Method for constructing high-resolution carbon emission inventory based on multi-source data fusion of gwr model

Abstract

This invention discloses a high-resolution carbon emission inventory construction method based on multi-source data fusion using the GWR model. It improves traditional downscaling methods by collecting carbon emission statistics from the power, industrial, transportation, and residential sectors, and combining them with various spatial proxy data such as nighttime light pollution, road density, population density, and point source emission locations. The GWR model is used to dynamically capture the spatial heterogeneity of carbon emissions, generating an enhanced spatial proxy index. Experimental results show that this method significantly outperforms traditional methods in terms of spatial resolution and prediction accuracy, providing technical support for precise carbon emission management and policy formulation.

Description

Technical Field

[0001] This invention relates to the field of carbon emission accounting technology, specifically to a method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model. Background Technology

[0002] With the increasing severity of global climate change, carbon emission control has become a crucial issue for global environmental protection and economic development. Accurate carbon emission inventories are essential for formulating reasonable emission reduction policies and assessing their effectiveness. However, traditional carbon emission inventories are typically based on administrative regions, resulting in low resolution and difficulty in reflecting the spatial heterogeneity of actual emission distribution, especially in highly urbanized areas where the spatial pattern of carbon emissions exhibits significant unevenness. Therefore, creating high-resolution carbon emission inventories is one of the significant technical challenges in current carbon emission research.

[0003] In recent years, data methods based on remote sensing and geographic information systems have provided new possibilities for the spatialization of carbon emissions research. Among these, nighttime light source data, population density data, and industrial emission location data have become important spatial proxy variables for studying the spatial distribution of carbon emissions. Nighttime light source data can be used as a proxy indicator of human activity intensity and has a strong correlation with carbon emissions; population density data reflects the spatial distribution of residential activities and energy consumption; road density data is closely related to traffic emission intensity; and point source emission location data provides direct spatial distribution information on high-polluting industries and facilities.

[0004] While the aforementioned data provides a wealth of foundational information for creating high-resolution carbon emission inventories, integrating this multi-source data and effectively improving prediction accuracy remains a key challenge. Geographically weighted regression (GWR) models are widely used tools in spatial statistical analysis. By introducing spatial location weights, they can effectively capture the spatial heterogeneity and local variation characteristics among variables. Therefore, utilizing GWR models in conjunction with nighttime light, population density, road density, and point source emission data can provide a more accurate and spatially resolved solution for carbon emission inventory creation.

[0005] Existing research primarily focuses on the spatialization of carbon emissions from single data sources or atmospheric inversion models. This results in insufficient resolution and accuracy in the generated inventories, failing to capture the spatial characteristics of different emission sources. A multi-source data integration method based on the GWR model, by leveraging the complementarity and synergy of different data sources, can significantly improve the accuracy of carbon emission inventories, providing a scientific basis for spatial management and policy support of carbon emissions. Furthermore, this method can adapt to application needs at different scales and regions, possessing strong universality and promotional value.

[0006] Currently, several mature and representative methods and databases have been developed for creating global and regional carbon emission inventories, including ODIAC, EDGAR, and MEIC. These inventories are widely used in carbon emission research and policy-making, but they still have limitations in practical applications. For example, ODIAC provides a global carbon emission inventory with a spatial resolution of 1 km. Although it has a high resolution on a global scale, it does not utilize spatial proxy data to characterize the linear emission characteristics of the transportation sector. Furthermore, due to the lack of data availability, the locations of industrial emissions in some regions are difficult to obtain, so the current emission inventories lack point source emission data and cannot meet the needs of precise emission reduction at the local scale. Similarly, the EDGAR and MEIC inventories use atmospheric inversion models to invert carbon emissions, resulting in a coarser spatial resolution and failing to reflect detailed emission characteristics such as transportation linear source emissions and industrial point source emissions.

[0007] Currently, many studies have explored methods for constructing high-resolution carbon emission inventories, but most employ simple weighting or linear regression methods, failing to fully integrate collaborative information from multiple sources such as nighttime light pollution, road density, and point source emission locations. This approach struggles to accurately reflect the spatial heterogeneity among different emission sources, leading to insufficient prediction accuracy. Particularly in regions with significant regional differences, the model results are prone to systematic errors. Summary of the Invention

[0008] Accurate and high-resolution carbon emission inventory creation has become a crucial aspect of carbon emission reduction research. More accurate spatial modeling of carbon emissions can facilitate more accurate atmospheric model simulations and emission source tracing. Therefore, the objective of this invention is:

[0009] 1. Improve the spatial resolution of carbon emission inventories and refine the emission characteristics of different regions;

[0010] 2. In the process of modeling multi-source emissions at points, lines, and surfaces, make full use of the collaborative information of various types of data to optimize the multivariate modeling method;

[0011] 3. Overcome the problem that existing models are not good at handling local heterogeneity and enhance the model's ability to characterize regional carbon emission differences;

[0012] This paper presents a method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model. By integrating statistical data and spatial proxy data, it achieves a 1km spatial resolution gridded allocation of carbon emissions, providing support for carbon emission research and management at different scales.

[0013] To achieve the above objectives, the present invention provides the following technical solution:

[0014] A high-resolution carbon emission inventory construction method based on multi-source data fusion using the GWR model includes the following steps:

[0015] S-1. Data Collection and Processing: Collect energy consumption and spatial distribution data from four sectors: electricity, industry, transportation, and civil use; multiply the energy consumption by the corresponding carbon emission factor to obtain the carbon emissions of each sector; sum the carbon emissions of the four sectors to obtain the total carbon emissions of the region.

[0016] S-2. Establishing a carbon emission downscaling model: A geographic weighted regression (GWR) model is used to construct the relationship between total carbon emissions and the spatial proxy index. Total carbon emissions are allocated to a 1km resolution grid, as shown in the following formula:

[0017]

[0018] in, This represents the carbon emission regression value of sector i within region k in year t; the subscripts i = 1, 2, 3, 4 correspond to the four sectors of electricity, industry, transportation, and civil use, respectively, and year t is a calendar year;

[0019] β n (u k ,v k The slope represents the sensitivity of carbon emissions to spatial distribution data, reflecting the increase in carbon emissions caused by each unit change in spatial distribution data; where (u) k ,v k () represents the geographic coordinates of region k; This represents the spatial distribution data of the electricity and industrial sectors obtained from the nighttime light index of the i-th sector in the k-th region in the t-th year; This represents the spatial distribution data of the transportation or civil sector of the i-th department in the k-th region for year t, obtained through formula coupling. 0.34 is the default threshold value for nighttime lighting.

[0020] S-3, Correction of carbon emission regression values: Adjust the regression values ​​obtained from the regression model using the following formula:

[0021]

[0022] in, This represents the corrected carbon emission grid value for the k-th region. This represents the high-resolution grid value of carbon emissions for the k-th region obtained using the GWR regression model; This represents the total predicted carbon emissions for the k-th region obtained using the GWR regression model. This means that the energy consumption is multiplied by the corresponding carbon emission factor to obtain the carbon emission calculation value of each sector in the kth region, thus obtaining the carbon emission inventory.

[0023] Preferably, the power sector uses publicly available global power plant location data to obtain spatial distribution data, extracts the nighttime light values ​​corresponding to the power plant locations, and uses the extracted nighttime light values ​​as the power sector's spatial distribution data; the nighttime light values ​​use global 500-meter resolution "NPP-VIIRS-like" nighttime light data.

[0024] Preferably, the location data of the industrial sector is obtained by using NASA's VIIRS 375m active fire product. The DBSCAN spatial clustering algorithm based on density noise is used to remove scattered fire points and initially screen out industrial locations that are close to each other. Then, combined with land use data, fire points on artificial surfaces are screened again to effectively distinguish industrial heat sources from other thermal anomalies. To avoid duplicate statistics, industrial heat source points within the power plant area are excluded, thereby obtaining accurate industrial sector locations. The nighttime light values ​​corresponding to the processed industrial sector locations are extracted, and the extracted nighttime light values ​​are used as the spatial distribution data of the industrial sector.

[0025] Preferably, the transportation department calculates road density data using publicly available global road data OpenStreetMap and couples it with nighttime light data to obtain spatial distribution data;

[0026] The residential sector obtains spatial distribution data by coupling publicly available global population density data and nighttime light data from the LandScan database.

[0027] Preferably, the spatial distribution data coupling formula for the transportation and residential sectors is as follows:

[0028]

[0029] Among them, D i NTL represents spatial distribution data, where i represents the transportation or residential sector, D represents road or population density data, and NTL represents the nighttime lighting index, with 0.34 being the default threshold for nighttime lighting values.

[0030] Preferably, region k represents a country, province, or city; the spatial distribution data is the nighttime light index.

[0031] Preferably, the density-based noise spatial clustering algorithm DBSCAN is used, wherein the clustering distance is set to 800 meters and the number of clusters for the minimum point feature is set to 5.

[0032] Compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. This invention takes into account both the integrity of statistical data and the refined characteristics of spatial proxy data. By optimizing traditional downscaling methods, it significantly improves the spatial resolution and prediction accuracy of carbon emission inventories.

[0034] 2. This invention fully utilizes multi-source spatial proxy data, such as nighttime light, population density, road density, and point source emission locations, to establish an efficient mapping relationship between carbon emission statistics and spatial proxy data. Compared to traditional downscaling methods based on single proxy data, the method of this invention achieves accurate allocation based on the spatial characteristics of carbon emissions from the power, industrial, transportation, and civil sectors at a spatial resolution of 1 km.

[0035] 3. This invention introduces a GWR model into the construction of carbon emission inventories. By using a weighted matrix to consider the influence of spatial location on model parameters, it can dynamically capture the spatial heterogeneity of carbon emissions in different regions. Compared with traditional global regression models, this method can better characterize the changes in carbon emission characteristics within a region, providing more refined carbon emission distribution data for areas with significant regional differences.

[0036] 4. Compared to atmospheric transport models and bottom-up statistical methods for calculating carbon emissions, this invention, through spatial proxy data calculation, significantly improves model operating efficiency. Furthermore, this method supports multi-scale applications, adapting to the needs of national, provincial, and municipal carbon emission inventory construction, and providing refined data support for carbon reduction policies at different management levels. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the overall method flow of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0039] Example:

[0040] Please see Figure 1 The present invention provides one of the technical solutions:

[0041] A high-resolution carbon emission inventory construction method based on multi-source data fusion using the GWR model includes the following steps:

[0042] S-1. Data Collection and Processing: Collect energy consumption and spatial distribution data from four sectors: electricity, industry, transportation, and civil use; multiply the energy consumption by the corresponding carbon emission factor to obtain the carbon emissions of each sector; sum the carbon emissions of the four sectors to obtain the total carbon emissions of the region.

[0043] The power sector uses publicly available global power plant location data to obtain spatial distribution data, extracts the nighttime light values ​​corresponding to the power plant locations, and uses the extracted nighttime light values ​​as the power sector's spatial distribution data; the nighttime light values ​​use global 500-meter resolution "NPP-VIIRS-like" nighttime light data.

[0044] The location data for the industrial sector was obtained using NASA's VIIRS 375m active fire product. The DBSCAN spatial clustering algorithm, which is based on density noise, was used to remove scattered fire points. The clustering distance of the algorithm was set to 800 meters, and the minimum number of clusters for each point feature was set to 5.

[0045] Initial screening identified industrial locations in close proximity. Combined with land use data, fire points on artificial surfaces were further screened to effectively distinguish industrial heat sources from other thermal anomalies. To avoid duplicate statistics, industrial heat sources within the power plant area were excluded, thus obtaining accurate industrial sector locations. The nighttime light values ​​corresponding to the processed industrial sector locations were extracted, and the extracted nighttime light values ​​were used as the spatial distribution data of the industrial sectors.

[0046] The transportation department calculates road density data using publicly available global road data from OpenStreetMap and couples it with nighttime light data to obtain spatial distribution data.

[0047] The residential sector obtained spatial distribution data by coupling publicly available global population density data and nighttime light data from the LandScan database.

[0048] The coupling formula for the spatial distribution data of the transportation and residential sectors is as follows:

[0049]

[0050] Among them, D i NTL represents spatial distribution data, where i represents the transportation or residential sector, D represents road or population density data, and NTL represents the nighttime lighting index, with 0.34 being the default threshold for nighttime lighting values.

[0051] S-2. Establishing a carbon emission downscaling model: A geographic weighted regression (GWR) model is used to construct the relationship between total carbon emissions and the spatial proxy index. Total carbon emissions are allocated to a 1km resolution grid, as shown in the following formula:

[0052]

[0053] in, This represents the carbon emission regression value of sector i within region k in year t; the subscripts i = 1, 2, 3, 4 correspond to the four sectors of electricity, industry, transportation, and civil use, respectively, and year t is a calendar year;

[0054] β n (u k ,v k The slope represents the sensitivity of carbon emissions to spatial distribution data, reflecting the increase in carbon emissions caused by each unit change in spatial distribution data; where (u) k ,v k () represents the geographic coordinates of region k; This represents the spatial distribution data of the electricity and industrial sectors obtained from the nighttime light index of the i-th sector in the k-th region in the t-th year; This represents the spatial distribution data of the transportation or civil sector of the i-th department in the k-th region for year t, obtained through formula coupling. 0.34 is the default threshold value for nighttime lighting.

[0055] S-3, Correction of carbon emission regression values: Adjust the regression values ​​obtained from the regression model using the following formula:

[0056]

[0057] in, This represents the corrected carbon emission grid value for the k-th region. This represents the high-resolution grid value of carbon emissions for the k-th region obtained using the GWR regression model; This represents the total predicted carbon emissions for the k-th region obtained using the GWR regression model. This means that the energy consumption is multiplied by the corresponding carbon emission factor to obtain the calculated carbon emission value for each sector in the k-th region.

[0058] This invention employs a multi-source data fusion method, organically combining traditional statistical data with high-resolution remote sensing data, providing a new technical solution for constructing high-precision carbon emission inventories. By combining the synergistic use of various spatial proxy data, it achieves refined spatial allocation of carbon emissions from the power, industrial, transportation, and civil sectors, filling the technical gap of insufficient spatial resolution in traditional carbon emission inventory production.

[0059] This invention proposes a downscaling method based on a geographically weighted regression model. It combines this model with the task of downscaling carbon emission statistics, offering a modeling approach that considers both global carbon emission statistical characteristics and captures local spatial heterogeneity. By introducing the weighting matrix of the geographically weighted regression, this invention can dynamically adjust model parameters, capturing the local variation characteristics of carbon emissions within a region, significantly improving the model's adaptability and prediction accuracy.

[0060] Furthermore, this invention is the first to combine NASA VIIRS fire point data with the DBSCAN clustering algorithm in point source emission identification. By combining this with methods for identifying industrial heat sources, the spatial location of industrial point sources is extracted, which improves the efficiency and accuracy of point source emission modeling and solves the problem of not being able to obtain the location of industrial emission point sources to a certain extent.

[0061] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model, characterized in that, The specific steps include: S-1. Data Collection and Processing: Collect energy consumption and spatial distribution data from four sectors: electricity, industry, transportation, and civil use; multiply energy consumption by the corresponding carbon emission factor to obtain the carbon emissions of each sector; sum the carbon emissions of the four sectors to obtain the total carbon emissions of the region. S-2. Establishing a carbon emission downscaling model: A geographic weighted regression (GWR) model is used to construct the relationship between total carbon emissions and the spatial proxy index. Total carbon emissions are allocated to a 1km resolution grid, as shown in the following formula: in, This represents the carbon emission regression value of sector i within region k in year t; the subscripts i = 1, 2, 3, 4 correspond to the four sectors of electricity, industry, transportation, and civil use, respectively, and year t is a calendar year; β n (u k ,v k The slope represents the sensitivity of carbon emissions to spatial distribution data, reflecting the increase in carbon emissions caused by each unit change in spatial distribution data; where (u) k ,v k () represents the geographic coordinates of region k; This represents the spatial distribution data of the electricity and industrial sectors obtained from the nighttime light index of the i-th sector in the k-th region in the t-th year; This represents the spatial distribution data of the transportation or civil sector of the i-th department in the k-th region in the t-th year, obtained through formula coupling previously. S-3, Correction of carbon emission regression values: Adjust the regression values ​​obtained from the regression model using the following formula: in, This represents the corrected carbon emission grid value for the k-th region. This represents the high-resolution grid value of carbon emissions for the k-th region obtained using the GWR regression model; This represents the total predicted carbon emissions for the k-th region obtained using the GWR regression model. This means multiplying the energy consumption by the corresponding carbon emission factor to obtain the calculated carbon emission value for each sector in the k-th region; this yields the carbon emission inventory.

2. The method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model according to claim 1, characterized in that: The power sector uses publicly available global power plant location data to obtain spatial distribution data, extracts the nighttime light values ​​corresponding to the power plant locations, and uses the extracted nighttime light values ​​as the power sector's spatial distribution data; the nighttime light values ​​use global 500-meter resolution "NPP-VIIRS-like" nighttime light data.

3. The method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model according to claim 1, characterized in that: The location data of the industrial sectors was obtained using NASA's VIIRS 375m active fire product. A density-based noise spatial clustering algorithm, DBSCAN, was used to remove scattered fire ignition points, initially identifying industrial sites with similar distances. This was then combined with land use data to further filter fire points on man-made surfaces, effectively distinguishing industrial heat sources from other thermal anomalies. To avoid duplicate statistics, industrial heat sources within the power plant area were excluded, thus obtaining accurate industrial sector locations. The nighttime light values ​​corresponding to the processed industrial sector locations were extracted, and these extracted nighttime light values ​​were used as the spatial distribution data of the industrial sectors.

4. The method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model according to claim 1, characterized in that: The transportation sector calculates road density data using publicly available global road data from OpenStreetMap and couples it with nighttime light data to obtain spatial distribution data; the civil sector obtains spatial distribution data by coupling publicly available global population density data and nighttime light data from the LandScan database.

5. The method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model according to claim 4, characterized in that: The spatial distribution data coupling formula for the transportation and civil sectors is as follows: Among them, D i NTL represents spatial distribution data, i represents the transportation or civil sector, and D i This represents road or population density data. NTL stands for Nighttime Lighting Index, and 0.34 is the default threshold for nighttime lighting values.

6. The method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model according to claim 1, characterized in that: The k region represents a country, province, or city; the spatial distribution data is the nighttime light index.

7. The method for constructing a high-resolution carbon emission inventory based on multi-source data fusion using the GWR model according to claim 2, characterized in that: The density-based noise spatial clustering algorithm DBSCAN is applied, with the clustering distance set to 800 meters and the number of clusters for the minimum point feature set to 5.

Images