Land resources are not only a natural resource on which human beings depend for their survival but also an important factor affecting changes in the ecological environment. Therefore, the rational use of land resources is an inevitable requirement for the sustainable development of cities (Haberl et al., 2004). From the "Western Development Strategy" to the "Belt and Road Initiative", rapid urban growth has placed immense strain on the surrounding cities, particularly in the arid areas of Northwest China (Li et al., 2012; Ahmad et al., 2018). At present, there is still a large area of land in the western of China that needs to be developed and utilized (Huang et al., 2020), and land allocation for high-quality urban construction is greatly concerning to managers and scholars (Yao et al., 2019). Agriculture is the main industry in Xinjiang Uygur Autonomous Region, China; however, in the process of urban development, there is a conflict between urban land and cultivated land, and irrational land use structure reduces cultivated land quality (Chen, 2007). In areas of cultivated land lacking irrigation from surface water sources, groundwater overexploitation for irrigation through driven wells leads to a decline in the groundwater table, accompanied by deterioration of the surrounding ecological environment (Huang et al., 2014). Simultaneously, population pressure and economic development demands have exacerbated environmental degradation through the conversion of land from high ecological value to construction land, resulting in urban problems such as soil erosion, overexploitation of water resources, and extreme weather conditions (Zhou et al., 2022), seriously constraining urban development. Future planning for the economic and environmental aspects of cities has a marked impact on urban land use patterns (Peng et al., 2020; Huo et al., 2022; Zeng et al., 2024). Moreover, multi-scenario simulations can predict and analyze future land use while evaluating ecological risks under varying scenarios, offering decision-makers recommendations for urban development planning that aims at reducing the occurrence of disasters (Wang et al., 2022b).

Simulation and prediction models for future land use are widely used across regions of all scales (Tong and Feng, 2020). The Conversion of Land Use and Its Effects to A Small Regional Extent (CLUE-S) is a widely used model for land use simulations (Mei et al., 2018). However, the CLUE-S model ignores possible spatial autocorrelation and self-organization in land use data (Wang et al., 2022a). The Cellular Automata (CA) is widely used for modeling predictions at specific spatial geographic unit scales, such as urban spatial growth, land use conversion, and natural disaster modeling (Li et al., 2017). The main advantages of CA as a dynamic model include: (1) simple implementation of calculation rules and fast calculation speeds; (2) scalability; and (3) combination with other models to improve accuracy. Over the last two decades, various CA-based models and platforms have been developed to simulate land use changes. Examples of these models include Slope, Land use, Excluded Areas, Urban Extent, Transport Networks, and Hillshade (SLEUTH), Dinamica Environment for Geoprocessing Objects (Dinamica EGO), Future Land Use Simulation (FLUS), Patch-generating Land Use Simulation (PLUS) and UrbanSim. The UrbanSim model focuses on simulating the impact of parameters on the spatial siting of subjects in an intra-city land (Patterson and Bierlaire, 2010). However, the UrbanSim model requires a large amount of base data; therefore, it is not applicable in areas where base information is incomplete. The SLEUTH model can predict various types of energy consumption in cities and is often used for large cities (Dadashpoor et al., 2019). The Dinamica EGO model is mainly used in research related to vegetation cover changes, such as deforestation (Stan and Sanchez-Azofeifa, 2019), revegetation of mining areas (Sonter et al., 2014), changes in vegetation cover in watersheds (Lima et al., 2014), and forest fires (Farfán Gutiérrez et al., 2020). The FLUS and PLUS models combine neural network algorithm and random forest algorithm, respectively, based on the CA model. The combination of these two models with satellite imagery data and Arc Geographic Information System (ArcGIS) is often used for land use change research in various scales (Li et al., 2024). The most obvious difference between the two models is that the FLUS model requires only one period of land use data to obtain the development probability of various land use types, whereas the PLUS model explores the mechanism of the drivers of land use change based on the land use dynamics of over two historical periods of urban development (Lin and Peng, 2022). It is noteworthy that the PLUS model has good flexibility as a stochastic model, which makes it possible to be widely applied in a variety of studies of land use change. However, the simulation process lacks specific criteria for selecting driving factors, which may result in logical inconsistencies even if the model's predictions meet accuracy standards. Therefore, this study used grey relational analysis to select the dataset with the highest correlation to the prediction year as the foundational data for PLUS model prediction. Furthermore, the model's algorithm is fixed for future trend changes, limiting its ability to promptly respond to certain policy plans and strategic goals, such as new urban area construction, ecological transition zone expansion, and new infrastructure development. Therefore, the PLUS model often needs to be combined with other algorithms or models for more accurate research. For example, assessing area habitat quality by incorporating with investment modeling (Li et al., 2024), studying the response relationship between runoff and land changes by incorporating with the Soil and Water Assessment Tool (SWAT) (Zhao et al., 2022), and evaluating regional ecological risks with Multi-objective Linear Programming (MOP) model (Guo et al., 2023).

Using the concept of sustainable development, the United States Environmental Protection Agency (USEPA) was the first to propose a framework for ecological risk assessment worldwide (Sorensen and Margolin, 1998). In the 1990s, research on ecological risk assessments began in China (Fu and Lu, 2006). Under the reform and opening-up policy, dominated by economic development, the rapid growth of Chinese economy has produced land use conflicts that cannot be ignored. Ecological risk assessments based on land use change have been applied in areas such as coastal, ecosystem fragility, and desert oasis areas (Ma et al., 2024). The land use ecological risk evaluation is mainly based on source-sink risks or landscape patterns. Compared with the traditional evaluation method based on source-sink risks, the method based on the evaluation of landscape patterns can not only directly reflect the ecological risk in the structure and composition of landscape patterns but also comprehensively evaluate the direct and cumulative effects of various landscape risks. However, evaluating ecological risks based solely on landscape patterns does not fully capture the impact of human activities on natural environment (Fan, 2022).

Through continuous development, ecological risk assessment has been integrated into landscape ecology theory (Turner, 2005). Risk sources and receptors have evolved from singular to multiple factors, including urbanization, human activities, meteorological variations, and land use changes, which are integral components of ecological risk assessments (Liu et al., 2021; Karimian et al., 2022). Evaluation indices have evolved from single indices to comprehensive multi-indices, and the evaluation index system has become richer (Landis, 2003). Recently, combining traditional models—such as the ecological footprint, hierarchical analysis, Pressure-State-Response (PSR) model, and fuzzy comprehensive evaluation method—with landscape pattern evaluation and then utilizing geostatistical analysis to establish a multi-indicator evaluation system has become a method of regional ecological risk evaluation that has received increasing attention (Larsen et al., 2016). Notably, in ecologically fragile and underdeveloped urban areas, the lack of socially productive resources and the difficulty in exploiting natural landscapes have led to a relative paucity of research on the coupling between land use changes and ecological risks. Fan et al. (2022) emphasized the increasing interaction between land use changes and ecological risks in Xiahuayuan District, Zhangjiakou City, Hebei Province, China. Liang et al. (2022) explored the transformation of land use and its relationship with ecological risks in the Three Gorges Reservoir area from the perspective of "production-living-ecological space". Deng et al. (2024) used the entropy weight method to analyze the impact of urban land use on the ecological environment and proposed measures to enhance ecological resilience in fragile areas. Huang et al. (2022) conducted an ecological security assessment and optimized the ecological landscape of Lhasa City, Xizang Autonomous Region, China by using the minimum cumulative resistance model, based on 2015 land use survey data and urban planning data. Ke et al. (2023) delineated priority areas for ecological restoration in Fuzhou City, Fujian Province, China by evaluating human disturbances and ecological security patterns based on land use survey data and social information. However, the relationship between land use changes and ecological risks from an engineering construction perspective still requires further research, especially in arid areas, where significant construction is necessary to overcome natural disadvantages. Therefore, studying the impact of engineering construction on the surrounding urban environment in the arid areas is crucial. Under this background, this study established a multi-indicator evaluation system based on PLUS model to conduct a comprehensive exploratory study on the changes in land use and high ecological risks of Turpan City by combining with information from local special water conservancy projects. Indicators were selected to evaluate the local ecological environment from four perspectives: urban expansion pressure, production pressure, landscape ecological risk, and ecological degradation pressure. Both urban expansion pressure and production pressure can reflect the pressure of human urban development on the environment (Chen et al., 2020). The other two aspects—landscape ecological risk and ecological degradation pressure—were used to quantify the ecological value of land use spatial patterns (Pan et al., 2021). In addition, considering that regional development is slow in the early stages (Tang et al., 2020), we took grid as unit to more effectively analyze how changes in land use within small areas affect the ecological environment at different scales.

Spatial autocorrelation has been widely employed to investigate the relationships between observed variables and location-based data (Liang et al., 2022). In addition to its utilization in resolving geographical issues, spatial autocorrelation analysis has progressively expanded into the fields of spatial econometrics (Getis, 2007), ecology (Valcu and Kempenaers, 2010), biology (Felizola Diniz-Filho and Bini, 2012), and others. The spatiotemporal characteristics of ecological risk index (ERI) in Turpan City were analyzed in this study using spatial autocorrelation. This study employed the grid method to analyze geographical features, human activity patterns, and urban construction characteristics at various spatial scales. Using the optimal grid scale method of Geodetector, the impact of hydraulic engineering on regional ecological environments at the most suitable spatial scale can be examined (Song et al., 2020). Additionally, this study incorporated the memory characteristics of Land Expansion Analysis Strategy (LEAS) module within the PLUS model (Xu et al., 2023). Additionally, this study incorporated the memory characteristics of LEAS module within the PLUS model. Turpan City, situated in Xinjiang Uygur Autonomous Region of Northwest China, primarily supports agriculture-based economy. The region suffers from high water consumption and severe groundwater over-exploitation, leading to considerable ecological degradation. As the city expands, these issues are further intensified by human activities, posing threats to overall ecological well-being. Aydingkol Lake, a critical ecological zone, faces threats from rising groundwater extraction levels. The expansion of groundwater over-exploitation zone poses serious threats to the lake and the broader ecosystem. The escalating ecological risks in these regions underscore the urgency of effective water resource management. This research, therefore, focuses analytically on the groundwater over-exploitation zones and the Aydingkol Lake (the goal region of Turpan City). Due to the very large area of unused land in Turpan City, it is difficult to clarify management approaches for ecological risks under different development plans. In other parts of the world, such as inland Australia (Allen et al., 2021), Arabian Peninsula (Alvarez et al., 2024), Northern Africa (Ndayiragije et al., 2022), and Congo Basin in Central Africa (Inogwabini, 2020), natural conditions in these areas are similar with Turpan City, and these areas face similar challenges in urban development. The objectives of this study include: (1) predicting the spatial distribution of land use and ecological risk of Turpan City in 2030 under different scenarios; and (2) analyzing the driving factors of land use change and ecological risk in Turpan City. The ecological research framework developed in this study offers a novel reference for sustainable development and ecological evaluation in arid areas.

To obtain accurate ecological risk predictions of Turpan City under the business as usual (BAU), rapid economic development (RED), ecological protection priority (EPP), and ecological-economic balance (EEB) scenarios in 2030, we undertaken the follow steps. Step 1: selecting the dataset with the highest correlation to the prediction year of 2030 as the foundational data for PLUS model prediction by grey relational analysis. Step 2: predicting land use under four scenarios in 2030 by MOP-PLUS model involving 14 driving factors (digital elevation model (DEM), slope, soil type, temperature, precipitation, population, gross domestic product rating (GDPR), distance from government (DFG), distance from primary road (DFPR), distance from secondary road (DFSR), distance from tertiary road (DFTR), distance from riverbed (DFRD), distance from reservoir (DFR), and distance from driven well (DFDW)). Step 3: constructing the ecological risk evaluation system that includes four dimensions of urban construction pressure, landscape ecological risk, production pressure, and ecological degradation pressure to calculate the ecological risk level (ERL) of each land use type and the ERI of optimal grid scale. Step 4: utilizing spatial autocorrelation to investigate the spatial distribution characteristics of ERI. And step 5: clarifying the explanatory power of the driving factors to ecological risk and the interaction between driving factors through Geodetector; the driving factors of ecological risk included slope, soil type, temperature, precipitation, population, GDPR, DFG, DFPR, DFSR, DFTR, DFRD, DFR, and DFDW.

From 2000 to 2020, Turpan City experienced significant land use changes, mainly reflected in the expansion of cultivated land and construction land, as well as a decrease in unused land and forest land, which contributed to a substantial increase in ecological risk (Li et al., 2022b). Between 2000 and 2010, the expansion of cultivated land and construction land was primarily concentrated around newly established driven wells and reservoirs, which undeniably supported agricultural and urban development (Lee et al., 2022). However, the conversion of unused land into construction land and cultivated land, coupled with the reduction in forest land, has weakened the region's ecological buffering capacity (Zhou et al., 2023). According to the results of land driving contributions, excessive groundwater extraction has led to water resource shortages, which has significantly increased the environmental pressure from unused land development (Orme et al., 2015). Between 2010 and 2020, although the intensity of land use changes decreased, construction land continued to expand rapidly. Despite increases in grassland, water body, and cultivated land, forest land significantly diminished, along with a slowdown in the development of unused land (Cheng et al., 2023; Lin and Wang, 2023).

As unused land development slowed and new driven wells and reservoirs were introduced, the ERL of unused land decreased, resulting in an overall decline in ERL. The driving factor analysis of ecological risk showed that between 2000 and 2010, the rise in GDPR and population not only facilitated the expansion of construction land but also heightened ecological risks due to intensified human activities (Gan et al., 2023; Zhu and Cai, 2023). From 2010 to 2020, the natural terrain factors (slope and soil type) became more influential on ecological risk. However, the slightly reduction in the ERL of unused land from 2010 to 2020 driven by DFDW and DFR suggested that identifying alternative water sources is a beneficial human intervention that can mitigate adverse effects of land use changes (Corwin et al., 2022). For sustainable development, the exploration of new water resources is recommended to support regional development and construction efforts (Choudhary et al., 2021).

Within this study's framework, the contributions of various driving factors to land use transitions in future scenarios aligned with the pattern from 2010 to 2020. This means that the positive effects of DFDW and DFR on land conversion from 2010 to 2020 will persist under future scenarios (Liu et al., 2024). However, under the RED scenario, extensive conversion of construction land led to a significant increase in ERL, indicating that the current strategy of adding new driven wells and reservoirs to find alternative water sources is insufficient to meet the sustainability demands of the RED scenario (Dhakal et al., 2021). In contrast, under the other three future scenarios in 2030, the ERL of each land use type declined compared with 2020, suggesting that current water resource planning can support sustainable development under the BAU, EEB, and EPP scenarios (Xie et al., 2021).

The ecological risk in Turpan City was closely linked to the conversion of unused land into construction land and cultivated land. A significant increase in cultivated land (52.30%) and construction land (75.02%) was observed between 2000 and 2010, and a notable increase in construction land (62.54%) was observed from 2010 to 2020. Observing the environmental changes from 2000 to 2020 in Turpan City, the ecological risk increased sharply from 2000 to 2010 due to the rapid rise in urban sprawl pressure (Ye et al., 2024). The main driving factors of ecological risk were GDPR, population, soil type, and precipitation. Under the RED scenario in 2030, the area at high ecological risk was larger than the other scenarios. In future ecological management of Turpan City, emphasis should be placed on the rational allocation of water resources and sustainable land use, particularly for groundwater over-exploitation zones and the Aydingkol Lake (Inostroza et al., 2013). Therefore, managing groundwater over-exploitation zone should focus on controlling the unregulated expansion of construction land, especially by optimizing spatial layouts to reduce the dependence of construction land on groundwater (Tao et al., 2016). Restricting development permits for new construction land and encouraging land use shifts towards low-water ecological restoration are recommended. In terms of water resource management, advanced water-saving irrigation techniques such as drip and micro-irrigation should be incorporated to decrease groundwater consumption in agriculture. Additionally, the exploration of alternative water sources, including rainwater harvesting and recycled water, should be encouraged to reduce groundwater dependency (John et al., 2019).

Under the EPP scenario, increased forest land and water bodies have markedly improved the overall ecological environment, offering valuable insights for managing the Aydingkol Lake. The presence of low-low ecological risk clusters around the Aydingkol Lake suggested a certain recovery capacity in its natural ecosystem (Yang et al., 2024). Therefore, vegetation restoration projects around the Aydingkol Lake, such as planting drought-tolerant species suited to arid climates, are recommended to improve the local microclimate through increased vegetation cover, while mitigating soil erosion and desertification (Gao et al., 2024). Gradual restoration of natural ecological functions in the Aydingkol Lake can potentially reduce ecological risk in surrounding areas without intensifying water resource demands. For areas adapting to the BAU, EEB, and EPP scenarios, guiding land use toward ecological protection and restoration based on current water resource planning can help continually lower the ecological risk (Liu et al., 2015). Furthermore, given the positive impact of DFDW and DFR on land transformation, establishing a long-term monitoring system based on these driving factors in the goal region of Turpan City is advisable (Guo et al., 2022). Regularly monitoring the changes of ecological risk index in areas near driven wells and reservoirs will enable the assessment of environmental impact of water resource facilities, allow for timely adjustments to water resource development plans to ensure alignment with sustainable development goals in construction and development.

Based on the MOP-PLUS model, this study presented a novel ecological risk management framework, which closely related ecological risks and land use changes, for urban development in ecologically fragile regions. As the case study, this paper analyzed the changes in land use and ecological risk in Turpan City from 2000 to 2020, and set up four scenarios (BAU, RED, EPP, and EEB) to simulate the relevant data in 2030. The results showed that from 2000 to 2020, the area of unused land dramatically decreased, while the areas of grassland, construction land, and cultivated land increased. Under the four predicted scenarios in 2030, the area of unused land, which is the largest land use type located mainly in the southeastern part of the city, was decreased under all four scenarios, and the area of construction land increased due to urban expansion. The main factors influencing the land use change in Turpan City were DEM and socio-economic factors such as GDPR and DFDW, and the environment deteriorated. Comparing to 2020, the ERL of all six land use types decreased under the BAU, EPP, and EEB scenarios in 2030, except for the RED scenario. The reason for the difference is that the RED scenario emphasizes economic benefit more than environment benefit. The goal region of Turpan City, where lots of residents live, was mainly at I-V levels of ecological risk. The dominant driving factors of ecological risk were GDPR, soil type, population, temperature, and DFRD. The interactions between these driving factor pairs amplified their influence on ecological risk. Under the present planning and management of water resources, it was not sufficient to support the sustainable development under the RED scenario. Although there was some improvement in the ecological environment under the BAU and EPP scenarios, it was still not sufficient to support the needs of sustainable development. After comprehensive comparison, this study recommended that the EEB scenario was the most suitable development scenario in Turpan City.

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

This study was financed by the Third Comprehensive Scientific Survey Project of Xinjiang (2021xjkk1003), the Youth Innovation and Cultivation Talent Project of Shihezi University (CXFZ202201, CXPY202201), the Annual Youth Doctoral Program of Xinjiang Uygur Autonomous Region 'Tianchi Elite' Introduction Plan (CZ002302, CZ002305), and the High Level Talent Research Launch Project of Shihezi University (RCZK202316, RCZK202321).

Conceptualization: LI Haocheng; Methodology: LI Haocheng; Writing - original draft preparation: LI Haocheng; Writing - review and editing: LI Haocheng, LI Junfeng, FARID Muhammad Arsalan; Software: Li Haocheng, CAO Zhiheng, MA Chengxiao, FENG Xueting; Validation: LI Junfeng, WANG Wenhuai, QU Wenying; Supervision: LI Junfeng; Project administration: LI Junfeng; Funding acquisition: LI Junfeng; Data curation: WANG Wenhuai, QU Wenying, CAO Zhiheng, MA Chengxiao, FENG Xueting; Visualization: CAO Zhiheng, MA Chengxiao, FENG Xueting; Investigation: CAO Zhiheng, MA Chengxiao, FENG Xueting. All authors approved the manuscript.