A Dynamic Evaluation Method of Urban Ecological Networks Combining Graphab and the FLUS Model

Abstract

Rapid urbanization has led to landscape fragmentation and habitat loss. As an organic structure integrating green space, an urban ecological network can effectively reduce ecological risks and protect biodiversity if its landscape connectivity is maintained. Chaoyang District in Beijing is facing the challenge of transformational development due to excessive urbanization. Taking this district as the study area, this study assessed the environmental impact of different development scenarios on landscape connectivity indices and explored the most relevant strategies for important green space patches by combining ecological network modeling (Graphab) and scenario simulation techniques (FLUS model). The results show that under the urban expansion scenario, the probability of connectivity (PC) decreases by 59.7%, while under the master plan scenario, it increases by 102.1%. Even under the ideal ecological scenario, the ecological network structure of the region faces structural problems. Patches and corridors with high delta probabilities of connectivity (dPC) are concentrated in the north, with no effective connection between the north and south. Finally, planning strategies and priorities for important patches under different urban development goals are proposed through a strategy matrix. Overall, this study proposes a framework for decision-makers to solve planning conflicts between urban expansion and biodiversity conservation, especially for cities in transition.

1. Introduction

Habitat loss due to urban sprawl is considered to be a major cause of biodiversity decline worldwide [1,2,3]. Urbanization leads to fragmentation, converting natural ecosystems into small fragments with complex shapes and affecting landscape connectivity [4,5,6,7]. Landscape connectivity means “the degree to which the landscape facilitates or impedes species movement among resource patches” [8] and depends on the spatial pattern of the landscape [9,10]. A complete ecological network can reduce the impact of landscape fragmentation on biodiversity, thus promoting the conservation of biodiversity [11,12,13]. Given the conflict between urbanization and ecological planning, the establishment of a complete urban ecological network has become a focus of decision-makers concerned with urban planning [14,15,16].

Graph theory serves as useful tool to easily identify ecological networks [17,18] compared to other methods for quantifying landscape connectivity, such as minimal cumulative resistance (MCR) [19], complex network theory [20], and circuit theory [21,22]. Based on topology theory, graph theory describes the spatial structure of patches, corridors, and matrices in the landscape system, focusing on the ecological process among landscape elements [23]. In a landscape graph, the ecological network is abstracted into a series of nodes and links. Habitat patches are defined as nodes, and the functional connections between habitat patches are represented by links [24]. This provides a mathematical research framework for analyzing landscape connectivity and ecological protection issues [25]. Graph theory has been applied to the study of forests [26], agricultural area [27], and urban ecological networks [28,29,30] to assess the impact of fragmentation and loss of landscape connectivity.

In most cases, the identification of ecological networks is based on land use and land cover [31]. Therefore, ecological networks can also be used in combination with land use simulation to assess the impact of different planning scenarios [32,33]. The simulation of the spatial distribution of land in future cities has been shown to be effective for developing sustainable urban planning scenarios with meta-automata CA models and their modifications, such as FLUS [34,35,36], PLUS [37,38], and CLUE-S [39,40], subject to the constraints of natural or socio-economic variables. In most land use simulation studies, urban areas are usually regarded as a threat factor, eroding the habitat space and exerting negative impacts on the natural ecosystem. Nevertheless, with increased awareness on ecological protection and the transformation of urban development goals, many regions are now shifting their priority to ecological development after strong urbanization, and urban areas are beginning to retreat to provide space for natural habitats [41,42]. In this context, few studies have been conducted on the relationship between urban land and habitats. Finding a means of maximizing the role of potential ecological land to enhance landscape connectivity and determining necessary actions for improving the urban ecological network remain crucial issues.

Chaoyang District in Beijing, China, is a city in transition from urban development to ecological development. In 2019, Beijing issued the Chaoyang District Planning (2017–2035) to promote the formation of a green, livable, and sustainable city. The region has been given greater urban ecological responsibility by the government.

Taking the case of Chaoyang District, this study aims to propose a methodological framework combining land use simulation models and graph theory to support biodiversity conservation and urban planning. First, the future state of urbanization under two different development scenarios was simulated using the FLUS model. Then, the landscape connectivity of the ecological network was assessed using the graph theory tool Graphab. Finally, a management strategy for important green space patches in the city is proposed to cope with different positions of urban development. This study also proposes recommendations on urban green space enhancement, conservation, and environmental management to promote urban ecological networks. The approach can also be applied to other urban areas for resolving planning conflicts between urban expansion and biodiversity conservation.

2. Materials and Methods

2.1. Study Area

Chaoyang District is located in the eastern part of the central city of Beijing (Figure 1), between the core area of the capital and the sub-center of the city (39°49’–40°5’ N, 116°21’–116°38’ E). With an area of 454.8 km² (excluding the capital airport), this district is the largest suburban area in Beijing. Chaoyang District has flat terrain and sufficient green space, and accounts for about half of the central city. With continuous ecological construction work such as “urban renewal” and “reserved land for green space”, the regional forest coverage has reached 23.93% and greening coverage has reached 48.05% as of September 2022. Located at the downstream of the catchment area of Beijing, it has a rich water network. The central part passes through the core area and sub-center of Tonghui River, and the northern and northeastern parts are bounded by the Qinghe River and Wenyu River, respectively. The total length of rivers in the region is approximately 150 km.

Chaoyang District has a strong economy and a large population, and is undergoing rapid urbanization. Over the past decade, its GDP has increased from CNY 214.405 billion to CNY 703.790 billion, an increase of 228.3%. In the context of limited land resources, the level of social construction in the region is extremely uneven, and the contradiction between urban construction and ecological protection is growing increasingly prominent. In response, the government put forward the transformational development goal in the Chaoyang District Planning (2017–2035). Chaoyang District is a corridor connecting the capital’s functional core area and the sub-center, which plays a key ecological role in supporting regional coordinated development. According to this goal, urban and rural construction land should be kept within 269 km², and the permanent population should be kept within 3.334 million people.

2.2. Simulation with the FLUS Model

The FLUS model (https://www.geosimulation.cn/FLUS.html, accessed on 11 December 2022) is used to simulate future land use change under human activities and natural influences [43]. It has been successfully applied to land use simulation research in China and around the world [44,45]. The FLUS model improves and optimizes the traditional cellular automata (CA) and introduces an adaptive inertial competition mechanism based on roulette selection. This mechanism can effectively handle the uncertainty and complexity of land use types when they are transformed under the joint influence of multiple driving forces and has higher simulation accuracy than the traditional CA.

The FLUS model has a built-in probability-of-occurrence module. The imported driving force factor atlas is used to calculate the occurrence probability of various land use types through an artificial neural network (ANN) algorithm. It is used to represent the intensity and direction of the driving natural, socio-economic, traffic, and human activity forces on regional development. The spatial datasets used in this study are listed in

Table 1. List of data used in this study and their sources.

Category	Data	Year	Resolution	Data Resource
Remote sensing images	Landsat image	2005, 2020	30 m	GSCloud (http://www.gscloud.cn/), accessed on 11 December 2022.
Socioeconomic data	GDP	2020	1 km	CAS (https://www.resdc.cn/), accessed on 11 December 2022.
	Population density	2020	1 km
Natural data	DEM		30 m	GSCloud
	Slope		30 m	Calculated from DEM
	Aspect		30 m	Calculated from DEM
	NDVI	2005, 2020	30 m	Calculated from Landsat image
Urban transportation	Road net	2020		OSM (https://www.openstreetmap.org/), accessed on 11 December 2022.
	Transit stations	2020
Planning data	Land use demand	2035		Chaoyang District Planning (2017–2035)
	Planning transit stations	2035
	Permanent basic farmland	2035

. All the datasets were resampled to a uniform resolution of 30 × 30 m².

Based on Landsat satellite images and ENVI 5.3 software, the land use and land cover map were obtained by supervised classification methods (maximum likelihood). The kappa coefficients in 2005 and 2020 were 0.83 and 0.81. The high-resolution image of Google Earth in the same period was referred to for manual visual calibration. According to the actual situation of Chaoyang District and the research purposes, land use was classified into the following six categories: forest, shrubland, grassland, farmland, water, and urban area. The land use data in 2005 were set as initial land use to predict land use in 2020. Comparing the actual land use in 2020 with the land use predicted by FLUS model, the kappa coefficient was 0.80, which proves that the simulation results are reliable [46,47]. After accuracy verification, the actual land use data in 2020 were used to simulate the land use pattern of Chaoyang District in 2035 under two scenarios.

In this study, two simulation scenarios with opposing developmental focus were developed, namely the urban expansion and master plan scenarios. Under the urban expansion scenario (UE_Scenario), rapid expansion will erode the surrounding green space, interrupt corridor connections between ecological nodes, and seriously lower the systemic performance and stability of the urban ecological network. It is used to study the feedback of green space in the face of huge threats as a representation of ecological early warning. Its land use demand comes from Markov Chain in 2005 and 2020. Compared with UE_Scenario, the master plan scenario (MP_Scenario) focuses on tapping the ecological potential. Its land use demand comes from the goals of Chaoyang District Planning (2017–2035), representing a sustainable urban development model.

In addition, two parameters are required for model processing. The cost matrix (

Table A1. The cost matrix in FLUS model simulation.

	UE_Scenario						MP_Scenario
	F	S	G	Fl	W	U	F	S	G	Fl	W	U
Forest (F)	1	1	1	1	0	1	1	1	1	0	0	0
Shrubland (S)	0	1	1	1	0	1	1	1	1	1	1	0
Grassland (G)	0	1	1	1	1	1	1	1	1	1	1	0
Farmland (Fl)	0	0	0	1	0	1	0	1	1	1	1	0
Water (W)	0	0	1	1	1	0	0	0	1	0	1	0
Urban area (U)	0	0	0	0	0	1	1	1	1	1	0	1

) represents the conversion rules for each land use type in the simulation. For a land use type that cannot be converted to another type, the corresponding value of the matrix between land use types is set to 0; when conversion is allowed, the value is set to 1. Under UE_Scenario, green space can be converted to urban area, indicating encroachment by urban area; under MP_Scenario, urban area can be converted to green space, indicating ecological development and the shrinkage of built-up land. The weights of neighborhood (

Table A2. The weight of neighborhood in FLUS model simulation.

	Forest	Shrubland	Grassland	Farmland	Water	Urban Area
Weight of neighborhood	1	0.8	0.5	0.3	0.3	1

) represent the difficulty of converting from one land use type to another. The closer the value to 1, the stronger the expansion capacity of the land use type. The weight of each land use type was set according to the research purpose and related literature [34,35].

2.3. Landscape Graph Analysis

Landscape graph analysis was conducted using Graphab 2.6 software (https://sourcesup.renater.fr/www/graphab/en/home.html, accessed on 11 December 2022). Graphab is a software based on graph theory for the construction and visualization of ecological networks and the calculation of landscape connectivity metrics [18,48].

According to the habitat preference of the indicator species, Graphab defines the corresponding land use types as ecological source patches. Amphibians are typical indicator species in urban ecosystems and can provide timely feedback on environmental changes. Bufo gargarizans is widely distributed in Beijing and is highly sensitive to ecological degradation [49]. Habitat fragmentation and loss in the last decade have led to a continuous decline in its population. Therefore, it was selected as an indicator species in the study to construct an ecological network of Chaoyang District.

Land use was categorized as optimal habitat, favorable, and unfavorable (

Table 2. List of land use categories and resistance values.

Category	Land Use	Resistance Value
Optimal habitat	Water	1
Favorable	Forest	10
	Shrubland
	Grassland
	Farmland
Unfavorable	1.86	100

) [32,50,51]. Water was assigned the lowest resistance value (1) as a habitat, as well as a preferred type of migratory foraging. Considering the close connection between its habitat and waterfront perimeter, non-building land within a 30 m buffer zone of water bodies was increased to reflect ecological characteristics. Favorable types such as forest and shrubland were assigned a low-cost value (10). Urban area is unfavorable for survival and access and was assigned a high-cost value (100). Referring to relevant ecological literature [52,53] and the landscape pattern of the study area, a minimum ecological source area of 3 hm² and a distance threshold of 1500 m were set in this study.

Four landscape connectivity indices were selected for the study, including probability of connectivity (PC), size of the largest component (SLC), mean size of the components (MSC) at the global level, and fractions of delta probability of connectivity (dPC) at the local level. The global level connectivity indices were used to represent the landscape connectivity of the regional ecological network, and the local level connectivity index was used to indicate the magnitude of the role played by individual patches or corridors in maintaining the regional landscape connectivity.

PC indicates the probability of connections between two randomly placed points in the study area [54]. It is used to evaluate the overall connectivity level of different development scenarios. The value ranges from zero to one, with higher values showing higher global level landscape connectivity of the ecological network. The formula is as follows:

$$\mathrm{PC}=\frac{1}{{\mathrm{A}}^2}{\displaystyle \sum }_{\mathrm{i}=1}^{\mathrm{n}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{a}}_{\mathrm{i}}{\mathrm{a}}_{\mathrm{j}}{\mathrm{e}}^{{-\mathsf{\alpha }\mathrm{d}}_{\mathrm{ij}}}$$

(1)

where ${\mathrm{a}}_{\mathrm{i}}$and ${\mathrm{a}}_{\mathrm{j}}$ are the areas of patches i and j, A is the total area of the study zone, and ${\mathrm{e}}^{{-\mathsf{\alpha }\mathrm{d}}_{\mathrm{ij}}}$ is the maximum probability of movement between these patches. ${\mathrm{d}}_{\mathrm{ij}}$ is the least-cost distance between i and j, and α expresses the intensity of decreasing probability resulting from the exponential function.

SLC indicates the largest capacity of components in the ecological network [55]. This is a measure of dominance of the core patch. Larger values indicate higher dominance of connected patch groups and higher resistance to disturbance of core ecological patches. It is calculated as follows:

$$\mathrm{SLC}=\max \left\{{\mathrm{ac}}_{\mathrm{k}}\right\}$$

(2)

where ${\mathrm{ac}}_{\mathrm{k}}$ is the capacity of component k (sum of the capacity of patches composing k).

MSC indicates the average of component capacities in the ecological network [56]. It is used to measure the degree of fragmentation of the ecological network. A larger value indicates lower fragmentation and higher stability of the overall network. Like SLC, MSC itself often has limited explanatory value because it does not transmit direct information about spatial distribution and connectivity. It is calculated as follows:

$$\mathrm{MSC}=\frac{1}{\mathrm{nc}}{\displaystyle \sum }_{\mathrm{k}=1}^{\mathrm{nc}}{\mathrm{ac}}_{\mathrm{k}}$$

(3)

where $\mathrm{nc}$ is the number of components and ${\mathrm{ac}}_{\mathrm{k}}$ is the capacity of component k.

This study quantified the local contribution of each node and its link to landscape connectivity through the nodes removal methodology [57]. dPC indicates the rate of variation between the values of PC and PC’ corresponding to the removal of patch i.

$${\mathrm{dPC}}_{\mathrm{i}}=\frac{{\mathrm{PC}-\mathrm{PC}}_{\mathrm{i}}^{’}}{\mathrm{PC}}$$

(4)

A higher value for a patch or corridor indicates that it has a larger contribution to the landscape connectivity of the overall ecological network and, thus, a higher conservation priority. The fractions were calculated from 0 to 1 and the results were multiplied by 100 to interpret the percentage. The natural breaks algorithm was used to classify dPC to very high, high, middle, low, and very low levels.

3. Results

3.1. Scenario Simulation

As shown in

Table 3. Area and proportion of each type of land use in 2020 and under two scenarios.

		Forest	Shrubland	Grassland	Farmland	Water	Urban Area
2020	Area (km2)	59.03	18.27	34.94	25.64	11.24	305.68
	Proportion (%)	12.98	4.02	7.68	5.64	2.47	67.21
UE_Scenario	Area (km2)	36.48	16.44	25.6	28.31	11.75	335.12
	Proportion (%)	8.02	3.61	5.63	6.22	2.58	73.94
MP_Scenario	Area (km2)	80.62	29.53	41.37	21.63	12.85	269.9
	Proportion (%)	17.73	6.49	9.1	4.51	2.83	59.34

, significant differences were observed in the characteristics of the area proportions of land use simulated under the two scenarios. Under UE_Scenario, the total amount of green space decreased significantly; on the contrary, under MP_Scenario, it increased significantly.

Under UE_Scenario, the urban area of Chaoyang District is 335.12 km², accounting for 73.94%. The total area of green space is 119.68 km², reflecting a decrease of 20.48% compared with 2020. The dominant type is forest land, with an area of 59.03 km², followed by farmland, grassland, and shrubland. The smallest type is water body, with an area of 11.75 km², accounting for 2.58%.

As shown in Figure 2, under UE_Scenario, the sprawling expansion of construction land in Chaoyang District presents a state of uncontrolled development, and green space is seriously eroded. The green space system of the first and second greenbelts in Chaoyang has clearly deteriorated, and the urban ecological network tends to be fragmented. Many rivers have been interrupted, most missing river sections lack a green space buffer, and the ecology is fragile. The driving factor of transportation location has a significant impact on urban land use simulation. Areas with high road density and a large number of stations near the central city in the west have a significant rate and scale of construction land expansion. In contrast, the eastern peri-urban areas have less outward expansion of construction land and less erosion of green space.

Under MP_Scenario, the urban area in Chaoyang District is 269.90 km², accounting for 59.34%, which basically meets the requirements of the planning goals. The total area of green space is 184.90 km², reflecting an increase of 24.73% compared to 2020. Forest land is still the largest type of green space with an area of 80.62 km², accounting for 17.73%. It is followed by grassland, with an area of 41.37 km², surpassing that of cropland. The area of water is 12.85 km², accounting for 2.83%.

As shown in Figure 2, under the MP_Scenario, the construction land in Chaoyang District is decreased, separated by the growing green space, showing a trend of decentralized development. The NDVI is considered to be the main trigger factor of this scenario. The patches grow continuously and significantly in areas with high green space density, especially in the northwest and southern edge areas. However, the area of isolated patches in the central city increases slightly. Under this scenario, the river system develops benignly, the river ecology tends to be stable, and the vegetation buffer zone of the local river corridor grows.

3.2. Landscape Connectivity

3.2.1. Global Level

According to

Table 4. Landscape connectivity at the global level in 2020 and under two scenarios.

Metrics	2020	UE_Scenario	MP_Scenario
PC (10−3)	5.78	2.33	11.68
SLC (km2)	75.71	29.96	104.06
MSC (km2)	3.35	1.86	9.11

, the change trends of the connectivity index of the urban ecological network under the two scenarios completely oppose each other, implying different ecological impacts on the development of the urban ecological network. Compared with 2020, the three indices of excessive urban expansion (PC, SCL, MSC) all decrease under UE_Scenario; on the contrary, the indices all increase under MP_Scenario.

The PC of Chaoyang District’s ecological network drops to 2.33 × 10⁻³ under UE_Scenario (approximately 60% less than that in 2020). The overall connectivity decreases significantly, indicating that the expansion and erosion of construction land have a serious negative impact on urban ecology. SLC is 29.96 km², showing a decrease of approximately 60%, similar to that of PC. MSC decreases from 3.35 km² in 2020 to 1.86 km² (a decrease of 44.48%). These changes show that the area of the core patch connector in the urban ecological network is greatly reduced, the degree of fragmentation is increased, and the overall stability is reduced.

Under MP_Scenario, the overall connectivity of the ecological network increases significantly. PC increases to 11.68 × 10⁻³ (approximately twice of that in 2020). SLC increases to 104.06 km² (an increase of 37.45%), and MSC increases to 9.11 km². These changes indicate that the growth of green space and the evacuation of construction land contribute to the structural enhancement of the ecological network, which significantly improves the urban ecological environment. The increase in the largest connected unit should be attributed to the establishment of new connections and integration of the original independent patches in the region. This enhances the dominance of core green patches, which, in turn, improves the overall level of landscape connectivity and strengthens the risk resistance and stability of the urban ecological network.

3.2.2. Local Level

As shown in Figure 3, patches with high levels of dPC are concentrated in the north under UE_Scenario. Compared with 2020, the urban ecological network of Chaoyang District is degraded and the main structure is destroyed. The largest landscape linkage in the north is divided into two parts. The number of isolated patches in the region has increased significantly. Smaller patches in the western part of Chaoyang are completely occupied by high-intensity urban construction and expansion activities due to their proximity to the central city. A large amount of highly scarce urban habitats disappear, resulting in a loss of biodiversity. The patches in the south show low dPC levels, contributing less to the overall regional landscape connectivity. The southern part shows a single network structure and no scale has formed yet.

In contrast, the network structure under MP_Scenario is clear and concise, and the northern part further develops to form a complete ring network with reference to the current situation. The number of isolated patches is reduced, indicating that the radiative linkage capacity of the core patches is significantly enhanced and the radiative level of the ecological network is improved. Although the green areas in the south form the prototype of the ecological network, the distribution of patches in the south continues to show low dPC levels and is not effectively improved. This shows that the problem of unbalanced development of the urban ecological network between the north and south of Chaoyang District is very serious.

From the aspect of corridors (Figure 4), under UE_Scenario, those with high dPC levels are distributed in the northern edge of the region. After excessive urbanization and expansion development, they continue to maintain a high level of landscape connectivity. This indicates their high ecological stability and resistance to ecological threats. The corridors of the two core patches in the northeast are disconnected and the structural nature of the network is severely disrupted, resulting in a significant decrease in the connectivity of the corridors emanating from the core. Under MP_Scenario, the number of corridors with high dPC levels increases. With the development of green space, the complete urban ecological network in the north is promoted. As with the patches, corridors in the south still show very low dPC levels. Although the river corridor of Tonghui River forms a connection, the three areas of north, middle, and south are still not ecologically connected.

4. Discussion

4.1. Management Strategies for Important Patches

We tracked the evolutionary state of the top twenty most important patches of the dPC index (Figure 5a and

Table A3. Data statistics of top 20 patches of dPC index.

2020			UE_Scenario				MP_Scenario
Rank	S	dPC	S	ΔS	dPC	ΔdPC	S	ΔS	dPC	ΔdPC
1	1770.93	0.38	976.95	−44.83	0.32	−15.88	3101.49	28.60	0.49	75.13
2	1238.76	0.27	1002.24	−19.09	0.32	20.23	1883.43	28.16	0.34	52.04
3	1113.75	0.25	1261.26	13.24	0.43	69.06	2304.90	86.49	0.47	106.95
4	426.24	0.10	329.58	−22.68	0.11	10.31	1019.43	102.96	0.20	139.17
5	313.02	0.05	271.17	−13.37	0.09	61.78	2304.90	789.36	0.47	636.34
6	232.56	0.05	168.21	−27.67	0.00	−95.77	307.89	32.78	0.06	32.39
7	211.50	0.05	175.86	−16.85	0.26	456.36	3101.49	963.99	0.49	1366.43
8	203.94	0.04	1261.26	518.45	0.43	880.95	2304.90	982.11	0.47	1030.19
9	215.37	0.04	176.85	−17.89	0.00	−92.63	1019.43	451.91	0.20	373.34
10	276.84	0.04	258.12	−6.76	0.07	99.80	294.48	45.71	0.05	6.37
11	170.55	0.03	124.20	−27.18	0.04	21.69	3101.49	1384.13	0.49	1718.52
12	160.83	0.03	329.58	104.92	0.11	247.52	1019.43	539.39	0.20	533.86
13	161.10	0.03	134.91	−16.26	0.04	42.75	2304.90	1492.71	0.47	1330.73
14	610.38	0.02	239.85	−60.71	0.02	−1.36	720.27	34.50	0.03	18.00
15	91.17	0.02	52.74	−42.15	0.01	−27.66	1883.43	1697.58	0.34	1965.84
16	82.44	0.02	55.89	−32.21	0.09	375.10	2304.90	2517.55	0.47	2695.85
17	71.19	0.01	54.54	−23.39	0.03	101.70	2304.90	3102.80	0.47	3137.67
18	63.27	0.01	14.13	−77.67	0.00	−99.93	75.60	64.14	0.02	19.49
19	89.01	0.01	38.61	−56.62	0.01	−6.00	1883.43	2401.10	0.34	2015.98
20	63.63	0.01	19.44	−69.45	0.01	−46.80	172.89	149.55	0.03	171.71

) through simulations under the two scenarios. These important patches are essential for the ecological network of Chaoyang District. The change rates of the patch area and dPC (ΔS and ΔdPC) index were counted; then, the strategy quadrant diagram (Figure 5b) was generated. The results showed a significant difference in the quadrant distribution of ΔS–ΔdPC under UE_Scenario and MP_Scenario. Under UE_Scenario, the area and dPC of most patch areas decrease, and the data points are concentrated in the second and third quadrants, with a small number in the first quadrant. In contrast, the area and dPC of all patches increase under MP_Scenario, and the data points are distributed in the first quadrant.

In the first quadrant, the patches are distributed in the upper and lower areas of the diagonal, representing different area–dPC conversion efficiencies. The upper area is mostly green space patches in development, which should be built to enhance the green space and fully stimulate the potential ecological value of small- and medium-sized patches. These patches have the highest construction priority. For example, the southern network is improved by upgrading patch No. 14, and a northern closed-loop network is formed by upgrading patch No. 18. Moreover, under MP_Scenario, patches No. 15, 16, and 17 realize direct connections through the new green space, effectively improving the landscape connectivity. The area–dPC conversion efficiency is relatively low in the lower region, such as in patches No. 1 and 2. Owing to the large-scale, mature, and stable green space, high dPC index, and small ecological potential of such patches, attention should be paid to the construction of surrounding corridors to improve the core radiation. Under UE_Scenario, only patches No. 3, 8, and 12 belong to the first quadrant and have excellent ecological resilience. Of these, patches No. 3 and 8 are distributed in the eastern edge of Chaoyang District and play an important role in the ecological connectivity of Chaoyang and the urban sub-center. Patch No. 12 has a locational value for realizing the ecological connection between the north and the central part of the city.

Patches in the second quadrant can withstand a certain range of area reduction to promote the optimization of network structure and improve landscape connectivity. The significance of this planning strategy is that no “blind” pursuit of the maximization of green space is required in the construction of urban ecological network, which is of great significance to future urban transformation and development. The reduced green space can be converted into opportunity land for urban construction, ensuring land resources for urban economic development and enhancing the efficient use of land. It can also be used to compensate for the lack of green space in the south of Chaoyang District and promote the coordinated development of the region.

Patches in the third quadrant have smaller areas and lower landscape connectivity. Therefore, protection strategies need to be implemented. In particular, patches No. 18, 6, and 9, with substantially reduced connectivity, have the highest protection priority. Their performance should be improved in order to achieve future connectivity between the north and the center. Patches in this quadrant were divided into two categories. The first type comprises the core ecological source sites of the region, such as patches No. 1, 2, and 4, which are at the top of the dPC index ranking. Such patches are better ecosystems in their own right, and the focus should be on limiting the high-intensity threat factors in the surrounding areas. Junction areas with frequent energy exchange should be regulated to mitigate the negative effects of edge effects and protect the ecological functions of green space [58]. The other type comprises patches with the lowest dPC index, namely No. 18, 19, and 20. Their ecosystems are fragile and vulnerable to erosion from the construction land expansion of surrounding areas. Consequently, a mature green space system cannot be built in the short term. Therefore, focus should be placed on the construction of buffer isolation zones to enhance resilience against ecological risks. Unlike the active optimization and promotion strategy of the first quadrant, the protection strategy is more passive.

4.2. Pros, Cons, and Future Perspectives

By combining ecological network modeling and land use scenario simulation, we effectively determined the priority and management modes of regional green space. More detailed and diverse strategies are also proposed for patches, rather than just protection and restoration. Some green spaces can also be optimized as urban areas, which can instead enhance landscape connectivity. Particularly in the context of limited land resources and rapid urbanization, this study provides a possible spatial reference for balancing urban development and ecological protection. In reality, the scale of protected areas is increasing globally, while habitat loss and conservation are decreasing [59]. In this regard, simply increasing the area of green space will not be entirely beneficial to conservation of biodiversity. Instead, it is necessary to optimize the structure of ecological networks and land use patterns [60]. Nevertheless, the form and scale of this green space transformation still require further research.

This study designed two urban development scenarios with opposing goals to assess the potential impacts on ecological networks. The proposed methodological framework can overcome the limitations of high precision data and high time costs. Decision-makers can quickly compare the evaluation results, identify habitat patches critical to ecological processes of the network, and then improve planning schemes by controlling land use and driving factors. By employing this approach, potential green spaces can be enhanced, threatened green spaces can be protected\, and areas can even be selected for mixed-use development. Urban development is often multi-objective, and a single scenario is not a feasible solution. The focus of this study was to explore the positive or negative impacts of future cities on green space under different visions, not to strictly predict urban land. We believe that this method can also be extended to the study of other similar cities to develop the best planning scheme by integrating different urban development goals and promoting the efficiency of land resource utilization.

Land use change is a complex issue influenced by many unpredictable variables. Moreover, theoretical assumptions applied to the model, as the well as the sensitivity and uncertainty of parameters, may affect the accuracy of the model. More detailed data are key to stable and reliable evaluation. First, a low resolution may oversimplify complex landscape features within the city, resulting in a certain amount of missing information. Further, ecological network modeling is based on land use and land cover. Considering the limited land use types, even the same type may feature significant differences in terms of its actual ecological nature. In addition, updating the model is also a way to improve accuracy. For example, the PLUS model is better at mining conversion rules and has higher simulation accuracy [61]. The graph-theory-based model also can be combined with species distribution models (SDM) to identify habitat patches and generate maps of habitat suitability [62].

5. Conclusions

By combining the ecological network model (Graphab) and the scenario simulation model (FLUS), this study assesses the impact of different urban development scenarios on the landscape connectivity of an ecological network and explores different development strategies for important green space patches. These strategies are (i) the construction priority strategy, where green space has high ecological potential; (ii) the opportunity land strategy, where green space can be transformed into urban construction land to ensure land resources for urban economic development but, at the same time, optimize the ecological network structure; and (iii) the conservation priority strategy, which protects green spaces facing high ecological threats. In general, we enriched the identification and evaluation framework of patches in the ecological network, aiming to seek a balance between urban development and ecological protection in the future, especially for cities in transition or with strong urban and ecological contradictions. Our method provides technical support and theoretical reference for governments to formulate detailed ecological management policies for promotion, construction, and protection.