Neighboring Effects on Ecological Functions: A New Approach and Application in Urbanizing China

Abstract

Rapid urbanization has widely induced fragmented landscapes and further negatively affected ecological functions. The edge effect is an approach commonly used to investigate these negative impacts. However, edge effect research tends to focus on the impacts that a certain landscape receives from its adjacent lands rather than to assess all the influences of the landscape edges in a region, even though the latter is critical for regional ecological planning. To fill in this gap, the concept of neighboring effect is raised and analyzed in this paper with a case study of Xintai City in Northern China. Results show that the neighboring effects are generally negative for ecological functions, especially in regions that experience rapid urbanization or heavy human activities. The U-shaped relationship between the neighboring effect of a patch and its distance to the nearest township center indicates that the border region of urban and built-up areas suffers the most negative influences due to the intense interactions between different land uses. The heterogeneous effects of influencing factors in urban and rural areas were revealed by the regression results. Socioeconomic development has more important influence on neighboring effects on ecological functions in rural areas than in urban areas, and local cadres’ support of environmental protection matters only in rural areas for a less ecological functional loss. This study quantitatively examined the negative ecological effects of landscape fragmentation during rapid urbanization and calls for more attention to ecological planning at the local scale.

1. Introduction

The fragmentation of landscapes has been widely acknowledged as detrimental to ecological conditions. It has drawn increasing attention in recent years, especially since the release of the UN’s Sustainable Development Goals in 2015 [1]. Numerous studies have demonstrated that human activities are the major causes of landscape fragmentation [2,3,4,5,6,7], which places the ecological planning aimed at protecting the environment and coordinating relationships between human and nature into an important position. However, prior studies have not adequately responded to what impacts landscape fragmentation might bring to ecological functions in theory or what local governments can do to minimize the negative influences in practice.

The edge effect is a theoretical framework commonly used to investigate the ecological consequences of landscape fragmentation. Landscape fragmentation is widely acknowledged to include the decrease in the area of the original landscape patch and the change in the spatial configurations of existing landscapes [8,9,10]. This perspective de facto obscures the two distinct ecological processes, which should be held separately for the sake of clarity [11]. Previous studies about landscape fragmentation mostly focus on the consequences of the changes in spatial configurations rather than the shrinkage of landscapes because the negative effects on ecological functions of the latter are obvious. Fragmentation in the aspect of spatial configuration can cause increasing edges between different landscapes. For example, in 2011, around 74% of forest area was situated within 100 m of the nearest edge in England [12]. This increase in edges might result in various ecological consequences, such as changes in microclimatic conditions, carbon storage capacity, and biodiversity changes [13,14,15,16]. These effects occurring on the edges of landscapes are termed “edge effects” [12] and have been long noticed and investigated by ecologists.

However, previous studies about edge effects mostly adopt the one-sided approach, which studies changes in ecological functions from the edge to the interior area of a certain landscape. The tremendous literature has both theoretically and empirically illustrated the abiotic and biotic impacts on certain landscapes, especially forests and wetlands, caused by their edges, including microclimatic changes, ecological vulnerability [16,17], plant dispersal and community structure, and species diversity and interaction [18,19]. However, the edges have two sides. This indicates that the edge effects should be seen as an interactive outcome by the landscapes of both sides [20], and it also means that the influences of edges on ecological processes are mutual, that is, surrounding lands from reserves, such as forests and wetlands, can also be affected by reserves. Although several recent studies have noticed the defects of considering edge effects only from one side and have provided some empirical evidence investigating the edge effects on both sides [21,22,23,24], these studies usually focus on certain types of landscapes and lack a holistic regional view, thereby inadequately supporting the local practice of ecological planning.

Regional planning needs an entire assessment of the region, but most of the prior studies only provided a mosaic understanding of interactions among different land uses. For example, related studies usually focus on the ecological impacts on forests and wetlands from surrounding land uses, whereas few studies investigate other types of land uses, such as grasslands and farmlands [22]. More attention is also paid to the edge effects between anthropogenic and natural land uses than those between two natural ones. Besides, the scales of prior studies are relatively diverse. Research at different spatial and temporal scales can reach distinct conclusions about the distance of the edge effect. Deleterious effects on ecological process can occur within meters to kilometers from the edges according to different studies. Moreover, scholars have argued that the affecting distance must vary with not only the research scales but also patch types and sizes [25], yet empirical studies have rarely examined the latter. These shortcomings of understanding can jeopardize the scientific assessment of the planning of regional ecological problems caused by landscape fragmentation during rapid urbanization.

Apart from the scantiness in the theoretical framework of the effects of increasing edges, data collection and processing are challenges for regional ecological planning. The existing methods of assessing the effects of landscape fragmentation are relatively data demanding, including the fundamental land use or land cover data, and other biotic and abiotic datasets, such as the physical and chemical attributions of soil and the conditions of key species. However, entirely collecting such miscellaneous and various types of data is difficult in regional ecological planning.

To address these gaps, the major objectives of this study were twofold. The first objective was to establish a novel approach based on easily accessible land use data for assessing the ecological consequences of landscape fragmentation at a relatively small scale. The second objective was to evaluate the effectiveness of this approach by applying it in a case study of Xintai City, Shandong Province, eastern coastal China, showing the assessment results and exploring possible influencing factors. Since the entry of the new century, China has experienced a very fast process of urbanization from 36.2% in 2000 to 64.7% in 2021, which has caused drastic land use changes and thereby degrading ecosystem services [26,27,28]. This phenomenon requires more ecological plannings and practices to avoid and alleviate the negative effects on the environment of urbanization [29,30]. However, under the institutional context of China, it is necessary to evaluate the ecological effects of the fragmented land use caused by fragmented authoritarianism for optimizing the regional ecological condition. Different kinds of ecological land use are governed by different divisions of government, and they used to have their own plannings, which might conflict with each other [31,32]. Thus, it is much harder to coordinate the relationships of the different land uses at a micro level. The recent work of Territorial Development Planning in China tried to incorporate various plannings into one for handling with the problems of fragmented authoritarianism in land use management. A holistic assessment about the ecological effects of fragmented land uses at the county level can provide some support for the planning.

The rest of this article is organized as follows. Section 2 shows the detailed methodology of assessing the neighboring effects on ecological functions among land uses, including the definition of neighboring effects, classification of land uses, and steps of calculating the assessment results. Section 3 presents the study area (Xintai City), data, and models used in this study. Section 4 discusses the assessment results of the neighboring effects in Xintai City and gives the explanations. Section 5 provides the main conclusions and some discussions.

2. Neighboring Effect and Measurement

2.1. Neighboring Effect

Landscape fragmentation is one of the major human-related causes of various ecological issues, such as increasing regional ecological vulnerability, loss of biodiversity, and shrinkage of species habitats [6,7,33,34,35,36]. It usually appears in two aspects, namely, the decrease in patches’ size and the complication of interfaces between different land uses. Regardless of which one of the two aspects appears, the effects of landscape fragmentation are generated by the increasing interactions between two adjacent land uses or ecosystems. These interactions act as various forms. For example, the boundaries of land use usually indicate abrupt changes in microclimatic conditions, including light, temperature, and moisture, which has been demonstrated to be deterrent to the survival of trees of forest land [37]. Some ecological processes, such as seed dispersal, carbon storage, and pollination, can also be influenced by the land use edges [38]. Although most of the prior research focused on the negative impacts of boundaries between land uses, the interactions through interfaces can be positive to ecological functions [39,40]. Scholars have found that the smaller woodlands in agricultural landscapes can provide more ecosystem services per area than those larger woodlands, which is thought to be caused by positive edge effects [41].

Previous studies have commonly viewed the edge effect as a one-way influencing process, without consideration of the possible reverse influence, which is in contradiction with the ubiquitous complex interactions between various ecosystems. As mentioned above, forests and wetlands are commonly regarded as the receivers of edge effects. Ecological and environmental processes in the interior side of these land uses receive more attention than those in the exterior side. This condition is in accordance with the common definition of edge effects, that is, the changes in a certain land use in biotic and abiotic features caused by its edges or boundaries. This definition originated from the conservation-related research that places extra emphasis on the species abundance and biodiversity protection, especially in important natural ecosystems, such as forests and wetlands.

However, not only natural reserves but also all other land uses in the region should be considered by regional planners and local decision makers. The ecological consequence that the edge induces is bilateral rather than monolateral. Prior studies using the concept of the one-sided edge effects usually fail to analyze the influences of edges on ecological functions at the regional scale, since they tend to focus on only a few landscapes and to overlook the interactions between different land uses. Therefore, in order to provide a holistic understanding of the ecological consequence of landscape fragmentation caused by urbanization in a region for planners and decision-makers, we used the concept of “neighboring effects” of land uses that stresses the “effect on the other side” of the edge to broaden the notion of “edge effects”. The neighboring effects can be defined as changes, either negative or positive ones, in ecological and environmental processes and features on the two sides of land use interfaces. These effects vary with the shape and length of the edges and with land uses on the two sides of the edges.

The shape and length of the edges are key indicators that define the spatial configuration of landscape patches and are thus crucial for examining the two-way edge effects. In landscape ecology, given a fixed area of patch, the round shape is usually thought to receive the minimum influences from surrounding lands and provide the best habitat conditions because it has the lowest edge to area ratio [42,43]. Vice versa, when we consider the exterior impacts on adjacent patches or the matrix of a certain patch, it also generates the minimum neighboring effects. Apart from the spatial configuration of a certain patch, the overall configuration of a landscape, that is the number of patches given a fixed total area, can intensively influence the edge length. More patches indicate longer edges and more intense neighboring effects. Figure 1 provides the illustrations of neighboring effects and the relationship with fragmentation. In Figure 1a, the interface of two different land uses is simply a straight line, and the neighboring effect only occurs within a certain distance on the two sides of the edge. Comparatively, in Figure 1b, two small round patches are surrounded by the matrix of the other land uses. Under this condition, the area that can be affected by the neighboring effect increases by two extra circular segments, even though the total area of each land use equals that in Figure 1a.

2.2. Classification of Land Uses

Apart from the spatial configuration of the landscape, land use types on the two sides of the edges can also influence the neighboring effects. For example, when forest land is adjacent to watery area and when forest land is adjacent to agricultural land, the neighboring effects of the two conditions may be completely different, indicating that we cannot regard all neighboring effects as homogenous impacts on ecological and environmental processes. Therefore, identifying and classifying land uses/ecosystems is a critical question for measuring the neighboring effect. Two crucial aspects must be focused on. The first one is intraclass homogeneity and interclass heterogeneity. A good classification should maximize the similarity in ecological processes and environmental conditions within an identified class and ensure the remarkable differences among classes. For patches in the same class, the boundaries between them should not generate remarkable changes in neither biotic nor abiotic characteristics. The second aspect is the feasibility in application and the simplicity of calculation. It requires that the number of identified classes should be relatively small. Although fine land use classification can help to identify interactions between different ecosystems scientifically, the calculation process will be extremely complicated to affect its practicability. Moreover, since the analysis of the neighboring effect is much more important at relatively smaller scales, such as counties and townships, than at larger scales and the land use types at such scales are not very diverse, it is necessary and reasonable to simplify the land use classifications.

Therefore, this study reclassified all land use types into two categories and nine classes based on the land use classification of the Third National Land Survey of China, as shown in

Table 1. Classification of land uses.

Categories	Classes	Corresponding Land Uses in the Third National Land Survey
Ecologically functional land	Forest land	Forest land
	Grassland	Grassland
	Garden land	Garden land
	Farming land	Farming land
	Watery surface	Watery surface and water facilities land; wetland
	Bare land	Other land (except facility agricultural land)
Ecologically disturbing land	Rural road	Rural road (in transportation land)
	Other roads	Transportation land (except rural road)
	Other construction land	Other land uses

. In accordance with the Third National Land Survey data, the land uses were classified into as many as 13 first-level classes and 56 second-level classes, which is obviously redundant to be used directly to analyze the neighboring effects at county and township levels. Those land uses that are adjacent with each other do not change their ecological processes substantially and should be combined as one when measuring the neighboring effects on ecological functions. Therefore, we first divided all the standard land uses dichotomously according to the dominant functions of different land uses, that is, the ecologically functional land and ecologically disturbing land [44,45]. The former refers to land uses that can provide ecological functions, and the latter is mainly anthropogenic land uses that exert negative impacts on the environment. Thus, the ecologically functional land uses are regarded as both the receiver and sender of neighboring effects, whereas the ecologically disturbing land uses are only considered as the sender. For the ecologically functional land category, based on the abovementioned principles and referring to the land use classification in related studies [46,47,48], we identified six classes (namely, forest land, grassland, garden land, farming land, watery surface, and bare land) in terms of the ecological processes and environmental conditions. As for the ecologically disturbing land category, roads and other construction lands were identified, and rural road was seen as an independent class from the roads since the rural roads have quite different impacts with other transportation lands on the ecological functions of surrounding land uses [49]. This classification of land uses can provide a solid data foundation for calculating the neighboring effect.

2.3. Influencing Distance and Intensity Matrix

As mentioned above, neighboring effects can vary with land uses on the two sides of the edges. This difference can be identified from two aspects, namely the influencing distance and influencing intensity, with reference to the research on edge effects [50]. The influencing distance refers to the maximum distance from the interface to the interior of the effect-receiving landscape where the neighboring effect can be observed. The intensity of the neighboring effect decreases with the increase in distance to the interface, and the decay function of this decrease will be significantly different for different adjacent land uses. Prior studies also demonstrated that the size of patches determines the influencing intensity and distance of neighboring effects [51,52,53,54,55]. If we include all the factors in the neighboring effect assessment, then the calculation process will be extremely arduous, which can obviously impede its application in regional planning work. Therefore, in this study, we assessed the neighboring effects based on the following three prerequisites. First, the influencing distance of neighboring effect on ecological functions does not vary with the size of the patches. Second, within the influencing distance, the influencing intensity of the neighboring effect is homogenous and does not decay with distance. Third, the neighboring effect always exists within the influencing distance regardless of whether the two land uses have a common edge. The three prerequisites can largely improve the feasibility of the approach to assess the neighboring effects on ecological functions raised in this study, and the errors they caused are acceptable in methodological and practical terms. At the regional scale, planners and decision makers aim to find a general pattern of the neighboring effect to support planning and decision making and to stress the importance of the practicability of approaches rather than obtaining an extremely precise result of the neighboring effect, as ecologists and conservationists do. Therefore, simplification is fairly important and necessary. The first two prerequisites are based on the reality that, at the scale of villages, towns, and counties, the size of ecological landscapes is relatively homogenous. For example, in Xintai City, the study area of this research, up to 75% of forest land patches have an area less than 1 km², more than 95% are less than 5 km², and approximately half of the farming land patches range from 0.1 km² to 1 km². This condition indicates that the influencing distance, intensity, and the decay function of the neighboring effects of these patches for the same land use are relatively similar. Therefore, the use of a fixed distance and a fixed intensity will not mislead the results. For the third prerequisite, it is reasonable to assume that neighboring effects also exist when two land uses are not adjacent directly, as long as parts of them are located within the influencing distance. This reasonability is supported by previous studies, concluding that land uses can affect and be affected by other land uses within a certain distance even though they do not have edges with each other [56].

The influencing distance and intensity matrices, as shown in

Table 2. Influencing distances of the neighboring effect (Unit: m).

	Effect Receiver	Forest Land	Grassland	Garden Land	Farming Land	Watery Surface	Bare Land
Effect Sender
Forest land		0	63	53	77	117	67
Grassland		23	0	33	47	63	37
Garden land		47	60	0	70	87	57
Farming land		47	50	50	0	100	43
Watery surface		67	77	83	117	0	67
Bare land		43	67	53	83	57	0
Rural road		53	67	60	200	63	33
Other roads		53	67	60	200	63	33
Other construction land		100	120	220	213	227	77

and

Table 3. Influencing intensity of the neighboring effect.

	Effect Receiver	Forest Land	Grassland	Garden Land	Farming Land	Watery Surface	Bare Land
Effect Sender
Forest land		1.00	1.27	1.17	1.27	1.43	1.60
Grassland		0.87	1.00	0.93	0.93	1.27	1.37
Garden land		0.83	1.00	1.00	1.27	1.23	1.20
Farming land		0.77	1.00	1.00	1.00	0.93	1.10
Watery surface		1.43	1.43	1.37	1.33	1.00	1.53
Bare land		0.70	0.83	0.73	0.77	0.73	1.00
Rural road		0.96	0.98	0.97	0.95	0.98	0.99
Other roads		0.87	0.93	0.90	0.83	0.93	0.97
Other construction land		0.63	0.70	0.70	0.77	0.63	0.90

, were generated on the basis of the abovementioned prerequisites and land use classification, where the values were calculated by the expert scoring method. Although the regional planning is a process that increasingly emphasizes the participation of multiple stakeholders, experts with related knowledge should still play important roles in this process by providing their professional opinions to facilitate the decision making. Therefore, five experts from different fields, including ecological planning and environmental management, urbanization and urban-rural development, and ecosystem service research, were selected and invited to participate in the scoring process to obtain interdisciplinary and planning-oriented professional opinions on the neighboring effects assessment. These experts included professors at top universities and urban planning institutes from Beijing, Shanghai, and Shandong. During the scoring process, experts were also allowed to discuss with their colleagues; therefore, the result of their final scores could represent a consensus in various research and practical fields to some extent. The five experts were sent two blank forms (the same with Table 2 and Table 3 but without values), the basic information of Xintai City, the definition of neighboring effects, the three aforementioned prerequisites, and explanations of the scoring process by email. It should be noted that, in the scoring forms sent to experts, there was no descriptions such as “ecologically functional land” or “ecologically disturbing land” we identified above, so as to avoid possible influences on the experts’ scoring process. In the evaluation of the influencing distance matrix, five experts were asked to give the maximum distance they thought the neighboring effect on ecological functions can penetrate from the sender to receiver between every two types of land uses. For the influencing intensity matrix, the experts gave the average effect intensity within the influencing distance they rated for every two different land use types. The experts were asked to assign the influencing intensity values from −1 to 1, where −1 denotes that the sender makes the receiver fully lose its ecological functions and 1 indicates an extreme increase in the ecological functions of the receiver from the neighboring effect. The mean value of every blank in the two matrices was calculated, with the exclusion of the maximum and the minimum ones, after collecting the scoring results from the five experts. Results with the mean values were then returned to experts, and we asked them to adjust the values in matrices accordingly. This procedure was repeated for several times until the five experts met an agreement. To simplify the calculation and make the result more interpretable, we transformed the values ranging from −1 to 1 to numbers from 0 to 2. After the rescaling, 1 denotes that the land use does not receive any neighboring effect from other land uses, and numbers larger than 1 and smaller than 1 indicate the increase and decrease in ecological functions caused by neighboring effects. By doing so, for a certain patch, all the intensity coefficients of received neighboring effects can be multiplied to generate a final intensity from all other adjacent land uses. It should be noted that, as the two matrices were obtained according to the scoring of selected experts based on the reality of Xintai City, they may not be directly applied to other places with geographical conditions that are significantly different from those in Xintai. Rather, it is the strategy of inviting multidisciplinary and experienced experts to determine the influencing distance and intensity of neighboring effects that can be repeated elsewhere to obtain an outcome fitting with local realities and major ecological issues and to better contribute to the planning compilation and implementation.

In accordance with the influencing distance and intensity shown in Table 2 and Table 3, the neighboring effects for every site can be obtained directly based on the land use in situ and its surrounding ones. As mentioned above, the total neighboring effects for land uses affected by more than one adjacent land can be calculated by multiplying the intensity coefficients. However, the patch level was somewhat too micro to analyze the spatial pattern of neighboring effects for finding general distributive rules. Therefore, we calculated the average neighboring effect at the village level. The calculation formula is:

$$I_{vil}=\frac{{\displaystyle \sum }\left(Are{a}_j\times I_j\right)}{{\displaystyle \sum }Are{a}_j}$$

(1)

where I_vil is the average neighboring effect in the village; I_j is the intensity of the neighboring effect for region j that every site in this region has the same intensity of neighboring effect; and Area_j is the area of region j.

3. Data and Methods

3.1. Study Area

The study area of this article was Xintai City, a county-level city under the administration of prefecture-level Tai’an City. It is located in central Shandong Province, east coastal region of China. With a total area of 1946 km² and 1.34 million people in 2020, it has a density of 688 inhabitants/km². It is now composed of 21 townships, namely Qingyun (QY), Xinwen (XW), Xinfu (XF), Dongdu (DD), Xiaoxie (XX), Zhaizhen (ZZ), Quangou (QG), Yangliu (YL), Guodu (GD), Xizhangzhuang (XZ), Tianbao (TB), Loude (LD), Yucun (YC), Gongli (GO), Guli (GL), Shilai (SL), Fangcheng (FC), Liudu (LI), Wennan (WN), Longting (LT), and Yuejiazhuang (YJ), with its administrative center located in Qingyun, as shown in Figure 2.

Xintai City was selected as the study case in this research for three reasons. First, minimizing and eliminating the negative ecological impacts on Xintai City are urgent, especially after the proposition of the development of ecological civilization in China. This condition is tightly correlated with the long history of coal exploitation in Xintai. Although, currently, plants related to coals in Xintai City have been widely shut down owing to the strict requirements for environmental protection from higher-level governments and the exhaustion of coal resources, the considerable abandoned mineshafts in rural areas can still have continuous impacts on ecological functions. Second, the topography in Xintai City is relatively various. It includes mountains, hills, and plains, and its altitude ranges from about 75 m to 989 m above the sea level. The longest river of Xintai City, the Chaiwen River, flows through the city from east to west, and its length in Xintai City reaches 93.9 km. The diversified topography of Xintai City can generate various landscape edges during the rapid urbanization processes, which can help strengthen the representativeness of the case study. Third, Xintai City is still in the mid-term of the urbanization process, with an urbanization rate of 57.9% in 2020. This indicates that the conflicts between humans and the environment are still intense and reflect on the pattern of land uses. Therefore, Xintai City was a suitable case for this study, and has a good representativeness of counties in the rapidly urbanizing stage.

3.2. Data

For the assessment of the neighboring effect, the land use data were derived from the Third National Land Survey (disanci quanguo guotu diaocha) in 2019. Mainly relying on the land use data can improve the practicality of the proposed assessment method for regional ecological planners by its simple calculation process and by ensuring the data availability. In accordance with the land use classification shown in Table 1, the land use map of Xintai City with this classification is shown as Figure 3. Different land uses exhibit distinct distributive patterns in Xintai City. The farming lands in Xintai can be considered the matrix, and other land uses can be regarded as patches because the farming lands are distributed widely in every township and have no significant spatial concentration. Forest lands are mainly located in the south, central west, and north parts of Xintai, where mostly their areas are mountains and hills. Garden lands are extremely concentrated in two towns situated in the northeast and northwest parts of Xintai, namely Longting and Tianbao. Watery surfaces in Xintai primarily contain three parts, namely Qingyun Lake on the boundaries of Qingyun, Wennan, and Longting, Guangming Reservoir across Liudu and Xiaoxie, and Chaiwen River, the main river in Xintai. For the ecologically disturbing lands, most of them surround the administrative centers of townships, with the largest patch being the urban areas in Qingyun, Xinfu, and Xinwen.

Several other data sources were used in the econometric models. The average slope was generated on the basis of the Shuttle Radar Topographic Mission digital elevation model data. The climate-related data, including the precipitation and temperature, were obtained by the Kriging spatial interpolation of meteorological station data from the China Meteorological Administration. The socioeconomic aspects, including the distances to the county center, to town centers, and to national highways, were calculated on the basis of the National Fundamental Geographic Information Database. Attributes of villages/communities (simply using “villages” hereafter), including population, industrial development, and other socioeconomic variables, were mainly obtained from the survey data acquired by Peking University in Xintai City in 2021. This survey includes mostly all the villages of Xintai City, except those in the urban areas and in Tianbao Town (which was transferred to Mount Culai and Wenhe River Scenic Area and was not under administrative control of Xintai City). Therefore, we excluded the villages out of the survey dataset in the later econometric model parts.

3.3. Specification of the Econometric Model

Multiple linear regression was used in this study to detect the influencing factors of the neighboring effects by village. In the econometric model, three groups of independent variables, namely the terrain factors, socioeconomic factors, and urbanization-related factors, were included. Thus, the formula of the model is as follows:

$$N{E}_i={\beta }_0+{\mathbf{\beta }}_{\mathbf{1}}\mathit{T}{\mathit{F}}_{\mathit{i}}+{\mathit{\beta }}_{\mathbf{2}}\mathit{S}\mathit{E}{\mathit{F}}_{\mathit{i}}+{\mathbf{\beta }}_{\mathbf{3}}\mathit{U}{\mathit{F}}_{\mathit{i}}+{\epsilon }_i$$

(2)

where $N{E}_i$ is the average neighboring effect of village i, whose calculation method will be elucidated in the next section; $\mathit{T}{\mathit{F}}_{\mathit{i}}$, $\mathit{S}\mathit{E}{\mathit{F}}_{\mathit{i}}$, and $\mathit{U}{\mathit{F}}_{\mathit{i}}$ are the vectors of terrain factors, socioeconomic factors, and urbanization factors of village i, respectively; β is the coefficient of corresponding independent variables; and ${\epsilon }_i$ represents the random error. The regression is estimated by using the ordinary least squares method.

Terrain factors contain the slope, altitude, and area of each village. The differences in terrain factors among villages can induce various land utility and configuration patterns [57], thereby generating heterogeneous neighboring effects on the ecological functions of villages.

Socioeconomic factors include the population density, the number of nonagricultural businesses, mining history, average income of residents, the maximum household income of residents, collective economy condition, and the views of village cadres on environmental protection. Population density is highly connected with the intensity of human activities in a certain region because the higher the population density, the more intensive the human needs and ecological protection conflict. The number of nonagricultural businesses and the mining history of the villages were used to reflect other disturbances to the environment. The other factors stated above are related with the adaptation capacity of the villages when facing ecological issues. With reference to prior studies, we hypothesized that the wealthier the residents and the villages, and the more emphasis on environmental protection the village cadres put, the less the impacts of human activities would be [58,59]. The distance to the nearest township center of each village was considered to reflect the location.

For the urbanization factor, the urban and rural classification codes given by the website of the National Bureau of Statistics of China were used to identify whether a village/community was urbanized or not. If the three-digit classification code began with 1, the village/community was regarded as an urbanized area, otherwise it was considered as a rural area. The specific clarification and descriptive statistics of the independent variables in 815 observations are shown in

Table 4. Descriptive statistics of independent variables.

Factor	Variable	Explanation of Variable	Unit	Ave.	Std. dev.	Min.	Max.
Terrain factors	Slope	Average slope of the village	°	6.17	2.77	3.05	20.36
	Altitude	Average altitude of the village	100 m	2.05	0.58	1.08	4.74
	Area	Total area of the village	km2	2.03	1.58	0.08	23.64
Socioeconomic factors	Popden	Population density of the village	1000/km2	0.77	0.77	0.05	14.98
	Nonag	The number of nonagricultural businesses in the village	100	0.13	0.29	0	6.34
	Mine	Whether there have been any mining activities in the village (1 = yes)	/	0.32	0.47	0	1
	Lginc	Average annual income per capita in the village (logarithm)	CNY	9.21	0.60	5.03	11.51
	Lginchmax	The maximum household annual income in the village (logarithm)	CNY	11.72	1.28	7.38	17.73
	Coleco	Whether there is any collective economy in the village (1 = yes)	/	0.25	0.43	0	1
	Overenv	Whether the village cadre outweighs environment than economy (1 = yes)	/	0.52	0.50	0	1
	Dtown	Distance to the nearest township government from the village	km	4.33	2.33	0	12.36
Urbanization factor	Urban	Whether the village is urbanized (1 = yes)	/	0.30	0.46	0	1

.

4. Results

4.1. Calculation Results of the Neighboring Effects

The distribution of the patch-level neighboring effect is shown in Figure 4. Five types of patches were identified on the basis of the calculated neighboring effect of every patch. Total neighboring effects smaller than 0.75 were regarded as major negative ones, and those ranging from 0.75 to 0.95 were classified as minor negative ones. Correspondingly, neighboring effects from 1.05 to 1.25 and larger than 1.25 were identified as minor positive and major positive ones, respectively. Other patches with neighboring effects ranging from 0.95 to 1.05 were where almost no neighboring effect was observed.

Patches with negative and subtle neighboring effects covered the most area of Xintai City, indicating that landscape fragmentation mostly brings negative effects to ecological functions. In accordance with the five classes identified above, we further calculated the total area of patches in each group, and the result is exhibited in Figure 5. As indicated, the total share in the area of patches with negative and subtle neighboring effect reached 81.5%, with approximately 50% being clearly negatively affected. This result is in line with the exorbitant focus on the negative influences of landscape fragmentation and the edge effects in previous studies. However, it is also worth noticing that nearly one-fifth of the total area was affected by a positive neighboring effect. Such positive effects might offset the negative ones caused by landscape fragmentation to some extent. This phenomenon possibly offers a new scope for ecological and environmental planning.

Patches with a positive neighboring effect are mainly located in places far from the centers of townships, whereas patches with a negative effect show the opposite pattern. Specifically, a negative neighboring effect is insignificant in the extremely close places to the centers but around the edges of the centers’ construction lands. This condition is because few ecologically functional lands exist in township and city centers, indicating that there are no intense interactions between ecologically functional lands and ecologically disturbing lands. However, things are different around the centers’ construction lands. The two sides of the edge of construction lands are ecologically functional and disturbing lands, causing frequent interplays between them and a severe negative neighboring effect on the outer side of construction lands. Besides, the expansion of construction lands, usually at the cost of the occupation of agricultural and other ecological lands in rural areas, tends to aggravate the fragmentation of landscapes [60]. The increase in fragmentation can generate longer edges, which will exacerbate the negative neighboring effects further. When the distance to centers becomes larger, the neighboring effects between ecologically functional lands become predominant with the decrease in ecologically disturbing lands. As shown in Table 4, the neighboring effects between ecologically functional lands are usually positive.

The total neighboring effects by villages show a similar pattern to those by patches, but the distributive rules are clearer. When the total effect of a village was smaller than 0.85, the village was classified with a major negative neighboring effect. The total effect ranging from 0.85 to 1 and larger than 1 corresponds to a minor negative effect and a positive effect. As illustrated in Figure 6, most villages in Xintai City are undergoing minor negative total neighboring effects. Villages experiencing major negative neighboring effects are mainly located near the township and city centers. They are mainly distributed in three regions: QY-DD-WN border region, GL-XX-XZ-GD border region, and LD-YC border region. The three border regions are all located around the concentrated construction lands, thereby facing intense interactions between ecologically disturbing and functional lands. Comparatively, villages with positive total neighboring effects are far from the centers. Most of them are situated in the north and south parts of Xintai City where lands are widely covered by forests.

4.2. The Relationship between Neighboring Effects and Urbanization

According to the spatial pattern of the estimated neighboring effects in Xintai, there seem to be some relationships between neighboring effects and urbanization. To more quantitatively reveal their connections in Xintai City, we first investigated the relationship between the neighboring effects and the distance to the township centers. Multiple nonoverlapping ring buffer strips were created at 250 m, centered on the township governments in ArcGIS software, and we calculated the average neighboring effects for every strip by using the same method with that of villages. The result is shown in Figure 7.

A clear U-shaped relationship was found between the neighboring effects and the distance to the township centers in Xintai City, as shown in Figure 7. With the increase in distance to the township centers, the negative neighboring effects become increasingly intense (which means the decline of the average effect value), reaching the peak at around 1250 m away from the township centers. After that, the negative neighboring effects gradually weaken (i.e., the rise of average effect value) with the distance to the township centers. These results are in accordance with the previous analysis based on the spatial pattern that negative neighboring effects mainly occur around the edges of cities and towns. In the very inner rings, construction lands cover the majority of the area, indicating that few interactions occur between ecologically functional lands and disturbing lands. Approaching the boundaries of the built-up areas, the interplays of the abovementioned two types of lands become more intense. However, the intensity of influence again weakens in the more outer rings, owing to the reduction of human activities. The results also show that the average neighboring effects of ring buffer strips are always less than 1, at least within the range of 5 km from township centers. This finding demonstrates that, although the effects are weak in the very inner and outer regions, the urbanization process accompanying various profound human activities brings negative influences on the ecological functions surrounding the cities and towns. Hence, regional planners and local governments should put particular emphasis on environmental protection, especially on the margins of urbanization.

The relationship between the average effect by village and the coverage of construction land can again confirm the abovementioned conclusions (Figure 8). The share of construction lands in a certain region can be used to partially reflect the extent of urbanization because the expansion of construction lands is a representative indication of urbanization [61,62]. In Figure 8, a U-shaped correlation is clearly illustrated between the percentage of construction land and the average neighboring effects by village, which is similar to the pattern shown in Figure 7. As the percentage of construction land in a village rises, the average neighboring effect value shows a trend of decreasing and then increasing, which means that the negative effects caused by edges in villages strengthen first and then weaken with the increase of the urbanization rate. In addition, the figure also shows that the average neighboring effects in villages with low construction land share are relatively diverse. With the increase in construction land proportion, the distinction among different villages reduces gradually. This suggests that regions with high urbanization tend to have identical patterns of various land types, resulting in similar average neighboring effects. However, in villages with lower percentages of construction lands, how different lands distribute and interact with each other is decisive for neighboring effects. Some villages can benefit from the land use distributive patterns, whereas others suffer negative impacts. Therefore, a proper arrangement of different land uses is crucial for maximally maintaining the ecological functions and minimizing the negative impacts of human activities, especially for regions undergoing rapid urbanization processes. The influencing factors of neighboring effects can be further analyzed to investigate the causes of the differentiation among villages, thereby providing empirical evidence for local planning and decision making.

4.3. Influencing Factors of Neighboring Effects

Three regressions were conducted with the aforementioned models, namely for all villages, for urbanized areas, and for rural areas, respectively, to further analyze the influencing factors of neighboring effects by village. Although the prior section clearly showed a U-shaped curving relationship between the percentage of construction land and the average neighboring effects by village, the models do not include the factor of construction land. The main consideration is that the percentage of construction land is not an exogenous variable to the neighboring effects because it is a crucial factor in the calculation process of neighboring effects. Regarding the percentage of construction land as an independent variable for detecting the influencing factors of neighboring effects is inappropriate. The regression results are given in

Table 5. Regression results of econometric models.

Variables	All Villages		Urbanized Areas		Rural Areas
	Coefficients	Robust t-stat.	Coefficients	Robust t-stat.	Coefficients	Robust t-stat.
Slope	−0.15	−1.24	0.50	1.59	−0.25 *	−1.73
Altitude	6.38 ***	3.41	20.63 ***	3.61	5.90 ***	2.62
Squared Altitude	−1.03 ***	−2.61	−5.27 ***	−3.30	−0.86 *	−1.93
Area	1.01 ***	6.02	0.94 ***	4.46	1.01 ***	4.62
Popden	−0.37 *	−1.67	−0.39 *	−1.93	−0.51	−0.82
Nonag	−0.87 **	−2.23	0.02	0.03	−1.03 **	−2.46
Mine	−0.82 **	−2.17	−0.19	−0.34	−1.23 **	−2.51
Lginc	−0.33	−1.13	0.21	0.46	−0.51	−1.41
Lginchmax	−0.47 ***	−3.42	−0.65 ***	−2.85	−0.37 **	−2.13
Coleco	−0.37	−0.90	−0.76	−1.29	−0.09	−0.17
Overenv	0.54	1.57	0.12	0.22	0.75 *	1.76
Dtown	0.32	1.19	−0.19	−0.49	0.87 **	2.14
Squared Dtown	0.01	0.55	0.05	1.06	−0.03	−0.85
Urban	−1.16 ***	−2.84
Constant	89.78 ***	24.33	72.04 ***	10.11	89.51 ***	19.42
Observations	815		247		568
R-squared	0.261		0.159		0.224

Note: *, **, and *** indicate significance at the 0.1, 0.05, and 0.01 levels, respectively.

.

Terrain factors, especially the altitude and area of the village, have significant influences on the average neighboring effects. An inverted U-shaped relationship was found between the average altitude and neighboring effect. This relationship is tightly related with the topographical features of Xintai City. On the one hand, the altitude of the central urban area of Xintai City is around the middle in the region, determining that the majority of edges between ecologically functional and disturbing lands are with the middle altitude. On the other hand, lands with two ends of altitudes in Xintai City are majorly farmlands and forest lands, causing a less negative or even positive neighboring effect. The area of village is another crucial factor for neighboring effects. The larger the village, the more positive average effect the village has. This phenomenon is mainly because when the area of a village is large, there are more ecologically functional lands outside of the constructed area. The effect of average slope on neighboring effect was insignificant for all villages after controlling other variables.

For the socioeconomic aspect, different factors play distinct roles in the neighboring effect. Population density, nonagricultural businesses, and mining activities, that can be regarded as sources of ecological disturbances, were all negatively correlated with the average neighboring effect of the village. Other variables reflecting the adaptation capacity and locational condition had no significant impacts on the neighboring effect, except for the maximum household annual income of the village. We supposed that the maximum household annual income would have a positive relationship with the average neighboring effect because it reflects the potential economic support for regulating environmental issues. However, beyond our expectation, they were negatively correlated. Several possible explanations are provided for this result. A higher maximum household annual income in a village might indicate that the nonagricultural businesses of the village are larger, given the same quantities as other villages, implying more needs for construction lands. The household with the maximum income in the village may neither be deeply engaged in decisions and plans of the village nor provide actual supports on the ecological protection for the village, thereby the high income cannot be converted to the adaptation capacity of the village. The distance to the nearest town did not show the same influences as analyzed in the prior sections. This condition is due to the difference in analyzing scales. When the scale is at the patch level, villages are separated into several different rings as calculated in the prior part. When we analyzed at the village level, only one distance to the nearest township center was given to every village rather than several different values. However, the distance to the township center of each site in the same village can be relatively different. Therefore, the scaling up of analysis might obscure the findings from lower-level analysis. However, the regression results indicate that at least at the village scale, the pattern of land uses can be optimized within the whole administrative area to eliminate the negative neighboring effects mainly caused by human activities.

The urbanization factor was significantly related to the average neighboring effect. An urbanized village will have a smaller value of the neighboring effect, that is, more of a negative neighboring effect than those in rural areas, by controlling other factors. This result demonstrates that the urbanization process has a negative influence on the environment and ecology, as proven by previous studies. Two separate regressions for villages in urban and rural areas were conducted to find the possible differences in the influencing factors of urban and rural areas. The results are shown in Table 5.

Most factors have the same effects on both villages in urban areas and in rural areas, but several factors that function heterogeneously are still found. For terrain factors, the average slope has a negative influence on the neighboring effect in rural areas. The reason can possibly be that villages with the steepest slope are mostly all in rural areas. Different land uses in villages with extremely steep slopes are usually interlaced on the constraint of topography, inducing complicated interplays between them and a more negative neighboring effect. For the socioeconomic aspect, all factors, except for population density, the maximum household annual income, and the collective economy, had significant impacts on rural areas. The attitude towards the environmental protection of the village cadre showed a positive impact on the average neighboring effect in rural areas but had no effect in urban areas. In other words, the urbanization process weakens the effect of village cadres on avoiding the loss of ecological functions. The land use planning or construction activities are not decided or influenced by the cadres in urban areas, but might be affected by opinions of the collective authorities in rural areas. Therefore, only in rural areas, the environmental protection consciousness of local cadre can promote positive neighboring effects.

5. Discussion and Conclusions

This study proposed an approach based on land use data to quantify the neighboring effects on ecological functions caused by the interplays between fragmented land uses under rapid urbanization. The effectiveness of the approach was demonstrated in the case study of Xintai City in northern China. The influencing factors of neighboring effects were further identified empirically. One advantage of this method is that it considers the neighboring effects on ecological functions between all of the land uses in a village. Studies using a one-sided edge effect methodology usually only present receiving or sending effects in ecological functions of certain land uses rather than offer a holistic evaluation of a region. Therefore, to some extent, a holistic regional neighboring effects analysis could be seen as a synthesis of a series of studies by using the edge effects approach. Such an intact estimation of the neighboring effect can provide a scientific evaluation result and practicable optimization for regional ecological planning. Another advantage of this approach is that its estimation requires only land use data rather than excessive data. This feature not only avoids the miscellaneous work of data collection during the regional ecological planning process, thereby saving much time, but it also lowers the data threshold of research, making it possible to conduct case studies in various contexts, comparative studies over time and across regions, and large-scale estimation. However, we argue that the neighboring effect approach raised in this study is more suitable for a relative short time period analysis at a more micro scale since it aims to assess the regional ecological consequences of landscape fragmentation caused by human activities and further to support regional ecological plannings and actions. Analysis within a long period cannot provide more useful information for regional planning than a short-term one. Besides, the process of assessing neighboring effects is thoroughly transparent, where all the key inputs, including the influencing distance and intensity between different land uses, are clear and can be adjusted in accordance with local conditions and land use data availability, thereby ensuring its feasibility in different places.

The results of the case study of Xintai City demonstrated the negative effects of urbanization and human activities on ecological function. According to the analysis at patch level, patches with clear negative neighboring effects on their ecological functions accounted for approximately a half of the total area. Comparatively, patches with clear positive ones only composed a quarter of the total area. Further, the distributive pattern of the neighboring effect in Xintai City as well as the relationship between the distance to the nearest township center and the average neighboring effect can clearly draw the conclusion that neighboring effects are the most negative on the edges of urban or constructed areas. Therefore, the urbanization process has an overall negative influence on ecological functions. As implied by the econometric models, the influencing factors of neighboring effects have differences between urban and rural areas. Although the nonagricultural activities have more significant impacts in rural than in urban areas, the attitude of local cadre plays a more important role in reducing ecological functional loss in rural areas than in urban communities. Thus, conducting regional ecological planning and raising the environmental protection awareness of local authorities are necessary, especially in rural areas, to minimize the negative effects of urbanization on the natural environment.

However, the proposed approach can still be improved in assessing the neighboring effects. First, there are few empirical studies that show whether two land uses are ecologically identical or not to support the simplification process of land use classification before analyzing the neighboring effects. Although the simplification used in this study refers to the common protocol, it might still cause some errors without evident research. Second, this study focused on the proportional change in ecological functions caused by interactions between land uses rather than the change in absolute ecological functional value. Although the calculation is simple, it might cause the underestimation or overestimation of the loss or gain of ecological functions. For example, in accordance with the influencing intensity matrix in this study, the neighboring effect intensity from rural roads was 0.96 for forest land and 0.95 for farming land. It seems that the rural roads have caused more severe ecological functional loss to farming land than to forest land. However, the loss of forest land is larger than that of farming land when the absolute functional value of each land use type is considered because the gross ecological function of forest land is larger. Considering the total amount of ecological function for each land use can definitely improve the accuracy of the neighboring effect assessment, which still needs further investigation. Third, this study used the fixed influencing distance and intensity for every land use, regardless of the size of each patch, for the convenience of calculation. Prior studies revealed that different patches with different sizes have varying influencing distance and intensity to their adjacent lands. Further studies are still needed to determine the specific relationship between the patch size of each land use and its influencing distance and intensity for adjacent lands to obtain more accurate estimation results of the neighboring effect, especially when applying the neighboring effects approach to a higher territorial level area, which might amplify the errors caused by the fixed influencing distance and intensity prerequisites. Finally, the scoring process of influencing distance and intensity matrices can also be improved when applying this approach in regional planning. In this article, the process of expert selection was still kind of arbitrary, even though we considered their different disciplinary backgrounds. To address this problem, measures including using a systematic and structured way to select more experts and expanding the participation of multiple stakeholders could be taken to achieve a more inclusive and democratic planning process.

Nonetheless, this study provided a simple and valid approach for regional ecological planning to probe the neighboring effects caused by landscape fragmentation in rapid urbanization. The case study in Xintai City confirmed the feasibility of the proposed approach. The scope of neighboring effects and the calculation process can be potentially used in the future evaluation of ecological functions and vulnerability to remedy the neglect of relationships between adjacent landscapes in present studies. However, a multi-scalar analysis of the neighboring effects on ecological functions is still needed, since the ecological planning and management is a multi-scalar affair in essence. It is necessary to consider the diversified ecological effects from different levels of processes, from various agencies, and even from other regions out of the study area, into the holistic regional ecological analysis. Besides, more comparative and dynamic empirical studies on the relationship between urbanization and neighboring effects are called. Different cities may follow distinct urbanization paths, such as compact city or sprawl mode, and may situate in different stages of urbanization. It is a critical question for regional planning whether these different urbanization modes and stages can engender diversified ecological consequences from the scope of neighboring effects. Such comparative and dynamic research can help us understand the impacts of urbanization on the environment more deeply, thereby improving the formulation and implementation of regional ecological planning.