1. Introduction
Cities are the most direct carriers of land-use/-cover change (LUCC) [
1], and urbanization is an inevitable result of global economic development—cities are regarded as the future environment for all human beings [
2]. Rapid urbanization is bound to produce urban ecological problems such as ecosystem decline, biodiversity reduction, and urban heat island effect, which directly or indirectly affect the quality of urban habitat and sustainable urban development [
3]. The emergence of these problems is closely related to land-use type conversion and landscape pattern changes in the region [
4]. In this process of rapid urbanization, the contradiction between the expansion of urban land and limited land resources—between rapid urbanization and the ecological environment—is gradually emerging, especially in developing countries. Therefore, analyzing the urban expansion and landscape pattern changes and driving forces can help to both solve these urban problems and provide a scientific basis for the optimization and upgrade of urban land-use structures.
Regarding the analysis of urban landscape pattern changes, scholars have conducted a large number of studies globally. Research on landscape patterns began in the 1950s, focusing on the characteristics and driving mechanisms of spatial and temporal changes in landscape patterns [
5] and the prediction of landscape patterns [
6]. With the rapid development of remote sensing and geo-information systems, the combination of the two methods has become an effective research method for analyzing changes in landscape patterns, mainly including dynamic monitoring and landscape index analysis—such explorations have changed in nature from qualitative research to quantitative research [
7]. Domestic research on landscape patterns started relatively late, but the results are fruitful. The relevant literature mainly focuses on the following aspects: The analysis and simulation research of landscape pattern evolution, using the CA–Markov model and the FLUS model, has been performed to simulate future land-use conditions, allowing the prediction of landscape patterns [
8,
9]. Secondly, researchers have combined landscape patterns and ecological risk models and have built an ecological network with the help of the least-resistance model to allow the optimization of landscape pattern formulation and ecological risk assessments in a given region [
10]. Thirdly, researchers have analyzed the driving factors of landscape patterns using geographic detectors, grey correlation analyses, and other methods to gain insight into the driving factors that affect the evolution of the landscape pattern in the region [
11,
12]; however, due to the different geographical locations of the study areas, there are huge differences in the driver databases chosen in these studies [
13]. Based on the above research results at home and abroad, we can see that current research mainly focuses on regional units, such as watersheds, forests, wetlands, and cities; meanwhile, fewer analyses have been performed on the landscape pattern evolution characteristics and driving forces of urban agglomerations in arid zones. Inspired by the above related studies, we consider the landscape pattern evolution characteristics of oasis urban agglomerations in arid zones: What are their driving forces? What are the links between landscape pattern evolution and land-use intensity in terms of spatial divergence patterns? Therefore, we selected the urban agglomeration on the northern slope of the Tianshan Mountains, in the arid region of Northwestern Xinjiang, China, as our research object; here, we analyzed the evolution characteristics and driving forces of its landscape pattern.
In recent years, human activities have gradually increased in the degree of disturbance for the ecological environment and the intensity of land use; thereby, the regional landscape pattern has changed, and the stable development of the ecological environment has been affected [
14]. As China’s economic and social development center gradually shifts to the west, small- and medium-sized cities have gradually grown into the mainstay of promoting national economic growth and urbanization [
15]. These small- and medium- sized cities have gradually developed into urban agglomerations. However, due to the constraints of geographical location, development scale, transportation conditions, resources and environment, and economic development level, the urbanization process of small- and medium- sized urban agglomerations may show different characteristics from those of large urban agglomerations [
16]. Recent studies have mainly focused on large and mature urban agglomerations, such as the Yangtze River Delta urban agglomeration [
17], the Beijing-Tianjin-Hebei urban agglomeration [
18], and the Pearl River Delta urban agglomeration [
19]; in these studies, insufficient attention has been paid to the rapidly urbanizing small- and medium-sized urban agglomerations. The Tianshan North Slope City agglomeration, located in the northern Tianshan Mountains in Xinjiang, Northwest China, is the most economically developed region in the northern Tianshan Mountains and is one of the most important nodal city clusters for the development of Western China. In the critical period of comprehensive urbanization in Xinjiang, the study on the landscape pattern and land-use intensity of urban agglomerations on the northern slope of the Tianshan Mountains not only enhances the mechanisms of the expansion pattern and the mechanisms of small- and medium-sized urban agglomerations, but also has guiding significance for the territorial spatial planning of small- and medium-sized urban agglomerations that are gradually developing in similar regions. Therefore, we took the urban agglomeration on the northern slope of Tianshan Mountains as the study area; we used multi-period land-use data and applied landscape pattern analysis, spatial autocorrelation analysis, and geographic probe methods to study landscape morphology and pattern changes within the urban agglomeration; finally, we quantitatively explored multiple factors in the dynamic changes of landscape fragmentation and their driving factors.
4. Results
4.1. Temporal and Spatial Changes in Land Use Types
Land use in UANSTM produced significant changes in time and space over the period 1995–2018. In general, the area of cropland and urban land continues to increase; the area of water, forest, and grassland decreases year by year; the area of unused land does not change much. From the viewpoint of counties and cities, urban land in all regions showed an increasing trend, among which the urban land area in Urumqi grew the fastest, with an average annual growth of 24.08 km2. The cropland area showed a decreasing trend only in Urumqi and Shihezi, while the remaining nine regions showed an increasing trend. Forest, grassland, and water areas all show decreases, but the trend is gradually leveling off.
4.2. Analysis on the Dynamic Degree of Land Use
As can be seen from
Figure 2 the area of urban land in UANSTM increased significantly from 1995 to 2018, followed by cropland, with increases of 23.55% and 13.75%, respectively—these are the two most dramatic types of land-use changes in the UANSTM. Additionally, the areas of forest, grassland, and water showed a decreasing state year by year. From each time period, it can be seen that the area of urban land and cropland grew the fastest from 2005 to 2010, with an average annual growth rate of 9.17% and 6.26%, respectively—these are much higher than in other time periods. There was a slow increase in water area from 2010 to 2018 (
Figure 3).
For the cities and counties in the UANSTM, the urban area of each region showed an increasing trend in all time periods, with the average annual growth rate of urban land in Urumqi peaking at 26.28% between 2000 and 2005. However, the change in cropland varies among cities and counties. Kuitun City, Wusu City, Manas County, Hutubi County, and Fukang City all showed a continuous growth trend in cropland area in all time periods, with Kuitun City having the highest average annual growth rate of 26.69% from 2005 to 2010 (
Figure 4).
4.3. Land-Use Transfer Matrix
As we can see in
Figure 5, the 11 regions in the UANSTM experienced more frequent land-use conversions over the 23-year period, and the continuous urbanization process led to a deepening fragmentation in the study area, affecting and changing the surrounding natural landscape. During the study period, the four land types of grassland, cropland, unused land, and urban land shifted more significantly, while forest and water areas shifted less significantly. Although the mutual transfer of water and forest areas with other land types was less obvious, there was still a small transfer of grassland to water area and forest land.
Table 5 allows us to see in more detail that there was a large transfer out as well as a large transfer in of land use within the UANSTM during this 23-year period. Among them, among the areas of urban land and cropland with the most significant changes, grassland and water areas were the main input types, with 5884.9 km
2 and 1700.4 km
2 transferred, respectively; cropland, with grassland as the main input type, showed an input area of 5988.4 km
2. Since the UANSTM regions are located on top of the same arid zone oasis, there is a similarity in their land-use type transfer; the urban land in these 11 regions is dominated by cropland and grassland transfer, and cropland is also dominated by grassland input.
4.4. Landscape Pattern Analysis
The landscape pattern index of the landscape level can clearly reflect the landscape characteristics of the whole region. The landscape characteristics of urban land use changed significantly with the change of land use in UANSTM during 1995–2018. As shown in
Figure 6, the PD, ED, and LPI of urban land all increased to varying degrees from 1995 to 2018, with the LPI showing the largest increase. This group of trend changes shows that the townscape of UANSTM tends to be fragmented, and the increase in LPI indicates that the inner city is gradually clustering, but because the urbanization of other counties and cities is still in the development stage, the growth of the number of urban landscape patches is accompanied by the increase in patch density, which shows that the number of patches increases with a greater landscape density, and the fragmentation of the landscape increases, indicating that the inner city landscape is becoming more complex. It can be seen that, over the past 23 years, the urban lands of the UANSTM have shown a trend of increasing landscape fragmentation, landscape heterogeneity, and complex patch shapes.
4.5. Degree of Comprehensive Land Use
The degree of comprehensive land use in the UANSTM increased during the study period, with significant differences in different directions and regions (
Figure 7). During these 23 years, the spatial distribution of the degree of land use comprehensively spread to the north, northeast, northwest, and southeast, respectively. The expansion was mainly to the north, but intensive growth also occurred within the cities.
The spatial and temporal distribution of the expansion of comprehensive land use in UANSTM is different; the key expansion areas are concentrated in Urumqi, Karamay, and Changji, and the urban land in these three cities shows a large-scale expansion trend. This type of land in Shihezi and Fukang cities is also in a state of rapid expansion, which is driven by economic development. Among them, Urumqi, Changji, and Shihezi cities have urban land as the dominant expansion type, while other areas are in the middle stage of urbanization development. Therefore, the distribution of areas with a high value of comprehensive land use is more scattered, and they show more reorganization stages of urban land use.
4.5.1. Spatial Clustering of the Degree of Comprehensive Land Use
The UANSTM‘s degree of comprehensive land use, assessed using the global Moran’s I, was calculated by the ArcGIS software. The results show that Moran’s I was positive for the period 1995–2018, with 0.304 and 0.188, respectively (
Figure 8), indicating that there is a high positive spatial correlation and significant spatial aggregation of the comprehensive land-use degree in the study area. The univariate local spatial autocorrelation LISA aggregation plot of land-use extent over the entire study period was also drawn based on the z-value test results (
Figure 9). Five types of spatial association are suggested as follows: high-high clustering type, high-low outlier type, low-high outlier type, low-low clustering type, and not significant. The analysis of the regional spatial agglomeration characteristics concluded that, in 1995, the high-high clusters were mainly distributed in the central part of the study area in a band-like manner, and the high-high clusters in 2018 showed a piecewise distribution with an increase in the distribution range compared with 1995. The low-low clusters during these 23 years were mainly distributed in the north and south sides of the study area, mainly because these two parts were mainly composed of bare land with low vegetation cover and poor ecological quality, which led to a decrease in land use.
4.5.2. Analysis of Driving Factors based on Geographic Detector
From the results of the factor detection (
Table 6), it can be seen that the explanatory power of different factors on the spatial and temporal distribution of the degree of comprehensive land use from 1995 to 2018 is, in descending order, as follows: NDVI > river density > temperature > sunshine hours > DEM > precipitation > slope > population > GDP. It can be seen that the most dominant factor affecting the spatial and temporal distribution of the degree of comprehensive land use in UANSTM is NDVI, indicating that the distribution of NDVI and river density becomes the biggest threat source of land-use/land-cover changes in the UANSTM. This also confirms that the UANSTM, as a typical oasis city agglomeration in an arid zone, can only rely on the oasis for development—urban development has certain limitations.
From the results of the interaction detection (
Figure 10), it can be seen that—taking the results of 2018 as an example (the results of the remaining years have the same mode of action)—the interactions of all the factors mutually reinforce each other, indicating that there is a clear correlation and coordination among them. Among the factors, river density and NDVI together provided a strong explanation. This indicates that the influence of human activities on oasis production is lower than that of other factors, and the contribution of natural factors is significantly higher than that of social and economic factors.
6. Conclusions
In summary, this study takes the UANSTM as the study area, combines land-use data, and calculates the landscape pattern index to explore the land-use change pattern and spatial characteristics of the landscape pattern of the UANSTM. The results indicate a gradual increase in intra-urban landscape heterogeneity over the study period. Urban land and cropland were the most significant land-use types that changed. This shows that urbanization is accelerating in this region, while safeguarding the amount of cropland to meet the needs of population growth. This result is feasible and reasonable. We calculated the land-use intensity of the UANSTM over a 23-year period and analyzed the spatial correlation of the index, finding a high positive spatial correlation and significant spatial aggregation in the degree of land use. In the analysis of the driving forces, it was found that UANSTM, as an oasis city cluster in the arid zone, can only rely on the oasis for its development, and that urban development is restrictive. Therefore, ecological protection should be used as the basis for future development to maintain the ecological security of the oasis and build a firm ecological security barrier for the construction of the urban cluster on the northern slopes of the Tianshan Mountains.