Introduction

With rapid urbanization, the transformation of natural surface by human activities has grown intensively, fragmenting the land surface, exacerbating land use intensity and leading to spatial conflicts (Winkler et al. 2021; Li et al. 2022). Ecologically vulnerable region is highly sensitive, easily affected by human activities and more prone to irreversible damage, requiring solutions with the worst-case-scenarios in mind for sustainable development (Ren et al. 2022; Li et al. 2021). Ecological security pattern (ESP), as a form of sustainable landscape pattern, can realize ecological conservation in vulnerable region through the improvement of landscape connectivity and safeguard human well-being by conserving ecosystem services provided by natural ecosystems (Yu 1996; Dong et al. 2022). ESP is a stable network structure formed by the most important ecological patches, and the best ecological corridors connecting these patches (Kong et al. 2021). In contrast to the concepts of sustainable landscape pattern such as ecological network for species migration and green infrastructure for ecological space conservation, ESP focuses on overall ecosystem conservation and landscape connectivity improvement, with the goal of achieving long-term regional ecological security and system conservation for ecologically vulnerable region (Peng et al. 2018a).

As an important concept in landscape ecology, landscape connectivity includes structural connectivity which describes ecological patches distribution, and functional connectivity which refers to ecological flow (Taylor et al. 1993; Ernst 2014). ESP takes ecological sources and ecological corridors as the core components, aiming to quantify the functional connectivity (Dong et al. 2020). In details, ecological sources are the key foundation of ESP construction, and should be capable of providing stable ecosystem services in a sustainable manner (Jiang et al. 2022). Ecological source identification takes landscape pattern (Albert et al. 2017) and ecosystem services (Peng et al. 2018b; Hofman et al. 2018) as the main evaluation indicators. On this basis, in order to achieve landscape sustainability, recent studies have taken the risks of geological hazards and landscape transformation into account (Peng et al. 2018a; Zhou et al. 2021). Ecological corridors help to achieve regional ecological security and landscape sustainability by connecting ecological sources. Circuit model, taking ecological sources and ecological resistance surface as inputs, is proposed based on graph theory and has been widely applied for ecological corridor extraction (Pomianowski and Solon 2020). Ecological resistance is a representation of the obstruction of landscape surface to ecological flow, which is generally assigned based on land use types, and modified by nighttime light intensity, impervious area ratio, and other indicators reflecting human activities (Peng et al. 2018a; Tang et al. 2023).

ESP optimization is an important transition from spatial modeling to planning practice. ESP prioritization, as an important procedure of ESP optimization, takes graph theory as the core approach, including multiple indicators and associated methods. For priority ranking of ecological sources, network topology indicators such as node degree and centrality measure the node’s position in the graph, which can represent the importance of ecological sources, usually being used as a ranking indicator (Qiu et al. 2022). On the other hand, there are studies that simulate the removal of a single node from a complete network and use the variation of network connectivity indicators as a representation of the node importance (Urban and Keitt 2001). For ecological corridors, current intensity based on circuit model can represent ecological flow, which has been used to measure the importance of ecological corridors. Gravity model can also quantify the importance of ecological corridors by calculating the interaction strength of the relating nodes (Xiao et al. 2020). The minimum spanning tree method can be used to extract the core ecological corridors by characterizing the tree structure with the minimum cumulative weights (Luo and Wu 2021). The above prioritization methods identify the relative importance of ecological sources and ecological corridors by abstracting them as nodes and edges of complex networks (Virtanen et al. 2020).

Ecological sources were usually identified based on landscape pattern or ecosystem services, without a comprehensive perspective. Ecosystem health represents the ability of an ecosystem to continuously provide valuable ecosystem services, taking both spatial pattern and human benefits into account (Costanza 1992; Peng et al. 2015), which can serve as a suitable indicator for identifying ecological sources. In terms of quantifying ecological resistance, the existing methods took a subset of human activities. As one of the hotspots of ecosystem conservation research, human footprint assesses the pressure of human activities on ecological processes and natural landscapes (Mu et al. 2022), which is suitable for ecological resistance quantification. The greater the human footprint, the larger the obstruction to ecological flow and thus the ecological resistance (Luo et al. 2020).

In the optimization of ESP, ecological sources and ecological corridors were usually abstracted as homogenized network elements. In this process, the structural composition and spatial pattern of ecological sources and ecological corridors were often ignored, which might lead to the mismatch between spatial planning and actual needs (Hersperger et al. 2006; Luo et al. 2020). Specifically, the land use within a patch of ecological sources or ecological corridors reflected the levels of anthropogenic disturbance and landscape fragmentation, which could affect patch stability. This was because high level of disturbance and fragmentation would be a great threat to biodiversity and ecosystem services, and thus patch stability for maintaining the initial status (Haddad et al. 2015). This highlights the need of considering the internal element structure of ESP.

The input of ecological conservation and restoration is limited from the perspective of spatial planning, while the goals and demands are diverse and complex (Teng et al. 2011; Liberati et al. 2020). For example, when determining the priority of ESP conservation, it is vital to manage the conservation of sensitive ecological sources and ecological corridors that are easily disturbed and damaged, in addition to the general maintenance of landscape network connectivity. Based on this context, the allocation of limited inputs to the conservation of key ecological elements has become an urgent issue in ESP optimization. Land use conflict can describe the patch stability by reflecting the landscape composition and spatial pattern of the patches (Jiang et al. 2021b). Thus, combining patch stability and network connectivity can aid in the resolution of this issue, and provide a better solution for ESP optimization (Pratiwi and Juerges 2022).

In order to provide an ESP scheme to conserve ecosystem and thereby safeguard human well-being, we aimed to develop a framework for ESP construction and optimization that takes both patch stability and network connectivity into account. We applied this framework to a case study of Ningxia Hui Autonomous Region, a typical ecologically vulnerable region. In details, the aims of this study are: (1) to construct ESP based on ecosystem health, human footprint and circuit model, (2) to explore the relationship between ESP optimization objectives in network interference scenarios based on land use conflict and network analysis, and (3) to optimize ESP with comprehensive priorities integrating patch stability and network connectivity.

Study area and datasets

Study area

Ningxia Hui Autonomous Region is located in the arid region of northwestern China (104°17′E − 107°39′E, 35°14′N − 39°23′N), and consists of five cities including Yinchuan, Shizuishan, Wuzhong, Guyuan and Zhongwei (Fig. 1). It is one of the most ecologically vulnerable regions in China, with over 40% area listed as ecological vulnerability. Ningxia lies entirely within the Yellow River Basin, and is surrounded by three deserts in the east, west and north, namely Mu Us, Tengger and Ulan Buh (Lyu et al. 2018). Ningxia has a substantially higher annual evaporation than precipitation and hence relies heavily on the Yellow River for water replenishment. The extensive wetlands contribute significantly to flood mitigation, climate regulation, and biodiversity conservation (Li et al. 2021).

The elevation varies from low in the north to high in the south, with major variation in the west and gentle fluctuations in the east. Ningxia is separated into three zones, including the northern green development zone (NGDZ), the central enclosed protected zone (CEPZ), and the southern water conservation zone (SWCZ). Helan Mountain in the north, Luo Mountain in the middle and Liupan Mountain in the south are important green ecological barriers of Ningxia. The population and construction land are mainly concentrated along the Yellow River, with the highest level of land use conflict in NGDZ. Protecting vegetation and biodiversity, improving water conservation and soil retention, building ecological corridors and constructing conservation networks are important goals of ecological conservation in Ningxia. The construction of ESP is an effective approach to conserving the vulnerable ecosystems of Ningxia and to solving the problems of landscape fragmentation, protected area isolation, and ecological connectivity reduction.

Fig. 1
figure 1

Geographical location and elevation (a), and land use types (b) of Ningxia Hui Autonomous Region

Datasets

The following datasets of 2020 were used in this study and transformed into Albers equal area projection with spatial resolution of 250 m: land use data with spatial resolution of 30 m from Resource and Environmental Science Data Center (https://www.resdc.cn/); soil property data with spatial resolution of 1 km from the Harmonized World Soil Database (https://www.fao.org/soils-portal/soil-survey/); precipitation data from China Meteorological Data Service Center (https://data.cma.cn/); MODIS Normalized Difference Vegetation Index (NDVI) data, Net Primary Productivity (NPP) data and Evaporation data with spatial resolution of 500 m from the United States Geological Survey (https://www.usgs.gov/); Digital Elevation Model (DEM) data with spatial resolution of 30 m from Resource and Environmental Science Data Center (http://www.resdc.cn/); annual nighttime light intensity data with spatial resolution of 500 m from Earth Observation Group (https://eogdata.mines.edu/products/vnl/) (Elvidge et al. 2021); population density data with spatial resolution of 100 m from WorldPop (https://www.worldpop.org/); and road data from OpenStreetMap (https://www.openstreetmap.org/).

Methods

We proposed a framework for ESP construction and optimization that combines patch stability and landscape network connectivity (Fig. 2). First of all, ecosystem health was evaluated in terms of physical health and ecosystem services to identify ecological sources. Then, human footprint was quantified using land use types, population density, road density and nighttime light intensity to form ecological resistance surface. Furthermore, the circuit model was used to extract ecological corridors to form ESP. Five indicators including closure, line-point rate, node connectivity, cost ratio and equivalent connectivity were used to evaluate the network connectivity of ESP. Finally, land use conflict analysis, connectivity importance analysis and network robustness analysis were applied to measure network connectivity variation under different scenarios, and to propose ESP optimization scheme.

Fig. 2
figure 2

Research framework for ecological security pattern construction and optimization

Ecological source identification based on ecosystem health assessment

Protecting key ecological sources is critical for maintaining regional ecological security (Jiang et al. 2022). Taking ecosystem health as an evaluation indicator to identify ecological sources can aid in achieving the goal of sustainable landscape pattern. According to the existing methods, patches with the top 30% value of ecosystem health and area of no less than 10 km2 were chosen as ecological sources (Jiang et al. 2021a). Ecosystem health included two dimensions, i.e. physical health and ecosystem services (Peng et al. 2015):

$$EH=\sqrt{PH\times ES}$$

where \(EH\) is ecosystem health; \(PH\) is physical health; and \(ES\) is the comprehensive importance of ecosystem services.

Physical health included three dimensions: vigor (\(V\)), organization (\(O\)) and resilience (\(R\)) (Costanza 1992):

$$PH=\sqrt[3]{V\times O\times R}$$

where \(V\) is represented by NDVI, measuring the vegetation growth; \(O\) denotes ecosystem structural stability determined by the spatial composition and pattern of the landscape, and was evaluated in terms of landscape heterogeneity (represented by Shannon diversity index, SHDI) and landscape connectivity (represented by contagion and patch cohesion, i.e. CONTAG and COHESION with equal weights) with equal weights; \(R\) was determined by land use types to represent the ability of the ecosystem to recover its original pattern and function after external disturbances (Peng et al. 2017).

Ecosystem service is a crucial indicator of ecosystem health in view of human well-being (Ouyang et al. 2016). According to China’s functional zoning, Ningxia is vulnerable to water and soil loss and also plays an important role in biodiversity conservation. Meanwhile, climate change mitigation has become a top priority for the world, of which the carbon peaking and carbon neutrality goals proposed by the Chinese government are getting attention. To focus on the most pertinent challenges to Ningxia, we quantified four kinds of ecosystem services including habitat maintenance, carbon sequestration, water conservation, and soil retention. The comprehensive importance of ecosystem services was calculated as the mean value of the four normalized ecosystem services.

In details, the service of habitat maintenance refers to the ability of ecosystem to provide living conditions suitable for individuals and populations, measured with the Habitat Quality module of the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model, which takes the availability of living resources, biological reproduction and presence quantity into account (NCP, 2022). Carbon sequestration is an important regulation service of terrestrial ecosystem, of capturing, collecting and storing carbon into carbon pools through different approaches, of which vegetation and soil are the most important (Chen et al. 2019). Water conservation service was calculated based on Water Yield module of the InVEST model, referring to the ability of natural ecosystem to intercept or store water resources from rainfall with the help of vegetation canopy, litter and roots, and retention, interception and storage of soil (Peng et al. 2018b). Soil retention service refers to the mitigation capacity of landscape to soil erosion, which was calculated using the Revised Universal Soil Loss Equation (RUSLE) (Renard et al. 1997).

Ecological resistance surface formation based on human footprint

Ecological resistance surface reflects the obstruction of landscape to ecological processes including material flow, energy flow and animal migration. It is a key material for ecological corridor extraction which is highly correlated with human activity intensity (Peng et al. 2018a). The resistance surface was formed using human footprint, which was calculated as the sum of the indicators of land use types, population density, road density, and nighttime light intensity (Sanderson et al. 2002a).

Land use types, reflecting human-driven natural land surface transformation, were assigned the values of 7, 10 and 2 for cropland, construction land and unused land respectively, and 1 for other land use types (Lin et al. 2018). Serving as the link between human and the environment, roads are closely tied to human activities (Trombulak and Frissell 2000). The kernel density estimation method was used to evaluate the impact of roads to natural ecosystem, and the values were stretched to the range of [0, 10] in order to avoid the saturation problem of the multiple ring buffer assignment approach in the high-density region (Bonato et al. 2019; Dong et al. 2020).

Nighttime light intensity reflects the intensity of electric power facilities, being an important representation of the pressure of human activities on the environment. The values of nighttime light intensity above 95% grids were set as 10, and the rest was linearly stretched to [0, 10]. Population density was used to represent the nonlinear impacts of human-induced environmental variation on natural ecosystem (Cincotta et al. 2000). The human footprint corresponding to population density was calculated as following:

$$P = \left\{ {\begin{array}{*{20}l} {3.333 \times {\text{lg}}\,\left( {pop + 1} \right),\,} \hfill & {0 \leq pop<1000} \hfill \\ {10,} \hfill & {\,pop \ge 1000} \hfill & {} \hfill & {} \hfill \\ \end{array} } \right.$$

where \(P\) is the human footprint corresponding to population density; and \(pop\) is population density.

Ecological security pattern construction and network connectivity assessment

Ecological corridors, as important channels connecting ecological sources, were extracted with circuit model (Peng et al. 2019). Ecological sources were regarded as nodes of the circuit. Non-ecological source area was regarded as resistors with different resistance values in the circuit branch. The current intensity \(I\) between any two nodes was calculated as following (Peng et al. 2018b):

$$I=\frac{V}{{R}_{eff}}$$

where \(V\) is the voltage; and \({R}_{eff}\) is the effective resistance, reflecting the distance between nodes at both ends of the branch. Ecological corridors were extracted using the Linkage Mapper toolbox of ArcGIS 10.2 software.

Network connectivity assessment was conducted in order to compare ESPs identified based on ecosystem health and ecosystem services. According to graph theory, ESP could be abstracted as a network composed of nodes and edges (Kong et al. 2010). Closure (\(\alpha\)), line-point rate (\(\beta\)), node connectivity (\(\gamma\)) and cost ratio (\(c\)) were used to characterize network connectivity, referring to occurring extent of loops, reachability, node connectivity degree and input-output relationship, respectively (Nie et al. 2021). The calculating formulas of the indicators were as following:

$$\alpha =\frac{l-v+1}{2v-5}$$
$$\beta =\frac{l}{v}$$
$$\gamma =\frac{l}{3(v-2)}$$
$$c=1-\frac{l}{d}$$

where \(l\) is the number of ecological corridors (edges); \(v\) is the number of ecological sources (nodes); and \(d\) is the total length of ecological corridors.

Furthermore, a connectivity indicator named equivalent connectivity was used to characterize the overall connectivity potential of the network in order to take both the area of ecological sources and cost distance of ecological corridors into account (Saura et al. 2011).

$$EC=\sqrt{\sum _{i=1}^{n}\sum _{j=1}^{n}{a}_{i}\times {a}_{j}\times {p}_{ij}}$$

where \(EC\) is equivalent connectivity; \(n\) is the total number of ecological sources; \({a}_{i}\) and \({a}_{j}\) represent the area of ecological source \(i\) and \(j\) respectively; and \({p}_{ij}\) is the maximum probability of the potential path between ecological source \(i\) and \(j\), calculated by exponential function.

Ecological security pattern optimization integrating patch stability and network connectivity

As ESP components, ecological sources and ecological corridors were considered to retain their own stability while simultaneously contributing to network connectivity improvement, resulting in a divergence between the two conservation objectives of patch stability and network connectivity. To clarify the relationship between the two objectives, patch stability and network connectivity should be evaluated. On one hand, based on the evaluation methods of probability of connectivity (PC) and patch importance (dPC), we extended the node removal method proposed by Xu et al. (2020) from ecological sources to ecological corridors, and measure network connectivity by simulating the impact on equivalent connectivity of removing a certain ecological source or ecological corridor from a complete ESP. On the other hand, we calculated the land use conflict index (\(LUCI\)) to characterize the patch fragmentation risk of ecological sources or ecological corridors, as a reverse representation of patch stability. \(LUCI\) was calculated from such three dimensions as spatial complexity index (\(SCI\)), spatial vulnerability index (\(SVI\)), and spatial fragmentation index (\(SFI\)) (Jiang et al. 2021b):

$$LUCI=SCI+SVI+SFI$$

where \(SCI\) reflects the disturbance of the neighborhood patch, calculated by the area-weighted mean patch fractal dimension (AWMPFD); \(SVI\) represents the ability of resisting external disturbance, assigned according to the related land use types; and \(SFI\) represents the competition intensity of different land use types, calculated by patch density (PD).

In this study, ESP was viewed as a complex network. Corresponding to conservation objectives of patch stability and network connectivity, three node/ edge removal scenarios were set to analyze the consistency of network robustness variation (Table 1). The hypothesis is as following: If the variation trends of network robustness indicators of Scenario 1 and 2 are consistent and lower than those of Scenario 3, the two objectives can achieve collaborative conservation and obtain higher outcome than random conservation without design. If the indicators’ variation trends of Scenario 1 and 2 are inconsistent, it indicates that trade-off exists between the two objectives, and conservation priority should be assessed under comprehensive consideration.

Table 1 Conservation objectives of node/ edge removal scenarios

We used connectivity robustness, global efficiency, and equivalent connectivity to assess the impact of ecological sources or ecological corridors removal on the network robustness. The three indicators have different connotations, focusing on network structure, network performance and global connectivity, respectively. Specifically, connectivity robustness represents the ratio of the maximum number of subgraphs composed of the remaining nodes and edges to the number of remaining nodes, which describes the ability of the network to maintain structural stability after being attacked. Global efficiency describes the transmission efficiency between nodes of the network, taking the length of the edges into account. Equivalent connectivity is a global connectivity indicator that considers the area of the nodes and the cost distance. Network robustness analysis was implemented in R version 4.1.2 with the R package igraph. The former two indicators were calculated as following (Luo et al. 2020):

$$R=\frac{{C}_{max}}{n}$$
$$E=\frac{1}{n(n-1)}\sum _{i\ne j\in G}\frac{1}{{d}_{ij}}$$

where \(R\) is connectivity robustness; \({C}_{max}\) is the maximum number of subgraphs; \(n\) is the number of nodes, which will change with the removal of nodes; \(E\) is global efficiency; \(G\) is the node set; and \({d}_{ij}\) is the shortest cost distance between nodes \(i\) and \(j\).

To integrate the conservation objectives of patch stability and network connectivity, a comprehensive importance index (\(CII\)) was proposed to measure the final priority in ESP optimization.

$$CII=PSI\times NCI$$

where \(PSI\) is patch stability importance, calculated as normalized value of land use conflict; and \(NCI\) is network connectivity importance, calculated as normalized value of equivalent connectivity change after removal.

Results

Spatial pattern of ecosystem health

High values of ecosystem vigor concentrated in SWCZ, cropland along the Yellow River, Helan Mountain and Luo Mountain (Fig. S1). The overall distribution of organization was uniformly distributed with higher values in the patches of higher connectivity. The resilience was higher in ecological land, while lower in artificial and semi-artificial ecosystems. Habitat maintenance was relatively high, with the highest values concentrated in Helan Mountain, Luo Mountain and Liupan Mountain. The large area of construction land and cropland in NGDZ posed a great threat to natural habitats, with 0.58 as the mean value of habitat maintenance service. Carbon sequestration showed significant spatial variation from the north to the south, with high values in Liupan Mountain and Nanhua Mountain and low values in the cropland along the Yellow River in the north. Helan Mountain in the north had the sunny slope at the eastern foot, with gravel as the main soil type and thus poor vegetation growth conditions. High values of water conservation concentrated in the southeast. Soil retention was high in Helan Mountain, Luo Mountain, Nanhua Mountain and Liupan Mountain, due to the strong soil retention ability of forest land and grassland.

The spatial patterns of physical health, ecosystem services and ecosystem health were overall similar, but different in Helan Mountain (Fig. 3). High values of ecosystem services and ecosystem health were significantly clustered, while those of physical health were relatively dispersed. The continuous cropland along the Yellow River had a high level of physical health, but were not able to provide high ecosystem services, resulting in low ecosystem health. The main urban centers in Ningxia had fragmented landscape and were dominated by artificial surfaces with low physical health and ecosystem services. Luo Mountain, Nanhua Mountain and Liupan Mountain had continuous distribution of forest land with high ecosystem health. As a result, these areas were identified as ecological sources.

Fig. 3
figure 3

Spatial patterns of physical health, ecosystem services and ecosystem health

Spatial patterns of ecological resistance surface and ecological security pattern

Ecological resistance surface formed based on human footprint was shown in Fig. 4. Ecological resistance had a mean value of 9.71 and spatial pattern with high values in the north and south, and low values in the middle. The mean values of ecological resistance of NGDZ, CEPZ and SWCZ were 9.99, 8.77 and 10.82, respectively. The northern Yinchuan Plain, including Yinchuan, Wuzhong, Shizuishan and Zhongwei, had high social and economic development, high intensity of human activities, and dense road, resulting in high ecological resistance. However, the northwest boundary was dominated by forest land and unused land, with low intensity of human activities, and thus low ecological resistance. The distribution of the roads in the southern region was more uniform, with large area of cropland, leading to higher ecological resistance.

Fig. 4
figure 4

Spatial patterns of ecological resistance and ecological security pattern

According to the ecosystem health, 71 ecological sources with a total area of 10970.25 km2 were identified, accounting for 21.31% of the total area of Ningxia (Fig. 4). The area of grassland, cropland and forest land accounted for 61.46%, 21.71% and 13.91% of the ecological sources, respectively. The area ratio of forest land in Ningxia was only 5.40%, while those of grassland and cropland were 45.31% and 33.90%, respectively. Ecological sources were concentrated in the southern part of Ningxia. There were also small ecological sources along the Yellow River, which had relatively high values of ecosystem health, and thus could act as stepping stones in ecological flow. The areas of ecological sources in SWCZ, CEPZ and NGDZ were 6397.49 km2, 3453.71 km2 and 1119.05 km2, accounting for 61.39%, 18.54% and 4.99% of each zone, respectively. On the whole, the border between Ningxia and the deserts in the north lacked of ecological sources and might face great challenge of ecological risk.

Based on ecological sources and ecological resistance surface, 150 ecological corridors with a total length of 2513.31 km were extracted through circuit model (Fig. 4). The total area of ecological corridors was identified as 3950.88 km2, accounting for 7.68% of the total area of Ningxia. Among them, the areas of ecological corridors of NGDZ, CEPZ and SWCZ were 2417.63 km2, 1517.81 km2 and 15.44 km2, accounting for 10.78%, 8.15% and 0.15% of the total area of each zone, respectively. Ecological corridors along the Yellow River ran east-west, while those in the central region ran north-south. The southern region had continuous distributed ecological sources with large area, and thus had few ecological corridors with small area. The ESP network connectivity indicators of \(\alpha\), \(\beta\), \(\gamma\), \(c\) and equivalent connectivity were 0.60, 2.14, 0.74, 0.94 and 342.80, respectively.

Trade-off between patch stability and network connectivity conservations

The removal priority of ecological sources in view of patch stability was high in the south and low in the north (Fig. 5). Two ecological sources with the highest land use conflict were located along the Yellow River, including Mingcui Lake, Zhaojia Lake and Yellow River National Wetland Park, which were small with high vulnerability and fragmentation. The largest ecological source had the third highest land use conflict and the highest complexity among all the ecological sources. These ecological sources were in urgent need of maintaining patch stability. Land use conflict in Helan Mountain and Luo Mountain was low, mainly due to the low vulnerability and fragmentation. Ecological corridors with the highest land use conflict were mainly distributed in the plain along the Yellow River, with narrow range, high vulnerability and fragmentation. Ecological corridors in the north had higher land use conflict, more fragmented pattern and thus were more susceptible to disturbance. Land use conflict of the connecting corridor between Luo Mountain and the largest ecological source in the south was also high.

Ecological sources with the highest removal priority in view of network connectivity included the largest ecological source in the south, Helan Mountain and Luo Mountain (Fig. 5). Network connectivity based removal priority of small ecological sources was low, since there were fewer corridors connecting them. The distribution of network connectivity based removal priority of ecological corridors was roughly opposite to that of patch stability. That was to say, the ecological corridors with high connectivity contribution were more stable, showing a potential trade-off between the two conservation objectives.

Fig. 5
figure 5

Spatial patterns of removal priorities of ecological sources and ecological corridors based on patch stability and network connectivity

Under different scenarios, variation trends of connectivity robustness, global efficiency and equivalent connectivity of ESP were significantly different (Fig. 6). With the increasing removal of ecological sources (nodes) and ecological corridors (edges), the connectivity robustness of ESP decreased from 1 to 0, from 0.29 to 0 for global efficiency, and from 342.80 to 0 for equivalent connectivity. When removing the nodes, the curves of connectivity robustness and global efficiency of Scenario 1 were higher than those of Scenario 2 but both fluctuated more sharply than those of Scenario 3, which slowly declined and then rose (Fig. 6a, c). Equivalent connectivity plunged significantly in Scenario 1 and 2 (Fig. 6e). Under Scenario 1, when removing 3, 32, 34, 56, 57 and 65 nodes according to land use conflict, connectivity robustness declined the fastest. Global efficiency declined the fastest when removing 3, 34, 56, 57 and 59 nodes. Equivalent connectivity declined the fastest when removing 1, 4 and 13 nodes.

Fig. 6
figure 6

Connectivity robustness (a, b), global efficiency (c, d) and equivalent connectivity (e, f) variation with removal of nodes or edges from the ecological security pattern under different scenarios

With the removal of edges, connectivity robustness and global efficiency of Scenario 1 were overall higher than those of the other two scenarios, with the fastest decline when 30 and 115 edges were removed (Fig. 5b, d). Equivalent connectivity of Scenario 2 was higher than that of Scenario 1 overall (Fig. 5f). Scenario 2 had the lowest values of connectivity robustness and global efficiency compared with Scenario 1 and 3, with the fastest decline when removing 58 edges. In general, the variation trends of the three indicators under Scenario 1 and 2 were inconsistent, suggesting trade-off between prioritizing conservation of ESP components with low patch stability and high contribution to network connectivity.

Discussion

Importance of landscape pattern in ESP construction and optimization

As an important component of ESP, ecological sources are identified using a variety of methods, including landscape integrity and conservation importance. Among them, ecosystem services are the most commonly used indicators (Jiang et al. 2021a). However, in this study we assessed ecosystem health rather than ecosystem services to construct ESP. The \(\alpha\), \(\beta\), \(\gamma\), \(c\) and equivalent connectivity of the ecosystem service based ESP were 0.59, 2.11, 0.73, 0.95 and 132.38, respectively, showing lower network connectivity and higher cost ratio than ESP identified based on ecosystem health (Fig. S2). It is suggested that considering landscape pattern when constructing ESP has a positive impact on connectivity enhancement. Compared with other related studies in Ningxia, it was found that Helan Mountain and the Yellow River were important ecological sources of the city belt along the Yellow River of Ningxia (Cheng et al. 2022). As a result, the ecological planning of Ningxia emphasized the importance of “three mountains and one river”, including Liupan Mountain, Helan Mountain, Luo Mountain and the Yellow River, which were all covered by our ESP.

ESP is a complex node-edge network (Zhou et al. 2021). Conserving nodes and edges against damage and maintaining patch stability are important guarantee for network performance (Albert et al. 2017). Network interference simulation and network robustness analysis have become effective methods for ESP optimization (Luo et al. 2020; Wimberly et al. 2018). In this process, ecological sources and ecological corridors are frequently abstracted as homogeneous nodes and edges (Sritongchuay et al. 2019; Xu et al. 2020). However, function evaluation is actually inseparable from spatial pattern clarity (Hersperger et al. 2021; Mitchell and Devisscher 2022). Ecological sources and ecological corridors have certain area and spatial range, and thus should not be simply abstracted without considering the internal structure, resulting in the loss of some key information, which may be crucial to spatial planning decisions (Fig. 7). Although weights can be assigned to represent the differences of various patches, they can only express limited information such as patch area in practice. However, studies have indicated that spatial fragmentation will pose a greater threat to ecosystem than area loss (Cottam et al. 2009; Moore et al. 2022). The stability of ecosystem structure cannot be maintained without spatial pattern optimization (Li et al. 2023). Therefore, the optimization of ESP should consider not only the contribution to network connectivity of ecological sources and ecological corridors, but also their spatial configuration and associated patch stability (Boeye et al. 2014).

Fig. 7
figure 7

Diagram of abstracting ecological sources and ecological corridors with internal structure

ESP optimization scheme with conservation strategy integrating patch stability and network connectivity

Determining conservation priorities in different scenarios with screening out patches of the highest conservation value is an important measure of ecological planning Zhang and Li 2022). The ecological conservation objectives of spatial planning are diversified, while the investments in funds, managers and resources are all limited, leading to the inevitable concentration of limited resources on the conservation projects with the greatest improvements of the ecosystem structure and function (Sanderson et al. 2002b; Chen et al. 2023). To enhance the conservation efficacy, it is necessary to clarify the potential trade-off and synergy between various conservation objectives. The difference between Scenario 1 and 2 indicated the conflict in conservation priority of ecological sources and ecological corridors between the objectives of high patch stability and high network connectivity. In the case of limited investment, the two objectives could not be achieved simultaneously, and the trade-off between the two objectives should be considered to maximize the effectiveness of ESP optimization (Pohjanmies et al. 2017; Zhang et al. 2022).

In both conservation objectives, the removal priority of patches represented their conservation priority. Thus, the ecological sources or ecological corridors could be divided into four clusters by ranking the conservation priorities of patch stability and network connectivity (Fig. 8), with diversified conservation strategies for different clusters. Taking ecological sources as an example, cluster I had high importance of network connectivity maintenance, but faced high patch stability risk, being the units of the first conservation priority. Conserving ecological sources in this cluster could realize both conservation objectives of patch stability and landscape network connectivity. Ecological sources above and below the diagonal were more important to each goal. Ecological sources in cluster II and III had higher conservation priorities in terms of reducing patch stability risk and sustaining landscape network connectivity, respectively. Ecological sources in cluster IV could be considered as a secondary conservation choice since both priority of network connectivity and patch stability were low. In spatial planning, different clusters of ecological sources and ecological corridors could be selected for conservation based on the characteristics and needs of different regions, such as highly urbanized region, ecologically vulnerable region, and key ecological function region.

We had also identified the comprehensive priority that integrated both objectives for ESP optimization (Fig. 8). For example, the largest ecological source in the south of Ningxia covered Liupan Mountain and had the highest comprehensive conservation priority, since it was important for network connectivity but faced high risk of landscape fragmentation. Helan Mountain and Luo Mountain were important for network connectivity and faced low patch stability risk, and thus their conservation needs were not the most urgent. Ecological corridors in the south of Yinchuan were in urgent need of comprehensive conservation due to the high connectivity importance and high landscape fragmentation risk.

Fig. 8
figure 8

Conservation priority clusters of patch stability and network connectivity, and spatial pattern of comprehensive importance index for ESP optimization

The difference of comprehensive importance index was similar to the results of the four clusters, but provided more specific guidance on comprehensive conservation objective. From the perspective of ecological sources, the patch with the highest conservation priority was the largest ecological source in the south, including Liupan Mountain, Nanhua Mountain, Yunwu Mountain, Huoshizhai National Geological Park, and other important protected areas. The patch covering Qingtongxia Yellow River Grand Canyon and Qingtongxia Reservoir was also of high priority. As an important wetland along the Yellow River, Qingtongxia was one of the six wetlands of Ningxia listed in the National Important Wetland List. With Qingtongxia as the core region, the near-nature restoration and connectivity improvement projects of wetlands along the Yellow River could be strengthened to maintain ecological security of the whole region. From the perspective of ecological corridors, the ecological corridor with the highest conservation priority was located in the northeast of Ningxia, passing through Baijitan and Haba Lake, two protected areas with high land use conflict and network connectivity importance. This ecological corridor effectively filled the conservation vacancy in the south of Yinchuan and the north of Wuzhong. It was a good choice to expand the connected ecological sources and set new protected areas along the corridor to enhance both ecosystem stability and landscape connectivity.

Furthermore, as an arid province, Ningxia takes the “three mountains and one river” as the most important ecological areas. The numerous wetlands along the Yellow River have played a key role in climate regulation and habitat quality improvement. However, due to the surrounding large sandy region, Ningxia is under increased risk of sandy desertification. Therefore, considering the ecological characteristics of Ningxia, it is suggested for systematic management and comprehensive quality improvement of waterbodies, especially the integrated conservation of the Yellow River. Waterbody connectivity projects should be implemented to enhance water mobility and quality. At the same time, wind-sand control projects should be carried out to lessen the risk of the surrounding sandy region. Because of the high evaporation and low precipitation of Ningxia, and grassland as the main vegetated land, ecological restoration should be carried out according to local social-ecological conditions, for example increasing grassland but not forest land.

Limitations and future research directions

In this study, we constructed ESP based on ecosystem health and human footprint, explored potential trade-off between different conservation objectives, and optimized ESP based on land use conflict and network robustness. However, there are still some limitations. Firstly, due to data availability, only four kinds of ecosystem services were included, and the service of windbreak and sand fixation should be considered in future studies. Secondly, based on the impact of landscape fragmentation, we used land use conflict to characterize patch stability, without considering other influencing factors of protected areas and urban development. Future studies should consider more indicators to quantify patch stability. Finally, we analyzed the trade-off between patch stability conservation and network connectivity maintenance through network robustness analysis, and proposed comprehensive importance index to optimize the ESP. However, the specific cost and benefit analysis, as well as the simulation of different interference levels, need to be further explored.

Conclusion

ESP is one of the effective approaches to safeguard ecologically vulnerable region and optimize landscape pattern with the worst-case-scenarios in mind. However, the current construction methods of ESP lack comprehensive consideration of landscape pattern and function. Network analysis in ESP optimization abstracts ecological sources and ecological corridors as homogeneous nodes and edges respectively, ignoring differences in land use conflict and patch stability. In this study, we constructed ESP based on ecosystem health and human footprint, and evaluated the importance of patch stability and network connectivity of ecological sources and ecological corridors by land use conflict and network robustness respectively. The ESP was optimized by comprehensive conservation priority considering both patch stability and network connectivity. The results showed that the ESP was composed of 71 ecological sources with a total area of 10970.25 km2 and 150 ecological corridors with a total area of 3950.88 km2. When simulating removal of ecological sources and ecological corridors, connectivity robustness, global efficiency, and equivalent connectivity decreased from 1, 0.29, and 342.80 to 0, with inconsistent variation under different scenarios, indicating the trade-off of conservation objectives between patch stability and network connectivity. This study proposed a new framework for conservation planners to construct and optimize ESP, which could prevent information loss caused by the abstraction of elements in network analysis, and provide guidance for the trade-off decision-making of different conservation objectives in ESP optimization.