Ecosystems support life on Earth, and thus play a vital role in human well-being, either directly or indirectly1. In recent years, anthropogenic activity has facilitated the invasion of ecosystems by nonnative species and natural hazards, leading to the worsening of various environmental problems, including the degeneration of ecosystem services and a sharp decline in biodiversity1,2. Low continuity between habitat patches caused by the fragmentation of ecological landscapes (i.e., natural or semi-natural habitats) represents the greatest threat to biodiversity conservation3,4,5. However, growing environmental awareness and an improved understanding of how human communities interact with their environment have led to growing concerns about enhancing habitat patch continuity within ecosystems6,7. Nevertheless, recent studies show that economic growth has actually made humans more dependent on ecosystem services and biodiversity8. Therefore, it is particularly important to maximize ecosystem service values by constructing networks that enhance the functionality of urban ecosystem services9,10.

Research on ecological network construction has been widely carried out on a global scale11,12. The ecological network is a representation of the biotic interactions in an ecosystem, in which ecological corridors link protected habitat patches13. A habitat patch is a set of landscape patches, while a landscape patch is the basic unit—a relatively homogenous mosaic of familiar land-use types that differ from the surrounding background—that formulates landscape patterns. Previous studies have suggested that habitat patches are areas where organisms aggregate, representing stepping stones for migration14,15, while ecological corridors are narrow bands of vegetation that promote biological migration between the two habitat patches, allowing wild animals to survive16. These two landscape types form the core of ecological networks. However, one of the main limitations of studies on ecological newtorks is that the movement data of target speices are frequently unavailable. Therefore, there have been an increasing number of simulation-based studies on migration and biodiversity conservation in an attempt to address such defects17,18,19,20,21,22. In particular, methods based on graph theory emphasize the functional connection between habitat patches, which is the effective relationship between components of an ecological object or process with corresponding characteristic scales16. Graph theory has been gradually introduced to conduct research on the ecological networks of various land-use types, including cities, farmlands, and forests20,23,24,25.

The most important step when using graph theory is the construction of resistance surfaces. The resistance surface is a raster map (mosaic with a value, a larger value corresponds to higher resistance) indicating the level of disturbance or degree of difficulty that target species are expected to encounter when moving between patches, which significantly affects the outputs of ecological network research26. Therefore, many studies construct resistance surfaces based on various methods, such as biological behavior resistance estimates, expert scoring, entropy weighing, and landscape development intensity indexing10,27,28,29. However, most studies have constructed landscape resistance surfaces based on expertise and overall ratings for certain land-use types, leading to the resulting landscape resistance surfaces being heavily dependent on grading factors5,26. In fact, there are differences between the same land-use types owing to their different locations and surroundings. As a result, previous studies have weakened the differences in resistance of the same land-use type.

In this study, we propose a habitat quality-based method that involves simulating the sensitivity of different land-use types to impact factors. Within this framework, we evaluated the habitat quality and functionality of ecosystem services for each assessment unit1. Pixel-scale habitat quality was used to characterize the optimal survival, reproduction, and energy flow conditions that an assessment unit provided to organisms. Besides facilitating studies on how human activities affect habitat quality and animal migration, this information is expected to help estimate the landscape resistance of different assessment units. Good habitat quality promotes the dispersal of animals, and thus corresponds to low resistance. Although several studies on urban ecological network construction do exist27,30,31, a limited number of studies have analyzed the effects of different resistance surfaces on corridor simulations. In addition, these investigations weaken the significance assessments of habitat patches and corridors, which may be not conducive to the implementation of actual protection measures.

The ecological protection planning of Changzhou, China—which contains habitat patches but no corridors—was implemented to improve ecosystem services and biodiversity. Thus, this study proposes a set of procedures that could be applied to ecological network research by identifying potential corridors in Changzhou. Based on landscape ecology, graph theory, and habitat quality assessments, this study aimed to (1) propose a method for constructing landscape resistance surfaces; (2) quantitatively characterize potential corridors in the study area by using a minimum cumulative resistance model; (3) evaluate the importance of habitat patches and corridors; and (4) compare the effects of different methods of constructing resistance surfaces on potential corridor simulations and conduct an importance assessment of habitat patches and corridors. This information is expected to provide quantitative scientific data to enhance environmental protection in Changzhou.

Study Area and Data Preparation

Overview of the study area

Changzhou (119°08′–120°12′E, 31°09′–32°04′N) is located in the Taihu Lake Plain, Yangtze River Delta, in the southeastern part of Jiangsu Province, China. It has a subtropical monsoon climate, with an annual average temperature of 16.5 °C and an annual precipitation of 1063.71 mm. From 2006 to 2014, the built-up area in the city increased by 25.68%, demonstrating clear urban expansion. However, the percentage of farmland, forest, grassland, and water body areas offering ecological services has decreased. The shrinkage of ecological land area, the fragmentation of habitats, and the obstruction of corridors are major contributors to the deterioration of ecosystem services in Changzhou, and present an increasing burden on environmental protection.

Source of data

The following datasets were provided by the Land and Resources Bureau: land-use change vector data (shapes with an attribute of its land-use type) from 2006 to 2014, administrative divisions, vector data of roads, vector data of built-up land, and ecological protection planning containing a nature reserve vector map of Changzhou. The nature reserve vector map is a highly important part of ecological protection planning in Changzhou. The map depicts the core area in Changzhou that could be protected by administrative means. Data regarding people per square kilometer of land area and Gross Domestic Product (GDP) per square kilometer in China (2010) were obtained from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences ( Certain factors (such as land-use type, level of human disturbance, and landscape function) were used to classify the landscape types in Changzhou as farmland, forests, built-up land, riparian green space, and other types. Because the riparian green space is always a strip of land along a river or lake, we integrated riparian green space and water bodies as riparian green space. This information was used to generate the landscape classification map of Changzhou.


We identified potential corridors based on graph theory. The three main steps were 1) the selection of habitat patches, 2) the construction of resistance surfaces, and 3) the identification of potential corridors. First, habitat patches were acquired from the nature reserve vector map. To analyze the effect of different resistance surfaces on corridors, we constructed resistance surfaces based on three different methods: (1) the habitat quality-based model, (2) the entropy coefficient method, and (3) the expert scoring method. Second, based on different resistance surfaces and the minimum cumulative resistance model, we identified all potential corridors in Changzhou. The results generated three groups of habitat patches and corridors reflecting the three methods we used. The resulting habitat patches, selected from the nature reserve vector map, were the same, while corridors differed as a result of differences in the resistance surfaces. Thus, we were able to compare the effects of the different methods of construction on the simulated potential corridors. Finally, the importance of both habitat patches and potential corridors were assessed. A flow chart of the techniques used is shown in Fig. 1.

Figure 1
figure 1

Technical flow chart.

Target species and the selection of habitat patches

Connectivity is a species-dependent property32. Therefore, selecting target species that may be used as the basis for the study of ecological networks is necessary27. In this study, we focused on the common species of Changzhou, which were mostly small mammals, such as weasels, hares, and badgers. These animals prefer to move in vegetated habitats, such as forests and riparian green spaces. Connectivity is a species-dependent property32. Thus, based on the nature reserve vector maps provided by the Changzhou Bureau, we selected habitat patches that contained forested areas and riparian green spaces, and excluded river habitat patches, which would not suitable for our target species. These habitat patches serve as “sources” of matter, energy flow, and animal migration, and are also the origins where ecological processes develop10,14.

Changzhou is one of the most developed cities in China. As a result, the habitat patches in this city are significantly affected by urbanization. Habitat patches are mostly distributed in the western and eastern parts of Changzhou, with fewer patches located in the north-northeastern part (Fig. 2(b)). The degree of ecological land fragmentation in Changzhou was evaluated using the patch density index33. The index grew from 1.42 in 2006 to 2.10 in 2012 (Fig. 2(c)), illustrating that habitat patches became increasingly fragmented. Thus, spatial fragmentation and uneven distribution lead to difficulties in constructing an effective ecological network. In this study, 20 ecological sources were selected, as shown in Table 1 and Fig. 2(a).

Figure 2
figure 2

(a) Habitat patches in Changzhou, (b) differences in the spatial distribution of different habitat patches (where, based on the center of gravity of Changzhou, study area was divided into eight equal areas. Then, the area of habitat patches located in different sectors was statistically calculated), and (c) variations in ecological land area (the total area of farmland, forests, and water bodies in Changzhou) and the patch density index in Changzhou during 2006–2014 (where the left axis indicates the ecological land area and the right axis indicates the patch density index values). (Created by ArcMap, version 10.2, Boundaries of Jiangsu province and Changzhou, land-use type data and habitat patches acquired from Changzhou Land Resources Bureau).

Table 1 Selected ecological sites in Changzhou.

Construction of landscape resistance surfaces

As shown in previous studies, the resistance surface has a major impact on the simulation of corridors26. Thus, to compare the effects of different landscape resistance surfaces on ecological network construction, and to verify the feasibility of applying the method proposed by this study landscape resistance, various methods used by previous studies were employed to construct different landscape resistance surfaces. Such methods included the entropy coefficient method34 and the expert scoring method27.

Habitat quality-based method

Research has shown that habitat quality is associated with the intensity of human activity and the land-use types of surrounding areas35,36. The InVEST (Integrated Valuation of Ecosystem Service and Tradeoffs) model was developed jointly by Stanford University, the University of Minnesota, The Nature Conservancy, and the World Wildlife Fund2. This model simulates the combined effects of land-use changes and human activities on ecosystems, and it provides a visual tool that quantitatively evaluates ecosystem service functions2,37,38. The habitat quality component of the biodiversity module of this model was adopted in this study to analyze how the intensity of human activity influences habitat quality in Changzhou. The principles for calculating this model are:

where Hj denotes the habitat suitability of land cover type j, denotes the degree of habitat degeneration of land cover type j in raster x, z is a model parameter with a value of 2.5, and k is the half-saturation coefficient. During this process, five raster layers and two tables were compiled. First, we created a land-use map in raster format at a resolution of one pixel = 30 m × 30 m. Second, roads and built-up land were created in raster format at a resolution of one pixel = 30 m × 30 m to indicate obstructions to animal migration. Then, data regarding people per square kilometer of land area and GDP per square kilometer in Changzhou were included to indicate the intensity of human activity. Finally, attributes of threats, habitat types, and target species sensitivity to different threats were created as shown in Tables 2 and 3.

Table 2 Attributes of threats.
Table 3 Habitat types and degree of sensitivity to each threat.

The habitat quality results produced by InVEST were presented as raster data in the range of 0 to 1. A larger value corresponds to better habitat quality, and vice versa. When InVEST is applied to the construction of landscape resistance surfaces, habitat quality is converted to landscape resistance. Units of better habitat quality have smaller landscape resistance, and vice versa. The equation governing conversion is:

where Qj is the result of habitat quality for raster unit j, and is the landscape resistance of the same unit. Within this process, the constructed resistance surface was completed based on habitat quality.

Entropy coefficient method

The entropy coefficient method is an objective weighting method in which weight is determined by entropy. Entropy is the amount of additional information needed to specify the exact physical state of a system. The greater the entropy, the more information it provides34. This process involves three main steps: 1) normalizing the initial information matrix, 2) calculating entropy weight, and 3) calculating resistance. In this study, landscape indexes and ecosystem service values were used as the ecological attributes for each land-use type in the weighted calculation (Fig. 3)29,39. Thus, the initial information matrix contains seven rows and five columns. To compare the different factors, the value of the factors was normalized by equations (3) and (4). i and j correspond to the number of rows and columns, respectively, in the initial information matrix. is the normalized result, xij is the original value, and and are the maximum and minimum value in ith row, respectively.

Figure 3: Landscape indexes of different land-use types (first seven columns from the left) and landscape resistance estimates (last column) used in the entropy coefficient method (the axis at the bottom of the figure is the value corresponding to the indexes shown at the top; the values of the indexes were calculated using Fragstats 4.2).
figure 3

(Created by Fragstats, version 4.2,

Through an integrated assessment of the volume of information carried by each factor, an integrated weight coefficient is calculated based on the abundance of information that a factor provides. The principles for the calculation are39:

where wi is the weighted coefficient of the corresponding factors. If fij = 0, we assumefij In fij = 0. Based on the normalized results of and wi, the final matrix could be calculated by . The resistance of different land-use types is defined as the sum of the columns in the final matrix.

Expert scoring method

The expert scoring method examines the sensitivity of each landscape to human disturbance, and provides a quantitative characterization of the magnitude of its resistance27,40. First, we based the different land-use types on those suggested by experts of the selected focal species in Changzhou. Then, the resistance values were primarily chosen according to vegetation type, in addition to the area of green space and the degree of anthropogenic disturbance30. The resistance values were set between 1 and 1000 to indicate the level of disturbance or degree of difficulty that target species would encounter when moving between patches. The reference landscape resistance values are shown in Table 4.

Table 4 Landscape resistance estimates based on the expert scoring method.

Simulation of potential corridors

Corridors are vital components of ecological networks. Corridors provide paths for animal migration, as well as conserving biodiversity and facilitating sustainable development41. As a result of increases in the intensity of human activity, ecological connectivity has sharply declined in Changzhou. In research on the construction of ecological networks, two corridor simulation methods are primarily used: (1) directly tracking animal migration paths and simulating ecological processes42, which is highly dependent on data availability, is difficult to perform, and is less commonly used in urban environment studies26; and (2) model simulation, which resolves the difficulties of data collection to some extent, resulting in this model being widely applied in ecological network research27. In particular, the minimum cumulative resistance (MCR) model is a raster-based optimization algorithm that was originally designed to find the least-expensive path for a link between two nodes, since we assumed that target species follow an optimum route to minimize exposure to intervening low-quality landscape patches. This model reveals obstructions and boosts underlying migration processes19. This model is often combined with landscape graph theory to reflect complex network structures and ecosystem-scale relationships between energy, matter, gene migration, and the underlying surface43. In this process, two layers are used. The first is the habitat patches layer (treated as the nodes). The second is the resistance surface layer. The resistance value for each potential corridor is determined by the number of raster units that are covered and the landscape resistance of each unit:

where MCR is the resistance of a potential corridor between two habitat patches, Dij is the number of raster units between two habitat patches, Rj is the resistance of a raster type, i is the number of habitat patches, and j is the number of raster units on the resistance surfaces.

Network continuity evaluation

Landscape continuity is the relationship between landscape elements in their spatial structures. This parameter is used to determine the structural characteristics of a landscape. It is commonly measured by continuity indexes, including the structural continuity index and the functional continuity index44. In recent years, quantitative assessments on continuity and the complexity of ecological networks based on landscape graph theory have increased45, with new indexes being developed, such as the overall network continuity index, based on graph theory46,47. Compared with traditional descriptive indexes on network complexity and continuity, graph theory-based evaluation indexes integrate the ecological attributes of a habitat into the calculation, giving these metrics ecological significance. Landscape continuity could be used to assess the importance of habitat patches and corridors. The evaluation indexes used in this study are shown in Table 5.

Table 5 Landscape continuity indexes45,46,47.

The results of habitat patch importance vary with the selection of different indexes. In this study, importance indexes dIIC and Dpc (Table 5) corresponded to the integral index of continuity (IIC) and probability of continuity (PC), respectively, and were applied to evaluate the significance of each habitat patch in upholding landscape continuity. Using the hierarchical clustering module in SPSS 21.0, each index was classified into three classes. As a result, the importance of habitat patches and corridors in the ecological network in Changzhou was determined. Thus, we determined the importance of habitat patches and corridors in the ecological network in Changzhou. At least one of the indexes ranked as “very important”, with such habitat patches being classified as a “primary patch”. When both indexes were ranked as “ordinary”, the habitat patch was considered to be a “tertiary patch.” In all other cases, habitat patches were considered to be a “secondary patch”. The same method was used to classify potential corridors.

Results and Discussion

Landscape resistance threshold recognition

Previous studies have made the assumption that ecological connection becomes effective when a potential corridor between habitat patches has a cumulative resistance that is smaller than the threshold48,49. Therefore, the threshold represents the highest cumulative resistance that limits the migration of species between two patches. The methods used to identify the threshold value can be divided into two parts. The first part tracks the maximum moving distance of the target species49. The second part uses landscape continuity indexes and statistical principles, catching the breaking point or stable value of the calculated index values at different thresholds48.

The second method was expected to be more suitable for the present study. All links between habitat patches were calculated using Linkage Mapper50 and Conefor Sensinode 2.6. Supporting previous studies, we found that calculating the threshold distance based on the highest number of landscape components (NC) is credible48. Thus, IIC and NC were used in this study. Two indexes were used to indicate connectivity between habitat patches. During this process, the values of IIC and NC can be calculated from a given hypothetical threshold. With a step defined as a landscape resistance value of 10000, the hypothetical threshold ranges from 0 to 400000, leading to variation in the index value. To compare the two indexes, the results of the calculation were normalized based on equation (3). For NC, a low value indicates good connectivity, while a high value indicates good connectivity for IIC. Thus, the green curve in Fig. 4 indicates the sum of these two indexes as calculated by:

Figure 4: Variations in the integral index of continuity (IIC), and number of landscape components (NC) based on landscape cumulative resistance.
figure 4

The left vertical axis represents the normalized results of IIC and NC, while the right vertical axis is the sum of normalized IIC and normalized NC.

where Sum is the sum value of these two indexes, NCnormalized is the normalized result of NC, and IICnormalized is the normalized result of IIC. Based on equation (7), we conclude that a high value of Sum indicates good connectivity.

The variation of IICnormalized and NCnormalized showed that, when landscape resistance reaches 340000, IIC no longer increases rapidly, but becomes steady. In comparison, NC no longer varies, with the sum of these two indexes producing a steady curve (Fig. 4). Therefore, a landscape resistance threshold of 340000 was confirmed to be applicable to Changzhou, with any values exceeding this threshold signifying disconnection.

The probability of continuity (PC), which is used to assess the importance of habitat patches, indicates the probability of a species migrating between habitat patches. If the value calculated by the PC is relatively high (nearly 1), habitat patches might be connected by the potential corridors. To compare IIC with PC, a coherent setting for IIC and PC is needed, which is calculated by a decreasing negative exponential function of distance51. In this study, a PC exceeding the threshold must be small, and was therefore set as 0.05.

Ecological network construction

An ecological network is the framework for environmental protection in a city. The identification of important habitat patches9 and the functional connection between them5,10 is critical for constructing ecological networks. The resistance surface covering the entire study area, together with the least-cost paths between pairs of habitat patches, is shown in Fig. 5. An ecological network consists of habitat patches and potential corridors. Spatially, habitat patches are composed of landscape patches along the path of minimum resistance52. The potential corridors are strips connecting habitat patches, of which the accumulated resistance is lowest between habitat patches of all potential connections. However, the width of corridors could vary in different environments27. In this study, we assumed that a corridor width of 30 m could accelerate migration, which represented the cell size of the resistance surface.

Figure 5
figure 5

Landscape resistance estimates based on (a) the habitat quality-based model, (b) the expert scoring method, and (c) the entropy coefficient method. Results of the potential corridor simulations, importance assessment of habitat patches, and potential corridors based on different models of resistance surfaces are presented according to (d) the habitat quality-based model (e) the entropy coefficient method, and (f) the expert scoring method. The legend corresponding to panels a–f is located at the bottom, where a yellow to blue scale indicates low to high resistance, respectively; the three types of patches are indicated by three shades of green; and the three types of corridors are indicated by light orange, orange, and red lines. (g) Principle of the importance evaluation of habitat patches. (Created by ArcMap, version 10.2,

Effects of landscape resistance surfaces

To compare the landscape resistance surfaces that were constructed, the average habitat quality of different landscapes was normalized, as illustrated in Table 6. The estimated resistance was smaller for forests, farmlands, and riparian green spaces, which are less disturbed, highly vegetated, and more favorable for animal migration. In contrast, clear obstructions to animal migration were identified for roads and built-up land. Nevertheless, as a result of rapid urban expansion and the degeneration of the habitat quality of farmlands in Changzhou, the habitat quality-based resistance estimate for farmlands was higher than the estimates obtained by the other two methods. The results showed that while different resistance surfaces led to significant spatial offsets in the simulation of potential corridors, these results demonstrated high consistency in the identification of important habitat patches and corridors (Fig. 5). Concerning assessment units, we found that the model constructed in our study is more sophisticated than the other two models, and is more suitable for research on ecological networks in fragmented landscapes1,2.

Table 6 Normalized results of landscape resistance and the composition of corridors based on different resistance surfaces.

In addition, we analyzed the composition of potential corridors, which also revealed the effect of landscape resistance surfaces. We found that the target species preferred to move in vegetative habitats and riparian green spaces. However, different landscape resistance surfaces generated different corridor compositions. For instance, the corridors that were simulated from resistance surfaces constructed by the habitat quality-based model (Table 6) consisted of 3.45% built-up land. This percentage was lower than those obtained by the expert scoring method and the entropy coefficient method (6.27% and 16.04%, respectively). As the main potential corridors are riparian green spaces, followed by forest and farmland, the results indicate that the resistance surfaces constructed by the habitat quality-based model are more suitable for ecological network analyses in Changzhou. Hence, under conditions with insufficiently detailed ecological data, or when it is difficult to obtain such data through experiments, it is more feasible to carry out ecological network analyses using the habitat quality-based model and the minimum cumulative resistance (MCR) model, rather than other methods, such as mark and capture53 and wireless wildlife tracking5.

Importance of habitat patches

Once the ecological corridors are defined, the assessment of habitat patch importance is fundamental for optimizing the ecological network to construct and conserve critical patches and corridors27. In Changzhou, the Ge Lake and Changdang Lake Wetland Conservation Areas ranked high when evaluating the importance of habitat patches (Fig. 5), as these areas contain extensive habitat patches and are well connected to natural landscapes in the central and southern parts of Changzhou. Wawushan Provincial Park and the Xinlong Ecological Forest belong to linear habitat patches and form natural corridors (Fig. 5). These two sites reduce landscape resistance against animal migration, and are essential for enhancing the continuity of the ecological network. Nature Reserves in the central and southern parts of Changzhou, such as the Qianzi Lake Wetland Conservation Area, Yancheng Park, the Songjian Lake Wetland Park, and the Shahe Reservoir Restoration Area, were smaller than the Ge Lake and Taihu Lake Conservation Areas. However, these areas contain many potential corridors that could serve as excellent stepping-stone patches, which are advantageous to biological dispersal between habitat patches in regions with intensive human activity54. The inclusion of stepping-stone patches reduced the threats to biodiversity owing to isolated habitat patches, and also reduced resistance against animal migration. The results suggested that fragmented patches rarely provide favorable habitats for organisms to survive and reproduce, but clearly facilitate animal migration30.

Importance of corridors

At present, the IIC of the ecological network in the study area is relatively low. The growth of IIC and resistance shows a logistic function (Fig. 4). When the resistance threshold was set as 340000, IIC became stable, with a value of just 0.007, signifying little potential for any connection between habitat patches. Among the potential corridors studied, the most important ecological connections were located between Taihu Lake and Ge Lake and between Ge Lake and the Changdang Lake Wetland Conservation Areas. The connections between Changdang Lake and the Qianzi Lake Conservation Areas and between Changdang Lake and the Liyang West Park were also of high importance. These four potential corridors combined with habitat patches to form a fundamental T-shaped habitat structure in Changzhou. This result is consistent with previous studies, which showed that rivers and coastal areas could be used as a means for ecological processes to enhance the connectivity of ecological networks and improve urban habitats in regions where ecological land is limited5.


This study suggested the use of model simulations to construct ecological networks. An ecological network was constructed for the city by integrating habitat quality, graph theory, and the minimum cumulative resistance model. Our results verify the reliability of using this method in ecological network research. It was concluded that (1) the proposed landscape resistance surface method provides a way of effectively overcoming research limitations caused by insufficient experimental data; (2) the use of different methods in constructing resistance surfaces considerably affects the delineation of potential corridors, but only slightly influences the evaluation of the importance of habitat patches and potential corridors; (3) the suggested procedure is reliable because the potential corridors are mostly composed of green spaces rather than built-up land; and (4) habitat patches of high importance and good quality should be prioritized in regions with limited green space. However, this study did not consider how corridor width and complexity affect the migration of different species. To promote regional sustainable development, optimal corridor widths for species migration and multi-scale compound ecological networks should be determined by future studies.

Additional Information

How to cite this article: Gao, Y. et al. Constructing Ecological Networks Based on Habitat Quality Assessment: A Case Study of Changzhou, China. Sci. Rep. 7, 46073; doi: 10.1038/srep46073 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.