Multiple Probability Ecological Network and County-Scale Management

Abstract

Ecological networks are threatened by rapid urbanization; the implementation of ecological network management and maintenance strategies is essential in the county units of urban agglomerations. This study focused on the Changsha–Zhuzhou–Xiangtan urban agglomeration (CZXUA) and objectively identified and evaluated the regional ecological networks. Connectivity distance thresholds were determined using goodness-of-fit experiments. The Probability of becoming an Ecological Source (PES) was constructed to investigate the ecological networks at different probabilities, and the network weight in each county was analyzed based on the gravity model. The results show that: (1) The distance thresholds were 2500, 2700, and 2300 m in 2000, 2010, and 2020, respectively. Source degradation and corridor disconnection were most significant at a PES > 50%. (2) At a PES > 50%, the high gravitational value increased from 39,972 to 31,4642, owing to fragmentation. Fourteen counties received weights, and fluctuations were most significant at a PES > 90%. (3) The trends in the PES and gravitational value were not always synergistic, and the negative impact of source degradation on the gravitational value was significantly less than the positive impact of the reduced distance between sources. The gravity center movement under the high PES determined the ecological network status of neighboring counties. The findings can quantify ecological information flow and provide a basis for county ecological management and restoration.

1. Introduction

The disturbance of ecological space due to urbanization and human activities can be broadly divided into short-term explicit and long-term implicit disturbances. Explicit disturbances are generally characterized by the encroachment of small-scale economic activities on woodlands and other ecological lands, significant transfers of land-use types [1], and negative changes in ecosystem services (habitat quality and climate regulation, among others) within a certain area. In contrast, implicit disturbances, such as the disruption of potential ecological network structures, are often long-term and subtle, and may not be visually reflected in small-scale ecological patches [2,3]. However, the patches are also affected by changes in neighboring patches and potential material–energy information flows at the macroscale; therefore, the ecological functions they perform in a given area also change indirectly [4,5]. For example, in the process of urbanization, the decline in connectivity between ecological patches at a large scale and the fragmentation of landscape patterns have led to the degradation of potential ecological networks, characterized by the decay of source areas and the break-up of corridors. The potential flow of biological information within spaces is hindered, seriously threatening the sustainable development of the ecological environment in areas of highly urbanized large urban agglomerations [6]. The implicit disturbances caused by urbanization, especially the impact and stress on regional ecological networks, should be given sufficient attention and focus, and a series of quantitative research methods for ecological networks is necessary [7,8]. Ecological sources and corridors are the main components of a network structure and have been identified and evaluated using various methods [9]. The identification and selection of ecological source units are based on the ability of the area to provide various ecosystem services, patch morphology, and connectivity [10,11,12]. Ecological corridors are dependent on ecological sources and migration resistance, and the minimum cumulative resistance model and circuit theory are the most widely used [13,14,15] when it is necessary to extract the elements of a linear ecological corridor. The minimum cumulative resistance is the sum of the resistance in the optimal migration path between two ecological sources, and the cost path tool of spatial analysis in the ArcGIS 10.2 platform can be used to quantify this research element objectively and scientifically based on this principle. The principle of circuit theory is similar, characterizing the relationship between electrical flow and electrical resistance to material–energy bioinformation flow and bio-diffusion resistance. Therefore, these two principles have become the most mainstream regional ecological corridor identification methods at this stage. Ecological network resilience, risk, and spatial spillover effects have also recently emerged [16,17]. Most of these evaluations are based on the characteristics of the network topology and apply elements, such as gravity models, and indices, such as network connectivity, connectivity rates, and the areas of pinch and barrier points. In terms of research scale, ecological networks are generally studied at a large scale at this stage, focusing on network identification and spatial evolution patterns at the scale of cities, urban agglomerations, or provinces [18,19]. Some studies not only focus on the spatial distribution of ecological networks but also on abstract network structures based on nodes and connecting paths [20,21,22].

However, as human geographical units, the evolutionary characteristics of ecological networks tend to vary considerably within different counties. In detail, at the urban agglomeration or city scale, the ecological network structure may be fractured and degraded over a long period of urbanization, resulting in the overall ecological network quality of a city being degraded. This macro-ecological network assessment may appear ambiguous when downscaled to the county. For a unit county, when the overall ecological network quality of the city is declining, the ecological network quality of the county is not necessarily declining, and the ecological network structure of the county is not necessarily disrupted. The degradation of ecological sources and destruction of corridors leads to changes in the network structure and shifts in the center of gravity, where one county loses the center of gravity of ecological sources, while a neighboring county may gain a new center of gravity [17]. When the investigation is downscaled to the county level, changes in the structure and quality of ecological networks per county during urbanization are not always negative. Therefore, the study of ecological network weights within counties is necessary, and it will be a fine-tuned and complementary work to the assessment of macroecological networks. Additionally, as ecological networks exist implicitly in urban spaces [23,24], the roughness of the identification method greatly affects the results of the network distribution, especially for the delta probability of connectivity (dPC) in the landscape connectivity indices. The dPC indicator calculation is based on patch areas and the probability of species dispersal between these patches in a unit region, which usually can be used to characterize the landscape connectivity of habitats. Also, the distance threshold is a very critical factor [25] in landscape connectivity evaluations. A relatively more accurate distance threshold can characterize the actual degree of landscape connectivity in the region, thus obtaining a more accurate distribution of ecological sources. Most existing studies have used the study area as the only reference factor to directly set the distance threshold. Although the identification of ecological sources may incorporate a variety of elements, it remains partially subjective. The objectification of identification methods and the optimization and improvement of related models remain important tasks worthy of intensive work at this stage [26].

In summary, this study used the Changsha–Zhuzhou–Xiangtan urban agglomeration (CZXUA), a highly urbanized region, as an example, and analyzed the ecological network evolution at multiple scales and county weights. Ecologically related indicators were cumulated as the Probability of becoming an Ecological Source (PES), and the gravity model was then used to analyze the evolution pattern of the ecological network weights for each county. Our goals were to investigate (1) the appropriate landscape connectivity distance thresholds in the three study periods, (2) the distribution of regional ecological networks under different PES, and (3) the change in the ecological network weights and optimization strategies for each county based on the gravity model. The results of this study provide a methodological reference for the objective identification of ecological networks in large urban agglomerations and evolution patterns of network weights at the county scale.

2. Materials and Methods

2.1. Study Area

Figure 1 shows the location of the CZXUA (26°03′–28°40′ N, 111°53′–114°15′ E). The CZXUA is the fastest-growing economic region in Hunan Province, with a total area of approximately 28,000 km² and a high level of urbanization. During the study period, a significant expansion of construction land in the central urban area of the CZXUA was observed, and the potential regional ecological network underwent rupture and transfer.

2.2. Data Sources and Processing

Land use/cover remote sensing monitoring data (30 m) were obtained from the Resource and Environment Science and Data Center (http://www.resdc.cn/, accessed on 21 November 2022). Based on this dataset and the secondary land-use types provided, the CZXUA area was further classified into six primary types: cultivated land, forest land, grassland, water area, construction land, and unused land. Specifically, the secondary types of construction land are urban construction land, rural settlement, and other construction land, which includes mining land, factories, and roads (Table S3). Digital elevation model data (30 m) were obtained from a geospatial data cloud (http://www.gscloud.cn/, accessed on 21 November 2022). Road data were obtained from OpenStreetMap (https://www.openstreetmap.org/, accessed on 15 October 2022), and the National Road Traffic Network Vector Map of the Peking University Geographic Data Platform (https://geodata.pku.edu.cn/, accessed on 15 October 2022). Vegetation data were obtained from the monthly 1 km Normalized Difference Vegetation Index (NDVI) spatial distribution dataset for China, collected by the Resource and Environment Science and Data Centre (http://www.resdc.cn/, accessed on 6 January 2023). Population density data were obtained from the World Population Density Map published by WorldPop (https://www. worldpop.org, accessed on 6 January 2023). Nighttime light data were obtained from the National Tibetan Plateau Data Center (http://www.tpdc.ac.cn/zh-hans/, accessed on 6 January 2023). Calculations were performed on a 1 × 1 km grid, with the 30 m resolution base data calculated at the location of the raster centroids when matched [27,28].

2.3. General Framework

This study used experiments to determine appropriate distance thresholds, which, in turn, led to more objective dPC values, to explore the distribution of ecological networks under the different PESs. We analyzed the evolution of the ecological network weights in each county in combination with the gravity model to obtain recommendations. Figure 2 illustrates the framework of this study.

2.4. Optimization of Ecological Source Identification Method

2.4.1. Probability of Becoming an Ecological Source

In this study, indicators closely related to ecological sources were selected to build a reserve of ecological sources for a more objective source selection. The habitat quality module is a widely used and proven model, based on InVEST [29,30], that can characterize regional biodiversity in terms of habitat suitability and sensitivity to threats. Morphological spatial pattern analysis is a method for analyzing patch types and combinatorial patterns based on graph theory [31,32]. The definitions of the patch types are listed in Table S1. Core areas are generally larger and morphologically intact ecological sites that provide more ecological source functions than the other patch types. dPC measures the dispersal of species and the exchange of material and energy information between regional ecological source patches and is closely related to the degree of tight connectivity of the ecological network structure [33]. Accordingly, we propose an ecological source selection method for a 1 × 1 km unit grid using the following formula:

$$PES=\frac{1}{2}\left[HQ+\frac{ARE{A}_c+dPC}{2}\right]\times 100\%,$$

(1)

where PES represents the Probability of becoming an Ecological Source, and HQ is the mean habitat quality per grid unit, with values ranging from 0 to 1. The HQ was calculated from Q_xj in the habitat quality module of InVEST (the principles and equations of which can be found in Table S2; habitat suitability and threat information can be found in Tables S3 and S4; and the habitat quality distribution is shown in Figure S1). Here, AREA_c is the percentage of the core area per unit grid and takes the value of 0–1 (Figure S2 shows the distribution of patch types based on the morphological spatial pattern analysis). Furthermore, dPC is the landscape connectivity of all core areas larger than 1 km² in the study area (the experimental methodology for the landscape connectivity distance threshold is described in Section 2.4.2). The results were normalized, taking values of 0–1. The distribution of the normalized dPC is shown in Figure S3. Therefore, PES was between 0 and 100%. In this study, complete ecological patches existing in the CZXUA at PES > 50%, PES > 70%, and PES > 90% were used as ecological sources to enhance the objectivity of ecological source identification using a multiple probability selection strategy.

2.4.2. Determination of Landscape Connectivity Distance Thresholds

The calculation of PES requires a series of indices characterizing connectivity [25]. Landscape connectivity indicators included the integral index of connectivity (IIC), probability of connectivity (PC), and delta of PC (dPC):

$$IIC=\frac{{\sum }_{i=1}^n{\sum }_{j=1}^n\frac{a_i{a}_j}{1+n{l}_{ij}}}{A_L^2},$$

(2)

$$PC=\frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{a}_i\cdot a_j\cdot p_{ij}^{*}}{A_L^2},$$

(3)

$$P_{ij}^{*}=e^{-\alpha d_{ij}},$$

(4)

$$dPC=\frac{PC-P{C}_{remove}}{PC}\times 100\%,$$

(5)

where n is the total number of patches; a_i and a_j are the areas of patches i and j, respectively; A_L² is the total patch area; $P_{ij}^{*}$ is the maximum likelihood of species dispersal between patches i and j; PC_remove is the landscape connectivity after removing random patch i; and d_ij is the minimum cost-distance threshold. The patch connectivity varies with d, as shown in Equation (4).

2.5. Identification of Resistance

2.5.1. Resistance Factors and Weights

In this study, factors related to human activity disturbance were selected as positive indicators of regional dispersal resistance, which include population density, nighttime lighting, and construction land percentage. The normalized difference vegetation index, which is closely related to biodiversity, was chosen as the negative indicator of resistance. The contributions of positive and negative resistance factors to the spatial heterogeneity of the PES were detected for the three study periods using a geographical detector (see Table S5 for principles, equations, and calculation procedure) [34,35,36]. The q values were used as the weight of the factors to construct a comprehensive resistance surface to partially avoid the subjectivity of traditional hierarchical analysis.

2.5.2. Minimum Cumulative Resistance Model

The minimum cumulative resistance model was used to identify the optimal pathway for species dispersal between ecological sources; this method is well established for application at this stage [15,37]:

$$MCR=\int min{\sum }_{j=n}^{i=m}{D}_{ij}\times R_i,$$

(6)

where MCR is the minimum cumulative resistance, D_ij is the spatial distance from the ecological source point j to i, and R_i is the resistance coefficient of i. Ecological corridors were calculated using the cost path tool of spatial analysis in ArcGIS 10.2.

2.6. Gravity Model

The gravity model can calculate the gravitational forces of interaction between ecological sources and the degree to which the sources and corridors are critical in the structure of the regional ecological network, based on the properties of the ecological sources and resistance in the land units where the corridors are located [25]:

$$G_{ij}=\frac{L_{max}^2\times \ln \left(S_i\right)\times \ln \left(S_j\right)}{L_{ij}^2\times P_i\times P_j},$$

(7)

where G_ij is the strength of the interaction force between patches i and j; S_i and S_j are the areas of patches i and j, respectively; P_i and P_j are the resistance values of patches i and j, respectively; L_ij is the cumulative resistance value of the corridor between patches i and j; and L_max is the sum of the cumulative resistance values of all the corridors in the study area.

2.7. Calculation of Ecological Network Weights for Each County

The gravity center of an ecological source is set to be located in a unit county [38]. There may be no gravity centers within a county, or multiple gravity centers may exist. The ecological network weight is the ecological network interaction force between counties, which can be characterized by the obtained gravity center of the sources and the gravitational force of interaction (G) between these sources [17]; the greater this force is, the more ecologically important this county is in the potential ecological network structure of the region. The calculated values represent both the ecological network weights of the counties themselves and the ecological network gravitational forces of interaction between counties (Figure 3).

3. Results

3.1. Evolution of Multiple Probability Ecological Network in the CZXUA

The distance threshold is a design parameter in connectivity calculations. It is the maximum distance that can be reached by ecological flows within a specific range and is used to evaluate the strength of ecological flows between habitats. As shown in Figure 4, core patches larger than 1 km² (the minimum size of the study grid) were extracted for experimentation. A total of 654, 690, and 712 patches were extracted in 2000, 2010, and 2020, respectively. A total of 10 distance thresholds of 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, and 5000 m were pre-set. We experimented with equal spacing for the distance thresholds. For large-scale landscape connectivity, we refer to studies [2,5,12,14,15,16], where the connectivity of less than 500 m is extremely high and may not fully account for factors such as human disturbance, and ecological networks at this level of connectivity may not be meaningful. For connectivity greater than 5000 m, connectivity is extremely low, and urban species can hardly migrate over such distances. Therefore, we repeated the experiment at equal intervals between 500 and 5000 m. The Number of landscape Links (NL) and Number of landscape Components (NC) in the experimental patches were calculated using Conefor 2.6 software for each distance threshold in 2000, 2010, and 2020. The Link number and Component number were calculated to reflect the changes in landscape connectivity under different distance thresholds and to determine a suitable range of distance thresholds for the CZXUA.

The Link number increased and the Component number decreased with increasing distance thresholds. In 2000, at a distance threshold of 500 m, the Link number value was 794, and the Component number value was 84, with a high degree of landscape fragmentation, making it impossible for ecological networks to exist. At a distance threshold of 2000 m, the Link number value increased, and landscape connectivity improved. When the distance threshold exceeded 3000 m, the Component number value gradually stabilized and reached saturation, and the landscape connectivity was too high, making the ecological network model too idealistic. Therefore, 2000–3000 m was a more reasonable distance threshold range in 2000. Similarly, the distance thresholds were determined to be 2000–3000 m for 2010 and 2020.

To further refine the distance thresholds, dIIC (delta integral index of connectivity) and dPC were selected for trend analysis. Sixty-seven large core patches (≥10 km²) in 2000 were selected, and a total of 11 distance thresholds of 2000, 2100, 2200, 2300, 2400, 2500, 2600, 2700, 2800, 2900, and 3000 m were set, based on the pre-defined range of 2000–3000 m. By comparing the R² value of 0.9739 and the highest relevance of the dPC and delta integral index of connectivity parameters among these core patches, a distance threshold of 2500 m was set for 2000, reflecting the connectivity of the patches in the CZXUA. Similarly, we determined the distance thresholds of 2700 m for 2010 and 2300 m for 2020. The appropriate distance thresholds were entered into the landscape connectivity analysis, and the connectivity probability was set to 0.5 to obtain the dPC for each ecological source.

Through repeated experiments, the appropriate distance thresholds for the CZXUA were 2500, 2700, and 2300 m in 2000, 2010, and 2020, respectively. The dPC values calculated from these thresholds were introduced into the PES equation to obtain the ecological network distributions for PES > 50%, 70%, and 90% (Figure 5). The PES distribution showed a general pattern of high values in the southeast and low values in the northwest. During the study period, the range of areas with low PES values expanded significantly in the central part of the CZXUA (Figure 5a). The trend in the range of high-resistance values was similar to that of the low-PES area, with relatively stable changes in the distribution of low-resistance areas (Figure 5b). From 2000–2010, the high-resistance value tended to expand outward in the central urban area; this phenomenon became more significant from 2010 to 2020. This indicates that with the advancement of urbanization and the mosaic infilling of construction land, the high-resistance value is expanding and spreading outward in the central urban area of the study area, which is strongly disturbed by human activities, thereby impeding the flow of biological information in the region.

As shown in Figure 5c, when PES > 50%, intact ecological source patches were distributed in a fragmented and homogeneous manner with a higher degree of uncertainty, and ecological corridor distributions with this probability ran through the study area. When PES > 70%, and especially when PES > 90%, the range of ecological sources was relatively more fixed and highly heterogeneous across the study area, with a low degree of uncertainty, and concentrated in large woodland patches in the southeast; in fact, ecological corridors under this probability were only present in the southeast.

From 2000 to 2020, when PES > 50%, the degradation trend of ecological sources distributed in the Wangcheng and Yuhu Districts was significant; the number and density of corridors were significantly reduced. This also verified that the expansion of construction land in the central area of the CZXUA greatly reduced the extent to which regional land could function as an ecological source, and was accompanied by a trend of breakage, extinction, and outgrowth of ecological corridors. When the PES was > 70%, there was a significant reduction in the degradation of the identified ecological sources. The corridors located in the northern part of the study area successively underwent outreach and breakage; those located in the southwest underwent outreach and partial breakage; and those located in the central part underwent diversion and breakage. When PES > 90%, the ecological sources and corridors changed less and were reflected only in the degradation at the edges. The shift in the ecological source gravity center was small, but the outgrowth of the corridor could still be observed in the northeast corner of Liuyang and in the eastern part of Yanling County. The break in the corridor in eastern Liling resulted in a thinner ecological network structure.

3.2. Interactions of Ecological Sources under Multiple PESs

These matrices demonstrate the comparative effects of the interactions between the ecological sources of the Changsha–Zhuzhou–Xiangtan urban agglomeration under three different PESs. They were used to express which ecological sources have strong forces, indicate the key hub counties of the ecological network structure, and form the basis for subsequent county weight analysis. Nine source interaction force matrices were derived from the ecological networks under multiple PESs over the three study periods. The number of ecological sources was sorted by area from largest to smallest (Figure 6). We observed that the number of sources was higher for PES > 50%. The number of sources was the lowest at PES > 70%, and the interaction forces were low. This may be due to the wide distribution of sources and their close spacing at a relatively high probability of 50%. Under the relatively strict condition of 90%, the sources were more fixed and concentrated in large ecological areas away from the disturbance. Despite fragmentation, the sources were extremely close and exhibited strong interaction forces. Thus, at PES > 70%, some of the fragmentation sources at PES > 50% were discarded, and the number was reduced to six, while the remaining sources were slightly larger, but not as highly aggregated as at PES > 90%, and therefore interacted weakly.

During the study period, at PES > 50%, the high values of gravity reached 39,972 in 2000 and 314,642 in 2020, which was attributed to the degradation of the central urban ecological network. Most notably, the splitting of the source into two or more, where the fragmentation of the sources reduced the distance between the sources in the later part of the study period, resulting in a grouped increase in gravity where the gravity center shifted. There was negligible change when PES > 70%. At PES > 90%, the gravity between the sources first weakened slightly, but, with the extinction of the source identified with a 90% probability in the central part of the study area in the Liling and Lukou District, the weak gravity values between the remaining sources, which were only located in the northeast and southeast, and the extinct source, which was further apart, also disappeared. The remaining sources were closer together and eventually showed very strong gravitational forces between multiple sources, reaching 131,827,212 in 2020.

3.3. Ecological Network Weights of Counties under Multiple PESs

Under multi-PESs, only 14 of the 23 counties in the CZXUA received ecological network weights during the study period, and all of these counties were far from the central part of the CZXUA, with the exception of the Yuelu and Yuhu Districts. The ecological network weighting matrix for these 14 counties is shown in Figure 7.

When PES > 50%, Ningxiang, Shaoshan, Lukou, and Liling gained weight in the latter part of the study, while Yuelu and Yuhu lost weight, similar to the trend in the ecological network structure, verified by the overall increase in the interaction between Xiangtan, Xiangxiang, and other counties. Xiangtan County’s own ecological network weight was 39,972, 84,501, and 314,642 in the three periods, respectively, showing a significant increase owing to the reduction in the spacing between the sources distributed in Xiangtan County after the network was extended. The dramatic increase in the interaction force between Changsha and Liuyang Counties to 186,684 was not only attributable to network outreach but also to an increase in the number of sources in Liuyang due to fragmentation. When PES > 70%, the counties changed less, with only Liling gaining weight and interacting with other counties in the later stages. When PES > 90%, Liling and You Counties lost weight. Lukou District first gained weight and then lost it. This indicates that under strict source selection conditions, the county-level ecological network weights still fluctuate. The weight of Liuyang changed to 8378, 9039, and 358,194; Yanling’s own weight changed to 170,208, 0, and 131,827,212; and Yanling’s interactions with Chaling changed to 107,232, 18,874, and 4007,423. They all fluctuated and then dramatically increased, indicating that the gravity center shifts in large ecological sources under a high PES greatly affected the position of the regional unit county in the ecological network structure.

4. Discussion

4.1. Large Variation in Structure of Potential Ecological Networks in a Given Area under Different PESs

Previous studies have verified [11,39] that the number of ecological sources and the proportion of their areas in the study area vary considerably under different selection principles. In this study, the distribution of sources was more widespread at a PES > 50%. At PES > 70%, the number and area of sources decreased, whereas, at PES > 90%, sources were present only in the east and south. This suggests the flow of potential ecological network information during the same period. Ecological networks should not be a fixed distribution pattern within a given region, but rather a multilayered flow of ecological information. When the selection principle is at a higher PES, it is not that there is no distribution of ecological networks in the northwestern region, where the PES is low, but rather that the flow of information in this region is weaker than that in the southeastern region, which is easily overlooked when identification conditions are stricter. Existing biological conservation approaches are mostly based on solidly defined protected areas; when ecological networks are identified with a high degree of constraint, it is relatively more difficult to implement targeted conservation strategies in areas with low biological mobility. Some studies have shown that areas with negligible human disturbance, such as agricultural land with extensive use (e.g., orchards), are important stepping stones for wildlife (e.g., birds) in urban areas [40,41]. Therefore, different ecological network identification and protection strategies should be implemented at different levels in rapidly developing urban agglomerations, avoiding the neglect of spaces, such as ecological stepping stones, in northwestern CZXUA. If the multiple probability approach is combined with field research and urban biodiversity monitoring in the future, it may help to quantify the ecological network information flow in low-PES areas and provide a basis for urban and rural greenway construction and ecological restoration.

The degradation and extinction of ecological sources and corridors varied considerably for different identification probabilities. In a fixed study area, negligible significant degradation of the ecological network was observed at PES > 90%. In contrast, at PES > 50%, the ecological network was severely stressed by urbanization over the twenty-year period. We can speculate that at lower PESs, we can identify changes in the “ecological information flow” close to the central urban area and that its fluctuations will be greater. This shows that differences in methods and probabilities significantly influence the definition of ecological degradation and vulnerable areas, delineation of planning boundaries for ecological restoration, and implementation of ecological connectivity measures in the unit area over a long period. The traditional single-probability identification method does not fully characterize the evolution and degradation of the ecological network. Potential ecological networks in the region should be multi-layered and nested, with complex evolutionary mechanisms. Ecological networks far from central urban areas have relatively more stable ecological information flows, more solid network structures [38], and weaker degradation.

4.2. Inconsistent Trends in PES, Ecological Network Degradation, and Gravity

The changes in PES are necessary to observe along with the trends in ecological network degradation, ecological source gravity, and county weights. From the results, the relationship between them may be trade-offs. PES at different values of 50%, 70%, and 90%, the changes in the degradation degree of the identified ecological network, and county gravity may be decreasing in the case of an increase in PES. At a PES > 50%, the variation in the interaction forces between ecological sources in the ecological network pattern was small in the early part of the study, whereas, in the later part, there was greater variation in the interaction forces between some individual sources. The change in forces was not significant when PES > 70% but was again significant when PES > 90%. Combined with the network distribution (Figure 5c), it can be concluded that, with a relatively medium probability of identification, unlike the high- and low-probability identification results, the ecological space in the central area was not identified as a source, and the ecological space in the northwest was not neglected, resulting in a relatively equal distribution of sources around the outer edge of the study area, with relatively stable spacing between them. This indicates the existence of a threshold. The variation in the strict degree of ecological network identification probability over a given study period did not always coincide with the variation in gravitational forces between regional ecological sources. There are parabolic cases with low probability–high gravity, medium probability–low gravity, and high probability–high gravity.

At PES > 90%, the source gravitational variation within the study area was significant, with multiple gravitational extremes occurring in the network pattern later in the study. This is because of the general occurrence, for example, when sources A and B are distant at the beginning and end of the study, source A undergoes a shift in the center of gravity and becomes source X. Source B is minimally degraded but fragmented into two sources (Y, Z). Within this given area, the spacing between X, Y, and Z is less than that between A and B at the beginning, thus showing a dramatic increase in the gravitational force between some of the sources, despite the fact that the areas of the source sites themselves do not change significantly. This partially suggests that the negative effects of source degradation and area reduction on gravity are much smaller than the positive effects of the reduced distance between sources.

Combined with previous findings, although the fragmentation of the ecological network was more severe at PES > 50%, the fluctuations in the gravitational force at this identification probability were not as large as at PES > 90%. This shows that the degree of degradation of ecological sources does not follow a consistent trend with the degree of change in the gravitational force per study area, which can be summarized as follows: low-PES, very significant degradation–generally significant gravitational change; medium-PES: generally significant degradation–no significant gravitational change; high-PES: no significant degradation–very significant gravitational change. These opposite results were observed for different ecological network identification probabilities.

4.3. Gain, Loss, and Value Fluctuations in County Ecological Network Weights More Significant under High-PES

Over the study period, counties in the central urban area lost ecological network weight at PES > 50% while outer-edge counties, such as Xiangtan, gained more weight, suggesting that when focusing on the county scale of the study area, the ecological network weight per county does not always decline with urbanization, but may instead increase significantly. At PES > 90%, the county ecological network weights fluctuated repeatedly, although the change in the network structure was small. This may be because, in the stable southeastern region of the CZXUA, county boundaries are mostly based on ridgelines and are generally far from construction land [42,43], resulting in the gravity center of sources being distributed mostly at the boundary of counties. When degradation of source edges occurs, and the center of gravity shifts repeatedly, counties in such areas are highly susceptible to constant changes in the gain and loss of ecological network weights. This suggests that shifts in the center of gravity of ecological sources under high-PESs are highly determinative of the ecological network status of regional counties.

Regional counties can also gain or lose weight due to this phenomenon. This is because, at a high PES, the sources themselves change less. However, owing to characteristics such as their large area and low resistance, they can generate and change larger gravitational forces, giving the regional counties extremely large values for their own weights and interactions. For example, the direct observation of an exposed pore in a large patch of intact woodland makes its explicit impact appear insignificant. However, its implicit impact is reflected in the fact that it symbolizes the degradation of one large ecological source in the region, with a shift in the center of gravity of the source, causing fluctuations in the weight and status of the ecological network of the two neighboring counties in an area assumed to be at the county boundary. Therefore, monitoring the combined changes in the structure of potential ecological networks in high-PES areas will help assess the status of regional ecological networks at the county scale and implement further conservation monitoring strategies.

4.4. Limitations and Prospects

In this study, the appropriate landscape connectivity distance threshold for the CZXUA was determined through repeated experiments, from which the dPC was derived. This subsequently led to the reserve of ecological sources to enhance the degree of objectivity in the identification of potential ecological networks with the idea of multiple probabilities and to observe changes in the ecological network weight in counties. Current distance thresholds are specific to the overall study area; future research should revise the experimental methods for thresholds for different species, identify the ecological networks of different species [44], and facilitate the development of biodiversity conservation strategies by species. Additionally, the setting of resistance surfaces must be expanded. When the most common and active wild bird species in urbanization is an established study species, the resistance factor does not include topographic elements. If terrestrial species are used as study targets, topographic factors may need to be supplemented. For urbanization and human disturbance factors, the areas of high values largely overlap with areas of gentle topography; although such indicators have been widely applied in existing studies, the redundancy of the indicators is subject to ongoing revision [45]. In future studies, county-scale greenway construction and ecological network restoration measures should be continuously improved, considering the results of studies on the changing status of ecological networks in different counties.

5. Conclusions

Focusing on the CZXUA region with high economic development, this study extracted appropriate landscape connectivity distance thresholds (2500, 2700, and 2300 m) for the region in 2000, 2010, and 2020, and combined connectivity, habitat quality, and morphological principles to construct the PES. We derived the distribution and evolution of the ecological network for the PES at three different strict levels of probabilities greater than 50, 70, and 90%, and the fluctuation of the ecological network weights for each county. Generally, the ecological network has experienced source degradation, center of gravity shift, corridor fracture, and outgrowth. The higher the PES, the less significant the tendency of the ecological network to degrade and fracture under the stress of urbanization. The significant difference between the ecological networks under high and low PES suggests the strength of the potential biological information flow. The potential ecological networks in the unit study area should not be solidly presented with a single identification strategy but require a more flexible and dynamic identification approach. The evolution of the source interactions varied considerably under different PESs. The differences in the reduction of the area and number of fragments cause a trend of network degradation, and the source gravitational forces do not always decrease in a consistent manner. The degradation of ecological sources, gravity center shifts, and the disappearance of corridors, as reflected in network changes, made the weight of ecological networks at the county level fluctuate greatly after downscaling. It is essential to focus on ecological monitoring of neighboring county boundaries, as the land in this area not only has explicit ecological functions but also determines the potential for implicit ecological flows in the region. Although this study has some limitations, it should provide new methodologies and strategies for landscape connectivity calculations, and ecological network identification and evaluation.