Introduction

The increasing human activity is threatening global biodiversity and has a detrimental impact on the stability of natural ecosystems (Luck 2007; Pereira et al. 2010; Crist et al. 2017). Construction land, which encompasses areas designated for buildings and structures, is one of the principal human-made sources of threat (Tang et al. 2015; Zhou et al. 2022). The expansion of construction land is an important driving factor for reshaping the regional landscape pattern (McKinney 2006; Huang et al. 2018), which disrupts multiple ecological processes across the entire landscape by converting natural habitat into impervious surfaces, and thereby affecting ecosystem services provision (Peng et al. 2021) and interfering with human well-being (Wu 2013, 2021). The rapid expansion of construction land, occurs not only in urban but also in rural areas. This raises concerns regarding its implications for landscape sustainability across the urban-rural continuum (Bohnet and Pert 2010; Wu 2017; Martellozzo et al. 2018).

The expansion of construction land directly and indirectly impacts landscape composition and configuration (Connors et al. 2013; van Vliet 2019). The direct impacts can be attributed to the occupation of construction land that reduces habitat area and biomass (He et al. 2014). The indirect impact can be more diverse, involving complex landscape patterns and ecological processes. Previous studies highlighted two mechanisms that can be attributed to the expansion of construction land: habitat fragmentation and habitat isolation. Habitat fragmentation refers to the splitting of integrated habitat areas into smaller patches due to the expansion of human activity, increasing the edge length and making the core habitat more susceptible to disturbance (Haddad et al. 2015; Li et al. 2022). This process essentially reduces the available living habitat for umbrella species and poses a threat to regional biodiversity (Newman et al. 2013). Habitat isolation occurs when the expansion of construction land reduces connectivity between different habitats (Perkl et al. 2018). A reduction of connectivity among habitat patches is often detrimental to species dispersal, nutrient cycling, and gene flow (Wilk et al. 2020). In cases where construction patches encroach upon critical pathways and key nodes essential for ecological processes, habitat isolation becomes a valid concern, increasing the vulnerability to land degradation. (Cerqueira et al. 2021).

When contrasting the direct impact of construction land on natural habitats with its indirect impacts, it becomes evident that the latter poses a significant threat to biodiversity conservation. (Ke et al. 2018; McDonald et al. 2020; Yang et al. 2020; Ren et al. 2022). This is attributed to the following two reasons. Firstly, vital natural habitats are protected by governmental regulations, making them less susceptible to direct occupation by construction land (Hu and Zhang 2020; He and Shi 2022). However, these protection policies inadvertently result in the increased susceptibility of smaller habitat patches to occupation, thereby impacting landscape connectivity and exacerbating the indirect influence of construction on natural habitats (Jiang et al. 2019). Secondly, the surrounding region of natural habitats is more attractive for development, rendering these areas more vulnerable to occupation by construction activities (Hansen and DeFries 2007). This issue is particularly pronounced in fast-urbanizing and industrializing regions, where the extensive expansion of construction land amplifies the pressure on natural habitats.

Recent studies evaluated the indirect impact of construction land on natural habitats from the perspective of landscape fragmentation. However, a reliable quantification approach for assessing habitat isolation at a large scale is currently lacking. The insulation model has been used to evaluate the isolation effect of urban expansion on natural reserve zones, with the Western Taihu Lake watershed in China as a case study, (Su et al. 2010). This model is grounded in landscape ecology principles (Forman 1995a), which proposes considering alterations in the spatial relationships between urban growth and natural habitats resulting in the isolation of habitat patches. The model considers two factors: the magnitude of urban growth and the Euclidean distance to habitat patches, emphasizing that both a larger size and a shorter distance contribute to a more pronounced isolation degree. Subsequent studies improved the model and applied it to quantify the isolation effect in fast-urbanizing regions. For instance, Ng et al. (2011) used the insulation model in a case study of the Shenzhen River catchment in China to quantify habitat isolation, encompassing additional factors such as the intensity of human activity and the loss of ecological value. Xun et al. (2014) developed an Urban Isolation Index, also based on the insulation model, which incorporated factors of urban expansion type, patch morphology, and habitat quality. Both studies demonstrated the adaptability of the insulation model in evaluating the isolation effect stemming from urban expansion. Meanwhile, other related studies characterized habitat isolation through buffer analysis and landscape indices (Seiferling et al. 2012; Liu et al. 2016; Rodríguez-Rodríguez and Martínez-Vega 2019). However, these studies primarily centered on the distance decay effect that highlights the land use/cover change surrounding habitat patches, while overlooking whether the construction patches impede ecological flows between different habitats. This aspect is pivotal to consider for a comprehensive understanding of habitat isolation.

The essence of habitat isolation caused by construction land expansion is that the change in landscape pattern driven by socioeconomic development interferes with the ecological flow between different habitats. This process involves the concept of an ecological network rooted in landscape connectivity, which is fundamental for constructing sustainable landscape patterns (Peng et al. 2021) and is also a research hotspot in the field of landscape ecology (Dong et al. 2022). The concept of an ecological network views the total landscape as a “patch-corridor-matrix” mosaic (Forman and Godron 1981). The patch signifies the habitat that served as a source of ecological flow, the corridor denotes the banded area connecting individual habitats, and the matrix is the landscape background that carries patches and corridors (Verbeylen et al. 2003). The identification approach of the ecological network, which can extract the key nodes and corridors responsible for maintaining ecological flows between different habitats, serves to address the limitations present in past studies concerning the quantification of habitat isolation.

As China continues to undergo rapid urbanization and industrialization processes, the expansion of construction land is also on the rise, thus affecting both urban and rural regions (Huang et al. 2015). The distinctive dual land system of China (Zhou et al. 2020), along with the ripple effect of socioeconomic development, has greatly contributed to this trend (Liu et al. 2015). According to land survey data published by the Ministry of Natural Resources of China, the extent of rural settlements is approximately twice that of urban areas, with rural settlements and urban areas expanded by 7,279 km2 and 21,820 km2 from 2009 to 2016, respectively (Chen et al. 2022). Under these circumstances, it becomes essential to consider the isolation effect induced by both urban areas and rural settlements. However, past studies primarily focused on the negative impacts of urban expansion on natural habitats (Dupras et al. 2016), while neglecting the corresponding impacts of rural settlements.

This study aims to address these research gaps by utilizing the ecological network approach to assess the habitat isolation resulting from construction land expansion and analyzing the differential contributions of urban areas and rural settlements. The Nanjing Metropolitan Area from 2000 to 2020 is taken as the case study region. To achieve these aims, we developed three research questions: (1) What methods can be employed to measure the degree of habitat isolation resulting from construction land? (2) To what degree do urban and rural construction areas contribute to habitat isolation? (3) How does the degree of habitat patch isolation vary across space and time? This study not only establishes appropriate indices for measuring habitat isolation but also holds practical significance for implementing sustainable landscape planning and governance.

Materials and methods

We developed a research framework (Fig. 1) composed of three steps. First, we extracted habitat patches by employing morphological spatial pattern analysis (MSPA), an image processing method based on mathematical morphological principles aimed at describing the integrity and connectivity of image components. Second, we identified ecological networks by constructing comprehensive resistance surfaces and applying circuit theory to obtain corridor extent. Third, two indices, namely the isolation effect index (IEI) and isolation degree (ID), were established. The IEI quantifies the isolation effect produced by each construction pixel, while the ID captures the degree of isolation suffered by each habitat patch. We further independently quantified the IEI for urban areas and rural settlements, followed by calculating the ID of each habitat patch.

Fig. 1
figure 1

Research framework of this study

Full size image

Study area

The Nanjing Metropolitan Area (NMA) is located in the central region of the Yangtze River Economic Belt (YREB) in China (Fig. 2a). Spanning the geographic coordinates of 117°09′E to 119°58′E longitude and 29°57′N to 34°06′N latitude, this region covers approximately 65,457 km2, encompassing portions of Jiangsu and Anhui provinces. This region consists of eight prefecture-level cities (Nanjing, Huaian, Yangzhou, Zhenjiang, Xuancheng, Wuhu, Maanshan, and Chuzhou) and two county-level cities (Jintan and Liyang). The elevation in this area ranges from 0 m to 1,740 m (Fig. 2b). The northern and central regions predominantly feature a flat terrain characterized by extensive croplands, while the southern reaches boast a mountainous landscape adorned with lush forests (Fig. 2c). The rural settlements are scattered throughout the study area, while the urban areas are mainly concentrated in the central zone.

The existing contradiction between socioeconomic development and ecological protection renders the selected study area suitable for our study objectives. From 2000 to 2020, construction land increased from 6,234.31 km2 to 9,067.17 km2, concomitant with a surge in GDP per capita from 10.62 thousand CNY to 123.98 thousand CNY. This expansion signifies not only a growth in the overall developmental area but also an escalation in development intensity. Although the area occupied by construction land remained small in relation to the overall forest and grassland area during this period, the rapid expansion and advancement of construction land bear the potential to disrupt ecological processes and fragment natural habitats. Moreover, this study also holds practical significance for future land-use planning and ecological protection within the NMA. This importance is accentuated by the unveiling of the Development Planning of the Nanjing Metropolitan Area in 2021, a document that emphasizes the collaborative governance essential for conserving biodiversity, particularly pertinent for habitats mandating cross-administrative protection (Liu et al. 2022).

Fig. 2
figure 2

Study area. a The location of the NMA in China. b The elevation characteristics of the study area. c Land use/cover status for 2000, 2010, and 2020 and their changes from 2000 to 2020

Full size image

Data sources

Land use/cover maps from 2000 to 2020 with a spatial resolution of 30 m×30 m were produced by the Resource and Environment Science and Data Center (Xu et al. 2018). The land use/cover was classified into seven categories: cropland, forest, grassland, waterbody, urban area, rural settlement, and bareland. The urban area and rural settlement, which refer to built-up land in urban and rural regions, together constitute the construction land. Additional datasets, including the Digital Elevation Model (DEM), Net Primary Productivity (NPP), nighttime lights, and road maps, were utilized to build comprehensive resistance surfaces for identifying ecological networks. The 30 m × 30 m DEM data were obtained from the Shuttle Radar Topography Mission (https://www.usgs.gov/centers/eros). The annual NPP data were collected from MOD17A3 (https://lpdaac.usgs.gov/products/mod13q1v006/), available via the Google Earth Engine (GEE) platform. The nighttime light data were obtained from global NPP-VIIRS-like nighttime light data (Chen et al. 2021). The road maps were extracted from the traffic network dataset provided by the National Geomatics Center of China (http://ngcc.sbsm.gov.cn/). Additionally, nighttime light data were also applied to construct the IEI and ID.

Extraction of habitat patches

According to the IUCN definition, habitats are characterized as land use/cover types dominated by non-artificial vegetation, including forests, grasslands, and waterbodies (He et al. 2014). In this study, we focused on terrestrial habitat patches (forests and grasslands) for two reasons. Firstly, terrestrial habitats exhibit a greater susceptibility to isolation effects compared to aquatic habitats. Generally, corridors between terrestrial habitats are more likely to be disturbed and interrupted by construction land, given their frequent occurrence within open spaces (Huang et al. 2021). Conversely, aquatic habitats maintain connections through waterways such as canals or rivers. The surrounding human activities are difficult to isolate and block hydrologic connectivity. Secondly, aquatic habitats represent relatively independent ecosystems, and their connections with other habitat types are ineffective since species colonization cannot be promoted (Xu et al. 2016).

To reduce the complexity of the extraction results, we further screened the key habitat patches endowed with core areas. This curation process was executed utilizing the MSPA method, which assesses the spatial topological relationships between target pixels and structural elements (Crouzeilles et al. 2013; Carlier and Moran 2019; An et al. 2021). Specifically, forests and grasslands were considered as the foreground, while cropland, waterbodies, urban areas, rural settlements, and barelands were treated as the background. As forests form the predominant component of ecological areas in our study, we set the edge width at 300 m (30 m×30 m, ten pixels on our map), a measure commonly defined and used by preexisting studies focusing on forest boundaries (Shapiro et al. 2016; Gatti et al. 2019). Furthermore, there is evidence indicating that the impact of fragmentation on biomass reduction can be detected up to this distance (Laurance et al. 1998; Albiero-Júnior et al. 2019).

We performed MSPA using Guidio 3.0 software, a platform for extracting information related to the geometry and connectivity of elements (Soille and Vogt 2009). The total landscape for each study year was classified into seven categories, including core, patch, pore, edge, bridge, ring, and branch (Vogt et al. 2007), as depicted in Fig. 3a. All of these categories carry significant ecological implications, with the core category being particularly acknowledged as the internal habitat which is important for species reproduction and ecosystem services (Ye et al. 2020). We thus treated habitat patches possessing core attributes as the subjects of analysis for isolation effects. These habitat patches contain rich biodiversity and represent key areas for maintaining regional ecological functions. Finally, we determined the final identification results by considering the spatial intersection of the habitat patches for each year. The results showed that 129 habitat patches were identified, covering an area of 14,510.48 km2 and accounting for 90-92% of the entire forest and grassland area (Fig. 3b).

Fig. 3
figure 3

Extraction results of habitat patches in the study area. a Results of MSPA for 2000, 2010, and 2020. b Spatial distribution of habitat patches

Full size image

Identification of ecological networks

Constructing comprehensive resistance surfaces

Construction land exerts isolation effects on habitat patches by disturbing the between-habitat ecological processes. We employed ecological corridors to depict the pathways through which ecological processes occur between paired habitat patches, which were identified through the ecological network approach. Firstly, we built resistance surfaces to characterize the impact of landscape heterogeneity on ecological flows. The resistance values were comprehensively assigned based on the resistive effect of natural and socioeconomic factors on interactions between habitat patches (Table 1). The natural factors include elevation degree (\({r}_{1}\)), relief amplitude (\({r}_{2}\)), and NPP (\({r}_{3}\)), which affect the dispersal of ecological function. The socioeconomic factors include the nighttime light index (\({r}_{4}\)), distance to roads (\({r}_{5}\)), and land use/cover type (\({r}_{6}\)), all of which reflect disturbances arising from human activities. The values of \({r}_{1}\sim{r}_{5}\) were reassigned from one to five based on the Jenks Natural Break method (Jenks 1967), whereas the values of \({r}_{6}\) were assigned based on expert knowledge concerning the intensity of human activity for different land use/cover types. The weights of these six factors were calculated using the Analytic Hierarchy Process (AHP), a method rooted in assessing the relative importance of different factors (Saaty 1988). All spatial data were resampled to a resolution of 500 m × 500 m, constituting the lowest resolution among the chosen factors within the utilized datasets. The equation for the resistance surfaces is shown as follows:

$$R_{i} = \sum\limits_{{k = 1}}^{6} {w_{k} \cdot r_{{i,k}} }$$
(1)

where \({R}_{i}\) is the resistance coefficient of pixel \(i\); \({r}_{i,k}\) is the resistance value of the \(k\)-th factor of pixel \(i\); and \({w}_{k}\) is the weight of the \(k\)-th factor.

Table 1 Resistance values of each factor
Full size table

Identifying ecological corridors

We then employed the Minimum Cumulative Resistance (MCR) model to identify the pathways of ecological corridors (Knaapen et al. 1992). Thereafter, circuit theory, which is based on the principles of the random walk of electrons within a certain circuit (McRae et al. 2008), was employed to identify the hotspot regions for ecological flows. To achieve this, we treated the habitat patches as electric circuit nodes and assigned the cumulative resistance as the electric resistance. For each pair of electric circuit nodes, one node was arbitrarily connected to a 1-amp current, while the other node was connected to the ground to iteratively estimate the effective resistances. Consequently, the spatial distribution of current density can be obtained, reflecting the migratory intensity of random walkers toward the destination node. A higher value of current density indicates a greater importance for ecological processes between habitat patches (Peng et al. 2018).

We used Circuitscape 4.0 software in conjunction with the Linkage Mapper Toolbox to perform these processes (McRae et al. 2013). A key parameter in the corridor identification process is the cumulative current threshold, which determines the width of corridors. In this study, we determined this parameter based on the amount of ecological area in the corridor extent. The corridor extent should encompass the ecological area adjacent to its pathway as extensively as feasible, warranting an ample presence of ecological stepping stones to facilitate ecological flows (Kong et al. 2021). We regarded the forest and grassland as ecological areas and tested various cumulative current thresholds ranging from 1,000 to 10,000 to determine the optimal width for ecological corridors. Figure 4 displays the ecological area in the corridor extent at different thresholds. We observed that the area of the corridor extent increased linearly with the threshold, while the ecological area in the corridor extent showed a logarithmic growth trend. The quantity of ecological area within the corridor extent displayed a marginal trend when the threshold value reached 6,000, suggesting that the majority of the ecological areas surrounding the corridor pathway had already been included in the corridor extent. Further enlarging the corridor width proved ineffectual as it led to an excessively large corridor extent, with the ecological area remaining same level. Furthermore, the ratio of the ecological area within the corridor extent surpassed 50% when the threshold was set to 6,000, guaranteeing a smooth flow of ecological processes. We thus chose 6,000 as the cumulative current threshold.

Fig. 4
figure 4

Identification of the suitable cumulative current threshold of ecological corridors. Note: Due to the similar features between different years, we only depict the identification results of the ecological corridor in 2000

Full size image

Quantification of the isolation effect index (IEI) and isolation degree (ID)

Quantifying the IEI of urban areas and rural settlements at the pixel level

The IEI was utilized to measure the isolation effect caused by urban areas and rural settlements at the pixel level. Three parameters were considered to manifest the sources of isolation. Firstly, the distance between the construction pixel and the habitat patch; the closer the distance, the higher the IEI. Secondly, the human activity intensity of the construction land, which exhibits a positive relationship with the IEI. These two parameters are general variables in the insulation degree model (Su et al. 2010; Ng et al. 2011; Xun et al. 2014), indicating that if the construction pixel is closer to a habitat patch and experiences a higher intensity of human activity, the disturbance to the habitat would be more significant. In addition to these two conventional factors, we further considered whether the construction land disrupts the ecological flow by incorporating a third parameter derived from the current density map generated by circuit theory. We used the current density within a certain buffer zone of the construction land pixel to reflect such disturbance. A higher value indicates that the construction pixel affects landscape connectivity more significantly. The IEI is a dimensionless and comprehensive metric, manifesting both the quantity and intensity of construction land expansion, as well as its impact on the ecological flow between habitat patches. The equation for the IEI is shown as follows:

$$\begin{array}{*{20}c} {IEI_{i} = \left( {\frac{{D_{i} }}{{\bar{D}}}} \right)^{{ - 1}} \times \left( {\frac{{NL_{i} }}{{\overline{{NL}} }}} \right) \times \frac{{C_{{i - buffer}} }}{{\bar{C}}}} \\ \end{array}$$
(2)

where \({IEI}_{i}\) is the isolation effect index of construction pixel \(i\); \({D}_{i}\) is the distance between construction pixel \(i\) and the nearest habitat patch; \(\stackrel{-}{D}\) is the mean nearest distance of all construction patches to all habitat patches in the study area; \({NL}_{i}\) is the time light intensity of construction pixel \(i\), while \(\stackrel{-}{NL}\) is the mean time light intensity of all construction pixels; and \({C}_{i-buffer}\) is the mean current density within the buffer area. We assigned the buffer area as 5 km based on previous studies that relate to the impact threshold of human activities on natural habitats (Skarin et al. 2015; Yang et al. 2019; Wu et al. 2020; Vasudev et al. 2021). \(\stackrel{-}{C}\) is the mean current density of the total buffer area. Each distance is defined as a core-to-core distance. Furthermore, we calculated the IEI separately for urban areas and rural settlements.

Quantifying the ID of each habitat patch

To evaluate the ID of each habitat patch, we considered the interrelationship between the isolation effect caused by construction pixels and the ecological networks, which involved three steps, as illustrated in Fig. 5.

Fig. 5
figure 5

The explanation of the calculation process. Note: A, B, and C are three habitat patches; the solid-green lines denote the corridor pathways; U denotes the urban pixel; R1 and R2 denote the rural settlement pixels; w is the weight of the corridor; the color light green means the corridor extent; the color grey indicates the construction pixels that do not have isolation effect; and the solid and dashed arrows denote the corridor and habitat patches impacted by construction land, respectively

Full size image
  1. (i)

    Establishment of spatial correspondence between habitat patches and construction pixels. We spatially associated construction pixels with the nearest ecological corridor. This enables us to determine the target habitat patches affected by construction pixels, which is based on the attributes of the paired habitat patches linked by the corresponding corridor. This step can be formulated as follows:

    $$\begin{array}{*{20}c} {\begin{array}{*{20}c} {\{ IEI_{{i,1}} ,i \in N_{1} \} } & \to & {\left( {HP_{k} ,HP_{{o_{1} }} } \right)} \\ {\{ IEI_{{i,2}} ,i \in N_{2} \} } & \to & {\left( {HP_{k} ,HP_{{o_{2} }} } \right)} \\ {\begin{array}{*{20}c} {\begin{array}{*{20}c} . \\ . \\ . \\ \end{array} } \\ {\{ IEI_{{i,s}} ,i \in N_{s} \} } \\ \end{array} } & {\begin{array}{*{20}c} {\begin{array}{*{20}c} . & . & . \\ \end{array} } \\ \begin{gathered} \hfill \\ \hfill \\ \to \hfill \\ \end{gathered} \\ \end{array} } & {\begin{array}{*{20}c} {\begin{array}{*{20}c} . \\ . \\ . \\ \end{array} } \\ {\left( {HP_{k} ,HP_{{o_{s} }} } \right)} \\ \end{array} } \\ \end{array} } \\ \end{array}$$
    (3)

assuming that a habitat patch \({HP}_{k}\) is linked by \(s\) corridors, while \({HP}_{{o}_{1}},{HP}_{{o}_{2}},\dots ,{HP}_{{o}_{s}}\) represent other habitat patches linked with \({HP}_{k}\); the \({HP}_{k}\) is affected by the construction pixel sets \({\{N}_{1},{N}_{2},\dots {N}_{s}\}\); and the \({\{IEI}_{i,s},i\in {N}_{s}\}\) represents a collection of IEIs that affect \({HP}_{k}\) through their association with the \(s\)-th corridor.

  1. (ii)

    Calculation of ecological corridor weights for each habitat patch. If a habitat patch was connected by a single corridor, the ID was entirely dependent on the construction pixels associated with that corridor. However, in the case of multiple connecting corridors, the IEI from construction land would be diluted based on the connectivity provided by each corridor. The greater the corridor connectivity is, the more pronounced the isolation effect of surrounding construction pixels on the habitat patch. Therefore, distinct weights were assigned to each corridor linked to a given habitat patch. We employed the gravity model to calculate these weights (Kong et al. 2010). The initial formula for the gravity model is expressed as follows:

    $$\begin{array}{*{20}c} {G = \frac{{N_{a} \cdot N_{b} }}{{D_{{ab}}^{2} }} = \frac{{L_{{max}}^{2} \cdot lnA_{a} \cdot lnA_{b} }}{{L_{{ab}}^{2} \cdot P_{a} \cdot P_{b} }}} \\ \end{array}$$
    (4)

where \(a\) and \(b\) are two habitat patches; \(G\) is the gravity value of the corridor that connects \(a\) and \(b\), reflecting the landscape connectivity between the two habitats; \({N}_{a}\) and \({N}_{b}\) denote the corresponding forces; \({D}_{ab}\) is the normalized resistance between \(a\) and \(b\); \({L}_{ab}\) denotes the cumulative resistance value of the corresponding corridor; and \({L}_{max}\) denotes the maximum cumulative resistance value of the corridors in the study region. \({P}_{a}\) and \({P}_{b}\) are the average self-resistance values of \(a\) and \(b\), while \({A}_{a}\) and \({A}_{b}\) are the corresponding patch areas. The larger the areas and the smaller the cumulative resistance, the higher the gravity value of the corridor.

The weights are calculated based on the following equation:

$$ \begin{array}{*{20}c} {\begin{array}{*{20}c} {\begin{array}{*{20}c} {w_{{k,j}} = } \\ \end{array} } \\ \end{array} } \\ \end{array} \frac{{G_{j} }}{{sum_{{Gk}} }} $$
(5)

where \({w}_{k,j}\) is the weight of the \(j\)-th corridor (\(j\le s\)) of the \({HP}_{k}{}_{}\); \({G}_{j}\) is the gravity value of the \(j\)-th corridor; and \({sum\_G}_{k}\) is the sum of gravity values of all the corridors linked with the \({HP}_{k}\). The sum of the \({w}_{k,j}\) equals 1.

  1. (iii)

    Quantification of the ID for each habitat patch. We conducted a weighted summation of the IEI values of construction pixels, producing a cumulative impact score for the corresponding habitat patch, as shown in the following equation:

    $$\begin{array}{*{20}c} {ID_{k} = \mathop \sum \limits_{{j = 1}}^{s} \left( {w_{{k,j}} \cdot \mathop \sum \limits_{{i = 1}}^{{n_{j} }} IEI_{{i,j}} } \right)} \\ \end{array}$$
    (6)

where \({ID}_{k{}_{}}\) denotes the isolation degree of \( HP_{k} \); \({w}_{k,j}\) is the weight of the \(j\)-th corridor expressed by Eq. (5). \({IEI}_{i,j}\) represents the isolation effect index of the \(i\)-th construction pixel associated with the \(j\)-th corridor, while \({n}_{j}\) is the number of construction pixels associated with the \(j\)-th corridor.

Results

Spatiotemporal characteristics of ecological corridors and construction land

We identified ecological corridors in this study for the years 2000, 2010, and 2020, as depicted in Fig. 6a. The results indicate that the spatial patterns of these corridors remained stable over time, with most pathways showing little change. These ecological corridors were evenly distributed across the study area, forming a network connecting various habitat patches. However, a reduction in the number of corridors was observed in the southeastern region, with two and three corridors lost between 2000 and 2010 and 2010–2020, respectively. Moreover, the total length of ecological corridors decreased over time, while the cumulative resistance value increased (Table 2). This suggests a growing disturbance within the ecological networks during the study period.

Table 2 Descriptive characteristics of ecological corridors
Full size table
Fig. 6
figure 6

Characteristics of the ecological corridors and construction land. a Identification of ecological corridors based on circuit theory. Note: The red circle denotes the major changing area of corridors. b The spatial relationship between ecological corridors and construction land

Full size image

At the same time, both the extent and intensity of urban areas and rural settlements experienced a significant increase from 2000 to 2020, as illustrated in Table 3. The urban areas expanded by more than threefold, while the area of rural settlements grew by 10.64% during the study period. The growth rate of the urban areas was higher in the first decade (2000–2010) compared to the second decade (2010–2020), while the rural settlements showed a consistent expansion process over this time frame. The development intensity of construction land, measured by the mean time light value, increased by 25.71% in urban areas and more than doubled in rural settlements from 2000 to 2020. These findings suggest an inverse relationship between the extent and intensity of urban areas and rural settlements. Approximately, 30–38% of the urban areas and nearly 25% of the rural settlements were located within the corridor extent over the twenty-year period, serving as the origin of the isolation effect (Fig. 6b). The mean time light value of construction land within the corridor extent was similar to that of the entire region, indicating that the intensity of human activities in construction land was not significantly affected by its location within or outside the corridor extent (Table 3).

Table 3 Descriptive characteristics of total construction land and construction land in corridor extent
Full size table

Isolation effects of construction land on habitat patches

We analyzed the distribution characteristics of the IEI values from 2000 to 2020, as depicted in Fig. 7a. The central region exhibited the highest IEI values, primarily concentrated in areas characterized by rapid urbanization. In contrast, lower IEI values were mainly associated with rural settlements and were mostly clustered around the habitat patches. We categorized the three-year ID of habitat patches into Level 1 to Level 5 based on the Jenks Natural Break Method, with higher levels indicating greater isolation. Habitat patches with high ID values were predominantly located in the central region, indicating that habitat patches in this area experienced more pronounced isolation compared to others. Habitat patches with low ID values were mainly distributed in the southern area, which is covered by large amounts of forests. Furthermore, we observed a rapid increase in both the IEI and ID during the study period, with the most substantial rise in the IEI occurring in the middle and southwest parts of the study area. While the majority of habitat patches were at Level 1 in 2000, a significant proportion of them had advanced to Level 3 by 2020. This indicates a substantial increase in the isolation effect of construction land on habitat patches, a trend observed not only in the central region but also in peripheral areas that are distant from the urbanization hotspot.

In Fig. 7b, we present the total IEI values of urban areas and rural settlements. We observed that the total IEI values increased approximately two times from 2000 to 2020, with contributions from both urban areas and rural settlements. In 2000, the cumulative IEI of rural settlements exceeded that of urban areas. By 2010, the total IEI caused by urban areas and rural settlements was roughly equal. In 2020, the IEI of urban areas surpassed that of rural settlements, indicating a faster increase in the isolation effect attributed to urban areas. Nonetheless, rural settlements still accounted for 40% of the total IEI value in 2020.

Figure 7c illustrates the descriptive characteristics of ID in each habitat patch and the corresponding contributions of urban areas and rural settlements. Significant differences were observed in the ID values of each habitat patch, with some patches had relatively low ID values while others exhibited extremely high ID values. This characteristic was more pronounced in the ID attributed to urban areas. Moreover, the ID values attributed to both urban areas and rural settlements displayed an upward trend from 2000 to 2020, with a more significant increase in urban areas. In 2000, the mean value of ID contributed by rural settlements was higher than that of urban areas. However, by 2010, the mean value of ID contributed by urban areas became equal to that of rural settlements and was higher in 2020. Figure 7d further confirms that the ID of most habitat patches was primarily attributed to rural settlements rather than urban areas. Nevertheless, the primary contributor of ID in an increasing number of habitat patches shifted towards urban areas. In 2000, only 15 habitat patches had ID primarily caused by urban areas, but this number increased to 33 in 2020.

Fig. 7
figure 7

Characteristics of the IEI and ID. a Spatiotemporal characteristic of the values of ID and IEI. b Characteristics of IEI values attributed to urban areas and rural settlements. c Characteristics of ID values attributed to urban areas and rural settlements. d Scatter plot of ID. Note: The area of scatter denotes the value of ID, and the dashed line denotes the equalization line of ID attributed to urban areas and rural settlements

Full size image

Discussion

The quantification procedure of the isolation effect

While previous studies have developed several indicators to quantify the habitat isolation resulting from urban expansion, there has been an overlook in considering the ecological flow into the assessment procedures. Specifically, existing indicators primarily focus on factors such as distance, patch area, and the intensity of human activities, assuming that only urban patches surrounding habitat patches can exert isolation effects. However, if urban patches exist within the ecological corridor extent but are not in proximity to habitat patches, ecological flows are likely to be disturbed (Xun et al. 2017). This condition can consequently lead to habitat isolation, which has not been adequately considered in previous studies. In addition, previous studies aggregated the isolation effect to the nearest habitat patch, assuming that the construction pixel only affects ecological flow from the nearest side, thus failing to account for the bi-directionality of the isolation effect (Su et al. 2010).

In this study, we addressed the aforementioned limitations by developing two indicators, the IEI and ID, which provide a more comprehensive focus on the ecological processes between different habitat patches. On the one hand, the IEI employs the ecological network approach considering not only factors such as distance and human activity intensity but also the disturbance impact of construction land on ecological flows. On the other hand, the calculation of ID considers the bidirectional flow of ecological processes between paired habitat patches, emphasizing that habitats on both sides of corridors can be isolated. These two considerations are supported by existing studies. For instance, Staddon et al. (2010) experimentally confirmed that local extinctions and disruption of ecosystem processes caused by habitat isolation were mitigated by the presence of corridors. Similar arguments have been put forth in large-scale studies, which have verified that the disruption of corridors isolates habitat patches and generally reduces the abundance of animals and plants (Tewksbury et al. 2002; Tian et al. 2011; Haddad et al. 2015).

The increasing isolation effect of construction land on natural habitats

Our study reveals that urban areas experienced a dramatic expansion during the study period, while the ecological areas presented only a modest decrease. This phenomenon lends support to the idea that urbanization may not directly lead to a decrease in ecological value and biomass (Golubiewski 2006; Kong et al. 2022; Zhang et al. 2022). This can be attributed to China’s Ecological Civilization Construction and Ecological Redline Policy, which rigorously safeguards natural habitats (Zhou et al. 2012). However, our results demonstrate that the indirect isolation effect of construction land expansion on natural habitats is notably strong, a phenomenon that is widely overlooked in current landscape planning and related policies. Both the IEI and ID increased significantly during the study period, indicating a growing disturbance to natural habitats due to human activities. Two reasons can explain their rise. Firstly, the intensity of human activities in existing construction areas continues to escalate over time (Xu et al. 2020), and these activities affect neighboring environments through radiation effects such as pollution discharge, resource acquisition, and heat island diffusion (Dong et al. 2020; Xian et al. 2021; Wang et al. 2022). The disruption to ecological flow becomes more severe as construction land becomes more intensified. Secondly, newly expanded construction land encroached upon key areas for ecological flows (Xie et al. 2020; Li et al. 2022). On the one hand, areas surrounded by natural habitats have a high probability of development due to the spatial spillover effect of ecosystem services (Gao et al. 2020). On the other hand, the disorderly expansion of construction land is likely to occupy or disturb ecological stepping stones that are critical for species dispersal (Bucher and Aramburú 2014), thereby interrupting connections between different habitats.

Urban development exerts negative impacts on ecological processes by altering landscape configuration and composition (Forman 1995b). Numerous studies have provided evidence that urban sprawl results in the fragmentation and isolation of natural habitats (Dupras et al. 2016), ultimately leading to a decline in biomass (Haddad et al. 2015). Our study confirms these findings from a geospatial perspective and further emphasizes that rural settlements also contribute to habitat isolation. Although recent studies have acknowledged the negative impact of human activities associated with rural settlements on surrounding biomass (Zhang et al. 2022), the isolation effect caused by these settlements has not been thoroughly evaluated. Our results demonstrate that rural settlements indeed contribute significantly to habitat isolation, underscoring their negative impact on natural habitats. The mechanism behind habitat isolation caused by urban areas and rural settlements differs. Urban areas primarily exert an isolation effect through the high intensity of human activities, indicating that habitats located near well-developed urban patches are more likely to be isolated. Comparatively, rural settlements are widespread but have scattered distribution in China, making them more likely to interrupt ecological flows. While human activities in a single rural settlement patch may be relatively low, the cumulative isolation effect is high due to the large number of pixels that isolate habitat from various directions. As a result, while urban areas strongly isolate several habitat patches, the isolation effect caused by rural settlements is more extensive and affects a large number of habitat patches.

Instead of solely focusing on the rapid expansion of urban areas driven by fast socioeconomic development, it is essential to recognize that rural settlements have also experienced significant growth, playing a crucial role in exacerbating the isolation effect. Despite a decline in the rural population during the urbanization process, evidence indicates that the areas occupied by rural settlements continue to expand (Zhou et al. 2022). This phenomenon can be primarily attributed to China’s dual-track system, which distinguishes the land market between urban and rural sectors (Long et al. 2007). As rural residents participate in urbanization, rural land cannot be directly transferred to urban areas (Tan et al. 2020). Under this circumstance, rural settlement exhibits an increasing trend due to the low cost of rural land, even though the population in rural areas is decreasing (Zhu et al. 2020). Consequently, the isolation effect resulting from the expansion of both urban areas and rural settlements will continue to increase if the current urban-rural dual system remains unchanged. These conditions highlight the necessity for formulating land use planning policies in response to the habitat isolation caused by the rapid expansion of both urban areas and rural settlements.

Policy implications

This study underscores the imperative need to address the increasing isolation of natural habitats, an issue that has not received adequate attention in current planning policies. With ongoing urbanization in China, natural habitats are increasingly isolated due to the rapid expansion of construction land. Our study reveals that the spatial positions of ecological corridors have remained relatively stable, highlighting the importance of protecting existing corridors to mitigate the isolation effect. Heightened isolation poses the risk of habitat degradation, as ecosystem stability becomes solely reliant on the internal habitat and cannot be compensated for by other habitats (Prugh et al. 2008). Our study developed two indices designed to identify isolation hotspots at a grid scale and habitats that are highly isolated. This information can be invaluable for land use planners and ecologists in pinpointing construction land that requires greening and habitats in need of improving landscape connectivity. The indices also facilitate differentiation between urban areas and rural settlements, aiding in the implementation of policies tailored to different types of construction land.

Our study proposes suggestions for policymakers from two aspects. Firstly, optimizing construction land to reduce the isolation effect requires limiting the intensity of urban areas that exert a high isolation effect on natural habitats, enhancing the creation of green and blue infrastructures, and guiding urban expansion planning to avoid disrupting connections between different habitat patches. Additionally, governments should implement centralized resettlement policies to consolidate fragmented rural settlements, particularly those in proximity to natural habitats or within the corridor extent. For instance, integrating depopulated rural settlements into central villages outside the corridor extent, and utilizing the vacated land for protection purposes, such as creating ecological stepping stones. Secondly, increasing landscape connectivity between habitats necessitates the implementation of ecological protection policies to address the growing habitat isolation in highly urbanized regions. Proactive measures, such as the establishment of heterogeneous landscapes, the construction of ecological stepping stones, and the planting of shelterbelt trees, can contribute to the optimization of ecological corridors.

Reliability analysis

The identification of the ecological network represents a crucial step in this study. To ensure its accuracy, we conducted a reliability analysis to validate the identification results for both habitat patches and ecological corridors (Fig. 8). Firstly, we compared the distribution of habitat patches with the national and local natural reserves obtained from the website of the Natural Reserve of China (http://www.zrbhq.com.cn/). The findings revealed that there are ten terrestrial natural reserve zones in the study area, nine of which are located within the habitat patches identified by this study. This indicates a spatial consistency between the identification results and the officially protected areas. Secondly, we tested the impact of spatial extent on the ecological corridors, as the extent of a landscape determines the number of ecological corridors. To do so, we expanded our study area to encompass the entire Yangtze River Delta, including three provinces (Jiangsu, Zhejiang, Anhui) and the city of Shanghai. By utilizing our study framework, we assessed whether there was a significant change in the number of ecological corridors within the original study area. The results demonstrated that only seven corridors, which were situated between the expanded area and the NMA, connected the habitat patches inside the study area with those outside it. These corridors had limited length within the study boundary and thus did not exert a substantial impact on the distribution of the original ecological network. Therefore, the ecological network identified by our study is reliable, given the validation and analysis of both habitat patches and ecological corridors.

It is crucial to select a study area that exhibits relative independence within the overall landscape when conducting ecological network analysis. This consideration ensures the reliability and accuracy of the identification results of ecological corridors on a regional scale. In this study, a high level of patch integrity was observed for natural habitats in the southern part of the NMA, while in other directions, the habitat patches were situated at a considerable distance from the study area. These conditions provide assurance that the number of ecological corridors within the NMA did not significantly increase when expanding the spatial extent, indicating that the NMA is a relatively reliable region for accurately identifying the ecological network.

Fig. 8
figure 8

Reliability analysis for the identification of ecological network. Note: Taking the ecological corridor identification results in 2000 as an example. The area with dark grey is the region of NMA, while the light grey denotes the expanded area. “Corridor-outside” represents the corridors located within the expanded area; “Corridor-inside” denotes the corridors located within the NMA; and “Corridor-edge” indicates the corridors located between the expanded area and the NMA.

Full size image

Limitations and future perspectives

There are two limitations in our study. Firstly, the width of ecological corridors is an important consideration for practical landscape planning, but there is currently no consensus on how to define it (Peng et al. 2017; Ford et al. 2020; Hou et al. 2022). Often, corridor width is either directly defaulted as a fixed value (Liu et al. 2021) or subjectively determined based on the requirements of ecological protection policies (Nie et al. 2021). In our study, we proposed a simple but robust method to determine corridor width, which emphasizes incorporating substantial ecological areas as stepping stones to warrant connectivity between habitat patches. However, future studies should explore applicable methodologies for identifying corridor width, considering the complex ecological processes related to species movement influenced by regional landscape features. Secondly, the simplification of the types of construction land and the threshold of human activity. Our study divided construction land into urban areas and rural settlements due to data limitations, but construction land can be further subdivided into residential, commercial, industrial, infrastructure, and other types based on functional perspectives. Different types of construction land may exert distinct isolation effects on habitats. Future studies should focus on examining the habitat isolation effects and mechanisms associated with different types of construction land based on detailed classification datasets. Moreover, the threshold for human activities used in our study assumed that the maximum influence distance of human activities on natural habitats was 5 km, which was based on existing studies. However, the impact of human activities is complex and related to various human behaviors. Future studies should consider these differences to improve the accuracy of assessments.

Conclusions

The isolation of natural habitats poses a significant risk to ecological degradation, which is detrimental to regional landscape sustainability. However, there is a lack of available indices that can accurately assess the isolation effect. In this regard, our study introduced two indices, namely the IEI and ID, to evaluate the isolation effect caused by construction land and the degree of isolation of natural habitats, respectively. Our study is the first attempt to use the ecological network approach in habitat isolation assessment, and the results are highly significant for landscape planning and ecological conservation. Two major conclusions can be made as follows:

Firstly, the indices developed in our study revealed the isolation effect of construction land on natural habitats and differentiated the contributions of urban areas and rural settlements. The findings indicate that both urban areas and rural settlements have a significant impact on habitat isolation. More attention should be directed toward addressing the negative effect of rural settlements on natural habitats. Secondly, practical landscape planning should consider the isolation effect of construction land since a slight decrease in the ecological area but a dramatic increase in habitat isolation. Policymakers should mitigate the isolation effect by controlling the development of construction land and safeguarding ecological corridors. Our indices and approaches possess strong interpretability and can be applied to other regions facing rapid urbanization.