Multi-target composite ecological network construction and local synergy method, device and equipment

Abstract

The invention relates to the field of urban planning and management, in particular to a multi-target composite ecological network construction and local synergy method, device and equipment, and innovatively provides a composite ecological network integrating double targets of an urban cold island effect and habitat quality. On the basis of a hierarchical optimization method of'macroscopic network pattern constraint-machine learning interpretation-spatial threshold regulation ', capturing a complex nonlinear relationship between a landscape index and a dual ecological target by utilizing a high-dimensional nonlinear fitting capability of a machine learning model in a heterogeneity composite ecological network; and the SHAP model is introduced to identify a synergistic dominant driving factor and a quantization threshold thereof in the composite ecological network, so that an abstract model prediction result is inverted into a planning control index with clear physical significance, a targeted local synergistic optimization strategy is generated, and the ecological restoration efficiency and the refined control effect are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of urban planning and management, and in particular to a method, apparatus and equipment for constructing and locally enhancing a multi-objective composite ecological network. Background Technology

[0002] Currently, research on mitigating urban heat islands largely focuses on analyzing the cooling effects of single landscape elements (such as water bodies and green spaces) or constructing cold island networks based on the "patch-corridor-matrix" theory. However, existing methods still have significant limitations: First, the construction of traditional cold island networks relies heavily on surface temperature as a physical indicator, lacking comprehensive consideration of the population dimension. This fails to organically combine thermal hazards with thermal exposure and vulnerability, potentially leading to "resource misallocation" in mitigation strategies—over-allocating cooling measures in low-population, high-exposure areas while neglecting high-risk areas with higher temperatures and population densities. Second, existing analyses of driving mechanisms often assume that urban thermal environment driving factors are spatially homogeneous and linear, ignoring the heterogeneous influence of different spatial functions. This fails to effectively distinguish the essential differences in dominant driving factors and thresholds among surfaces as thermal risks, corridors as heat transport channels, and barriers as thermal barriers. Therefore, there is an urgent need for a comprehensive evaluation framework that can integrate "thermal hazards, thermal exposure, and thermal vulnerability," and a technical method that can construct high / low thermal risk networks based on this framework, while accurately analyzing the nonlinear morphological thresholds under different network functions, in order to support the precision and effectiveness of urban thermal risk regulation. Summary of the Invention

[0003] This invention provides a method, apparatus, and equipment for constructing and locally enhancing a multi-objective composite ecological network, which solves the technical problems mentioned above.

[0004] A first aspect of this invention provides a method for constructing a multi-objective composite ecological network, comprising the following steps:

[0005] Step 1: Acquire multi-source remote sensing data of the target city and perform preprocessing. The multi-source remote sensing data includes surface temperature data and geospatial data.

[0006] Step 2: Construct a threat source parameter table and a habitat type sensitivity table based on the geospatial data, calculate habitat quality using the InVEST model, and generate a habitat quality index distribution map;

[0007] Step 3: Perform reverse and forward normalization and spatial overlay on the surface temperature data and the habitat quality index distribution map, respectively, and extract potential composite ecological source areas that simultaneously satisfy the dual attributes of biological habitat and urban cold source;

[0008] Step 4: Perform morphological spatial pattern analysis on the potential composite ecological source areas to identify target composite ecological source areas, and calculate the comprehensive resistance surface to identify composite ecological corridors and composite obstacle points, and construct a multi-objective composite ecological network.

[0009] A second aspect of this invention provides a local enhancement method based on the multi-objective composite ecological network, comprising the following steps:

[0010] Step 5: Identify the dominant driving factors and corresponding threshold ranges of the multi-objective composite ecological network based on an interpretable machine learning model;

[0011] Step 6: Based on the dominant driving factors and their threshold ranges, formulate spatial optimization strategies for key areas of the multi-objective composite ecological network to achieve synergistic improvement of the cold island effect and habitat quality.

[0012] A third aspect of the present invention provides a local enhancement device, including a computer-readable storage medium and a processor, wherein the processor executes a computer program on the computer-readable storage medium to implement the steps of the local enhancement method described above.

[0013] A fourth aspect of this invention provides a local enhancement device, including a data processing module, a habitat calculation module, a network construction module, a threshold identification module, and a strategy generation module.

[0014] The data processing module is used to acquire and preprocess multi-source remote sensing data of the target city, including surface temperature data and geospatial data.

[0015] The habitat calculation module is used to construct a threat source parameter table and a habitat type sensitivity table based on the geospatial data, calculate habitat quality using the InVEST model, and generate a habitat quality index distribution map.

[0016] The network construction module is used to perform reverse and forward normalization and spatial overlay on the surface temperature data and the habitat quality index distribution map, respectively, and to perform morphological spatial pattern analysis to extract composite ecological source areas that simultaneously satisfy the dual attributes of biological habitat and urban cold source; and to calculate the comprehensive resistance surface based on the composite ecological source areas to identify composite ecological corridors and composite obstacle points, and to construct a multi-objective composite ecological network.

[0017] The threshold identification module is used to identify the dominant driving factors and corresponding threshold ranges of the multi-objective composite ecological network based on an interpretable machine learning model.

[0018] The strategy generation module is used to formulate spatial optimization strategies for key areas of the multi-objective composite ecological network based on the dominant driving factors and their threshold ranges, so as to achieve synergistic improvement of the cold island effect and habitat quality.

[0019] The beneficial effects of this invention are as follows: This invention provides a method, apparatus, and device for constructing and locally enhancing a multi-objective composite ecological network, which has the following beneficial effects compared with the prior art:

[0020] (1) It innovatively constructs a composite ecological network that integrates the dual objectives of "urban cooling island effect" and "habitat quality", breaking through the limitations of traditional single-objective planning, achieving a win-win situation of cooling and biodiversity protection within a limited urban space, and improving land use efficiency.

[0021] (2) By combining the complex ecological network with the interpretable machine learning model, a complete technical path from macro pattern identification to micro local efficiency enhancement is proposed. By identifying the source areas, corridors and obstacle points of the complex ecology, the area that needs priority restoration can be accurately located, which greatly improves the efficiency and implementation effect of ecological restoration.

[0022] To make the above-mentioned objects, features and advantages of the invention more apparent and understandable, preferred embodiments of the invention are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart illustrating the method for constructing a multi-objective composite ecological network provided in an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of the integrated resistance surface in an embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram of the composite ecological source area, composite ecological corridor, and composite obstacle point in an embodiment of the present invention.

[0027] Figure 4 A schematic flowchart of the local enhancement method provided in an embodiment of the present invention;

[0028] Figure 5 This is a diagram showing the global feature importance and local interpretation of the landscape index in this embodiment of the invention;

[0029] Figure 6 This is a feature dependency diagram of surface temperature and habitat quality in an embodiment of the present invention;

[0030] Figure 7 This is a schematic diagram of the structure of the local enhancement device provided in the embodiment of the present invention;

[0031] Figure 8 This is a schematic diagram of the structure of the local enhancement device provided in the embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0033] It should be noted that, unless otherwise specified, the various features in the embodiments of this invention can be combined with each other, all of which are within the protection scope of this invention. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this invention do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0034] This invention provides a method, apparatus, and equipment for constructing and locally enhancing a multi-objective composite ecological network. It aims to address the technical challenge of synergistically optimizing thermal environment mitigation and habitat protection in high-density cities. The invention innovatively proposes a composite ecological network integrating the dual objectives of "urban cold island effect" and "habitat quality." Based on a hierarchical optimization method of "macro-network pattern constraints—machine learning interpretation—spatial threshold regulation," within a heterogeneous composite ecological network, the high-dimensional nonlinear fitting capability of machine learning models is used to capture the complex nonlinear relationship between landscape indices and the dual ecological objectives. Furthermore, a SHAP model is introduced to identify the dominant driving factors and their quantification thresholds for synergistic enhancement within the composite ecological network. This transforms abstract model predictions into planning control indicators with clear physical meaning, generating targeted local enhancement optimization strategies, significantly improving the efficiency of ecological restoration and the effectiveness of refined management.

[0035] Figure 1 This is a flowchart illustrating a method for constructing a multi-objective composite ecological network as provided in Example 1. Figure 1 As shown, it includes the following steps:

[0036] Step 1: Acquire multi-source remote sensing data of the target city and perform preprocessing. The multi-source remote sensing data includes surface temperature data and geospatial data.

[0037] Step 2: Construct a threat source parameter table and a habitat type sensitivity table based on the geospatial data, calculate habitat quality using the InVEST model, and generate a habitat quality index distribution map;

[0038] Step 3: Perform reverse and forward normalization and spatial overlay on the surface temperature data and the habitat quality index distribution map, respectively, and simultaneously extract potential composite ecological source areas that satisfy the dual attributes of biological habitat and urban cold source.

[0039] Step 4: Perform morphological spatial pattern analysis on the potential composite ecological source areas to identify target composite ecological source areas, and calculate the comprehensive resistance surface to identify composite ecological corridors and composite obstacle points, and construct a multi-objective composite ecological network.

[0040] The above embodiments provide a method for constructing a multi-objective composite ecological network, which innovatively constructs a composite ecological network that integrates the dual objectives of "urban cooling island effect" and "habitat quality". This breaks through the limitations of traditional single-objective planning, achieves a win-win situation of cooling and biodiversity protection within a limited urban space, and improves land use efficiency.

[0041] The following specific embodiments will be used to describe each step of the above method in detail.

[0042] For example, in a preferred embodiment, step 1, acquiring and preprocessing the surface temperature data, includes the following steps:

[0043] Step 101: Filter the initial remote sensing images that meet the preset conditions in Google Earth Engine;

[0044] Step 102: Mask the low-quality areas in the initial remote sensing image using the quality assessment band, and synthesize single-image surface temperature data using the median.

[0045] Step 103: The single-frame surface temperature data is initially filled using a preset interpolation strategy, and the optimal semi-variation model is selected for secondary filling using the average error and / or standardized root mean square error to generate seamless target surface temperature data.

[0046] Specifically, taking a certain city as an example, firstly, Landsat 8 OLI / TIRS remote sensing images from typical months, such as the hot summer months of July and August, are selected in Google Earth Engine (GEE), and cloud cover is controlled to be within 20%. Then, for low-quality areas in the images, such as cloud-covered areas, quality assessment bands (QA) are used for masking, and data from contemporaneous images of nearby years are used for filling. For areas with missing data, co-kriging interpolation is used. In co-kriging interpolation, environmental factors such as Normalized Difference Vegetation Index (NDVI), DEM, and slope are selected as covariates. Based on the strong correlation between surface temperature and these environmental factors, the interpolation accuracy is improved, ensuring the spatial continuity of the data. Finally, the optimal semi-variogram model is selected by comparing average error close to 0 and standardized root mean square error close to 1, such as spherical, exponential, or Gaussian models, ultimately generating seamless, high-precision target surface temperature data.

[0047] As those skilled in the art will know, the geospatial data includes at least one or more of the following: land use / land cover data from the same or similar time period, digital elevation models, local climate zones, nighttime light data, population data, and road data.

[0048] Specifically, a 10-meter resolution Land Use / Land Cover (LULC) dataset was used, with an overall accuracy of 84.35% ± 0.92%. The Digital Elevation Model (DEM) was derived from a 30-meter resolution dataset obtained from radar topographic mapping tasks, and slope was further calculated using the DEM in ArcGIS. The Local Climate Zone (LCZ) map systematically divides the urban surface into 10 built-up areas and 7 natural cover areas, aiming to provide high-resolution and physically meaningful basic data for urban climate, environment, and energy research.

[0049] The spatial resolution of the nighttime light data is 130m. For ease of storage, the floating-point data was magnified by 10^10 and then stretched to INT32 for storage. Therefore, the radiance of the data was further converted. Population data was selected from the WorldPop population dataset, which has a spatial resolution of 100m. Road data was obtained through OpenStreetMap, and road density was calculated to characterize the intensity of anthropogenic heat emissions.

[0050] The collection of the above data provides a solid data foundation for subsequent habitat quality assessment. Specifically, habitat quality is assessed using the Habitat Quality module of the InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) model.

[0051] First, a threat source parameter table and a habitat type sensitivity table are constructed. As shown in Table 1, the threat source parameter table contains seven key threat factors (threat sources), and for each key threat factor, an impact parameter value is set to represent the stress intensity on different habitat types. The key threat factors include impermeable surfaces, roads, farmland, bare land, nighttime lights, population, and railways. The impact parameter values ​​include maximum impact distance, weight, and spatial decay function form.

[0052] Table 1 Threat Source Parameter Table

[0053] Among them, impermeable surfaces, as the most prominent feature of urbanization, are set as the threat source with the highest weight of 1, with a maximum impact distance of 10 kilometers, and follow an exponential decay law. Similarly, nighttime light and population factors, which represent the intensity of human activities, are set with a weight of 0.8, with a maximum impact distance of 10 kilometers, and also adopt an exponential decay model. Roads, farmland, railways, and bare land are set to a linear decay mode, with railways and roads, and farmland having weights of 0.8 and 0.6, respectively, and bare land having the lowest weight of 0.3, reflecting the differences in the degree of disturbance to natural habitats by different artificial or semi-artificial landscape elements.

[0054] Based on this, as shown in Table 2, the habitat type sensitivity table includes a suitability score (between 0 and 1) for different habitat types as biological habitats and their sensitivity to the aforementioned key threat factors. In this parameter system, forests and water bodies are considered extremely important ecological bases for the region, with the highest habitat suitability value of 1. Shrubs and grasslands are next, at 0.8 and 0.6 respectively. Cultivated land frequently disturbed by human activities and bare land with poor natural conditions have suitability values ​​of only 0.3 and 0.1 respectively. Impermeable surfaces are considered non-habitats, with a suitability value of 0.

[0055] Table 2 Habitat Type Sensitivity Table

[0056] In terms of sensitivity, land use types with high ecological value tend to be more vulnerable to external disturbances. For example, forests have sensitivity coefficients as high as 0.9, 0.8, and 0.8 for roads, impermeable surfaces, and nighttime light, respectively. This means that once such habitats are exposed to the radiation range of threat sources, their habitat quality will significantly degrade. Conversely, impermeable surfaces, as the disturbance source themselves, have a sensitivity of 0 to all types of threats. By coupling the spatial attenuation mechanism of threat sources defined in Table 1 with the habitat sensitivity response mechanism determined in Table 2, this invention can achieve refined simulation and calculation of the spatial pattern of habitat degradation and habitat quality in the study area, and generate a habitat quality index distribution map with a value range of 0-1. The higher the value, the stronger the biodiversity maintenance capacity.

[0057] Then, based on surface temperature data and habitat quality data, a composite ecological network that simultaneously satisfies the dual attributes of biological habitat and urban cold source is constructed. This composite ecological network includes composite ecological source areas, composite ecological corridors, and composite barrier points. The identification method for each element is explained in detail below.

[0058] In a preferred embodiment, for example, identifying a complex ecological source area includes the following steps:

[0059] Step 301: The surface temperature data is reverse normalized using the reverse form of the range normalization method, the habitat quality data is forward normalized, and the normalized cold island intensity raster map and habitat quality raster map are weighted and superimposed to calculate a composite ecological index.

[0060] Specifically, firstly, dimensionless normalization was performed on both land surface temperature (LST) and habitat quality data to eliminate differences in physical dimensions. Considering the negative impact of the urban heat island effect, the inverse form of the range standardization method was used when constructing the cold island intensity index, i.e., through the formula... Surface temperature was reverse-normalized to provide a high numerical value characterizing the strong cold island effect. Simultaneously, habitat quality data were forward-normalized. Based on this, using spatial overlay analysis, the normalized cold island intensity raster layer and habitat quality raster layer were assigned equal weights and weighted overlaid to calculate a composite ecological index reflecting the comprehensive effectiveness of "thermal environment regulation—biodiversity maintenance" in the study area.

[0061] Step 302: The composite ecological index is reclassified using the quantile classification method, and high-value areas are selected as potential composite ecological source areas. These areas exhibit significant dual characteristics of biological habitat and urban cold source in terms of spatial function.

[0062] Step 303: Extract the core patches of the potential complex ecological source area through morphological spatial pattern analysis, and filter out the small and fragmented interfering patches through an area screening mechanism.

[0063] This step involves performing morphological spatial pattern analysis in Guidos Toolbox 3.1 software, with foreground connectivity defined as 8 neighborhoods and edge width as 1, to identify landscape types in the high-value areas selected above. Specifically, high-value areas are defined as foreground elements, and other areas are defined as background elements. Through a series of image processing morphological operations such as erosion, dilation, and reconstruction, the landscape pattern is precisely divided into seven mutually exclusive landscape types: core area, edge area, pore area, bridging area, roundabout area, branch area, and isolated area. In this process, the "core area," which has strong internal stability and anti-interference capabilities, is extracted as the main candidate for composite ecological source areas. An area screening mechanism is set to remove fragmented interference patches with an area of ​​less than 100,000 square meters to ensure that the composite ecological source areas have sufficient habitat scale and radiation effect.

[0064] Finally, step 304 is executed to extract the preferred core patches with high connectivity based on the possible connectivity index and / or the overall connectivity index, and to use them as the target composite ecological source areas.

[0065] Specifically, to further quantify the importance of each candidate patch in the regional ecological network, the Overall Connectivity Index (IIC) and Possible Connectivity Index (PC) were selected as evaluation indicators. The specific values ​​were calculated using Conefor 2.6 software, thereby quantitatively measuring the landscape connectivity of the core area patches after screening. The calculation methods for the IIC and PC indices are as follows:

[0066]

[0067]

[0068] Where n is the core quantity of the cold island or habitat quality, a i and a j It is the area of ​​cores i and j, nl ij A is the number of shortest paths between cores i and j. L It is the largest area of ​​the cold island or habitat quality. It represents the maximum product probability of all paths between core i and j of the cold island or habitat quality.

[0069] Then, the IIC and PC are superimposed using an equal-weighting method to obtain a comprehensive connectivity index, and the top 50 highly connected patches in terms of comprehensive importance are identified. These patches play a key "stepping stone" role in maintaining the ecological flow of the entire region, promoting species migration and cold air diffusion. They not only have extremely high habitat quality and significant cold island effect, but also occupy a pivotal position in the spatial topology. Therefore, they are formally defined as target composite ecological source areas, laying a solid spatial data foundation for the subsequent construction of highly connected composite ecological corridors and optimization of the urban ecological security pattern.

[0070] In a preferred embodiment, as exemplified, identifying composite ecological corridors and composite obstacle points includes the following steps:

[0071] Step 401: Construct the first resistance factor affecting the cold island effect and the second resistance factor affecting habitat quality. Use the factor detector in the geographic detector to obtain the explanatory power q value of each resistance factor on the spatial differentiation characteristics of surface temperature and habitat quality. After normalizing the q value, use it as the weight of each resistance factor and generate a comprehensive resistance surface by weighted superposition.

[0072] Specifically, the first resistance factor includes at least the Local Climate Zone (LCZ), Normalized Difference Vegetation Index (NDVI), Improved Normalized Difference Water Index (MNDWI), Normalized Difference Building Index (NDBI), road density, DEM, and slope. The second resistance factor includes at least land use type, i.e., habitat type, land use / land cover change (LULC), distance from road, distance from impervious surface, nighttime light, DEM, and slope.

[0073] First, the factor detector module from the geographic detector model is introduced to conduct a quantitative attribution analysis of potential resistance factors. Specifically, this step objectively reveals the strength of the resistance factors' influence on the two major ecological processes by calculating the explanatory power (i.e., q-value) of each environmental covariate on the spatial differentiation characteristics of surface temperature and habitat quality. The specific calculation formula is as follows:

[0074]

[0075]

[0076] Where L represents the stratification of LST or impact factor, i.e., classification, and N... h σ and N are the number of classifications in layer h and the entire region, respectively. h 2 and σ 2, h, and SST represent the variances of LST or habitat quality for stratum h and the entire region, respectively. SSW and SST represent the sum of variances within the stratum and the total variance of LST or habitat quality for the entire region, respectively. The q-value ranges from [0,1]. The closer it is to 1, the greater the explanatory power of the influencing factor on the spatial heterogeneity of LST, i.e., the greater the explanatory power... .

[0077] Then, based on the magnitude of the q value, each resistance factor is objectively assigned a corresponding weight coefficient, thereby constructing a single resistance surface that reflects the flow resistance of the urban cold island effect and the resistance to maintaining habitat quality, as shown in Table 3.

[0078] Table 3. Resistance surface coefficients and weights for surface temperature and habitat quality

[0079]

[0080] Subsequently, to eliminate differences in data dimensions and achieve synergistic optimization of dual ecological objectives, the two types of single-item resistance surfaces were standardized and normalized. Then, using an equal-weighted superposition spatial analysis technique, a comprehensive resistance surface integrating the dual constraints of thermal environment regulation and biodiversity conservation was generated. This resistance surface spatially represents the comprehensive cost of ecological flow operation, such as... Figure 2 As shown.

[0081] Then, step 402 is executed, taking the target composite ecological source area as the source point and the comprehensive resistance surface as the cost basis, and calculating the minimum resistance path based on the minimum cumulative resistance model to identify the composite ecological corridor.

[0082] Specifically, the Linkage Mapper toolkit's core Minimum Cumulative Resistance (MCR) model and cost-weighted distance algorithm are used for path optimization calculations, and the identified composite ecological corridors are prioritized. The MCR calculation formula is as follows:

[0083]

[0084] Where MCR is the minimum cumulative resistance model, D ij R is the distance between surfaces i and j of a cold island or high-quality habitat. i It is the connection resistance of the surface i in a cold island or high-quality habitat.

[0085] These corridors form a key link between fragmented habitats and urban cold island patches in terms of spatial topology. They are not only low-resistance pathways for the migration, dispersal and gene exchange of biological species, but also efficient corridors for the transport of urban cold air and heat relief. Thus, in terms of spatial entity, they establish the optimal path network that simultaneously meets the needs of biodiversity conservation and leverages the urban cold island effect.

[0086] Finally, step 403 is executed, which involves scanning the entire area of ​​the composite ecological corridor and its buffer zone based on the preset obstacle detection model and the Barrier Mapper tool. The obstacle points are divided into three levels according to the obstacle value using the quantile classification method, and the obstacle point with the highest obstacle value level is taken as the composite obstacle point.

[0087] Specifically, in the precise identification and diagnosis of complex obstacle points, the preferred embodiment is based on an obstacle point detection model that integrates landscape ecology and circuit theory, using the Barrier Mapper analysis module in the Linkage Mapper tool for quantitative calculations. The core logic of this step lies in simulating the obstruction mechanism of ecological flow in heterogeneous landscapes. By setting a specific search radius and moving window, a full-area scan is performed on the extracted complex ecological corridors and their buffer zones. The aim is to calculate and quantify the resistance value of removing or improving specific grid cells, and the marginal contribution rate for reducing the cumulative distance consumed between ecological source areas and improving overall landscape connectivity. This algorithm can accurately locate key obstacle points that pose a significant bottleneck effect on ecological flow and designate them as target complex obstacle points. For example, the spatial distribution of a complex ecological source area, a complex ecological corridor, and complex obstacle points is as follows: Figure 3 As shown.

[0088] like Figure 4 The present invention also provides a local enhancement method based on the multi-objective composite ecological network described above, comprising the following steps:

[0089] Step 5: Identify the dominant driving factors and corresponding threshold ranges of the multi-objective composite ecological network based on an interpretable machine learning model;

[0090] Step 6: Based on the dominant driving factors and their threshold ranges, formulate spatial optimization strategies for key areas of the multi-objective composite ecological network to achieve synergistic improvement of the cold island effect and habitat quality.

[0091] The local enhancement method provided in the above embodiments combines complex ecological networks and interpretable machine learning models, and proposes a complete technical path from macro-pattern identification to micro-local enhancement. By identifying complex ecological sources, corridors and obstacle points, it can accurately locate areas that need priority restoration, which greatly improves the efficiency and effectiveness of ecological restoration.

[0092] For example, in a preferred embodiment, step 5 identifies the dominant driving factor and the corresponding threshold range, including the following steps:

[0093] Step 501: Construct and train multiple interpretable machine learning models. The input of the interpretable machine learning model is the landscape index of the multi-objective composite ecological network, and the output is the surface temperature and habitat quality.

[0094] Step 502: Perform grid search and hyperparameter tuning on the multiple interpretable machine learning models, and select the machine learning model with the highest accuracy as the optimal prediction model based on preset evaluation metrics.

[0095] Step 503: Establish landscape index response curves at different granularities and obtain the optimal grid resolution that meets the preset numerical stability conditions;

[0096] Step 504: Calculate the SHAP value of each landscape index in the optimal prediction model at the best grid resolution using the SHAP model, i.e., the feature contribution.

[0097] Step 505: Summarize the average absolute values ​​of all samples to show the global importance ranking and local interpretation of each landscape index on surface temperature and habitat quality, and obtain the dominant driving factors.

[0098] Step 506: Establish the SHAP local dependency graph of the dominant driving factor, and generate the key threshold of the dominant driving factor based on the SHAP local dependency graph. The key threshold is the inflection point where the SHAP value changes from negative to positive or from positive to negative.

[0099] Specifically, the landscape indices include landscape percentage (PLAND), maximum patch index (LPI), patch density (PD), edge density (ED), landscape shape index (LSI), average patch area (AREA_MN), average shape index (SHAPE_MN), similarity adjacency percentage (PLADJ), and connectivity index (COHESION), etc.

[0100] Considering the complex nonlinear interactions among elements within urban ecosystems, this invention preferably employs multiple interpretable machine learning models, including Random Forest (RF), Lightweight Gradient Boosting Machine (LightGBM), and Extreme Gradient Boosting (XGBoost), for modeling and analysis. In the model building and training phases, a sample dataset is first extracted and randomly divided into a 70% training set and a 30% test set. Then, 5-fold cross-validation is introduced to perform grid search and hyperparameter tuning on the hyperparameter space of each model to ensure its generalization ability and robustness. Finally, the coefficient of determination (R²) on the test set is used to... 2The comprehensive comparison results of evaluation indicators such as root mean square error (RMSE) and mean square error (MSE) show that the LightGBM model has the best fitting effect. Therefore, the LightGBM model is used for subsequent mechanism explanation and feature analysis.

[0101] In a preferred embodiment, to address the scale effect problem commonly encountered in landscape ecology analysis when establishing the spatial analysis unit of the aforementioned driving factor analysis model, this invention innovatively introduces a suitable granularity discrimination method based on landscape index response curves. Specifically, by utilizing the inflection point characteristics of landscape indices in response curve analysis to determine the optimal grid size, this method effectively elucidates the influence mechanism of landscape indices on surface temperature and habitat quality. The inflection point reflects the sensitivity of landscape indices to spatial scale, aiming to identify and maximize the capture of landscape index information while minimizing data uncertainty. In one embodiment, landscape index response curves at different granularities were generated within a scale range of 100 to 600 meters, with a step size of 100 meters. The analysis results show that most landscape indices exhibit significant numerical stability characteristics when the grid resolution reaches 400 meters × 400 meters. Accordingly, the above embodiment ultimately selected 400 meters as the optimal grid resolution for subsequent spatial analysis and feature extraction, thereby ensuring the spatial accuracy of driving factor quantification and the reliability of its ecological significance.

[0102] To address the interpretability challenge of optimal prediction models, this invention further integrates the SHAP interpretation framework to analyze model results. This model, derived from Shapley value theory in cooperative game theory, can accurately quantify the marginal contribution, dominant role, and threshold effect of each landscape index feature on LST or habitat quality prediction results. In the specific analysis, on the one hand, the global importance of driving factors is quantified and ranked by calculating the average absolute SHAP value of all sample features, identifying the core driving elements affecting the cold island effect and habitat quality—the dominant driving factors. On the other hand, the SHAP local dependency graph is used to analyze the nonlinear response structure of feature variables to the prediction target, thereby transforming the complex black-box output of the machine learning model into response patterns with clear ecological significance, achieving an effective link between the ranking of model driving factors and ecological response patterns in cold island and high habitat quality networks. In one embodiment, the global feature importance and local interpretation graph of the SHAP model are shown below. Figure 5 As shown, the feature dependency graph is as follows Figure 6 As shown.

[0103] Finally, based on the identified dominant driving factors and threshold ranges, the optimization strategy is implemented in specific spatial units.

[0104] As those skilled in the art know, areas with significant cold island effects and areas with high habitat quality exhibit a high degree of spatial coupling and jointly follow a spatial configuration of "matrix-corridor-barrier point". This spatial consistency provides a solid foundation for implementing coordinated ecological optimization.

[0105] For example, in one specific embodiment, the multi-objective composite ecological network constructed by this invention includes 115 composite ecological corridors, 268 composite obstacle points, and composite ecological source areas composed of interconnected blue-green spaces such as lake clusters and forest belts, collectively constructing a gridded "source area-corridor-obstacle point" structure. Compared to the single-objective networks proposed in the prior art, this invention demonstrates enhanced spatial overlap and functional synergy. More importantly, the substantial technical contribution of this invention lies in transforming abstract ecological indicators into operable biodiversity conservation and urban climate adaptation management strategies with clearly defined spatial attributes.

[0106] Regarding the optimization of the morphological configuration of blue-green spaces, embodiments of this invention have confirmed through quantitative analysis that the connectivity and geometric complexity of water bodies are key parameters for maximizing synergistic effects. Figure 5 and Figure 6 Specifically, this invention identifies three key thresholds and their corresponding technical strategies: First, for the similar adjacency percentage (PLADJ), a threshold exceeding 96.59% is set. This indicator indicates that water bodies must be constructed as highly interconnected networks rather than isolated patches to simultaneously trigger robust cooling effects and habitat quality improvement. Based on this, the proposed management strategy prioritizes the restoration of connectivity in river and lake systems to ensure the flow of water and cold air and the continuity of species migration corridors.

[0107] Secondly, regarding the water edge density (ED), it was determined that it needs to be maintained above 94.05 m / hectare. This value is significantly higher than the 75.00 m / hectare required to meet only the cooling needs, indicating that simple geometric boundaries are insufficient to support high-quality habitats. Therefore, optimization strategies should focus on the restoration and naturalization of ecological revetments. By creating complex shoreline structures that include serrated shorelines and interwoven aquatic vegetation, the area of ​​the water-land transition zone can be increased to enhance the evaporative cooling effect and provide diversified ecological niches.

[0108] Third, for the Maximum Patch Index (LPI), this invention sets a threshold greater than 55.26, emphasizing the irreplaceable role of large, continuous core water bodies as regional ecological anchors, and proposes a strategy of implementing strict ecological red line management to prevent habitat fragmentation caused by urban infrastructure construction, so as to maintain the region's thermal buffer capacity and biodiversity pool.

[0109] Regarding the quantitative regulation of landscape component composition and spatial layout, this invention sets precise control indicators based on the baseline capacity of blue-green infrastructure to ensure the effective supply of ecosystem services. Within key functional zones, water coverage must exceed 47.00% to ensure the simultaneous realization of thermal environment and ecological efficiency. This requires a dual-path approach in planning: strictly protecting existing wetlands in suitable areas. Simultaneously, the proportion of forest landscape must be maintained above 27.38%. Given that this threshold is a critical point for supporting habitat function while also considering cooling benefits, urban renewal actions in core urban areas should actively implement a "greening in available spaces" strategy, achieving this threshold through the construction of pocket parks and vertical greening facilities. Furthermore, this invention also discovered a positive synergistic threshold for water patch density (PD), namely, above 6.25 patches / 100 hectares. This reveals that, outside of large water bodies, a high-density matrix composed of small micro-wetlands and rain gardens is crucial for constructing ecological stepping stones, not only promoting species migration but also achieving localized cooling and effectively mitigating the heat island effect. By strictly adhering to the above six quantitative thresholds, this invention achieves a leap from vague ecological concepts to precise spatial intervention technologies.

[0110] To systematically optimize composite ecological corridors, this invention proposes a hierarchical governance scheme with tiered management. For the 115 identified composite ecological corridors, differentiated management is implemented as channels for cold island transmission and species migration. Specifically, to enhance the actual effectiveness of the corridors, this invention proposes strict ecological red line delineation measures for tertiary corridors to prevent erosion and encroachment by urbanization. For primary and secondary corridors connecting gaps in semi-urbanized areas, a "stepping stone" strategy is adopted, reducing ecological resistance through the planning and construction of linear parks and green corridors, thereby building a resilient ecological network system with complete structure and synergistic functions.

[0111] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0112] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the multi-objective composite ecological network construction method and the local efficiency enhancement method described above.

[0113] Figure 7 This is a schematic diagram of the structure of the local enhancement device provided in the embodiment, such as... Figure 7 As shown, it includes a data processing module 100, a habitat calculation module 200, a network construction module 300, a threshold recognition module 400, and a strategy generation module 500.

[0114] The data processing module 100 is used to acquire and preprocess multi-source remote sensing data of the target city, including surface temperature data and geospatial data.

[0115] The habitat calculation module 200 is used to construct a threat source parameter table and a habitat type sensitivity table based on the geospatial data, calculate habitat quality using the InVEST model, and generate a habitat quality index distribution map.

[0116] The network construction module 300 is used to perform reverse and forward normalization and spatial overlay on the surface temperature data and the habitat quality index distribution map, respectively, and to perform morphological spatial pattern analysis to extract composite ecological source areas that simultaneously satisfy the dual attributes of biological habitat and urban cold source; and to calculate the comprehensive resistance surface based on the composite ecological source areas to identify composite ecological corridors and composite obstacle points, and to construct a multi-objective composite ecological network.

[0117] The threshold identification module 400 is used to identify the dominant driving factors and corresponding threshold ranges of the multi-objective composite ecological network based on an interpretable machine learning model.

[0118] The strategy generation module 500 is used to formulate spatial optimization strategies for key areas of the multi-objective composite ecological network based on the dominant driving factors and their threshold ranges, so as to achieve synergistic improvement of the cold island effect and habitat quality.

[0119] This invention also provides a local enhancement device, including a computer-readable storage medium and a processor, wherein the processor executes a computer program on the computer-readable storage medium to implement the steps of the multi-objective composite ecological network construction method and the local enhancement method described above.

[0120] Figure 8 This is a schematic diagram of the structure of the local efficiency enhancement device provided in an embodiment of the present invention, as shown below. Figure 8 As shown, the local enhancement device 8 of this embodiment includes: a processor 80, a readable storage medium 81, and a computer program 82 stored in the readable storage medium 81 and executable on the processor 80. When the processor 80 executes the computer program 82, it implements the steps in the various method embodiments described above, for example... Figure 1 , Figure 4 The steps shown. Alternatively, when the processor 80 executes the computer program 82, it implements the functions of each module in the above-described device embodiments, for example... Figure 7 The functions of the module shown.

[0121] This invention overcomes the shortcomings of traditional ecological planning, which relies more on qualitative analysis than quantitative analysis and employs a "global homogenization strategy," by introducing interpretable machine learning. Through machine learning predictions and deep analysis using SHAP (Shape-Based Approach), this invention not only tells planners "where" needs protection (complex ecological networks) but also "how" to protect them (specific landscape index thresholds). It provides scientific, objective, and actionable guidance for refined ecological restoration in urban renewal, effectively solving the spatial coordination challenge of mitigating urban heat islands and protecting biodiversity.

[0122] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0123] The present invention is not limited to the description in the specification and embodiments, and thus other advantages and modifications can be readily realized by those skilled in the art. Therefore, the present invention is not limited to the specific details, representative devices and illustrated examples shown and described herein without departing from the spirit and scope of the general concept as defined by the claims and their equivalents.

Claims

1. A method for constructing a multi-objective composite ecological network, characterized in that, Includes the following steps: Step 1: Acquire multi-source remote sensing data of the target city and perform preprocessing. The multi-source remote sensing data includes surface temperature data and geospatial data. Step 2: Construct a threat source parameter table and a habitat type sensitivity table based on the geospatial data, calculate habitat quality using the InVEST model, and generate a habitat quality index distribution map; Step 3: Perform reverse and forward normalization and spatial overlay on the surface temperature data and the habitat quality index distribution map, respectively, and perform morphological spatial pattern analysis to extract composite ecological source areas that simultaneously satisfy the dual attributes of biological habitat and urban cold source. Step 4: Calculate the comprehensive resistance surface based on the composite ecological source area to identify composite ecological corridors and composite obstacle points, and construct a multi-objective composite ecological network.

2. The method for constructing a multi-objective composite ecological network according to claim 1, characterized in that, The geospatial data includes at least one or more of the following: land use / land cover data, digital elevation models, local climate zones, nighttime light data, population data, and road data.

3. The method for constructing a multi-objective composite ecological network according to claim 1, characterized in that, Acquiring and preprocessing the surface temperature data includes the following steps: Step 101: Filter the initial remote sensing images that meet the preset conditions in Google Earth Engine; Step 102: Mask the low-quality areas in the initial remote sensing image using the quality assessment band, and synthesize single-image surface temperature data using the median. Step 103: The single-frame surface temperature data is initially filled using a preset interpolation strategy, and the optimal semi-variation model is selected for secondary filling using the average error and / or standardized root mean square error to generate seamless target surface temperature data.

4. The method for constructing a multi-objective composite ecological network according to claim 1, characterized in that, The threat source parameter table contains multiple key threat factors, and sets an impact parameter value for each key threat factor to represent the stress intensity of different habitat types. The key threat factors include impermeable surfaces, roads, farmland, bare land, nighttime lights, population distribution, and railways. The impact parameter values ​​include maximum impact distance, relative weight, and spatial decay function form. The habitat type sensitivity table includes a suitability score for different habitat types as habitats for organisms and their sensitivity to the key threat factors.

5. The method for constructing a multi-objective composite ecological network according to any one of claims 1-4, characterized in that, The identification of complex ecological source areas includes the following steps: Step 301: The surface temperature data is reverse normalized using the reverse form of the range normalization method, the habitat quality data is forward normalized, and the normalized cold island intensity raster map and the habitat quality raster map are weighted and superimposed to calculate and generate a composite ecological index. Step 302: The composite ecological index is reclassified using the quantile classification method, and high-value areas are selected as potential composite ecological source areas. Step 303: Extract the core patches of the potential complex ecological source area through morphological spatial pattern analysis, and filter out the small and fragmented interference patches through an area screening mechanism; Step 304: Extract preferred core patches with high connectivity based on the possible connectivity index and / or the overall connectivity index, and use them as target composite ecological source areas.

6. The method for constructing a multi-objective composite ecological network according to claim 5, characterized in that, The identification of composite ecological corridors and composite obstacle points includes the following steps: Step 401: Construct the first resistance factor affecting the cold island effect and the second resistance factor affecting habitat quality. Use the factor detector in the geographic detector to obtain the explanatory power q value of each resistance factor on the spatial differentiation characteristics of surface temperature and habitat quality. After normalizing the q value, use it as the weight of each resistance factor and generate a comprehensive resistance surface by weighted superposition. Step 402: Using the target composite ecological source area as the source point and the comprehensive resistance surface as the cost basis, and calculating the minimum resistance path based on the minimum cumulative resistance model, the composite ecological corridor is identified. Step 403: Based on the preset obstacle detection model and the Barrier Mapper tool, perform a full-area obstacle scan of the composite ecological corridor and its buffer zone, and take the obstacle with the highest obstacle value level as the composite obstacle.

7. A method for local synergy enhancement, based on a multi-objective composite ecological network constructed according to any one of claims 1-6, characterized in that, Includes the following steps: Step 5: Identify the dominant driving factors and corresponding threshold ranges of the multi-objective composite ecological network based on an interpretable machine learning model; Step 6: Based on the dominant driving factors and their threshold ranges, formulate spatial optimization strategies for key areas of the multi-objective composite ecological network to achieve synergistic improvement of the cold island effect and habitat quality.

8. The local enhancement method according to claim 7, characterized in that, Step 5 involves identifying the dominant driving factor and its corresponding threshold range, including the following steps: Step 501: Construct and train multiple interpretable machine learning models. The input of the interpretable machine learning model is the landscape index of the multi-objective composite ecological network, and the output is the surface temperature and habitat quality. Step 502: Perform grid search and hyperparameter tuning on the multiple interpretable machine learning models, and select the machine learning model with the highest accuracy as the optimal prediction model based on preset evaluation metrics. Step 503: Establish landscape index response curves at different granularities and obtain the optimal grid resolution that meets the preset numerical stability conditions; Step 504: Calculate the SHAP value of each landscape index in the optimal prediction model at the best grid resolution using the SHAP model, i.e., the feature contribution. Step 505: Summarize the average absolute values ​​of all samples to show the global importance ranking and local interpretation of each landscape index on surface temperature and habitat quality, and obtain the dominant driving factors. Step 506: Establish the SHAP local dependency graph of the dominant driving factor, and generate the key threshold of the dominant driving factor based on the SHAP local dependency graph. The key threshold is the inflection point where the SHAP value changes from negative to positive or from positive to negative.

9. A local enhancement device, characterized in that, It includes a data processing module, a habitat computing module, a network construction module, a threshold identification module, and a policy generation module. The data processing module is used to acquire and preprocess multi-source remote sensing data of the target city, including surface temperature data and geospatial data. The habitat calculation module is used to construct a threat source parameter table and a habitat type sensitivity table based on the geospatial data, calculate habitat quality using the InVEST model, and generate a habitat quality index distribution map. The network construction module is used to perform reverse and forward normalization and spatial overlay on the surface temperature data and the habitat quality index distribution map, respectively, and to perform morphological spatial pattern analysis to extract composite ecological source areas that simultaneously satisfy the dual attributes of biological habitat and urban cold source; and to calculate the comprehensive resistance surface based on the composite ecological source areas to identify composite ecological corridors and composite obstacle points, and to construct a multi-objective composite ecological network. The threshold identification module is used to identify the dominant driving factors and corresponding threshold ranges of the multi-objective composite ecological network based on an interpretable machine learning model. The strategy generation module is used to formulate spatial optimization strategies for key areas of the multi-objective composite ecological network based on the dominant driving factors and their threshold ranges, so as to achieve synergistic improvement of the cold island effect and habitat quality.

10. A local enhancement device, comprising a computer-readable storage medium and a processor, characterized in that, When the processor executes the computer program on the computer-readable storage medium, it implements the steps of the local enhancement method described in claim 7 or 8 above.

Images