Iterative deviation correction method considering non-uniformity of satellite instrument probe element

Abstract

The invention discloses an iterative deviation correction method considering satellite instrument probe element heterogeneity, which comprises the following steps of: correcting amplitude by deviation, constructing a cost function and initializing, constructing a deviation operator, clearing probe element deviation, and updating an atmospheric quality deviation correction BC coefficient according to an observation background deviation O-B, the updated atmospheric quality deviation correction BC coefficient is adopted to calculate the observation background deviation O-B residual error as the detection element deviation, the observation background deviation O-B of the next round is calculated according to the detection element deviation of the previous round, the detection element deviation calculation step is repeated, iteration is conducted for multiple times till the atmospheric quality deviation correction BC coefficient is stable, and the detection element deviation is obtained. And outputting the final air quality deviation correction BC coefficient and the detection element deviation to carry out deviation correction on the observation data. According to the method, both the air mass deviation correction coefficient and the detector deviation tend to be stable, the normality of deviation of initial guess observation and simulation results is improved, the peak value and the average value of the deviation are closer to zero, and the method can be used for solving the problem of non-uniformity of detection elements of satellite instruments and infrared hyperspectral detectors.

Description

Technical Field

[0001] This invention relates to the field of satellite data processing technology, and in particular to an iterative deviation correction method that takes into account the non-uniformity of satellite instrument detectors. Background Technology

[0002] In satellite remote sensing data processing, instrument system bias is one of the key factors restricting the accuracy of observational data, and its correction effect directly affects the quality of satellite data assimilation in numerical weather prediction. Traditional bias correction methods have certain limitations when dealing with satellite instrument detector biases with complex non-uniform characteristics. Problems such as overcorrection or inability to effectively distinguish biases from different sources may occur. For example, the 128 detectors of the Geostationary Infrared Interferometer (GIIRS) exhibit significant non-uniformity, posing a challenge to accurate bias correction. Therefore, this invention proposes an iterative bias correction method that considers the non-uniformity of satellite instrument detectors. Through an iterative process, both the air mass bias correction coefficient and the detector bias tend to stabilize, effectively handling detector non-uniformity and avoiding overcorrection. This helps improve the application value of observational data in downstream remote sensing research and satellite data assimilation studies. Summary of the Invention

[0003] The purpose of this invention is to provide an iterative deviation correction method that takes into account the non-uniformity of satellite instrument detectors.

[0004] To achieve the above objectives, the present invention is implemented according to the following technical solution: This invention includes the following steps: The constrained bias correction (CBC) method is used to limit the bias correction magnitude, and a cost function is constructed and initialized. A bias operator is constructed by combining atmospheric state dependence bias and detector bias; the atmospheric state dependence bias includes atmospheric state prediction factor and atmospheric quality bias correction BC coefficient. The probe bias is cleared to zero. The initial guess of the observed background deviation OB is calculated based on the original observed brightness temperature and the simulated brightness temperature of the background field. The atmospheric mass deviation correction coefficient BC is updated based on the observed background deviation OB. The atmospheric state dependence bias is calculated using the updated atmospheric quality deviation correction BC coefficient. The atmospheric state dependence bias is subtracted from the initial guess of the observed background deviation OB to obtain the OB residual as the probe bias. Calculate the background deviation OB for the next round of observations based on the probe deviation from the previous round. Repeat the probe deviation calculation steps and iterate multiple times until the atmospheric quality deviation correction BC coefficient is stable. Output the final atmospheric quality deviation correction BC coefficient and probe deviation to correct the deviation of the observation data.

[0005] Furthermore, the cost function is specifically as follows: ; in Let cost function be Background state The BC vector is used to correct for atmospheric quality deviation, and the BC coefficient is used to correct for atmospheric quality deviation. composition, To observe the background deviation OB, The original observed brightness temperature, To simulate brightness temperature for the background field, For the bias operator, The observation error covariance matrix, Prior estimation of probe bias The error covariance matrix, For regularization parameters; Furthermore, the deviation operator is specifically: ; ; in This is an atmospheric state-dependent bias. For probe bias, It is a constant for a given channel. This is an atmospheric state predictor, derived from the background state.

[0006] Furthermore, the method for initializing the cost function includes: Prior estimation of probe bias Error covariance matrix Equal to the observation error covariance matrix Set the regularization parameter to ; The peak value of the observed background deviation OB was obtained by performing routine bias correction on the instrument, and the prior estimate of the probe bias was taken. It equals the peak value of the observed background deviation OB.

[0007] Furthermore, the atmospheric state prediction factor corresponds to the air mass thickness at different altitudes.

[0008] Furthermore, the condition for updating the atmospheric quality deviation correction BC coefficient based on the observed background deviation OB is that the cost function is minimized.

[0009] Furthermore, the method for calculating the observation background deviation OB of the next round based on the probe deviation of the previous round includes: The initial estimate of the observed background deviation OB is subtracted from the probe bias of the previous round to obtain the observed background deviation OB calculated in the next round. The expression is as follows:

[0010] in For the first The observed background deviation OB calculated in rounds, For the first The probe deviation calculated in rounds.

[0011] Furthermore, the criterion for determining the stability of the atmospheric quality deviation correction BC coefficient is that the norm of the change in the atmospheric quality deviation correction BC coefficient is less than a preset threshold.

[0012] The beneficial effects of this invention are: This invention is an iterative deviation correction method that takes into account the non-uniformity of satellite instrument detectors. Compared with the prior art, this invention has the following technical advantages: This invention, through constructing a cost function and bias operator, correcting the BC coefficient and calculating atmospheric state-dependent bias, calculating the observation background deviation OB residual to update the detector bias, and iterative steps, can effectively limit the magnitude of bias correction in the iterative bias correction of detector non-uniformity of satellite instruments, prevent over-correction and avoid introducing background bias. It can stabilize both the air mass bias correction coefficient and the detector bias, improve the normality of the deviations in the initial observation and simulation results, and make their peak and average values ​​closer to zero. It can be used to solve the detector non-uniformity problem of satellite instruments and infrared hyperspectral detectors, thereby improving the application value of observation data in downstream applications such as remote sensing research and satellite data assimilation research. Attached Figure Description

[0013] Figure 1 This is a flowchart of the iterative deviation correction method for considering the non-uniformity of satellite instrument detectors according to the present invention; Figure 2 This is an iterative flowchart of an iterative deviation correction method for satellite instrument detectors that takes into account non-uniformity of the present invention. Figure 3 A graph showing the iterative changes in the BC coefficient and probe bias for atmospheric quality deviation correction; Figure 4 A graph showing the deviation data obtained by the iterative deviation correction method for different channels and probes; Figure 5 Plots of probability density function and statistical peak index data for the background deviation OB before and after instrument bias correction; Figure 6 This is an array diagram of statistical indicators for instrument probe deviation. Detailed Implementation

[0014] The present invention will be further described below through specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0015] The present invention provides an iterative bias correction method considering the non-uniformity of satellite instrument detectors, comprising the following steps: like Figure 1 As shown, this embodiment includes the following steps: The constrained bias correction (CBC) method is used to limit the bias correction magnitude, and a cost function is constructed and initialized. A bias operator is constructed by combining atmospheric state dependence bias and detector bias; the atmospheric state dependence bias includes atmospheric state prediction factor and atmospheric quality bias correction BC coefficient. The probe bias is cleared to zero. The initial guess of the observed background deviation OB is calculated based on the original observed brightness temperature and the simulated brightness temperature of the background field. The atmospheric mass deviation correction coefficient BC is updated based on the initial guess of the observed background deviation OB. S4. The atmospheric state dependence bias is calculated using the updated atmospheric quality deviation correction BC coefficient. The atmospheric state dependence bias is subtracted from the atmospheric state dependence bias to obtain the residual of the observation background deviation OB as the probe bias. Calculate the background deviation OB for the next round of observations based on the probe deviation from the previous round. Repeat the probe deviation calculation steps and iterate multiple times until the atmospheric quality deviation correction BC coefficient is stable. Output the final atmospheric quality deviation correction BC coefficient and probe deviation to correct the deviation of the observation data.

[0016] like Figure 2 The figure shown is an iterative flowchart of an iterative deviation correction method for satellite instrument detectors that takes into account non-uniformity of the present invention.

[0017] In this embodiment, bias correction is carried out using the China Meteorological Administration Global Forecast System (CMA-GFS) model and the GIIRS instrument on the Fengyun-4B satellite as examples; First, the constrained bias correction (CBC) method is used to limit the bias correction magnitude. A cost function is constructed and initialized to quantify and minimize the deviation between model predictions and actual observations. To ensure that the cost function can work effectively in practical applications, the key parameters need to be accurately initialized. The core of this is to balance prior estimation information and observation information to obtain the optimal atmospheric state estimate. The cost function is specifically: ; in Let cost function be Background state The BC vector is used to correct for atmospheric quality deviation, and the BC coefficient is used to correct for atmospheric quality deviation. composition, To observe the background deviation OB, The original observed brightness temperature, To simulate brightness temperature for the background field, For the bias operator, The observation error covariance matrix, Prior estimation of probe bias The error covariance matrix, For regularization parameters; The cost function is initialized as follows: Prior estimation of probe bias Error covariance matrix Set to equal to the observation error covariance matrix This setting is based on the assumption that, in the initial stage, the error of the prior bias has similar statistical characteristics to the error of the observed data. This helps to provide a reasonable starting point for the model in the early stages of iteration and avoids convergence difficulties due to the imbalance of error weights. regularization parameters Setting it to 0.3, the regularization parameter is used to control model complexity and avoid overfitting. Its value determines the degree of influence of the bias correction term on the final result. 0.3 is an empirical value, which applies weak constraints to the model and can effectively correct bias while maintaining model stability. Meta-bias prior estimation The peak value of the observed background deviation OB is set as the initial offset. The OB distribution is the probability density function of the difference between the observed value and the background field (model prediction). Setting its peak value as the initial offset can ensure that the initial bias correction term can quickly capture the most significant systematic bias, thereby accelerating the convergence of the entire correction process. Then, a bias operator is constructed by combining atmospheric state-dependent bias and probe bias. It is a key bridge connecting observations and model biases. It not only considers biases related to atmospheric conditions (i.e., atmospheric condition-dependent biases), but also includes fixed biases of the instrument probe itself (probe biases). Atmospheric state-dependent bias is related to the physical state of the atmosphere itself (such as temperature, humidity, and air pressure). In different air masses, the difference between satellite observations and the model background field may vary systematically. In order to accurately describe atmospheric state-dependent bias, it is necessary to select appropriate predictors to characterize the current atmospheric environment. In this invention, drawing on the experience of the CMA-GFS four-dimensional variational assimilation system, four different air mass thicknesses at different altitudes were selected as predictors: 1000-300 hPa thickness, 200-50 hPa thickness, 50-10 hPa thickness, and 10-2 hPa thickness. These thickness values ​​can effectively reflect the vertical structure and energy distribution of the atmosphere and are powerful indicators for characterizing the atmospheric state. By associating these predictors with bias operators, the model can adaptively adjust the bias correction according to the real-time atmospheric state, thereby improving the accuracy and robustness of the correction. The deviation operator is specifically: ; ; in This is an atmospheric state-dependent bias. For probe bias, It is a constant for a given channel. These are atmospheric state predictors, derived from background states. Finally, ocean surface satellite observation data from clear-sky regions were selected (the atmospheric conditions in these regions are relatively stable, and the background field, sea surface temperature, and surface emissivity are more uniform). The cloud mask product of the Fengyun-4B geostationary orbital radiometric imager (AGRI) was used to filter the GIIRS clear-sky data, and the data was iteratively processed. Initialization: Clear the probe bias to zero, and calculate the observed background deviation OB based on the original observed brightness temperature and the simulated brightness temperature of the background field. , The initial guess for the observed background deviation OB is given; the procedure is to provide a clean starting point, and all deviations are temporarily regarded as atmospheric state deviations. Calculate the atmospheric quality deviation correction coefficients (BC coefficients): Based on the probability distribution of the observed background deviation (OB), the cost function is minimized using atmospheric state prediction factors with four different air mass thicknesses to calculate the four atmospheric quality deviation correction coefficients (BC coefficients). Bias Separation and Update: Atmospheric state predictors are derived from the background state, and atmospheric state dependence bias is calculated by combining the four atmospheric quality bias correction BC coefficients. The atmospheric state dependence bias is obtained by subtracting the atmospheric state dependence bias from the initial guess of the observed background deviation OB, which is then used as the probe bias (i.e., , For the first The probe deviation calculated by the round, For the first The atmospheric state-dependent bias of the next round of calculation is calculated by subtracting the detector bias of the previous round from the observed background deviation OB. , For the first (Observational background deviation OB calculated in rounds) Iteration and convergence judgment: Based on the observed background deviation OB calculated in this round and the probe deviation calculated in the previous round, update the atmospheric quality deviation correction BC coefficient and the probe deviation in this round. Repeat the above process until the atmospheric quality deviation correction BC coefficient is stable. When the algorithm is determined to have converged to the optimal solution, the final atmospheric quality deviation correction BC coefficient and probe deviation are output. The criterion for determining the stability of the atmospheric quality deviation correction BC coefficient is that the norm of the change in the atmospheric quality deviation correction BC coefficient is less than 1 / 1000. Figure 2 The process of calculating atmospheric state-dependent bias and Fengyun-4B satellite GIIRS probe bias based on iterative bias correction method is demonstrated. Figure 3 The changes of four atmospheric state deviation correction coefficients and probe deviation during the iteration process are shown. (a)-(d) represent thicknesses of 1000-300 hPa, 200-50 hPa, 50-10 hPa, and 10-2 hPa, respectively, and (e) represents the probe deviation. Figure 4 The iterative bias correction method is used to show the biases of different water vapor channels (including upper / middle / lower strata water vapor channels) and different probes. Figure 5 The diagrams show the comparison of probability density functions and statistical peak indices of the observed background deviation OB before and after instrument bias correction. (a)-(c) are the comparison of probability density functions of the observed background deviation OB of each water vapor channel before and after instrument bias correction, and (d) is the comparison of statistical peak indices of the observed background deviation OB of each water vapor channel before and after instrument bias correction. Figure 6 The normalized mean and standard deviation distribution characteristics of the detector bias of the Fengyun-4B satellite GIIRS instrument in different wavelength channels (long-wave infrared channel and mid-wave infrared channel) are shown in the 16×8 detector array. Once the iterative process converges, the final estimated values ​​of the atmospheric quality deviation correction BC coefficient and probe deviation for each channel's clear sky observation data are obtained. These estimated values ​​are used to perform deviation correction processing on the observation data to eliminate systematic deviations caused by the non-uniformity of the instrument probes. The data after deviation correction will be more accurate and reliable, thereby improving the quality of subsequent data assimilation, weather forecasting, or climate analysis.

[0018] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An iterative bias correction method considering non-uniformity of a satellite instrument's footprint, characterized in that, The method comprises the following steps: S1, limiting the bias correction range by using the constrained bias correction method, constructing a cost function and initializing the cost function; S2, constructing a bias operator by combining the atmospheric state dependent bias and the probe bias; the atmospheric state dependent bias comprises an atmospheric state prediction factor and an atmospheric quality bias correction BC coefficient; S3, clearing the probe bias, calculating the initial guess value of the observation background deviation O-B according to the original observed brightness temperature and the background field simulated brightness temperature, and updating the atmospheric quality bias correction BC coefficient according to the observation background deviation O-B; S4, calculating the atmospheric state dependent bias by using the updated atmospheric quality bias correction BC coefficient, subtracting the atmospheric state dependent bias from the initial guess value of the observation background deviation O-B to obtain the O-B residual as the probe bias; S5, calculating the observation background deviation O-B of the next round according to the probe bias of the last round, repeating the probe bias calculation step, and iterating multiple times until the atmospheric quality bias correction BC coefficient is stable, and outputting the final atmospheric quality bias correction BC coefficient and the probe bias to correct the observation data; The cost function is specifically: ； wherein is a cost function, is a background state, is an atmospheric mass bias corrected BC vector, where is an atmospheric mass bias corrected BC coefficient comprises, is an observation background departure O-B, is a raw observed brightness temperature, is a background field simulated brightness temperature, is a bias operator, is an observation error covariance matrix, is an error covariance matrix of a priori bias estimate of a sounding element , and is a regularization parameter; The bias operator is specifically: ； ； where is the atmospheric state dependent bias, is the probe bias, is constant for a channel, is the atmospheric state predictor derived from the background state.

2. The method according to claim 1, wherein, The method for initializing the cost function comprises: Taking probe bias priori estimation error covariance matrix equal to observation error covariance matrix , the regularization parameter is set to ; The peak of the observed background deviation O-B is obtained by the routine bias correction of the instrument, and the prior estimate of the probe bias is taken is equal to the peak of the observed background deviation O-B.

3. The method of claim 1, wherein the method further comprises: The atmospheric state prediction factor corresponds to the thickness of the air mass at different altitudes.

4. The method of claim 1, wherein the method further comprises: The condition for updating the atmospheric quality bias correction BC coefficient according to the observation background deviation O-B is that the cost function is minimum.

5. The method of claim 1, wherein the non-uniformity of the satellite instrument's footprint is considered in the iterative bias correction. The method for calculating the observation background deviation O-B of the next round according to the probe bias of the last round comprises: The initial guess value of the observation background deviation O-B is subtracted from the probe bias of the last round to obtain the observation background deviation O-B calculated in the next round, and the expression is: ； wherein is the observation background deviation O-B for the first wheel calculation, is the observation background deviation O-B for the first wheel calculation.

6. The method of claim 1, wherein the method further comprises: The determination criterion for the stability of the atmospheric quality bias correction BC coefficient is that the norm of the change of the atmospheric quality bias correction BC coefficient is less than a preset threshold.

Images