Methane column concentration remote sensing inversion method and system of high spatial resolution satellite load

Abstract

The application provides a methane column concentration remote sensing inversion method and system of a high spatial resolution satellite load, comprising: calculating unit absorption characteristic spectrum of atmospheric gas components in a short-wave infrared band, and screening the best channel for atmospheric methane increment inversion; calculating spectral radiance data for inversion according to short-wave infrared reflectivity data and the best channel, constructing a spectral similar pixel set of a pixel to be inverted, and calculating a reference background spectrum and an error covariance matrix of the pixel to be inverted; calculating a methane increment initial value of the pixel to be inverted, adjusting the spectral radiance value of the pixel in the spectral similar pixel set, updating the reference background spectrum and the error covariance matrix, and performing optimization iteration on a near-surface methane increment result; inverting a background atmospheric methane concentration vertical profile; adjusting the obtained background atmospheric methane concentration vertical profile, calculating a methane dry air mixing ratio concentration, performing geometric correction on the obtained result, and generating a standard grid product of atmospheric methane column concentration.

Description

Technical Field

[0001] This invention relates to the field of greenhouse gas remote sensing monitoring, specifically to a method and system for remote sensing inversion of methane column concentration using a high spatial resolution satellite payload; more specifically, to a method and system for remote sensing inversion of atmospheric methane column concentration using a high spatial resolution satellite payload. Background Technology

[0002] Methane, as a potent greenhouse gas, has significant impacts on regional atmospheric ecology and global climate change. Therefore, quantitatively monitoring the spatial distribution trend of atmospheric methane concentration is crucial for regional environmental governance and improving energy efficiency. Satellite remote sensing technology provides an effective means to quickly and accurately monitor atmospheric methane levels. Furthermore, high spatial resolution satellite payloads have the potential to provide even more refined information on the spatial distribution of atmospheric methane. However, due to the influence of other gases in the atmosphere and the complex surface, remote sensing inversion of atmospheric methane column concentration still requires consideration of various factors, making atmospheric methane column concentration inversion for such satellite payloads a challenging problem.

[0003] When using satellite remote sensing platforms with high spatial and spectral resolution to retrieve atmospheric methane column concentration, neither traditional all-physical inversion algorithms nor machine learning inversion algorithms based on spectral features can achieve accurate and efficient retrieval. Specifically, for all-physical inversion algorithms, the satellite remote sensing platforms often have meter-level resolution detection capabilities, and the time cost of pixel-by-pixel inversion using all-physical algorithms is enormous, making them unsuitable for the task of retrieving atmospheric methane column concentration. For machine learning inversion algorithms based on spectral features, the lack of sufficient and accurate prior inversion results as training targets for machine learning prevents the correct construction of the machine learning model, thus also making them unsuitable for the task of retrieving atmospheric methane column concentration. To address these issues, this invention proposes a high-resolution remote sensing inversion method for atmospheric methane column concentration that combines all-physical inversion and incremental methane inversion techniques.

[0004] Patent document CN113624694A (application number: 202111178987.2) provides a method and apparatus for inverting atmospheric methane concentration, relating to the technical field of environmental monitoring. The method includes: acquiring sample input data and sample calibration data of the area to be monitored; using the PCA algorithm to perform dimensionality reduction processing on the sample input data to obtain dimensionality-reduced sample input data; using the dimensionality-reduced sample input data and sample calibration data to train a preset XGBoost model to obtain a target XGBoost model; after acquiring the current input data of the area to be monitored, using the PCA algorithm to perform dimensionality reduction processing on the current input data to obtain dimensionality-reduced current input data; inputting the dimensionality-reduced current input data into the target XGBoost model to obtain the current atmospheric methane column concentration of the area to be monitored. This method solves the technical problems of large inversion errors and low inversion efficiency in existing atmospheric methane concentration inversion methods. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for remote sensing inversion of methane column concentration using high spatial resolution satellite payloads.

[0006] A remote sensing inversion method for methane column concentration from a high spatial resolution satellite payload, provided by the present invention, is characterized by comprising:

[0007] Step S1: Calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band and screen the optimal channel for atmospheric methane increment inversion.

[0008] Step S2: Based on the shortwave infrared band reflectivity data and the optimal channel, calculate the spectral radiance data used for inversion, construct the spectral similarity pixel set of the pixel to be inverted, and calculate the reference background spectrum and error covariance matrix of the pixel to be inverted.

[0009] Step S3: Calculate the initial value of methane increment for the pixel to be inverted, adjust the spectral radiance value of the pixel in the spectral similarity pixel set, update the reference background spectrum and error covariance matrix, and perform optimization iteration of the near-surface methane increment results.

[0010] Step S4: Based on the reference background spectrum, invert the vertical profile of the background atmospheric methane concentration;

[0011] Step S5: Based on the atmospheric methane increment inversion results, adjust the vertical profile of the background atmospheric methane concentration obtained from the inversion, calculate the methane dry air mixing ratio concentration, perform geometric correction on the obtained results, and generate a standard grid product of atmospheric methane column concentration.

[0012] Preferably, in step S1:

[0013] A hyperspectral radiance simulation experiment was conducted using a radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band. Based on the combined bubble sorting method, the optimal channel for atmospheric methane increment inversion was screened.

[0014] Using an atmospheric radiative transfer model with shortwave infrared hyperspectral radiance simulation characteristics, the simulation conditions required for the radiative transfer mode were set. The concentration levels of the prior vertical profiles of atmospheric gas components were varied to simulate and calculate the radiance values ​​observed by the satellite payload under different concentration backgrounds. Based on the simulation results, the unit absorption characteristic spectra of atmospheric gases in the shortwave infrared band were calculated using the least squares fitting method. Using the gas unit absorption characteristic spectra, and based on the combined bubble sorting method, the bands with strong methane absorption characteristics and weak absorption characteristics of other gases were selected as the optimal channels for incremental atmospheric methane retrieval.

[0015] The proposed parameters for radiative transfer model simulation conditions include: spectral characteristics of the satellite payload, satellite observation conditions, atmospheric parameters, and surface reflectivity;

[0016] The spectral characteristics of the satellite payload include: center wavelength, full peak width at half maximum (FWHM), and spectral response function.

[0017] Satellite observation conditions include: solar zenith angle, satellite zenith angle, and relative azimuth angle;

[0018] Atmospheric parameters include: aerosol parameterization scheme, aerosol optical thickness, and vertical profiles of the concentrations of absorbing atmospheric gas components;

[0019] The atmospheric gaseous components include: water vapor, methane, nitrogen oxides, nitrous oxide, nitrogen dioxide, and carbon dioxide;

[0020] The combined bubble sort method involves arranging the absorption characteristics of atmospheric gas components in each band according to the multivariate bubble sort method.

[0021] Preferably, in step S2:

[0022] Based on satellite-observed shortwave infrared reflectance data and the aforementioned optimal channel, spectral radiance data for inversion is calculated. A set of spectrally similar pixels to the pixel to be inverted is constructed based on this data. The reference background spectrum and error covariance matrix of the pixel to be inverted are then calculated.

[0023] Step S2.1: Based on the satellite observation shortwave infrared band reflectance data and the nearest channel for methane inversion, calculate the band spectral radiance used for inversion, and smooth the spectral radiance value using a sliding spectral window;

[0024] Step S2.2: Construct a set of spectral similar pixels to the pixel to be inverted. The set of spectral similar pixels consists of pixels whose geographic spatial distance from the pixel to be inverted is less than a preset standard, whose correlation coefficient is higher than a preset standard, and whose spectral angle and overall deviation are less than a preset standard.

[0025] Step S2.3: Perform principal component decomposition on the shortwave infrared spectral radiance of the pixels in the spectral similarity pixel set, and generate the reference background spectrum of the pixel to be inverted based on the cumulative variance interpretation rate of the principal components.

[0026] Step S2.4: Calculate the average shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and use the average shortwave infrared spectral radiance μ to calculate the error covariance matrix C of the pixel to be inverted;

[0027] The formula for calculating the error covariance matrix C is as follows:

[0028]

[0029] Where N is the number of pixels in the spectrally similar pixel set, and L is the shortwave infrared spectral radiance of the pixels in the spectrally similar pixel set.

[0030] Preferably, in step S3:

[0031] The initial value of methane increment for the pixel to be inverted is calculated based on the matched filtering method. The spectral radiance value of the pixel in the spectral similarity pixel set is adjusted based on the methane increment result. The reference background spectrum and error covariance matrix are updated again. The near-surface methane increment result is optimized iteratively again using the matched filtering method.

[0032] The matched filtering method uses the following formula to calculate the methane increment:

[0033]

[0034] Where α is the initial value of methane increment, L′ is the difference between the observed radiance and the reference background spectrum, C is the error covariance matrix, and t′ is the target characteristic spectrum for stretching and updating the unit absorption characteristic spectrum of methane using the reference background spectrum.

[0035] The spectral radiance values ​​of each pixel in the spectral similarity pixel set are adjusted using:

[0036] L′ i =L i -α i ·μ·t

[0037] Among them, L′ i L is the adjusted spectral radiance value of the i-th pixel. i Let α be the observed radiance of the i-th pixel. iLet μ be the initial value of the methane increment for the i-th pixel, μ be the reference background spectrum, and t be the target feature spectrum.

[0038] The optimization iteration of the methane increment results is as follows: when the update amount of the inverted methane increment results is less than the preset value, it is determined that the optimization iteration has converged, and the optimization iteration is stopped.

[0039] Preferably, in step S4:

[0040] Based on the calculation of the global reference background spectrum, the vertical profile of background atmospheric methane concentration is retrieved using an optimal estimation algorithm:

[0041] The global reference background spectrum is calculated by averaging the reference background spectrum of each pixel.

[0042] The global reference background spectrum is approximated using the following formula:

[0043] y = F(x, b) + ε y +ε F

[0044] Where y represents the initial reference spectrum, F represents the forward model (radiative transfer model), the state vector x represents the atmospheric state of the spectrum to be inverted, which can be specifically quantified using variables, including methane profiles, temperature profiles, humidity profiles, pressure profiles, aerosol parameters, and surface pressure, the state vector b represents atmospheric state constants, the atmospheric state that does not need to be inverted, and ε y Represents the instrument observation error, ε F Represents the forward model error;

[0045] The optimal estimation algorithm performs the inversion using the following steps:

[0046] Step S4.1: Define the objective function x of the algorithm. 2 for:

[0047]

[0048] Where, x a Let y represent the prior state variable. a S represents the spectral radiance value used in radiative transfer simulations using prior state variables. a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y.

[0049] Step S4.2: Solve the objective function using the Levenberg-Marquardt iterative method. The iterative form is as follows:

[0050]

[0051] Where K is the Jacobian matrix, calculated by the radiative transfer model, i is the iteration number identifier, i+1 is the next generation of i, and x i With x i+1 To progressively update the iterative state variables, S a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y, and γ is the Levenberg-Marquardt parameter, which is adjusted according to Rodgers' update strategy after each iteration of x.

[0052] Step S4.3: Iteratively update the state vector x until the update amount of the state vector x is lower than the preset value and convergence is achieved. The state vector x records the atmospheric state information inverted from the reference background spectrum.

[0053] Preferably, in step S5:

[0054] Based on the atmospheric methane increment inversion results, the vertical profile of the background atmospheric methane concentration obtained from the inversion is adjusted. The methane-dry-air mixing ratio concentration is calculated according to surface pressure and water vapor content. The results are then geometrically corrected to generate a standard grid product of atmospheric methane column concentration.

[0055] The methane dry air mixture concentration XCH4 was calculated using:

[0056]

[0057] in The total amount of methane in the column is obtained by integrating the methane profile. and These represent the total dry air column and the total water vapor column, respectively, P s Let g be the surface air pressure, and g be the average gravitational acceleration of the cylinder. and These are the molecular masses of water vapor and dry air, respectively.

[0058] Geometric correction is performed by using the geographic location information of each pixel in the original observation data to perform geometric correction on the methane dry air mixing ratio concentration results.

[0059] A high spatial resolution satellite payload methane column concentration remote sensing inversion system according to the present invention includes:

[0060] Module M1: Calculates the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band and screens the optimal channel for atmospheric methane increment inversion;

[0061] Module M2: Based on the shortwave infrared band reflectivity data and the optimal channel, calculate the spectral radiance data used for inversion, construct a set of spectrally similar pixels of the pixel to be inverted, and calculate the reference background spectrum and error covariance matrix of the pixel to be inverted.

[0062] Module M3: Calculates the initial value of methane increment for the pixel to be inverted, adjusts the spectral radiance value of the pixel in the spectral similarity pixel set, updates the reference background spectrum and error covariance matrix, and performs optimization iteration of the near-surface methane increment results.

[0063] Module M4: Retrieves the vertical profile of background atmospheric methane concentration based on the reference background spectrum;

[0064] Module M5: Based on the atmospheric methane increment inversion results, adjust the vertical profile of the background atmospheric methane concentration obtained from the inversion, calculate the methane dry air mixing ratio concentration, perform geometric correction on the obtained results, and generate a standard grid product of atmospheric methane column concentration.

[0065] Preferably, in module M1:

[0066] A hyperspectral radiance simulation experiment was conducted using a radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band. Based on the combined bubble sorting method, the optimal channel for atmospheric methane increment inversion was screened.

[0067] Using an atmospheric radiative transfer model with shortwave infrared hyperspectral radiance simulation characteristics, the simulation conditions required for the radiative transfer mode were set. The concentration levels of the prior vertical profiles of atmospheric gas components were varied to simulate and calculate the radiance values ​​observed by the satellite payload under different concentration backgrounds. Based on the simulation results, the unit absorption characteristic spectra of atmospheric gases in the shortwave infrared band were calculated using the least squares fitting method. Using the gas unit absorption characteristic spectra, and based on the combined bubble sorting method, the bands with strong methane absorption characteristics and weak absorption characteristics of other gases were selected as the optimal channels for incremental atmospheric methane retrieval.

[0068] The proposed parameters for radiative transfer model simulation conditions include: spectral characteristics of the satellite payload, satellite observation conditions, atmospheric parameters, and surface reflectivity;

[0069] The spectral characteristics of the satellite payload include: center wavelength, full peak width at half maximum (FWHM), and spectral response function.

[0070] Satellite observation conditions include: solar zenith angle, satellite zenith angle, and relative azimuth angle;

[0071] Atmospheric parameters include: aerosol parameterization scheme, aerosol optical thickness, and vertical profiles of the concentrations of absorbing atmospheric gas components;

[0072] The atmospheric gaseous components include: water vapor, methane, nitrogen oxides, nitrous oxide, nitrogen dioxide, and carbon dioxide;

[0073] The combined bubble sort method involves arranging the absorption characteristics of atmospheric gas components in each band according to the multivariate bubble sort method.

[0074] In module M2:

[0075] Based on satellite-observed shortwave infrared reflectance data and the aforementioned optimal channel, spectral radiance data for inversion is calculated. A set of spectrally similar pixels to the pixel to be inverted is constructed based on this data. The reference background spectrum and error covariance matrix of the pixel to be inverted are then calculated.

[0076] Module M2.1: Based on satellite observation of shortwave infrared band reflectance data and the nearest channel for methane inversion, calculate the band spectral radiance used for inversion, and smooth the spectral radiance values ​​using a sliding spectral window;

[0077] Module M2.2: Constructs a set of spectral similar pixels to the pixel to be inverted. The set of spectral similar pixels consists of pixels whose geographic spatial distance from the pixel to be inverted is less than a preset standard, whose correlation coefficient is higher than a preset standard, and whose spectral angle and overall deviation are less than a preset standard.

[0078] Module M2.3: Performs principal component decomposition on the shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and generates the reference background spectrum of the pixel to be inverted based on the cumulative variance interpretation rate of the principal components.

[0079] Module M2.4: Calculate the average shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and use the average shortwave infrared spectral radiance μ to calculate the error covariance matrix C of the pixel to be inverted;

[0080] The formula for calculating the error covariance matrix C is as follows:

[0081]

[0082] Where N is the number of pixels in the spectrally similar pixel set, and L is the shortwave infrared spectral radiance of the pixels in the spectrally similar pixel set.

[0083] Preferably, in module M3:

[0084] The initial value of methane increment for the pixel to be inverted is calculated based on the matched filtering method. The spectral radiance value of the pixel in the spectral similarity pixel set is adjusted based on the methane increment result. The reference background spectrum and error covariance matrix are updated again. The near-surface methane increment result is optimized iteratively again using the matched filtering method.

[0085] The matched filtering method uses the following formula to calculate the methane increment:

[0086]

[0087] Where α is the initial value of methane increment, L′ is the difference between the observed radiance and the reference background spectrum, C is the error covariance matrix, and t′ is the target characteristic spectrum for stretching and updating the unit absorption characteristic spectrum of methane using the reference background spectrum.

[0088] The spectral radiance values ​​of each pixel in the spectral similarity pixel set are adjusted using:

[0089] L′ i =L i -α i ·μ·t

[0090] Among them, L′ i L is the adjusted spectral radiance value of the i-th pixel. i Let α be the observed radiance of the i-th pixel. i Let μ be the initial value of the methane increment for the i-th pixel, μ be the reference background spectrum, and t be the target feature spectrum.

[0091] The optimization iteration of the methane increment results is as follows: when the update amount of the inverted methane increment results is less than the preset value, it is determined that the optimization iteration has converged, and the optimization iteration is stopped.

[0092] In module M4:

[0093] Based on the calculation of the global reference background spectrum, the vertical profile of background atmospheric methane concentration is retrieved using an optimal estimation algorithm:

[0094] The global reference background spectrum is calculated by averaging the reference background spectrum of each pixel.

[0095] The global reference background spectrum is approximated using the following formula:

[0096] y = F(x, b) + ε y +ε F

[0097] Where y represents the initial reference spectrum, F represents the forward model (radiative transfer model), the state vector x represents the atmospheric state of the spectrum to be inverted, which can be specifically quantified using variables, including methane profiles, temperature profiles, humidity profiles, pressure profiles, aerosol parameters, and surface pressure, the state vector b represents atmospheric state constants, the atmospheric state that does not need to be inverted, and ε y Represents the instrument observation error, ε F Represents the forward model error;

[0098] The optimal estimation algorithm performs the inversion using the following steps:

[0099] Module M4.1: Defines the objective function χ of the algorithm. 2 for:

[0100]

[0101] Where, x a Let y represent the prior state variable. a S represents the spectral radiance value used in radiative transfer simulations using prior state variables. a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y.

[0102] Module M4.2: The Levenberg-Marquardt iterative method is used to solve the objective function. The iterative form is as follows:

[0103]

[0104] Where K is the Jacobian matrix, calculated by the radiative transfer model, i is the iteration number identifier, i+1 is the next generation of i, and x i With x i+1 To progressively update the iterative state variables, S a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y, and γ is the Levenberg-Marquardt parameter, which is adjusted according to Rodgers' update strategy after each iteration of x.

[0105] Module M4.3: Iteratively updates the state vector x until the update amount of the state vector x is lower than the preset value and convergence is reached. The state vector x records the atmospheric state information inverted from the reference background spectrum.

[0106] Preferably, in module M5:

[0107] Based on the atmospheric methane increment inversion results, the vertical profile of the background atmospheric methane concentration obtained from the inversion is adjusted. The methane-dry-air mixing ratio concentration is calculated according to surface pressure and water vapor content. The results are then geometrically corrected to generate a standard grid product of atmospheric methane column concentration.

[0108] The methane dry air mixture concentration XCH4 was calculated using:

[0109]

[0110] in The total amount of methane in the column is obtained by integrating the methane profile. and These represent the total dry air column and the total water vapor column, respectively, P s Let g be the surface air pressure, and g be the average gravitational acceleration of the cylinder. and These are the molecular masses of water vapor and dry air, respectively.

[0111] Geometric correction is performed by using the geographic location information of each pixel in the original observation data to perform geometric correction on the methane dry air mixing ratio concentration results.

[0112] Compared with the prior art, the present invention has the following beneficial effects:

[0113] This invention proposes a remote sensing technique for calculating atmospheric methane column concentration based on atmospheric methane increment inversion results, which solves the problem that the original algorithm cannot be applied to the task of inverting atmospheric methane column concentration using satellite remote sensing observation platforms with high spatial resolution and high spectral resolution. Attached Figure Description

[0114] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0115] Figure 1 This is a flowchart of a remote sensing inversion method for atmospheric methane column concentration used in high spatial resolution satellite payloads;

[0116] Figure 2 This is a schematic diagram of the atmospheric methane column concentration inversion results from a high spatial resolution satellite payload. Detailed Implementation

[0117] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0118] Example 1:

[0119] This invention provides a remote sensing inversion method for atmospheric methane column concentration using high spatial resolution satellite payloads. The method includes: calculating the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band using a radiative transfer model; selecting the optimal channel for atmospheric methane increment inversion; constructing a set of spectrally similar pixels for the pixels to be inverted; calculating the reference background spectrum and error covariance matrix of the pixels to be inverted; performing optimized iterative inversion of atmospheric methane increments using matched filtering; inverting the vertical profile of atmospheric methane concentration in the global reference background spectrum using an optimal estimation method; calculating the atmospheric methane dry air mixing ratio concentration grid-by-grid based on the atmospheric methane increment inversion results and the atmospheric methane concentration vertical profile; and performing geometric correction on the methane dry air mixing ratio concentration results to generate a standard grid product of atmospheric methane column concentration. This invention can be used for remote sensing inversion of atmospheric methane column concentration with high spatial resolution, and in particular, it solves the key technical problem that all-physics and machine learning inversion algorithms are not applicable to atmospheric methane column concentration inversion tasks relying on high spatial resolution and high spectral resolution satellite remote sensing observation platforms.

[0120] According to the present invention, a remote sensing inversion method for methane column concentration from a high spatial resolution satellite payload is provided, characterized in that, as Figures 1-2 As shown, it includes:

[0121] Step S1: Calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band and screen the optimal channel for atmospheric methane increment inversion.

[0122] Specifically, in step S1:

[0123] A hyperspectral radiance simulation experiment was conducted using a radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band. Based on the combined bubble sorting method, the optimal channel for atmospheric methane increment inversion was screened.

[0124] Using an atmospheric radiative transfer model with shortwave infrared hyperspectral radiance simulation characteristics, the simulation conditions required for the radiative transfer mode were set. The concentration levels of the prior vertical profiles of atmospheric gas components were varied to simulate and calculate the radiance values ​​observed by the satellite payload under different concentration backgrounds. Based on the simulation results, the unit absorption characteristic spectra of atmospheric gases in the shortwave infrared band were calculated using the least squares fitting method. Using the gas unit absorption characteristic spectra, and based on the combined bubble sorting method, the bands with strong methane absorption characteristics and weak absorption characteristics of other gases were selected as the optimal channels for incremental atmospheric methane retrieval.

[0125] The proposed parameters for radiative transfer model simulation conditions include: spectral characteristics of the satellite payload, satellite observation conditions, atmospheric parameters, and surface reflectivity;

[0126] The spectral characteristics of the satellite payload include: center wavelength, full peak width at half maximum (FWHM), and spectral response function.

[0127] Satellite observation conditions include: solar zenith angle, satellite zenith angle, and relative azimuth angle;

[0128] Atmospheric parameters include: aerosol parameterization scheme, aerosol optical thickness, and vertical profiles of the concentrations of absorbing atmospheric gas components;

[0129] The atmospheric gaseous components include: water vapor, methane, nitrogen oxides, nitrous oxide, nitrogen dioxide, and carbon dioxide;

[0130] The combined bubble sort method involves arranging the absorption characteristics of atmospheric gas components in each band according to the multivariate bubble sort method.

[0131] Step S2: Based on the shortwave infrared band reflectivity data and the optimal channel, calculate the spectral radiance data used for inversion, construct the spectral similarity pixel set of the pixel to be inverted, and calculate the reference background spectrum and error covariance matrix of the pixel to be inverted.

[0132] Specifically, in step S2:

[0133] Based on satellite-observed shortwave infrared reflectance data and the aforementioned optimal channel, spectral radiance data for inversion is calculated. A set of spectrally similar pixels to the pixel to be inverted is constructed based on this data. The reference background spectrum and error covariance matrix of the pixel to be inverted are then calculated.

[0134] Step S2.1: Based on the satellite observation shortwave infrared band reflectance data and the nearest channel for methane inversion, calculate the band spectral radiance used for inversion, and smooth the spectral radiance value using a sliding spectral window;

[0135] Step S2.2: Construct a set of spectral similar pixels to the pixel to be inverted. The set of spectral similar pixels consists of pixels whose geographic spatial distance from the pixel to be inverted is less than a preset standard, whose correlation coefficient is higher than a preset standard, and whose spectral angle and overall deviation are less than a preset standard.

[0136] Step S2.3: Perform principal component decomposition on the shortwave infrared spectral radiance of the pixels in the spectral similarity pixel set, and generate the reference background spectrum of the pixel to be inverted based on the cumulative variance interpretation rate of the principal components.

[0137] Step S2.4: Calculate the average shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and use the average shortwave infrared spectral radiance μ to calculate the error covariance matrix C of the pixel to be inverted;

[0138] The formula for calculating the error covariance matrix C is as follows:

[0139]

[0140] Where N is the number of pixels in the spectrally similar pixel set, and L is the shortwave infrared spectral radiance of the pixels in the spectrally similar pixel set.

[0141] Step S3: Calculate the initial value of methane increment for the pixel to be inverted, adjust the spectral radiance value of the pixel in the spectral similarity pixel set, update the reference background spectrum and error covariance matrix, and perform optimization iteration of the near-surface methane increment results.

[0142] Specifically, in step S3:

[0143] The initial value of methane increment for the pixel to be inverted is calculated based on the matched filtering method. The spectral radiance value of the pixel in the spectral similarity pixel set is adjusted based on the methane increment result. The reference background spectrum and error covariance matrix are updated again. The near-surface methane increment result is optimized iteratively again using the matched filtering method.

[0144] The matched filtering method uses the following formula to calculate the methane increment:

[0145]

[0146] Where α is the initial value of methane increment, L ′ To observe the difference between radiance and the reference background spectrum, C is the error covariance matrix, t ′ The target characteristic spectrum is used to stretch and update the unit absorption characteristic spectrum of methane using a reference background spectrum;

[0147] The spectral radiance values ​​of each pixel in the spectral similarity pixel set are adjusted using:

[0148] L i =L i -α i ·μ·t

[0149] Among them, L′ i L is the adjusted spectral radiance value of the i-th pixel. i Let α be the observed radiance of the i-th pixel. i Let μ be the initial value of the methane increment for the i-th pixel, μ be the reference background spectrum, and t be the target feature spectrum.

[0150] The optimization iteration of the methane increment results is as follows: when the update amount of the inverted methane increment results is less than the preset value, it is determined that the optimization iteration has converged, and the optimization iteration is stopped.

[0151] Step S4: Based on the reference background spectrum, invert the vertical profile of the background atmospheric methane concentration;

[0152] Specifically, in step S4:

[0153] Based on the calculation of the global reference background spectrum, the vertical profile of background atmospheric methane concentration is retrieved using an optimal estimation algorithm:

[0154] The global reference background spectrum is calculated by averaging the reference background spectrum of each pixel.

[0155] The global reference background spectrum is approximated using the following formula:

[0156] y = F(x, b) + ε y +ε F

[0157] Where y represents the initial reference spectrum, F represents the forward model (radiative transfer model), the state vector x represents the atmospheric state of the spectrum to be inverted, which can be specifically quantified using variables, including methane profiles, temperature profiles, humidity profiles, pressure profiles, aerosol parameters, and surface pressure, the state vector b represents atmospheric state constants, the atmospheric state that does not need to be inverted, and ε y Represents the instrument observation error, ε F Represents the forward model error;

[0158] The optimal estimation algorithm performs the inversion using the following steps:

[0159] Step S4.1: Define the objective function χ of the algorithm. 2 for:

[0160]

[0161] Where, x a Let y represent the prior state variable. a S represents the spectral radiance value used in radiative transfer simulations using prior state variables. a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y.

[0162] Step S4.2: Solve the objective function using the Levenberg-Marquardt iterative method. The iterative form is as follows:

[0163]

[0164] Where K is the Jacobian matrix, calculated by the radiative transfer model, i is the iteration number identifier, i+1 is the next generation of i, and x i With x i+1 To progressively update the iterative state variables, S a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, Se The observation error covariance matrix is ​​calculated from the observed spectral radiance y, and γ is the Levenberg-Marquardt parameter, which is adjusted according to Rodgers' update strategy after each iteration of x.

[0165] Step S4.3: Iteratively update the state vector x until the update amount of the state vector x is lower than the preset value and convergence is achieved. The state vector x records the atmospheric state information inverted from the reference background spectrum.

[0166] Step S5: Based on the atmospheric methane increment inversion results, adjust the vertical profile of the background atmospheric methane concentration obtained from the inversion, calculate the methane dry air mixing ratio concentration, perform geometric correction on the obtained results, and generate a standard grid product of atmospheric methane column concentration.

[0167] Specifically, in step S5:

[0168] Based on the atmospheric methane increment inversion results, the vertical profile of the background atmospheric methane concentration obtained from the inversion is adjusted. The methane-dry-air mixing ratio concentration is calculated according to surface pressure and water vapor content. The results are then geometrically corrected to generate a standard grid product of atmospheric methane column concentration.

[0169] The methane dry air mixture concentration XCH4 was calculated using:

[0170]

[0171] in The total amount of methane in the column is obtained by integrating the methane profile. and These represent the total dry air column and the total water vapor column, respectively, P s Let g be the surface air pressure, and g be the average gravitational acceleration of the cylinder. and These are the molecular masses of water vapor and dry air, respectively.

[0172] Geometric correction is performed by using the geographic location information of each pixel in the original observation data to perform geometric correction on the methane dry air mixing ratio concentration results.

[0173] Example 2:

[0174] Example 2 is a preferred embodiment of Example 1, and is used to illustrate the present invention in more detail.

[0175] The present invention also provides a methane column concentration remote sensing inversion system for high spatial resolution satellite payloads. The methane column concentration remote sensing inversion system for high spatial resolution satellite payloads can be implemented by executing the process steps of the methane column concentration remote sensing inversion method for high spatial resolution satellite payloads. That is, those skilled in the art can understand the methane column concentration remote sensing inversion method for high spatial resolution satellite payloads as a preferred embodiment of the methane column concentration remote sensing inversion system for high spatial resolution satellite payloads.

[0176] A high spatial resolution satellite payload methane column concentration remote sensing inversion system according to the present invention includes:

[0177] Module M1: Calculates the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band and screens the optimal channel for atmospheric methane increment inversion;

[0178] Specifically, in module M1:

[0179] A hyperspectral radiance simulation experiment was conducted using a radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band. Based on the combined bubble sorting method, the optimal channel for atmospheric methane increment inversion was screened.

[0180] Using an atmospheric radiative transfer model with shortwave infrared hyperspectral radiance simulation characteristics, the simulation conditions required for the radiative transfer mode were set. The concentration levels of the prior vertical profiles of atmospheric gas components were varied to simulate and calculate the radiance values ​​observed by the satellite payload under different concentration backgrounds. Based on the simulation results, the unit absorption characteristic spectra of atmospheric gases in the shortwave infrared band were calculated using the least squares fitting method. Using the gas unit absorption characteristic spectra, and based on the combined bubble sorting method, the bands with strong methane absorption characteristics and weak absorption characteristics of other gases were selected as the optimal channels for incremental atmospheric methane retrieval.

[0181] The proposed parameters for radiative transfer model simulation conditions include: spectral characteristics of the satellite payload, satellite observation conditions, atmospheric parameters, and surface reflectivity;

[0182] The spectral characteristics of the satellite payload include: center wavelength, full peak width at half maximum (FWHM), and spectral response function.

[0183] Satellite observation conditions include: solar zenith angle, satellite zenith angle, and relative azimuth angle;

[0184] Atmospheric parameters include: aerosol parameterization scheme, aerosol optical thickness, and vertical profiles of the concentrations of absorbing atmospheric gas components;

[0185] The atmospheric gaseous components include: water vapor, methane, nitrogen oxides, nitrous oxide, nitrogen dioxide, and carbon dioxide;

[0186] The combined bubble sort method involves arranging the absorption characteristics of atmospheric gas components in each band according to the multivariate bubble sort method.

[0187] Module M2: Based on the shortwave infrared band reflectivity data and the optimal channel, calculate the spectral radiance data used for inversion, construct a set of spectrally similar pixels of the pixel to be inverted, and calculate the reference background spectrum and error covariance matrix of the pixel to be inverted.

[0188] In module M2:

[0189] Based on satellite-observed shortwave infrared reflectance data and the aforementioned optimal channel, spectral radiance data for inversion is calculated. A set of spectrally similar pixels to the pixel to be inverted is constructed based on this data. The reference background spectrum and error covariance matrix of the pixel to be inverted are then calculated.

[0190] Module M2.1: Based on satellite observation of shortwave infrared band reflectance data and the nearest channel for methane inversion, calculate the band spectral radiance used for inversion, and smooth the spectral radiance values ​​using a sliding spectral window;

[0191] Module M2.2: Constructs a set of spectral similar pixels to the pixel to be inverted. The set of spectral similar pixels consists of pixels whose geographic spatial distance from the pixel to be inverted is less than a preset standard, whose correlation coefficient is higher than a preset standard, and whose spectral angle and overall deviation are less than a preset standard.

[0192] Module M2.3: Performs principal component decomposition on the shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and generates the reference background spectrum of the pixel to be inverted based on the cumulative variance interpretation rate of the principal components.

[0193] Module M2.4: Calculate the average shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and use the average shortwave infrared spectral radiance μ to calculate the error covariance matrix C of the pixel to be inverted;

[0194] The formula for calculating the error covariance matrix C is as follows:

[0195]

[0196] Where N is the number of pixels in the spectrally similar pixel set, and L is the shortwave infrared spectral radiance of the pixels in the spectrally similar pixel set.

[0197] Module M3: Calculates the initial value of methane increment for the pixel to be inverted, adjusts the spectral radiance value of the pixel in the spectral similarity pixel set, updates the reference background spectrum and error covariance matrix, and performs optimization iteration of the near-surface methane increment results.

[0198] Specifically, in module M3:

[0199] The initial value of methane increment for the pixel to be inverted is calculated based on the matched filtering method. The spectral radiance value of the pixel in the spectral similarity pixel set is adjusted based on the methane increment result. The reference background spectrum and error covariance matrix are updated again. The near-surface methane increment result is optimized iteratively again using the matched filtering method.

[0200] The matched filtering method uses the following formula to calculate the methane increment:

[0201]

[0202] Where α is the initial value of methane increment, L′ is the difference between the observed radiance and the reference background spectrum, C is the error covariance matrix, and t′ is the target characteristic spectrum for stretching and updating the unit absorption characteristic spectrum of methane using the reference background spectrum.

[0203] The spectral radiance values ​​of each pixel in the spectral similarity pixel set are adjusted using:

[0204] L′ i =L i -α i ·μ·t

[0205] Among them, L′ i L is the adjusted spectral radiance value of the i-th pixel. i Let α be the observed radiance of the i-th pixel. i Let μ be the initial value of the methane increment for the i-th pixel, μ be the reference background spectrum, and t be the target feature spectrum.

[0206] The optimization iteration of the methane increment results is as follows: when the update amount of the inverted methane increment results is less than the preset value, it is determined that the optimization iteration has converged, and the optimization iteration is stopped.

[0207] Module M4: Retrieves the vertical profile of background atmospheric methane concentration based on the reference background spectrum;

[0208] In module M4:

[0209] Based on the calculation of the global reference background spectrum, the vertical profile of background atmospheric methane concentration is retrieved using an optimal estimation algorithm:

[0210] The global reference background spectrum is calculated by averaging the reference background spectrum of each pixel.

[0211] The global reference background spectrum is approximated using the following formula:

[0212] y = F(x, b) + ε y +ε F

[0213] Where y represents the initial reference spectrum, F represents the forward model (radiative transfer model), the state vector x represents the atmospheric state of the spectrum to be inverted, which can be specifically quantified using variables, including methane profiles, temperature profiles, humidity profiles, pressure profiles, aerosol parameters, and surface pressure, the state vector b represents atmospheric state constants, the atmospheric state that does not need to be inverted, and ε y Represents the instrument observation error, ε F Represents the forward model error;

[0214] The optimal estimation algorithm performs the inversion using the following steps:

[0215] Module M4.1: Defines the objective function χ of the algorithm. 2 for:

[0216]

[0217] Where, x a Let y represent the prior state variable. a S represents the spectral radiance value used in radiative transfer simulations using prior state variables. a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y.

[0218] Module M4.2: The Levenberg-Marquardt iterative method is used to solve the objective function. The iterative form is as follows:

[0219]

[0220] Where K is the Jacobian matrix, calculated by the radiative transfer model, i is the iteration number identifier, i+1 is the next generation of i, and x i With x i+1 To progressively update the iterative state variables, S a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y, and γ is the Levenberg-Marquardt parameter, which is adjusted according to Rodgers' update strategy after each iteration of x.

[0221] Module M4.3: Iteratively updates the state vector x until the update amount of the state vector x is lower than the preset value and convergence is reached. The state vector x records the atmospheric state information inverted from the reference background spectrum.

[0222] Module M5: Based on the atmospheric methane increment inversion results, adjust the vertical profile of the background atmospheric methane concentration obtained from the inversion, calculate the methane dry air mixing ratio concentration, perform geometric correction on the obtained results, and generate a standard grid product of atmospheric methane column concentration.

[0223] Specifically, in module M5:

[0224] Based on the atmospheric methane increment inversion results, the vertical profile of the background atmospheric methane concentration obtained from the inversion is adjusted. The methane-dry-air mixing ratio concentration is calculated according to surface pressure and water vapor content. The results are then geometrically corrected to generate a standard grid product of atmospheric methane column concentration.

[0225] The methane dry air mixture concentration XCH4 was calculated using:

[0226]

[0227] in The total amount of methane in the column is obtained by integrating the methane profile. and These represent the total dry air column and the total water vapor column, respectively, P s Let g be the surface air pressure, and g be the average gravitational acceleration of the cylinder. and These are the molecular masses of water vapor and dry air, respectively.

[0228] Geometric correction is performed by using the geographic location information of each pixel in the original observation data to perform geometric correction on the methane dry air mixing ratio concentration results.

[0229] Example 3:

[0230] Example 3 is a preferred example of Example 1, and is used to illustrate the present invention in more detail.

[0231] The purpose of this invention is to provide a remote sensing inversion method for atmospheric methane column concentration for high spatial resolution satellite payloads, in order to solve the problems mentioned in the background art.

[0232] To achieve the above objectives, the present invention provides the following technical solution: a remote sensing inversion method for atmospheric methane column concentration for high spatial resolution satellite payloads, comprising the following steps:

[0233] Step S1: Use the radiative transfer model to conduct a hyperspectral radiance simulation experiment, calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band, and screen the optimal channel for atmospheric methane increment inversion based on the combined bubble sorting method.

[0234] Step S2: Based on the satellite-observed shortwave infrared band reflectivity data and the optimal channel, calculate the spectral radiance data for inversion, construct a set of spectral similar pixels of the pixel to be inverted based on the data, and calculate the reference background spectrum and error covariance matrix of the pixel to be inverted.

[0235] Step S3: Calculate the initial value of methane increment for the pixel to be inverted based on the matched filtering method. Adjust the spectral radiance value of the pixels in the spectral similarity pixel set based on the methane increment result. Update the reference background spectrum and error covariance matrix again. Optimize and iterate the near-surface methane increment result again using the matched filtering method.

[0236] Step S4: Calculate the global reference background spectrum based on the method in step S2, and use the optimal estimation algorithm to invert the vertical profile of the background atmospheric methane concentration.

[0237] Step S5: Based on the atmospheric methane increment inversion results of Step S3, adjust the background atmospheric methane concentration vertical profile obtained from Step S4, calculate the methane dry air mixing ratio concentration according to surface pressure and water vapor content, perform geometric correction on the obtained results, and generate a standard grid product of atmospheric methane column concentration.

[0238] Preferably, step S1 involves: using an atmospheric radiative transfer model with shortwave infrared hyperspectral radiance simulation characteristics to set the simulation conditions required for the radiative transfer mode; changing the concentration level of the prior vertical profile of atmospheric gas components to simulate and calculate the radiance values ​​observed by the satellite payload under different concentration backgrounds; calculating the unit absorption characteristic spectrum of atmospheric gases in the shortwave infrared band using the least squares fitting method based on the simulation results; and using the aforementioned gas unit absorption characteristic spectrum, selecting the band with strong methane absorption characteristics and weak absorption characteristics of other gases using the combined bubble sorting method, which is the optimal channel for atmospheric methane incremental inversion.

[0239] The proposed parameters for the radiative transfer mode simulation conditions include: spectral characteristics of the satellite payload (center wavelength, full width at half maximum, spectral response function, etc.), satellite observation conditions (solar zenith angle, satellite zenith angle, relative azimuth angle), atmospheric parameters (aerosol parameterization scheme, aerosol optical thickness, vertical profile of the concentration of absorbing atmospheric gas components), and surface reflectivity.

[0240] The atmospheric gas components include: water vapor, methane, nitrogen oxides, nitrous oxide, nitrogen dioxide, carbon dioxide, etc.

[0241] The combined bubble sorting method involves arranging the absorption characteristics of atmospheric gas components in each band according to a multivariate bubble sorting method.

[0242] Preferably, step S2 employs:

[0243] Step S2.1: Based on the satellite observation shortwave infrared band reflectance data and the nearest channel for methane inversion, calculate the band spectral radiance used for inversion, and smooth the spectral radiance value using a sliding spectral window.

[0244] Step S2.2: Construct a set of spectral similar pixels to the pixel to be inverted. The set of spectral similar pixels consists of pixels that are close to the pixel to be inverted in geospatial distance, have a high correlation coefficient, and have small spectral angle and overall deviation.

[0245] Step S2.3: Perform principal component decomposition on the shortwave infrared spectral radiance of the pixels in the spectral similarity pixel set, and generate the reference background spectrum of the pixel to be inverted based on the cumulative variance interpretation rate of the principal components.

[0246] Step S2.4: Calculate the average shortwave infrared spectral radiance of pixels in the spectral similarity pixel set, and use the average shortwave infrared spectral radiance μ to calculate the error covariance matrix C of the pixel to be inverted.

[0247] The formula for calculating the error covariance matrix C is as follows:

[0248]

[0249] Where N is the number of pixels in the spectrally similar pixel set, and L is the shortwave infrared spectral radiance of the pixels in the spectrally similar pixel set.

[0250] Preferably, step S3 employs the following methods:

[0251] The matched filtering method uses the following formula to calculate the methane increment:

[0252]

[0253] Where α is the initial value of methane increment, L ′ To observe the difference between radiance and the reference background spectrum, C is the error covariance matrix, t ′ The target characteristic spectrum is used to stretch and update the unit absorption characteristic spectrum of methane using a reference background spectrum.

[0254] The spectral radiance values ​​of each pixel in the adjusted spectral similarity pixel set are adopted as follows:

[0255] L′ i =L i -α i ·μ·t

[0256] Among them, L′ i L is the adjusted spectral radiance value of the i-th pixel. i Let α be the observed radiance of the i-th pixel. iLet μ be the initial value of the methane increment for the i-th pixel, μ be the reference background spectrum, and t be the target feature spectrum.

[0257] The optimization iteration of the methane increment results is as follows: when the update amount of the inverted methane increment results is less than a preset value, it is determined that the optimization iteration has converged, and the optimization iteration is stopped.

[0258] Preferably, step S4 employs the following methods:

[0259] The global reference background spectrum is calculated by averaging the reference background spectra of each pixel.

[0260] The global reference background spectrum is approximated using the following formula:

[0261] y = F(x, b) + ε y +ε F

[0262] Where y represents the initial reference spectrum, F represents the forward model (i.e., the radiative transfer model), the state vector x represents the atmospheric state of the spectrum to be inverted, which can be specifically quantified by variables such as methane profile, temperature profile, humidity profile, pressure profile, aerosol parameters, and surface pressure, the state vector b represents the atmospheric state constant, i.e., the atmospheric state that does not need to be inverted, and its accurate value can be obtained from other data sources, ε y Represents the instrument observation error, ε F This represents the forward model error.

[0263] The optimal estimation algorithm performs the inversion using the following steps:

[0264] Step S4.1: Define the objective function χ of the algorithm. 2 for:

[0265]

[0266] Where x a Let y represent the prior state variable. a S represents the spectral radiance value used in radiative transfer simulations using prior state variables. a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y.

[0267] Step S4.2: Solve the objective function using the Levenberg-Marquardt iterative method. The iterative form is as follows:

[0268]

[0269] Where K is the Jacobian matrix, obtained from simulations using the radiative transfer model, and x i With x i+1 To progressively update the iterative state variables, S a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e γ is the observation error covariance matrix, calculated from the observed spectral radiance y, and γ is the Levenberg-Marquardt parameter, which is adjusted according to Rodgers' update strategy after each iteration of x.

[0270] Step S4.3: Iteratively update the state vector x according to the above formula until the update amount of the state vector x is lower than the preset value and convergence is reached. The final state vector x records the atmospheric state information inverted by the reference background spectrum.

[0271] Preferably, step S5 employs the following:

[0272] The vertical profile of background atmospheric methane concentration obtained from the incremental atmospheric methane inversion is adjusted pixel by pixel. The methane dry air mixing ratio concentration is calculated based on surface pressure and water vapor content. After geometric correction, a standard grid product of atmospheric methane column concentration is generated.

[0273] The atmospheric methane increment inversion result adopts the atmospheric methane increment result obtained from the iterative inversion in step 3.

[0274] The background atmospheric methane concentration vertical profile is obtained from the methane profile information in the atmospheric state variables in step 4.

[0275] The calculation of the methane dry air mixing ratio concentration uses:

[0276]

[0277] in The total amount of methane in the column is obtained by integrating the methane profile. and These represent the total dry air column and the total water vapor column, respectively, P s Let g be the surface air pressure, and g be the average gravitational acceleration of the cylinder. and These are the molecular masses of water vapor and dry air, respectively.

[0278] The geometric correction method involves using the geographic location information of each pixel in the original observation data to perform geometric correction on the methane dry air mixing ratio concentration result.

[0279] Example 4:

[0280] Example 4 is a preferred example of Example 1, which is used to illustrate the present invention in more detail.

[0281] Step 1: Conduct shortwave infrared hyperspectral radiance simulation experiments using the radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band, and screen the optimal channel for atmospheric methane increment inversion based on this.

[0282] A) Download and install atmospheric radiative transfer models with shortwave infrared hyperspectral radiance simulation capabilities, including but not limited to SCIATRAN and LBLTRAN models;

[0283] B) Setting up simulation conditions for the radiative transfer model requires specifying parameters including: spectral characteristics of the satellite payload (center wavelength, full width at half maximum, spectral response function, etc.), satellite observation conditions (solar zenith angle, satellite zenith angle, relative azimuth angle), atmospheric parameters (aerosol parameterization scheme, aerosol optical thickness, vertical profile of the concentration of absorbing atmospheric gas components), and surface reflectivity, etc.

[0284] The concentration levels of the a priori vertical profile of atmospheric gas components are gradually changed, and the radiance values ​​observed by the satellite payload under different concentration backgrounds are simulated and calculated. The atmospheric gas components include: water vapor, methane, nitrogen oxides, nitrous oxide, nitrogen dioxide, carbon dioxide, etc.

[0285] C) Based on the above simulation results, the absorption rate of each gas component in the shortwave infrared is calculated using the least squares fitting method, that is, the unit absorption characteristic spectrum of atmospheric gas components in the shortwave infrared band is calculated.

[0286] D) Using the gas unit absorption characteristic spectrum mentioned above, and according to the combined bubble sorting method, the bands with strong methane absorption characteristics and weak absorption characteristics of other gases are selected, which are the best channels for atmospheric methane increment inversion. The number of bands is marked as b.

[0287] Step 2: Construct the reference background spectrum of the pixel to be inverted, and calculate the error covariance matrix of the spectral similar pixel set.

[0288] A) Obtain the L1 level reflectivity data of the hyperspectral high spatial resolution satellite payload, perform radiometric calibration on the data of each band according to the provided calibration coefficients, and smooth the spectral radiance values ​​after calibration using a sliding spectral window.

[0289] B) Calculate the spectral angle, correlation coefficient, overall deviation and spatial distance between the spectrum of the pixel to be inverted and the spectra of other pixels in the same scene image, and select the top 100 pixels that are close to the pixel to be inverted in space, have high correlation coefficient and small spectral angle to form a set of spectrally similar pixels;

[0290] C) Based on the spatial sparsity assumption of the anomalous increase in methane gas, principal component decomposition is performed on the spectral matrix within the spectral similarity pixel set. The top n principal components with a cumulative variance explanation rate exceeding 90% are selected for matrix reconstruction to generate the reference background spectrum μ of the pixel to be inverted. 1×b ;

[0291] D) Calculate the average short-wave infrared spectral radiance of pixels in the spectral similarity pixel set, and use the average short-wave infrared spectral radiance μ to calculate the error covariance matrix C of the pixel to be inverted. b×b ;

[0292]

[0293] Where N is the number of pixels in the spectrally similar pixel set, and L is the shortwave infrared spectral radiance of the pixels in the spectrally similar pixel set.

[0294] Step 3: Use the matched filtering method to perform iterative inversion of atmospheric methane increment.

[0295] A) For any pixel x to be inverted, calculate its observed radiance L. 1×b Compared with the reference background spectrum μ 1×b The difference between:

[0296] L′ 1×b =L 1×b -μ 1×b

[0297] B) Utilizing the reference background spectrum μ 1×b The characteristic absorption spectrum of methane 1×b Stretch the image, calculate the dot product of the two values, and update the target feature spectrum t′. 1×b :

[0298] t′ 1×b =S 1×b ⊙μ 1×b

[0299] C) The initial value of methane increment is retrieved pixel-by-pixel using a matched filtering method;

[0300]

[0301] D) Adjust the spectral radiance values ​​of each pixel in the spectral similarity pixel set according to the initial methane increment value α;

[0302] L′ i =L i -α i ·μ·t

[0303] E) Repeat step 2, based on the adjusted spectral radiance values ​​L′ of each pixel. i Update the reference background spectrum μ1×b And error covariance matrix C b×b ;

[0304] F) Repeat the ae sub-step in step 3 to perform iterative optimization of methane increment inversion until the iteration converges, thus obtaining the pixel-by-pixel atmospheric methane increment inversion result.

[0305] Step 4: Calculate the global background reference spectrum based on the results of Step 2, and use the full physical inversion algorithm to invert the vertical profile of the background atmospheric methane concentration.

[0306] A) Calculate the average spectrum of the reference background spectrum for each pixel as the global background reference spectrum;

[0307] B) For the aforementioned global background reference spectrum μ 1×b It can be approximated by the following formula:

[0308] y = F(x, b) + ε y +ε F

[0309] Where y represents the initial reference spectrum, F represents the forward model (i.e., the radiative transfer model), the state vector x represents the atmospheric state of the spectrum to be inverted, which can be specifically quantified by variables such as methane profile, temperature profile, humidity profile, pressure profile, aerosol parameters, and surface pressure, the state vector b represents the atmospheric state constant, i.e., the atmospheric state that does not need to be inverted, and its accurate value can be obtained from other data sources, ε y Represents the instrument observation error, ε F Represents the forward model error;

[0310] C) Define the objective function χ of the all-physics inversion algorithm. 2 for:

[0311]

[0312] Where x a Let y represent the prior state variable. a S represents the spectral radiance value used in radiative transfer simulations using prior state variables. a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y.

[0313] D) The objective function is solved using the Levenberg-Marquardt iterative method, with the iterative form as follows:

[0314]

[0315] Where K is the JacObian matrix, obtained from simulation calculations using the radiative transfer model, x i With x i+1 To progressively update the iterative state variables, S a The prior covariance matrix is ​​formed by the prior state variable x. a Calculated, S e The observation error covariance matrix is ​​calculated from the observed spectral radiance y, and γ is the Levenberg-Marquardt parameter, initially set to 10, and adjusted according to Rodgers' update strategy after each iteration of x.

[0316] E) Iteratively update the state vector x according to the above formula until the update amount of the state vector x is less than 0.01% and convergence is reached. The final state vector x records the atmospheric state information inverted by the reference background spectrum.

[0317] Step 5: Based on the atmospheric methane increment inversion results described in Step 3 and the atmospheric methane concentration vertical profile described in Step 4, calculate the grid-by-grid atmospheric methane column concentration.

[0318] A) Extract the methane profile information from the atmospheric state variables described in step 4;

[0319] B) Based on the atmospheric methane increment inversion results described in step 3, adjust the obtained methane profile;

[0320] C) The methane-dry-air mixing ratio concentration is calculated based on surface pressure and water vapor content:

[0321]

[0322] in The total amount of methane in the column is obtained by integrating the methane profile. and These represent the total dry air column and the total water vapor column, respectively, P s Let g be the surface air pressure, and g be the average gravitational acceleration of the cylinder. and These are the molecular masses of water vapor and dry air, respectively.

[0323] D) Using the geographic location information of each pixel in the original observation data, the methane dry air mixing ratio concentration result is geometrically corrected to generate a standard grid product of atmospheric methane column concentration.

[0324] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0325] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A methane column concentration remote sensing retrieval method for high spatial resolution satellite payloads, characterized in that, The method comprises the following steps: Step S1: calculating the unit absorption characteristic spectrum of atmospheric gas components in the short-wave infrared band, and screening the best channel for atmospheric methane increment inversion; Step S2: calculating the spectral radiance data for inversion according to the short-wave infrared reflectivity data and the best channel, constructing a spectral similar pixel set of the pixel to be inverted, calculating the reference background spectrum and error covariance matrix of the pixel to be inverted; Step S3: calculating the initial value of the methane increment of the pixel to be inverted, adjusting the spectral radiance value of the pixel in the spectral similar pixel set, updating the reference background spectrum and error covariance matrix, and performing optimization iteration on the near-surface methane increment result; Step S4: inverting the background atmospheric methane concentration vertical profile based on the reference background spectrum; Step S5: adjusting the background atmospheric methane concentration vertical profile obtained by inversion based on the atmospheric methane increment inversion result, calculating the methane dry air mixing ratio concentration, and performing geometric correction on the obtained result to generate a standard grid product of atmospheric methane column concentration; The spectral similar pixel set is composed of pixels with a geographical spatial distance less than a preset standard, a correlation coefficient higher than a preset standard, and a spectral angle and overall deviation less than a preset standard; The initial value of the methane increment of the pixel to be inverted is calculated based on the matched filter method; The background atmospheric methane concentration vertical profile is inverted based on the optimal estimation algorithm.

2. The method for remote sensing inversion of methane column concentration of high spatial resolution satellite load according to claim 1, characterized in that, In the step S1: A hyperspectral radiance simulation experiment is performed by using a radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the short-wave infrared band, and the best channel for atmospheric methane increment inversion is screened based on the combined bubble sorting method; An atmospheric radiative transfer model with short-wave infrared hyperspectral radiance simulation characteristics is used to set the simulation conditions required by the radiative transfer model; The concentration level of the prior vertical profile of the atmospheric gas component is changed to simulate and calculate the radiance value observed by the satellite payload under different concentration backgrounds; according to the simulation results, the least square fitting method is used to calculate the unit absorption characteristic spectrum of the atmospheric gas in the short-wave infrared band; The waveband with strong methane absorption characteristic and weak absorption characteristic of other gases is screened out as the best channel for atmospheric methane increment inversion based on the combined bubble sorting method using the unit absorption characteristic spectrum of the gas; The parameters of the simulation conditions of the radiative transfer model include: spectral characteristic information of the satellite payload, satellite observation conditions, atmospheric parameter conditions, and surface reflectivity; The spectral characteristic information of the satellite payload includes: center wavelength, full peak half-width, and spectral response function; The satellite observation conditions include: solar zenith angle, satellite zenith angle, and relative azimuth angle; The atmospheric parameter conditions include: aerosol parameterization scheme, aerosol optical thickness, and concentration vertical profile of the absorbing atmospheric gas component; The atmospheric gas components include: water vapor, methane, nitrogen oxide, nitrous oxide, nitrogen dioxide, and carbon dioxide; The combined bubble sorting method is used to arrange the absorption characteristics of the atmospheric gas components in each waveband according to the multivariate bubble sorting method.

3. The method for remote sensing inversion of methane column concentration of high spatial resolution satellite load according to claim 1, characterized in that, In the step S2: According to the satellite observation short-wave infrared band reflectivity data and the best channel, the spectral radiance data used for inversion is calculated, the spectral similar pixel set of the pixel to be inverted is constructed based on the data, the reference background spectrum of the pixel to be inverted is calculated, and the error covariance matrix is calculated: Step S2.1: According to the satellite observation short-wave infrared band reflectivity data and the methane inversion best channel, the band spectral radiance used for inversion is calculated, and the spectral radiance value is smoothed by using a sliding spectral window; Step S2.2: Constructing a spectral similar pixel set of the pixel to be inverted, the spectral similar pixel set is composed of pixels with a geographical spatial distance less than a preset standard, a correlation coefficient higher than a preset standard, and a spectral angle and overall deviation less than a preset standard; Step S2.3: The short-wave infrared spectral radiance of the pixel in the spectral similar pixel set is decomposed by principal component, and the reference background spectrum of the pixel to be inverted is generated according to the cumulative variance explanation rate of the principal component; Step S2.4: calculating the average of the short-wave infrared spectral radiance of the pixels in the spectral similar pixel set, using the average of the short-wave infrared spectral radiance calculating the error covariance matrix of the pixel to be inverted ; The error covariance matrix Calculation formula: wherein, is the number of pixels in the spectrally similar pixel cluster, is the short-wave infrared spectral radiance of the pixels in the spectrally similar pixel cluster.

4. The method for remote sensing inversion of methane column concentration of high spatial resolution satellite load according to claim 1, characterized in that, In the step S3: Based on the matched filter method, the methane increment initial value of the pixel to be inverted is calculated, the spectral radiance value of the pixel in the spectral similar pixel set is adjusted based on the methane increment result, the reference background spectrum and the error covariance matrix are updated again, and the matched filter method is used again to optimize and iterate the near-surface methane increment result; The matched filter method calculates the methane increment by using the following formula: wherein, is a methane increment initial value, is a difference between the observed radiance and the reference background spectrum, is an error covariance matrix, is a target feature spectrum updated by stretching the methane unit absorption feature spectrum using the reference background spectrum; The adjustment of the spectral radiance value of each pixel in the spectral similar pixel set uses: wherein, is the adjusted spectral radiance value for the i-th pixel, is the observed radiance for the i-th pixel, is the initial value of the methane increment for the i-th pixel, is the reference background spectrum, is the target feature spectrum; The optimization and iteration of the methane increment result uses: when the update amount of the inverted methane increment result is less than a preset value, it is judged that the optimization and iteration converges, that is, the optimization and iteration is stopped.

5. The method for remote sensing inversion of methane column concentration of high spatial resolution satellite load according to claim 1, characterized in that, In the step S4: Based on the calculation of the global reference background spectrum, the background atmospheric methane concentration vertical profile is inverted by using the optimal estimation algorithm: The global reference background spectrum uses: the average spectrum of the reference background spectrum of each pixel is calculated; The global reference background spectrum is approximately expressed by using the following formula: wherein, denotes an initialization reference spectrum, denotes a forward model, a radiative transfer model, a state vector denotes an atmospheric state in which the spectrum to be inverted is located, which can be quantified using variables including a methane profile, a temperature profile, a humidity profile, a pressure profile, an aerosol parameter and a surface air pressure, a state vector denotes atmospheric state constants, atmospheric states that do not need to be inverted, denotes an instrument observation error, denotes a forward model error; The optimal estimation algorithm uses the following steps for inversion: Step S4.1: Defining the objective function of the algorithm is: wherein represents a prior state variable, represents a spectral radiance value simulated using the prior state variable for radiative transfer, is a prior covariance matrix, calculated from the prior state variable is an observation error covariance matrix, calculated from the observed spectral radiance ;​ Step S4.2: The Levenberg-Marquardt iteration method is used to solve the objective function, and the iteration form is: where, is the Jacobian matrix, computed from the radiative transfer model, i is the iteration number identifier, i+1 is the next generation of i, is the state variable of the iteration, is the step-wise updated state variable of the iteration, is the prior covariance matrix, computed from the prior state variable , is the observation error covariance matrix, computed from the observed spectral radiance , is the Levenberg-Marquardt parameter, adjusted after each iteration of according to the update strategy of Rodgers. Step S4.3: Iteratively update the state vector until the state vector converges when the update quantity is below a preset value, the state vector Records the atmospheric state information retrieved from the reference background spectrum.

6. The method for remote sensing inversion of methane column concentration of high spatial resolution satellite load according to claim 1, characterized in that, In the step S5: Based on the atmospheric methane increment inversion result, the background atmospheric methane concentration vertical profile obtained by inversion is adjusted, the methane dry air mixing ratio concentration is calculated according to the surface air pressure and water vapor content, the obtained result is geometrically corrected, and the standard grid product of the atmospheric methane column concentration is generated: Calculating the methane dry air mixture ratio concentration Adopt: where is the total amount of methane column, which is obtained by integrating the methane profile, and are the total amounts of dry air column and water vapor column, respectively, is the surface air pressure, is the average gravitational acceleration of the column, and are the molecular masses of water vapor and dry air, respectively; The geometric correction uses: the geographical position information of each pixel in the original observation data is used to perform geometric correction on the methane dry air mixing ratio concentration result.

7. A methane column density remote sensing retrieval system for a high spatial resolution satellite payload, characterized in that, It includes: Module M1: Calculate the unit absorption characteristic spectrum of atmospheric gas components in the short-wave infrared band, and select the best channel for atmospheric methane increment inversion; Module M2: According to the short-wave infrared band reflectivity data and the best channel, the spectral radiance data used for inversion is calculated, the spectral similar pixel set of the pixel to be inverted is constructed, the reference background spectrum of the pixel to be inverted is calculated, and the error covariance matrix is calculated; Module M3: calculating the methane increment initial value of the pixel to be inverted, adjusting the spectral radiance value of the pixel in the spectral similar pixel set, updating the reference background spectrum and error covariance matrix, and performing optimization iteration of the near-surface methane increment result; Module M4: based on the reference background spectrum, inverting the background atmospheric methane concentration vertical profile; Module M5: based on the atmospheric methane increment inversion result, adjusting the background atmospheric methane concentration vertical profile obtained by inversion, calculating the methane dry air mixing ratio concentration, and performing geometric correction on the obtained result to generate the standard grid product of atmospheric methane column concentration; The spectral similar pixel set is composed of pixels with a geographical spatial distance less than a preset standard, a correlation coefficient higher than a preset standard, and a spectral angle and overall deviation less than a preset standard from the pixel to be inverted; The methane increment initial value of the pixel to be inverted is calculated based on the matched filter method; The background atmospheric methane concentration vertical profile is inverted based on the optimal estimation algorithm.

8. The methane column concentration remote sensing inversion system of a high spatial resolution satellite load according to claim 7, characterized in that: In the module M1: A high spectral radiance simulation experiment is performed using a radiative transfer model to calculate the unit absorption characteristic spectrum of atmospheric gas components in the short-wave infrared band, and the best channel for atmospheric methane increment inversion is selected based on the combined bubble sorting method; An atmospheric radiative transfer model with short-wave infrared high spectral radiance simulation characteristics is used to set the simulation conditions required by the radiative transfer model; The concentration level of the prior vertical profile of atmospheric gas components is changed to simulate and calculate the radiance values observed by the satellite load under different concentration backgrounds; according to the simulation results, the least square fitting method is used to calculate the unit absorption characteristic spectrum of atmospheric gas in the short-wave infrared band; The waveband with strong methane absorption characteristics and weak absorption characteristics of other gases is selected as the best channel for atmospheric methane increment inversion based on the combined bubble sorting method using the unit absorption characteristic spectrum of the gas; The parameters of the simulation conditions of the radiative transfer model include: spectral characteristic information of the satellite load, satellite observation conditions, atmospheric parameter conditions, and surface reflectivity; The spectral characteristic information of the satellite load includes: center wavelength, full peak half-width, and spectral response function; The satellite observation conditions include: solar zenith angle, satellite zenith angle, and relative azimuth angle; The atmospheric parameter conditions include: aerosol parameterization scheme, aerosol optical thickness, and concentration vertical profile of absorbing atmospheric gas components; The atmospheric gas components include: water vapor, methane, nitrogen oxide, nitrous oxide, nitrogen dioxide, and carbon dioxide; The combined bubble sorting method is used: the absorption characteristics of each waveband atmospheric gas component are arranged according to the multivariate bubble sorting method; In the module M2: Based on the satellite observed short-wave infrared waveband reflectivity data and the best channel, the spectral radiance data for inversion is calculated, the spectral similar pixel set of the pixel to be inverted is constructed based on the data, and the reference background spectrum and error covariance matrix of the pixel to be inverted are calculated: Module M2.1: Calculate the spectral radiance of the band used for inversion according to the satellite observed short-wave infrared band reflectivity data and the methane inversion optimal channel, and smooth the spectral radiance value using a sliding spectral window; Module M2.2: Construct a spectral similar pixel set of the pixel to be inverted, which is composed of pixels with a geographical spatial distance less than a preset standard, a correlation coefficient higher than a preset standard, and a spectral angle and overall bias less than a preset standard; Module M2.3: Perform principal component decomposition on the short-wave infrared spectral radiance of the pixels in the spectral similar pixel set, and generate a reference background spectrum for the pixel to be inverted according to the cumulative variance explanation rate of the principal components; Module M2.4: calculating the average value of the short-wave infrared spectral radiance of the pixels in the spectral similar pixel set, using the average value of the short-wave infrared spectral radiance calculating the error covariance matrix of the pixel to be inverted ; The error covariance matrix Calculation formula: wherein, is the number of pixels in the spectral similar pixel set, is the short-wave infrared spectral radiance of the pixel in the spectral similar pixel set.

9. The methane column concentration remote sensing inversion system of a high spatial resolution satellite load according to claim 7, characterized in that: In the module M3: Based on the matched filter method, calculate the methane increment initial value of the pixel to be inverted, adjust the spectral radiance value of the pixels in the spectral similar pixel set based on the methane increment result, update the reference background spectrum and error covariance matrix again, and use the matched filter method again for optimization iteration of the near-surface methane increment result; The matched filter method calculates the methane increment using the following formula: wherein, is a methane increment initial value, is a difference between the observed radiance and the reference background spectrum, is an error covariance matrix, is a target feature spectrum updated by stretching the methane unit absorption feature spectrum with the reference background spectrum; Adjusting the spectral radiance value of each pixel in the spectral similar pixel set uses: wherein, is the adjusted spectral radiance value for the i-th pixel, is the observed radiance for the i-th pixel, is the initial value of the methane increment for the i-th pixel, is the reference background spectrum, is the target feature spectrum; The optimization iteration of the methane increment result uses: When the update amount of the inverted methane increment result is less than a preset value, it is judged that the optimization iteration converges, i.e. stop the optimization iteration; In the module M4: Based on the calculation of the global reference background spectrum, use the optimal estimation algorithm to invert the background atmospheric methane concentration vertical profile: The global reference background spectrum uses: Calculate the average spectrum of each pixel reference background spectrum; The global reference background spectrum is approximately expressed using the following formula: wherein, denotes an initialization reference spectrum, denotes a forward model, a radiative transfer model, a state vector denotes an atmospheric state in which the spectrum to be inverted is located, which can be quantified using variables including a methane profile, a temperature profile, a humidity profile, a pressure profile, an aerosol parameter, and a surface air pressure, a state vector denotes atmospheric state constants, atmospheric states that do not need to be inverted, denotes an instrument observation error, denotes a forward model error; The optimal estimation algorithm uses the following steps for inversion: Module M4.1 : Defining the objective function of the algorithm is: wherein represents a prior state variable, represents a spectral radiance value simulated using the prior state variable for radiative transfer, is a prior covariance matrix, derived from the prior state variable computed, is an observation error covariance matrix, derived from the observed spectral radiance computed. Module M4.2: Use the Levenberg-Marquardt iteration method to solve the objective function, and the iteration form is: in, The Jacobian matrix is ​​calculated using the radiative transfer model, where i is the iteration number and i+1 is the next generation of i. and To update the state variables iteratively, The prior covariance matrix is ​​derived from the prior state variables. Calculated, The observation error covariance matrix is ​​derived from the observed spectral radiance. Calculated, For Levenberg-Marquardt parameters, in After each iteration, adjustments are made according to Rodgers' update strategy; Module M4.3: Iteratively update the state vector until the state vector converges when the update is below a preset value, the state vector records the atmospheric state information retrieved from the reference background spectrum.

10. The methane column concentration remote sensing retrieval system of high spatial resolution satellite payload according to claim 7, characterized in that, In the module M5: Based on the atmospheric methane increment inversion result, adjust the background atmospheric methane concentration vertical profile obtained by inversion, calculate the methane dry air mixing ratio concentration according to the surface air pressure and water vapor content, perform geometric correction on the obtained result, and generate a standard grid product of atmospheric methane column concentration: Calculating the methane dry air mixture ratio concentration Adopt: where is the total amount of methane column, which is obtained by integrating the methane profile, and are the total amounts of dry air column and water vapor column, respectively, is the surface air pressure, is the average gravitational acceleration of the column, and are the molecular masses of water vapor and dry air, respectively; Geometric correction uses: Use the geographical location information of each pixel in the original observation data to perform geometric correction on the methane dry air mixing ratio concentration result.

Images