An ecological protection red line zoning method based on an ecologically important area

Abstract

The application discloses an ecological protection red line zoning method based on ecological important areas and belongs to the technical field of ecological protection red line demarcation. First, a first ecological important area is identified by analyzing the spatial and temporal pattern of ecological service functions of a target area by using an InVEST physical model. Then, a positive sample is constructed by spatially superimposing a nature reserve and ecological factors, a negative sample is determined by combining an outlier detection model, and a machine learning model is trained to identify a second ecological important area. The spatial and temporal quantification of the two types of ecological important areas is carried out, the time sequence trend and the spatial pattern are determined, and then the ecological importance pattern partition and the ecological change trend partition are formed by respectively fusing and clustering. The identification result is obtained by superimposing the two types of partitions, the types and grades are divided, and the ecological protection red line zoning scheme is formed. The application establishes a zoning method system integrating a physical model and a machine learning model, forms a multi-level red line partition method coupling the spatial distribution of ecological importance and the time sequence trend, and provides a targeted protection strategy.

Description

Technical Field

[0001] This invention belongs to the field of ecological protection red line delineation technology, specifically relating to a method for delineating ecological protection red lines based on ecologically important areas. Background Technology

[0002] The accurate identification of ecological protection red lines depends on the scientific assessment of regional ecosystem service functions. Reasonable functional zoning helps clarify protection priorities and limiting measures. Furthermore, red line zoning needs continuous updating based on ecosystem changes and scientific research progress. This urgently requires a scientifically sound framework for red line delineation and dynamic assessment and adjustment. With the continuous advancement of Earth observation technology and computer science, the ability to acquire and process long-term time-series ground observation information has been greatly enhanced. Ecophysical modeling based on remote sensing observations can play a crucial role in red line delineation.

[0003] Currently, there is no unified methodology for delineating ecological protection red lines in China. Most related studies are based on national policies on ecological red line delineation, and the research methods mainly focus on aspects such as ecosystem service functions and ecological security patterns. In terms of research methods, many scholars rely on remote sensing technology to conduct research on ecological protection red line delineation, mainly including remote sensing information extraction and time series analysis. The results are then used to guide ecological protection work. However, previous studies have not fully explored the potential of remote sensing data and lack information extraction and time series analysis of factors specific to ecological red line delineation.

[0004] Therefore, existing ecological protection red line zoning techniques are still in their infancy, relying mainly on qualitative assessments of the current status of regional ecosystem services, and lacking long-term, large-scale quantitative analysis, especially in large areas with complex topographic features.

[0005] To address the aforementioned issues, it is necessary to comprehensively consider the ecological characteristics of the region, including topography, climate, hydrology, and human activities. This involves combining multi-source long-term time-series data with data-driven machine learning algorithms, supplemented by ground-based measurement methods in areas with complex terrain. The goal is to construct a comprehensive ecological and environmental database, conduct scientific and accurate ecological protection red line zoning and effectiveness assessments, and achieve scientific and effective ecological protection and natural resource management. Summary of the Invention

[0006] To address the aforementioned shortcomings in existing technologies, the ecological protection red line zoning method based on ecologically important areas provided by this invention solves the problems of singular ecological characteristics and incomplete assessment systems in the delineation and quantitative evaluation of red lines in geomorphically complex areas. It provides a methodological reference for the optimization and adjustment of provincial red lines and the evaluation of protection effectiveness, and offers technical reference for the ecological protection and development of large regions and important ecological spaces.

[0007] To achieve the aforementioned objectives, the technical solution adopted by this invention is as follows: a method for delineating ecological protection red lines based on ecologically important areas, comprising the following steps:

[0008] S100. Based on the InVEST physical model, conduct spatiotemporal pattern evolution analysis of different ecosystem service functions in the target area to identify the first ecologically important zone in the target area.

[0009] S200. Spatially overlay nature reserves and ecological factors within the target area to construct positive samples, and then analyze and identify negative samples using an outlier detection model; construct a dataset based on the positive and negative samples, and train a machine learning model to identify the second ecologically important area within the target area;

[0010] S300: Spatiotemporal quantification is performed on the first and second ecologically important zones respectively to determine the temporal trends and spatial patterns of the first and second ecologically important zones;

[0011] S400: Integrate and cluster the spatial patterns of the first and second ecologically important zones to form ecological importance pattern zones; integrate and cluster the temporal trends of the first and second ecologically important zones to form ecological change trend zones;

[0012] By combining S500 with overlaying ecological importance pattern zoning and ecological change trend zoning, the ecological protection red line identification results are obtained. The ecological protection red line types and levels are then classified to form an ecological protection red line zoning scheme.

[0013] Further, step S100 includes the following sub-steps:

[0014] S101. Construct a long-term time-series ecological factor dataset for the target region;

[0015] S102. Based on the long-term ecological factor dataset, the InVEST physical model is used to perform long-term time-series calculations on the ecological service functions of the target area, and to analyze the spatial pattern and temporal trend of each ecological service function.

[0016] The ecosystem services include water conservation, soil retention, carbon sequestration services, and habitat quality.

[0017] S103. Conduct spatial autocorrelation analysis on the spatial pattern and temporal trend analysis results of each ecosystem service function;

[0018] S104. Based on the results of spatial autocorrelation analysis, use a geographic detector to analyze the ecological factors of each ecosystem service function, and then determine the driving factors of each ecosystem service function and quantitatively assess their role in the corresponding ecosystem service function.

[0019] S105. Combining the influence of each ecosystem service function and its driving factors, the first ecologically important zone in the target area is obtained.

[0020] Furthermore, in step S103, the spatial pattern and temporal trend analysis results of each ecosystem service function are used as indicators, and the corresponding global autocorrelation coefficient, Moran's I index scatter distribution and local spatial autocorrelation clustering analysis are calculated to obtain the spatial autocorrelation analysis results.

[0021] Further, in step S104, the geographic detector includes a differentiation and factor detector and an interaction detector, and the method for determining the driving factors is as follows:

[0022] The first and second influence values ​​of ecological factors are analyzed by differentiation and factor detectors and interaction detectors, respectively. The third influence value of ecological factors is calculated when the differentiation and factor detectors and interaction detectors are used in interaction analysis. The comprehensive influence value of ecological factors is determined by combining the first, second and third influence values, and then the driving factors are determined.

[0023] The calculation formulas for the first, second, and third influence values ​​of the ecological factors are the same, and are all expressed as:

[0024]

[0025] In the formula, This indicates the first, second, or third influence value of an ecological factor on ecosystem service functions. Indicates the first [unit] within the target area Number of cells in a layer, σ 2 and σ h 2 Let represent the variance of the h-th layer and the variance of the ecosystem service function, respectively. h represents the number of categories or partitions of the ecosystem factor, and N is the number of units of the ecosystem factor within the region.

[0026] Further, step S200 includes the following sub-steps:

[0027] S201. Spatially overlay the nature reserves and ecological factors within the target area. Randomly generate several positive sample points within the nature reserves as positive samples of the dataset, and generate several negative sample points outside the nature reserves. Use the positive sample points and negative sample points to extract the raster values ​​of the ecological factors respectively.

[0028] S202. Train the outlier detection model using the ecological factor raster values ​​extracted from positive sample points. Input the ecological factor raster values ​​extracted from negative sample points into the trained outlier detection model, extract the corresponding ecological factor raster values, perform similarity analysis, and sort to determine the number of negative sample points equal to the number of positive sample points, which are then used as negative samples in the dataset.

[0029] S203. Use the dataset to train and validate the machine learning model to obtain the ecologically important area identification model;

[0030] S204. Divide the target area into several grid units, extract the ecological factor raster value of the center point of each grid unit, and input it into the ecological importance zone identification model to quantify and output the ecological importance value of each location in the target area.

[0031] S205. Use the inverse distance weighting method to interpolate the ecological importance values ​​to generate a fine-resolution raster, thus obtaining the second ecological importance zone of the target area.

[0032] Furthermore, in step S201, the ecological factor is an ecological factor that has a significant impact on ecosystem service functions, obtained through information entropy and collinearity screening.

[0033] The ecological factors include elevation, bedrock depth, soil erodibility, available water content for vegetation, potential evapotranspiration, land use, and average precipitation.

[0034] Furthermore, in step S300, the spatiotemporal quantization includes time dimension quantization and spatial dimension quantization:

[0035] For time dimension quantification, the first and second ecological importance zones are spatially superimposed respectively, and the Mann-Kendall trend test is used to calculate the Theil-Sen Slope value of each ecological service function in the time series, so as to obtain the time series trend of each ecological service function in the first and second ecological importance zones.

[0036] For spatial dimension quantification, the first and second ecologically important zones are spatially superimposed, and the mean value of each cell in the raster is calculated in time series to obtain the spatial pattern of each ecological service function in the first and second ecologically important zones.

[0037] Further, step S400 includes:

[0038] The spatial patterns of various ecological service functions in the first and second ecologically important zones are integrated, and unsupervised clustering is performed in space through a self-organizing mapping network to form an ecological importance pattern partition that represents the distribution of ecological importance levels in the target area.

[0039] The temporal trends of various ecosystem service functions in the first and second ecologically important zones are integrated, and unsupervised clustering is performed in time through a self-organizing mapping network to form an ecological change trend partition that represents the distribution of ecological change trend levels in the target area.

[0040] Further, step S500 includes the following sub-steps:

[0041] S501. Overlay the ecological importance pattern zoning and the ecological change trend zoning, and remove overlapping patches and patches smaller than the set threshold to obtain the ecological protection red line identification results.

[0042] S502. Based on the dominant ecological service functions and geographical locations of different zones in the ecological protection red line identification results, divide the types and levels of ecological protection red lines to form an ecological protection red line pattern.

[0043] S503. The pattern of ecological protection red lines is sampled and filtered to form an ecological protection red line zoning scheme.

[0044] Further, step S503 includes:

[0045] Nearest neighbor resampling is used to improve the image resolution of the ecological protection red line pattern to a uniform scale. Iterative voting filtering is used to filter out independent grid patches in the ecological protection red line pattern after the resolution is improved, so as to obtain an ecological protection red line zoning scheme with clear ecological change trends and ecological protection red line boundaries.

[0046] The ecological protection red line zoning scheme represents the distribution of ecological importance and the distribution of ecological change trends within the target area.

[0047] The beneficial effects of this invention are as follows:

[0048] (1) This invention is based on the InVEST model to carry out long-term time series calculations of water conservation, soil conservation, carbon sequestration services and habitat quality for the core ecological functions of the region, and to modify each factor in combination with the current ecological status of the region to improve regional adaptability; then, spatial autocorrelation analysis is carried out to explore the spatiotemporal evolution trend of each ecological service function in depth, and the driving factor analysis is carried out using a geographic detector; each step is designed according to the progressive logic of "basic → calculation → analysis → attribution → conclusion", with strong logical closed loop and tight connection between links, forming a complete technical chain from source data to the final output of ecologically important areas, avoiding fragmented analysis.

[0049] (2) This invention proposes a data-driven, collaborative method for accurate identification of ecologically important zones. It applies machine learning technology to the direct identification of ecologically important zones, constructing an intelligent, collaborative method for identifying ecologically important zones. This enables data-driven identification of ecologically important zones under complex terrain conditions, effectively improving identification efficiency and accuracy. Specific advantages include:

[0050] The semi-supervised OC-SVM model was used to optimize the negative samples, and the information gain method and multicollinearity analysis were used to screen important ecological factors, which improved the accuracy of subsequent machine learning models in identifying ecologically important areas.

[0051] By directly fitting the nonlinear relationship between ecological factors and important ecological zones using machine learning models, the complex physical modeling process is simplified, and the accuracy of the algorithm and the contribution of factors are quantitatively evaluated.

[0052] (3) This invention establishes an ecological protection red line zoning method system that integrates physical models and machine learning models. Based on the long-term time series ecological importance zone identification results, a multi-level red line zoning method is formed that couples the spatial distribution and temporal trend of ecological importance, and provides targeted protection strategies for the zoning results. This method takes into account both spatial pattern and temporal trend, which can effectively improve the scientificity and accuracy of red line zoning and is an important supplement to traditional ecological zoning methods. Attached Figure Description

[0053] Figure 1 The present invention provides a method for delineating ecological protection red lines based on ecologically important areas. Detailed Implementation

[0054] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0055] This invention provides a method for delineating ecological protection red lines based on ecologically important areas. It comprehensively considers the spatial pattern and temporal trend of ecologically important areas and constructs an ecological protection red line delineation scheme that reflects the distribution of ecological importance and the distribution of ecological change trends.

[0056] refer to Figure 1 This includes the following steps:

[0057] S100. Based on the InVEST physical model, conduct spatiotemporal pattern evolution analysis of different ecosystem service functions in the target area to identify the first ecologically important zone in the target area.

[0058] S200. Spatially overlay nature reserves and ecological factors within the target area to construct positive samples, and then analyze and identify negative samples using an outlier detection model; construct a dataset based on the positive and negative samples, and train a machine learning model to identify the second ecologically important area within the target area;

[0059] S300: Spatiotemporal quantification is performed on the first and second ecologically important zones respectively to determine the temporal trends and spatial patterns of the first and second ecologically important zones;

[0060] S400: Integrate and cluster the spatial patterns of the first and second ecologically important zones to form ecological importance pattern zones; integrate and cluster the temporal trends of the first and second ecologically important zones to form ecological change trend zones;

[0061] By combining S500 with overlaying ecological importance pattern zoning and ecological change trend zoning, the ecological protection red line identification results are obtained. The ecological protection red line types and levels are then classified to form an ecological protection red line zoning scheme.

[0062] In this embodiment, step S100 includes the following sub-steps:

[0063] S101. Construct a long-term time-series ecological factor dataset for the target region;

[0064] S102. Based on the long-term ecological factor dataset, the InVEST physical model is used to perform long-term time-series calculations on the ecological service functions of the target area, and to analyze the spatial pattern and temporal trend of each ecological service function.

[0065] Among them, ecosystem services include water conservation, soil retention, carbon sequestration services, and habitat quality;

[0066] S103. Conduct spatial autocorrelation analysis on the spatial pattern and temporal trend analysis results of each ecosystem service function;

[0067] S104. Based on the results of spatial autocorrelation analysis, use a geographic detector to analyze the ecological factors of each ecosystem service function, and then determine the driving factors of each ecosystem service function and quantitatively assess their role in the corresponding ecosystem service function.

[0068] S105. Combining the influence of each ecosystem service function and its driving factors, the first ecologically important zone in the target area is obtained.

[0069] In step S101, the data in the long-term ecological factor dataset include Zhang coefficient, total annual rainfall, potential evapotranspiration, available water content of plants, soil erodibility, rainfall erosivity, vegetation cover factor, soil and water conservation measures factor, aboveground biomass carbon density, litter organic carbon density, soil organic carbon density, land use type, and stress factor.

[0070] In step S102, for each ecosystem service function, water conservation ecosystem service function refers to the ability of an ecosystem to intercept, infiltrate, store, and purify precipitation through its structure and processes, thereby providing a continuous and stable water resource for human society. This is a key support service of the ecosystem for human society. In this embodiment, the water production module in the InVEST model is used, combined with rainfall, topography, and land use, to calculate the water conservation capacity based on the water balance method. The specific formula is as follows:

[0071]

[0072] In the formula, Represents pixels The actual evaporation, Represents pixels Annual precipitation, Represents pixels Water conservation capacity.

[0073] For land use types with vegetation, the evapotranspiration of the water balance The calculation formula is:

[0074]

[0075] In the formula, Potential evaporation Potential evaporation is a non-physical parameter characterizing natural climate-soil properties. Defined as:

[0076]

[0077] In the formula, The potential evaporation of pixel x. For land use and pixels The relevant vegetation evapotranspiration coefficient.

[0078] For land use types without vegetation, the actual evapotranspiration is also calculated from the reference evapotranspiration, using the following formula:

[0079]

[0080] In the formula, For reference evaporation rate, Evaporation factor for land use.

[0081] Soil conservation's ecological service function effectively protects water conservancy projects, reduces flood risk, extends project lifespan, and ensures water quality safety by reducing sediment content in rivers and water systems. Simultaneously, it maintains soil fertility, reduces nutrient loss, and provides necessary nutrients and a suitable growing environment for plants. The effectiveness of soil conservation is also reflected in the interception and storage capacity of surface cover for rainfall, slowing the speed and intensity of surface runoff, promoting rainwater infiltration, thereby reducing the external driving force of soil erosion and achieving soil protection. This embodiment calculates and analyzes soil conservation using the SDR module within InVEST, calculating soil retention using a modified general soil loss equation. The specific calculation formulas for potential soil erosion under bare land conditions are as follows:

[0082]

[0083] In the formula, Potential soil erosion, , These are soil erodibility factors and rainfall erosivity factors, respectively. , These are vegetation cover and management factors, and slope and slope length factors.

[0084] When vegetation cover and engineering measures are in place, the actual soil erosion volume is calculated using the following formula:

[0085]

[0086] In the formula, This represents the actual amount of soil loss. Indicates engineering measures factors.

[0087] Soil retention capacity is calculated by subtracting actual soil erosion from potential soil erosion capacity.

[0088]

[0089] Carbon sequestration services are a crucial indicator of terrestrial ecosystem service function and a key factor in mitigating global greenhouse gas emissions. Terrestrial ecosystems can fix carbon in soil, vegetation, and leaf litter. This embodiment uses the carbon module in the InVEST model, which estimates carbon sequestration based on land use, aboveground biomass, belowground biomass, soil, and dead organic matter. This calculation method is characterized by model-driven data acquisition, ease of operation, and strong result visualization. The evaluation of carbon sequestration services mainly utilizes the carbon storage module within InVEST, employing dead organic matter, soil carbon pool, aboveground biomass, and belowground biomass in the calculation process. Specifically, it can be represented as follows:

[0090]

[0091]

[0092] In the formula, For carbon density, , These represent the aboveground and underground biomass carbon densities of plants, respectively. , These represent the soil organic carbon density within the litter and the soil layer, respectively; C total Represents total carbon reserves; It represents the total area covered by each land type, and n represents the total number of land use types.

[0093] In the InVEST model, habitat quality and ecosystem service functions are quantified by assessing the sensitivity of landscape types and the intensity of external threats. A habitat quality index is calculated by reflecting habitat suitability and degradation levels. High habitat quality indicates high biodiversity. The habitat quality index based on the InVEST model provides a quantitative method to map the spatial distribution of regional biodiversity service functions. This embodiment uses the habitat quality module from the InVEST model. This index is a dimensionless comprehensive indicator. It generally ranges from 0 to 1, and its value describes habitat quality; higher values ​​correspond to higher habitat quality. Specifically, it can be expressed as:

[0094]

[0095] In the formula, This is the habitat quality index, where r represents the raster layer, y represents all the rasters within r, and j represents a specific land use type. This represents the biomass within grid r. Represents the sum of biomass of all grid cells within r; Distance functions representing habitat and stress sources; This represents the effect of coercion r within y; This represents the sensitivity of j to r; This represents the degree of protection that x receives.

[0096] This yields the habitat quality score at location x. As shown below:

[0097]

[0098] In the formula, z is a constant. It is the half-saturation constant. Score the habitat type corresponding to land use type j

[0099] In step S103, the spatial pattern and temporal trend analysis results of each ecosystem service function are used as indicators, and the corresponding global autocorrelation coefficient, Moran's I index scatter distribution and local spatial autocorrelation cluster analysis are calculated as the spatial autocorrelation analysis results.

[0100] Specifically, spatial autocorrelation analysis is an important development direction of spatial statistical analysis. Its core lies in establishing statistical relationships between data through geographic spatial locations, making the spatial characteristics of data visible. Spatial distribution relationships and global spatial autocorrelation are used to characterize the spatial characteristics of attributes within the entire region.

[0101] Using Moran's I index as a correlation indicator can determine the similarity of attribute values ​​of spatially adjacent units; using hot and cold spots as indicators can evaluate whether there is a phenomenon of low-value clustering and high-value clustering in spatial changes.

[0102] In a specific example of the present invention, the GeoDA tool is used to analyze the spatial autocorrelation and spatial clustering of water conservation, soil retention, carbon sequestration services, and habitat quality.

[0103] In the process of spatial autocorrelation analysis, a spatial connectivity matrix is ​​needed to measure the potential interaction between spatial units. In this invention, the Rook adjacency matrix is ​​used to calculate spatial autocorrelation.

[0104] In step S104, the geographic detector is used to detect the spatial stratification and variation characteristics and patterns of geographic phenomena, thereby identifying the interaction relationships between multiple elements and solving the spatial dependence and heterogeneity problems caused by scale changes that traditional statistics cannot handle.

[0105] To analyze the driving factors of ecosystem service functions and whether there are synergistic or competitive effects among these factors, the geographic detectors used in this embodiment include differentiation and factor detectors and interaction detectors. Among them, differentiation and factor detection can quantitatively analyze the degree of influence of factors on target variables, interaction detection can reveal whether different factors strengthen or weaken the influence of spatial differentiation, risk area detection is used to identify high-risk areas or hotspots of specific phenomena in space, and ecological detection is used to detect the spatial differentiation of ecologically related phenomena and their driving factors.

[0106] Based on the selected geographic detectors, the method for determining the driving factors is as follows:

[0107] The first and second influence values ​​of ecological factors are analyzed by differentiation and factor detectors and interaction detectors, respectively. The third influence value of ecological factors is calculated when the differentiation and factor detectors and interaction detectors are used in interaction analysis. The comprehensive influence value of ecological factors is determined by combining the first, second and third influence values, and then the driving factors are determined.

[0108] The formulas for calculating the first, second, and third influence values ​​of ecological factors are the same, and they are all expressed as:

[0109]

[0110] In the formula, This indicates the first, second, or third influence value of an ecological factor on ecosystem service functions. Indicates the first [unit] within the target area Number of cells in a layer, σ 2 and σ h 2 Let represent the variance of the h-th layer and the variance of the ecosystem service function, respectively. h represents the number of categories or partitions of the ecosystem factor, and N is the number of units of the ecosystem factor within the region.

[0111] Specifically, the differentiation and factor detector calculates the explanatory power of the independent variable factor X on the dependent variable factor Y by analyzing the spatial differentiation of the dependent variable factor Y, i.e., the driving force of the driving factors on geographical phenomena. The interaction detector is used to identify the interaction between different factors, i.e., to assess whether the combined effect of factors X1 and X2 will affect the explanatory power of the dependent variable Y.

[0112] The evaluation method first calculates the q-values ​​of the two factors for Y: q(X1) and q(X2) respectively, then calculates the q-value for their interaction: q(X1∩X2), and compares q(X1), q(X2), and q(X1∩X2). The relationships between the two factors can be divided into 5 categories, as shown in Table 1.

[0113] Table 1: Criteria for Judging Interaction Detection Assessment Factors

[0114]

[0115] Physical models typically target single ecosystem service functions, leading to biases in assessing the importance of ecologically important zones. Therefore, a strategy that integrates multiple ecosystem service functions needs to be considered. Given the powerful data-driven fitting capabilities of data-driven algorithms, the complex nonlinear relationships between designated ecologically important zones and basic ecological factors can be fitted. Based on this, step S200 designs a framework based on a machine learning model to directly identify ecologically important zones through nature reserves.

[0116] Specifically, step S200 of this embodiment includes the following sub-steps:

[0117] S201. Spatially overlay the nature reserves and ecological factors within the target area. Randomly generate several positive sample points within the nature reserves as positive samples of the dataset, and generate several negative sample points outside the nature reserves. Use the positive sample points and negative sample points to extract the raster values ​​of the ecological factors respectively.

[0118] S202. Train the outlier detection model using the ecological factor raster values ​​extracted from positive sample points. Input the ecological factor raster values ​​extracted from negative sample points into the trained outlier detection model, extract the corresponding ecological factor raster values, perform similarity analysis, and sort to determine the number of negative sample points equal to the number of positive sample points, which are then used as negative samples in the dataset.

[0119] S203. Use the dataset to train and validate the machine learning model to obtain the ecologically important area identification model;

[0120] S204. Divide the target area into several grid units, extract the ecological factor raster value of the center point of each grid unit, and input it into the ecological importance zone identification model to quantify and output the ecological importance value of each location in the target area.

[0121] S205. Use the inverse distance weighting method to interpolate the ecological importance values ​​to generate a fine-resolution raster, thus obtaining the second ecological importance zone of the target area.

[0122] In step S201, since nature reserves contain almost all ecologically important areas including water conservation, soil conservation, carbon sequestration services, and biodiversity, nature reserves are selected as positive samples. Positive sample points are randomly generated inside the nature reserves, and multiple negative sample points are generated outside the nature reserves.

[0123] Furthermore, ecological factors are those that have a significant impact on ecosystem service functions, obtained through screening using information entropy and collinearity.

[0124] Specifically, the screening of ecological factors is a crucial prerequisite for the accurate identification of ecologically important areas. Due to the collinearity and differences in information content among factors, the pre-selected factor data needs to be screened to improve the quality of the dataset. Generally, there is no direct correlation between model accuracy and the number of ecological factors; simply increasing the number of ecological factors may not yield more accurate evaluation results and could even affect computational efficiency and increase computational costs. Furthermore, model accuracy is also affected by data redundancy; redundant data is highly susceptible to noise, making it difficult to guarantee model stability and result reliability.

[0125] Considering the above issues, it is necessary to rationally analyze and select ecological factors to ensure that the influence of collinearity among factors is eliminated. This invention analyzes and screens the initially selected ecological factors from the perspectives of information entropy and collinearity, removing factors with low contribution to ecological modeling and high collinearity to optimize basic ecological factors and improve sample quality.

[0126] The information gain ratio is used to evaluate the information content of all pre-selected ecological factors, determining their importance to ecological modeling. The information gain ratio is a statistical method based on information theory to measure the impact of a factor on the target variable. It assesses the discriminative power of ecological factors by calculating the information gain, i.e., the reduction in uncertainty, that a factor brings to the target variable. The information gain ratio of ecological factor A in dataset D is expressed as:

[0127]

[0128] In the formula, Indicates ecological factors The inherent value, This represents the information gain measurement function.

[0129] Based on the above formula, the information gain ratio of each ecological factor can be obtained. This indicator describes the magnitude of its impact on the ecology. If the value is high, it can be considered that a more significant impact has been produced. If the information gain ratio is less than or equal to 0, it means that the ecological factor has made a small contribution to the identification of ecologically important areas.

[0130] Furthermore, if significant linear correlations exist among the factors, it can easily reduce the model's predictive accuracy or prevent effective parameter estimation, adversely affecting subsequent processes. This invention uses the variance inflation factor (VIF) to test for multicollinearity among eight ecological factors. The variance inflation factor test includes two important indicators: the variance inflation factor (VIF) and tolerance (TOL), and these two are inversely related.

[0131] The ecological factors obtained through the above process include elevation, bedrock depth, soil erodibility, available water content for vegetation, potential evapotranspiration, land use, and average precipitation.

[0132] In step S202, a one-class support vector machine (OC-SVM) is used as the outlier detection model to detect outliers and to perform negative sample selection.

[0133] Specifically, single-class support vector machines map samples to a high-dimensional feature space based on kernel functions. When the samples are well clustered, the optimal hyperplane is obtained, and the output result can be obtained. In this way, specific outliers can be detected.

[0134] Based on the detected outliers and the data characteristics of basic ecological factors, the similarity between the negative and positive samples to be screened is examined. Samples with large inter-class differences are set as negative samples to avoid extracting negative samples from potentially important ecological areas, thereby improving the quality of the dataset.

[0135] In step S203, during the training of the machine learning model, for the constructed dataset, seven ecological factors, namely elevation, bedrock depth, soil erodibility, vegetation available water content, potential evapotranspiration, land use, and annual average precipitation, are used as subsequent input factors for the machine. Among them, elevation, bedrock depth, soil erodibility, and vegetation available water content are static data, while potential evapotranspiration, land use, and annual average precipitation are annual dynamic data.

[0136] Furthermore, positive and negative samples are merged to extract factor raster values, and the annual ecological factor data and labels are divided into training and test sets in a 5:5 ratio. Since the generated random points are spatially ordered, the dataset is randomly divided to avoid bias caused by spatial clustering. Finally, both the training and test sets are in the form of two-dimensional arrays. Before inputting the data into the machine learning model, the data in each column of the two-dimensional array needs to be Max-Min normalized to eliminate the influence of differences in the scale of different factors.

[0137] In a specific example of this invention, the selected machine learning models are LGR (Logistic Regression), RF (Random Forest), SVM (Support Vector Machine), and MLP (Multilayer Perceptron), and their specific hyperparameter settings are shown in Table 2:

[0138] Table 2: Hyperparameter Settings for Machine Learning Models

[0139]

[0140] In step S204, the ecological factor raster values ​​of the entire target area are input into the trained machine learning model to generate the ecological importance value identification results of the entire target area, which are presented in point format.

[0141] In step S205, the inverse distance weighting method is used to interpolate the identification results into a 1km resolution grid to generate a refined ecological importance map, thus obtaining the second ecological importance zone of the target area.

[0142] Given that InVEST often focuses on single ecosystem service functions when modeling ecological physical processes, machine learning models, while possessing strong data fitting capabilities and accurate identification, lack interpretability. Selecting only one outcome may lead to omissions in the identification of ecologically important areas. Therefore, a comprehensive approach is needed, starting with spatial patterns and temporal changes, to analyze their regularities and conduct precise zoning of ecological protection red lines.

[0143] Based on the identification results of the first and second ecologically important regions obtained from the aforementioned steps over several consecutive years, in step S300 of this embodiment, spatiotemporal quantization includes temporal dimension quantization and spatial dimension quantization:

[0144] For time dimension quantification, the first and second ecological importance zones are spatially superimposed respectively, and the Mann-Kendall trend test is used to calculate the Theil-Sen Slope value of each ecological service function in the time series, so as to obtain the time series trend of each ecological service function in the first and second ecological importance zones.

[0145] For spatial dimension quantification, the first and second ecologically important zones are spatially superimposed, and the mean value of each cell in the raster is calculated in time series to obtain the spatial pattern of each ecological service function in the first and second ecologically important zones.

[0146] Specifically, in the time dimension, the Mann-Kendall (MK) trend test can detect whether there is a significant monotonic trend in time series data. Its basic idea is to compare the data points in the time series to determine whether there is a significant upward or downward trend in the data.

[0147] It should be noted that in the above spatiotemporal quantification process, when superimposing in space, the first ecologically important area (second ecologically important area) identified in different years is superimposed to obtain a temporal trend and spatial pattern of the ecologically important area that can be reflected over time.

[0148] Step S400 in this embodiment includes:

[0149] The spatial patterns of various ecological service functions in the first and second ecologically important zones are integrated, and unsupervised clustering is performed in space through a self-organizing mapping network to form an ecological importance pattern partition that represents the distribution of ecological importance levels in the target area.

[0150] The temporal trends of various ecosystem service functions in the first and second ecologically important zones are integrated, and unsupervised clustering is performed in time through a self-organizing mapping network to form an ecological change trend partition that represents the distribution of ecological change trend levels in the target area.

[0151] Specifically, considering that physical models are based on ecological process modeling, and the quantitative assessment results of ecosystem service functions take into account the regional ecological evolution characteristics and have considerable interpretability, while machine learning models can accurately fit the complex nonlinear relationship between ecological factors and ecological importance, and can more accurately characterize the ecological importance zone features of a region compared to a single ecosystem service. Therefore, this invention uses a self-organizing map algorithm to perform unsupervised clustering partitioning of the ecological assessment results based on the InVEST model and the machine learning model in space, so as to fully couple the interpretability and identification accuracy of both, and obtain more accurate redline partitioning results.

[0152] Among them, the Self-Organizing Map (SOM) is an unsupervised learning automatic clustering algorithm used to map high-dimensional data to a low-dimensional (usually 2-dimensional) space while preserving the topological structure of the data.

[0153] In response to the problems of large areas of ecologically important zones, unclear protection boundaries, and difficulty in implementing strict management, an assessment of the importance of ecosystem service functions is conducted. This involves analyzing the regional distribution patterns of ecosystem service functions, classifying the importance of ecosystem service functions into levels, and designating areas with high importance levels and upward trends as ecological protection red lines.

[0154] Based on this, step S500 of this embodiment includes the following sub-steps:

[0155] S501. Overlay the ecological importance pattern zoning and the ecological change trend zoning, and remove overlapping patches and patches smaller than the set threshold to obtain the ecological protection red line identification results.

[0156] S502. Based on the dominant ecological service functions and geographical locations of different zones in the ecological protection red line identification results, divide the types and levels of ecological protection red lines to form an ecological protection red line pattern.

[0157] S503. The pattern of ecological protection red lines is sampled and filtered to form an ecological protection red line zoning scheme.

[0158] Step S503 above includes:

[0159] Nearest neighbor resampling is used to improve the image resolution of the ecological protection red line pattern to a uniform scale. Iterative voting filtering is then used to filter out independent raster patches in the ecological protection red line pattern after the resolution is improved, resulting in an ecological protection red line zoning scheme with clear ecological change trends and ecological protection red line boundaries. The ecological protection red line zoning scheme characterizes the distribution of ecological importance and ecological change trends within the target area.

[0160] This invention couples the spatial distribution and temporal trend of ecological importance. Based on the long-term ecological importance zone identification result dataset identified by physical models and machine learning models, the Mann-Kendall test is used to quantify the temporal trend. Then, the results of the two datasets are coupled by a self-organizing map network, and the ecological importance and trend changes are superimposed to form the ecological protection red line identification result. The "five zones and ten cores" ecological protection red line pattern is constructed, and targeted protection measures are explored in combination with the current status of the ecological service functions of each region.

[0161] In a specific embodiment of the present invention, taking Sichuan Province as the target area as an example, the above method is used to delineate ecological protection red line zones.

[0162] In this embodiment, the ecological factor dataset of Sichuan Province was processed based on the aforementioned steps S100~S500, and the ecological protection red line identification results are shown in Table 3. The area of ​​the ecological protection red line is 285042.30 km². 2 .

[0163] Table 3: Identification Results of Ecological Protection Red Lines in Sichuan Province

[0164]

[0165] The results of ecological protection red line zoning in Sichuan Province, combining machine learning and the InVEST model, indicate that the red line areas are mainly distributed in the high-altitude plateau region of northwestern Sichuan and the surrounding mountains of the basin. The distribution pattern can be roughly described as "five axes and ten cores." The "five axes" refer to the Qinling-Bashan Mountains, the southeastern Sichuan karst region, the dry-hot valley of the lower reaches of the Jinsha River, and the Shaluli Mountains, where the ecological protection red lines are distributed in a belt-like pattern. The "ten cores" refer to the Ruoergai Grassland Wetland, the Qiangtang-Yalong River source area, the Dadu River source area, the Sichuan-Yunnan dry-hot valley, the Chengdu Plain basin urban area, the Anning River Valley, the Yalong River source area, the Dadu River source area, the Shaluli Mountains, the Min Mountains, the Daxueshan-Sanjiangyuan area, the Jinpingshan-Hengduan Mountains, the Minshan-Hengduan Mountains, the Liangshan-Xiangling Mountains, and the Wuling Mountains, etc. The red line zoning shows the characteristic of concentrated and contiguous distribution (with mountain and river systems as the framework).

[0166] Based on the above-mentioned "five axes and ten cores" ecological protection red line identification results, the ecological protection red lines are divided into various types, including the Qinba biodiversity conservation ecological protection red line, the Chengdu Plain basin urban carbon sequestration ecological protection red line, the northeastern Sichuan parallel ridge and valley soil conservation ecological protection red line, the southeastern Sichuan karst region soil conservation ecological protection red line, the Wuling Mountain biodiversity conservation ecological protection red line, the Anning River Valley biodiversity conservation and soil conservation ecological protection red line, the Sichuan-Yunnan dry-hot river valley soil conservation ecological protection red line, the Qiangtang-Yalong River source biodiversity conservation and water conservation ecological protection red line, the Shaluli Mountain biodiversity conservation ecological protection red line, the Ruoergai grassland wetland water conservation ecological protection red line, the Dadu River source water conservation ecological protection red line, the Liangshan-Xiangling biodiversity conservation and soil conservation ecological protection red line, the Daxueshan-Sanjiangyuan biodiversity maintenance and water conservation ecological protection red line, the Jinpingshan-Hengduan Mountains biodiversity conservation and soil conservation ecological protection red line, and the Minshan-Hengduan Mountains biodiversity conservation ecological protection red line.

[0167] The ecological protection red line zoning scheme of Sichuan Province formed by the above-mentioned ecological protection lines is shown in Table 4.

[0168] Table 4: Ecological Protection Red Line Zoning Scheme of Sichuan Province

[0169]

[0170] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

[0171] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for delineating ecological protection red lines based on ecologically important zones, characterized in that, Includes the following steps: S100. Based on the InVEST physical model, conduct spatiotemporal pattern evolution analysis of different ecosystem service functions in the target area to identify the first ecologically important zone in the target area. S200. Spatially overlay nature reserves and ecological factors within the target area to construct positive samples, and then analyze and identify negative samples using an outlier detection model; construct a dataset based on the positive and negative samples, and train a machine learning model to identify the second ecologically important area within the target area; S300: Spatiotemporal quantification is performed on the first and second ecologically important zones respectively to determine the temporal trends and spatial patterns of the first and second ecologically important zones; S400: The spatial patterns of the first and second ecologically important zones are integrated and clustered to form ecologically important pattern zones; The temporal trends of the first and second ecologically important zones are integrated and clustered to form ecological change trend zones. By combining S500 with overlaying ecological importance pattern zoning and ecological change trend zoning, the ecological protection red line identification results are obtained. The ecological protection red line types and levels are then classified to form an ecological protection red line zoning scheme.

2. The method for ecological protection red line zoning based on ecologically important zones according to claim 1, characterized in that, Step S100 includes the following sub-steps: S101. Construct a long-term time-series ecological factor dataset for the target region; S102. Based on the long-term ecological factor dataset, the InVEST physical model is used to perform long-term time-series calculations on the ecological service functions of the target area, and to analyze the spatial pattern and temporal trend of each ecological service function. The ecosystem services include water conservation, soil retention, carbon sequestration services, and habitat quality. S103. Conduct spatial autocorrelation analysis on the spatial pattern and temporal trend analysis results of each ecosystem service function; S104. Based on the results of spatial autocorrelation analysis, use a geographic detector to analyze the ecological factors of each ecosystem service function, and then determine the driving factors of each ecosystem service function and quantitatively assess their role in the corresponding ecosystem service function. S105. Combining the influence of each ecosystem service function and its driving factors, the first ecologically important zone in the target area is obtained.

3. The method for ecological protection red line zoning based on ecologically important zones according to claim 2, characterized in that, In step S103, the spatial pattern and temporal trend analysis results of each ecosystem service function are used as indicators to calculate the corresponding global autocorrelation coefficient, Moran's I index scatter distribution and local spatial autocorrelation clustering analysis, which are then used as the spatial autocorrelation analysis results.

4. The method for ecological protection red line zoning based on ecologically important zones according to claim 2, characterized in that, In step S104, the geographic detector includes a differentiation and factor detector and an interaction detector. The method for determining the driving factors is as follows: The first and second influence values ​​of ecological factors are analyzed by differentiation and factor detectors and interaction detectors, respectively. The third influence value of ecological factors is calculated when the differentiation and factor detectors and interaction detectors are used in interaction analysis. The comprehensive influence value of ecological factors is determined by combining the first, second and third influence values, and then the driving factors are determined. The calculation formulas for the first, second, and third influence values ​​of the ecological factors are the same, and are all expressed as: In the formula, This indicates the first, second, or third influence value of an ecological factor on ecosystem service functions. Indicates the first [unit] within the target area Number of cells in a layer, σ 2 and σ h 2 Let represent the variance of the h-th layer and the variance of the ecosystem service function, respectively. h represents the number of categories or partitions of the ecosystem factor, N is the number of units of the ecosystem factor in the region, and L represents the total number of categories or partitions of the ecosystem factor.

5. The method for ecological protection red line zoning based on ecologically important zones according to claim 1, characterized in that, Step S200 includes the following sub-steps: S201. Spatially overlay the nature reserves and ecological factors within the target area. Randomly generate several positive sample points within the nature reserves as positive samples of the dataset, and generate several negative sample points outside the nature reserves. Use the positive sample points and negative sample points to extract the raster values ​​of the ecological factors respectively. S202. Train the outlier detection model using the ecological factor raster values ​​extracted from positive sample points. Input the ecological factor raster values ​​extracted from negative sample points into the trained outlier detection model, extract the corresponding ecological factor raster values, perform similarity analysis, and sort to determine the number of negative sample points equal to the number of positive sample points, which are then used as negative samples in the dataset. S203. Use the dataset to train and validate the machine learning model to obtain the ecologically important area identification model; S204. Divide the target area into several grid units, extract the ecological factor raster value of the center point of each grid unit, and input it into the ecological importance zone identification model to quantify and output the ecological importance value of each location in the target area. S205. Use the inverse distance weighting method to interpolate the ecological importance values ​​to generate a fine-resolution raster, thus obtaining the second ecological importance zone of the target area.

6. The method for ecological protection red line zoning based on ecologically important zones according to claim 5, characterized in that, In step S201, the ecological factor is an ecological factor that has a significant impact on ecosystem service functions, obtained through information entropy and collinearity screening. The ecological factors include elevation, bedrock depth, soil erodibility, available water content for vegetation, potential evapotranspiration, land use, and average precipitation.

7. The method for ecological protection red line zoning based on ecologically important zones according to claim 2, characterized in that, In step S300, the spatiotemporal quantization includes time dimension quantization and spatial dimension quantization: For time dimension quantification, the first and second ecological importance zones are spatially superimposed respectively, and the Mann-Kendall trend test is used to calculate the Theil-Sen Slope value of each ecological service function in the time series, so as to obtain the time series trend of each ecological service function in the first and second ecological importance zones. For spatial dimension quantification, the first and second ecologically important zones are spatially superimposed, and the mean value of each cell in the raster is calculated in time series to obtain the spatial pattern of each ecological service function in the first and second ecologically important zones.

8. The method for ecological protection red line zoning based on ecologically important zones according to claim 1, characterized in that, Step S400 includes: The spatial patterns of various ecological service functions in the first and second ecologically important zones are integrated, and unsupervised clustering is performed in space through a self-organizing mapping network to form an ecological importance pattern partition that represents the distribution of ecological importance levels in the target area. The temporal trends of various ecosystem service functions in the first and second ecologically important zones are integrated, and unsupervised clustering is performed in time through a self-organizing mapping network to form an ecological change trend partition that represents the distribution of ecological change trend levels in the target area.

9. The method for ecological protection red line zoning based on ecologically important zones according to claim 1, characterized in that, Step S500 includes the following sub-steps: S501. Overlay the ecological importance pattern zoning and the ecological change trend zoning, and remove overlapping patches and patches smaller than the set threshold to obtain the ecological protection red line identification results. S502. Based on the dominant ecological service functions and geographical locations of different zones in the ecological protection red line identification results, divide the types and levels of ecological protection red lines to form an ecological protection red line pattern. S503. The pattern of ecological protection red lines is sampled and filtered to form an ecological protection red line zoning scheme.

10. The method for ecological protection red line zoning based on ecologically important zones according to claim 9, characterized in that, Step S503 includes: Nearest neighbor resampling is used to improve the image resolution of the ecological protection red line pattern to a uniform scale. Iterative voting filtering is used to filter out independent grid patches in the ecological protection red line pattern after the resolution is improved, so as to obtain an ecological protection red line zoning scheme with clear ecological change trends and ecological protection red line boundaries. The ecological protection red line zoning scheme represents the distribution of ecological importance and the distribution of ecological change trends within the target area.

Images