Study on the Extraction Method for Ecological Corridors under the Cumulative Effect of Road Traffic

Abstract

Research on ecological corridor extraction methods has made some progress and has been gradually applied to the planning and construction of regional ecological corridors, which play a role in biodiversity conservation efforts. However, the factors affecting species migration in ecological environments are very complex, especially anthropogenic disturbances, typically including noise pollution. Their effects on species habitats, reproduction, predation, and other activities are currently underestimated. In this paper, we propose an algorithm for superposition analysis of multiple road impacts and construct an ecological corridor extraction method that considers landscape pattern, habitat quality, remote sensing ecology, and road traffic resistance to address the shortcomings of current ecological corridor extraction methods that underestimate the potential impacts of road traffic. An extraction of ecological corridors was completed in Wuhan, and a quantitative comparative analysis was conducted from multiple perspectives. The results show that the improved method was effective, with the proportion of ecological corridors not re-identified due to road traffic impacts being 0.45% and the proportion of ecological corridors with significant changes in spatial location, represented by regions far from roads or high road network density, being 22.15% in the whole of Wuhan.

1. Introduction

“Ecological corridor” is a term in landscape ecology that refers to a type of ecosystem that is laid out in a linear or ribbon-like fashion in an ecological setting, capable of connecting spatially isolated and dispersed ecological units [1]. Ecological corridors can prevent soil erosion, increase carbon storage, and promote biodiversity conservation and ecosystem stability [2]. Within a certain period of time, humans have carried out a series of development activities on the earth to create a better life, which have disturbed the regional ecological environment to different degrees, directly leading to the fragmentation of ecological sources and the fracture of ecological corridors.

Urbanized areas, agricultural lands, and roads are the three most common anthropogenic features that influence the intensity of human footprints in ecological regions [3]. Anthropogenic disturbances have been pointed out as among the main causes of the world biodiversity crisis [4]. In terms of future world population trends, urbanization is irreversible. Urban expansion and road construction directly lead to fragmentation of many wildlife habitats, scattered in different locations in cities (e.g., new roads into forested landscapes often lead to economic development, as well as deforestation and habitat fragmentation). There are indications that some species are less active during the day and avoid more developed areas [5]. On the one hand, these disturbances directly alter the ground cover; e.g., the construction of roads inevitably alters the original land cover, produces edge effects, and exacerbates landscape fragmentation [6]. On the other hand, they indirectly affect biological habitats in the form of noise, artificial light, and dust, and they ultimately have far-reaching effects on the whole ecosystem. For example, in general, some species seem to abandon the area when noise occurs and return once the noise dissipates. Of all these disturbances, urbanization and road construction cause the fastest and most radical changes to regional ecosystems, with extremely high costs of recovery at a later stage. In particular, the presence of a large number of hardened surfaces has, to a certain extent, hindered the exchange of plants and animals across the hardened surfaces and isolated the material and energy exchange between the surface and the subsurface.

Based on the data on land cover, land use, and topography, a series of methods of ecological corridor extraction have been studied, with relatively good results. For example, many scholars in landscape ecology have tried to construct ecological networks at different spatial scales. The minimum cumulative resistance (MCR) model is a practical method for identifying the most efficient movement paths in landscape ecology [7,8]. The study [9] combined MCR with DO indices (Duranton and Overman indices) to construct an ecological safety network for the urban cluster around Poyang Lake in China. Different from the MCR model, which represents ecological corridors by finding the path of least cumulative resistance, the study [10] proposed the circuit theory, which considers that species have multiple alternative paths in the migration process and do not necessarily choose the optimal path. It considers the landscape surface as a conducting surface and uses the random movement of electrons in a circuit to simulate the migration process of species while considering all possible paths in the landscape [11]. It distinguishes ecological sources and ecological corridors according to their importance to the ecological network. As a key aspect of ecological corridor construction, ecological source identification has often been performed by subjective methods in previous studies, and new methods are needed to objectively identify ecological sources. Morphological spatial pattern analysis (MSPA) can objectively and accurately identify ecological sources by relying only on land cover data, and this method has been widely used in ecological source identification [12,13] to improve the rationality of ecological source selection. In constructing ecological corridors, the magnitude of spatial resistance to be overcome to connect the target landscape is generally calculated by establishing a resistance surface, and many studies have used land use as the basis for resistance computation. However, even for the same land cover class, ecological resistance differences still exist. In the study [14], influencing factors such as nighttime light intensity and altitude were identified according to the characteristics of the study area, so that the results could better reflect the ecological resistance differences in the study area.

The construction of a road completely changes the status of ground cover on the one hand, and on the other hand, with the opening and uninterrupted use of the road, noise pollution, light pollution, etc., will continue to be generated, all of which will have a huge impact on the surrounding environment. At the landscape scale, the main ecological impacts of road networks are the disruption of landscape processes and loss of biodiversity. Road networks interrupt horizontal ecological flows and alter the spatial patterns of the landscape, thereby suppressing important interior species [15]. Meanwhile, visual and auditory systems mediate basic behaviors, including foraging, predator avoidance, territorial defense, and mating decisions [16]. Many animal species, especially birds, rely heavily on acoustic signals for intraspecific communication [17]. Anthropogenic noise and artificial lighting are sensory pollutants that have been increasing in recent decades, and they constitute a global environmental challenge to terrestrial and aquatic environments [18]. There is growing evidence that noise and nighttime lighting have strong ecological consequences [19]. Noise, characterized by a high-amplitude and low-frequency spectrum, is typical of habitats in and around human-altered landscapes [20,21]. Studies have shown that the species richness and occurrence of birds are significantly reduced and the species composition of birds is significantly altered at sites near highways [22]. Noise pollution causes cognitive impairment, distraction, stress, and behavioral and physiological changes that directly affect wildlife and humans [23]. A large body of literature studying the ecological effects of roads has repeatedly identified noise as a causal factor [24]. Of all the human disturbance impacts, noise pollution is a potentially underestimated threat that is expected to increase with urban expansion in the coming decades. It also has the potential to drive local extinctions and possible extinctions, as evidenced by declines in diversity and avoidance of noisy areas [25,26]. In addition, noise pollution that alters the distribution or behavior of key species can have cascading effects on ecosystem integrity [27].

Birds are commonly used to assess environment–biodiversity relationships in urban and natural areas [28]. We analyzed and studied the optimal extraction of regional ecological corridors using birds as target species. In the case of birds, factors that directly contribute to population decline include habitat loss and degradation, as well as mortality due to motor vehicle collisions. Artificial light pollution from streetlights and car headlights at night may also influence nest-site selection of breeding birds, prompting them to expand their daytime range [29]. Due to birds’ reliance on acoustic communication, noise may disrupt their acoustic communication, interfere with the detection of warning signals, and increase stress levels [20]. Males of territorial species use song to inform females of their breeding status [30] and provide them with indicators of their quality as potential mates [31,32]. Two research groups [33,34,35] have presented evidence that anthropogenic noise, as a major source of long-distance effects on bird densities (affecting distances from 120 to 1200 m), may be an important factor driving bird species away from cities or roads. In addition, vehicles often hit vertebrates attracted by spilled grain, roadside plants, insects, basking animals, small mammals, road salt, or dead animals [36,37]. In the United States, an estimated one million vertebrates are killed on roads every day.

Currently, roads are widely distributed over the Earth’s land surface, covering a staggering number of miles and aiding the movement of countless vehicles. However, in the study of ecological corridor extraction and analysis, we pay insufficient attention to the noise impact of road traffic and ignore the superposition effect of noise propagation. In this study, facing the long-distance propagation characteristics of sound and its significant ecological interference ability, we consider the difference in regional road traffic noise interference under different road network densities, introduce noise superposition effect analysis, and improve the traditional ecological corridor extraction and analysis methods. We hope to objectively respond to the effects of different distances to roads and different road network densities on the extraction results for potential ecological corridors through methodological improvements.

2. Method

The traditional method of ecological corridor analysis is extended by emphasizing the continuous impact of road traffic while integrating the use of landscape connectivity and regional ecological quality. Firstly, the MSPA model is used to identify landscape types, and then the Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model is used to calculate the habitat quality index. The principal component analysis method is used to calculate the remote sensing ecological index. The road traffic resistance surface is constructed according to the sound propagation and attenuation law. Secondly, using hierarchical analysis, the respective weights of habitat quality, remote sensing ecology, and road traffic resistance in relation to ecological corridors are quantified to form a combined resistance surface. Finally, the ecological sources extracted by the integrated resistance surface and MSPA model are imported, and the ecological corridor extraction is completed by using the MCR model. The specific research method and process are shown in Figure 1.

2.1. Ecological Landscape Pattern Analysis

In the process of ecological landscape pattern analysis, two landscapes with high ecological value and less susceptible to human disturbance in the MSPA method, i.e., woodland and wetland, were used as foreground, and other land use classes were designated as background to reclassify the land use data. The reclassified image data were imported into GuidosToolbox software for analysis, and seven landscape categories that do not overlap with each other were obtained. They are core, isolated island, loop, bridge, branch, pore, and edge.

Ecological sources are the main ecological sites that have good ecological quality and can ensure the ecological safety of the region [38]. With this hypothesis, ecological sources are identified through the MSPA model, a customized sequence of mathematical morphological operators that analyzes the spatial morphological patterns of the landscape, with the aim of describing the geometry and connectivity of patches [39]. Among the seven landscape categories, core and isolated-island areas are ecological sources or have the potential to become ecological sources with the highest connectivity weighting, while loops, connecting bridges, and branches can be considered as potential ecological corridors. Other landscape categories are less connected. Core areas are the larger and better-connected habitat patches among the foreground elements in MSPA analysis, which can provide larger habitats for species and are important for biodiversity conservation. They are generally used as the candidate set of ecological sources in the study area. Among the identified core areas, ecological sources can be screened according to two points: (1) ecological sources being too densely distributed will reduce the comprehensive benefits of ecological networks, so try to make them distributed in different areas when screening ecological sources; (2) ecological sources are highly fragmented and more susceptible to external environmental influences, so give priority to core areas with larger areas as ecological sources.

2.2. Resistance Surface Construction

Ecological corridors are an important part of a complete ecosystem, helping to maintain biodiversity and providing protection for biological migration and habitat. “Resistance surface” refers to the spatial distribution of the ease of species migration in the landscape. In the natural environment, there are many factors that influence animal migration, and the degree of influence varies from one factor to another. Due to immature technical methods, high cost of information acquisition, and unclear influence mechanisms, only approximate simulations of resistance surfaces can be produced at present. Of course, with the development of science and technology, respecting the laws of nature, more comprehensive consideration, finer quantification, and other initiatives will ensure that the simulated resistance surface is constantly close to the real situation.

2.2.1. Estimation of Ecological Resistance

In the approach of our research design, two main factors, ecological quality and landscape pattern, are considered when constructing the ecological resistance surface.

The migration resistance of species between different patches is mainly influenced by natural and anthropogenic disturbances. Areas with high habitat quality tend to have better ecological conditions, which usually represent less disturbance from outside, and such areas play an important role in conserving biodiversity and maintaining the stability and integrity of ecosystems. The remote sensing ecological index (RSEI) can be calculated from remotely sensed imagery, which focuses more on the actual ecological condition of the surface of a single pixel area. It can reflect the possible differences between the same feature classes while showing the ecological status of the study area, thus compensating for the lack of land use data. The combination of remote sensing ecological index and habitat quality index can reflect the regional ecological resistance status more comprehensively. In addition, to maximize the use of existing ecological patch linkages, the landscape pattern is calculated by the MSPA model and incorporated into the resistance surface calculation.

(1) Habitat quality index calculation.

The InVEST model is a series of tools used to assess natural services. Using the habitat quality calculation tools in the InVEST model, habitat quality can be quantified by measuring the sensitivity of landscape classes and the intensity of external threats. In addition, the model assumes that if the habitat quality is good, then the biodiversity of the area is high. Habitat quality scores range from 0~1, and the higher the regional habitat quality value, the higher the score.

The habitat quality $Q_{xj}$ is calculated as follows.

$$i_{rxy}=1-\left(\frac{d_{xy}}{d_{r\mathrm{m}\mathrm{a}\mathrm{x}}}\right)$$

(1)

where $i_{rxy}$ is the impact of the threat source on each raster in the ecoregion, $d_{xy}$ is the distance between raster x and threat source y in the ecoregion, and $d_{r\mathrm{m}\mathrm{a}\mathrm{x}}$ is the impact range of threat source r.

$$D_{xj}={\sum }_{r=1}^R{\sum }_{y=1}^{Y_r}(\frac{{\omega }_r}{{\sum }_{r=1}^R{\omega }_r})r_y{i}_{rxy}{\beta }_x{S}_{jr}$$

(2)

where $D_{xj}$ is the degree of habitat degradation of raster x at position j, r is the threat source in the ecoregion, y is the influence range of threat source r, ${\omega }_r$ is the weight of threat source r, $r_y$ is the intensity of threat source y, ${\beta }_x$ is the level of habitat resistance to disturbance, and $S_{jr}$ is the relative sensitivity of each habitat to different threat sources.

$$Q_{xj}=H_j\left[1-\left(\frac{D_{xj}^z}{D_{xj}^z+k^z}\right)\right]$$

(3)

where $H_j$ is the habitat suitability of land category j, k is the half-saturation constant, i.e., half of the degradation maximum, and z is the default parameter of the model.

(2) Remote sensing ecological index calculation.

The remote sensing ecological index calculation is entirely based on remote sensing technology and natural factors, and it is used for rapid monitoring and evaluation of regional ecological conditions, which can quantitatively evaluate and compare the quality of regional ecological environment. The index uses the principal component analysis technique and integrates four evaluation indexes: vegetation index, humidity, surface temperature, and building index, which represent four ecological elements of greenness, humidity, heat, and dryness, respectively [40].

$$RSEI=f\left(NDVI,Wet,LST,NDBSI\right)$$

(4)

where NDVI is the vegetation index, Wet is the humidity component, LST is the surface temperature, and NDBSI is the building and bare soil index. Humidity is obtained from the tassel cap conversion, surface temperature is obtained from the surface temperature product in Landsat8 L2 data product, dryness is obtained from the building index and bare soil index together, and the remote sensing ecological index calculation results of the study area are obtained by carrying out principal component analysis and normalization on the four components.

2.2.2. Road Traffic Resistance Estimation

Road traffic hinders the communication of reptiles on both sides due to the constant passage of vehicles, on the one hand, and causing animals to be reluctant to approach due to noise and dust on the other hand. With the increase in road network density, the impact of noise and dust will accumulate and intensify, and the impact map is shown in the following figure.

As shown in Figure 2, a and b are two roads, and the thick boxes A (dashed line) and B (solid line) correspond to the impact areas of a and b, respectively. The figure contains the following elements:, (1) spatial distribution of road a, b; (2) impact region of road a; (3) impact region of road b; (4) superimposed impact region of roads a and b, divided into three regions (O, P, Q) from top to bottom, impact region (O) of road a, common impact region of roads a and b (the middle short vertical line covering the region P), and impact region (Q) of road b. The influence degree of region P is the cumulative amount of a and b influence values, and this calculation method is applicable to the spatial propagation of noise, dust, lighting, and other influence factors. For a single road, the closer to the road, the greater the impact, and then as the distance between roads gradually increases, the impact gradually decreases until it disappears. As an extreme example, if road a can only be passed during the day, road b can only be passed at night, and the sound of vehicles moving can only be heard in region O during the day and in region Q at night, while noise can be heard in region P throughout the day. This superimposed effect must be noted.

F is the impact value of the road, f is the impact value as a function of distance, and dis is the linear distance between the target point and the road.

$$F=f\left(dis\right)$$

(5)

From Formula (5), roads a and b’s respective impact functions can be obtained, as shown in Formulas (6) and (7).

$$F_a=f\left({dis}_a\right)$$

(6)

where $F_a$ is the influence value of road a and ${dis}_a$ is the linear distance between the target point and road a.

$$F_b=f\left({dis}_b\right)$$

(7)

where $F_b$ is the influence value of road b and ${dis}_b$ is the linear distance between the target point and road b.

In the range where both road a and b can be influenced (such as the above P region), their joint influence can be expressed as $F_{ab}$.

$$F_{ab}=f\left({dis}_a\right)+f\left({dis}_b\right)$$

(8)

Figure 3 shows a map of the spatial distribution of road traffic resistance in a part of Wuhan. The maximum impact distance of road traffic is set at 1000 m, and it can be seen that the denser the roads, the greater the resistance value. More typical is the vicinity of the road intersection, whose resistance value is significantly higher than other surrounding areas. In addition, the more bifurcations there are, the higher the resistance value will be.

2.2.3. Combined Resistance Estimation

The magnitude of the resistance value reflects the willingness or likelihood of a species to migrate in an area [41]. The resistance value is calculated in the form of raster. The larger the value, the more difficult it is for the species to migrate in the region, and conversely, the more suitable the raster is for animal migration. In this study, four indicators, namely landscape category, habitat quality, remote sensing ecology, and road traffic resistance, were selected to achieve a combined resistance analysis. The resistance values and weights were set as shown in

Table 1. Resistance Values and Weights.

Resistance Factor	Weight	Class	Resistance Value
Landscape (LA)	0.389	core	5
		isolated island	10
		bridge	20
		loop	30
		branch	40
		edge	50
		pore	70
		background	80
Habitat quality index (HQI)	0.167	/	$\frac{Q_{\mathrm{M}\mathrm{a}\mathrm{x}}-Q}{Q_{\mathrm{M}\mathrm{a}\mathrm{x}}-Q_{\mathrm{M}\mathrm{i}\mathrm{n}}}\times 100$
Remote sensing ecological index (REI)	0.167	/	$\frac{{RSEI}_{\mathrm{M}\mathrm{a}\mathrm{x}}-RSEI}{{RSEI}_{\mathrm{M}\mathrm{a}\mathrm{x}}-{RSEI}_{\mathrm{M}\mathrm{i}\mathrm{n}}}\times 100$
Road traffic resistance (RTR)	0.277	/	$\frac{F-F_{\mathrm{M}\mathrm{i}\mathrm{n}}}{F_{\mathrm{M}\mathrm{a}\mathrm{x}}-F_{\mathrm{M}\mathrm{i}\mathrm{n}}}\times 100$

. The resistance value of the landscape category was set with reference to the study [42]. The weights of four factors, namely landscape category, remote sensing ecology, habitat quality, and road traffic resistance, were determined by using hierarchical analysis.

Then, we can calculate the combined resistance CR determined by four factors: landscape, habitat quality, remote sensing ecology, and road traffic resistance.

$$CR=LA\times W_{la}+HQI\times W_{hqi}+REI\times W_{rei}+RTR\times W_{rtr}$$

(9)

where LA represents landscape resistance, HQI represents habitat quality resistance, REI represents remote sensing ecological resistance, RTR represents road traffic resistance, and $W_{la}$, $W_{hqi}$, $W_{rei},$ and $W_{rtr}$ correspond to the weights of the above four factors, respectively.

2.3. Minimum Resistance Path Extraction

In the approach of our research design, minimum resistance paths are extracted based on MCR, which is a GIS-based platform that simulates the minimum cumulative resistance paths between ecological source areas. Based on the assumption that the minimum resistance area has good ecological quality, our research uses Linkage Mapper software to connect the existing core areas and calculate the minimum cumulative resistance pathway as the best dispersal pathway for species migration [43]. The minimum cumulative resistance is calculated as follows.

$$MCR=f_{min}{\sum }_{j=n}^{i=m}\left(D_{ij}\times R_i\right)$$

(10)

where $MCR$ is the minimum cumulative resistance, $D_{ij}$ is the spatial distance from source point j to space unit i, $R_i$ is the resistance value of space unit i, and f is the positive correlation between minimum accumulative resistance and ecological process, reflecting the distance relationship from any point in space to all sources. Minimum resistance path can be approximated as an ecological corridor.

2.4. Data Collecting and Pre-Processing

The land cover data were obtained from GlobeLand30, the 30 m resolution global land cover data product, which was developed by China, and the latest version is GlobeLand30 2020. GlobeLand30 includes 10 classes of land cover: cultivated land, forest, grassland, shrubland, wetland, water bodies, tundra, artificial surfaces, bare land, and permanent snow and ice.

The remote sensing image data were obtained from Landsat8 L2 data product. Our research uses imagery from May 2020 to September 2020, and it completes the fusion process and removes the impacts of clouds and noise before use.

The road network data are obtained from OpenStreetMap, firstly extracting the high-grade roads of fclass as “primary”, “motorway”, or “trunk” and then processing the adjacent parallel multi-road in a single line and the multi-segment roads without bifurcation in a single segment. The processed results are used as the basic data for road traffic resistance analysis.

3. Case Study and Analysis

Our research includes the following two steps. Firstly, we completed the collection and pre-processing of basic data. Secondly, ecological corridors were extracted separately using the before-and-after optimization method, and separate preliminary analyses were conducted.

3.1. Study Area

Wuhan, which is in the province of Hubei, is selected as the study area. Located in the middle reaches of the Yangtze River, Wuhan is the capital of Hubei Province and a megacity in central China, as well as an important transportation hub in the central region. The Yangtze River and the Han River meet in Wuhan and divide it into three, forming the towns of Wuchang, Hankou, and Hanyang. Wuhan has a large population and an urbanization rate of 80.2%. According to different development directions and resource distribution, Wuhan is divided into one central district and six suburban districts. The suburbs include Huangpi District, East–West Lake District, Caidian District, Jiangxia District, Hannan District, and Xinzhou District. Due to the implementation of the “Rise of Central China” strategy, Wuhan is experiencing rapid urbanization, which has brought great pressure on the ecological landscape construction in Wuhan.

3.2. Ecological Corridor Extraction and Analysis

First, the GlobeLand30, Landsat8 L2, and OpenStreetMap road network data are analyzed and calculated using the method in Section 2.2 to obtain information on the landscape, habitat quality, remote sensing ecology, and road traffic resistance in Wuhan. Then, the combined resistance is calculated using the weights and resistance values in Table 1 and the method in Section 2.2.3. The calculation results of the combined resistance are detailed in Figure 4, where the resistance values gradually increase from blue to red. In the central city of Wuhan, the combined resistance is larger, and the further the distance away from the city center, the smaller the resistance value becomes. However, there is a general abrupt change in the resistance values, mainly due to the abrupt change from the artificial surface to the natural surface, and the ecological resistance values of these two surfaces differ greatly.

Using the method in Section 2.1 to analyze and calculate the GlobeLand30 data, we extracted the ecological source areas of Wuhan, as shown in the green areas in Figure 5. We can see that the ecological source areas are mainly distributed in the peripheral areas of Wuhan, especially in the northern and northeastern mountainous areas, and the ecological source areas consist of multiple plots with large areas and concentrated distribution. The ecological source areas in the central and southern areas are smaller in area and sporadically distributed.

According to Table 1, the weights of landscape, habitat quality, and remote sensing ecology were calculated as 0.538, 0.231, and 0.231, respectively, by using hierarchical analysis without considering the influence of road traffic, and then the combined resistance was calculated by using the method in Section 2.2. Finally, the ecological corridors were extracted by using the method in Section 2.3 without considering the influence of road traffic, as shown by the blue path in Figure 5. Each ecological source area is connected to at least one other ecological source area through one or more ecological corridors.

According to Table 1, the ecological corridors under the influence of road traffic were extracted using the method in Section 2.3, as detailed by the red paths in Figure 6. It can be seen that ecological corridors remain basically the same before and after the improvement of the extraction method, but there are still some obvious improvements.

3.3. Difference between Two Methods

In order to analyze the differences between the two results more intuitively, firstly, we superimposed the ecological corridors extracted by the two methods, so that some relatively subtle changes could be easily found. Then, the ecological corridors extracted by pre- and post-improvement methods were classified according to the specific change motives.

Comparing the results before and after the improvement of the extraction method, as shown in Figure 7, it can be found that differences in the results extracted by the improved method relative to the method before the improvement can be divided into three changes. The first is the disappearance of ecological corridors, such as ①–③. The second is the adjustment of ecological corridors, which can be interpreted as a change in the spatial location of the ecological corridor, such as ④–⑨. The third is that new ecological corridors appear, such as ⑩, but they do not lead to the disappearance of other ecological corridors,. All these changes in the ecological corridors extracted by the improved method are as far as possible from the roads, or from the dense areas of roads. Among them, the blue paths are the ecological corridors extracted with the pre-improved method. The red paths are the ecological corridors extracted by the improved method. The blue and red overlapping paths are the ecological corridors extracted by the pre- and post-improvement methods, which overlap or change very little, i.e., road traffic has no or negligible impact on them.

3.4. Statistical Analysis

According to the study results, classification statistics and comparative calculations are carried out to quantify and analyze the differences in extraction results between the pre- and post-improvement methods.

First, the ecological corridors extracted by the two methods were compared, analyzed, and categorized by four types: unchanged, new, disappeared, and location adjustment.

Second, the total lengths of ecological corridors before and after the improvement were measured, and the lengths of ecological corridors were measured separately by type of change, as shown in

Table 2. Statistics of ecological corridors extracted by two methods (Wuhan, China, 2020).

Class	Value
Length of corridors extracted by improved method	799.12 km
Length of corridors extracted by pre-improved method	842.67 km
Length of new corridors	28.59 km
Length of disappeared corridors	3.76 km
Length of changed corridors	186.64 km

.

Finally, the change proportion of ecological corridor length was calculated, as shown in

Table 3. Change ratio of ecological corridors extracted by two methods (Wuhan, China, 2020).

Class	Value
Change percentage of total length	5.17%
Length percentage of new corridors	3.58%
Length percentage of disappeared corridors	0.45%
Length percentage of changed corridors	22.15%

.

As shown in Table 2, the total length of ecological corridors extracted by the improved method decreased by 43.55 km compared with that before the improvement. There are two main reasons for this: one is that some potential ecological corridors were not re-identified, and the other is that some potential ecological corridors chose a shorter path through the resistance surface by adjusting their spatial location. Specifically, a new ecological corridor of 28.59 km was added, which should be just a coincidence, because an ecological corridor with the same function exists nearby. In practical application, the retention can be determined according to the needs of the project and some related constraints. A total of 3.76 km of ecological corridors were not re-identified by the improved method. A total of 186.64 km of ecological corridors showed significant changes in spatial location, mainly in terms of distance from regions with high road network density. These changes are the result of the influence of road traffic, and the absolute number of these changes reflects the actual effect of the improved extraction method.

As shown in Table 3, the length of the ecological corridors extracted by the improved method was shortened by 5.17% compared with that before the improvement. In total, 0.45% of the lengths of ecological corridors were not identified by the improved method, and 22.15% of the spatial locations of ecological corridors were adjusted to different degrees. The length of the new z-increased ecological corridors was 3.58% of the corridor length extracted by the improved h method. The statistical results in Table 2 show that the improved method is useful. On the other hand, the statistical results in Table 3 show that the biggest change lies in the fine-tuning of the spatial location of the ecological corridors to minimize the potential impact of road traffic. This improvement is an optimization of the previous method rather than a fundamental rejection.

4. Discussions

Road traffic has different degrees of influence on biological habitat and migration. The purpose of this study is to explore the extraction method for ecological corridors, considering the influence of road traffic. The extraction results of the study area also confirm the scientific validity and practicality of the improved method. As can be seen from Figure 5 and Figure 6, the number and spatial distribution of ecological corridors extracted with the two methods before and after the improvement are roughly the same, and each ecological source area has at least one ecological corridor connected to it. At the same time, some obvious differences can be found. Of course, due to the difficulty of data collection and other limitations, there are some shortcomings in the methodology designed for the study, and further in-depth research is needed in the future.

4.1. Analysis of Road Traffic Impact

Because of the effects of noise, dust, and other pollution, various types of organisms generally have the instinct to reside and migrate away from roads, and the ecological corridors extracted by pre- and post-improvement methods differ significantly in this regard.

In Figure 8, the blue paths near positions ①–③ are ecological corridors extracted with the pre-improvement method. When extracted using the new method, they are no longer identified as ecological corridors due to the influence of road traffic resistance. The other surrounding corridors do not change significantly but only show the disappearance of individual ecological corridors. Such changes are manifested in the abandonment of some potential ecological corridors due to road traffic pollution, which makes the extraction results more consistent with the natural patterns of biological habitats and migrations.

In Figure 9, the blue paths near positions ④–⑨ are ecological corridors extracted with the pre-improvement method. When extracted with the new method, they are no longer identified as ecological corridors due to the influence of road traffic resistance, but new ecological corridors are found in the vicinity. This change is mainly manifested in the fact that the new ecological corridors are located as far away from the roads as possible, the length of the ecological corridors is shortened, and the areas where the road network is sparser are preferred for crossing as opposed to the correspondingly disappearing ecological corridors. These changes can also effectively reduce the impact of road traffic on species migration along ecological corridors. Specifically, when extracted with the new method, a section of the ecological corridor near ④ that overlapped with the road disappeared. The new ecological corridor was oriented north–south and intersected the road almost perpendicularly, and the road traffic resistance was significantly reduced. The section of ecological corridor near ⑤, which overlaps with the road and is located in an area with higher road network density, often resulting in more frequent and higher levels of pollution (one of the important influencing factors in population decline, habitat loss, and degradation) due to the presence of pollution superposition effects, disappears when extracted with the new method. A new east–west ecological corridor was added due south, and the density of the road network in the area it traverses was significantly reduced. Relative to the original ecological corridor, the new ecological corridor near ⑥ exhibits an obvious vertical crossing with the road, a situation that can reduce the accumulated combined resistance to some extent. The new ecological corridor near ⑦ is significantly shorter, and the number of intersecting roads is reduced. The distance between the new ecological corridor near ⑧ and the roughly parallel roads increases significantly. The new ecological corridor near ⑨ is far from the high-road-network-density area. These changes objectively respond to the reality that both habitat and migration of organisms have to be as far away from road traffic pollution as possible.

In Figure 10, a long ecological corridor is emerging, which may be a coincidence. Due to the inclusion of the influence factor in road traffic, the resistance surface originally constructed based on the habitat quality index and remote sensing ecological index has changed to different degrees, leading to the emergence of two connected ecological corridors between two points with essentially equal cumulative resistance values when using the MCR method to extract ecological corridors. In the actual planning and construction process, if there are multiple ecological corridors with basically the same function in the extraction results, they can be further screened according to the constraints such as funding, and a final decision can be made to keep or delete some of them.

4.2. Limits of the Study

Road traffic has various important direct or indirect impacts on the ecological environment, especially animal migration. Currently, a considerable number of green belts are built on both sides of roads, especially on highways. Therefore, the traditional extraction method tends to consider the green belts on both sides as part of the ecological corridor. However, we still need to face the negative impacts of road traffic and cannot directly consider the green belts on both sides as the most effective ecological corridors. This requires that road traffic impacts be included in the study of ecological corridor extraction methods to promote scientific ecological corridor planning and protection.

In response to the existing research on ecological corridor extraction methods that underestimate the potential impacts of road traffic, the approach of our research design considers the superimposed impacts of road traffic, typically represented by noise pollution, and determines their impact weights using hierarchical analysis. The ecological corridor extraction model was extended and experimentally verified in Wuhan, and the improvement effect was obvious. At the same time, there are some shortcomings in our research. Firstly, GlobalLand30 data are based on the surface coverage extracted from 30 m resolution remote sensing images, which is coarse in granularity; only has a large artificial surface category of urban built-up areas, with low classification accuracy; and has some influence on corridor extraction. Secondly, considering that the potential impact of high-grade roads is more important, this study only uses roads with fclass as “primary”, “motorway”, and “trunk” in the OpenStreetMap road network, and the same 1000 m impact distance is used. Thirdly, the potential pollution impact of road traffic is closely related to the brand and speed of passing vehicles, actual traffic flow, and shaded houses and trees on both sides. For example, high traffic volumes often mean constant, more severe noise pollution, which will have a greater impact on the habitat and migration of birds. However, because these data are not available, they are not considered in this research. Fourthly, the target species studied in our research are birds, and the species diversity of birds is ignored. Bird species’ tolerance to road traffic noise may vary very greatly, so the ecological corridors were extracted in a way that weakened the ecological corridors crossing the roads. Fifthly, water sources are indispensable for all organisms’ habitats, but the distance dependence of different organisms on water sources (e.g., rivers) varies, and this study only indirectly considers them in the remote sensing ecological index through humidity. In the future, it is still necessary to collect finer data as comprehensively as possible according to the actual situation, to include the whole road network, to consider the traffic flow and the spatial distribution of shelters, to construct resistance functions taking into account specific groups of birds along with their preferences, to set different influence distances and weights, to conduct finer ecological corridor extraction research and experimental verification, and to achieve a more scientific and accurate completion of ecological corridor extraction.