The impact of gradient expansion of urban–rural construction land on landscape fragmentation in typical mountain cities, China

ABSTRACT

Hillside urbanization currently provides the most economical solution for urban expansion. However, limited attention was paid to the characteristics, processes, and patterns of gradient expansion of construction land (GEC) in mountainous regions and its impact on landscape fragmentation. This study suggests an approach to identify GEC in mountainous regions and quantify its spatial impact on the fragmentation of cultivated land and ecological land. The results are as follows: (1) GEC can be categorized as ‘unidirectional’, ‘bidirectional’, or ‘multidirectional’ expansion, influenced by the proportion of topography in different directions of the cities. The magnitude of expansion is correlated with the city's economic level. (2) GEC leads to an increasing trend of landscape fragmentation, with nearly a three-fold increase in arable land fragmentation and a two-fold increase in ecological land fragmentation within the focus areas over 20 years. (3) The impact of GEC on the fragmentation of cultivated and ecological land diminishes with distance, primarily influencing within 3 km. The results of this study provide a methodological reference for identifying key areas and characteristics of GEC, and serve as a theoretical basis for reducing landscape ecological risks in the process of mountainous urban development and promoting sustainable development.

KEYWORDS:

• Urban sprawl
• ecological land
• environmental impact
• land use
• urbanization
• SDGs

1. Introduction

Mountainous areas account for 22% of the global land surface and are home to over 900 million people. The sustainable development of mountain resources is an important part of global sustainable development (FAO 2015; United Nations 2016). China has the widest distribution of mountains, with mountainous areas accounting for about two-thirds of the total area. The number of mountain towns accounts for about half of the towns, and mountain resources have great development potential. However, the complex terrain conditions and unique geographical environment have exacerbated regional development imbalances, and the overall economic development of mountain towns has long lagged behind the national average level (Deng, Cheng, and Wen 2008). With the advancement of China's industrialization and urbanization process, the urbanization process in mountainous areas is gradually accelerating. However, in this process, the topography and landforms as basic conditions have always guided and constrained the expansion of cities. Faced with the increasingly intensifying ‘human-land’ contradiction, many cities have carried out a series of transformation activities on mountain environments, such as ‘cutting mountains and reclaiming land’ and ‘flattening mountains and leveling peaks’ (Wu et al. 2019; Wang, Yan, and Su 2021; Pu et al. 2023). With the development of the economy and the expansion of the population, mountain cities are constantly expanding to the surrounding hilly slopes, which is visually manifested in the expansion of construction land to higher terrain gradients. However, scholars have different names and definitions for this phenomenon. Some refer to it as the ‘development of low hills and gentle slopes’, which refers to the process of urban development in hilly zones with elevations below 300 m and slopes between 6° and 25° (Peng et al. 2015). Others metaphorically refer to it as ‘urban slope-climbing’, which refers to the process of urban construction through engineering transformation in relatively small-scale hilly areas with a topographic relief of less than 200 m and an average slope of less than 25°, predominantly consisting of unused land (Zhou et al. 2021b). By summarizing these definitions, we believe that this phenomenon is essentially the gradient expansion of construction land (GEC) in mountainous urban areas, which is the process of gradually encroaching on mountain areas with a topographic relief of more than 200 m or slopes of more than 6° or the evident increase in slope within the built-up area.

The changes in landscape patterns resulting from urban expansion are a significant driver of alterations to the urban and surrounding ecological environment (Das et al. 2021; Wei et al. 2022). Effectively quantifying and analyzing the impact of landscape patterns during the GEC in mountainous urban areas is crucial for comprehending and enhancing the mountainous construction and development process. Fragmentation is a pivotal aspect of natural landscape evolution on the Earth's surface, exerting a substantial influence on biodiversity and ecosystem functioning. Consequently, it has become a research focus within disciplines such as geography and landscape ecology (Das and Basu 2020; De Montis et al. 2017). Landscape fragmentation refers to the process in which originally continuous and complete patches become complex and discrete due to factors such as human activities and natural disasters (Zou, Wang, and Bai 2022). Presently, research on landscape fragmentation incorporates three primary components: (1) the causes and mechanisms of landscape fragmentation. Researchers frequently employ landscape indices, such as patch density (PD), landscape division index (DIVISION), splitting index (SPLIT), and landscape percentage (PLAND), to evaluate the natural fragmentation processes in wetlands, forests, grasslands, cultivated lands, water bodies, and green spaces, as well as the impacts of human activities on landscape fragmentation (Kubacka et al. 2022; Zhou et al. 2024; Chettry 2023). (2) The influence of landscape fragmentation on biodiversity and ecosystem functioning. The ‘island effect’ resulting from landscape fragmentation leads to the dispersal and isolation of biological populations, thereby restricting species migration and gene flow, with severe negative ramifications for biodiversity and ecosystem functioning (Ghosh and Das 2020; Gibson et al. 2011). (3) Ecological restoration and recovery strategies for addressing landscape fragmentation. At present, researchers commonly employ various scenarios to simulate and forecast future land conditions, allowing for a comparison of the impact of different landscape patterns on the ecological environment. This comparison facilitates the development of sustainable urban development strategies and ecological restoration plans (Basu et al. 2021; Ahmadi Nadoushan 2022).

Currently, most research on the impact of urban expansion on landscape fragmentation primarily focuses on the horizontal expansion of cities in plain areas and the resulting ecological risks for natural landscapes. However, the GEC in mountainous urban areas occurs simultaneously in both vertical and horizontal directions, leading to significant disturbances to mountain ecosystems and landscape patterns. Only a limited number of studies have addressed the ‘third dimension’ of topographic gradients. These studies mainly employ elevation, slope, and terrain position index to delineate the gradients and investigate the effects of urban expansion on land use patterns and ecological environments (Liang et al. 2023; Xue et al. 2019). These topographic factors vary in the form of continuous gradients rather than discrete structures, which limit observations to characteristics and impacts of urban expansion within a specific gradient range, without considering individual objects. Therefore, when exploring the GEC on landscape fragmentation, it is crucial to first define the spatial extent of the gradient expansion. Although several studies have proposed spatial delimitation methods for the gradient expansion of construction land (Yang et al. 2021; Shi, Wu, and Liu 2022; Shi et al. 2023), these methods currently have issues with poor adaptability and ignore the neighborhood characteristics of gradient expansion. Furthermore, the existing research on the gradient expansion of construction land predominantly focuses on the spatial distribution and impacts of land encroachment during a single period, thereby lacking a comprehensive analysis of multi-period spatiotemporal evolution (Yang et al. 2022).

Above all, the following problems still need to be explored to understand the impact of GEC on landscape fragmentation: (1) What is the definitive spatial scope and influence scope of GEC? (2) What characterizes the fragmentation of regional landscapes? (3) What quantifies the impact of landscape and topographic features of GEC on the spatial heterogeneity of landscape fragmentation? Therefore, we proposed a method to identify the GEC, and selected ten factors: elevation, slope, topographic relief, PLAND, largest patch index (LPI), PD, mean patch area (AREA_MN), edge density (ED), mean Euclidean nearest neighbor distance (ENN_MN), and landscape shape index (LSI). With the statistical methods and multi-buffer analysis, we analyzed the effects of GEC of nine typical mountain cities in China on the fragmentation of ecological land and cultivated land from 2000 to 2020. The study provides a basis for protecting the surrounding environment of mountainous cities and standardizing urban construction, and it helps promote the green and sustainable development of mountainous cities.

2. Data and methods

2.1. Study area

China is a mountainous country, with a mountainous area of about 622.39 × 104 km², accounting for about two-thirds of the country’s total area. Against the background of the vast mountainous environment, cities carry out a series of projects to open up space for development, and thus Chinese cities show unique distribution patterns and evolution patterns in the three-dimensional space of the terrain. China’s elevation is high in the west and low in the east, and the terrain gradually declines from west to east, so the terrain can be divided into the first, second and third gradient terrains. The first and second gradient terrains and the junction of them have large topographic relief, and they are the areas where the plateau of China is mainly situated, with a large number of mountainous cities distributed. In order to explore the spatial and temporal patterns of the GEC and its related impacts on the landscape pattern, we selected mountainous cities located in the first and second gradient, and the boundaries of these gradients where obvious gradient expansion occurs as the main research objects (Zhou et al. 2021a). It includes the main types of mountainous cities in China (Huang 2021), namely hilly mountainous cities such as Chongqing, Panzhihua, Chengdu, Shiyan, and Guiyang (basin and hilly landforms), and Dali (plateau and hilly landforms); valley mountainous cities such as Lanzhou and Xining; and ravine mountainous city like Yan'an. The spatial location and topographic features of the nine cities in China are shown in Figure 1.

Figure 1. Study areas.

2.2. Data source and preprocessing

The land use data selected for the study originated from the Resource and Environment Science and Data Centre with a spatial resolution of 30 m. Five periods of data, 2000, 2005, 2010, 2015 and 2020, were used for the experiment. The study reclassified this secondary classification of land use into a primary classification through ArcGIS software, including six types: cultivated land, forest, grass, water, construction land and unused land, of which forest and grass are the main types of ecological land used in this study. The digital elevation model is derived from the NASA Shuttle Radar Topography Mission (SRTM) dataset, collected in 2000, and has a spatial resolution of 30 meters. The study calculated the slope and relief amplitude based on DEM data as input parameters for the identification method of the GEC. In addition, we also selected the spatial distribution data of Chinese landform types with a resolution of 1 km from the Resource and Environment Science and Data Centre to analyze the geomorphic environment of each mountainous city (Table 1).

Table 1. Research data.

Data	Spatial resolution	Usage
DEM	30 m	Used to calculate terrain factors such as slope and topographic relief
Multi-period Land Use Remote Sensing Monitoring Dataset in China (CNLUCC)	30 m	Monitoring the dynamic process of urban expansion
Spatial distribution data of landform types in China	1 km	Calculate the percentage of different landforms in each direction of the city

2.3. Methods

2.3.1. Identification method for GEC in mountain cities

The current method for identifying the GEC primarily involves setting thresholds based on topographic data. Any new construction land falling within these thresholds is designated as part of the GEC These methods are suitable for quickly detecting large-scale areas with less accurate data, their results may lack the necessary precision to portray the GEC pattern effectively. Moreover, the threshold method lacks adaptivity. Setting the threshold too high may overlook urban expansion in low-slope hilly regions, while setting it too low may incorporate non-gradient expansion areas. To address these limitations, we propose a GEC identification method that combines the traditional threshold approach with the concept of GEC, which the emphasizes noticeable growth of urban construction land on the topographic gradient compared to the previous period (Yang et al. 2022), as shown in Figure 2:

Figure 2. Methods for GEC identification.

Based on Figure 2, we divided the GEC into three main categories: (1) New construction land is located in mountainous areas with topographic relief more than 200 m. (2) The new construction is located in a sloping area with topographic relief less than 200 m and slope more than 6°. (3) When the new construction land in flat and gently sloping areas with topographic relief less than 200 m and slope less than 6°, it is a GEC area if the average slope of the new patch is larger than the average slope of the nearest patch in the previous period. Then, we clustered the patches by spatial distance, used the minimum bounding rectangle for spatial division, and removed the areas less than 5 km² to get the key areas of GEC as the spatial unit for our study.

2.3.2. Measures of landscape fragmentation

Landscape fragmentation refers to the process of transforming a landscape from a simple and homogeneous state into a complex and heterogeneous one, caused by natural or human influences. This transformation involves a shift from individual, continuous patches to discontinuous and diverse patches. To assess the impact of GEC on the fragmentation of both cultivated land and ecological land, we utilized Fragstats to calculate three metrics at the class scale: Patch Density (PD), Landscape Shape Index (LSI), and Landscape Division (DIVISION) (Fan and Myint 2014). These three metrics can reflect the landscape characteristics and ecological information from the perspective of fragmentation, complexity and connectivity. To further investigate the influence of GEC on landscape fragmentation, we took into account the varying scales of the significant GEC areas. We evaluated the rate of change of each indicator within a given period, including the rate of change in Patch Density (C_PD), Landscape Shape Index (C_LSI), and Landscape Division (C_DIVISION), to gauge changes in landscape fragmentation. Meanwhile, in order to portray the spatial and morphological characteristics of the GEC, we also selected some landscape metrics, including PLAND, LPI, PD, AREA_MN, ED, ENN_MN, and LSI.

2.3.3. Analysis of the drivers of landscape fragmentation

Redundancy analysis (RDA) is a technique used to examine asymmetric or directional correlations between multiple sets of variables. Its purpose is to identify the influential variables among a set of known variables that have the greatest impact on a target variable (Wang and Cheng 2023). It concentrates the influence of the influential variables on the target variable on several sorting axes, also known as canonical axes, which are linear combinations of the influential variables. The RDA is based on a certain gradient model of the influence variables to express how the target variable responds to continuously changing influences, reflecting the relationship between the influences and the target variable under certain environmental conditions. In our study, we selected key areas of GEC as the sampling sites and extracted sample attributes to establish a database. C_PD, C_LSI, and C_DIVISION of cultivated land and ecological land were used to form the target variable data matrix, and average slope, average elevation, average topographic relief, PLAND, LPI, PD, AREA_MN, ED, ENN_MN and LSI of the GEC were used to form the influencing factor data matrix. Finally, the redundancy analysis of the target and influence variable matrices and the calculation of Spearman’s correlation coefficients were carried out using the R language ‘vegan’ package.

3. Results

3.1. Spatiotemporal evolution patterns of gradient expansion of construction land

We identified areas of gradient expansion in nine typical mountain cities based on the identification method of GEC (Figure 3). The analysis revealed significant spatial heterogeneity in the patterns of GEC across each region. In general, the GEC can be categorized into two main types: edge-type expansion and enclave-type expansion. The dominant type is edge-type expansion, which gradually extends from the main urban areas to the surrounding slopes, observed in all the cities under study. Enclave-type expansion primarily involves the establishment of new cities. In situations where the terrain around a city is unsuitable for development or when there is limited available land, the government designates new areas in the mountains for the purpose of ‘bulldozing mountains to build new cities’, such as Lanzhou New Area, Yan'an New Area, and Liangjiang New Area. Then, we clustered patches of GEC based on distance, then used a minimum bounding rectangle to divide the spatial extent of each group and excluded the area smaller than 5 km² to identify key areas for GEC. Figure 4(b) presents the results of the identification of key gradient expansion areas in Lanzhou from 2015 to 2020, which will be used as the unit of analysis for our study.

Figure 3. The identification results of GEC.

Figure 4. Results of identifying key areas of GEC (taking Lanzhou as an example): (a) The scenario where areas less than 5 km² are not excluded, (b) The scenario where areas less than 5 km² are omitted.

To quantify the overall scale of GEC and the evolution law of topographic features, we counted the area and average slope of GEC in the four periods in nine cities (Figure 5), respectively, and found that there were significant differences among them. In terms of time trends, the scale of GEC showed a fluctuating upward trend, and the overall development intensity was positively related to economic levels. Chongqing had the largest total area of GEC during 2000-2020, totaling 1,059.70 km², while Xining had the smallest, with only 51.33 km², which was a difference of nearly 20 times. The area of GEC in Chengdu reached a maximum of 266.11 km² in 2015–2020. The area of GEC in Lanzhou shows a significant growth trend, with the area of only 10.39 km² from 2000 to 2005, and as high as 289.29 km² in 2015-2020, which expanded nearly 29 times. In addition, the average slope of GEC in the Panzhihua was the highest of all nine cities, and the overall slope was maintained at about 17°. The GEC slopes in Shiyan, Chongqing, and Xining were stable, at about 15°, 13°, and 6°, respectively.

Figure 5. The scale and slope characteristics of the GEC.

The analysis of topographic area proportions and gradient expansion directions in mountainous cities (Figures 6 and 7) revealed that the majority of gradient expansion occurs in areas with high proportions of platforms, hills, and low undulating mountain terrains, while avoiding regions with high undulation. Yan'an and Lanzhou exhibit similar expansion patterns, both characterized as ‘unidirectional’. These cities primarily expand towards the northern regions, with platforms dominating the topography in northern Lanzhou and hills in Yan'an. Similarly, cities such as Chongqing, Dali, Panzhihua, and Shiyan demonstrated similarities in expansion directions, following a primarily ‘bidirectional’ growth pattern over the years. On the other hand, the main expansion directions of Chengdu, Guiyang, and Xining vary over time, indicating a lack of fixed directionality. This variation can be classified as ‘multidirectional’. Additionally, in the ‘multidirectional’ gradient expansion cities, the distribution of platforms and hills in different directions was relatively uniform, providing cities with more flexibility in selecting the direction of gradient expansion.

Figure 6. Gradient expansion direction change in construction land.

Figure 7. Proportion of topographic area in different directions in mountainous cities.

3.2. Landscape fragmentation characteristics of key regions of GEC

Correlation heat maps (Figure 8) were generated for each city to analyze the relationship between landscape features and topographic characteristics in key areas of GEC. Among the nine cities, a consistent pattern emerged, showing a strong positive correlation (correlation coefficients > 0.9) between the slope and topographic relief of the GEC. Additionally, LPI and PLAND demonstrated a positive correlation (correlation coefficients > 0.7). Conversely, AREA_MN, ENN_MN, and PD displayed a negative correlation, with AREA_MN and PD in Yan'an, Xining, and Panzhihua exhibiting correlation coefficients below −0.75. In the cities of Chengdu, Shiyan, and Yan'an, AREA_MN exhibited a positive correlation with slope, topographic relief, and elevation. Notably, the correlation coefficient between AREA_MN and elevation in Chengdu was remarkably high at 0.99, indicating a significant increase in the average GEC patch area with rising elevation. In Xining, ENN_MN demonstrated a strong negative correlation with slope and topographic relief, suggesting that higher slopes and relief contributed to enhanced connectivity, reduced fragmentation, and increased aggregation. Moreover, PLAND and LPI showed a negative correlation with a average slope in Chongqing, Lanzhou, Panzhihua, Shiyan, and Xining, but a positive correlation in Yan'an.

Figure 8. Correlation between the features of GEC.

By comparing the fragmentation of cultivated land and ecological land in the key areas of GEC in different years, a clear temporal disparity in landscape fragmentation resulting from GEC was identified (Figure 9). Analysis of the cultivated land fragmentation metrics showed that the average LSI increased from 4.19 in 2000 to 6.04 in 2020, indicating an increasing complexity in the shape of cultivated land. The average PD was 0.38, and the average DIVISION was 0.54 in 2020, and increased to 0.94 and 0.96 respectively in 2020, indicating that the GEC had become more and more drastic in recent years, and fragmentation of cultivated land had significantly increased. Regarding changes in the fragmentation of ecological land, the average PD increased from 0.35 in 2000 to 0.69 in 2020; the average LSI increased from 3.64 in 2000 to 5.28 in 2015, while slightly decreasing to 5.21 by 2020; and the average DIVISION fluctuated a lot, but the overall value has been higher than 0.88.

Figure 9. Temporal evolution of landscape fragmentation.

3.3. The spatial heterogeneity impacts of GEC on landscape fragmentation

To quantify the impact of GEC on landscape fragmentation, the correlation between characteristics of GEC and landscape fragmentation was established based on RDA and Spearman’s correlation coefficient. The topographic relief was excluded from the matrix of influence variables because of the high correlation between slope and topographic relief. RDA then was applied to the characteristics of the gradient expansion patches and the characteristics of the fragmentation of cultivated land and ecological land (Figure 10). In the key areas of GEC, the slope and elevation of the gradient expansion patches showed a significant negative correlation with C_PD, C_LSI and C_DIVISION of cultivated land, indicating that the fragmentation of the cultivated land decreased as the average slope and elevation of the GEC increased. As shown by the Spearman correlation coefficient analysis results (Table 2), slope had the greatest effect on C_DIVISION, while elevation had a greater effect on C_LSI. The fragmentation of cultivated land was significantly and positively correlated with the PD of GEC (p < 0.01), while it was significantly and negatively correlated with AREA_MN (p < 0.01). That indicated that the more fragmented the GEC patches were, the more fragmented the cultivated land would be, and the larger the average area of GEC patches was, the less fragmented the cultivated land would be.

Figure 10. The Results of redundancy analysis.

Table 2. Relationships between fragmentation and features of gradient expansion.

Parameter	Cultivated land			Ecological land
Correlation coefficient	C_PD	C_LSI	C_DIVISION	C_PD	C_LSI	C_DIVISION
Slope	−0.26 (**)	−0.19 (**)	−0.48 (**)	0.32 (**)	0.29 (**)	0.53 (**)
Elevation	−0.16 (*)	−0.25 (**)	−0.24 (**)	0.31 (**)	0.24 (**)	0.26 (**)
PLAND	0.13 (*)	0.12	0.21 (**)	0.01	−0.10	0.03
LPI	0.07	0.06	0.16 (*)	0.04	−0.04	0.08
PD	0.27 (**)	0.26 (**)	0.30 (**)	−0.29 (**)	−0.26 (**)	−0.26 (**)
AREA_MN	−0.20 (*)	−0.20 (**)	−0.20 (**)	0.27 (**)	0.20 (**)	0.24 (**)
ED	0.19 (*)	0.18 (**)	0.22 (**)	−0.11	−0.15 (*)	0.01
ENN_MN	−0.24 (**)	−0.23 (**)	−0.25 (**)	0.19 (**)	0.27 (**)	0.06
LSI	0.09	0.07	0.02	−0.14	−0.19 (**)	−0.11

The fragmentation of ecological land showed a diametrically opposite pattern to that of cultivated land. The topographic features of GEC showed a significantly positive correlation with the fragmentation of ecological land (p < 0.01), and the correlation coefficients of slope with C_PD, C_LSI, and C_DIVISION were 0.32, 0.29, and 0.53, while the correlations of elevation with these three fragmentation indicators were 0.31, 0.24, and 0.26. It could be seen that the ecological land fragmentation increased gradually as the slope and elevation of the GEC increased. Moreover, the fragmentation (PD) of construction land showed a significant negative correlation with the fragmentation of ecological land, and the more fragmented the construction land was, the higher the aggregation of ecological land patches was.

In order to portray the spatial heterogeneity influence of GEC on the fragmentation of cultivated land and ecological land, we analyzed the spatial evolution of fragmentation by setting up multi-distance buffers for the key areas of GEC in nine cities (Figure 11). Considering the variations in the buffer area, we selected PD (number of patches per unit area) as an indicator to portray the fragmentation of cultivated land and ecological land, as it is not influenced by the area. Analysis of Figure 11 revealed a decrease in fragmentation of both cultivated land and ecological land as the distance from the key area of GEC increased. Specifically, within a 5 km radius of the key area of GEC, the fragmentation of cultivated land was found to be higher compared to that of ecological land. However, as the distance from the key area increased, the fragmentation of cultivated land and ecological land gradually approached each other. Therefore, we infer that the rate of change in fragmentation of cultivated land under the influence of GEC is greater than that of ecological land. Based on the previous results, we believe that the impact of GEC on the fragmentation of cultivated land is greater than its impact on the fragmentation of ecological land. Ecological land is further constrained by topographical conditions. Furthermore, once the distance from the key area of GEC exceeded 3 km, the fragmentation of cultivated land and ecological land stabilized gradually, with the average PD remaining around 0.25. Hence, the influence range of GEC on landscape fragmentation is generally limited to within 3 km. The properties of the box-plot revealed that as the area of GEC got closer, the greater the fluctuations in landscape fragmentation.

Figure 11. The spatial evolution law of landscape fragmentation.

4. Discussion

4.1. The differential effects of GEC on landscape fragmentation at different spatial scales and extents

We observed a positive correlation between the slope and elevation of the GEC and ecological land fragmentation. This suggests that as the expansion of construction land increases, the fragmentation of ecological land becomes more pronounced. Landscape fragmentation is influenced by various environmental disturbances at different spatial scales, resulting in distinct patterns (Bar-Massada et al. 2012; Safaei et al. 2022). Therefore, when analyzing the process of landscape fragmentation and its influencing factors, the spatial scale effect is an issue that cannot be ignored. When the scale is too large, it will result in disturbing factors that cannot be eliminated, and when the scale is too small, it will result in some information absent. To mitigate interference factors and retain more valid information, we clustered the identified GEC patches based on distance, utilized the minimum bounding rectangle to determine the spatial scope, and excluded areas smaller than 5 km².

Based on the concept of landscape fragmentation and current research, the factors affecting landscape fragmentation mainly include natural and anthropogenic factors. The influence of natural and anthropogenic factors on landscape fragmentation is often closely related to the characteristics of the research object, the spatial scale and the environmental background. For example, in some studies anthropogenic activities have a greater effect on the ecological landscape than other natural factors such as topography and climate (Liu et al. 2011), while in other studies natural factors have a more significant impact on landscape patterns, with human activities playing only a weakly interfering role (Shi, Yin, and Gao 2021). In general, in regions where human activities are more intense, such as around the city and on both sides of the road, the impact of human activities on landscape fragmentation is huger, while in remote areas or ecologically fragile areas, climatic conditions and topographic elements play a more significant role in changing the landscape pattern (Lin et al. 2019; Liu et al. 2021). Our findings align with previous research, demonstrating that the impact of construction activities on ecological and cropland landscape fragmentation diminishes with distance from anthropogenic areas.

From the information presented above, it is evident that the impact of GEC on landscape fragmentation is influenced by both the surrounding natural environment and human activities, which give rise to various forms of land competition and compromise (Figure 12). In the key areas of GEC, human activities have a significantly greater impact on landscape fragmentation than natural factors. Consequently, the competition for construction land surpasses that of other land types, resulting in immense pressure on regional landscape ecology. As the distance from the key areas of GEC increases, the impact of human activities on landscape fragmentation progressively diminishes, and the landscape pattern is gradually dominated by natural factors. In light of our study's findings, a rational layout of construction land in sloping areas can mitigate the fragmentation of both cultivated and ecological land caused by construction activities. Therefore, planners need to strategize the distribution of construction land patches within the GEC inner region, aiming for maximum regularity. Additionally, when designating gradient development areas, it is advisable to maximize the distance between regions, ensuring a minimum separation distance of 3 km to prevent the overlapping of multiple influences and increase the pressure on landscape security patterns.

Figure 12. Factors influencing landscape fragmentation due to hillside urbanization.

4.2. Suggestions for sustainable planning

Results showed that the GEC would lead to the fragmentation of cultivated land and ecological land. Increased fragmentation will have a negative impact on regional agricultural production and the ecological environment (Figure 13). It has been found that the more fragmented cultivated land led to higher negotiation costs for infrastructure, lower accessibility of cultivated land, greater difficulty in continuous operation, and lower levels of agricultural specialization and quality of cultivated land (Pang et al. 2023). In addition, mountainous areas have a short growing season and have difficulty in carrying out mechanized agriculture on a large scale because of high altitude and complex terrain (Song, Prishchepov, and Song 2022; Klima et al. 2020). Therefore, as cultivated land shifts towards highland and the loss and fragmentation of cultivated land, the agricultural production efficiency in mountainous areas will decrease and the potential risk of regional food security will increase (Li et al. 2021). The impact of ecological land fragmentation on the overall ecological functioning is significant and can lead to a reduction in biodiversity (Haddad et al. 2015) and enhanced isolation of ecosystems (Ng, Xie, and Yu 2011). Moreover, the mountainous environment is complex, and the productivity value of ecosystems is high. When human activities shift towards highland gradually, it will have a profound impact on carbon storage and land cover (Yu et al. 2021). Therefore, based on the characteristics of the gradual expansion of construction land in different mountainous cities and their own environmental conditions, adopting different development models is crucial for reducing landscape fragmentation and promoting the sustainable development of mountainous cities.

Figure 13. Impacts of landscape fragmentation caused by GEC.

Hilly mountain cities have relatively gentle terrain, with small undulations and gradual slopes, which are conducive to urban expansion and development. Additionally, due to the relatively gentle terrain, these cities have larger scales and higher land use efficiency. Therefore, hilly mountain cities can engage in small-scale slope development, adjust the spatial layout and structure of the cities, and construct a network-type urban system, wherein multiple small and medium-sized towns replace a single central metropolis. To address slope development in these areas, it is important to incorporate the concept of compact city theory, optimize the urban structure, reduce land occupation, and achieve sustainable development (Wu et al. 2023). Valley mountain cities are characterized by valley and hillside terrain, with significant undulations. These cities have complex topography, limited space for expansion, and a strong demand for change in the regional terrain to meet urban development needs. However, the gradient expansion of such cities tends to encroach upon ecological spaces. Therefore, the concept of urban renewal can be applied, starting with the three-dimensional development of the city, increasing land use intensity, including vertical expansion of buildings and the utilization of underground space. If the population growth demands cannot be met, it is necessary to avoid major forest areas, wetlands, and other ecologically vital zones, while incorporating ecological security pattern theory for ecological monitoring and timely restoration (Xiang et al. 2023). In the case of ravine mountain cities, characterized by rugged terrain composed of gullies and ridges, these cities have complex topography, smaller scales, and are prone to geological hazards such as settlement, landslides, and mudslides. Therefore, large-scale gradient development is not suitable for ravine cities, especially those situated in loess areas like Yan'an, as blind vertical growth could exacerbate surface deformation. For such cities, it is necessary to optimize the land use structure and plan the direction and scale of urban expansion rationally. Large-scale construction should be avoided in ecologically sensitive areas, areas prone to geological hazards, and water conservation zones.

4.3. Limitation and prospects

Based on the traditional threshold method, we proposed a multi-temporal identification method for the GEC by combining the neighborhood characteristics of GEC. We then utilized the spatial distance clustering method to divide the key areas of GEC into statistical units, which served as a basis for our analysis. From the findings obtained through this research, we identified a significant influence of GEC on landscape fragmentation, along with determining the threshold distance of this influence. However, it is important to note that our quantification of the spatial impacts of GEC was limited due to inconsistency in the size of the key areas. As a representative measure that is not affected by the area of buffers, we selected PD to evaluate landscape fragmentation, even though it represents only one aspect. In addition, the GEC has a certain impact on the geological environment, which can lead to landslides, mudslides, geological subsidence and other geological disasters. We can focus on analyzing the diversity impacts of GEC and developing advanced models and producing accurate data to monitor suitable construction areas for gradient development in mountain cities. This will help to realize the coordinated development of economy and environment, improve agricultural production, and form a new model of urban development with mountain characteristics.

5. Conclusions

Starting from the geographic phenomenon of GEC of typical mountainous cities in China, we proposed a method for identifying multi-period GEC and drew the key areas of GEC by spatial clustering based on distance. Besides, we quantified the impacts of the GEC on the fragmentation of cultivated land and ecological land based on RDA and multi-buffer analysis. Based on the questions we raised, the main conclusions are as follows:

1. Our proposed methodology for GEC identification reveals the actual spatial extent of GEC. From the perspective of expansion patterns, the GEC mainly consists of two types: edge-type expansion and enclave-type expansion. From the perspective of the expansion direction, there are three types of GEC: ‘unidirectional’, ‘bidirectional’, and ‘multidirectional’, which are determined by the terrain in different directions of the city. The overall GEC scale exhibits a correlation with the city's economic status. Chongqing, as one of the economic centers in the southwest region, has the largest expansion area, reaching 1,059.70 km². In contrast, Xining, located in the northwest, has an expansion area of only 51.33 km². Combined with the multi-distance buffer analysis, we also found that the impact scope of GEC on regional landscape fragmentation is within 3 km. When the distance exceeds 3 km from the key areas of GEC, the average PD of cultivated land and ecological land remains below 0.3, while the fragmentation of the central area is triple.

2. With the landscape pattern indices such as PD, LSI, and DIVISION, we have depicted the long-term changes in the landscape characteristics within the key areas of GEC. It is found that in the key areas of GEC, the degree of fragmentation of ecological land and cultivated land has gradually increased. From 2000 to 2020, a rise in the average PD of cultivated land from 0.38 to 0.94, in the average LSI from 4.19 to 6.04, and in the average DIVISION from 0.54 to 0.96 was observed. Concurrently, the average PD of ecological land increased from 0.35 to 0.69, the average LSI value from 3.64 to 5.21, and the average DIVISION value was consistently over 0.85. Overall, the degree of fragmentation of cultivated land is slightly higher than that of ecological land.

3. Through statistical analysis, we found that the laws of the impact of GEC's landscape features and terrain features on cultivated land and ecological land are opposite. Specifically, with increasing average slope and altitude of GEC, the fragmentation of cultivated land shows a decline, whereas the ecological land fragmentation intensifies. Contrarily, within key areas of GEC, heightened fragmentation is observed with increased dispersal of construction land patches and cultivated land, while the fragmentation degree of ecological land decreases.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability Statement

The land use data that support the findings of this study are available in [CNLUCC] at [10.12078/2018070201]. The DEM data that support the findings of this study are available in [NASA EOSDIS Land Processes Distributed Active Archive Center] at https://doi.org/[10.5067/MEaSUREs/SRTM/SRTMGL1.003]. The Spatial distribution data of landform types in China that support the findings of this study are openly available in ‘Resources and Environment Science and Data Center’ at https://www.resdc.cn/data.aspx?DATAID=124.

Additional information

Funding

This study was supported by the National Natural Science Foundation of China [grant number: 42271214], National Key R&D Program of China [grant number: 2022YFC3800700], Natural Science Foundation of Gansu Province [grant number: 21JR7RA281, 21JR7RA278], the CAS’ Light of West China’ Program [grant number: 2020XBZGXBQNXZ-A], Basic Research Top Talent Plan of Lanzhou Jiaotong University [grant number: 2022JC01], the Gansu Province “Innovation Star” Project [grant number: 2023CXZX-541].