TVOC emission remote sensing monitoring method based on multi-source data driving

Abstract

The invention discloses a TVOC emission remote sensing monitoring method based on multi-source data driving, and the method comprises the following steps: analyzing related environmental factor data affecting TVOC emission according to the total concentration of TVOC measured on the ground in combination with various environmental factors; estimating and fitting an atmospheric methane vertical profile, and calculating a ratio of the surface methane concentration to the whole column methane concentration; establishing a TVOC emission model by using a Gaussian process regression algorithm by acquiring and matching each parameter; and measuring and calculating the TVOC emission amount of each grid in the monitoring range by using the model to obtain TVOC emission data fully covered by the monitoring area. The beneficial effects of the invention are that the method can obtain a high daily temporal-spatial resolution, greatly improves the spatial coverage of VOC emission data, meets the demands of regional refined monitoring, reduces the consumption of computing resources, improves the computing efficiency, can rapidly obtain the VOC emission data, and improves the reliability of the VOC emission data. The cost is greatly reduced; and the production efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and in particular to a remote sensing monitoring method for TVOC emissions based on multi-source data. Background Technology

[0002] Total Volatile Organic Compounds (TVOCs) are organic compounds that have high saturated vapor pressure, low boiling points, and are easily volatile under standard conditions. They typically include non-methane hydrocarbons and organic compounds containing oxygen, chlorine, nitrogen, and sulfur. TVOCs are important precursors to the formation of fine particulate matter (PM2.5) and ozone (O3) in the atmosphere, negatively impacting both the environment and health. TVOCs have a wide range of sources, which can be broadly categorized into natural sources and anthropogenic sources. Natural sources mainly include natural processes such as plant metabolism and volcanic eruptions; while anthropogenic sources are concentrated in industrial activities, traffic emissions, and building decoration. For example, solvents and coatings frequently used in chemical, printing, and surface coating processes release large amounts of TVOCs.

[0003] Currently, the main methods for estimating TVOC emissions include the emission factor method and physical model simulation. These two methods can be used to generate emission inventories at different scales, covering countries, regions, cities, and industrial parks. However, existing emission inventories have the following shortcomings: 1. Low spatial resolution: Most existing emission inventories are based on statistical results from administrative divisions or have a spatial resolution of 0.05° or even lower. This means that they can only provide overall TVOC emissions over large spatial areas and cannot accurately reflect emission differences in minute areas.

[0004] 2. Lack of monitoring and validation: Existing model results are mostly estimated data, lacking effective validation against actual emissions. For example, the traditional emission factor method relies on emission source activity data and corresponding emission factors. While it can provide statistical results over a wide range, it is difficult to conduct refined regional emission monitoring. Although emission models based on physical mechanisms, such as Megan's, can simulate VOC emission processes from natural sources, their spatial resolution is also limited, and they lack support from actual observational data.

[0005] Therefore, current TVOC emission monitoring and estimation methods still have considerable room for improvement in terms of accuracy and verifiability, especially in obtaining accurate TVOC emission data in higher resolution areas, which has become an urgent technical problem to be solved. Summary of the Invention

[0006] To address the problems of low spatial resolution, computational complexity, long processing time, and lack of reliable verification in existing TVOC emission monitoring technologies, this invention provides a remote sensing monitoring method for TVOC emissions based on multi-source data. By combining satellite remote sensing observation data, ground-based measured data, and atmospheric chemical model data, a Gaussian process regression model is established. This enables the rapid and efficient acquisition of high-precision, regionally comprehensive VOC emission data, meeting the requirements for refined TVOC emission data and providing data support for quantifying and monitoring TVOC emissions.

[0007] To achieve the above-mentioned objectives, this invention provides a remote sensing monitoring method for TVOC emissions based on multi-source data, the monitoring method comprising the following steps: Step S1: Based on the total TVOC concentration measured on the ground, analyze the relevant environmental factors affecting TVOC emissions in conjunction with various environmental factors; Step S2: Acquire and process the methane column average dry air mixture ratio concentration data and methane stratified concentration data, estimate and fit the atmospheric methane vertical profile, and calculate the ratio of surface methane concentration to whole column methane concentration through the atmospheric methane vertical profile. Convert the methane column average dry air mixture ratio concentration data into surface methane concentration data. Step S3: By acquiring and matching the total TVOC concentration data measured on the ground, the surface methane concentration data, and the relevant environmental factor data affecting TVOC emissions, a TVOC emission model is established using the Gaussian process regression algorithm, and 10-fold cross-validation is performed. Step S4: Based on the TVOC emission model established in Step S3, the TVOC emission of each grid within the monitoring range is calculated to obtain TVOC emission data covering the entire monitoring area.

[0008] The relevant environmental factors affecting TVOC emissions analyzed in step S1 include temperature factor, wind speed factor, water vapor factor, topography factor, and atmospheric methane concentration factor.

[0009] The TVOC data measured on the ground was obtained by mobile measurement with the measuring instrument fixed on a mobile vehicle, and has a spatial resolution of 0.0001°.

[0010] The methane column mean dry air mixing ratio concentration data are from the second-level (L2) offline (OFFL) data product of satellite methane column concentration from TROPOMI; the methane stratification concentration data are from the methane stratification concentration data simulated by the ECMWF atmospheric chemistry model CAMS; the temperature factor is from MODIS surface temperature data; the wind speed factor is from ERA horizontal and vertical wind speed data; the water vapor factor is from MODIS atmospheric water vapor content data; the topographic factor is from USGS digital elevation model (DEM) data; the atmospheric methane stratification concentration data are from CAMS EGG4 stratified methane concentration data; the atmospheric methane vertical column concentration factor is from TROPOMI XCH4 data product; in addition, the spatial resolution of all data is uniformly resampled to 0.001° × 0.001°, and the time range is resampled to days.

[0011] The TROPOMI methane column mean dry air mixture ratio concentration data is an integral average from the Earth's surface to the top of the atmosphere. The ratio of surface methane concentration to the overall column methane concentration is calculated using an atmospheric methane vertical profile, converting the TROPOMI methane column mean dry air mixture ratio concentration data into surface methane concentration data. The atmospheric methane vertical profile is obtained by fitting methane concentration data from the Earth's surface to the top of the atmosphere.

[0012] Step S3 specifically involves: Step S301: Obtain the modeling dataset by matching the TVOC data measured by ground mobile survey with the surface methane concentration data and related environmental factor data; Step S302: Model TVOC emissions using a Gaussian process regression model. The response function of the Gaussian process regression model is defined according to the following formula.

[0013] In the formula y i It is x i The corresponding objective function, σ 2 is the error variance, and β is the coefficient, both calculated from the sample data.

[0014] The TVOC emission model was validated using 10-fold cross-validation. During the validation process, the dataset was randomly divided into 10 equal parts, with 9 parts used for training and 1 part used for validation each time.

[0015] The raster in step S4 is the corresponding raster data obtained by spatial rasterization of temperature, wind speed, water vapor, topographic factor, atmospheric methane concentration factor, TROPOMI methane column concentration data and CAMS methane stratification concentration data. The spatial resolution of the raster data is uniformly 0.001°×0.001°.

[0016] The beneficial effects of this invention are as follows: Based on satellite remote sensing landing data and ground-based measured data, this invention, through the establishment of a Gaussian process regression model, can obtain VOCs with a daily spatiotemporal resolution of up to 0.001° (approximately 100m). Compared with traditional emission inventory methods, this invention significantly improves the spatial coverage of VOC emission data, meeting the needs of refined regional monitoring. Compared with emission estimation methods based on physical model simulations, this invention uses on-site mobile measured data as the true value and utilizes a Gaussian process regression model, which not only reduces the consumption of computational resources but also improves computational efficiency, enabling rapid acquisition of VOC emission data, greatly reducing costs and improving production efficiency. This invention uses ground-based measured data as the true value adjustment model, significantly improving the spatial resolution and reliability of VOC emission data, and is particularly suitable for the accurate monitoring of TVOC emissions in the petrochemical industry, providing sufficient technical support for the industry's environment. Attached Figure Description

[0017] Figure 1 The flowcharts are for Embodiments 1 and 2 of the present invention.

[0018] Figure 2 This is the effect of simulating atmospheric methane profile based on CAMS data in Embodiment 2 of the present invention.

[0019] Figure 3 This is the accuracy result of the ten-fold cross-validation of the Gaussian process regression model in Embodiment 2 of the present invention.

[0020] Figure 4 This is the spatial distribution of estimated regional TVOC emissions data in Embodiment 2 of the present invention.

[0021] Figure 5 The data shows the BVOC emission results in the comparative examples of this invention. Detailed Implementation

[0022] To clearly illustrate the technical features of this solution, the following detailed implementation method will be used to explain the solution.

[0023] Example 1 This invention provides a remote sensing monitoring method for TVOC emissions based on multi-source data, the monitoring method comprising the following steps: Step S1: Based on the ground-measured total TVOC concentration, analyze the relevant environmental factors affecting TVOC emissions, combining various environmental factors; among them, the analyzed relevant environmental factors affecting TVOC emissions include temperature factor, wind speed factor, water vapor factor, topographic factor, and atmospheric methane concentration factor. The ground-measured TVOC data is obtained by mobile measurement with the measuring instrument fixed on a mobile vehicle, and has a spatial resolution of 0.0001°.

[0024] Step S2: Acquire and process the methane column average dry air mixture ratio concentration data and methane stratified concentration data, estimate and fit the atmospheric methane vertical profile, and calculate the ratio of surface methane concentration to whole column methane concentration through the atmospheric methane vertical profile. Convert the methane column average dry air mixture ratio concentration data into surface methane concentration data. Step S3: By acquiring and matching the total TVOC concentration data measured on the ground, the surface methane concentration data, and the relevant environmental factor data affecting TVOC emissions, a TVOC emission model is established using the Gaussian process regression algorithm, and 10-fold cross-validation is performed. Step S4: Based on the TVOC emission model established in Step S3, the TVOC emission of each grid within the monitoring range is calculated to obtain TVOC emission data covering the entire monitoring area.

[0025] The methane column mean dry air mixing ratio concentration data are from the second-level (L2) offline (OFFL) data product of satellite methane column concentration from TROPOMI; the methane stratification concentration data are from the methane stratification concentration data simulated by the ECMWF atmospheric chemistry model CAMS; the temperature factor is from MODIS surface temperature data; the wind speed factor is from ERA horizontal and vertical wind speed data; the water vapor factor is from MODIS atmospheric water vapor content data; the topographic factor is from USGS digital elevation model (DEM) data; the atmospheric methane stratification concentration data are from CAMS EGG4 stratified methane concentration data; the atmospheric methane vertical column concentration factor is from TROPOMI XCH4 data product; in addition, the spatial resolution of all data is uniformly resampled to 0.001° × 0.001°, and the time range is resampled to days.

[0026] The TROPOMI methane column mean dry air mixture ratio (MAG) concentration data is an integral value from the Earth's surface to the top of the atmosphere. The ratio of surface methane concentration to the overall column methane concentration is calculated using an atmospheric methane vertical profile, converting the TROPOMI Methane column mean dry air mixture ratio (MAG) concentration data into surface methane concentration data. The atmospheric methane vertical profile is obtained by fitting methane concentration data from the Earth's surface to the top of the atmosphere.

[0027] Step S3 specifically involves: Step S301: Obtain the modeling dataset by matching the TVOC data measured by ground mobile survey with the surface methane concentration data and related environmental factor data; Step S302: Model TVOC emissions using a Gaussian process regression model. The response function of the Gaussian process regression model is defined according to the following formula.

[0028] In the formula y i It is x i The corresponding objective function, σ 2 is the error variance, and β is the coefficient, both calculated from the sample data.

[0029] The TVOC emission model was validated using 10-fold cross-validation. During the validation process, the dataset was randomly divided into 10 equal parts, with 9 parts used for training and 1 part used for validation each time.

[0030] The raster in step S4 is the corresponding raster data obtained by spatial rasterization of temperature, wind speed, water vapor, topographic factor, atmospheric methane concentration factor, TROPOMI methane column concentration data and CAMS methane stratification concentration data. The spatial resolution of the raster data is uniformly 0.001°×0.001°.

[0031] Example 2 To enhance the monitoring of TVOC emissions from the petrochemical industry and provide data support for filling gaps in the emission inventory, this embodiment provides a multi-source data-driven remote sensing monitoring technology for TVOC emissions from the petrochemical industry. It integrates multi-source data, including but not limited to ground-based mobile survey data, XCH4 data from TROPOMI, CAMS data from ECMWF, temperature data from MODIS, and atmospheric water vapor content data, as sample data. The atmospheric methane profile function is calculated using CAMS stratified methane concentration data to estimate the ratio of surface methane concentration to the total methane column concentration from the surface to the top of the atmosphere. The XCH4 data observed by TROPOMI is converted into corresponding surface methane concentration data as the main variable factor, and combined with temperature, water vapor, and other data to construct a training set. A Gaussian process regression model is used to establish a surface TVOC emission estimation model to estimate regional TVOC emissions, thereby facilitating the monitoring of TVOC emissions from the petrochemical industry.

[0032] See Figure 1 An embodiment of the present invention provides a remote sensing monitoring method for TVOC emissions in the petrochemical industry driven by multi-source data, comprising the following steps: Step 1: Starting with the overall TVOC concentration measured by ground-based mobile surveys, and combining the anthropogenic and natural sources of atmospheric TVOC, as well as the influencing factors of TVOC, an analysis was conducted to obtain relevant influencing factor data, including temperature, wind speed, water vapor, topographic factors, and atmospheric methane concentration factors as relevant variable factors. Specifically, the following data were selected: Level 2 (OFFL) satellite methane column concentration data from Tropomi, digital elevation model (DEM, 90m) data from USGS, surface temperature and atmospheric water vapor content data from MODIS, methane stratification concentration data from ECMWF atmospheric chemistry model simulation (CAMS), and horizontal and vertical wind speed data from ERA. The ground-based mobile survey data, obtained by moving the instrument while it was fixed to a vehicle, has a spatial resolution of up to 0.0001°. The remaining collected data have a maximum spatial resolution of approximately 90m. Therefore, all data were resampled to 0.001° × 0.001° (approximately 100m) and resampled to the daily level.

[0033] Step 2: Based on CAMS data, obtain methane concentration data and corresponding altitude data for different altitude layers from the Earth's surface to the top of the atmosphere, and estimate the vertical profile of atmospheric methane at the fitted point, such as... Figure 2 The TROPOMI full-coverage methane column mean dry air mixture ratio (MAM) concentration data is an integral value from the Earth's surface to the top of the atmosphere. By using the atmospheric methane vertical profile, the ratio of surface methane concentration to whole column methane concentration can be calculated. Based on this, the TROPOMI Methane Column Mean Dry Air Mixture Ratio (MAM) concentration data is converted into surface methane concentration data, serving as an important indicator of VOC emissions from the petrochemical industry.

[0034] Step 3: Obtain the modeling dataset by matching the ground-based mobile survey TVOC data with the data of each variable factor. Model the model using the Gaussian process regression algorithm, adjust the parameters in the model, and validate the model results using a 10-fold difference validation. Specifically, the obtained variable factor data is matched with the ground-based mobile survey TVOC data to obtain the modeling dataset. The model selected in this invention is the Gaussian process regression model, a nonparametric statistical model based on Bayesian inference, commonly used for nonlinear modeling. The Gaussian process regression model has advantages such as strong generalization ability, simple model training, hyperparameter adaptiveness, interpretability, and robustness, and is particularly suitable for solving regression problems with nonlinear, high-dimensional, and small sample data, making it suitable for the limited data volume in this invention.

[0035] (1)

[0036] In the formula, f(x) follows a Gaussian distribution. , This is the covariance function used to define a Gaussian process, where h(x) represents the transformation function that converts sample features x into target features h(x), which is also the basis function matrix. Therefore, the response function of Gaussian process regression is described as: (2)

[0037] In the formula, y i It is x i The corresponding objective function, is the error variance, and β is the coefficient, both calculated from the sample data; According to the above formula, the main considerations in Gaussian process regression models are the setting of the kernel function and the basis functions. By choosing different kernels, Gaussian process regression models can characterize different statistical features. Common kernels include the quadratic exponential kernel, the rational quadratic kernel, the Matern 3 / 2 kernel, the Matern 5 / 2 kernel, and the exponential kernel.

[0038] The squared exponential kernel is represented as follows: (3)

[0039] In the formula, It is a feature length scale. It is the standard deviation of the signal.

[0040] A rational quadratic kernel is represented as follows: (4)

[0041] In the formula, α is a positive proportional mixing parameter.

[0042] The Matern 3 / 2 core is represented as follows: (5)

[0043] In the formula, r represents the Euclidean distance between samples. (6)

[0044] The Matern 5 / 2 core is represented as follows: (7)

[0045] The Exponential Kernel is represented as follows: (8)

[0046] Furthermore, the definition of basis functions also needs to be considered. Basis functions are typically expressed as... , where β is the coefficient of the corresponding sample, and h is the basis matrix, which is usually of several types, such as constant, linear, and pureQuadratic, or empty; empty means that the basis function is defined as an empty matrix; constant represents the basis matrix:

[0047] In the formula, H is a vector representing the size of the sample data; linear represents the basis matrix:

[0048] In the formula, X is virtual sample expansion data; pureQuadratic represents the basis matrix:

[0049] In the formula It is the covariance matrix of X.

[0050] Through model tuning, in this example, the basis functions are defined as constants, and a rational quadratic kernel is chosen as the kernel function.

[0051] The model was validated using 10-fold cross-validation to avoid potential overfitting. The training set was randomly split into 10 equal parts, with 9 parts used for training and the remaining part used for testing. The measured results were compared with the original data results. This process was repeated ten times until every part of the data was tested, ultimately yielding the measured values ​​for all data.

[0052] Meanwhile, statistical indicators such as correlation coefficient (R), root mean square error (RMSE), and mean absolute error (MAE) were used to evaluate the model accuracy. The formulas for R, RMSE, and MAE are shown below: (12)

[0053] in, This represents the TVOC emissions calculated by the model. This represents the TVOC emission data measured by ground-based mobile monitoring. This represents the average value of TVOC emissions, and n represents the number of measured values.

[0054] (13)

[0055] in, This represents the TVOC emissions calculated by the model. This represents the TVOC emission data measured by ground-based mobile monitoring.

[0056] (14)

[0057] in, This represents the TVOC emissions calculated by the model. This represents the TVOC emission data measured by ground-based mobile monitoring, where n represents the number of measured values.

[0058] The accuracy of the model was verified using the above accuracy metrics, and the results are as follows: Figure 3 As shown. The direct fitting results obtained are: R=0.84, RMSE=117.4 (ug / m). 3 MAE = 78.2 (ug / m) 3 The validation results show that, during the 10-fold cross-validation, the TVOC emission data fitted by the model is highly consistent with the TVOC emission data measured by ground-based mobile data collection, indicating that the model is running well and can be used to calculate TVOC emissions.

[0059] Step 4: Perform standard spatial gridding on all variable factor data to obtain the corresponding raster data. Then, using the established Gaussian process regression model, calculate the TVOC emissions of each raster within the grid to obtain the regional full-coverage TVOC emission data, such as... Figure 4 Using the methods described above, a high spatial resolution TVOC emission dataset covering the entire region can be obtained, providing support for TVOC emission monitoring in the petrochemical industry.

[0060] Comparative example: To highlight the advantages of the method of this invention, a comparative example is provided. This comparative example mainly uses the traditional emission inventory method and the physical model simulation method. The specific steps and results of this method are as follows: Step 1: Determine the scope of the study. In this comparative study, natural source volatile organic compounds (TVOC) are used as the research object. TVOC emission species are divided into three categories: isoprene, monoterpenes and other TVOCs. TVOC species mainly include alcohols, aldehydes, ketones, organic acids, low-carbon alkanes and alkenes.

[0061] Step 2: Collect data on the categories of emission sources and their corresponding activity and emission factor data. TVOC is mainly emitted from natural sources such as vegetation. Therefore, emission factor data for different vegetation types will be collected based on measured emission factors from domestic and international studies and related research. Simultaneously, data on leaf biomass, temperature, and solar radiation that influence the emission process will be collected as activity data.

[0062] Step 3: Obtain VOC emission data by calculating using the emission factor method or simulating using the MEGAN physical model. The emission factor method is calculated as follows: (15)

[0063] In the formula, T represents the activity data. denoted as emission factor, and i represents different types of emission sources.

[0064] In calculating TVOC emissions, the MEGAN physical model simulated emissions from all naturally occurring sources in all ecosystems. The main input data included leaf area index (LAIv), vegetation function type (PFT), emission factor (EF), and meteorological data. EF used a globally universal emission factor with a spatial resolution of 0.05°, encompassing emission levels and net primary emissions for 15 vegetation function types. It considered the dry deposition of the plant canopy and differentiated between different species compositions and species-specific regions. Finally, a weighted average was calculated based on the area of ​​all ecological regions. LAIv used the MODIS leaf area index standard product MOD15A2H with a spatial resolution of 500m and a temporal resolution of 8 days. PFT used the MODIS land use standard product MCD12Q1 with a spatial resolution of 500m and employed IGBP classification. Based on the Land Use Product Guidelines and the ERA-Interim climatological reanalysis dataset, the IGBP classification was mapped to the PFT classification of the Community Land Model (CLM) to obtain the PFT data used as model input. The obtained TVOC emission results are as follows: Figure 5 As shown, its spatiotemporal resolution is relatively low. This can be seen by comparing the results of the embodiments of the present invention (…). Figure 4 ) and comparative results ( Figure 5 It can be seen that the method of the present invention can obtain a high daily spatiotemporal resolution, greatly improve the spatial coverage of VOC emission data, and meet the needs of regional fine-grained monitoring.

[0065] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A remote sensing monitoring method for TVOC emissions based on multi-source data, characterized in that, Includes the following steps: Step S1: Based on the total TVOC concentration measured on the ground, analyze the relevant environmental factors affecting TVOC emissions in conjunction with various environmental factors; Step S2: Acquire and process methane column average dry air mixture ratio concentration data and methane stratified concentration data, estimate and fit atmospheric methane vertical profile, calculate the ratio of surface methane concentration to whole column methane concentration through atmospheric methane vertical profile, and convert methane column average dry air mixture ratio concentration data into surface methane concentration data. Step S3: By acquiring and matching the total TVOC concentration data measured on the ground, the surface methane concentration data, and the relevant environmental factor data affecting TVOC emissions, a TVOC emission model is established using the Gaussian process regression algorithm, and 10-fold cross-validation is performed. Step S4: Based on the TVOC emission model established in Step S3, the TVOC emission of each grid within the monitoring range is calculated to obtain TVOC emission data covering the entire monitoring area.

2. The monitoring method according to claim 1, characterized in that, The relevant environmental factors affecting TVOC emissions analyzed in step S1 include temperature factor, wind speed factor, water vapor factor, topography factor, and atmospheric methane concentration factor.

3. The monitoring method according to claim 1, characterized in that, The TVOC data measured on the ground was obtained by conducting mobile measurements with the measuring instrument fixed on a mobile vehicle.

4. The monitoring method according to claim 2, characterized in that, The methane column mean dry air mixing ratio concentration data are from the secondary offline data product of satellite methane column concentration from TROPOMI; the methane stratification concentration data are from the methane stratification concentration data CAMS simulated by the ECMWF atmospheric chemistry model.

5. The monitoring method according to claim 2, characterized in that, The temperature factor is derived from MODIS surface temperature data; The wind speed factor is derived from ERA's horizontal and vertical wind speed data; The water vapor factor is derived from atmospheric water vapor content data from MODIS; The terrain factors are derived from the USGS Digital Elevation Model (DEM) data. The atmospheric methane stratification concentration data are from CAMS's EGG4 stratified methane concentration data; The atmospheric methane vertical column concentration factor is derived from TROPOMI's XCH4 data product.

6. The monitoring method according to claim 1, characterized in that, The TROPOMI methane column average dry air mixing ratio concentration data is an integral average value from the surface to the top of the atmosphere. The ratio of surface methane concentration to whole column methane concentration is calculated by using the atmospheric methane vertical profile, and the TROPOMI methane column average dry air mixing ratio concentration data is converted into surface methane concentration data.

7. The monitoring method according to claim 1, characterized in that, The atmospheric methane vertical profile was obtained by fitting methane concentration data from the Earth's surface to the top of the atmosphere.

8. The monitoring method according to claim 1, characterized in that, Step S3 specifically involves: Step S301: Obtain the modeling dataset by matching the TVOC data measured by ground mobile survey with the surface methane concentration data and related environmental factor data; Step S302: Model TVOC emissions using a Gaussian process regression model. The response function of the Gaussian process regression model is defined according to the following formula. In the formula y i It is x i The corresponding objective function, σ 2 is the error variance, and β is the coefficient, both calculated from the sample data.

9. The monitoring method according to claim 1, characterized in that, The raster in step S4 is the corresponding raster data obtained by spatial rasterization processing of temperature, wind speed, water vapor, topographic factor, atmospheric methane concentration factor, TROPOMI methane column concentration data and CAMS methane stratification concentration data.

Images