A method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observations

Abstract

The present invention discloses a method for rapidly retrieving and inverting atmospheric pollution component profiles based on hyperspectral observations, comprising: using solar scattering spectra obtained at different observation angles as a data source, obtaining the total amount of observed differential slant paths of atmospheric pollution components, and performing rapid statistical simulation and iteration in combination with a differential atmospheric mass factor lookup table constructed according to an atmospheric radiation transfer model, ultimately obtaining the vertical distribution of atmospheric pollution components, greatly improving the profile inversion speed while ensuring the rationality and reliability of the inversion results, and at the same time, providing a quality control scheme for the inversion results.

Description

Technical Field

[0001] The invention belongs to the technical field of optical measurement, and in particular relates to a method for quickly retrieving and inverting atmospheric pollution component profiles based on hyperspectral observation. Background Art

[0002] With the rapid economic development and rapid urbanization, the atmospheric environment will inevitably undergo some overall and regional changes. The deteriorating environment is gradually recognized by all countries, and the key and necessity of solving environmental problems are gradually becoming a consensus. Therefore, more complete environmental monitoring technology is needed. However, at present, the monitoring of pollutants in the atmospheric environment is mostly focused on point chemical measurement, which is complex and inefficient, and can only reflect the air conditions at the observation point, which is difficult to meet the complex and changeable environmental monitoring needs.

[0003] Optical remote sensing uses the method of collecting solar spectra to obtain the observed spectra after being absorbed by various components in the atmosphere, and the components in the atmosphere can be obtained by inversion. Compared with the traditional sampling method, it has the advantages of rapid data collection, diverse observation of pollutant components, wide measurement range, real-time measurement, and no need for personnel on duty, which has great advantages in the field of environmental inspection and monitoring. At present, the instruments used in the field of optical remote sensing pollutant monitoring mainly include laser radar, Fourier infrared spectrometer and hyperspectrometer. The laser radar transmits a detection signal (laser beam) to the target, and then compares the received signal reflected from the target (target echo) with the transmitted signal to obtain relevant information about the target. It is more effective in medium and long-distance detection, and a blind spot appears when the distance is close. The Fourier infrared spectrometer consists of an optical detection part and a computer part. Due to the absorption of energy of certain frequencies, the obtained interference pattern intensity curve will produce some changes accordingly. Through Fourier transform technology, each frequency on the interference pattern can be converted into a corresponding light intensity, and the entire infrared spectrum can be obtained. According to the different characteristics of the spectrum, the relative concentration of atmospheric pollutants can be detected. The hyperspectrometer collects solar scattering spectra, and differentiates the collected spectrum from the reference spectrum according to the different absorption of the solar spectrum by different substances. Based on the Lambert-Beer law and the different absorption characteristics of spectra of different wavelengths by different substances, the content of different components in the atmosphere can be inverted. It has the characteristics of measuring many components, a wide measurement range, and the ability to obtain the vertical distribution of pollutants.

[0004] There are several methods for inverting the spectrum collected by the hyperspectrometer, such as the optimal estimation method and the parameterization method. These processing methods all use the radiation transfer model for complex calculations and need to be run on high-performance computer equipment. They have high hardware performance requirements and high hardware costs, and there are many inversion singular values ​​and long inversion time. These problems limit the integration of instrument acquisition to data image acquisition. At present, there is no mature hyperspectrometer that integrates spectrum acquisition to pollutant distribution in the same device. Summary of the invention

[0005] In view of the above, the purpose of the present invention is to provide a method for quickly retrieving and inverting the profile of atmospheric pollution components based on hyperspectral observations. By establishing a differential atmospheric mass factor lookup table instead of a radiation transfer model for profile inversion, this method can quickly invert the vertical distribution gradient of atmospheric pollutant components.

[0006] In order to achieve the above-mentioned invention object, the present invention provides the following technical solutions:

[0007] A method for quickly retrieving and inverting atmospheric pollution component profiles based on hyperspectral observations includes the following steps:

[0008] Step 1, using the radiation transfer model to establish a differential atmospheric mass factor lookup table I for aerosol inversion and a differential atmospheric mass factor lookup table II for trace gas inversion;

[0009] Step 2, performing dimensionality reduction on the differential atmospheric quality factor lookup table I and the differential atmospheric quality factor lookup table II with respect to the station dimension and the observation month to obtain the station lookup table I and the station lookup table II;

[0010] Step 3, collecting observation data, and using the solar scattering spectrum contained in the observation data to calculate the total amount of observed differential slant range of atmospheric pollution components;

[0011] Step 4, extracting observation sequence information based on the observation data and the total amount of observed differential slant range, and performing dimensionality reduction on the site lookup table I and the site lookup table II according to the observation sequence information to obtain the sequence lookup table I and the sequence lookup table II;

[0012] Step 5, inverting aerosol profile and trace gas profile according to the observed sequence information and sequence lookup table I and sequence lookup table II.

[0013] In one embodiment, a radiation transfer model is used to simulate the Jacobian matrix of light intensity, and the influence factors of different observation elevation angles and different altitudes are calculated based on the Jacobian matrix of light intensity. The differential atmospheric mass factors of different observation elevation angles are constructed based on the influence factors to form a differential atmospheric mass factor lookup table I for aerosol inversion and a differential atmospheric mass factor lookup table II for trace gas inversion.

[0014] In one embodiment, the influence factor IF(z) at different heights z is calculated using the radiation transfer model, and then the atmospheric mass factor AMF at different observation elevation angles is calculated using the following formula based on the influence factor IF(z) and the profile p(z) under different profile parameters: α , with the atmospheric mass factor AMF at 90 degrees zenith 90° AMF αDo the difference to get the differential atmospheric mass factor dAMF α :

[0015]

[0016] In one embodiment, the differential atmospheric mass factor lookup table I for aerosol inversion includes parameters of eight dimensions: observation dimension, observation month, solar zenith angle, relative azimuth, observation elevation angle, aerosol total extinction, aerosol profile height parameter, and aerosol profile shape factor;

[0017] The differential atmospheric mass factor lookup table II for trace gas inversion contains parameters in ten dimensions: observation dimension, observation month, solar zenith angle, relative azimuth, observation elevation angle, aerosol total extinction, aerosol profile height parameter, aerosol profile shape factor, trace gas profile height parameter, and trace gas profile shape factor.

[0018] In one embodiment, the extracted observation sequence information includes: observation time, total observed differential slant range of O4 and other trace gases and the error of total observed differential slant range, solar zenith angle at that time, relative azimuth angle, and observation elevation angle.

[0019] In one embodiment, the site lookup table I and the site lookup table II are respectively reduced in dimension according to the observed sequence information to obtain the sequence lookup table I and the sequence lookup table II, including:

[0020] For the station lookup table I corresponding to the aerosol, one-dimensional interpolation and dimensionality reduction are performed on the three parameter dimensions of solar zenith angle, relative azimuth angle, and observation elevation angle to obtain the sequence lookup table I;

[0021] For the station lookup table II corresponding to trace gases, one-dimensional interpolation and dimensionality reduction are performed on the six parameter dimensions of solar zenith angle, relative azimuth, observation elevation angle, total aerosol extinction, aerosol profile height parameter, and aerosol profile shape factor to obtain the sequence lookup table II, among which the total aerosol extinction, aerosol profile height parameter, and aerosol profile shape factor are obtained by inversion of the aerosol profile.

[0022] In one embodiment, inverting an aerosol profile according to observation sequence information and sequence lookup table I includes:

[0023] Step 5-1, randomly generate simulated profiles within the range of initial profile parameters, find the differential atmospheric mass factor corresponding to the remaining profiles after pre-screening the simulated profiles in the reduced-dimensional sequence lookup table I, and then calculate the total simulated differential slope range using the following formula:

[0024]

[0025] Among them, dAMF α is the differential atmospheric mass factor found from the reduced-dimensional sequence lookup table I, VCD is the inverted vertical total, where VCD is the VCD of O4;

[0026] Step 5-2, selecting the simulated profile parameters according to the matching degree of the simulated differential slant path, during which the correction factor sf is added to correct the simulated differential slant path corresponding to the 30° angle;

[0027] Step 5-3, redefine the range of profile parameters according to the simulated profile parameters obtained by screening, and repeat the inversion process of steps 5-1 and 5-2, and after multiple iterations and convergence, output the simulated profile parameters obtained by screening;

[0028] Step 5-4, outputting the final simulated profile information and profile-related intermediate information according to the simulated profile parameters obtained by screening; wherein the final simulated profile information includes: the best matching profile, the average profile, the weighted average profile with the weight set by the root mean square error, and the total mean of the simulated slope range;

[0029] Step 5-5, perform quality control inspection on the simulated profile based on the intermediate information related to the profile, and mark the abnormal profile.

[0030] In one embodiment, inverting the trace gas profile according to the observed sequence information and the sequence lookup table II includes:

[0031] Step 5-1', randomly generate simulated profiles within the range of initial profile parameters, find the differential atmospheric mass factor corresponding to the remaining profiles after pre-screening the simulated profiles in the reduced-dimensional sequence lookup table II, and then calculate the total simulated differential slope range using the following formula:

[0032]

[0033] Among them, dAMF α is the differential atmospheric mass factor found from the reduced-dimensional sequence lookup table II, VCD is the inverted vertical total, and is obtained by the formula Calculated, N is the total observed elevation angle.

[0034] Step 5-2', screening the simulated profile parameters according to the matching degree of the total amount of simulated differential slope range;

[0035] Step 5-3', redefine the range of profile parameters according to the simulated profile parameters obtained by screening, and repeat the inversion process of step 5-1' and step 5-2', after multiple iterations and convergence, output the simulated profile parameters obtained by screening;

[0036] Step 5-4', outputting the final simulated profile information and profile-related intermediate information according to the simulated profile parameters obtained by screening; wherein the final simulated profile information includes: the best matching profile, the average profile, the weighted average profile with the weight set by the root mean square error, and the total mean of the simulated slope range;

[0037] Step 5-5', perform quality control inspection on the simulated profile according to the intermediate information related to the profile, and mark the abnormal profile.

[0038] In one embodiment, the simulated profile parameters are screened according to the matching degree of the simulated differential slope total amount, including:

[0039] Calculate the residual absolute value of the total amount of simulated differential slope and the total amount of observed differential slope α The minimum value of the RMS corresponding to all simulated profiles is marked as RMSmin, and RMSmin is used as the screening parameter. At the same time, the preset proportional coefficients F1(i), F2, and F3(i) are combined, and the corresponding simulated profiles are screened by the preset screening criteria presented by the following formula:

[0040] RMS≤RMSmin*F1(i)

[0041] residual α ≤F2*F3(i)

[0042] Among them, the two formulas represent the screening RMS to satisfy RMS≤RMSmin*F1(i), and residual α Meet residual α The simulated profile with ≤F2*F3(i) is regarded as the simulated profile that meets the requirements. i represents the number of iterations. In each iteration, the values ​​of the proportional coefficients F1(i) and F3(i) are different. The proportional coefficient F2 is determined according to the following formula:

[0043]

[0044] Where dSCDerr is the error of the total differential slope, generating dSCD α Generate synchronously.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] Combined with the differential atmospheric mass factor lookup table constructed according to the atmospheric radiation transfer model, rapid statistical simulation and iteration are carried out to finally obtain the vertical distribution of atmospheric pollution components. This greatly improves the profile inversion speed while ensuring the rationality and reliability of the inversion results. At the same time, a quality control plan for the inversion results is also given. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 It is a flow chart of a method for quickly retrieving and inverting atmospheric pollution component profiles based on hyperspectral observation provided by an embodiment of the present invention;

[0049] Figure 2 The aerosol vertical profile of the atmospheric pollution component observation site provided by the embodiment of the present invention;

[0050] Figure 3 A vertical distribution diagram of four atmospheric trace gases at an atmospheric pollution component observation site provided by an embodiment of the present invention;

[0051] Figure 4 A comparison and verification diagram of the aerosol inversion results provided by the embodiment of the present invention and the external independent data sun photometer. In the figure, the horizontal axis is time, the vertical axis is the total aerosol extinction, the cross-shaped points represent the total aerosol extinction results of the sun photometer, and the triangles represent the total aerosol extinction results of the hyperspectral instrument inversion;

[0052] Figure 5 The four figures are the comparison and verification diagram of the NO2 inversion results provided by the embodiment of the present invention and the external independent data of the tower data of the Beijing Institute of Atmospheric Physics. The four figures are the comparison diagram of the inverted NO2 concentration and the height of 100m, 200m, and 300m of the tower of the Beijing Institute of Atmospheric Physics and the comparison diagram of the average results. The horizontal axis is the inversion result and the vertical axis is the tower data of the Beijing Institute of Atmospheric Physics. DETAILED DESCRIPTION

[0053] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.

[0054] Figure 1 FIG. 1 is a flow chart of a method for quickly retrieving and retrieving atmospheric pollution component profiles based on hyperspectral observations provided in an embodiment. Figure 1 As shown, the method for quickly retrieving and inverting atmospheric pollution component profiles based on hyperspectral observation provided in the embodiment includes the following steps:

[0055] Step 1: Use the radiation transfer model to establish the differential atmospheric mass factor lookup table I for aerosol inversion and the differential atmospheric mass factor lookup table II for trace gas inversion.

[0056] In the embodiment, the atmospheric pollution components mainly include aerosols and trace gases, wherein the trace gases include nitrogen dioxide, sulfur dioxide, formaldehyde, glyoxal, methylglyoxal, nitrous acid, water vapor, etc. The concentration of O4 in the atmosphere is proportional to the square of the oxygen concentration, and the oxygen concentration is relatively stable in the atmosphere, mainly depending on the temperature and pressure. Therefore, the vertical profile of O4 can be used as a known quantity, and the size of the total amount of O4 oblique path measured in the hyperspectral observation is mainly affected by the photon transmission path, and aerosol extinction is the main factor affecting photon transmission. Therefore, the change of aerosol in the atmosphere can be reflected by the change of the total amount of O4 oblique path. However, the influence of trace gases on the optical path is small, and the photon path is mainly affected by the aerosol condition. The determination of the differential atmospheric mass factor of trace gases must first determine the aerosol condition. Based on this, the differential atmospheric mass factor lookup table I of aerosols and the differential atmospheric mass factor lookup table II of trace gases are constructed separately for aerosols and trace gases.

[0057] In the embodiment, when constructing the differential atmospheric mass factor lookup table I for aerosol inversion, three types of parameters are considered, namely, observation parameters, geometric parameters, and aerosol profile parameters, wherein the observation parameters include observation latitude (LAT) and observation month (MON); the geometric parameters include solar zenith angle (SZA), relative azimuth angle (RAA), and observation elevation angle (ELEV, i.e., α); and the aerosol profile parameters include aerosol total extinction (c aero ), aerosol profile height parameter (h aero ), aerosol profile shape factor (s aero The aerosol differential atmospheric mass factor lookup table I is equivalent to the eight-dimensional matrix Table(LAT,MON,SZA,RAA,ELEV,c aero ,h aero ,s aero ), in the embodiment, the values ​​of each dimension are shown in Table 1.

[0058] In the embodiment, when constructing the differential atmospheric mass factor lookup table II for trace gas inversion, four types of parameters are considered, namely, observation parameters, geometric parameters, aerosol profile parameters, and trace gas profile parameters, wherein the specific parameters included in the observation parameters, geometric parameters, and aerosol profile parameters are the same as those described above. The trace gas profile parameters include the trace gas profile height parameter (h trace ) and trace gas profile shape factor (s trace ). The trace gas differential atmospheric mass factor lookup table II is equivalent to the ten-dimensional matrix Table(LAT,MON,SZA,RAA,ELEV,c aero ,h aero ,s aero ,h trace ,s trace), in the embodiment, the values ​​of each dimension are shown in Table 1.

[0059] Table 1

[0060]

[0061] In the embodiment, the corresponding relationship between the profile of the atmospheric pollution component and the profile parameter includes:

[0062] When the shape factor s is between 0 and 1, it means that the atmospheric pollution components are high at the bottom and low at the top:

[0063]

[0064] When the shape factor s=1, it means that the atmospheric pollution components are mainly concentrated in the bottom layer:

[0065]

[0066] When the shape factor s is between 1 and 2, it represents the distribution of atmospheric pollution components in the elevated layer:

[0067]

[0068] It should be noted that for aerosols, the shape factor s is s aero , the height parameter h is h aero , c is c aero , z is a highly stratified array, p(z) is the aerosol extinction coefficient corresponding to the highly stratified array; for trace gases, the shape factor s is s trace , the height parameter h is h trace , c is the molecular column concentration of trace gas c trace , z is the highly stratified array, p(z) is the trace gas concentration corresponding to the highly stratified array. Since the trace gas has little effect on the optical path, the differential atmospheric mass factor of the trace gas is related to the molecular column concentration c of the trace gas. trace It is irrelevant, so c is set to 1 when establishing the lookup table.

[0069] When applied, the above profile parameters s, h, c and z are substituted into the above formulas (1)-(3) to calculate the profiles of atmospheric pollution components, namely the aerosol profile and trace gas profile.

[0070] In the embodiment, some settings of the radiation transfer model are as follows: the instrument field of view is set to 0.045D0, the number of fine grids is set to 5, the wavelength band is set to 360-360, the surface albedo is set to 0.05, and the atmospheric radiation transfer model used in the simulation adopts a pseudo-spherical scattering atmospheric model as the simulation type. The input temperature and pressure profile adopts the temperature and pressure profile information assimilated by the regional air quality model that is fully coupled online with the global atmospheric chemical transport model, meteorological model and chemical model.

[0071] Due to the large number of parameters in the lookup table, it takes a long time to directly use the atmospheric radiation transfer model simulation method to invert the differential atmospheric mass factor dAMF corresponding to all parameters. The speed is too slow and cannot be used in field observations. Therefore, in the embodiment, the radiation transfer model does not directly calculate the atmospheric mass factor, but calculates the influence factor IF(z) at different heights z. Based on the influence factor, the atmospheric mass factor at different observation elevation angles is calculated according to formula (1), and the atmospheric mass factor AMF at a 90-degree angle at the zenith is taken as 90° AMF α Do the difference to get the differential atmospheric mass factor dAMF α :

[0072]

[0073] dMF α =AMF α -AMF 90° (5)

[0074] Among them, p(z) is the profile under different profile parameters.

[0075] When constructing the differential atmospheric mass factor lookup table I, p(z) in formula (4) is changed to the vertical profile of the known O4, and the dAMF of O4 under three different parameters can be obtained. The aerosol lookup table I is constructed based on the dAMF of O4.

[0076] When looking up the differential atmospheric mass factor in Table II, substitute the value of c=1 into formulas (1)-(3) to calculate p(z), and then use the calculated p(z) to calculate the differential atmospheric mass factor dAMF of the trace gas through formulas (4)-(5): α , to construct the differential atmospheric quality factor lookup table II.

[0077] Step 2: Reduce the dimensions of the differential atmospheric quality factor lookup table I and the differential atmospheric quality factor lookup table II with respect to the site dimension and the observation month to obtain the site lookup table I and the site lookup table II.

[0078] In an embodiment, after constructing the differential atmospheric quality factor lookup tables I and II, for different sites, interpolating the lookup tables according to the local observation dimension lat and the observation month mon can reduce the size of the lookup tables, reduce the memory usage during subsequent inversion, reduce the hardware requirements of the computer, reduce data calculation, and speed up the inversion.

[0079] It has been verified by experiments that for the differential atmospheric quality factor lookup tables I and II, the six-dimensional station lookup table I Table (lat, mon, SZA, RAA, ELEV, c aero ,h aero ,s aero ) and eight-dimensional site lookup table II Table(lat,mon,SZA,RAA,ELEV,c aero ,h aero ,s aero ,h trace ,s trace ) takes less than 2 minutes, while the traditional method of obtaining an applicable lookup table requires millions of radiation transfer model calculations, which takes about 5 to 10 days on a high-performance computer.

[0080] Step 3: collect observation data and use the solar scattering spectrum contained in the observation data to calculate the total amount of observed differential slant range of atmospheric pollution components.

[0081] A hyperspectral instrument is used to collect the solar scatter spectrum according to a certain elevation observation sequence to obtain observation data. The observation data includes longitude and latitude, solar zenith angle, relative azimuth, observed solar scatter spectrum, observation time, observation elevation, etc. In the embodiment, an elevation observation sequence of 1°, 2°, 3°, 4°, 5°, 6°, 8°, 10°, 15°, 30°, and 90° is used for spectrum collection. In the embodiment, the solar scatter spectrum is collected by hyperspectral observation, and the collected solar scatter spectrum is subjected to nonlinear correction and instrument dark background filtering. Based on the Lambert-Beer law, the total amount of observed differential slant path of the atmospheric pollution components is calculated using the least squares algorithm. This process is a prior art, and reference may be made to the vertical distribution inversion method of atmospheric pollution components based on optical remote sensing disclosed in Application No. CN202010611508.0. Among them, the total amount of observed differential slant path at each observation angle is expressed as dSCD α .

[0082] Step 4: extract observation sequence information based on the observation data and the total amount of observed differential slope range, and perform dimensionality reduction on site lookup table I and site lookup table II according to the observation sequence information to obtain sequence lookup table I and sequence lookup table II.

[0083] In the embodiment, the complete elevation observation sequence is 1°, 2°, 3°, 4°, 5°, 6°, 8°, 10°, 15°, 30°, 90°. Before the inversion begins, the algorithm extracts the observation data according to the observation sequence. The extracted observation sequence information includes: observation time, total amount of differential slant range dSCD of O4 and other trace gases α And the error of the total amount of observed differential slant range dSCDerr, the solar zenith angle θ, the relative azimuth angle β, the observed elevation angle α, etc.

[0084] In the embodiment, the dimensions of the site lookup tables I and II obtained in step 2 are still too high, and the speed of computer processing high-dimensional interpolation is greatly reduced as the dimension increases. Therefore, before inversion is performed on each set of data, the site lookup table obtained in step 2 is applied to perform dimensionality reduction processing again according to the observed sequence to obtain sequence lookup tables I and II, which specifically include:

[0085] For the station lookup table I obtained in step 2, one-dimensional interpolation and dimensionality reduction are performed on the three parameters of the solar zenith angle θ, the relative azimuth angle β, and the observation elevation angle α, that is, the aerosol sequence lookup table ITable (lat, mon, θ, β, α, c aer° ,h aer° ,s aer° ).

[0086] The station lookup table II obtained in step 2 is analyzed with respect to the solar zenith angle θ, relative azimuth angle β, observation elevation angle α, c mean 、h mean 、s mean The one-dimensional interpolation and dimensionality reduction of the six parameters are used to obtain the trace gas sequence lookup table II (lat, mon, θ, β, α, c mean ,h mean ,s mean ,h trace ,s trace ). Where c mean 、h mean 、s mean are the average values ​​of aerosol profile parameters obtained by inverting the aerosol profile of the current sequence.

[0087] Step 5, inverting aerosol profile and trace gas profile according to the observed sequence information and sequence lookup table I and sequence lookup table II.

[0088] Step 5-1, randomly generate simulated profiles within the range of initial profile parameters, find the differential atmospheric mass factor corresponding to the remaining profiles after pre-screening the simulated profiles in the reduced-dimensional sequence lookup table I and sequence lookup table II, and then calculate the total amount of simulated differential slope range

[0089] In the embodiment, each set of observation data generates multiple sets (exemplarily 125,000 sets) of simulated profile parameters of aerosols or multiple sets (exemplarily 2,500 sets) of simulated profile parameters of trace gases according to the parameter dimension to be searched within the range of the initial profile parameters. It should be noted that before inversion, the simulated profile parameters of aerosols should be pre-screened, specifically including: pre-screening the simulated profiles above a preset height (e.g., 5 km) and with aerosol extinction greater than a preset concentration (e.g., 1e-5) to obtain the simulated profile parameters for inversion. For the range of the initial profile parameters, see Table 2. The simulated profile parameters of trace gases are not pre-screened.

[0090] Table 2

[0091] Finding process configuration parameters Parameter settings O4 VCD 1.276329494696885e+43 <![CDATA[Initial c aero Parameter random range]]> [0.05,5] <![CDATA[Initial h aero Parameter random range]]> [0.1,5] <![CDATA[Initial s aero Parameter random range]]> [0.2,1.8] <![CDATA[Initial h trace Parameter random range]]> [0.1,5] <![CDATA[Initial s trace Parameter random range]]> [0.2,1.8] F1 [3.0,2.0,1.3] F3 [1.5,1.0,1.0]

[0092] In the embodiment, in the sequence lookup tables I and II, the differential atmospheric mass factor corresponding to the remaining profile after the simulated profile is pre-screened is searched, and then the simulated differential slope total amount is calculated using formula (6):

[0093]

[0094] Among them, dAMF α It is the differential atmospheric mass factor of Tables I and II from the reduced-dimensional sequence lookup. VCD is the inverted vertical total. When inverting aerosols, VCD is taken as the VCD of O4. When inverting trace gases, the approximate value of VCD is obtained using formula (7).

[0095]

[0096] Step 5-2, the simulation profile parameters are selected according to the matching degree of the total amount of simulated differential slant path, during which the correction factor sf is added to correct the total amount of simulated differential slant path corresponding to an angle of 30° when performing aerosol profile inversion.

[0097] In the embodiment, the simulated profile parameters are screened according to the matching degree of the simulated differential slope range total, including:

[0098] Calculate the total amount of simulated differential slant distance and the total amount of observed differential slant distance dSCD α The absolute value of the residual at each observation angle α And the overall root mean square error RMS. The minimum value of the RMS corresponding to all simulated profiles is marked as RMSmin, and RMSmin is used as the screening parameter. At the same time, combined with the preset proportional coefficients F1(i), F2, F3(i), the corresponding simulated profile parameters are screened through formulas (8) and (9):

[0099] RMS≤RMSmin*F1(i) (8)

[0100] residual α ≤F2*F3(i) (9)

[0101] Where i represents the number of iterations. In each iteration, the values ​​of the proportional coefficients F1(i) and F3(i) are different. Taking three iterations as an example, the corresponding values ​​of F1(i) are 3.0, 2.0, and 1.3, and the corresponding values ​​of F3(i) are 1.5, 1.0, and 1.0, as shown in Table 2. The values ​​of the three iterations are empirical values ​​determined through exploration. The proportional coefficient F2 is determined according to formula (10):

[0102]

[0103] Among them, dSCDerr is the observation error, generating dSCD α Generate synchronously.

[0104] Since the high-elevation-angle data of some stations have abnormal fitting effects when fitting the aerosol profile, in order to solve this problem, a fitting correction factor is introduced to the total observed differential slant path dSCD at an observation elevation of 30° during aerosol inversion. 30° In the embodiment, the initial correction factor sf is set to 1, and the total amount of simulated differential slope of 30° determined by equation (6) is corrected using equation (12). Then, take the 30° angle corresponding to The average value is dSCD mean30 , set the new correction factor by the given ratio β. The specific new correction factor sf is determined by formula (11), and then the uncorrected sf determined by formula (6) is calculated again using formula (12). To make a correction:

[0105]

[0106]

[0107] According to the corrected dSCD model Repeat step 5-2 and calculate the new correction factor sf, and repeat this cycle twice.

[0108] Step 5-3, redefine the range of profile parameters according to the simulated profile parameters obtained by screening, and repeat the inversion process of steps 5-1 and 5-2. After multiple iterations and convergence, output the simulated profile parameters obtained by screening.

[0109] The simulated profile is gridded according to the preset profile parameter grid. If all simulated profiles in a certain grid do not meet the pre-screening conditions in step 5-1, the grid is deleted and the profile parameter range defined by the remaining grid is used as the new profile parameter range. Repeat the inversion process of steps 5-1 and 5-2. After multiple iterations and convergence, the remaining simulated profile parameters and their corresponding residual absolute values, root mean square errors, simulated differential slope distances, etc. are output.

[0110] Step 5-4, outputting the final simulated profile information and profile-related intermediate information according to the screened simulated profile parameters.

[0111] In the embodiment, the remaining profile parameters are restored to profiles according to the given height layers according to equations (1)-(3), and then the final profile information of the inversion is calculated. The final profile information includes: the best matching profile; the average profile; the weighted average profile with the weight set by the root mean square error; the total mean of the simulated slope, etc. The molecular column concentration c is used in the trace gas inversion. trace Substituting for c in equations (1)-(3), the specific column concentration is determined by equation (13).

[0112]

[0113] Where molarmass is the molar mass of the gas and NA is Avogadro's constant, which is taken as 6.02214076e16.

[0114] In the embodiment, the outputted profile-related intermediate information includes:

[0115] The average value of the remaining simulated profile parameters (c mean ,h mean ,s mean ), median (c median ,h median ,s median ), variance (c var ,h var ,s var ), kurtosis (c kurt ,h kurt ,s kurt ), skewness (c skew ,h skew ,s skew ), the kurtosis of the final profile kurt ) and skewness (profile skew ); the number of remaining simulated profiles; the root mean square error between the total amount of simulated differential slant range of the remaining simulated profiles and the total amount of observed differential slant range; the root mean square error of the total amount of observed differential slant range itself; the error between the total amount of simulated differential slant range at each elevation angle and the total amount of observed differential slant range, etc.

[0116] Step 5-5, perform quality control inspection on the simulated profile based on the intermediate information related to the profile, and mark the abnormal profile.

[0117] In the embodiment, the simulated profile is quality controlled and tested according to the judgment methods such as whether the profile parameters obtained by inversion are abnormal, whether the profile shape is abnormal, whether the profile inversion error is too large, etc., and the abnormal profile is marked.

[0118] The embodiment uses the method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observation provided by the present invention to invert the vertical gradient distribution of aerosols and trace gases at the observation site of the Chinese Academy of Meteorological Sciences. Figure 2 and Figure 3 , Figure 2 and Figure 3 It can be seen that the inversion results show a distribution characteristic of high at the bottom and low at the top from near the ground to high altitude, which is consistent with the fact that aerosol particles are heavy and not easy to float high, and trace gas pollutants are basically generated near the ground. HONO shows an obvious high value near the ground in the morning, which is consistent with the fact that HONO is generated near the ground and is easily photolyzed by light.

[0119] This example compares and verifies the AOD and aerosol inversion results of an external independent data sun photometer (located at 39.9472N, 116.3206E). Figure 3 , the correlation between the inverted aerosol optical depth and the aerosol optical depth measured by the sun photometer R 2 At the same time, this embodiment compares and verifies the inversion results of the NO2 data of the Beijing Institute of Atmospheric Physics tower and the NO2 profile at 0-100m, 100-200m, and 200-300m. Figure 3 , the correlation between the inverted NO2 vertical distribution at 0-100m, 100-200m, 200-300m and the tower results R 2 The average correlation R 2 The results show that the inversion results can well reproduce the total amount and vertical distribution of aerosols and atmospheric pollution components on the transmission channel, indicating that the research results can fully meet the requirements for the analysis of the vertical gradient distribution and evolution of atmospheric pollution components of hyperspectral observation instruments.

[0120] The above embodiment provides a method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observations. The use of a pre-inversion lookup table can simplify the inversion calculation amount, and a statistical simulation method is used to achieve rapid search, which reduces hardware requirements and has higher real-time performance. It is expected to achieve real-time distribution of aerosols and atmospheric trace gases after spectral acquisition, which has great advantages in optical remote sensing.

[0121] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observation, characterized in that: The following steps are involved: Step 1, using the radiation transfer model to establish a differential atmospheric mass factor lookup table I for aerosol inversion and a differential atmospheric mass factor lookup table II for trace gas inversion; Step 2, reducing the dimensions of the differential atmospheric quality factor lookup table I and the differential atmospheric quality factor lookup table II with respect to the site dimension and the observation month to obtain the site lookup table I and the site lookup table II, specifically including: for different sites, interpolating the lookup table according to the local observation dimension and observation month to reduce the volume of the lookup table and achieve dimensionality reduction; Step 3, collecting observation data, and using the solar scattering spectrum contained in the observation data to calculate the total amount of observed differential slant range of atmospheric pollution components; Step 4, extracting observation sequence information based on the observation data and the total amount of observed differential slant range, and performing dimensionality reduction on the site lookup table I and the site lookup table II according to the observation sequence information to obtain the sequence lookup table I and the sequence lookup table II; Step 5, inverting aerosol profiles and trace gas profiles according to the observation sequence information and sequence lookup table I and sequence lookup table II, including: The aerosol profile is inverted based on the observation sequence information and sequence lookup table I, specifically including: Step 5-1, randomly generate simulated profiles within the range of initial profile parameters, find the differential atmospheric mass factor corresponding to the remaining profiles after pre-screening the simulated profiles in the reduced-dimensional sequence lookup table I, and then calculate the total simulated differential slope range using the following formula: Among them, dAMF α is the differential atmospheric mass factor found from the reduced-dimensional sequence lookup table I, VCD is the inverted vertical total, where VCD is the VCD of O4; Step 5-2, selecting the simulated profile parameters according to the matching degree of the simulated differential slant path, during which the correction factor sf is added to correct the simulated differential slant path corresponding to the 30° angle; Step 5-3, redefine the range of profile parameters according to the simulated profile parameters obtained by screening, and repeat the inversion process of steps 5-1 and 5-2, and after multiple iterations and convergence, output the simulated profile parameters obtained by screening; Step 5-4, outputting the final simulated profile information and profile-related intermediate information according to the simulated profile parameters obtained by screening; wherein the final simulated profile information includes: the best matching profile, the average profile, the weighted average profile with the weight set by the root mean square error, and the total mean of the simulated slope range; Step 5-5, performing quality control inspection on the simulated profile according to the intermediate information related to the profile, and marking the abnormal profile; According to the observation sequence information and sequence lookup table II, the trace gas profile is inverted, including: Step 5-1', randomly generate simulated profiles within the range of initial profile parameters, find the differential atmospheric mass factor corresponding to the remaining profiles after pre-screening the simulated profiles in the reduced-dimensional sequence lookup table II, and then calculate the total simulated differential slope range using the following formula: Among them, dAMF α is the differential atmospheric mass factor found from the reduced-dimensional sequence lookup table II, VCD is the inverted vertical total, and is obtained by the formula Calculated, N is the total observed elevation angle; Step 5-2', screening the simulated profile parameters according to the matching degree of the total amount of simulated differential slope range; Step 5-3', redefine the range of profile parameters according to the simulated profile parameters obtained by screening, and repeat the inversion process of step 5-1' and step 5-2', after multiple iterations and convergence, output the simulated profile parameters obtained by screening; Step 5-4', outputting the final simulated profile information and profile-related intermediate information according to the simulated profile parameters obtained by screening; wherein the final simulated profile information includes: the best matching profile, the average profile, the weighted average profile with the weight set by the root mean square error, and the total mean of the simulated slope range; Step 5-5', perform quality control inspection on the simulated profile according to the intermediate information related to the profile, and mark the abnormal profile.

2. The method for rapidly retrieving and inverting atmospheric pollution component profiles based on hyperspectral observation according to claim 1 is characterized in that: The influence factor IF(z) at different heights z is calculated using the radiation transfer model. Then, the atmospheric mass factor AMF at different observation elevation angles is calculated using the following formula based on the influence factor IF(z) and the profile p(z) under different profile parameters: α , the atmospheric mass factor AMF at 90 degrees zenith 90° AMF α Do the difference to get the differential atmospheric mass factor dAMF α : The differential atmospheric mass factors at different observation elevation angles form the differential atmospheric mass factor lookup table I for aerosol inversion and the differential atmospheric mass factor lookup table II for trace gas inversion.

3. The method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observation according to claim 1 is characterized in that: In step 1, the differential atmospheric mass factor lookup table I used for aerosol inversion contains parameters of eight dimensions: observation dimension, observation month, solar zenith angle, relative azimuth, observation elevation angle, aerosol total extinction, aerosol profile height parameter, and aerosol profile shape factor; The differential atmospheric mass factor lookup table II for trace gas inversion contains parameters in ten dimensions: observation dimension, observation month, solar zenith angle, relative azimuth, observation elevation angle, aerosol total extinction, aerosol profile height parameter, aerosol profile shape factor, trace gas profile height parameter, and trace gas profile shape factor.

4. The method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observation according to claim 1 is characterized in that: In step 4, the extracted observation sequence information includes: observation time, total observed differential slant range of O4 and other trace gases and the error of total observed differential slant range, solar zenith angle at that time, relative azimuth angle, and observation elevation angle.

5. The method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observation according to claim 4 is characterized in that: In step 4, the site lookup table I and the site lookup table II are respectively reduced in dimension according to the observed sequence information to obtain the sequence lookup table I and the sequence lookup table II, including: For the station lookup table I corresponding to the aerosol, one-dimensional interpolation and dimensionality reduction are performed on the three parameter dimensions of solar zenith angle, relative azimuth angle, and observation elevation angle to obtain the sequence lookup table I; For the station lookup table II corresponding to trace gases, one-dimensional interpolation and dimensionality reduction are performed on the six parameter dimensions of solar zenith angle, relative azimuth, observation elevation angle, total aerosol extinction, aerosol profile height parameter, and aerosol profile shape factor to obtain the sequence lookup table II, among which the total aerosol extinction, aerosol profile height parameter, and aerosol profile shape factor are obtained by inversion of the aerosol profile.

6. The method for rapid retrieval and inversion of atmospheric pollution component profiles based on hyperspectral observation according to claim 1 is characterized in that: The method of screening the simulated profile parameters according to the matching degree of the total amount of simulated differential slope range includes: Calculate the residual absolute value of the total amount of simulated differential slope and the total amount of observed differential slope α The minimum value of the RMS corresponding to all simulated profiles is marked as RMSmin, and RMSmin is used as the screening parameter. At the same time, the preset proportional coefficients F1(i), F2, and F3(i) are combined, and the corresponding simulated profiles are screened by the preset screening criteria presented by the following formula: RMS≤RMSmin*F1(i) residual α ≤F2*F3 (i) Among them, the two formulas represent the screening RMS to satisfy RMS≤RMSmin*F1(i), and residual α Meet residual α The simulated profile with ≤F2*F3(i) is regarded as the simulated profile that meets the requirements. i represents the number of iterations. In each iteration, the values ​​of the proportional coefficients F1(i) and F3(i) are different. The proportional coefficient F2 is determined according to the following formula: Where dSCDerr is the error of the total differential slope, generating dSCD α Generate synchronously.

Images