The Expansion Mechanism of Rural Residential Land and Implications for Sustainable Regional Development: Evidence from the Baota District in China’s Loess Plateau

Rural residential land is the main space of a farmer’s life, rural culture, and social relations. Prior research of rural residential land has focused more on its evolvement in plain and traditional agricultural areas. Yet, there is no clear picture of rural residential land expansion, especially in ecologically fragile areas. This study analyzed the characteristics of rural residential land expansion based on 30 m spatial resolution land-use datasets of the Baota District of Yan’an City, Shannxi Province, and further explored the influencing factors and mechanisms of rural residential land expansion through binary logistic regression (BLR) modeling. Our findings indicated that the area of rural residential land in the Baota District increased by 116.16% during 1990–2015. More than 75% of the residential land expansion came from the occupation of cropland. Moreover, rural residential land expansion was heterogeneous in the rural regional system. The expansion scale, speed, and mode diversity of rural residential land decreased with the increased distance to urban built-up areas. Geographical conditions and resource endowments are the primary internal driving factors; urbanization and policy implementation are two major external driving forces. The authors suggest that the realization of regional sustainable development in ecologically fragile areas should strengthen urban–rural integration, focus on constructing central towns, and ensure ecological protection measures.


Introduction
Rural residential land is the main space of a farmer's life, rural culture, and social relations [1], reflecting the human-land relationship, historical background, and sociopolitical relations in rural areas [2,3]. Since the 19th century, geographers have carried out a range of research on rural residential land [4]. Prior research mainly focused on the spatial distribution of rural residential land and the impact on the natural environment [5,6]. Researchers gradually explored rural residential land's spatiotemporal evolving patterns, influencing factors, and dynamic modeling [7][8][9] with the rapid development of "3S" technology and mathematical models. The structural equation model, geographic detector, and logistic regression are the methods most frequently used to detect influencing factors [10][11][12]. Meanwhile, attention has also been paid to rural landscape characteristics [13,14], the impact of rural residential land change on the ecosystem [15], the evolution of rural residential land's function [16,17], and rural residential land reconstruction [18,19]. Researchers generally agree that significant changes have taken place in rural residential land over the past decades.
The hyper-fast change of rural residential land leads to the saliency of sustainable development and urban-rural integration, which have become the focal points for decision-Land 2021, 10, 172 2 of 16 makers and rural planners [20,21]. The rapid growth of residential land is usually at the cost of occupying cultivated land and ecological land [22,23], which requires land planners to tradeoff the conflicts among the human living, production, and ecological land. For instance, the expansion of residential land provides residents with more living space, but this might also result in a deficiency of industrial development and job opportunity [24]. Therefore, decision-makers and rural planners strive to achieve sustainable growth and urban-rural integration through effective management of rural residential land expansion.
To achieve sustainable regional development and urban-rural integration, it is crucial to recognize the evolution progress and formation mechanism of rural residential land expansion. However, extant research has only explored rural residential land spatial distribution and area change [25,26], ignoring the change characteristics and mechanisms of newly added residential land [27]. In addition, most of these studies focused on plain areas [28,29] and traditional agricultural areas [30], but less on mountainous areas and ecologically fragile areas [31].
As the world's largest developing country, China's rapid urbanization has driven unprecedented rural residential land changes over the past few decades. From 2007 to 2015, China experienced a 257.27 million hectare increase in rural residential land [25]. Some problems such as population outflow, eco-environmental pollution, and economic recession, appeared in rural development with the rapid progress of urbanization [32,33]. More importantly, the contradiction between population outflow and rural residential land expansion is becoming increasingly prominent, which seriously restricts sustainable regional development in rural areas. In 2012, China implemented ecological civilization construction as a national development strategy. Therefore, the question of how to realize urban-rural integrated development in ecologically fragile areas has become an urgent problem to be addressed. This paper aims to examine the spatiotemporal evolution of rural residential land expansion in the Baota District of Yan'an City, Shannxi Province, and explore how and why the expansion differs across rural areas. This research's findings help deepen our understanding of rural residential land expansion and its formation mechanism, especially in ecologically fragile areas. Further, this research makes strategic recommendations for sustainable regional development and urban-rural integrated development in ecologically fragile regions.

The Context of the Chinese Rural Homestead System
Rural residential land reflects the socio-political relations in rural areas, and the rural homestead system also affects the development of rural residential land [2]. The current Chinese rural homestead property rights system originates from rural land reform in the 1950s and can be divided into four stages [34]. From 1949 to 1962, after the completion of rural land reform, the central government confirmed that rural homesteads and houses were owned by farmers, allowing farmers to buy, sell, inherit, and rent homesteads and houses freely. During this period, dwellers mainly applied for homesteads to meet their own housing needs, and the area of residential land increased slowly. From 1963 to 1981, People's Communes were established in rural areas. The central government recognized farmers' long-term ownership and the use rights of their homesteads, and allowed homesteads to be sold and leased. Meanwhile, the Economic Reform and Open up policy in 1978 promoted a house-building craze in the countryside due to farmers' increased farming incomes and their accompanying demand for better living conditions [35]. This housebuilding craze triggered the phenomenon of occupying cropland for newly added rural residential land. From 1982 to the end of 1996, the central government stipulated that both urban and rural residents could obtain the right to use rural homesteads, but mainly rural residents. During this period, a large amount of high-quality cropland was occupied and thus threatened national food security. For this reason, the Chinese central government implemented strict policies to protect the limited cropland. After 1997, the national law stipulated that rural collectives own rural residential land, and only rural dwellers have Land 2021, 10, 172 3 of 16 the right to use it [36]. Consequently, all rural residential land transactions between urban dwellers and village collectives were outlawed [35]. However, the rapid urbanization process had brought challenges to the implementation of the rural housing system. Many rural dwellers and even rural collectives were lured by the substantial income from renting or selling their houses on collective land to urban dwellers [37]. This challenge resulted in the development of the informal land markets, including small-property-rights housing and "urban villages".
In addition, the poor operation of the rural homestead withdrawal mechanism had led to a prominent problem of building new houses without demolishing old ones. Statistics showed that there were 287 million migrant workers in China [38]. A large proportion of these migrant workers' rural houses were either under-used or entirely left idle before they return to rural areas. Under the current rural house system, the redistribution and circulation of abandoned or under-used houses were discouraged. Therefore, eligible new households were more inclined to apply for new residential land. As a result, the abandonment and expansion of residential land, which coexisted in China's rural areas, exacerbated the human-land relationship conflicts.

An Explanatory Framework for Rural Residential Land Expansion
Rural residential land change is affected by a variety of factors, including geographical conditions, resource endowments, urbanization, industrialization, urban-rural relations, and policy systems [28][29][30][31], etc. (Figure 1). Topography and physiognomy are the primary factors affecting rural settlements size and location choice [39], especially in mountainous and hilly areas. In general, the scales of residential land in plain areas are large, and the scales of residential land in mountainous areas are relatively small. Elevation and slope have greatly influenced the safety and convenience of residents living and production conditions [40]. Resource endowments include natural resources (light, temperature, cropland, etc.) and social resources (population, technology, knowledge, etc.). Cropland and human resources play a vital role in the change of residential land. Cropland resources determine the upper limit of the population carrying capacity in a particular area, thus restricting the development scale of residential land [41]. Human resources directly affect the changes in residential land through housing demand [42]. Additionally, the effect of natural resources such as light, temperature, and precipitation on residential land changes is mainly reflected on the macro scale [43]. residents. During this period, a large amount of high-quality cropland was occupied and thus threatened national food security. For this reason, the Chinese central government implemented strict policies to protect the limited cropland. After 1997, the national law stipulated that rural collectives own rural residential land, and only rural dwellers have the right to use it [36]. Consequently, all rural residential land transactions between urban dwellers and village collectives were outlawed [35]. However, the rapid urbanization process had brought challenges to the implementation of the rural housing system. Many rural dwellers and even rural collectives were lured by the substantial income from renting or selling their houses on collective land to urban dwellers [37]. This challenge resulted in the development of the informal land markets, including small-property-rights housing and "urban villages". In addition, the poor operation of the rural homestead withdrawal mechanism had led to a prominent problem of building new houses without demolishing old ones. Statistics showed that there were 287 million migrant workers in China [38]. A large proportion of these migrant workers' rural houses were either under-used or entirely left idle before they return to rural areas. Under the current rural house system, the redistribution and circulation of abandoned or under-used houses were discouraged. Therefore, eligible new households were more inclined to apply for new residential land. As a result, the abandonment and expansion of residential land, which coexisted in China's rural areas, exacerbated the human-land relationship conflicts.

An Explanatory Framework for Rural Residential Land Expansion
Rural residential land change is affected by a variety of factors, including geographical conditions, resource endowments, urbanization, industrialization, urban-rural relations, and policy systems [28][29][30][31], etc. (Figure 1). Topography and physiognomy are the primary factors affecting rural settlements′ size and location choice [39], especially in mountainous and hilly areas. In general, the scales of residential land in plain areas are large, and the scales of residential land in mountainous areas are relatively small. Elevation and slope have greatly influenced the safety and convenience of residents′ living and production conditions [40]. Resource endowments include natural resources (light, temperature, cropland, etc.) and social resources (population, technology, knowledge, etc.). Cropland and human resources play a vital role in the change of residential land. Cropland resources determine the upper limit of the population carrying capacity in a particular area, thus restricting the development scale of residential land [41]. Human resources directly affect the changes in residential land through housing demand [42]. Additionally, the effect of natural resources such as light, temperature, and precipitation on residential land changes is mainly reflected on the macro scale [43].  The relationship between rural living and production space has changed with the advance of urbanization, industrialization, and the enhanced level of economic development [39]. Firstly, the rapid development of urbanization and industrialization provide more employment opportunities for rural residents and attract laborers to concentrate in cities and towns. Villages near cities have been affected more significantly [7,44]. Consequently, farmers' willingness to renovate and expand their original houses has gradually increased with the non-agriculturalization of the employment structure and increased family income. However, some factories and enterprises began to move to rural areas due to rising urban land rents. The rapid expansion of urban areas will also encroach on surrounding rural areas, resulting in a decrease in rural residential land [45].
With the transformation of urban-rural relations and the strengthening of rural supporting policies, China's rural residential land has entered a new transformation stage. The long-standing urban-rural dual system in China has led to an imbalance of urban-rural relations, with rural development elements flowing unidirectionally to cities and many villages beginning to decline and die [46]. Since the Third Plenary Session of the 16th Central Committee of the Communist Party of China proposed the coordinated development of urban and rural areas in 2003, the State has successively implemented rural supporting policies such as small-town construction, new countryside building, and rural revitalization. Implementing these policies is conducive to enhancing rural development's vitality and realizing the reasonable flow of urban and rural elements [47].

Study Area
Baota District (36 • 10 N-37 • 02 N, 109 • 14 E-110 • 50 E) locates in the middle of the Loess Plateau. With the national strategy of western regional development and ecological civilization construction, this district has become one of the most typical cities in China's fragile ecological areas [19]. The district covers approximately 3539 km 2 (0.55% of the Loess Plateau's whole). With its perennial average sediment transportation at 1.6 × 10 8 t (4.33% of the Loess Plateau's whole), this city is recognized as a typical fragile ecological area for severe soil erosion in China's Loess Plateau ( Figure 2). It is also considered a typical valley city. The gully density of the Baota District is 3.04-5.01 km per km 2 , and the geographical area of the valley accounts for 41-46% of its total coverage. Compared with the plain area, the development space of the Baota District is very limited.
The relationship between rural living and production space has changed with the advance of urbanization, industrialization, and the enhanced level of economic development [39]. Firstly, the rapid development of urbanization and industrialization provide more employment opportunities for rural residents and attract laborers to concentrate in cities and towns. Villages near cities have been affected more significantly [7,44]. Consequently, farmers' willingness to renovate and expand their original houses has gradually increased with the non-agriculturalization of the employment structure and increased family income. However, some factories and enterprises began to move to rural areas due to rising urban land rents. The rapid expansion of urban areas will also encroach on surrounding rural areas, resulting in a decrease in rural residential land [45].
With the transformation of urban-rural relations and the strengthening of rural supporting policies, China's rural residential land has entered a new transformation stage. The long-standing urban-rural dual system in China has led to an imbalance of urbanrural relations, with rural development elements flowing unidirectionally to cities and many villages beginning to decline and die [46]. Since the Third Plenary Session of the 16th Central Committee of the Communist Party of China proposed the coordinated development of urban and rural areas in 2003, the State has successively implemented rural supporting policies such as small-town construction, new countryside building, and rural revitalization. Implementing these policies is conducive to enhancing rural development's vitality and realizing the reasonable flow of urban and rural elements [47].

Study Area
Baota District (36°10′ N-37°02′ N, 109°14′E-110°50′ E) locates in the middle of the Loess Plateau. With the national strategy of western regional development and ecological civilization construction, this district has become one of the most typical cities in China's fragile ecological areas [19]. The district covers approximately 3,539 km 2 (0.55% of the Loess Plateau's whole). With its perennial average sediment transportation at 1.6 × 10 8 t (4.33% of the Loess Plateau's whole), this city is recognized as a typical fragile ecological area for severe soil erosion in China's Loess Plateau ( Figure 2). It is also considered a typical valley city. The gully density of the Baota District is 3.04-5.01 km per km 2 , and the geographical area of the valley accounts for 41-46% of its total coverage. Compared with the plain area, the development space of the Baota District is very limited.  The district contains 5 streets, 12 towns, and 623 administrative villages, with a population of 478,782. During the process of urbanization and industrialization, the conflicts between population growth and land supply in Baota have intensified. From 1990 to 2015, the average annual growth rate of the rural population in Baota was 0.7%, while the average annual growth rate of rural residential land was 3.1%. The growth rate of rural residential land was much faster than the population growth rate, which restricted sustainable regional development. Hence, the Baota District is a good sample for studying the rural residential land expansion of ecologically fragile areas.

Data Sources and Preprocessing
The land-use datasets of Baota (1990Baota ( , 1995Baota ( , 2000Baota ( , 2005Baota ( , 2010, and 2015; 30 m spatial resolution) were acquired from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC) [48]. Land-use types were classified as cropland, forestland, grassland, water bodies, urban built-up land, rural residential land, industrial land, and unused land. We also adopted field survey and image interpretation to verify the precision of the classified land-use map by randomly selecting 600 points (100 points each year) from the classified land-use map. Rural residential land in this paper is defined as the land for building houses, other constructions for living needs, and the necessary infrastructure.
Socio-economic data referring to labor employment structure, farmer income, and rural population, etc., were mainly collected from the 2015 Statistical Yearbook of Baota District, village statistical data, and partly from the practical survey. Digital elevation model (DEM) data was collected from the International Scientific Data Mirroring Website of the Computer Network Information Center of the Chinese Academy of Sciences (http://www.datamrror.csdb.cn, at 16 October 2018). Elevation, slope, and aspect were obtained based on DEM with ArcGIS 10.2 software (Redlands, CA, USA).

Rural Residential Land Expansion Modes
This paper used the landscape expansion index (LEI) to identify rural residential land expansion modes [27,49]. The LEI of each new rural residential land patch was calculated using Equation (1).
where A 1 was the intersection area between the buffer zone and the previously existing rural residential land patch, and A 2 was the intersection area between the buffer zone and previously non-rural residential land. Based on the LEI, rural residential land expansion could be defined into three modes: edge expansion, leapfrog, and infilling. The edge expansion mode was defined as a new patch of rural residential land spreading from an edge, and the LEI was between 0 and 50. If a new patch was isolated from an old patch and the LEI was equal to 0, we defined this mode as leapfrog. The infilling mode referred to the gaps among existing rural residential land being filled with newly added rural residential land, and the LEI was between 50 and 100.

Getis-Ord Gi*
The Getis-Ord Gi* model was used to analyze the spatial agglomeration characteristics of the rural residential land expansion scale. The Euclidean distance was selected as the spatial distance calculation method. The null hypothesis for the pattern analysis tool was complete spatial randomness of either the features themselves or the values associated with those features. The resultant Z(Gi*) and p-values indicate where the features with high or low values were spatially clustered [50].
Gi*(d) was normalized to obtain Z(Gi*). E(Gi*) and var(Gi*) were the mathematical expectation and variance of Gi*(d), respectively. If Z(Gi*) was positive and statistically significant, it indicated location i as a hot area of high-value clustering. If Z(Gi*) was negative and statistically significant, it indicated location i as a cold area of low-value clustering.

Potential Influencing Factors Selection
The above analysis about rural residential land evolution in the Baota District found that rural residential land evolution was driven by the interaction between internal and external factors, which could be categorized as basic geographical conditions, resource endowment, urbanization industrialization, and policy regulation (Table 1). Existing research indicated that topography and resource endowment were the main internal driving factors that affected the change of rural residential land [51], given the effect of light, temperature, precipitation, technology, and knowledge on residential land changes is mainly reflected on the macro scale [43]. Therefore, the topography (elevation, slope, and aspect), geographical location (distance to cropland, distance to rivers), and resource endowment (total area of cropland, per capita cropland, total population, population density) were selected as internal driving factors of rural residential land expansion. In a similar vein, urbanization, industrialization, and policy implication were the main external factors that affect the evolution of rural residential land [52]. As the Baota District is a typical valley city in China's Loess Plateau, traffic conditions play a crucial role in the process of rural development. Besides, the Baota District has successively implemented the policies of small-town construction and new countryside construction since 2003. The implementation of these policies is conducive to enhancing the vitality of rural development, but they were scantly noticed in previous research. Therefore, we selected economic locations (distance to urban area, distance to state road, distance to provincial road), rural industrial transformation (proportion of non-agricultural labor), economic output (per capita income), and policy implementation (small-town construction and new countryside building) as external factors of rural residential land expansion.

Binary Logistic Regression (BLR)
The BLR model is suitable for categorical variables whose dependent variable is binary [53]. In this paper, we defined whether the land patch was an expansion patch of rural residential land as a dependent variable (Yes = 1, No = 0). The formula of BLR is as follows: ln where P was the probability of rural residential land expansion; x k was the influencing factor; β k was the undetermined parameters; and β 0 was the intercept.

Temporal Change of Rural Residential Land
As shown in Figure 3, the rural residential land area in Baota increased significantly during 1990-2015. The rural residential land area increased by 116.16% from 1990 to 2015 with an average annual change rate (AACR) of 3.13%, and 84.3% of the newly added rural residential land was distributed below a slope of 15 • . Specifically, edge expansion was the primary pattern of rural residential land expansion in Baota during 1990-2015, at over 2.5 times the total area of leapfrog and infilling. The edge expansion pattern amounted to 757.34 ha, accounting for 71.76% of the total expanded area in rural residential land. Leapfrog and infilling patterns were 211.92 ha and 86.05 ha, accounting for 20.08% and 8.16% of the total expanded area in rural residential land, respectively. Besides, the residential areas with leapfrog and infilling patterns increased significantly from 5. The BLR model is suitable for categorical variables whose dependent variable is binary [53]. In this paper, we defined whether the land patch was an expansion patch of rural residential land as a dependent variable (Yes = 1, No = 0). The formula of BLR is as follows: where P was the probability of rural residential land expansion; xk was the influencing factor; βk was the undetermined parameters; and β0 was the intercept.

Temporal Change of Rural Residential Land
As shown in Figure 3, the rural residential land area in Baota increased significantly during 1990-2015. The rural residential land area increased by 116.16% from 1990 to 2015 with an average annual change rate (AACR) of 3.13%, and 84.3% of the newly added rural residential land was distributed below a slope of 15°. Specifically, edge expansion was the primary pattern of rural residential land expansion in Baota during 1990-2015, at over 2.5 times the total area of leapfrog and infilling. The edge expansion pattern amounted to 757.34 ha, accounting for 71.76% of the total expanded area in rural residential land. Leapfrog and infilling patterns were 211.92 ha and 86.05 ha, accounting for 20.08% and 8.16% of the total expanded area in rural residential land, respectively. Besides, the residential areas with leapfrog and infilling patterns increased significantly from 5.   Cropland was the main land-use type transferred to rural residential land in Baota, accounting for 75.52% of the total area of rural residential land expansion between 1990 and 2015 ( Table 2). The area of grassland transformed to rural residential land was 161.65 ha, or 15.32% of the total expansion in rural residential land. Areas of woodland were reduced by approximately 67 ha, or about 6% of the total rural residential land expansion. Less than 28 ha of river and pond were transferred to rural settlement land, accounting for nearly 3% of the total area of rural residential land expansion. Edge expansion, leapfrog, and infilling modes of rural residential land in Baota caused a cropland loss of 556. 16

Spatial Agglomeration Changes
Using the Getis-Ord Gi* method, we analyzed the spatial agglomeration change characteristic of rural residential land expansion in Baota during 1990-2015. The spatial agglomeration of rural residential expansion was different during each period ( Figure 4). Overall, the expansion of rural residential land was mainly concentrated around the city and in the river valley, and there was very little expansion of residents in deep gullies.

Uneven Expansion of Rural Residential Land
To further explore expansion characteristics, this paper analyzed the change of scale, speed, and expansion mode diversity of rural residential land in buffer zones with different distances to urban areas ( Figure 5). With the increase of buffer distance, the expansion scale, speed, and expansion mode diversity of rural residential land showed the characteristics of decreasing volatility, especially in expansion scale. The reason for the sudden increase of rural residential land expansion scale within the 30-35 km buffer zone may be ascribed to the implementation of the population concentration project in Nanniwan Town, which stimulated the increase of rural residential land. Specifically, the newly added rural residential land was mainly concentrated within the buffer zone's 10 km range and accounted for 80% of the total expansion scale. In terms of expansion speed, the AACR of rural residential land within 10 km of the buffer zone was 5.3%, which was more than twice the AACR of the whole study area. From the perspective of expansion mode, the expansion mode of residential land gradually changed from diversification (edge expansion, infilling, leapfrog) to single (edge expansion). Overall, the expansion of rural residential land was mainly concentrated around the city and in the river valley, and there was very little expansion of residents in deep gullies.

Uneven Expansion of Rural Residential Land
To further explore expansion characteristics, this paper analyzed the change of scale, speed, and expansion mode diversity of rural residential land in buffer zones with different distances to urban areas ( Figure 5). With the increase of buffer distance, the expansion scale, speed, and expansion mode diversity of rural residential land showed the characteristics of decreasing volatility, especially in expansion scale. The reason for the sudden increase of rural residential land expansion scale within the 30-35km buffer zone may be ascribed to the implementation of the population concentration project in Nanniwan Town, which stimulated the increase of rural residential land. Specifically, the newly added rural residential land was mainly concentrated within the buffer zone's 10km range and accounted for 80% of the total expansion scale. In terms of expansion speed, the AACR of rural residential land within 10km of the buffer zone was 5.3%, which was more than twice the AACR of the whole study area. From the perspective of expansion mode, the expansion mode of residential land gradually changed from diversification (edge expansion, infilling, leapfrog) to single (edge expansion).

Identifying the Driving Factors of Rural Residential Land Expansion
The regression results of the BLR of rural residential land expansion were presented in Table 3. The results showed good explanatory power of the model for rural residential land expansion. The adjusted value of R 2 was more than 0.5, and the p-value of the Hosmer-Lemeshow test was more than 0.5. Table 3. Coefficients of driving factors of rural residential land expansion.

Identifying the Driving Factors of Rural Residential Land Expansion
The regression results of the BLR of rural residential land expansion were presented in Table 3. The results showed good explanatory power of the model for rural residential land expansion. The adjusted value of R 2 was more than 0.5, and the p-value of the Hosmer-Lemeshow test was more than 0.5. One-way ANOVA (** = p < 0.01; * = p < 0.05). Table 3 demonstrated that elevation and slope had a significant negative effect on the rural residential land expansion (p < 0.01), which indicates that mountain landform had insufficient space for rural residential land expansion and limited their development. The variable "total area of cropland" had a negative effect on the rural residential land expansion (p < 0.01), and the "distance to cropland" had a positive effect on it (p < 0.05). These findings appeared to confirm that the restriction of agricultural resource conditions on newly added rural residential land had weakened [54]. Besides, the coefficient signs of the total population and population density were positive (p < 0.01). This result was in line with the findings that larger villages tend to have more services than smaller villages [55], and this was also the case in villages with higher population density.
The variable "distance to urban areas" had a negative effect on the rural residential land expansion (p < 0.01), whereas "proportion of non-agricultural labor" and "per capita income" had a positive effect on it (p < 0.05). These findings suggested that the rising urbanization rate has caused a change in demographic structure, especially the increase in non-agricultural employment opportunities, which have significantly promoted the nonagricultural transition of the rural population. Also, there was a strong correlation between rural residential land expansion and economic locations, whereas the sign of the coefficient was the opposite, which implied that the location advantage showed a significant distance attenuation effect. The transportation network was the bridge for the flow of production factors between urban and rural areas. Better traffic conditions effectively promoted the flow of elements between urban and rural areas, thereby promoting rural development. For convenient transportation and more job opportunities, most of the newly added residential land was located near urban areas and main roads. Moreover, the "small-town construction" implementation had positive effects (p < 0.05) on rural residential land expansion, though the effect of the "new countryside building" variable was not significant (p > 0.05). This finding indicated that government policies and assistance had facilitated rural residential land expansion, and townships played a crucial central role in rural areas [43].

Analysis of the Formation Mechanism of Rural Residential Land Expansion
The above analysis confirmed that rural residential land expansion was driven by internal and external factors. On the one hand, the geographical conditions are direct, stable, and primary internal influencing factors of rural residential land expansion. On the other hand, rural resource conditions are another type of main internal driving factors, especially land resources and human resources. Urbanization, industrialization, and policy implementation are major external driving forces. In addition, the imbalance of urbanrural development exacerbates the gap between urban and rural areas and strengthens the impact of urbanization on rural residential land expansion. However, given that rural residential land expansion's formation mechanism is a complicated and systematic process, this paper focuses on the economic location, rural society, rural support policies, and planning.
As a spatial feature, economic location is closely related to urbanization's influence intensity on rural residential land expansion. Benefiting from economic location advantages, the dual effect of urbanization and policy implementation prompts the radical shift of demographic and employment structure in the villages [47]. Residents earn more income through non-agricultural work and realize wealth accumulation, especially in peri-urban areas [56,57]. This also provides an economic guarantee for residents to pursue better living conditions [58]. For example, the proportion of the non-agricultural employment population in suburban villages of the Baota District was 13.2 percentage points higher than the average level, and the per capita net income in suburban villages was 1367 CNY (Chinese yuan) higher than the average level. Furthermore, the regional income disparities attract residents to gather in high-income areas [59]. China's urban-rural division and city priority development policy further increase the income gap. Thus, 80% of new rural residential land was distributed within 10km of the urban built-up area.
Population growth and family structure miniaturization increase the demand for residential land. From 1990 to 2015, the total population of the Baota District increased by 1.48 times, and the total number of households by 2.41 times. Rural population growth inevitably leads to residential land expansion [60]. This view was also confirmed by the BLR regression results on rural residential land expansion. With the advancement of people's viewpoints on family size, the rural family structure gradually turns to miniaturization. The average household size in the Baota District decreased from 4.1 in 1990 to 2.5 in 2015. This family structure's miniaturization trend accelerates the growth of numbers of rural households and intensifies rural residential land expansion [52].
The implementation of national and local rural development policies has different effects on rural residential land expansion. In 1979, the State Council promulgated the "Provisions on Several Issues Concerning the Development of Communal Enterprises", which provided supporting policies for communal enterprises' development. In terms of tax policy, the State applied low tax rates and tax exemption policies, which also promoted the development of communal enterprises. During this period, the township enterprise number in the Baota District grew from 536 in 1979 to 5384 in 1993, and the rural residential land increased from 908.46 hm 2 in 1980 to 1113.21 hm 2 in 1995. Before 2002, the local government occupied a large amount of dam land for construction land development and achieved farmland balance through the reclamation of slope farmland. On 6 December 2002, the 66th executive meeting of the State Council adopted the Regulations on Conversion of Farmland to Forests to strengthen ecological and environmental protection. This policy limited the reclamation of sloping land and reduced the new farmland source. As a result, the rate of rural residential land expansion slowed. In addition, small-town construction promoted surrounding residents from villages to towns by supporting industrial parks and improving infrastructures such as transportation, education, and medical care. Although these new countryside infrastructures obviously improved rural residents' living and production conditions, they also accelerated the rural population loss due to the development gap among regions [61]. Compared with townships, small villages' endogenous development motivation is insufficient, especially in terms of industry. Therefore, the implementation of new countryside construction renders the difficulty of attracting or even retaining residents in rural villages due to the lack of employment opportunities [62].
Rural planning decisions guide the direction of new rural residential land distribution. As a result of the severe rainfall disaster in 2013, traditional dwellings (cave dwellings) in the Baota District were seriously damaged. Local policymakers strictly control the housing structures and the new residential land location to cope with potentially heavy rainfall disasters in the future. On the one hand, it restricts the approval and construction of the homestead in gully areas; on the other hand, it encourages rural residents to live in the new communities. This planning decision facilitates the centralized distribution of new residential land.

Implications for Sustainable Regional Development
The comprehensive impact of internal and external environmental factors stimulated the uneven expansion of rural residential land. This spatial difference is more salient, especially in the context of unbalanced urban-rural relations. China implemented a series of rural support policies such as new countryside construction, and these policies have not changed the decline in some rural areas. Hence the rural residential land change is the result of the spatial reorganization process among various factors, ranging from population, land, to the industry in the rural regional system. It is difficult to rebuild the declining countryside, especially in China. However, innovative ways could be explored based on rural development trends for the purpose of sustainable regional development. In light of this, this current research suggests policy implications for sustainable regional development in ecologically fragile areas as follows: urban-rural integration, central towns reconstruction, and ecosystem protection.

•
New urban-rural integration. Speeding rural modernization cannot only depend on rural areas. Urban-rural integration is the only way to realize rural modernization [46]. Previous understandings surrounding the urban-rural relationship present a kind of antagonistic and unequal relationship. In fact, the urban-rural relationship is similar to a mother-child relationship, i.e., the countryside gave birth to a city, and the city is contained in the countryside. Urban and rural are closely related to each other in people, goods, capital, and other social transactions. Rural development needs to be supported by urban markets and services. However, the expected trickledown effect remains dissatisfied [63], and industrial support was also absent in the practice of new countryside construction. Policymakers are encouraged to employ pertinent intervention to create a long-term mechanism for urban areas helping rural areas: for instance, (1) promoting urban-rural integration planning and realizing the complementary functions in urban and rural areas; (2) facilitating the integration of urban-rural infrastructure and public facilities, improving the production and living conditions of farmers; and (3) advancing the integration of urban-rural industries and employment, and realizing the balance of housing and work. • Central town reconstruction. Central towns play significant prominent roles in the rural regional system, providing necessary public services for surrounding areas, and they are also the gathering centers of regional resource elements [43,64]. The results of this study also indicated that the "small-town construction" implementation had positive effects on rural residential land expansion. However, the construction progress of central towns in the Baota District is slow, and the central towns in the southern area of the Baota District have not yet been formed. On the one hand, due to the dualistic structure system of urban and rural areas, migrant workers entering the city cannot settle down, and their empty homesteads remain in the village [58], which hinders the relocation and integration of villages. On the other hand, the currently paid exit mechanism for rural homesteads needs to change, and the circulation of the rural land element of the market is not frictionless [65]. Moreover, the construction and management of central towns require talented people, but the loss and shortage of rural talents cannot support the construction of central towns [66]. Therefore, central town construction in the Baota District should first break the institutional bottleneck and achieve a reasonable flow of urban and rural elements. Secondly, policies should be introduced to attract talents back to villages so as to improve the capacity of endogenous development of villages. Finally, multi-center town structures should be set up, especially for central town construction in the south of the Baota District. In addition, in the process of multi-center town construction, decision-makers also need to ensure the employment supply in central towns [67]. • Ecological conservation area construction. Loess Plateau is one of the regions most seriously affected by soil erosion in the world [68,69]. The Chinese government implemented the "grain for green" projection to protect the local eco-environment. However, a large amount of grassland and woodland is still used for residential land construction. There were 161.65 ha of grassland and 66.71 ha of woodland transformed into residential land in the Baota District during 1990-2015. Therefore, it is necessary to construct ecological conservation areas. The results of this study indicated that the expansion of rural residential land was mainly concentrated around the city and in the river valley, and 84.3% of the newly added rural residential land was distributed below a slope of 15 • . Previous studies have found that the soil erosion-sensitive areas on the Loess Plateau mainly occur in the zones with a slope of 8 • -15 • and 15 • -25 • [70]. Therefore, rural planners could plan rural life and ecological space according to the spatial difference characteristics of residential land expansion. Specifically, strict control measures such as residential land boundaries should be implemented in areas with a slope of 8 • to 15 • . The ecological protection region should concentrate on the regions with a slope above 15 • .

Conclusions
This research analyzed the spatiotemporal changes of rural residential land in a typical ecologically fragile area: the Baota District. It also examined the driving factors of these changes based on the BLR modeling. To further explore the expansion characteristics, we analyzed the change of rural residential land in buffer zones with different distances. The results indicated that: (1) rural residential land increased significantly in the Baota District, and cropland was the primary land-use type transferred to rural residential land.
(2) Rural residential land expansion was imbalanced in the rural regional system. The expansion scale, speed, and expansion mode diversity of rural residential land decreased, along with the increase in distance to urban built-up areas. (3) The uneven expansion of rural residential land was driven by the interaction and reorganization of external and internal factors. The geographical conditions and resource endowment were the main internal driving factors, while urbanization and policy implementation were two major external driving forces. Additionally, the imbalance of urban-rural relations exacerbated the gap between urban and rural areas and strengthened the impact of urbanization on rural residential land expansion. Furthermore, we argued that the realization of sustainable regional development in ecologically fragile areas should strengthen the urban-rural integration, focus on the construction of central towns, and ensure ecological protection measures. These findings and implications are applicable to a large number of ecologically fragile cities in mountainous areas.
Previous studies mainly focused on rural residential land spatial distribution and area change, ignoring the spatial difference of newly added residential land, and the research on ecologically fragile areas was also insufficient. Theoretically, this research adds to the literature regarding the formation mechanism of rural residential land expansion in ecologically fragile areas. We also found that "small-town construction" had a very prominent impact on rural residential land expansion, which was seldom examined in previous studies. Practically, a deep, quantifiable understanding of rural residential land expansion characteristics offers important insights for the strategic planning of a specific region to stimulate sustainable regional development. Additionally, there are still some limitations for this research: we analyzed only a limited number of policy driving factors due to data acquisition limitation at the village-scale. Potential policy determinants should be incorporated in the following studies to advance the knowledge of the driving forces of rural residential land expansion. Furthermore, the land-use data resolution in this research is 30 m, which can meet mesoscale and macro-scale research. The empirical study of residential land changes at a micro-scale in small watersheds is merited.