A method for constructing local carbon emission zoning to overcome scale effect

Abstract

The application provides a local carbon emission partition construction method overcoming scale effect, and belongs to the technical field of urban low-carbon planning. The method is realized by the following technical scheme: firstly, multi-source city space data is integrated to calculate multi-dimensional city form factors; secondly, a data collapse method based on limited size scaling theory is introduced to perform a rescaling transformation on the factor probability density function under different spatial resolutions, and the collapsible city form factors with cross-scale statistical rules are screened out to overcome the scale effect caused by the variable area unit problem; then, an optimal parameter geographic detector model is applied to perform double optimization of the screening factor on the discretization method and factor combination, and a local carbon emission partition construction scheme with the highest explanatory force for the specific city CO2 emission heterogeneity is identified; finally, the Kruskal-Wallis non-parametric test and the coefficient of variation are used to evaluate the inter-group difference and the intra-group homogeneity of the partition. The application effectively solves the technical bottleneck of strong scale dependence and insufficient explanation of regional CO2 emission heterogeneity in the existing partition method, and can provide precise and customizable spatial unit division basis for low-carbon spatial planning and carbon emission reduction policy making.

Description

Technical Field

[0001] This invention belongs to the field of urban environmental planning and low-carbon management technology. Specifically, it relates to a technical framework that uses data collapse method to overcome scale effect and combines optimal parameter geographic detector to construct local carbon emission zones. It can be used for urban low-carbon spatial planning and precise carbon emission control. Background Technology

[0002] In the context of global climate change, the relationship between urban spatial morphology and CO2 emissions is central to low-carbon city research. However, existing spatial zoning methods face two major drawbacks in research and practical planning applications: First, they are highly scale-dependent. Traditional zoning methods typically rely on statistical aggregation based on administrative boundaries or fixed grid scales, making them heavily influenced by the variable area unit problem and exhibiting significant scale effects. The same morphological factor may show drastically different statistical relationships at different scales, leading to unstable zoning boundaries and difficulty in generalization. Second, they lack sufficient explanatory power and specificity for regional CO2 emission heterogeneity. Existing models often employ a uniform and fixed set of factors and discretization rules for urban CO2 emissions, ignoring the unique characteristics of different cities in terms of historical development, transportation networks, and spatial structures. When a fixed set of factors is imposed on a general model, it easily obscures the specific morphological trajectory of a city, resulting in low explanatory power for CO2 emission heterogeneity and difficulty in translating it into actionable planning policies. Summary of the Invention

[0003] Purpose of the invention: To address the shortcomings of the existing technologies, this invention provides a method for constructing local carbon emission zones that overcomes scale effects. The method aims to eliminate scale effect interference through the finite size scaling theory in physics, extract cross-scale stability factors, and automatically identify the optimal combination of factors and discretization thresholds suitable for specific cities using an optimal parameter geographic detector, thereby constructing local carbon emission zones with high explanatory power and strong internal homogeneity.

[0004] The technical solution adopted in this invention includes the following steps: Step 1: Data Preparation and Factor Calculation. Acquire urban built-up area boundary data, urban CO2 emission grid data, 3D building outline data, transportation network data, and land use data. Calculate multidimensional urban morphology factors, including buildings, transportation, and land use, within a unified baseline spatial grid. These factors include: Patch Density (PD), Number of Patches (NP), Number of Road Intersections (INTER), Non-Motorized Vehicle Lane Length (NMRL), Number of Public Transportation Stations (STAT), Functional Mix Entropy (FME), Park Area (PA), Shrub Area (SA), Motor Vehicle Lane Length (MRL), Building Coverage Ratio (BCR), Building Shape Factor (BSC), Spatial Crowding (SCD), Green Space Area (GSA), Average Building Height (MBH), Average Building Area (MBA), Average Building Volume (MBV), Forest Area (FA), and Maximum Patch Index (LPI). Step 2: Using a data collapse method based on finite size scaling theory, calculate the probability density function of each morphological factor at different spatial resolutions (1 km and 4 km). A Gaussian kernel is used as the kernel function, and the bandwidth is determined using the Silverman rule. Where σ is the standard deviation, IQR is the interquartile range, and n is the sample size. A power-law relationship is introduced for rescaling:

[0005] in It is the mean of variable V at grid scale l, and parameters a and b satisfy the normalization condition. F(x) is a general scaling function that satisfies the condition F(x)→0 as x→∞ and F(x) approaches a constant as x→0. The optimal power-law parameters are determined by minimizing the cumulative residual area of ​​the probability density function curves across different scales.

[0006] Traversal The optimal parameter is b, corresponding to the minimum value of E(α). The dual criteria for determining a collapsible factor are: (a) the optimal power-law parameter satisfies a≈1 and b≈1 (deviation ≤0.15); (b) the statistical fitting residual E(α) <0.1. Factors that simultaneously meet both criteria are identified as collapsible factors, while non-collapsible factors are eliminated if the power-law parameter deviates from the predetermined range or the residual exceeds the threshold. After the above screening, 11 collapsible factors were identified: forest area (FA), green space area (GSA), number of road intersections (INTER), average building volume (MBV), motor vehicle lane length (MRL), non-motor vehicle lane length (NMRL), number of patches (NP), park area (PA), patch density (PD), shrub area (SA), and number of public transport stops (STAT). The eliminated non-collapsible factors include: building coverage (BCR), building shape factor (BSC), functional mix entropy (FME), maximum patch index (LPI), average building height (MBH), and spatial crowding (SCD). The data was aggregated to a 4 km resolution, and the collapse process was repeated. The set of collapsible factors was consistent with the results at the 1 km resolution (still the same 11 factors), verifying the robustness of cross-scale screening. Step 3: City-specific model optimization. For the screened collapsible factors, an optimal parameter geographic detector model was applied. The system traversed various discretization methods, classifying the factors into high, medium, and low levels; then, under specific factor combinations, the q-values ​​of each combination were calculated, and the optimal combination and discretization parameters with the highest explanatory power for CO2 emission heterogeneity were selected. Step 3: For the 11 selected collapsible factors, apply the optimal parameter geographic detector model. The system iterates through five discretization methods (equal interval method, geometric breakpoint method, natural breakpoint method, quantile method, and standard deviation method) to classify the factors as high, medium, and low. The factor combination is limited to three factors (total...). (Several combinations), and calculate the q value for each parameter configuration for each combination:

[0007] Where N is the total number of samples in the study area, and σ² is the total variance. j and σ j ² represents the sample size and variance of the j-th stratum, respectively. The statistical significance of the q-value is determined using the F-test:

[0008] When p < 0.05, the factor combination is considered to have significant explanatory power. The optimal three-factor combination and its corresponding discretization parameters with the highest explanatory power for CO2 emission heterogeneity are selected.

[0009] Step 4: Using the optimal three-factor combination obtained in Step 3 and the high, medium, and low discretization thresholds for each factor, perform the following operations sequentially for each grid cell in a 1km × 1km spatial grid: (a) Read the values ​​of the grid cell on the three selected factors; (b) Determine the level of the grid cell on each factor according to the high / medium / low thresholds, encoding them as 3 (high), 2 (medium), and 1 (low) respectively. Theoretically, a maximum of [number] values ​​can be generated. (c) Merge adjacent grid cells with the same code into the same local carbon emission zone; if the codes are different, use the grid boundary as the partition boundary. This ultimately generates a spatial distribution of local carbon emission zones specific to each city. Figure 1 .

[0010] Step 5: Use the Kruskal-Wallis nonparametric test to verify the significant differences in CO2 emissions among different types of local carbon emission zones. The Kruskal-Wallis statistic H is defined as:

[0011] Where n is the total number of grid cells in the study area, n i Let R be the sample size of the i-th partition type. i Let be the rank sum of the i-th partition type. The effect size uses... measure:

[0012] when The data was determined to have a large effect size. Simultaneously, the coefficient of variation (CV) within each zone was calculated as CV = σ / μ (σ is the standard deviation, μ is the mean), and compared with the global baseline CV to verify intra-group homogeneity. When the zone CV was significantly lower than the global CV, a significant improvement in intra-group homogeneity was confirmed. To address the statistical masking effect caused by extreme point sources in the absolute emission data, the logarithm of the CO2 emission data was taken to control for intra-layer variance, and the optimal parameter geospatial detector model was reapplied for testing. This confirmed that taking the logarithm significantly increased the explanatory power of the geospatial detector, thus verifying that urban morphology factors have a robust explanatory power for the carbon emission gradient.

[0013] Step Six: Use Getis-Ord Statistical analysis identifies hot and cold spots in CO2 emissions:

[0014] in x j For the attribute value of feature j, w i,j The spatial weights between features i and j S is the mean of all feature attribute values, and S is the standard deviation. The calculated... The z-score is used to define hotspots (high-emission clusters) with a significant positive z-score (p<0.05) and coldspots (low-emission clusters) with a significant negative z-score. The distribution characteristics of different local carbon emission zones in hotspot and coldspot areas are analyzed, and differentiated low-carbon planning strategies targeting specific morphological factors are output. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the spatial distribution of local carbon emission zones provided in an embodiment of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to specific embodiments.

[0017] Example 1: Construction of Localized Carbon Emission Zones in Typical Global Metropolitan Areas Step 1: Data Preparation. Four cities—London, New York, Paris, and Sydney—were selected as the study area. CO2 emission data were obtained using the ODIAC2023 1 km × 1 km resolution grid; 3D building data were obtained using the 3D-GloBFP sub-meter resolution 3D building outline dataset; traffic and land use data were obtained from Open Street Map. Within each CO2 data grid, 18 urban morphological factors were calculated, including Patch Density (PD), Number of Patches (NP), Number of Road Intersections (INTER), Non-Motorized Lane Length (NMRL), Number of Public Transport Stations (STAT), Functional Mixed Entropy (FME), Park Area (PA), Shrub Area (SA), Motorized Lane Length (MRL), Building Coverage (BCR), Building Shape Factor (BSC), Spatial Crowding (SCD), Green Space Area (GSA), Average Building Height (MBH), Average Building Area (MBA), Average Building Volume (MBV), Forest Area (FA), and Maximum Patch Index (LPI).

[0018] The second step, cross-scale factor selection, involves aggregating the data to 1 km and 4 km resolutions respectively, and constructing probability density functions using kernel density estimation. Gaussian kernel density estimation is used to construct the probability density functions, with the bandwidth determined using the Silverman rule. By using data collapse methods, traversing... Minimize the cumulative residual area The optimal power-law parameters were determined by minimizing the cumulative residual area using a data collapse method. Eleven collapsible factors (FA, GSA, INTER, MBA, MBV, MRL, NMRL, NP, PA, PD, SA, STAT) were selected that met the requirements of residuals below 0.1 and power-law parameters. Non-collapsible factors (BCR, BSC, FME, LPI, MBH, SCD) that lacked cross-scale consistency and were susceptible to scale effects were removed. The collapse process was repeated at a 4 km resolution, and the set of collapsible factors was completely consistent with the 1 km result, verifying the robustness of the selection.

[0019] The third step, model optimization, involves applying an optimal parameter geographic detector model to the 11 collapsible factors. Five discretization methods are tested: equal interval, geometric breakpoint, natural breakpoint, quantile, and standard deviation. Factor combinations are limited to three factors. For London, the optimal combination is INTER+PD+STAT, with standard deviation as the discretization method and a q-value of 0.026. For New York, the optimal combination is STAT+MBV+PA, with natural breakpoint as the discretization method and a q-value of 0.126. For Paris, the optimal combination is MBV+MRL+PA, with quantile as the discretization method and a q-value of 0.0330. For Sydney, the optimal combination is MBV+NMRL+PD, with geometric breakpoint as the discretization method and a q-value of 0.247.

[0020] The fourth step is to generate regional zones: based on the high, medium and low thresholds of the optimal combination of cities, spatial overlay is performed. London generates 27 local carbon emission zones, New York generates 18, Paris generates 9 and Sydney generates 17.

[0021] Step 5: Statistical validation: The Kruskal-Wallis test showed that the asymptotic significance of all four cities was 1×10⁻⁶. -6 The effect sizes all exceeded 0.14 (0.143-0.233), indicating significant differences between groups. Within-group homogeneity tests showed that the average within-group coefficient of variation for each city decreased significantly compared to the global baseline coefficient of variation (a reduction of over 80%). For example, London's global coefficient of variation was 8.051, while the average within-group coefficient of variation dropped to 1.853, with 96.30% of the partitions achieving variance compression, demonstrating that the scaling effect was effectively overcome. To overcome the statistical masking effect caused by extreme point sources in the absolute carbon emission data, the logarithm of the CO2 emission data was further taken to control for intra-layer variance, and the optimal parameter geospatial detector model was reapplied for validation. The results show that taking the logarithm significantly increased the explanatory power of the geospatial detector; for example, the highest q-value for London reached 0.459, and the highest q-value for Paris reached 0.322, confirming that the selected optimal factor combination has a robust explanatory power for the step gradient of carbon emissions.

[0022] Step 6: Strategy Output: Through hot and cold spot analysis, identify the dominant local carbon emission zone types in each city's low-carbon spaces. For example, for the 1-1-1 type (low intersections, low building volume, low park area) that dominates in New York's hot and cold spots, propose strategies to optimize transportation network connectivity and increase vertical greening to reduce emissions.

Claims

1. A method for constructing local carbon emission zones to overcome scale effects, characterized in that... Includes the following steps: Step 1, data preparation, includes acquiring urban built-up area boundary data, urban CO2 emission spatial distribution data, three-dimensional building footprint data, transportation network data and land use data, and calculating multi-dimensional urban morphology factors related to buildings, transportation and land use within a unified spatial grid; Step 2: Based on the urban morphology factors obtained in Step 1, a data collapse method based on the finite size scaling theory is used to rescale the probability density functions of the factors under different spatial resolutions, and to screen out the collapsible morphology factors that satisfy cross-scale statistical laws in order to overcome the scale effect caused by the variable area unit problem. Step 3: Based on the collapsible morphology factors selected in Step 2, apply the optimal parameter geographic detector model to test various discretization methods and select a three-factor combination to maximize the explanatory power for the heterogeneity of CO2 emissions. Step 4: Based on the optimal three-factor combination obtained in Step 3 and the high, medium and low discretization threshold parameters of each factor, the factor level is determined and the three-factor code combination is performed on each grid cell in turn. Adjacent grid cells with the same type of code are merged and spatially superimposed to generate a local carbon emission zone. Step 5: Use the Kruskal-Wallis nonparametric test, coefficient of variation and logarithmic control layer variance method to verify the intergroup significance and intragroup homogeneity of the morphological partitions obtained in Step 4. Step 6: Based on the optimal factor combination obtained in Step 3, and combined with the spatial autocorrelation analysis results of CO2 emission hotspots and colds, tailor differentiated low-carbon planning intervention measures for specific morphological factors according to the distribution characteristics of different types of local carbon emission areas in cold / hotspot regions.

2. The method according to claim 1, characterized in that... In step 2, the data collapse method is based on the finite-size scaling theory, specifically including: dividing the city into grid cells with a side length of l and calculating the target city morphology factor V; constructing probability density functions at different spatial resolutions using kernel density estimation, with the kernel function being a Gaussian kernel and the bandwidth determined by the Silverman rule: Where σ is the standard deviation, IQR is the interquartile range, and n is the sample size; a power-law relationship is introduced: ； in It is the mean of variable V at grid scale l, and parameters a and b satisfy the normalization condition. F(x) is a general scaling function; it is achieved by minimizing the cumulative residual area between probability density function curves at different spatial resolutions. Determine the optimal power-law parameters and iterate through... The optimal parameter is b, which corresponds to the minimum value of E(α). The optimal power law parameter satisfies a≈1 and b≈1 (deviation ≤0.15). Factors with statistical fitting residuals E(α) <0.1 and power law parameters within the above predetermined range are identified as collapsible urban morphology factors, and non-collapsible factors whose power law parameters deviate from the predetermined range are eliminated.

3. The method according to claim 1, characterized in that... In step 3, the discretization method includes five types: equal interval method, geometric breakpoint method, natural breakpoint method, quantile method, and standard deviation method; the selection method for the optimal three-factor combination is: exhaustively enumerating all three-factor combinations (total) among the collapsible factors selected in step 2. (where m is the number of collapseable factors), and for each combination, apply five discretization methods to calculate the q value under each parameter configuration: ； Where N is the total number of samples in the study area, and σ² is the total variance. j and σ j ² represents the sample size and variance of the j-th stratum, respectively; the parameter configuration with the highest q-value is selected as the optimal solution, and its statistical significance is verified by the F-test: ； When p < 0.05, the explanatory power of the factor combination is considered statistically significant.

4. The method according to claim 1, characterized in that... In step 4, the specific method for generating local carbon emission zones by spatial overlay is as follows: In a unified spatial grid (1 km × 1 km), perform the following operations on each grid cell in sequence: (a) read the values ​​of the grid cell on three selected factors; (b) determine the level of the grid cell on each factor according to the high / medium / low thresholds determined in step 3, and encode them as 3 (high), 2 (medium), and 1 (low) respectively; (c) merge adjacent grid cells with the same code into the same local carbon emission zone, and use the grid boundary as the partition boundary when the codes are different.

5. The method according to claim 1, characterized in that... In step 5, the statistic H of the Kruskal-Wallis nonparametric test is defined as: ； Where n is the total number of grid cells in the study area, n i Let R be the sample size of the i-th partition type. i The rank sum of the i-th partition type; the effect size is... measure: ； when The large effect size indicates that the zoning scheme has significant inter-group differences. The coefficient of variation verification includes calculating the CO2 emission coefficient of variation (CV) = σ / μ (where σ is the standard deviation and μ is the mean) within each zoning and comparing it with the global baseline coefficient of variation of the unzoned city. When the zoning CV is significantly lower than the global CV, it confirms a significant improvement in internal homogeneity. In addition, it also includes taking the logarithm of the CO2 emission data to control the intra-layer variance in order to address the statistical masking effect caused by extreme point sources in the absolute emission data, and reapplying the optimal parameter geospatial detector model for testing. It is confirmed that the explanatory power of the geospatial detector is significantly increased after taking the logarithm, thereby verifying that the urban morphology factor has a robust explanatory power for the carbon emission gradient.

6. The method according to claim 1, characterized in that, The method also includes step 6, in which the spatial autocorrelation analysis of CO2 emission hotspots is performed using Getis-Ord. Statistic: ； Where x j w is the attribute value of feature j. i,j The spatial weights between features i and j S is the mean of all feature attribute values, and S is the standard deviation; the calculated... The z-score is used to determine whether a positive z-score indicates a hotspot (high-emission cluster) or a negative z-score indicates a coldspot (low-emission cluster). Based on the distribution characteristics of different local carbon emission zones in hotspot and coldspot areas, differentiated low-carbon planning intervention strategies are output for specific morphological factors.

Images