial oases were all considered as transfer-in types, and the increased parts were also all majorly converted from bare land.

(2)

Grassland was viewed as the key land type in the whole transition process due to the high out-degree values, in-degree values and betweenness centrality. This means that most land-type area change and indirect conversion can be related to grassland. Thus, grassland had the most vital status in the entire oasis system for controlling the efficiency of land-type transitions. Cropland also showed a higher betweenness centrality, and its area should be controlled at a dynamic balanced level.

(3)

From 1990 to 2020, the average path length of the oasis transition network was decreased, which means two arbitrary land types could be more easily converted directly, and the structural stability of the oasis transition was reduced. In this context, more attention should be paid to controlling the expansion scale of the oasis and regulating the transition coordination among shrubland, grassland, and forests.

(4)

Among the natural driving factors, the temperature correlated positively with natural oasis changes and had insignificant effects on artificial oases. Among the human driving factors, the population, the total power of agricultural machinery, the raw coal production, and the total agricultural output facilitated the expansion of shrubland and the artificial oasis. Built-up areas showed the most positive response to human driving factors.

This study clarified the oasis transition pattern and process and provided scientific references for future oasis studies and management.