Ecological Network Construction Methods Based on Habitat Accessibility and Availability under Climate Change

Abstract

This invention discloses a method for constructing ecological networks based on habitat accessibility and availability under climate change, comprising: screening bioclimatic variables of a target area to obtain a first set of bioclimatic variables; assessing future habitat availability based on species distribution points, non-climatic variables, and the first set of bioclimatic variables in the target area; screening the first set of bioclimatic variables to obtain a second set of bioclimatic variables, and using the second set of bioclimatic variables to assess future habitat accessibility; constructing a composite index based on availability and accessibility assessment results, and constructing a potential range transfer network based on accessibility assessment results, weighting each path in the potential range transfer network with the corresponding composite index value; using network analysis to identify key locations and paths in the potential range transfer network that require priority protection, and constructing an ecological network using the key locations and paths. This invention maximizes the protection of species' survival sustainability under climate change conditions.

Description

Technical Field

[0001] This invention relates to the field of ecological protection planning and management technology, and in particular to a method for constructing ecological networks based on habitat accessibility and availability under climate change. Background Technology

[0002] In recent years, the situation regarding climate change has become increasingly severe. Its impacts have caused irreversible damage to terrestrial, freshwater, and marine ecosystems, extensively affecting ecosystem structure, species geographic distribution, and phenology. It has altered the geographic distribution, seasonal dynamics, and composition of plant and animal species, and has even led to the extinction of some species, becoming one of the greatest threats to biodiversity in the coming decades. Therefore, developing conservation strategies adapted to climate change to maintain the sustainability of species and ecological conditions has become a crucial issue for biodiversity conservation in the context of climate change.

[0003] Establishing ecological networks is an effective way to conserve biodiversity and adapt to climate change. Ecological networks can maintain and increase habitat connectivity, protect biodiversity, and enhance inter-patch connections. In particular, building ecological networks in high-value conservation priority areas can promote the migration and exchange of species within the region. While various modeling methods exist for identifying conservation priority areas and constructing ecological networks, most are based on species distribution within a specific timeframe. However, species distribution is dynamic and easily affected by external environmental pressures. Especially with the intensifying impacts of climate change in recent years, more and more species are forced to migrate to areas with more suitable climates to avoid extinction. Therefore, traditional ecological networks based on static models may not effectively promote the sustainable survival of species. It is urgent to consider the dynamics of species migration under climate change, assess the migration potential of species, predict distribution changes, and construct climate change-adaptive ecological networks, which will help address the dynamic challenges to the long-term survival of species.

[0004] Currently, most existing methods for identifying priority protected areas and constructing ecological networks only consider the single factor of future habitat availability or accessibility, without taking into account the impact of environmental change on the overall dynamic process of species range migration. Although some methods consider both factors simultaneously, they are mostly simple superpositions of habitat suitability and climate rate analyses, ignoring the species dynamic range transfer processes implied by these analyses. Furthermore, current conservation strategies related to climate change adaptation largely focus on identifying climate refuges or climate connectivity areas, and have not yet been able to construct complete ecological networks that can help address climate change. Summary of the Invention

[0005] The technical problem to be solved by this invention is: in view of the technical problems existing in the prior art, this invention provides an ecological network construction method based on habitat accessibility and availability under climate change, which can construct an ecological network according to the potential distribution and spread of species under the influence of climate change, and maximize the protection of species survival and sustainability under climate change environment.

[0006] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0007] A method for constructing ecological networks based on habitat accessibility and availability under climate change includes the following steps:

[0008] S1) Obtain target data for the target area, and filter the bioclimatic variables in the target data to obtain the first bioclimatic variable group;

[0009] S2) Analyze habitat suitability based on species distribution points, non-climate variables, and the first bioclimate variable group in the target data to assess future habitat availability;

[0010] S3) The first bioclimatic variable group is screened to obtain the second bioclimatic variable group. The second bioclimatic variable group is used to calculate the climate rate and simulate potential range transfer paths to assess the accessibility of future habitats.

[0011] S4) Construct a composite index based on the assessment results of the availability and accessibility of future habitats, and construct a potential range transfer network based on the assessment results of the accessibility of future habitats. Weight each path in the potential range transfer network with the corresponding composite index value.

[0012] S5) Use network analysis to identify key locations and paths in the potential range transfer network that need priority protection, and use the key locations and paths to construct an ecological network.

[0013] Furthermore, the specific steps of step S1 include:

[0014] S11) Collect species distribution records within the target area, and filter out records that match the target spatial resolution from the species distribution records to obtain species distribution point data;

[0015] S12) Obtain bioclimatic variable data for the current and future periods of the target area, resample the bioclimatic variable data according to the target spatial resolution, and perform preliminary screening of the sampling results to obtain the first bioclimatic variable group.

[0016] S13) Obtain non-climate variable data within the target area and resample the non-climate variable data according to the target spatial resolution.

[0017] Furthermore, when filtering the bioclimatic variables in the target data to obtain the first bioclimatic variable group, the following steps are included:

[0018] Pearson correlation analysis and variance inflation factor test were performed on the variables, and pairs of variables with high correlation were selected based on the results of the correlation analysis and variance inflation factor test.

[0019] For each variable, construct a univariate generalized linear model with linear and quadratic terms, and calculate the AUC value of the univariate generalized linear model.

[0020] The less important variables are excluded from pairs of highly correlated variables, and variables with AUC values ​​less than a first preset threshold are also deleted.

[0021] Furthermore, the specific steps of step S2 include:

[0022] S21) Based on species distribution point data, land cover and topography from the first bioclimatic variable group and non-climatic variable data are used as environmental prediction variables. Species distribution models are fitted for each species. Species distribution models that exceed the predetermined prediction performance threshold are selected from all species distribution models. The selected species distribution models are used to predict the continuous habitat suitability map for each species. Then, the continuous habitat suitability maps of all species are converted into binary maps to obtain binary habitat suitability maps.

[0023] S22) Stack the continuous habitat suitability maps of all species and calculate the average value to obtain a summary continuous habitat suitability map. At the same time, stack and sum the binary habitat suitability maps of all species to obtain a summary binary habitat suitability map. In the summary binary habitat suitability map, the grids with values ​​greater than 1 are considered as suitable habitats available to the species in the future.

[0024] Furthermore, step S21) also includes: using the weighted mean of the AUC of the selected species distribution models to generate an integrated species distribution model; again selecting integrated species distribution models with AUC values ​​greater than the second preset threshold from the integrated species distribution models; selecting the target variable from the first bioclimatic variable group; calculating the prediction results of the selected integrated species distribution models and the selected integrated species distribution models without the target variable; calculating the Peirce correlation between the prediction results; and obtaining the importance correlation value corresponding to the target variable.

[0025] Step S22 further includes: calculating the overall contribution percentage of each variable in the first bioclimatic variable group when simulating the presence or absence of a species, based on the importance correlation value of the first bioclimatic variable group in the species distribution model integrated by each species.

[0026] When filtering the first bioclimatic variable group to obtain the second bioclimatic variable group in step S3, the following steps are included: removing variables from the first bioclimatic variable group whose overall contribution percentage is lower than a third preset threshold.

[0027] Furthermore, the specific steps of step S3 include:

[0028] S31) Perform principal component analysis on the second bioclimatic variable group to obtain the principal component scores PCA1 and PCA2 of the current period climate variables, as well as the load of each climate variable, and generate gridded principal component scores PCA1 and PCA2 of the future period climate variables based on the load.

[0029] S32) Divide the principal component scores PCA1 and PCA2 of the current period climate and the principal component scores PCA1 and PCA2 of the future period into equal-width partitions. The PCA1 and PCA2 generated after each partition are uniquely combined into a climate type. Identify the grids in the object area that belong to the same climate type. All grids belonging to the same climate type are climate-similar grids.

[0030] S33) Interpolate the principal component scores PCA1 and PCA2 of the current climate and the principal component scores PCA1 and PCA2 of the future climate. For each climate type in the current period, create an intermediate cost surface for each interpolation grid. Determine the cost value based on the similarity between the principal component score value of the corresponding climate type and other grids. Finally, summarize and retain the minimum cost value of each grid in all time periods and create a minimum cost surface based on the minimum cost value.

[0031] S34) Based on the minimum cost surface, the minimum cost algorithm is used to identify climate-similar grid pairs between the current period and the future period, and the minimum cost path is generated according to the path with the minimum cumulative climate difference between the climate-similar grid pairs.

[0032] S35) The climate rate is obtained by dividing the length of the least cost path by the time interval between the current period and the future period. The climate rate is then used to assess the accessibility of future habitats.

[0033] Furthermore, the specific steps of step S4 include:

[0034] S41) Calculate the composite index value of the climate-similar grid pair based on the climate rate value corresponding to the source grid in the climate-similar grid pair and the habitat suitability value of the target grid in the climate-similar grid pair.

[0035] S42) Using the source and target rasters of a climate-similar raster pair as the starting and target nodes, respectively, the minimum cost path between the climate-similar raster pairs is used as the edge, and the composite index value of the climate-similar raster pair is used as the edge weight. Based on the starting node, target node, and edges, a weighted directed potential range transfer network is constructed within the future habitat suitability area of ​​the target region. Considering that the possibility of species migration is low when the future habitat availability is poor, the potential range transfer network is only constructed within the future habitat suitability area in the summarized binary habitat suitability map.

[0036] Furthermore, the expression for the composite index value in step S41 is as follows:

[0037] RSP s(c-f) =CV-1 s(c-f) ×HS d(f)

[0038] Among them, RSP s(c-f) This refers to the range transfer potential of species from source grids to corresponding target grids under current and future climate change conditions; CV -1 s(c-f) The HS is the reciprocal of the source grid climate rate values ​​under current and future climate change conditions. d(f) It is the continuity habitat suitability value of the target grid in the future period.

[0039] Furthermore, the specific steps of step S5 include:

[0040] Calculate node metrics for each node in the network, then iteratively remove nodes. After each iteration, calculate global network metrics for the newly generated temporary network to evaluate the overall characteristics of the network until the network collapses. Analyze the relationship between the removed nodes and the decline in network metrics, and identify important nodes as critical locations. Use the edges connected to important nodes as critical paths, and construct an ecological network based on the critical locations and critical paths.

[0041] Furthermore, when analyzing the relationship between removed nodes and the decline in network metrics and identifying important nodes as key locations, the following criteria are included: if the global network metrics drop to less than half of the original network after the current node is removed, then the current node is an important node.

[0042] Compared with the prior art, the advantages of the present invention are as follows:

[0043] 1) Based on the mechanism of the impact of climate change on species distribution and migration, this invention assesses the future accessibility and availability of habitats through climate rate and habitat suitability analysis. Subsequently, it effectively integrates the comprehensive assessment results of habitat accessibility and availability to construct a composite index, thereby quantifying the potential of species distribution and migration under climate change and improving the accuracy of predicting the potential distribution and spread of species under the influence of climate change.

[0044] 2) Based on network theory, this invention extracts climate-similar grid pairs and potential range transfer paths as nodes and corridors, respectively. Each corridor is assigned a range transfer potential value, and a weighted directed potential range transfer network is constructed. The network robustness analysis method is used to determine the priority protection ecological nodes and corridors to construct the ecological network, and to delineate the accurate range of species' successful transfer and distribution under climate change. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram illustrating a case of resistance surface construction and climate-similar grid pair identification during climate rate analysis in an embodiment of the present invention.

[0047] Figure 3 This is a schematic diagram illustrating the process of constructing composite indicators in an embodiment of the present invention.

[0048] Figure 4 These are examples and simplified diagrams related to the four node metrics used in network analysis in this embodiment of the invention. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0050] Before introducing the specific implementation methods, let's first explain the relevant concepts.

[0051] Habitat availability and accessibility: The potential for species to shift their distribution range in response to change depends on two key factors: future habitat availability and accessibility. Consequently, two main approaches exist to predict species range shifts under climate change and identify priority protected areas to construct ecological networks. The most common is a model-based approach based on habitat availability assessment. This approach emphasizes the availability of future suitable habitats, assuming that species are more likely to continue existing or establish new populations in climate-similar suitable habitats (Hamann & Aitken 2013). The most commonly used tool in this approach is the species distribution model, which, based on niche theory, uses species distribution data and environmental variables to predict and project the potential distribution area of ​​species according to specific mathematical algorithms (Liang et al. 2018). Species distribution models were first applied in the 1990s to expand the spatial census data of species as a basis for selecting protected areas. They have since been widely used to predict the spatial distribution shifts of species caused by climate change, thereby identifying priority protected areas and constructing ecological networks to develop targeted conservation strategies (Cabeza & M Miranen 2001, Hamann & Aitken 2013, Ma Mtinez-L M Mez et al. 2021, Maz et al. 2021).

[0052] Model-based approaches to habitat accessibility assessment address connectivity issues between isolated habitat patches by establishing ecological corridors, thereby helping species cope with environmental change. The most widely adopted technique involves using species distribution models to analyze habitat suitability within a study area over a specific time period (current or future), and employing minimum cumulative resistance models to construct potential ecological dispersal corridors to propose habitat connectivity solutions for that area over a specific time period (current or future) (Teng et al. 2022). However, under the context of climate change, species ranges may deviate from their current distribution locations. Therefore, recent research has proposed the concept of "climate connectivity"—a measure of the degree to which landscapes facilitate or hinder species from tracking grids similar to current climate conditions in the context of predicted climate change (McGuiMe et al. 2016M, CaMMMll et al. 2018M, PaMks et al. 2020). This method emphasizes accessibility from a species' current distribution area to its future suitable habitat. It assumes that under climate change, if some regions are conducive to dispersal and range transfer, barriers to species movement between habitat patches can be effectively reduced, enhancing inter-regional connectivity (Littlefield et al. 2017, M. CaMMMll et al. 2018). Climate rate is one of the most commonly used indicators for assessing climate connectivity (L. MaMie et al. 2009, M. Hamann et al. 2015).

[0053] Example

[0054] This embodiment proposes an ecological network construction method based on habitat accessibility and availability under climate change. This method is a framework approach that combines climate rate analysis with habitat suitability analysis to quantify future habitat availability and accessibility under climate change. Then, network analysis is used to assess the migration potential of species within the study area and identify priority conservation areas, thus constructing an ecological network adapted to climate change. Figure 1 As shown, it mainly includes the following five steps:

[0055] S1) Obtain target data for the target area, including species distribution points, bioclimatic variables, and non-climatic variables (such as land use and elevation data). Filter the bioclimatic variables in the target data and identify a set of important bioclimatic variables for each species (hereinafter referred to as the first bioclimatic variable group) to reduce variable correlation and prevent overfitting.

[0056] S2) Analyze habitat suitability based on species distribution points, non-climate variables, and the first bioclimate variable group in the target data to assess future habitat availability;

[0057] S3) The first set of bioclimatic variables is screened to obtain a set of bioclimatic variables that are most important to all species (hereinafter referred to as the second set of bioclimatic variables). The second set of bioclimatic variables is used to calculate climate rates and simulate potential range transfer paths to assess future habitat accessibility.

[0058] S4) Construct a composite index based on the assessment results of the availability and accessibility of future habitats, and construct a potential range transfer network based on the assessment results of the accessibility of future habitats. Weight each path in the potential range transfer network with the corresponding composite index value.

[0059] S5) Use network analysis to identify key locations and pathways in the potential range shift network that need priority protection, and use these key locations and pathways to construct an ecological network that can cope with climate change.

[0060] The following is a detailed explanation of each step.

[0061] Step S1 in this embodiment aims to collect the data required for analysis, and its specific steps include:

[0062] S11) Collect species distribution records within the target area, using the spatial resolution used for analysis as the target spatial resolution. Filter the species distribution records to find those matching the target spatial resolution, obtaining species distribution point data. Specifically, collect species distribution records within the target area through citizen science databases or field survey data sources, and remove records with spatial uncertainty greater than the spatial resolution used for analysis. To reduce sampling bias and spatial autocorrelation, delete duplicate geographic location records for each species and filter the dataset to match the used spatial resolution, ensuring only one record is retained for each grid (this step can be implemented using the R package "sMThin"). Summarize and generate species distribution point data.

[0063] S12) Obtain multiple bioclimatic variable data for the current and future periods of the target area. Resample the bioclimatic variable data according to the target spatial resolution. Perform preliminary screening on the sampling results to obtain the first bioclimatic variable group. In this embodiment, obtain 19 bioclimatic variable data for species in the current and future periods of the target area. Resample the original climate data grid to the spatial resolution used for analysis using bilinear interpolation. Given the common multicollinearity problem in climate data, it is necessary to perform preliminary screening on the 19 bioclimatic variables to identify the most important bioclimatic variables that have the greatest impact on each species. First, perform Pearson correlation analysis and variance inflation factor test on the variables to generate M-values ​​and MIF values ​​to check the correlation and collinearity strength between variables. The M-value generated by Pearson correlation analysis represents the correlation coefficient between variables, reflecting the magnitude of the correlation. The larger the absolute value of the correlation coefficient, the higher the correlation. The variance inflation factor test uses the input variables to construct a multiple linear regression model. The generated MIF value is used to measure the severity of multicollinearity in the model. The larger the value, the more severe the multicollinearity. It is generally believed that a MIF greater than 10 indicates that the model has a serious collinearity problem. Secondly, the importance of all bioclimatic variables in explaining species distribution is quantified. A univariate generalized linear model with linear and quadratic terms is constructed for each variable, and the importance of each variable is determined based on the AUC value of its respective univariate model. Then, using a stepwise iterative procedure, in each iteration, less important variables are excluded from pairs of highly correlated variables (i.e., M ≥ 0.8), while variables with MIF > 10 and AUC < 0.6 are also removed. Finally, the iteration continues until the correlation between variables falls below a threshold (M < 0.8) or the minimum number of variables threshold (6) is reached, thus determining the important bioclimatic variables for each species. Therefore, the process of filtering the bioclimatic variables in the target data to obtain the first bioclimatic variable group includes the following steps:

[0064] Pearson correlation analysis and variance inflation factor test were performed on the variables, and pairs of variables with high correlation were selected based on the results of the correlation analysis and variance inflation factor test.

[0065] For each variable, construct a univariate generalized linear model with linear and quadratic terms, and calculate the AUC value of the univariate generalized linear model.

[0066] The less important variables are excluded from the pairs of highly correlated variables, and variables with AUC values ​​less than the first preset threshold (0.6) are also deleted.

[0067] S13) Obtain non-climate variable data within the target area, including land cover and elevation data. Land use data is mainly divided into six types: forest, grassland, cultivated land, urban land, water bodies, and unused land. To maintain consistency with climate data, elevation and land cover data need to be resampled to a spatial resolution consistent with the climate variable data. That is, the non-climate variable data is resampled according to the target spatial resolution.

[0068] Step S2 in this embodiment aims to assess the availability of future habitats, and its specific steps include:

[0069] S21) Based on species distribution point data, land cover and topography from the first bioclimatic variable group and non-climatic variable data are used as environmental prediction variables. On the R integrated model platform "SSDM", a species distribution model is fitted for each species. Species distribution models that exceed the predetermined prediction performance threshold are selected from all species distribution models.

[0070] Considering the uncertainty of simulation by a single algorithm, this embodiment uses 9 algorithms to fit the model for each species: General Additive Model (GAM), Generalized Linear Model (GLM), Multivariate Adaptive Regression Spline (MARS), Classification Tree Analysis (CTA), Generalized Reinforcement Model (GBM), Maximum Entropy (Maxent), Artificial Neural Network (ANN), Random Forest (RF), and Support Vector Machine (SMM). Each algorithm is repeated 10 times to generate 90 models.

[0071] Each algorithm employs a default pseudo-missing data selection strategy (e.g., performing 10 runs of 1000 randomly selected pseudo-missing samples on GLM). For each model, 70% of the initial data is randomly sampled for training, and the remaining 30% is used for model performance evaluation. Only models exceeding a predetermined predictive performance threshold (i.e., AUC ≥ 0.75) are retained. These remaining models are then used to predict habitat suitability for each species in the current and future periods to obtain habitat suitability maps for each species. An ensemble species distribution model (ESDM) is generated using the AUC-weighted mean of the selected species distribution models. The ESDM performance for each species is examined, and ensemble models with AUC values ​​greater than 0.8 and their associated species are retained for subsequent analysis.

[0072] To assess variable importance, the Peirce correlation between the full model and the predictions without the variable is calculated, where a high value indicates a significant impact of the variable on the model. Furthermore, the sensitivity-specificity equation (SES) is used to determine a threshold to convert the habitat suitability map from a continuous (0-1) map to a binary (0,1) map, resulting in a binary habitat suitability map where 0 represents unsuitable habitat and 1 represents suitable habitat.

[0073] Therefore, after screening out models that exceed the predetermined prediction performance threshold, the selected species distribution models are used to predict the continuous habitat suitability map for each species. Then, the continuous habitat suitability maps of all species are converted into binary maps to obtain binary habitat suitability maps.

[0074] After selecting integrated species distribution models with an AUC value greater than the second preset threshold (0.8) from the integrated species distribution models, target variables are selected from the first bioclimatic variable group. For the same selected integrated species distribution model, the prediction results of the model containing all variables and the model without the target variable are calculated, and the Peirce correlation between the prediction results is obtained to obtain the importance correlation value corresponding to the target variable. After predicting the habitat suitability map of each species according to the species distribution model, the continuous habitat suitability maps of all species are converted into binary to obtain binary habitat suitability maps.

[0075] S22) Generate a summary habitat suitability map by combining habitat suitability maps of all species to assess future habitat availability. Specifically, stack the continuous habitat suitability maps of all species and calculate the average of the continuous habitat suitability map for each species at each grid cell to generate a summary continuous habitat suitability map that reflects the overall habitat availability of species in the current and future periods. Simultaneously, sum the binary habitat suitability maps of all species to obtain a summary binary habitat suitability map, where grid cells with values ​​greater than 1 are considered suitable habitats for the species' future use. That is, grid cells with values ​​greater than 1 in the summary binary habitat suitability map are considered suitable habitats for the species' future use. Furthermore, based on the variable importance obtained from the ESDM (Integrated Species Distribution Model) for each species (i.e., the corresponding Peirce correlation values), the average importance of each variable is calculated. Then, the ratio between the square of the average importance and the sum of squares is calculated to obtain the weight of each variable's importance, which represents the overall contribution percentage of each variable in simulating the presence or absence of a species. In other words, based on the importance correlation values ​​of the first bioclimatic variable group in the integrated species distribution model, the overall contribution percentage of each variable in the first bioclimatic variable group in simulating the presence or absence of a species is calculated. Unlike previous methods that assessed future habitat availability based on habitat suitability analysis results of a few species, this invention can simultaneously analyze populations containing a large number of species in a species-specific manner, generating an assessment of future habitat availability for the entire population.

[0076] Step S3 of this embodiment aims to use the most important bioclimatic variables for all species to calculate climate rates and simulate potential range shift paths in order to assess future habitat accessibility. Specifically, it includes the following steps:

[0077] S31) Principal component analysis of climate variables. Before analysis, variables with an overall contribution percentage of less than 1% are removed, retaining only the most important bioclimatic variables for predicting the distribution points of all modeled species. That is, when selecting the second bioclimatic variable group from the first bioclimatic variable group, variables with an overall contribution percentage of less than a third preset threshold (1%) are removed from the first bioclimatic variable group;

[0078] Next, principal component analysis (PCA) is performed on the second bioclimatic variable group to reduce variable dimensionality, transforming the original multiple variables into two new variables. The principal component scores PCA1 and PCA2 for the current period's climate variables, along with the loadings for each climate variable, are obtained. Based on these loadings, gridded principal component scores PCA1 and PCA2 for the future period's climate variables are generated. Subsequent climate rate analysis is based on the gridded principal component scores PCA1 and PCA2 for the current and future periods. Principal component analysis is a commonly used method in this field, and the specific calculation derivation process will not be elaborated upon here.

[0079] S32) Determine climate-similar grids. First, preprocess the gridded principal component scores PCA1 and PCA2 of the current and future climates. Rescale the principal component scores PCA1 and PCA2 of the current climate to the range of 0-100, and readjust the principal component scores PCA1 and PCA2 of the future climate using the same scaling factor. Then, divide the principal component scores PCA1 and PCA2 of the current climate and the principal component scores PCA1 and PCA2 of the future climate into equal-width divisions, specifically by equal-width units of 1 (±0.5). Each division generates a unique combination of PCA1 and PCA2 to form a climate type. Next, for the current and future climates, determine the grids belonging to the same climate type. All grids belonging to the same climate type are considered climate-similar grids. In this embodiment, the range of the habitat suitability map and the grid map used for the climate rate are consistent.

[0080] S33) Establish the minimum cost surface. Interpolate the principal component scores (PCA1 and PCA2) of the current climate and the principal component scores (PCA1 and PCA2) of the future climate. For each climate type in the current period, create an intermediate cost surface for each interpolated grid. The cost value is determined based on the similarity between the PCA value of that climate type and other grids. Specifically, the cost value is determined based on the similarity between the principal component score of the corresponding climate type and other grids. For the two PCA values—PCA1 and PCA2—included in that climate type, calculate the cost value for each grid in the current, future, and intermediate periods to generate the intermediate cost surface. Then, add the two intermediate cost surfaces for each period to generate the intermediate cost surface for that climate type in each period. Finally, summarize and retain the minimum cost value for each grid in all periods to create the minimum cost surface (i.e., the final cost surface). In other words, create the minimum cost surface based on the minimum cost value. Figure 2 The case study further illustrates this point. For simplicity, the process of establishing a minimum cost surface from a single PCA1 raster graph is shown below: (e.g.) Figure 2 As shown, the PCA1 raster surface for the current climate period includes raster values ​​ranging from 12 to 20, while the PCA1 raster surface for the future climate period includes raster values ​​ranging from 18 to 26. First, to reflect the change of PCA values ​​over time, the PCA1 raster surfaces for the current and future climate periods are interpolated to produce the PCA1 raster surface for the intermediate climate period (see...). Figure 2 a); Then, for each climate type in the current period (PCA1 = 18 or 19), according to the following expression (1), the cost value of each grid is calculated based on the principal component score PCA1 grid surface of the current, future and intermediate periods, respectively, to generate the intermediate cost surface for each period (see Figure 2(b and 2c); Finally, for each grid cell, the minimum cost value that retains the intermediate cost surface for all periods is summarized to establish the minimum cost surface (see b and 2c); Figure 2 d).

[0081] cost (i) =1+p×|PCA-PCA (i) | (1)

[0082] Among them, cost () is the cost value of raster i, p is the penalty parameter for dissimilarity (1 in this example), and PCA is the principal component score (PCA) value corresponding to the focused climate type (18 or 19 in the above case). () It is the PCA value corresponding to grid i.

[0083] S34) Identification of Climate-Similar Raster Pairs and Potential Range Transfer Paths. Based on the minimum cost surface, a minimum cost algorithm is used to identify climate-similar raster pairs between the current and future periods—that is, raster pairs with similar climate environments (this step is implemented using the package "gdistance" in R). Specifically, for each raster in the current period (i.e., the source raster), the single raster with the minimum cumulative climate difference (i.e., the target raster) is determined based on the minimum cost surface. A source raster and its target raster form a climate-similar raster pair. A minimum cost path (i.e., a potential range transfer path) is generated based on the path with the minimum cumulative climate difference between this pair of climate-similar raster pairs, see [link to relevant documentation]. Figure 2 e;

[0084] S35) Calculate the climate rate. Divide the length of the minimum-cost path by the time interval between the current and future periods to obtain the climate rate. Use the climate rate to assess future habitat accessibility. The mathematical expression for the climate rate is as follows:

[0085]

[0086] Climate rate is measured in km / yeaM. This value is used to assess the accessibility of future habitats. A higher climate rate indicates that it takes longer for a species to migrate to a suitable habitat within a given timeframe, and may even exceed the species' migration limit, resulting in poorer accessibility.

[0087] Step S4 of this embodiment aims to construct a composite index—range migration potential—based on the assessment of future habitat availability and accessibility, and to build a potential range migration network, where each path is weighted by the composite index value. Specific steps include:

[0088] 1) S41) Based on the climate rate value corresponding to the source grid in the climate-similar grid pair and the habitat suitability value of the target grid in the climate-similar grid pair, calculate the composite index value of the climate-similar grid pair. Specifically, for the feasibility and accessibility of usable habitats in the future, construct a new composite index—Range Transfer Potential (RSP), such as... Figure 3 As shown, this quantifies the potential for species to successfully shift their distribution range in response to climate change. The formula for calculating the range shifting potential (RSP) is as follows:

[0089] RSP s(c-f) =CV -1 s(c-f) ×HS d(f) (3)

[0090] Among them, RSP s(c-f) Under current (c) and future (f) climate change conditions, the range transfer potential of species from source grids (s) to their corresponding target grids (d) is CV. -1 s(c-f) It is the reciprocal of the climate rate values ​​of the source grid(s) under climate change in the current (c) and future (f) periods, HS d(f) It is the continuous habitat suitability value (range 0-1) of the species in the target grid (d) corresponding to the future period (f), which is the value of the corresponding grid in the summary continuous habitat suitability map obtained in step S22;

[0091] S42) Based on complex network theory, the climate-similar grid pairs and potential range transfer paths generated in step 3 are treated as nodes and edges in the network, respectively, and a potential range transfer network is constructed. Specifically, the source grid in the current period and the target grid in the future period of the climate-similar grid pair are taken as the starting node and the target node, respectively. The minimum cost path between the climate-similar grid pairs is taken as the edge, and the composite index value of the climate-similar grid pairs is taken as the weight value of the edge. Based on the starting node, target node, and edge, a weighted directed potential range transfer network is constructed within the future habitat suitability area of ​​the target region. The future habitat suitability area is the region composed of grids with a value greater than 1 in the summarized binary habitat suitability map obtained in step S22. Considering that the possibility of species migration is small when the future habitat availability is poor, the potential range transfer network is only constructed within the future habitat suitability area in the summarized binary habitat suitability map.

[0092] Step S5 of this embodiment aims to use network analysis to identify key locations and pathways requiring priority protection through robustness analysis, and on this basis, construct an ecological network capable of responding to climate change. Specific steps include:

[0093] S51) Using network analysis, the importance of each node is measured by calculating and analyzing local node indices and global network robustness during node removal (this step is performed using GeMhi 0.8.2 and the R package "igMaMh"). Network analysis theory has been widely applied in the field of ecological networks, where nodes generally refer to habitats, and corridors generally refer to functional connections between habitats (such as dispersal and gene flow). However, unlike previous methods that were typically used for ecological network connectivity analysis under current habitat configurations, this embodiment uses it to simulate changes in the connectivity and robustness of potential range transfer networks under the influence of climate change. This helps identify key nodes and pathways that are crucial for species to cope with climate change and require priority protection, and to construct ecological networks, thereby improving the ability of ecological networks to cope with climate change. It is assumed that the removal of a node in a potential range transfer network is equivalent to the destruction of that node's habitat, so network robustness is defined as the network's ability to maintain normal operation for nodes that may lose their usability and accessibility due to disturbances.

[0094] S52) Four node centrality-related metrics are used: degree centrality, in-degree centrality, betweenness centrality, and eigenvector centrality. Each metric measures the importance of a node from different aspects, as shown in Table 1 and... Figure 4 As shown.

[0095] Table 1. Four node indicators used in this invention a and 3 network indicators b Definition and calculation formula

[0096]

[0097]

[0098] a The four node metrics are degree centrality, in-degree centrality, betweenness centrality, and eigenvector centrality.

[0099] b The three network metrics are connectivity robustness, global efficiency, and average path length.

[0100] S53) Calculate node metrics for each node in the network, then iteratively remove nodes. After each iteration, calculate global network metrics for the newly generated temporary network to evaluate the overall characteristics of the network, until the network collapses. Analyze the relationship between the removed nodes and the decline in network metrics to identify important nodes as critical locations. Specifically, to simulate the impact of habitat destruction and identify critical nodes, an iterative node removal method is used for network robustness analysis. The node removal order is determined by four local node metrics. First, calculate the four node metrics for each node in the initial network, then iteratively remove nodes, starting from n=1, removing one node at a time in each iteration until the network collapses. During each iteration, the node removal order includes four scenarios: deleting nodes in descending order of degree centrality, deleting nodes in descending order of in-degree centrality, deleting nodes in descending order of betweenness centrality, and deleting nodes in descending order of eigenvector centrality. After each iteration, calculate three global network metrics for the newly generated temporary network to evaluate the overall characteristics of the network: connectivity robustness, global efficiency, and average path length. When a removed node has a significant impact on network connectivity and resilience, three network metrics will drop sharply, thus identifying key locations where node destruction may cause potential range-shifting networks to collapse due to climate change. Therefore, when analyzing the relationship between removed nodes and the decline in network metrics and identifying important nodes as key locations, the following criteria are used: if the global network metrics drop to less than half of the original network after the current node is removed, then the current node is an important node.

[0101] S54) Edges connecting to important nodes are designated as critical paths, and an ecological network is constructed based on critical locations and critical paths. Specifically, critical locations are determined by the threshold at which the global network index drops below half due to a decrease in the index of a sensitive node (i.e., the node whose index drops the fastest). Edges connecting to important nodes are considered critical paths, which are reclassified into three categories based on climate rate values: low (0-5km), medium (5-10km), and high (above 10km). Based on the identified critical locations and paths, an ecological network capable of responding to climate change is constructed.

[0102] In summary, this invention considers the impact mechanisms of climate change on species distribution and migration, effectively integrates the assessment and analysis of future habitat availability and accessibility under the background of climate change, and constructs an ecological protection network adapted to climate change, thereby maximizing the protection of species' survival and sustainability in the context of climate change. Its advantages over existing technologies are as follows:

[0103] 1) While some models and methods that partially consider habitat availability and accessibility identify priority protected areas and construct ecological networks by overlaying habitat suitability and climate rate analyses, these methods neglect the dynamic range transfer processes implied by these two analyses. This invention, based on the impact mechanism of climate change on species distribution and migration, assesses the future accessibility and availability of habitats within the study area through climate rate and habitat suitability analyses: climate rate analysis identifies climate-similar grid pairs and climate exposure status, indicating the extent to which waterbirds can reach similar future climate habitats from their current distribution areas under climate change; habitat suitability analysis assesses the future suitability of the habitat, reflecting the degree to which similar future climate habitats can sustainably support species survival under climate change. Subsequently, by effectively integrating the comprehensive assessment results of habitat accessibility and availability, a composite index of "climate rate + habitat suitability"—range transfer potential—is constructed. This quantifies the potential for species distribution and migration under climate change, qualitatively elucidates the different risks and opportunities of migration under climate change, and provides a new method and perspective for predicting the potential distribution and dispersal of species under the influence of climate change.

[0104] 2) Current conservation strategies related to climate change adaptation mostly focus on identifying climate refuges or climate connectivity areas. There is currently no technology to combine these two approaches to determine priority protection locations and pathways, forming a complete ecological network capable of responding to climate change. This invention, based on previous assessments of future habitat accessibility and availability within the study area and the quantification of species range shift potential under climate change, employs network analysis to identify key nodes and pathways in the species range shift process as priority protection areas. It constructs a network-based conservation method of "priority protected habitats - range shift pathways," considering the dynamics of species migration. Based on network theory, this invention extracts climate-similar grid pairs and potential range shift pathways as nodes and corridors, respectively. Each corridor is assigned a range shift potential value, constructing a weighted directed potential range shift network. Network robustness analysis is then used to determine priority protected ecological nodes and corridors to construct the ecological network, which is beneficial for promoting successful species range shifts under climate change, thereby avoiding species extinction. Simultaneously, this invention provides a conceptually simple, widely applicable, species-specific, and biologically significant method to address the issue of conservation prioritization. This technology enables data processing and spatial analysis using minimal software such as AMgGIS, GeMhi, and R programming, and is applicable to any region and time frame. The resulting conservation strategies are spatially clearer, providing direct scientific guidance for policymakers. It plays a crucial role in determining the importance of ecological spaces, identifying priority areas for ecological protection, and constructing ecological networks, thereby maximizing the survival and sustainability of species in the context of climate change.

[0105] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention should fall within the protection scope of the present invention.

Claims

1. A method for constructing ecological networks based on habitat accessibility and availability under climate change, characterized in that, Includes the following steps: S1) Obtain target data for the target area, and filter the bioclimatic variables in the target data to obtain the first bioclimatic variable group; S2) Analyze habitat suitability based on species distribution points, non-climate variables, and the first bioclimate variable group in the target data to assess future habitat availability; S3) The first bioclimatic variable group is screened to obtain the second bioclimatic variable group. The second bioclimatic variable group is used to calculate the climate rate and simulate potential range transfer paths to assess the accessibility of future habitats. S4) Construct a composite index based on the assessment results of future habitat availability and accessibility, and construct a potential range transfer network based on the assessment results of future habitat accessibility. Weight each path in the potential range transfer network using the corresponding composite index value. Specific steps include: S41) Based on the climate rate value corresponding to the source grid in the climate-similar grid pair and the habitat suitability value of the target grid in the climate-similar grid pair, calculate the composite index value of the climate-similar grid pair. The expression for the composite index value is as follows: in, RSP s(c-f) It refers to the potential for species to migrate from source grids to corresponding target grids under current and future climate change conditions. CV -1 s(c-f) It is the reciprocal of the source grid climate rate values ​​under current and future climate change conditions. HS d(f) It is the continuity habitat suitability value of the target grid in the future period; S42) Using the source and target graticules of the climate-similar graticule pair as the starting node and target node respectively, the minimum cost path between the climate-similar graticule pair as the edge, and the composite index value of the climate-similar graticule pair as the weight value of the edge, a weighted directed potential range transfer network is constructed in the future habitat suitable area of ​​the object region based on the starting node, target node and edge. S5) Use network analysis to identify key locations and paths in the potential range transfer network that need priority protection, and use the key locations and paths to construct an ecological network.

2. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 1, characterized in that, The specific steps of step S1 include: S11) Collect species distribution records within the target area, and filter out records that match the target spatial resolution from the species distribution records to obtain species distribution point data; S12) Obtain bioclimatic variable data for the current and future periods of the target area, resample the bioclimatic variable data according to the target spatial resolution, and perform preliminary screening of the sampling results to obtain the first bioclimatic variable group. S13) Obtain non-climate variable data within the target area and resample the non-climate variable data according to the target spatial resolution.

3. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 1, characterized in that, When filtering the bioclimatic variables in the target data to obtain the first bioclimatic variable group, the following steps are included: Pearson correlation analysis and variance inflation factor test were performed on the variables, and pairs of variables with high correlation were selected based on the results of the correlation analysis and variance inflation factor test. For each variable, construct a univariate generalized linear model with linear and quadratic terms, and calculate the AUC value of the univariate generalized linear model. The less important variables are excluded from pairs of highly correlated variables, and variables with AUC values ​​less than a first preset threshold are also deleted.

4. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 1, characterized in that, The specific steps of step S2 include: S21) Based on species distribution point data, land cover and topography from the first bioclimatic variable group and non-climatic variable data are used as environmental prediction variables. Species distribution models are fitted for each species. Species distribution models that exceed the predetermined prediction performance threshold are selected from all species distribution models. The selected species distribution models are used to predict the continuous habitat suitability map for each species. Then, the continuous habitat suitability maps of all species are converted into binary maps to obtain binary habitat suitability maps. S22) Stack the continuous habitat suitability maps of all species and calculate the average value to obtain a summary continuous habitat suitability map. At the same time, stack and sum the binary habitat suitability maps of all species to obtain a summary binary habitat suitability map. In the summary binary habitat suitability map, the grids with values ​​greater than 1 are considered as suitable habitats available to the species in the future.

5. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 4, characterized in that, Step S21) further includes: using the weighted mean of the AUC of the selected species distribution models to generate an integrated species distribution model; again selecting integrated species distribution models with AUC values ​​greater than the second preset threshold from the integrated species distribution models; selecting the target variable from the first bioclimatic variable group; calculating the prediction results of the selected integrated species distribution models and the selected integrated species distribution models without the target variable; calculating the Peirce correlation between the prediction results; and obtaining the importance correlation value corresponding to the target variable. Step S22 further includes: calculating the overall contribution percentage of each variable in the first bioclimatic variable group when simulating the presence or absence of a species, based on the importance correlation value of the first bioclimatic variable group in the species distribution model integrated by each species. When filtering the first bioclimatic variable group to obtain the second bioclimatic variable group in step S3, the following steps are included: removing variables from the first bioclimatic variable group whose overall contribution percentage is lower than a third preset threshold.

6. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 1, characterized in that, The specific steps of step S3 include: S31) Perform principal component analysis on the second bioclimatic variable group to obtain the principal component scores PCA1 and PCA2 of the current period climate variables, as well as the load of each climate variable, and generate gridded principal component scores PCA1 and PCA2 of the future period climate variables based on the load. S32) Divide the principal component scores PCA1 and PCA2 of the current climate and the principal component scores PCA1 and PCA2 of the future climate into equal-width partitions. The PCA1 and PCA2 generated after each partition are uniquely combined into a climate type. Identify the grids in the object area that belong to the same climate type. All grids belonging to the same climate type are climate-similar grids. S33) Interpolate the principal component scores PCA1 and PCA2 of the current climate and the principal component scores PCA1 and PCA2 of the future climate. For each climate type in the current period, create an intermediate cost surface for each interpolation grid. Determine the cost value based on the similarity between the principal component score value of the corresponding climate type and other grids. Finally, summarize and retain the minimum cost value of each grid in all time periods and create a minimum cost surface based on the minimum cost value. S34) Based on the minimum cost surface, the minimum cost algorithm is used to identify climate-similar grid pairs between the current period and the future period, and the minimum cost path is generated according to the path with the minimum cumulative climate difference between the climate-similar grid pairs. S35) The climate rate is obtained by dividing the length of the least cost path by the time interval between the current period and the future period. The climate rate is then used to assess the accessibility of future habitats.

7. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 1, characterized in that, The specific steps of step S5 include: Calculate node metrics for each node in the network, then iteratively remove nodes. After each iteration, calculate global network metrics for the newly generated temporary network to evaluate the overall characteristics of the network until the network collapses. Analyze the relationship between the removed nodes and the decline in network metrics, and identify important nodes as critical locations. Use the edges connected to important nodes as critical paths, and construct an ecological network based on the critical locations and critical paths.

8. The method for constructing ecological networks based on habitat accessibility and availability under climate change according to claim 1, characterized in that, When analyzing the relationship between removed nodes and the decline in network metrics, and identifying important nodes as key locations, the following criteria are used: if the global network metrics drop to less than half of the original network after the current node is removed, then the current node is an important node.

Images