FY-4 / GIIRS data clear sky infrared spectrum reconstruction method under cloud influence

Abstract

The invention belongs to the technical field of hyperspectral data processing, and particularly relates to a method for eliminating the influence of cloud in GIIRS data of a part of cloud view fields and realizing clear sky infrared spectrum reconstruction so as to effectively improve the data utilization rate. Aiming at GIIRS cloud view field observation data of a static rail platform, observation data of a radiation imager (AGRI) with higher spatial resolution on the same platform is utilized, and on the basis of a traditional clear sky correction algorithm, the GIIRS and AGRI data characteristics are combined, so that the GIIRS cloud view field observation data of the static rail platform is obtained. A three-step optimization algorithm is provided, namely AGRI observation data information of a plurality of time sequences is utilized, channel information of a plurality of infrared windows is comprehensively considered, and reconstruction results of a plurality of auxiliary view fields are compared to reduce errors, so that a clear sky infrared spectrum high-precision reconstruction result of a GIIRS part cloud view field is obtained, the utilization rate of satellite data is increased, and the reconstruction efficiency of the GIIRS part cloud view field is improved. And data support is further provided for data assimilation and meteorological analysis.

Description

Technical Field

[0001] The present invention belongs to the technical field of hyperspectral data processing, and specifically aims to eliminate the influence of clouds in GIIRS data of a partial cloud field of view, realize clear sky infrared spectrum reconstruction, and thus effectively improve data utilization. Background Art

[0002] Fengyun-4 (FY-4) is my country's second-generation geostationary meteorological satellite, currently including A (FY-4A) and B (FY-4B). The Geostationary Interferometric Infrared Sounder (GIIRS) carried by the Fengyun-4 satellite has high spectral and high temporal resolution, but its spatial resolution is low (16km for FY-4A and 12km for FY-4B), resulting in a high probability of cloud cover in the observation area. The strong absorption and scattering effects of clouds in the infrared band significantly reduce the infrared radiation in the area below the clouds, thereby affecting the acquisition of clear sky infrared spectra. Therefore, it is necessary to reconstruct the clear sky infrared spectrum for the data with cloud fields to improve data utilization.

[0003] At present, the clear sky correction method is commonly used internationally to deal with cloud fields in polar-orbit platform observation data. The clear sky correction method converts the radiation values ​​of part of the cloud field into equivalent clear sky radiation by using the infrared observation data of adjacent parts of the cloud field, thereby eliminating the influence of clouds and obtaining the clear sky equivalent radiation value. However, unlike polar-orbit instruments, the GIIRS detector and radiation imager (Advanced Geosynchronous Radiation Imager, AGRI) carried on geostationary satellites have high temporal resolution but relatively low spatial resolution, resulting in a small number of AGRI pixels contained in the GIIRS field of view, which increases the error. In addition, the reconstruction parameters are highly dependent on different infrared windows, and the selection of different windows will cause differences in reconstruction accuracy. Finally, since the clear sky correction method usually only uses the information of a single field of view or window, it fails to fully integrate the data information of multiple fields of view and multiple windows, which further limits the accuracy of clear sky radiation reconstruction and affects the optimality of the calculation results. Summary of the invention

[0004] In order to solve the inaccuracy problems of existing clear sky correction methods in field of view selection and reconstruction parameters, the present invention is based on the characteristics of GIIRS and AGRI data, uses the observation data of AGRI, a radiation imager with higher spatial resolution on the same platform, and optimizes the traditional clear sky correction algorithm using a "three-step optimization" algorithm - introducing multiple time-series AGRI observation data information, comprehensively considering multiple infrared window channel information, and comparing the reconstruction results of multiple auxiliary fields of view to reduce errors, thereby obtaining high-precision reconstruction results of the clear sky infrared spectrum of part of the cloud field of view of GIIRS.

[0005] A method for reconstructing clear sky infrared spectra of FY-4 / GIIRS data under cloud influence, characterized by comprising the following steps:

[0006] Step 1: align the GIIRS and AGRI data in terms of time, space and spectrum dimensions;

[0007] Step 2: Based on the cloud detection product matching method of different instruments, the cloud mask of GIIRS is obtained by using the registered AGRI cloud detection product;

[0008] Step 3: Based on the GIIRS cloud detection results obtained in step 2, use the AGRI infrared band band12 (central wavelength of 10.8 μm) as the window channel in the GIIRS main field of view to calculate the average AGRI clear sky radiance.

[0009]

[0010] Where np is the number of AGRI clear sky viewing points, w is the inverse distance weight value, d is the actual distance from the AGRI pixel center to the center of the GIIRS main viewing field, and w is the reciprocal of d. is the radiance value of the i-th AGRI data;

[0011] Step 4: According to the result of step 3, obtain the clear sky radiance of window channel W After that, the reconstruction parameter N * Available through The radiometric observations of the GIIRS field of view window channel after spectral matching To calculate, get N * Then, the clear-sky equivalent radiance over the entire long-wave infrared spectral range of GIIRS is calculated.

[0012]

[0013] Where v i represents the spectral frequency, corresponds to the frequency v iThe equivalent radiation value, They represent the given frequencies v of the adjacent fields of view of GIIRS respectively. i The equivalent radiation value of

[0014] Step 5: Based on the results obtained in step 4, the AGRI radiance is interpolated according to the observation time of GIIRS to further accurately calculate the average clear sky radiance;

[0015] Step 6: Based on the results obtained in step 5, using multi-time series information, take the weighted average of the radiances of all clear-sky AGRI pixels in the GIIRS main field of view at three moments, and increase the AGRI clear-sky field of view points that can be used in the GIIRS main field of view calculation;

[0016] Step 7: Based on the results of step 6, the three infrared window channels are used together to calculate the difference between the average clear sky radiance of each channel of AGRI and the radiation obtained by spectral convolution of the GIIRS clear sky equivalent radiation value to construct the loss function J(N * ), minimize the loss function and get N * The analytical solution of

[0017] Step 8: Based on the result of step 7, multiple auxiliary fields of view are introduced. For each field of view pair, multiple moments and multiple channels are considered for spectral reconstruction to obtain different loss function values ​​J. k (NN * ), find the auxiliary field of view corresponding to the minimum value of the cost function, and the parameters corresponding to the field of view and the clear sky equivalent radiation value are the optimal solution;

[0018] Step 9: Optimize the optimal solution Perform quality control, calculate quality control functions,

[0019]

[0020] where I is the number of AGRI channels used for spectral reconstruction, Is to connect AGRI channel w i The radiance is converted into a function of brightness temperature. When RMSE TB When the clear sky spectrum is reconstructed successfully, the root mean square error between the average clear sky brightness temperature observed by AGRI and the equivalent clear sky brightness temperature value of GIIRS after spectral convolution is less than 1K.

[0021] Beneficial effects achieved by the present invention:

[0022] 1. In order to solve the problem that the number of AGRI pixels contained in the GIIRS field of view is small, resulting in inaccurate clear sky equivalent radiation estimation, the present invention proposes a multi-time series optimization method. By introducing AGRI observation data at multiple times, using the observation results at different time points for weighted averaging, increasing the number of AGRI clear sky field points used for estimation, reducing the error caused by insufficient number of pixels, and thus improving the estimation accuracy of clear sky equivalent radiation.

[0023] 2. In order to solve the error problem caused by improper window selection when using a single infrared band window, the present invention proposes a multi-channel window optimization method. The joint loss function is constructed by using multiple infrared band window data at the same time, and the cost function is minimized to optimize the calculation of reconstruction parameters, thereby reducing the error caused by single window selection, thereby improving the accuracy and reliability of spectral reconstruction.

[0024] 3. In order to solve the error caused by a single field of view, the present invention proposes a multi-field of view optimization method. By introducing data from multiple auxiliary fields of view, the cost function is constructed by comprehensively considering the data from different fields of view, and the auxiliary field of view corresponding to the minimum value of the cost function is found, thereby optimizing the estimation of the clear sky equivalent radiation value, reducing the error caused by field of view differences, and improving the accuracy and robustness of spectral reconstruction.

[0025] 4. In order to solve the limitations of traditional optimization methods in radiation estimation, window and field of view selection, this paper proposes a method of joint optimization of multi-time series, multi-channel window and multi-field of view. By combining the three optimization strategies, the error can be minimized, the basic model of GIIRS&AGRI clear sky infrared spectrum reconstruction is optimized, and the overall accuracy of spectrum reconstruction is improved.

[0026] 5. Most of the existing clear sky radiation reconstruction algorithms are designed for polar-orbiting satellites and cannot fully utilize the advantages of geostationary satellites. To solve this problem, the present invention proposes an optimization algorithm specifically suitable for geostationary satellites. Unlike polar-orbiting satellites, geostationary satellites have high temporal resolution and long-term stable observation capabilities, and can perform continuous and multi-time observations of the same area. The present invention makes full use of the spatial stability and temporal consistency of geostationary satellites, and through the joint optimization of multi-time series data, multi-channel windows and multi-fields of view, it realizes high-precision clear sky radiance reconstruction under geostationary satellite data, and expands the scope of application of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Technical Flowchart DETAILED DESCRIPTION

[0028] A clear sky infrared spectrum reconstruction method based on GIIRS and AGRI data, the detailed implementation method includes the following steps:

[0029] Step 1: Take 15 minutes as the threshold of observation time difference, that is, if the observation time of GIIRS and AGRI in the same area is less than 15 minutes apart, their observation data are considered to be matched in time;

[0030] Step 2: Perform spatial registration based on the field of view vector fast matching method to calculate the vector from the satellite position to the GIIRS and AGRI observation pixel positions on the surface of the Earth ellipsoid. Calculate the angle θ between two vectors,

[0031]

[0032] Respectively represent vector When the angle θ is less than half of the instantaneous field of view of GIIRS, that is, in is the instantaneous field of view of GIIRS, and it is considered that the AGRI pixel falls within the instantaneous field of view of GIIRS, that is, the two are spatially matched;

[0033] Step 3: interpolate the spectral response function of the imager AGRI to the spectral channel of the detector GIIRS to match the spectral dimension.

[0034]

[0035] In the formula, is the radiation value obtained by convolution, specifically the radiation observed by GIIRS convolved with the radiation of the AGRI imager, v1, v2 represent the spectral frequencies, f(ν) is the spectral response function of AGRI, R H (ν) is the radiation spectrum of the corresponding band of GIIRS;

[0036] Step 4: Based on the cloud detection product matching method of different instruments, the cloud mask of GIIRS is obtained using the registered AGRI cloud detection product. When there are only clear sky or possible clear sky AGRI pixels in the field of view of GIIRS, GIIRS is considered to be completely clear; when there are only cloudy or possible cloudy AGRI pixels in the field of view of GIIRS, GIIRS is considered to be completely covered by clouds; when there are both clear sky and cloudy AGRI fields in the field of view of GIIRS, GIIRS is considered to be partially covered by clouds, that is, partially cloudy;

[0037] Step 5: Based on the GIIRS cloud detection results obtained in step 4, use the AGRI infrared band band12 (central wavelength of 10.8 μm) W as the window channel in the GIIRS main field of view to calculate the average AGRI clear sky radiance.

[0038]

[0039] Where np is the number of AGRI clear sky viewing points, w is the inverse distance weight value, d is the actual distance from the AGRI pixel center to the center of the GIIRS main viewing field, and w is the reciprocal of d. is the radiance value of the i-th AGRI data;

[0040] Step 6: Based on the result of step 5, obtain the clear sky radiance of the window channel After that, the reconstruction parameter N * Available through The radiometric observations of the GIIRS field of view window channel after spectral matching To calculate, get N * Then, the clear-sky equivalent radiance over the entire long-wave infrared spectral range of GIIRS is calculated.

[0041]

[0042]

[0043] Where v i represents the spectral frequency, corresponds to the frequency υ i The equivalent radiation value, They represent the given frequencies v of the adjacent fields of view of GIIRS respectively. i The equivalent radiation value of

[0044] Step 7: Based on the results obtained in step 6, use the interpolation method to calculate the equivalent radiance of AGRI according to the observation time of GIIRS. In the AGRI data, find the time closest to t GIIRS Two moments t k and t k+1 Satisfy t k <t GIIRS <t k+1 , using the interpolation formula,

[0045]

[0046] This allows for a more accurate calculation of the average clear sky radiance of AGRI.

[0047] Step 8: Based on the results obtained in step 7, introduce multiple time series of AGRI observation data to recalculate Consider the current time T t Main time series, first 15 minutes T t-1 , after 15 minutes T t+1To assist the timing, the radiance of all completely clear sky AGRI pixels in the GIIRS main field of view at three moments is taken. The weighted average of

[0048]

[0049] Obtain more accurate average clear sky radiance values ​​from AGRI data

[0050] Step 9: Based on the results of step 8, introduce multiple infrared window channels to optimize the spectrum reconstruction parameters. Construct the loss function,

[0051]

[0052] W in the infrared channel i The equivalent radiation of the GIIRS clear sky equivalent radiation after spectral convolution is N * Substitute into In the calculation of Minimize the loss function,

[0053]

[0054] Get N * The analytical solution of NN * ;

[0055] Step 10: Based on the result of step 9, multiple auxiliary fields of view are introduced to further optimize the spectral reconstruction parameters. Consider all 8 field of view points in the "8 neighborhoods" of the main field of view. These k (k≤8) auxiliary points can form k reconstruction combinations with the main field of view points. For each field of view pair, considering multiple moments and multiple channels for spectral reconstruction, different cost function values ​​J will be obtained. k (NN * ). Find the auxiliary field of view k corresponding to the minimum value of the cost function m ,

[0056]

[0057] Field of view k m The corresponding parameter NN * and clear sky equivalent radiation value R CCR This is the optimal solution;

[0058] Step 11: Optimize the optimal solution Perform quality control. Calculate the quality control function,

[0059]

[0060] where I is the number of AGRI channels used for spectral reconstruction, Is to connect AGRI channel w i The radiance is converted into a function of brightness temperature. When RMSE TB When the clear sky spectrum is reconstructed successfully, the root mean square error between the average clear sky brightness temperature observed by AGRI and the equivalent clear sky brightness temperature value of GIIRS after spectral convolution is less than 1K.

Claims

1. A clear sky infrared spectrum reconstruction method for FY-4 / GIIRS data under cloud influence, characterized in that: It includes the following steps: Step 1: align the data of the Geostationary Interferometric Infrared Sounder (GIIRS) and the Advanced Geosynchronous Radiation Imager (AGRI) from the time dimension, spatial dimension and spectral dimension; Step 2: Based on the cloud detection product matching method of different instruments, the cloud mask of GIIRS is obtained by using the registered AGRI cloud detection product; Step 3: Based on the GIIRS cloud detection results obtained in step 2, use the AGRI infrared band band12 (central wavelength of 10.8 μm) as the window channel in the GIIRS main field of view to calculate the average AGRI clear sky radiance. Where np is the number of AGRI clear sky viewing points, w is the inverse distance weight value, d is the actual distance from the AGRI pixel center to the center of the GIIRS main viewing field, and w is the reciprocal of d. is the radiance value of the i-th AGRI data; Step 4: According to the result of step 3, obtain the clear sky radiance of window channel W After that, the reconstruction parameter N * Available through The radiometric observations of the GIIRS field of view window channel after spectral matching To calculate, get N * After that, the clear-sky equivalent radiation over the entire long-wave infrared spectral range of GIIRS can be calculated. Where v i represents the spectral frequency, corresponds to the frequency v i The equivalent radiation value, They represent the given frequencies v of the adjacent fields of view of GIIRS respectively. i The equivalent radiation value of Step 5: Based on the results obtained in step 4, the AGRI radiance is interpolated according to the observation time of GIIRS to further accurately calculate the average clear sky radiance; Step 6: Based on the results obtained in step 5, using multi-time series information, take the weighted average of the radiances of all clear-sky AGRI pixels in the GIIRS main field of view at three moments, and increase the number of AGRI clear-sky field points that can be used in the GIIRS main field of view calculation; Step 7: Based on the results of step 6, the three infrared window channels are used together to calculate the difference between the average clear sky radiance of each channel of AGRI and the radiation obtained by spectral convolution of the GIIRS clear sky equivalent radiation value to construct the loss function J(N * ), minimize the loss function and get N * The analytical solution of Step 8: Based on the result of step 7, multiple auxiliary fields of view are introduced. For each field of view pair, multiple moments and multiple channels are considered for spectral reconstruction to obtain different loss function values ​​J. k (NN * ), find the auxiliary field of view corresponding to the minimum value of the cost function, and the parameters corresponding to the field of view and the clear sky equivalent radiation value are the optimal solution; Step 9: Optimize the optimal solution Perform quality control, calculate quality control functions, where I is the number of AGRI channels used for spectral reconstruction, Is to connect AGRI channel w i The radiance is converted into a function of brightness temperature. When RMSE TB When the clear sky spectrum reconstruction is successful, the root mean square error between the average clear sky brightness temperature observation of AGRI and the spectrally convolved GIIRS clear sky equivalent brightness temperature is less than 1K.

2. The method according to claim 1, characterized in that: Step 1: Time dimension matching: 15 minutes is taken as the threshold of observation time difference. If the observation time of GIIRS and AGRI in the same area is less than 15 minutes apart, their observation data are considered to be matched in time.

3. The method according to claim 1, characterized in that: Step 1: spatial dimension matching is to perform spatial registration based on the field of view vector fast matching method, and calculate the vector from the satellite position to the GIIRS and AGRI observation pixel positions on the surface of the earth ellipsoid. Calculate the angle θ between two vectors, Respectively represent vector When the angle θ is less than half of the instantaneous field of view of GIIRS, that is, in is the instantaneous field of view of GIIRS, and it is considered that the AGRI pixel falls within the instantaneous field of view of GIIRS, that is, the two are spatially matched.

4. The method according to claim 1, characterized in that: Step 1: Spectral dimension matching is to interpolate the spectral response function of the imager AGRI to the spectral channel of the detector GIIRS to perform spectral dimension matching. In the formula, is the radiation value obtained by convolution, specifically the radiation observed by GIIRS convolved with the radiation of the AGRI imager, v1, v2 represent the spectral frequencies, f(v) is the spectral response function of AGRI, R H (v) is the radiation spectrum of the corresponding band of GIIRS.

5. The method according to claim 1, characterized in that: In step 2, the cloud mask is obtained using the cloud detection product of AGRI after registration. When there are only clear sky or possibly clear sky AGRI pixels in the GIIRS field of view, GIIRS is considered to be completely clear. When there are only cloudy or possibly cloudy AGRI pixels in the GIIRS field of view, GIIRS is considered to be completely covered by clouds. When there are both clear sky and cloudy AGRI fields in the GIIRS field of view, GIIRS is considered to be partially covered by clouds, that is, partially cloudy.

6. The method according to claim 1, characterized in that Step 5: Calculate the equivalent radiance of AGRI using the interpolation method based on the observation time of GIIRS. In the AGRI data, find the time closest to t GIIRS Two moments t k and t k+1 Satisfy t k <t GIIRS <t k+1 , using the interpolation formula, This allows for a more accurate calculation of the average clear sky radiance of AGRI.

7. The method according to claim 1, characterized in that Step 6: Introduce multiple time series of AGRI observation data to recalculate Consider the current time T t Main time series, first 15 minutes T t-1 , after 15 minutes T t+1 To assist the timing, the radiance of all completely clear sky AGRI pixels in the GIIRS main field of view at three moments is taken. The weighted average of Obtain more accurate average clear sky radiance values ​​from AGRI data 8. The method according to claim 1, characterized in that Step 7 introduces multiple infrared window channels to optimize the spectrum reconstruction parameters, and the constructed loss function is: is the infrared channel W i The equivalent radiation of the GIIRS clear sky equivalent radiation after spectral convolution is N * Substitute into In the calculation of Minimize the loss function, Get N * The analytical solution of NN * .

9. The method according to claim 1, characterized in that Step 8 introduces multiple auxiliary fields of view to further optimize the spectral reconstruction parameters. Consider all 8 field points of the main field of view "8 neighborhoods". These k (k≤8) auxiliary points can form k reconstruction combinations with the main field of view points respectively. For each field of view pair, consider multiple moments and multiple channels for spectral reconstruction to obtain different cost function values ​​J. k (NN * ), find the auxiliary field of view k corresponding to the minimum value of the cost function m , Field of view k m The corresponding parameter NN * and clear sky equivalent radiation value R CCR This is the optimal solution.

Images