A Cross-Regional Carbon Ecological Compensation Optimization Method and System Based on Spatial Resistance Constrained Carbon Sequestration Service Flow Simulation
By using a spatial resistance-constrained carbon sequestration service flow simulation method, the shortcomings in identifying spatial connections and determining compensation standards in cross-regional carbon ecological compensation have been addressed. This has enabled a more accurate compensation scheme, strengthened the incentives for regions with high ecological contributions and the responsibilities of regions with high carbon emissions, and formed a complete technology chain.
Patent Information
- Application Number
- CN202611283759.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-24
- Publication Date
- 2026-11-17
AI Technical Summary
Existing carbon ecological compensation technologies are insufficient in identifying spatial linkages of cross-regional carbon sequestration services and determining compensation standards. They cannot accurately reflect the actual linkage intensity and the differences in ecological contributions between supply and demand regions under complex ecological and geographical conditions, leading to a deviation between compensation results and actual contributions and responsibilities.
A carbon sequestration service flow simulation method based on spatial resistance constraints is adopted. By constructing a comprehensive spatial resistance surface and introducing a distance decay function from a comparative ecological radiation force model, combined with machine learning models and interpretability analysis, the intensity of ecological linkages and compensation amounts are determined to achieve differentiated compensation.
Accurately identifying the spatial connections of cross-regional carbon ecological compensation improves the rationality and feasibility of compensation schemes, reflects the incentives of regions with high ecological contributions and the responsibilities of regions with high carbon emissions, and forms a complete technical chain to support the formulation of cross-regional ecological compensation schemes.
Smart Images

Figure CN123117438A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of ecological environment and carbon compensation technology, specifically involving a cross-regional carbon ecological compensation optimization method and system based on spatial resistance-constrained carbon sequestration service flow simulation. Background Technology
[0002] Carbon sequestration services are an important type of ecosystem service. Terrestrial ecosystems fix carbon dioxide from the atmosphere through vegetation photosynthesis, playing a vital role in regulating climate change and stabilizing the global carbon cycle. Due to significant differences in regional natural endowments and economic activity levels, ecosystem carbon sequestration supply and regional carbon emission demand are generally spatially separated, forming a spatial mismatch pattern of concentrated supply and concentrated demand. This leads to cross-regional ecological externalities and environmental responsibility allocation issues.
[0003] Ecological compensation is an important policy tool for correcting ecological externalities. Existing research on carbon ecological compensation is mainly based on regional carbon budget accounting. It identifies carbon surplus and carbon deficit regions by calculating the difference between the supply of carbon sequestration services and the demand for carbon emissions, and determines the compensation amount by multiplying the carbon surplus or deficit by the carbon price. This type of method is a static supply and demand assessment based on statistics, which can reflect the carbon supply and demand imbalance at the regional scale, but it is difficult to characterize the spatial transfer process of carbon ecological services. It cannot determine which carbon surplus regions actually provide the carbon sequestration services needed by carbon deficit regions, nor can it determine the actual contribution of different carbon surplus regions to the same carbon deficit region, nor can it determine the distribution relationship of carbon sequestration service contributions among multiple carbon surplus regions and multiple carbon deficit regions. Furthermore, this type of method implicitly assumes that the theoretical regional surplus can be completely transferred outwards, failing to consider the distance attenuation and geographical barriers encountered during the spatial transfer of carbon sequestration services, which can easily lead to biases in the identification of ecological contributions.
[0004] To characterize the spatial connections between ecosystem service supply and benefit regions, existing research has introduced ecosystem service flow methods, among which the Comparative Ecological Radiation Force (CERF) model has been used to describe the strength of ecological connections between supply and demand regions (Chen et al., 2014). This model, drawing inspiration from gravity models, posits that the strength of ecological connections between regions is influenced by supply capacity, demand levels, and spatial distance, and employs an exponential distance decay function to describe the decrease in ecosystem service impact with increasing distance. However, traditional CERF models primarily rely on Euclidean distance for distance decay calculations, failing to adequately consider the spatial barrier effects caused by differences in land use types, topography, and climate between different regions. Under complex ecological and geographical conditions, these models struggle to accurately reflect the actual strength of connections between ecosystem service supply and demand regions.
[0005] Regarding the determination of compensation standards, existing horizontal ecological compensation models typically calculate compensation amounts directly based on carbon sequestration services and carbon prices, mainly reflecting differences in the quantity of ecosystem services. They do not fully consider the differences in ecological contribution capacity among different supply regions, as well as the differences in responsibility among different demand regions due to carbon demand levels and dependence on external carbon sinks. This makes it difficult for compensation standards to simultaneously reflect ecological contributions on the supply side and payment responsibilities on the demand side, resulting in deviations between compensation results and actual ecological contributions and payment responsibilities in the region.
[0006] Therefore, existing technologies still have shortcomings in identifying spatial connections of cross-regional carbon sequestration services and in rationally determining compensation standards, and a carbon ecological compensation technology solution that can overcome the above problems is needed. Summary of the Invention
[0007] To address the shortcomings and deficiencies of existing technologies, this invention provides a cross-regional carbon ecological compensation optimization method and system based on spatial resistance-constrained carbon sequestration service flow simulation. This invention identifies carbon surplus and carbon deficit regions based on carbon sequestration service supply and carbon emissions, designating them as supply and demand regions, respectively. A comprehensive spatial resistance surface is constructed to determine the spatial resistance coefficient between each supply and demand region. This coefficient is then incorporated into the distance decay function of a comparative ecological radiation force model. The product of geometric distance and spatial resistance coefficient is used as the equivalent transmission distance for distance decay calculation, yielding the ecological connection strength. Under the constraints of carbon sequestration capacity in the supply region and carbon demand in the demand region, the ecological connection strength is converted into the actual amount of carbon sequestration services provided by each supply region to each demand region. Machine learning models are trained using regional carbon sequestration capacity and regional carbon emissions as prediction targets. Interpretability analysis is used to determine key driving factors and their contribution weights, thereby constructing a supply region ecological contribution index and a demand region payment responsibility index. A two-factor adjustment coefficient is constructed based on the deviation of these two indices from their respective overall average levels. This coefficient is used to differentiate the basic compensation amount determined based on carbon sequestration service volume and carbon price, outputting compensation scheme data. This achieves spatial identification and differentiated compensation of cross-regional carbon sequestration service contribution relationships.
[0008] To achieve the above objectives, the present invention adopts the following technical solution:
[0009] A cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation, the method being executed by a computer system, includes the following steps:
[0010] The study obtains the carbon sequestration service supply and carbon emissions of each spatial unit within the study area. The difference between the two is determined as the net carbon sequestration supply and demand. Spatial units with positive net carbon sequestration supply and demand are identified as carbon surplus areas and serve as carbon sequestration service supply areas, while spatial units with negative net carbon sequestration supply and demand are identified as carbon deficit areas and serve as carbon sequestration service demand areas. Carbon sequestration service supply characterizes the ecosystem carbon sequestration level of a spatial unit, while carbon emissions characterize the carbon sequestration service demand level of a spatial unit. Net carbon sequestration supply and demand reflects the balance between the two; a positive value indicates that the corresponding spatial unit can provide carbon sequestration services to other spatial units after meeting its own needs, while a negative value indicates that the corresponding spatial unit needs to rely on carbon sequestration services provided by other spatial units to make up for the demand gap.
[0011] A comprehensive spatial resistance surface is constructed by selecting resistance factors that influence the spatial transmission of carbon sequestration services. Based on this comprehensive spatial resistance surface, the spatial resistance coefficient between each supply region and each demand region is determined. Carbon sequestration services are hindered by conditions such as land use, topography, and climate during spatial transmission. The comprehensive spatial resistance surface is used to characterize the spatial distribution of these hindering effects, and the spatial resistance coefficient is used to characterize the strength of the comprehensive barriers encountered in the spatial transmission of carbon sequestration services between supply and demand regions.
[0012] By incorporating the spatial resistance coefficient into the distance attenuation function of the comparative ecological radiation force model, and using the product of the geometric distance between the supply and demand areas and the spatial resistance coefficient as the equivalent transmission distance for distance attenuation calculation, the ecological connection strength between each supply and demand area is obtained. Thus, the spatial barrier between supply and demand areas is converted into an amplified equivalent transmission distance. The distance attenuation calculation simultaneously reflects geometric distance attenuation and geographical environmental barriers. Compared to the traditional comparative ecological radiation force model that only calculates distance attenuation based on Euclidean distance, it can more accurately reflect the actual connection strength between supply and demand areas under complex ecological and geographical conditions.
[0013] Based on the aforementioned ecological linkage strength, and combining the carbon sequestration capacity constraints of the supply areas and the carbon demand constraints of the demand areas, the actual amount of carbon sequestration services provided by each supply area to each demand area is determined. The carbon sequestration capacity constraint of the supply area means that the amount of carbon sequestration services output by the supply area does not exceed the amount of carbon sequestration it can provide. The carbon demand constraint of the demand area means that the amount of carbon sequestration services received by the demand area does not exceed its own carbon demand. Through these two constraints, the allocation of carbon sequestration services simultaneously satisfies both supply feasibility and demand rationality.
[0014] Machine learning models were trained using regional carbon sequestration capacity and regional carbon emissions as prediction targets, respectively. Interpretability analysis was used to determine key driving factors and their contribution weights, thereby constructing an ecological contribution index for the supply region and a payment responsibility index for the demand region. The interpretability analysis was used to analyze the contribution of each feature variable used in training the machine learning models to the prediction results, with key driving factors being the feature variables with higher contribution levels. The ecological contribution index characterizes the relative level of carbon sequestration contribution capacity in the supply region, while the payment responsibility index characterizes the relative ecological compensation responsibility of the demand region due to carbon emission levels and dependence on external carbon sinks.
[0015] A two-factor adjustment coefficient is constructed based on the deviations of the ecological contribution index from the average ecological contribution index of all supply areas and the deviations of the payment responsibility index from the average payment responsibility index of all demand areas. Combined with the carbon sequestration service volume and carbon price, the compensation amount paid by each demand area to each supply area is calculated, outputting compensation scheme data that includes the correspondence between supply areas, demand areas, carbon sequestration service volume, and compensation amount. The deviation reflects the degree of difference between the ecological contribution and payment responsibility of each region relative to the regional average. The two-factor adjustment coefficient differentially adjusts the basic compensation amount determined based on carbon sequestration service volume and carbon price, strengthening the compensation incentive for supply areas with high ecological contributions while reflecting the relative ecological responsibility that demand areas with high carbon emissions and high dependence on external carbon sinks should bear.
[0016] Furthermore, the comprehensive spatial resistance surface is constructed as follows: Multiple resistance factors influencing the spatial transfer of carbon sequestration services are selected, including land use type, elevation, slope, and wind speed. Land use type characterizes the relative spatial constraints under different land cover and human activity backgrounds; elevation and slope characterize the relative spatial constraints formed by topographic complexity; and wind speed characterizes the relative conditions of spatial connections for carbon sequestration services under different atmospheric environmental backgrounds. A judgment matrix is constructed by pairwise comparisons of the multiple resistance factors using the analytic hierarchy process (AHP). After a consistency test, the weights of each factor are determined, where the consistency test determines whether the consistency ratio of the judgment matrix is less than a preset threshold, for example, 0.1. The graded resistance values of each resistance factor are determined, and the graded resistance values of each resistance factor are weighted and superimposed according to the weights to obtain the comprehensive spatial resistance surface. Each resistance factor is assigned a corresponding resistance value according to a preset grading standard. The resistance value characterizes the degree of relative spatial constraint formed by different environmental conditions on the ecological connections between carbon sequestration service supply and demand areas; the larger the resistance value, the stronger the relative spatial constraint.
[0017] Furthermore, the spatial resistance coefficient is the average value of the combined resistance surface value of each grid along the line connecting the geometric centers of the supply area and the demand area. The spatial resistance coefficient characterizes the amplification factor of the equivalent transmission distance, and its value ranges from, for example, 1 to 7. When the combined resistance surface value of each grid along the line is at the lowest resistance level, the spatial resistance coefficient takes the lowest value, and the distance attenuation calculation degenerates to the case of only considering geometric distance attenuation.
[0018] Furthermore, the strength of the ecological link is calculated using the following formula:
[0019]
[0020] in, For the supply area With demand area The strength of ecological connections between them For the supply area With demand area The Euclidean distance between them The space drag coefficient is... This represents the maximum distance between all supply and demand pairs within the study area. For the supply area Net supply and demand for carbon sequestration For demand area The absolute values of net carbon sequestration supply and demand. The exponential term of the formula describes the combined attenuating effect of geometric distance and spatial resistance, while the denominator reflects the moderating effect of the relative structure of supply and demand on the strength of ecological linkages.
[0021] Further, determining the actual carbon sequestration services provided by each supply region to each demand region includes: for each supply region, normalizing the intensity of its ecological connections to each demand region to obtain an allocation ratio, and allocating the net carbon sequestration supply and demand of the supply region to each demand region according to the allocation ratio to obtain an initial allocated service volume; for each demand region, when the sum of the initial allocated services provided by all supply regions to the demand region is greater than the absolute value of the net carbon sequestration supply and demand of the demand region, compressing each initial allocated service volume proportionally according to the ratio of the absolute value of the net carbon sequestration supply and demand to the sum of the initial allocated services; when the sum of the initial allocated services is not greater than the absolute value of the net carbon sequestration supply and demand of the demand region, keeping the initial allocated services unchanged; the compressed or unchanged initial allocated services volume is taken as the actual carbon sequestration services provided by each supply region to each demand region. Therefore, the sum of carbon sequestration services output by any supply region to each demand region shall not exceed the net carbon sequestration supply and demand of that supply region, and the sum of carbon sequestration services received by any demand region shall not exceed the absolute value of the net carbon sequestration supply and demand of that demand region. The spatial allocation of carbon sequestration services is constrained by both the carbon sequestration capacity on the supply side and the carbon demand on the demand side.
[0022] Furthermore, the machine learning model is a CatBoost model. Before training, Pearson correlation analysis is used to screen candidate features: based on the correlation between each candidate feature and the prediction target, candidate features with weak explanatory power are eliminated. Weak explanatory power refers to a low degree of correlation between the corresponding candidate feature and the prediction target. When the correlation coefficient between any two candidate features is greater than a preset collinearity threshold, only candidate features with a higher correlation to the prediction target are retained to avoid multicollinearity among features. The preset collinearity threshold is, for example, 0.8. When regional carbon sequestration capacity is the prediction target, net primary productivity and soil heterotrophic respiration are removed from the candidate features before the screening. This is because there is a deterministic algebraic relationship between net primary productivity and soil heterotrophic respiration and regional carbon sequestration capacity. Retaining them would cause the machine learning model to fit this deterministic algebraic relationship rather than the true driving mechanism of regional carbon sequestration capacity.
[0023] Furthermore, the interpretability analysis employs the SHAP method, including: calculating the average SHAP importance of each feature variable used in training the machine learning model; selecting key driving factors with high importance rankings, such as the top three key driving factors; normalizing the SHAP importance values of each key driving factor to obtain relative weights; and weighting and summing the relative weights with the standardized values of the corresponding key driving factors after min-max normalization to construct the ecological contribution index and the payment responsibility index, respectively, where min-max normalization maps the values of the corresponding key driving factors to the [0,1] interval. Thus, the selection of key driving factors and the determination of their weights are based on their contribution to the prediction results, avoiding the subjectivity caused by manually setting indicator weights.
[0024] Furthermore, the two-factor adjustment coefficient is calculated using the following formula:
[0025]
[0026] in, For the supply area With demand area The two-factor adjustment coefficient between them; and The adjustment weights for the ecological contribution index and the payment responsibility index are respectively. and equal; For the supply area Ecological contribution index. The average ecological contribution index of all supply areas; For demand area Payment responsibility index This represents the average payment responsibility index across all demand areas. and These are the lower and upper limits of the adjustment coefficient, respectively. This is a truncation function, used when the value of the independent variable is less than... or greater than The corresponding boundary value is taken at the time; the compensation amount is the product of the basic compensation amount and the dual-factor adjustment coefficient, the basic compensation amount is the product of the carbon sequestration service volume and the carbon price, and the carbon price can be determined based on the carbon trading market price. and Equal factors ensure a balance between the regulatory effects of ecological contribution and payment responsibility, preventing compensation results from being biased towards the supply or demand side. When the ecological contribution index of a supply area is higher than the average of all supply areas, the dual-factor adjustment coefficient is greater than 1, and the compensation level of that supply area increases accordingly. When the payment responsibility index of a demand area is higher than the average of all demand areas, the amount paid by that demand area increases accordingly.
[0027] Furthermore, and All are taken as 0.5, the lower limit Take 0.5, the upper limit We set the value to 1.5. Under the condition that the values of the key driving factors are mapped to the [0,1] interval after min-max normalization, the absolute values of the deviations of the ecological contribution index and the payment responsibility index from their respective average values do not exceed 1. After multiplying by the adjustment weight of 0.5, the single-factor adjustment amount does not exceed 0.5. The theoretical range of the dual-factor adjustment coefficient in the extreme case of the two factors moving in the same direction is [0,2]. By using the truncation function, the dual-factor adjustment coefficient is constrained to [0.5,1.5] so that the adjustment range of the basic compensation amount does not exceed 50%, ensuring that the compensation result is mainly based on the actual contribution of carbon sequestration services and remains stable.
[0028] This invention also provides a cross-regional carbon ecological compensation optimization system based on spatial resistance-constrained carbon sequestration service flow simulation, comprising:
[0029] The carbon supply and demand identification module is used to obtain the carbon sequestration service supply and carbon emissions of each spatial unit, determine the net supply and demand of carbon sequestration, and identify supply areas and demand areas.
[0030] The integrated resistance surface construction module is used to construct an integrated spatial resistance surface and determine the spatial resistance coefficient between each supply area and each demand area.
[0031] The carbon sequestration service flow simulation module is used to introduce the spatial resistance coefficient into the distance decay function of the comparative ecological radiation force model, and use the product of geometric distance and spatial resistance coefficient as the equivalent transmission distance to calculate distance decay, so as to obtain the ecological connection strength between each supply area and each demand area.
[0032] The carbon sequestration service conversion module is used to determine the actual amount of carbon sequestration services provided by each supply area to each demand area based on the ecological linkage strength, combined with the carbon sequestration capacity constraints of the supply area and the carbon demand constraints of the demand area.
[0033] The dual-index construction module is used to train machine learning models and determine key driving factors and their contribution weights through interpretability analysis, thereby constructing an ecological contribution index and a payment responsibility index.
[0034] The compensation amount calculation module is used to construct a two-factor adjustment coefficient based on the deviation of the ecological contribution index from the average ecological contribution index of all supply areas and the deviation of the payment responsibility index from the average payment responsibility index of all demand areas. Based on the two-factor adjustment coefficient, the carbon sequestration service volume and the carbon price, the compensation amount is calculated, and compensation scheme data containing the correspondence between supply areas, demand areas, carbon sequestration service volume and compensation amount is output.
[0035] Each of the above modules executes the corresponding steps in the aforementioned method, and their specific implementation can be found in the corresponding description of the aforementioned method.
[0036] Compared to existing technologies, this invention and its preferred embodiments offer at least the following advantages: This invention incorporates the spatial connection between carbon supply and demand regions into the compensation calculation process, transforming it from static carbon supply and demand assessment to spatialized carbon responsibility identification. It simultaneously identifies the compensation subject, the compensation recipient, the contribution of carbon sequestration services, and the compensation amount, avoiding the inaccurate matching of compensation recipients caused by directly calculating compensation amounts based solely on regional carbon surplus or carbon deficit. This invention incorporates land use, topography, and climate environmental factors into the calculation process of spatial connections for carbon sequestration services. Compared to traditional models that calculate distance attenuation based solely on geometric distance, the simulated carbon sequestration service flow relationships better reflect the actual ecological transmission process under complex geographical conditions. Furthermore, after spatial resistance constraint correction, the intensity of each ecological connection flow decreases, and the degree of decrease increases with the increase of spatial resistance level, indicating that the correction results have a clear spatial response law rather than a uniform proportional reduction. This invention, through supply-side proportions… The conversion method, which combines allocation with demand-side truncation constraints, ensures that the output of supply areas does not exceed their carbon sequestration capacity and the reception of demand areas does not exceed their carbon demand, thus improving the rationality and feasibility of carbon sequestration service flow allocation. This invention uses a data-driven approach to identify key drivers of carbon sequestration capacity and carbon emissions and constructs ecological contribution and payment responsibility indices, improving the objectivity of indicator construction. This invention differentiates the basic compensation amount through a two-factor adjustment coefficient with truncation constraints to strengthen compensation incentives for supply areas with high ecological contributions and to reflect the relative ecological compensation responsibility that demand areas with high external carbon sink dependence should bear, achieving a reasonable redistribution of compensation responsibility while maintaining the overall compensation pattern basically stable. This invention also forms a complete technical chain from carbon supply and demand identification, carbon sequestration service flow simulation, carbon sequestration service volume conversion, index construction to compensation amount calculation, which can directly support applications such as cross-regional horizontal ecological compensation scheme preparation, regional carbon sink capacity assessment, and cross-regional ecological responsibility delineation. Attached Figure Description
[0037] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;
[0038] Figure 2 This is a system architecture diagram of an embodiment of the present invention;
[0039] Figure 3 This is a graph showing the relationship between spatial resistance levels and the degree of weakening of ecological connectivity flows in an embodiment of the present invention.
[0040] Figure 4 This is a comparison chart of ecological compensation payments and received amounts in various provinces across four major regions, as presented in this embodiment of the invention.
[0041] Figure 5 This is a feature importance diagram of the SHAP feature of the CatBoost model for carbon sequestration supply capacity in an embodiment of the present invention.
[0042] Figure 6 This is a feature importance graph of the CatBoost model for carbon emissions in an embodiment of the present invention. Detailed Implementation
[0043] To make the features and advantages of the present invention more apparent and understandable, specific embodiments are described below in detail:
[0044] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.
[0045] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0046] This embodiment provides a cross-regional carbon ecological compensation optimization method and system based on spatial resistance-constrained carbon sequestration service flow simulation. Each step of the above method can be executed by a computer system.
[0047] like Figure 1As shown, this method includes the following main steps. First, constructing a regional carbon sequestration supply and demand database: The net primary productivity of regional vegetation is calculated based on the CASA model, and the net ecosystem productivity is estimated by combining soil heterotrophic respiration, then converted into carbon dioxide equivalents as the supply of carbon sequestration services. County-level carbon emissions are obtained based on provincial carbon emission accounting and a spatial downscaling method driven by characteristic variables, serving as the demand for carbon sequestration services. Net carbon sequestration supply and demand are calculated, and carbon surplus and carbon deficit regions are identified. Second, constructing a carbon sequestration service flow simulation model considering spatial resistance constraints: Four factors—land use type, elevation, slope, and wind speed—are selected, and a comprehensive spatial resistance surface is constructed through hierarchical assignment and weighting using the analytic hierarchy process (AHP). The spatial resistance coefficient between supply and demand regions is determined, and this coefficient is introduced into the distance decay function of the CERF model to construct an R-CERF model, calculating the ecological connection strength between different carbon surplus regions and different carbon deficit regions. Third, a dual-constraint allocation method is established to convert ecological linkage intensity into actual carbon sequestration services: on the supply side, carbon surplus is allocated according to the proportion of ecological linkage intensity; on the demand side, the input services are truncated according to the carbon deficit scale, resulting in the actual carbon sequestration services provided by each supply region to each demand region. Fourth, an ecological contribution index and a payment responsibility index are constructed based on SHAP interpretable machine learning: CatBoost models are trained with positive carbon sequestration capacity and carbon emissions as prediction targets, respectively; key driving factors are analyzed using the SHAP method; and two types of indices are constructed by weighting and summing the standardized values of the key driving factors with normalized SHAP weights. Fifth, a two-factor carbon ecological compensation model is constructed: the basic compensation amount is determined based on carbon sequestration services and carbon price; a two-factor adjustment coefficient with truncation constraints is constructed using the ecological contribution index and the payment responsibility index to optimize and adjust the basic compensation amount, outputting a cross-regional carbon ecological compensation scheme.
[0048] This method can simultaneously output the compensation subject, the compensation recipient, the service contribution, and the differentiated compensation amount, realizing the transformation from regional static carbon accounting to spatial carbon responsibility identification and differentiated compensation calculation.
[0049] like Figure 2 As shown, as a preferred implementation, this embodiment also provides a system based on the above method, including a carbon supply and demand identification module, a comprehensive resistance surface construction module, a carbon sequestration service flow simulation module, a carbon sequestration service quantity conversion module, a dual-index construction module, and a compensation amount calculation module.
[0050] The carbon supply and demand identification module is used to obtain the carbon sequestration service supply and carbon emissions of each spatial unit in the study area, and the difference between the two is determined as the net carbon sequestration supply and demand. Spatial units with positive net carbon sequestration supply and demand are identified as supply areas, and spatial units with negative net carbon sequestration supply and demand are identified as demand areas.
[0051] The comprehensive resistance surface construction module is used to select resistance factors that affect the spatial transmission of carbon sequestration services to construct a comprehensive spatial resistance surface, and to determine the spatial resistance coefficient between each supply area and each demand area based on the comprehensive spatial resistance surface.
[0052] The carbon sequestration service flow simulation module is used to introduce the spatial resistance coefficient into the distance decay function of the comparative ecological radiation force model, and use the product of geometric distance and spatial resistance coefficient as the equivalent transmission distance to calculate distance decay, so as to obtain the ecological connection strength between each supply area and each demand area.
[0053] The carbon sequestration service conversion module is used to determine the actual amount of carbon sequestration services provided by each supply area to each demand area based on the ecological linkage strength, combined with the carbon sequestration capacity constraints of the supply area and the carbon demand constraints of the demand area.
[0054] The dual-index construction module is used to train machine learning models with regional carbon sequestration capacity and regional carbon emissions as prediction targets, respectively. Through interpretability analysis, key driving factors and their contribution weights are determined, and the ecological contribution index of the supply area and the payment responsibility index of the demand area are constructed accordingly.
[0055] The compensation amount calculation module is used to construct a two-factor adjustment coefficient based on the deviation of the ecological contribution index from the average ecological contribution index of all supply areas and the deviation of the payment responsibility index from the average payment responsibility index of all demand areas. Based on the two-factor adjustment coefficient, the carbon sequestration service volume and the carbon price, the compensation amount is calculated, and compensation scheme data containing the correspondence between supply areas, demand areas, carbon sequestration service volume and compensation amount is output.
[0056] As a further preferred implementation, the above modules can be further subdivided into nine sub-modules according to functional granularity: data acquisition module, carbon supply and demand identification module, comprehensive resistance surface construction module, carbon sequestration service flow simulation module, carbon sequestration service volume conversion module, ecological contribution index construction module, payment responsibility index construction module, compensation amount calculation module, and result output module. Each sub-module executes the corresponding steps in the aforementioned method.
[0057] The key technologies of this embodiment are described in detail below:
[0058] (I) Construction of carbon sequestration service supply and demand data and identification of carbon surplus and carbon deficit regions
[0059] Ecological and environmental data and socioeconomic data for each spatial unit within the study area were acquired. Ecological and environmental data included vegetation productivity, land use, topography, and climate data; socioeconomic data included carbon emissions, population, and regional economic development data. After preprocessing such as unified projection, spatial overlay, resolution resampling, and scale matching, the above multi-source data were used to construct an analytical dataset at the county scale.
[0060] Carbon sequestration service supply was calculated based on vegetation net primary productivity. A modified CASA model (Zhu et al., 2007) was used to calculate the monthly net primary productivity of the study area:
[0061]
[0062] In the formula, Indicates spatial location; Indicates the month; For pixels In the Net primary productivity in months ( ); For pixels In the Photosynthetically active radiation absorbed per month ( ); For pixels In the Actual solar energy utilization efficiency in months ( The CASA model is driven by remote sensing data such as temperature, precipitation, solar radiation, and vegetation type, and is a well-known method in the field.
[0063] Soil respiration was estimated using the generalized spatial mean scale relation model (GSMSR) (Chuai et al., 2018; Li et al., 2023):
[0064]
[0065] In the formula, Breathing for the lunar soil ( ); Soil organic carbon density in the 0–20 cm soil layer ( ); The average monthly temperature ( ); Monthly precipitation (cm). Annual soil respiration. This is the sum of soil respiration for each month.
[0066] Soil heterotrophic respiration was determined by linear fitting of total soil respiration (Chuai et al., 2018):
[0067]
[0068] In the formula, For soil heterotrophic respiration ( ); Total annual soil respiration ( ).
[0069] Net ecosystem productivity is the difference between net primary productivity and soil heterotrophic respiration.
[0070]
[0071] In the formula, Net ecosystem productivity ( ).when When the carbon fixed by vegetation exceeds the carbon emitted by soil microbial respiration, the ecosystem exhibits carbon sink characteristics; when When this occurs, it indicates that the ecosystem is acting as a carbon source.
[0072] The net ecosystem productivity per unit area of each pixel within the county is multiplied by the pixel area and summed. After unit conversion, the total carbon mass of carbon sequestration services in the county, expressed in tons of carbon, is obtained. This is then converted to carbon dioxide equivalent based on the molar mass ratio of carbon to carbon dioxide.
[0073]
[0074] In the formula, For the region Total carbon mass of carbon sequestration services (tons of carbon). For the region Carbon sequestration service supply (million tons) The coefficient 44 / 12 represents the molar mass ratio of carbon to carbon dioxide. This is the dimensional conversion factor between tons and millions of tons.
[0075] The demand for carbon sequestration services is characterized by county-level carbon emissions. County-level carbon emission data is obtained using a combination of provincial accounting constraints, characteristic variable-driven methods, and spatial downscaling (Wang et al., 2025): First, the total carbon dioxide emissions of each province are calculated based on IPCC guidelines and provincial energy statistics, serving as the upper-level constraint for county-level emission estimation. Second, characteristic variables that characterize regional differences in energy consumption and human activities, such as nighttime light intensity, population size and structure, construction land distribution, climate conditions, and technological innovation levels, are selected to depict the spatial heterogeneity of provincial carbon emissions. Based on this, the relative distribution characteristics of these variables at the county level are used to weight the provincial emission results and allocate them to county-level units, achieving spatial downscaling from the provincial to the county level, ultimately obtaining national county-level carbon emission data (in millions of tons). ).
[0076] The regional carbon supply and demand status is determined by the difference between the supply and demand of carbon sequestration services:
[0077]
[0078] In the formula, For the region Net carbon sequestration supply and demand (million tons) ); For the region Carbon sequestration service supply (million tons) ); For the region Demand for carbon sequestration services (million tons) ).when When the supply in a region can meet its own needs while providing carbon sequestration services to surrounding areas, it is considered a carbon surplus region, i.e., a carbon sequestration service supply region; when This indicates that the region needs to rely on external supply to meet the demand gap, making it a carbon deficit region, i.e., a region with a demand for carbon sequestration services.
[0079] For ease of subsequent calculations, the supply area will be... The carbon surplus is denoted as (million tons) (Take a positive value), and divide the demand area The carbon deficit is denoted as (million tons) (Take a positive value).
[0080] (II) Simulation Model of Carbon Sequestration Service Flow under Spatial Resistance Constraints
[0081] To address the issue that traditional CERF models only consider geometric distance attenuation and ignore spatial environmental differences, this embodiment introduces spatial resistance constraints to construct a carbon sequestration service flow simulation model that takes spatial resistance constraints into account.
[0082] First, a comprehensive spatial resistance surface was constructed. Four factors were selected: land use type, elevation, slope, and wind speed. Land use type was used to characterize the differences in ecological function among different landscape units; elevation and slope reflected the moderating effect of topographic conditions on inter-regional ecological connections; wind speed, as an atmospheric environmental factor, characterized the influence of climatic conditions on carbon sequestration services during spatial operations. Lower wind speeds resulted in weaker atmospheric turbulent diffusion and convective transport, thus limiting the spatial range of carbon sequestration services, and consequently, lower wind speeds corresponded to higher resistance values. These factors, from the three dimensions of landscape structure, topographic conditions, and climatic environment, jointly characterized the comprehensive spatial constraints that inter-regional ecological connections may face.
[0083] The graded resistance values for each factor were determined based on the strength of their blocking effect on the spatial transmission of carbon sequestration services, referencing the parameter system in existing carbon sequestration service flow network studies (Liang Huaqiu, 2024), as shown in Table 1. Each factor was mapped to a dimensionless resistance value according to the graded index; the larger the resistance value, the stronger the constraint on the spatial transmission of carbon sequestration services under that landscape condition.
[0084] Table 1. Resistance Factor Classification, Resistance Values, and Weights
[0085]
[0086] Factor weights were determined using the analytic hierarchy process (AHP). Based on expert experience and a reference parameter system, pairwise comparisons were made between the four types of factors to construct a judgment matrix, as shown in Table 2.
[0087] Table 2. Pairwise comparison and judgment matrix of resistance factors
[0088]
[0089] Calculations show that the largest eigenvalue of the judgment matrix is 4.0435, and the consistency ratio CR = 0.0163 < 0.1 (the random consistency index RI is taken as 0.89), therefore the judgment matrix meets the consistency requirements. The weights of each factor are determined using the eigenvector method, resulting in the comprehensive resistance surface.
[0090]
[0091] In the formula, The total resistance surface value (dimensionless); , , , The values are the graded resistance values for elevation, slope, land use, and wind speed factors (dimensionless, mapped according to the graded indexes in Table 1); the coefficients 0.0955, 0.0955, 0.2495, and 0.5596 are the weights of the corresponding factors in the analytic hierarchy process, and the sum of the weights is 1.
[0092] Based on this, the spatial resistance coefficient between supply zone i and demand zone j is determined according to the comprehensive spatial resistance surface. The spatial resistance coefficient is obtained by arithmetically averaging the surface resistance values of all grid cells along the line connecting the geometric centers of supply zone i and demand zone j.
[0093]
[0094] In the formula, is the spatial resistance coefficient between supply zone i and demand zone j (dimensionless, ranging from 1 to 7); n is the number of grids on the connecting line; This represents the total resistance surface value of the k-th grid cell on the connection line.
[0095] The physical meaning of this space drag coefficient is the equivalent transmission distance magnification factor. When When the total resistance surface value of all grids along the entire line is at its lowest value of 1, meaning that the four factors of elevation, slope, land use, and wind speed are all at the lowest resistance level, the R-CERF model degenerates into a traditional CERF model that only considers geometric distance attenuation; when When the distance is greater than the minimum level, it indicates that there is a spatial barrier above the minimum level, and the equivalent transmission distance is correspondingly magnified to the product of the geometric distance and the drag coefficient.
[0096] The Comparative Ecological Radiation Force (CERF) model, drawing inspiration from the gravity model, posits that the strength of ecological linkages between regions is jointly determined by supply intensity, demand intensity, and distance, and assumes that the impact of ecosystem services decreases with increasing distance (Chen et al., 2014). The traditional CERF model calculates the ecological linkage strength between supply and demand zones as follows:
[0097]
[0098] In the formula, Supply area calculated for traditional CERF model With demand area The strength of the ecological connections between them (dimensionless); For the supply area With demand area The Euclidean distance (km) between the geometric centers; The maximum distance (km) between all supply and demand point pairs within the study area is used to make the distance term dimensionless; For the supply area carbon surplus (million tons) ); For demand area carbon deficit (million tons) ).
[0099] The denominator reflects the moderating effect of the relative structure of supply and demand on the strength of the connection, and its form is derived from the well-known breakpoint formula (Converse, 1949).
[0100] The carbon surplus in the supply region is the net supply and demand of carbon sequestration in that region, while the carbon deficit in the demand region is the absolute value of the net supply and demand of carbon sequestration in that region.
[0101] By incorporating the space drag coefficient into the distance decay function of the CERF model, and combining the geometric distance effect with the space drag effect, an R-CERF model is constructed:
[0102]
[0103] In the formula, To comprehensively consider distance attenuation and spatial resistance constraints, the supply area With demand area The strength of the ecological connections between them (dimensionless); The coefficient represents the spatial drag coefficient; the meanings of the other parameters are the same as in the previous formula. This model converts spatial barriers into equivalent transmission distances by multiplying the geometric distance by the spatial drag coefficient, thereby correcting the spatial constraints on the distance attenuation relationship and obtaining the contribution relationship of carbon sequestration services from different carbon surplus regions to different carbon deficit regions.
[0104] (III) Conversion of Actual Carbon Sequestration Services Based on Ecological Linkage Intensity
[0105] Ecological link strength only reflects the relative magnitude of the carbon sequestration service transfer relationship between different supply and demand regions, and does not directly represent the actual amount of carbon sequestration services transferred. This embodiment establishes a method for converting ecological link strength into actual carbon sequestration service volume, simultaneously incorporating the carbon sequestration capacity constraints of the supply region and the carbon demand constraints of the demand region, thereby achieving spatial allocation of carbon sequestration service flows.
[0106] First, supply-side constraints are allocated. For each carbon supply region, the intensity of its ecological connections to different demand regions is normalized to obtain its relative contribution ratio:
[0107]
[0108] In the formula, For the supply area Demand-oriented area The allocation ratio (dimensionless, summed to 1 for all demand areas in the same supply area); Ecological link strength calculated by the R-CERF model; denominator is the supply area. The sum of the ecological connectivity strengths across all demand areas.
[0109] The initial allocation of services is calculated based on supply-side constraints, that is, all output carbon sequestration services in the supply area are allocated outward according to different ecological linkage intensities:
[0110]
[0111] In the formula, For the supply area To the demand area Initial carbon sequestration services allocated (million tons) ); For the supply area carbon surplus (million tons) ).
[0112] Then, demand-side constraints are adjusted. Since different carbon demand regions have different carbon deficit sizes, their actual acceptable carbon sequestration services should not exceed their own demand. Therefore, a demand constraint coefficient is defined as follows:
[0113]
[0114] In the formula, For demand area Demand constraint coefficient (dimensionless) ); For demand area carbon deficit (million tons) ); For all supply areas to demand areas The sum of the initial service volume provided (million tons) When the initial total supply exceeds the demand, The input service volume is compressed proportionally; when the initial total supply is no greater than the demand, Maintain the initial supply level and do not artificially increase it.
[0115] The final actual carbon sequestration service volume is:
[0116]
[0117] In the formula, For the supply area To the demand area Actual carbon sequestration services provided (million tons) ).
[0118] Through the aforementioned dual constraint mechanism, the amount of carbon sequestration services output by any supply region does not exceed its actual carbon sequestration capacity, and the amount of carbon sequestration services received by any demand region does not exceed its actual carbon demand, thus satisfying the requirement. and .
[0119] (iv) Construction of Ecosystem Contribution Index Based on SHAP Interpretable Machine Learning
[0120] Traditional ecological compensation methods typically determine ecological contributions directly based on regional carbon sequestration, neglecting the differences in carbon sequestration capacity formation mechanisms across different regions due to variations in natural environmental conditions, vegetation structure, and land use patterns. This embodiment utilizes a machine learning model to establish a nonlinear mapping relationship between ecological environmental factors and regional carbon sequestration capacity. Furthermore, interpretability analysis is used to analyze the contribution of different environmental factors to the model's prediction results, thereby constructing an ecological contribution index. As a preferred implementation, the interpretability analysis employs the SHAP (SHapley Additive exPlanations) method.
[0121] The sample includes regions with positive carbon sequestration capacity, i.e., regions with net ecosystem productivity greater than zero. The region was selected as the modeling sample, with regional carbon sequestration capacity as the prediction target, and regional carbon sequestration capacity was characterized by net ecosystem productivity. Multi-source ecological and environmental variables that can characterize the formation process of regional carbon sequestration capacity were selected as candidate input features, including five categories: natural topography and climate factors, socio-economic factors, land use structure factors, vegetation ecology factors, and landscape pattern factors, totaling 35 candidate features.
[0122] Natural topographic and climatic factors include: total slope. That is, the sum of the slopes of all pixels within the county. , Pixel slope; average slope , , Number of pixels within the county; average elevation , , Pixel elevation (m); average annual temperature , , For the first The average temperature of the day ( ); average annual precipitation , , For the first Monthly precipitation (mm).
[0123] Socioeconomic factors include: GDP per capita , , The total GDP of the county (in ten thousand yuan). Total population of the county (in people); population density , , The total area of the county ( ); added value of the secondary industry (in ten thousand yuan); Total resident population (Persons); Total urban population (People); Urbanization rate , Total value of nighttime lights This refers to the sum of nighttime light brightness values within the county.
[0124] Land use structure factors include: cultivated land area With the proportion of arable land , Forest area With forest land ratio , grassland area With grassland ratio , Forest and grassland area ratio , Forest and grassland area and , and the proportion of forest and grassland area , Water area proportion of water body , Construction land area Construction land ratio , All area units mentioned above are... .
[0125] Vegetation ecological factors include: net primary productivity ( (The annual value is obtained by accumulating the data monthly based on the CASA model); Summer NDVI mean. , Winter NDVI mean , Average summer vegetation cover Compared with the average winter vegetation cover The average vegetation cover was taken in June, July, August and November, December and January of the following year, respectively; soil heterotrophic respiration was measured. ( ).
[0126] Landscape pattern factors include: patch density , , The number of patches within the county; clustering index. , , This represents the number of adjacent edges between similar patches. The maximum number of adjacent nodes of this type of patch can reach, ranging from 0 to 100, with higher values indicating higher clustering; Landscape segmentation index. , , The first in the county The total area of patches; the higher the value, the greater the degree of fragmentation; landscape shape index. , , The total length (m) of the patch boundaries within the county. The larger the value, the more complex the shape.
[0127] For model training and validation, the CatBoost model is used to establish a nonlinear mapping relationship between ecological and environmental factors and carbon sequestration capacity. In this embodiment, due to net primary productivity... heterotrophic respiration in soil There is a deterministic algebraic relationship between the predicted target net ecosystem productivity and ( To avoid the model learning trivial fits to deterministic algebraic relationships instead of the true driving mechanisms, both were removed from the candidate feature set. Subsequently, Pearson correlation analysis was used to screen candidate features. First, variables with weak explanatory power were removed based on their correlation with the target variable. Second, to avoid multicollinearity, when the correlation coefficient between any two features was greater than 0.8, only features with a higher correlation to the target variable were retained. During training, the samples were divided into training and test sets at a ratio of 0.7:0.3, and the optimal parameter combination was determined using five-fold cross-validation to determine the coefficients. With root mean square error Evaluate the model's predictive ability. After the model training and validation are completed, use the SHAP method to analyze the internal feature contributions of the model.
[0128] Regarding the construction of the ecological contribution index, based on the trained CatBoost model, the SHAP method is used to calculate the marginal contribution of each ecological and environmental variable to the model's prediction results, obtaining the average SHAP importance. According to the ranking of average SHAP importance, the key driving factors with the highest contribution are selected as variables for constructing the ecological contribution index. The SHAP importance values of each factor are normalized to obtain relative weights. The normalized weights are then weighted and summed with the standardized values of the corresponding factors in the region to construct the ecological contribution index.
[0129]
[0130] In the formula, For the supply area Ecological contribution index (dimensionless). The key driver index is selected from the top three key drivers in terms of importance in SHAP. For the first Normalized SHAP weights of the three key drivers (dimensionless, the sum of the three weights is 1). For the supply area In the Standardized values for key driving factors. As a preferred implementation, standardization employs min-max normalization, mapping the values of each factor to the [0,1] interval.
[0131] (v) Construction of a payment responsibility index based on SHAP interpretable machine learning
[0132] Existing carbon compensation methods typically determine demand responsibility based on regional carbon emissions or the size of the carbon deficit, reflecting only the quantitative differences in regional carbon demand and failing to fully consider the differences in the multidimensional driving factors that lead to regional carbon demand. This embodiment uses regional carbon emissions as the prediction target, constructs a nonlinear mapping relationship between multi-source environmental and socio-economic factors and carbon emission levels, and uses interpretability analysis to analyze the contribution of different driving factors to the formation of regional carbon emission demand, thereby constructing a payment responsibility index; the interpretability analysis also employs the SHAP method.
[0133] Using regional carbon emissions as the prediction target of the machine learning model, multidimensional feature variables influencing the formation process of regional carbon emissions are selected as model inputs. The feature variable set is the same as the aforementioned candidate feature set. A CatBoost model is used to establish the nonlinear relationship between regional carbon emissions and multi-source driving factors. The training and validation methods are consistent with those described above, i.e., the training and test sets are divided at a ratio of 0.7:0.3, and the optimal parameter combination is determined through five-fold cross-validation. and Evaluate the model's predictive ability. In the carbon emission prediction model, there is no deterministic algebraic relationship between the candidate features and the target variable; therefore, there is no need to eliminate specific features. Feature selection is still completed using Pearson correlation analysis and collinearity tests. After model training and validation, the SHAP method is used to analyze the feature contributions.
[0134] The construction method of the payment responsibility index is similar to that of the ecological contribution index. Based on the average SHAP importance ranking, the key driving factors with the highest contribution are selected to construct the payment responsibility index. The SHAP importance values of each factor are normalized to obtain relative weights. The normalized weights are then weighted and summed with the standardized values of the corresponding factors in the demand area.
[0135]
[0136] In the formula, For demand area The payment responsibility index (dimensionless). The key driver index is selected from the top three key drivers in terms of importance in SHAP. For the first Normalized SHAP weights of the three key drivers (dimensionless, the sum of the three weights is 1). For demand area In the Standardized values for each key driving factor. As a preferred implementation, standardization also employs min-max normalization, mapping each factor value to the [0,1] interval.
[0137] The payment responsibility index is used to characterize the relative payment responsibility of a demand region due to its carbon emission levels and dependence on external carbon sequestration services. The formation mechanism of regional carbon emissions is closely related to the intensity of regional economic activity. An index constructed using carbon emission driving factors can reflect the region's carbon emission demand and its dependence on external carbon sequestration services. The higher the carbon emission demand and the stronger the dependence on external carbon sequestration services, the greater the compensation responsibility that the demand region should bear for benefiting from them.
[0138] (vi) Two-factor carbon ecological compensation model
[0139] First, determine the basic compensation amount based on the carbon sequestration service flow:
[0140]
[0141] In the formula, For demand area Should be supplied to the supply area The basic compensation amount to be paid (million yuan); Carbon trading price (yuan / ton) As the preferred implementation method, the value of 42.85 yuan / ton was taken from the first compliance cycle report of the national carbon emission trading market. ; For the supply area To the demand area Actual carbon sequestration services provided (million tons) Since the service volume is measured in millions of tons, the calculated compensation amount is in millions of yuan. In the example statistics, 100 million yuan is converted to 100 million yuan.
[0142] Further construct the two-factor adjustment coefficient:
[0143]
[0144] In the formula, This is a two-factor adjustment coefficient (dimensionless), used to comprehensively reflect the ecological contribution of the supply side and the payment responsibility of the demand side; This is a truncation function that takes the corresponding boundary value when the value of the independent variable is less than the lower limit or greater than the upper limit; and The adjustment weights for the ecological contribution index and the payment responsibility index are respectively set to 0.5, which is the preferred implementation method. For the supply area Ecological contribution index; This is the average value of the ecological contribution index for all carbon surplus regions. For demand area The payment responsibility index; This represents the average of the payment responsibility index for all carbon deficit regions. and These are the lower and upper limits of the adjustment coefficient, respectively, and are preferably set to 0.5 and 1.5.
[0145] The following factors were considered when setting the adjustment weights: Ecological contribution and payment responsibility represent supply responsibility and demand responsibility in the compensation relationship, respectively, and the two have complementary effects; using equal weights can avoid artificially reinforcing one side of the factors, causing the compensation results to be biased towards the supply side or the demand side; in the absence of prior policy preferences, equal weights can maintain a basic balance between ecological contribution responsibility and payment responsibility.
[0146] The following factors are considered when setting the upper and lower limits of the adjustment coefficient. The two-factor adjustment serves only as an optimization correction to the basic compensation result and should not alter the original ecosystem service calculation results; the adjustment range should be limited to a reasonable extent to avoid abnormal expansion or contraction of the compensation result due to extreme differences in regional indicators; the compensation result should remain stable so that the final compensation amount is still primarily based on the actual contribution of carbon sequestration services. Under the optimal condition that both types of indices are standardized in the [0,1] interval, and The absolute values of all factors do not exceed 1. After multiplying by a weight of 0.5, the single-factor adjustment amount does not exceed 0.5. In the extreme case of two factors moving in the same direction, the theoretical range of the adjustment coefficient is [0,2]. In order to limit its impact on the basic compensation amount, the adjustment coefficient is truncated to [0.5,1.5] so that the adjustment range of the basic compensation amount does not exceed 50%. That is, the lower limit of the adjustment coefficient corresponds to a maximum reduction of 50% in the basic compensation amount, and the upper limit corresponds to a maximum increase of 50%.
[0147] The final compensation amount is calculated using the following formula:
[0148]
[0149] In the formula, For demand area to supply area The final compensation amount paid (in millions of yuan).
[0150] Through the aforementioned adjustment mechanism, targeting the supply area, when When the ecological contribution of the region is higher than the average level, the adjustment coefficient is greater than 1, and the corresponding compensation level is increased to incentivize higher ecological contributions; conversely, the compensation is appropriately reduced. For regions with high demand, when... When the ecological contribution rate is higher than the average, the adjustment coefficient is greater than 1, and the payment amount increases; conversely, the payment pressure decreases. This model achieves a compensation structure that matches ecological contribution responsibility with payment responsibility through a dual-factor adjustment of ecological contribution and payment responsibility.
[0151] (vii) Implementation Examples and Effect Verification
[0152] This study focuses on county-level units within 31 provincial-level administrative units in mainland China (excluding Hong Kong, Macau, and Taiwan) in 2020. Data sources include: Normalized Difference Vegetation Index (NDVI) from MODIS NDVI product MOD13Q1 (16-day composite, 250 m resolution); precipitation from CHIRPS diurnal precipitation product; temperature and solar radiation from ERA5-Land reanalysis dataset; elevation from SRTM 30 m product; land use from CLCD global 30 m land use dataset; nighttime light from the cross-sensor calibrated DMSP-OLS nighttime light dataset (1 km resolution); population density from WorldPop data; resident population, urban population, and secondary industry added value from the National Bureau of Statistics; and GDP data from the China Resources and Environmental Sciences Data Center. All data were uniformly projected to the Albers equal-area conic projection and resampled to 100 m spatial resolution.
[0153] A county-level carbon sequestration service supply and demand database was constructed using the aforementioned method to identify carbon surplus and carbon deficit counties in 2020. The results show that there are 546 carbon surplus counties, mainly distributed in ecological functional areas such as the southwestern mountains, the eastern edge of the Qinghai-Tibet Plateau, and the northeastern forest region, mostly located in areas with relatively low population density; there are 2304 carbon deficit counties, mainly distributed in the eastern coastal areas and some resource development areas, concentrated in densely populated and economically active areas east and south of the Hu Huanyong Line. China's total carbon emissions in 2020 were approximately 10726.6 million tons. The carbon sequestration amount is approximately 2822.6 million tons. The number of carbon surplus counties has decreased significantly compared to 2000 (1055), while the number of carbon deficit counties has increased significantly compared to 2000 (1795), indicating a continued strengthening of the supply-demand mismatch. In terms of regional distribution, the western region has the highest proportion of carbon surplus counties, while the eastern and central regions have the highest proportion of carbon deficit counties within their respective regions.
[0154] Based on the spatial combination of supply and demand regions, all potential ecosystem service transfer pathways were constructed, resulting in a total of 546 × 2304 = 1,257,984 ecosystem link flows. Strong link flows mainly occur between supply and demand regions that are spatially close and have low spatial resistance. A relatively dense link network forms between the carbon surplus region in the southwest mountains and the carbon deficit regions in central and eastern China, and between the forest region in northeast China and the carbon deficit region in north China. However, the link strength between supply and demand pairs crossing high-resistance terrain units (such as large mountain ranges and plateau barriers) is significantly reduced. For all ecosystem link flows, the ecological link strength was calculated using both the traditional CERF model and the R-CERF model.
[0155] To compare the differences in the calculation results between the two models, the rate of change of R-CERF relative to the traditional CERF model was calculated for each ecological link flow:
[0156]
[0157] In the formula, The relative change rate of ecological link intensity (%) Ecological link strength calculated for the R-CERF model; The ecological link strength calculated using the traditional CERF model.
[0158] The results showed that, compared with the traditional CERF model, the intensity of all 1,257,984 ecological link flows calculated by the R-CERF model decreased, with an average change rate of -67.23%. Among the typical ecological link flows, the supply area, Hekou Yao Autonomous County (… ) and the demand area Ezhou District, Ezhou City ( The strength of ecological links between the two ecosystems decreased from 0.1216 in the traditional CERF model to 0.0420 in the R-CERF model, a reduction of 65.46%. These results indicate that the R-CERF model can correct for spatial resistance constraints on all ecological link flows, and this correction is not based on a uniform ratio, but rather on differentiated adjustments according to the spatial resistance conditions corresponding to different ecological link paths.
[0159] To verify whether the correction effect conforms to the spatial transmission law of ecosystem services, such as Figure 3 As shown, all ecological link flows are divided into four levels—low resistance, low-to-medium resistance, medium-to-high resistance, and high resistance—based on their spatial resistance coefficients. The average weakening effect of the R-CERF model relative to the CERF model is then statistically analyzed for each resistance level. The results indicate that the R-CERF model's correction of ecological link strength gradually increases with increasing spatial resistance. The average weakening effects for the low-to-low resistance, medium-to-high resistance, and high resistance levels are approximately 40%, 65%, 76%, and 87%, respectively. This suggests that ecological link pathways with higher spatial resistance experience more significant limitations on their actual carbon sequestration service transfer capacity.
[0160] Furthermore, the Spearman rank correlation method was used to test the monotonic relationship between resistance level and the degree of weakening of ecological link flow. With resistance level as the ranking variable and ecological link flow weakening index as the response variable, the results showed that the rank correlation coefficient was 0.9664, p<0.001, indicating that there is a highly significant positive correlation between resistance level and the degree of weakening of ecological link flow. This demonstrates that the R-CERF model can adaptively adjust the strength of ecological link according to the degree of spatial obstruction in the process of ecosystem service transmission.
[0161] To identify the key drivers influencing the supply and demand of carbon sequestration services in counties, the predictive performance of various machine learning models was compared, including CatBoost, XGBoost, random forest, BP neural network and convolutional neural network. The results are shown in Table 3.
[0162] Table 3 Comparison of Machine Learning Model Performance
[0163]
[0164] The results show that CatBoost achieves the highest coefficient of determination (R²=0.852) and the lowest root mean square error (RMSE=0.383) on the independent test set. Its test set performance is slightly better than XGBoost (R²=0.846, RMSE=0.391), and significantly better than Random Forest, Convolutional Neural Networks, and Backpropagation Neural Networks. Five-fold cross-validation results further demonstrate that CatBoost has the highest average validation performance with the least fluctuation. 2 The hyperparameters were 0.856±0.018, higher than XGBoost (0.850±0.022), Random Forest (0.783±0.023), Convolutional Neural Network (0.782±0.033), and Backpropagation Neural Network (0.773±0.035), indicating that CatBoost has relatively better generalization performance and cross-fold consistency under different data partitioning conditions, and can better characterize the complex nonlinear relationship between multidimensional driving factors and county-level carbon sequestration service patterns. Therefore, CatBoost was selected as the core model for subsequent SHAP interpretability analysis. The key hyperparameters of the model were optimized using learning curves and cross-validation methods to reduce generalization error and improve model reliability.
[0165] Carbon sequestration supply in counties ( In the SHAP contribution analysis of carbon demand (carbon emissions) and carbon emissions, the top three key driving factors in terms of contribution for each year are identified, and an ecological contribution index and a payment responsibility index are constructed accordingly. For example... Figure 5 As shown, the top three factors in terms of importance for the carbon sequestration supply capacity (SHAP) model are sum of forest and grassland area (FGS), sum of slope (SSLOPE), and forest area (FLA); Figure 6 As shown, the top three factors in terms of importance in the SHAP carbon emission model are the proportion of built-up land (BUAR), secondary industry added value (SI), and arable land area (CLA). Specific results for 2020 are shown in Table 4.
[0166] Table 4 Key Drivers and Weights of the Ecological Contribution Index and Payment Responsibility Index in 2020
[0167]
[0168] The ecological contribution index is composed of three key driving factors: the sum of forest and grassland area, the sum of slope, and forest area. It reflects the dominant role of ecological structure and topographic conditions in the formation of county-level carbon sequestration capacity. The payment responsibility index is composed of three key driving factors: the proportion of construction land, the added value of the secondary industry, and the area of cultivated land. It reflects the driving role of the intensity of regional economic activity in the formation of carbon emission demand. The key driving factors have been slightly adjusted according to the development stage in different years. In 2000 and 2010, the payment responsibility index was dominated by the proportion of construction land, the added value of the secondary industry, and population density, while the ecological contribution index was dominated by the sum of forest and grassland area, the sum of slope, and forest area (vegetation coverage in 2010). The SHAP weights of the factors in each year were determined according to the aforementioned normalization method.
[0169] Based on the aforementioned ecological contribution index and payment responsibility index, the payment amount for carbon deficit counties and the compensation amount for carbon surplus counties in China in 2020 were calculated using the two-factor carbon ecological compensation model, and compared with the traditional compensation model (calculated only based on service volume and carbon price). At the national level, the total payment amount for carbon deficit counties under the traditional model was 69.2589 billion yuan, which increased to 70.2809 billion yuan after introducing the two-factor adjustment, representing a slight increase in the compensation scale. Spatially, the payment amount in most provinces in the eastern and central regions increased under the two-factor model compared to the traditional model, with a significant strengthening of payment responsibility in economically developed areas. On the compensation side, the compensation amount in the western region generally increased, with significant improvements in compensation benefits for ecological function zones such as Qinghai, Tibet, Xinjiang, and Inner Mongolia. Comparing the net change amount (the difference between net compensation and net payment after the improved model) for each province, it can be seen that the eastern and central provinces generally have negative values, indicating an increase in their net payment responsibility; while the western provinces consistently have positive values, indicating an increase in their compensation benefits.
[0170] like Figure 4 As shown in Table 5, the statistics of compensation payments and received amounts in various regions under the traditional model and the two-factor model in 2000, 2010 and 2020 are as follows. The comparison by province further shows that the two-factor adjustment made the increase in the amount of compensation received in western provinces significantly higher than that in eastern and central provinces, with the net increase in compensation in ecological functional areas such as Tibet and Inner Mongolia being the most prominent.
[0171] Table 5. Statistics on ecological compensation payments and receipts in four major regions of China (traditional model and two-factor model, in RMB 100 million)
[0172]
[0173] As shown in Table 5, the dual-factor regulation has led to a structural shift in compensation responsibility from economically developed regions to ecological functional zones, which has improved the fairness of compensation responsibility among regions while maintaining the overall compensation pattern in a basically stable manner.
[0174] Furthermore, to test the stability of the carbon supply and demand identification results, a combination of scenario perturbation and Monte Carlo simulation was used for analysis. When ±10% perturbation was applied to net ecosystem productivity or carbon emissions, the changes in the number of carbon surplus and carbon deficit counties in each year were controlled between 1.0% and 8.5%, and the spatial pattern remained basically unchanged. Monte Carlo simulation showed that most counties maintained the same carbon supply and demand type in multiple simulations, with only a small number of counties in the transition zone where net supply and demand were close to zero experiencing classification fluctuations. The proportions of counties with stable carbon supply and demand classifications in 2000, 2010, and 2020 were 96.14%, 94.95%, and 94.81%, respectively, indicating that the carbon supply and demand identification results have high temporal stability and input robustness.
[0175] Regarding data reliability, the net primary productivity estimated by the CASA model was compared with the MODIS NPP product. At the county level, the coefficients of determination for 2000, 2010, and 2020 were 0.594, 0.553, and 0.515, respectively, while at the provincial level, they increased to 0.739, 0.749, and 0.686, indicating strong robustness of the model at macroscale. The estimated net ecosystem productivity ranged from -469.39 to 1182.50. The average value is 80.62. The result is within a reasonable range of existing research, indicating that the estimation results of carbon sequestration service supply are reliable.
[0176] Compared with the prior art, the solution of this embodiment has the following beneficial effects.
[0177] 1. Achieving precise spatial identification of carbon sequestration service contribution relationships. This embodiment constructs a carbon sequestration service flow simulation model that considers spatial resistance constraints, incorporating the spatial connection between carbon supply regions and carbon demand regions into the compensation calculation process. This enables the simultaneous determination of the compensation subject, the compensation recipient, the service contribution amount, and the differentiated compensation amount, realizing a shift from static carbon supply and demand assessment to spatial carbon responsibility identification. This avoids the problem of inaccurate matching of compensation objects caused by directly calculating compensation amounts based solely on regional carbon deficits or carbon surpluses.
[0178] 2. Reducing the simulation bias of carbon sequestration service flows caused by traditional distance attenuation models. Traditional CERF models describe the strength of ecological connections between regions based solely on geometric distance. This embodiment introduces spatial resistance constraints, incorporating land use type, topographic conditions, and climatic environmental factors into the calculation process of carbon sequestration service connections. This makes the simulated carbon sequestration service flow relationships more consistent with the actual ecological transmission process under complex geographical conditions. Implementation examples show that the R-CERF model achieves differentiated spatial resistance correction for all ecological connection flows, and there is a highly significant positive correlation between resistance level and attenuation degree, proving that the correction results have a clear spatial response law.
[0179] 3. Achieve spatial allocation of carbon sequestration services under dual constraints. This embodiment uses a conversion method that combines supply-side proportional allocation with demand-side truncation constraints to convert the intensity of ecological linkages into actual carbon sequestration services. At the same time, it ensures that the output of any supply area does not exceed its carbon sequestration capacity and the reception of any demand area does not exceed its carbon demand, thereby improving the spatial rationality and feasibility of carbon sequestration service flow allocation.
[0180] 4. Achieving objective quantification of ecological contribution and payment responsibility. This embodiment utilizes the SHAP interpretable machine learning method to identify key driving factors of carbon sequestration capacity and carbon emissions in a data-driven manner, and constructs ecological contribution index and payment responsibility index accordingly. This replaces the traditional subjective weighting method of manually setting indicator weights, improving the objectivity and interpretability of indicator construction.
[0181] 5. Achieving differentiated and dynamic adjustment of compensation standards. This embodiment constructs a dual-factor adjustment coefficient based on ecological contribution and payment responsibility, and uses a cutoff function to limit the adjustment range. On the one hand, it ensures that carbon sink areas with high ecological contributions receive compensation commensurate with their contributions; on the other hand, it ensures that areas with higher payment responsibilities bear corresponding compensation responsibilities, achieving a synergistic match between ecological contribution responsibility and payment responsibility. The implementation example shows that, while maintaining the overall compensation pattern basically stable, the dual-factor adjustment achieves a structural shift of compensation responsibility from economically developed regions to ecologically functional zones, improving the fairness of inter-regional compensation.
[0182] 6. Establish a complete technical process for cross-regional carbon governance. This embodiment consists of a complete technical chain, from carbon supply and demand identification, carbon sequestration service flow simulation, carbon sequestration service allocation, dual-index construction to dual-factor compensation calculation. The output results can directly support application scenarios such as cross-regional horizontal ecological compensation scheme preparation, regional carbon sink capacity assessment and compensation object identification, and cross-regional ecological responsibility delineation in environmental impact assessment.
[0183] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.
Claims
1. A cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation, characterized in that, The method is executed by a computer system and includes: The carbon sequestration service supply and carbon emissions of each spatial unit within the study area are obtained. The difference between the two is determined as the net carbon sequestration supply and demand. Spatial units with positive net carbon sequestration supply and demand are identified as carbon surplus areas as supply areas, and spatial units with negative net carbon sequestration supply and demand are identified as carbon deficit areas as demand areas. A comprehensive spatial resistance surface is constructed by selecting resistance factors that affect the spatial transmission of carbon sequestration services, and the spatial resistance coefficient between each supply area and each demand area is determined based on the comprehensive spatial resistance surface. The spatial resistance coefficient is introduced into the distance attenuation function of the comparative ecological radiation force model. The product of the geometric distance and the spatial resistance coefficient is used as the equivalent transmission distance for distance attenuation calculation to obtain the ecological connection strength between each supply area and each demand area. Based on the aforementioned ecological linkage strength, and in conjunction with the carbon sequestration capacity constraints of the supply area and the carbon demand constraints of the demand area, the actual amount of carbon sequestration services provided by each supply area to each demand area is determined. Machine learning models were trained with regional carbon sequestration capacity and regional carbon emissions as prediction targets, respectively. Key driving factors and their contribution weights were determined through interpretability analysis, and the ecological contribution index of the supply area and the payment responsibility index of the demand area were constructed accordingly. A two-factor adjustment coefficient is constructed based on the deviation of the ecological contribution index from the average ecological contribution index of all supply areas and the deviation of the payment responsibility index from the average payment responsibility index of all demand areas. The compensation amount paid by each demand area to each supply area is calculated in combination with the carbon sequestration service volume and carbon price. The compensation scheme data containing the correspondence between supply areas, demand areas, carbon sequestration service volume and compensation amount is output.
2. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation as described in claim 1, characterized in that: The comprehensive spatial resistance surface is constructed by selecting multiple resistance factors that affect the spatial transmission of carbon sequestration services, including land use type, elevation, slope, and wind speed. The multiple resistance factors are compared pairwise using the analytic hierarchy process to construct a judgment matrix, and the weight of each factor is determined after a consistency test. The graded resistance values of each resistance factor are determined separately, and the graded resistance values of each resistance factor are weighted and superimposed according to the weight to obtain the comprehensive space resistance surface.
3. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation as described in claim 1, characterized in that: The spatial resistance coefficient is the average value of the combined resistance surface value of each grid on the line connecting the geometric centers of the supply area and the demand area.
4. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation as described in claim 1, characterized in that: The formula for calculating the strength of the ecological link is: in, For the supply area With demand area The strength of ecological connections between them For the supply area With demand area The Euclidean distance between them The space drag coefficient is... This represents the maximum distance between all supply and demand pairs within the study area. For the supply area Net supply and demand for carbon sequestration For demand area The absolute value of net carbon sequestration supply and demand.
5. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation according to claim 1, characterized in that: The actual amount of carbon sequestration services provided by each supply region to each demand region includes: For each supply area, the intensity of its ecological connection with each demand area is normalized to obtain the allocation ratio. The net carbon sequestration supply and demand of the supply area is allocated to each demand area according to the allocation ratio to obtain the initial allocation service volume. For each demand region, when the sum of the initial allocation services provided by all supply regions to that demand region is greater than the absolute value of the net carbon sequestration supply and demand of that demand region, the initial allocation services are compressed proportionally according to the ratio of the absolute value of the net carbon sequestration supply and demand to the sum of the initial allocation services; when the sum of the initial allocation services is not greater than the absolute value of the net carbon sequestration supply and demand of that demand region, the initial allocation services remain unchanged. The initial allocation of services, whether compressed or unchanged, will be used as the actual amount of carbon sequestration services provided by each supply region to each demand region.
6. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation according to claim 1, characterized in that: The machine learning model is the CatBoost model. Before training, Pearson correlation analysis is used to screen candidate features. Based on the correlation between each candidate feature and the prediction target, candidate features with weak explanatory power are removed. When the correlation coefficient between any two candidate features is greater than the preset collinearity threshold, only the features with higher correlation to the prediction target are retained. Among them, when the regional carbon sequestration capacity is the prediction target, net primary productivity and soil heterotrophic respiration are removed from the candidate features before the screening.
7. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation according to claim 1, characterized in that: The average SHAP importance of each feature variable used to train the machine learning model is calculated through SHAP interpretability analysis. The top three key driving factors in terms of importance are selected, and the SHAP importance values of each key driving factor are normalized to obtain relative weights. The relative weights are then weighted and summed with the standardized values of the corresponding key driving factors after min-max normalization to construct the ecological contribution index and the payment responsibility index, respectively.
8. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation according to claim 1, characterized in that: The formula for calculating the two-factor adjustment coefficient is as follows: in, For the supply area With demand area The two-factor adjustment coefficient between them; and The adjustment weights for the ecological contribution index and the payment responsibility index are respectively. and equal; For the supply area Ecological contribution index. The average ecological contribution index of all supply areas; For demand area Payment responsibility index This represents the average payment responsibility index across all demand areas. and These are the lower and upper limits of the adjustment coefficient, respectively. This is a truncation function, used when the value of the independent variable is less than... or greater than The corresponding boundary value is taken at the time; the compensation amount is the product of the basic compensation amount and the dual-factor adjustment coefficient, and the basic compensation amount is the product of the carbon sequestration service volume and the carbon price.
9. The cross-regional carbon ecological compensation optimization method based on spatial resistance-constrained carbon sequestration service flow simulation according to claim 8, characterized in that: and All values are 0.
5. Take 0.5, Take 1.
5.
10. A cross-regional carbon ecological compensation optimization system based on spatial resistance-constrained carbon sequestration service flow simulation, characterized in that, include: The carbon supply and demand identification module is used to obtain the carbon sequestration service supply and carbon emissions of each spatial unit, determine the net supply and demand of carbon sequestration, and identify supply areas and demand areas. The integrated resistance surface construction module is used to construct an integrated spatial resistance surface and determine the spatial resistance coefficient between each supply area and each demand area. The carbon sequestration service flow simulation module is used to introduce the spatial resistance coefficient into the distance decay function of the comparative ecological radiation force model, and use the product of geometric distance and spatial resistance coefficient as the equivalent transmission distance to calculate distance decay, so as to obtain the ecological connection strength between each supply area and each demand area. The carbon sequestration service conversion module is used to determine the actual amount of carbon sequestration services provided by each supply area to each demand area based on the ecological linkage strength, combined with the carbon sequestration capacity constraints of the supply area and the carbon demand constraints of the demand area. The dual-index construction module is used to train machine learning models and determine key driving factors and their contribution weights through interpretability analysis, thereby constructing an ecological contribution index and a payment responsibility index. The compensation amount calculation module is used to construct a two-factor adjustment coefficient based on the deviation of the ecological contribution index from the average ecological contribution index of all supply areas and the deviation of the payment responsibility index from the average payment responsibility index of all demand areas. Based on the two-factor adjustment coefficient, the carbon sequestration service volume and the carbon price, the compensation amount is calculated, and compensation scheme data containing the correspondence between supply areas, demand areas, carbon sequestration service volume and compensation amount is output.
