Ozone profile retrieval method based on thermal infrared remote sensing

Abstract

The application provides an ozone profile inversion method based on thermal infrared remote sensing and belongs to the technical field of atmospheric environment remote sensing. The technical scheme is as follows: S1, data collection and processing; S2, an ozone prior profile model; S3, a temperature and humidity profile error correction model; S4, thermal infrared ozone profile optimization inversion; and S5, determination of a proportional factor. The application has the beneficial effect that the stratospheric column total amount and the tropospheric column total amount are respectively used for constraint in the optimization inversion, so that the accuracy of the ozone profile, especially the accuracy of the lower troposphere, is improved, and the deficiency of the thermal infrared sensor in the nadir observation mode is made up.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of atmospheric environment remote sensing, and particularly relates to an ozone profile inversion method based on thermal infrared remote sensing. BACKGROUND

[0002] In the thermal infrared channel, the main signal source comes from temperature, and there is continuous absorption of water vapor, so that the temperature and humidity profile errors affect the radiation and the weight function of ozone in the inversion process, and further affect the inversion accuracy of the ozone profile. Figure 2 Based on the HIRAS channel sensitivity analysis, when the temperature profile is +1K, the humidity profile is Q-10%, and the O3 concentration is -10%, the radiance change at the HIRAS entrance pupil.As can be seen from the above table, the ozone absorption band is disturbed by the temperature and humidity profile. Figure 2

[0003] Figure 2 The HIRAS channel sensitivity analysis located near 9.6 μm, when the temperature profile is +1K, the humidity profile is Q-10%, and the O3 concentration is -10%, the radiance change at the HIRAS entrance pupil (based on the ECMWF profile sample library).

[0004] The thermal infrared adopts the brightness temperature difference to invert the gas concentration, and in the lower troposphere, the brightness temperature difference between the atmosphere and the ground is very small, and the radiation information available for inversion is very limited; secondly, the ultraviolet scattering spectrum is affected by factors such as the absorption line band dependence and the surface reflectivity, and the absorption of the tropospheric aerosol to the ultraviolet radiation often leads to underestimation of the total amount of ozone, and the error caused by the absorbing dust aerosol is greater than the inversion error caused by the non-absorbing aerosol, and the ultraviolet is limited to the vertical distribution information of ozone. The simple joint inversion cannot solve the inversion uncertainty caused by the surface reflectivity and the absorption line band dependence of the ultraviolet and the temperature error of the infrared, and the inversion model has a part of effect on improving the information freedom degree of the tropospheric ozone, but has little improvement on improving the inversion accuracy. Since the troposphere concentrates 10% of the ozone, the spaceborne sensor observed in the nadir mode is used to detect the ozone, whether in the ultraviolet band or in the infrared band, only the whole layer ozone and the ozone above the high layer of the troposphere is sensitive, thereby leading to great uncertainty of the tropospheric ozone inversion. SUMMARY

[0005] The present application aims to provide an ozone profile inversion method based on thermal infrared remote sensing, and this part is mainly based on the data of the thermal infrared sensor, and the optimization method is used to invert the ozone profile based on the stratospheric column total amount and the tropospheric column total amount respectively, and the optimization model is used to improve the accuracy of the ozone profile in the nadir observation mode, especially the accuracy of the middle and lower layers of the troposphere.

[0006] In order to achieve the above-mentioned purpose of the application, the technical scheme is specifically as follows: An ozone profile inversion method based on thermal infrared remote sensing, comprising the following steps:

[0007] Step (1): data collection and processing: obtaining satellite thermal infrared spectral data (for example, FY3D / E / F-HIRAS, NOAA20-CrIS), cloud product data, temperature and humidity profile data of reanalysis ERA5 data, tropopause data, total ozone column data and tropospheric ozone column data (for example, Sentinel5P-TROPOMI), removing clouds from the thermal infrared spectral data using cloud product data (MERSI cloud product), and performing spatiotemporal matching of the reanalysis data, tropopause data, total ozone column data and tropospheric column data based on the thermal infrared spectral data;

[0008] Thermal infrared hyperspectral data: the infrared hyperspectral detection instrument HIRAS is carried on the FY-3D (launched in 2017), FY-3E (launched in 2021) and FY-3F (planned to be launched in 2022) satellites. The HIRAS carried on the FY3D is the first infrared hyperspectral detection instrument of China's polar orbit meteorological satellite, which has three wavelength ranges of long-wave infrared (8.80-15.39 μm), medium-wave infrared (5.71-8.26 μm) and short-wave infrared (3.92-4.64 μm), and the spectral resolution and spatial resolution are 0.625 cm-1 and 16 km respectively. Currently, the first-order spectral data of FY3D-HIRAS can be downloaded from the National Satellite Meteorological Center (http: / / www.nsmc.org.cn / NSMC / Home / Index.html). The medium-resolution spectral imager MERSI realizes high-precision quantitative inversion of atmospheric, land and ocean parameters such as clouds, water vapor and land surface characteristics through continuous observation of micro-light and infrared band multi-spectrum.

[0009] Reanalysis data: ERA5 is a new generation of reanalysis data, which is built by Copernicus Climate Change Service (C3S) operated by the European Centre for Medium-Range Weather Forecasts (ECMWF). The spatial resolution is 62 km, and the time resolution is 1 hour. It can provide more accurate real-time atmospheric information for users. At present, the Integrated Global Radiosonde Archive (IGRA) is distributed in various regions of the world, with a total of 1500 stations, of which about 92 are in China. The detection can provide meteorological parameters including temperature, dew point temperature, pressure, potential height, tropopause, and ozone column concentration data, covering different pressure ranges from the surface to the lower stratosphere.

[0010] TROPOMI (TROPOspheric Monitoring Instrument) is a sensor carried on the Sentinel 5 (S5P, Sentinel-5 Precursor), which was successfully launched on October 13, 2017, and is expected to have a 7-year on-orbit life. The payload has global daily coverage detection capability in the ultraviolet-visible-near-infrared band. The spectral resolution of the ultraviolet band is 0.5 nm, and TROPOMI can provide high spatial resolution of tropospheric trace gases such as sample column concentration, tropospheric ozone column concentration, NO2, SO2, HCHO, and aerosol information.

[0011] Step (2): Ozone prior profile model: The ozone profile is influenced by many factors, and its changes have great uncertainty. In this study, the LSTM model is used to generate the ozone prior profile. LSTM is called long short-term memory network, which is the development and optimization of recurrent neural network (RNN). LSTM network improves the shortcomings of gradient disappearance and gradient explosion of RNN network, and is suitable for solving the problem of long time series. At the same time, LSTM network also solves the long dependence problem of RNN, by adding an adaptive "forget gate" in the network, so that the LSTM model can automatically learn and judge whether the previous information has been stored. This project uses the characteristics of LSTM to learn and model the complex nonlinear relationship, and establishes the relationship between various factors to establish the ozone profile prior estimation model: First, collect ERA5 meteorological data (temperature, humidity, potential vorticity, wind speed, wind direction, etc.), infrared and ultraviolet hyperspectral data near the ozone absorption band (use long time series of thermal infrared and ultraviolet hyperspectral data (such as AIRS and OMI on Aqua), ozone precursors and industrial production data, ozone data from lidar, sounding, ground stations, ERA5 reanalysis data; Second, unify the different data sets in space and time, and resample to the same vertical grid; use random forest method to analyze the importance of the collected data features, and get the important features of the model input; Finally, use LSTM to train and test the data, select the appropriate hidden layer and node number, get the optimal network structure, and get the best generalization ozone profile prior model.

[0012] Based on ground station data, sounding data, meteorological data, ozone precursor data, etc. to update the atmospheric background profile library, use deep learning model to get ozone prior profile and ozone prior covariance matrix, covariance matrix B represents the background field covariance matrix or the error covariance of the initial profile. The background covariance matrix is obtained by statistical analysis of the prior profile library:

[0013]

[0014] Where X is the profile in the profile library, is the statistical mean of the profile library.

[0015] Step (3): Temperature and humidity profile error correction model: The temperature and humidity profile correction factor is obtained by using the ground-based temperature and humidity profile data and the ERA5 data input into the radiation transfer model. According to the climate characteristics, the ERA5 temperature and humidity profile products and the IGRA ground-based data are statistically analyzed in different regions to obtain the temperature and humidity profile error at different heights. Then, based on the forward radiation transfer model RTTOV, the influence of temperature and humidity profile error on radiation is studied, and appropriate parameters are selected for correction to reduce the uncertainty of temperature and humidity profile error.

[0016] Step (4): Optimization of thermal infrared ozone profile inversion: The ozone absorption band at 9.6 μm in the thermal infrared absorption band is used for inversion. The column total amount of ozone and the column total amount of troposphere data are introduced to constrain the inversion results of tropospheric and stratospheric ozone in the inversion process, and the Levenberg-Marquardt (LM) model is used to invert the tropospheric ozone profile. The cost function with the introduction of the column total amount of ozone is redefined as:

[0017]

[0018] Where the vector X = X(O3) is the n-layer ozone concentration profile, the Y vector is the brightness temperature observed by the m-inversion channel satellite, γ is the introduced Lagrange multiplier, λ and ε are the proportional coefficients, TO3 and TO tropo are the column total amount of ozone and the column total amount of tropospheric ozone, and tropo represents the number of layers of the tropopause pressure. F(X) is the atmospheric forward radiation transfer model. The norm here is the Mahalanobis distance, so the cost function can be written as:

[0019]

[0020] R is the error covariance matrix of Y m ; B represents the initial value error covariance matrix of the sought variable O3 profile. After defining the cost function, the iteration formula is derived and the iteration formula is strengthened according to the Levenberg-Marquardt model. In the Levenberg-Marquardt method, the positive definite matrix and the γ factor are adjusted adaptively through the diagonal elements of the Hessian matrix and the nonlinearity;

[0021] According to the steepest descent principle, let J'(X) = 0 to obtain the linear solution form of this problem, that is:

[0022]

[0023] Where K is the first derivative of Y with respect to X, also known as the Jacobian matrix. Due to the nonlinearity of the inversion problem, Newton's nonlinear iteration can improve the inversion accuracy and convergence speed; that is, differentiating the cost function again yields:

[0024] 0.5J (X) = γB -1 +K T R -1 K (5)

[0025] J (X) is called the Hessian matrix. Substituting it into formula (4) yields:

[0026]

[0027] or:

[0028]

[0029] This method requires a relatively accurate estimate of the covariance of the background error and the setting of an appropriate Lagrange multiplier γ. The Lagrange multiplier is affected by human factors to some extent and inverts the variance of the error.

[0030] To address the two drawbacks of the Gauss-Newton iteration method: first, the Hessian matrix in the algorithm can become singular, making it impossible to determine successor points; second, non-singular Hessian matrices can also be non-positive definite, causing model invalidity. Many methods exist to solve the problem of non-positive definite Hessian matrices. The basic idea is to modify the matrix to construct a symmetric positive definite matrix. Based on the Levenberg-Marquardt method for solving least squares problems, a positive definite matrix is ​​constructed, letting...

[0031]

[0032] Here, H is a diagonal matrix, and w is redefined in each iteration to ensure the cost function decreases rapidly. When w is large, the steepest descent direction is reached; when w is small, the Gauss-Newton descent direction is followed. Therefore, the Levenberg-Marquardt (LM) method is the steepest descent method when the iteration point is far from the minimum, and the Gauss-Newton method when it is close to the minimum. These two methods complement each other; the former converges quickly when far from the minimum, and the latter converges quickly when close to the minimum. During the iteration process, w is the harmonizing factor between these two methods. After defining the cost function, this paper derives the iterative formula using the Gauss-Newton nonlinear iterative method and strengthens the constraints on the iterative formula based on the Levenberg-Marquardt model. In addition to strengthening the constraints on the iterative formula using the LM model, this section constrains the iterative results using the total stratosphere column and the total tropospheric column separately.

[0033] Step (5): Determination of scaling factors: The scaling factors λ and ε are determined by experimental methods. Existing ground-based, radiosonde, and lidar ozone profile data are used as benchmarks. The root mean square error and root mean square relative error of the ozone profiles obtained by different scaling factors in different regions compared with the benchmark values ​​are analyzed to obtain reasonable allocation of scaling factors λ and ε.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] 1. In the optimization inversion, this invention utilizes the total amount of the stratospheric column and the total amount of the tropospheric column to constrain the ozone profile, thereby improving the accuracy of the ozone profile, especially the accuracy of the lower troposphere, thus compensating for the shortcomings of thermal infrared sensors in the nadir observation mode.

[0036] 2. The present invention can effectively improve the accuracy and reliability of thermal infrared ozone profile inversion by constraining the total amount of tropospheric and stratospheric columns in the optimization inversion. At the same time, it is not sensitive to ozone in the lower troposphere by the nadir observation method, thus reducing the uncertainty of tropospheric ozone inversion.

[0037] 3. In traditional thermal infrared ozone profile inversion, information about the ozone profile in the lower troposphere often comes from prior knowledge, resulting in low inversion accuracy. However, constraining the total tropospheric column increases the degree of freedom of information in the troposphere, making the inversion results for the lower troposphere closer to reality. Constraining the stratosphere also significantly improves the accuracy and degree of freedom of the stratospheric inversion results. Therefore, combining stratospheric and tropospheric column constraints with thermal infrared observations can improve the accuracy of ozone profile inversion.

[0038] 4. In practical applications, the thermal infrared ozone profile obtained by constraining the total amount of ozone in the stratosphere and troposphere can provide more accurate information on the vertical distribution of ozone, providing important data support for atmospheric chemistry and climate research. Accurate ozone profiles are crucial for understanding atmospheric circulation, climate change, and the impact of human activities on the atmospheric environment. Furthermore, they are significant for understanding the horizontal and vertical transport of ozone between tropospheric regions, and provide high-precision, wide-range ozone profile information for PM2.5 and ozone pollution control. Attached Figure Description

[0039] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0040] Figure 1 This is a schematic diagram illustrating the detailed inversion results of the Legionowo station in the mid-latitude region of the Northern Hemisphere on July 4, 2019, in an embodiment of the present invention.

[0041] (a) is a schematic diagram of the inversion results with total column constraints at the Legionowo site; (b) is a schematic diagram of the inversion results without constraints at the Legionowo site.

[0042] Figure 2 This is a schematic diagram of the process for an ozone profile inversion method based on thermal infrared remote sensing provided by the present invention.

[0043] Figure 3 This is a schematic diagram of the prior art based on HIRAS channel sensitivity analysis described in the background section of this invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] Example 1

[0046] See Figures 2 to 3 This embodiment provides a technical solution: an ozone profile inversion method based on thermal infrared remote sensing. This embodiment compares the results with those obtained after constraining the total ozone column in the troposphere and stratosphere with the unconstrained method. The comparison results are as follows. Figure 1 As shown: Figure 1 The inversion results for the Legionowo site with total column volume constraints (a) and without constraints (b) are shown in the left figure. The comparison of radiosonde data, smoothed radiosonde data, prior profile and inversion results is shown in the middle figure. The average kernel function of the inversion results at different pressures is shown in the right figure.

[0047] Figure 1 This section details the inversion results for selected days at the Legionowo site in the mid-latitudes of the Northern Hemisphere. Before constraint, the inversion results at pressures >500 hPa at the Legionowo site largely originated from prior profiles. After column total constraint, the accuracy of the inversion and smoothing results improved by approximately 15% compared to before constraint. However, the accuracy of the inversion results with column total constraint at 200-300 hPa decreased slightly by approximately 12% compared to before constraint. Furthermore, the maximum peak value of the average kernel function after column total constraint at 200-390 hPa was approximately 0.98, while the maximum peak value before constraint was approximately 1.2. The maximum peak value of the average kernel function from 103 hPa to near-surface was at 200-300 hPa, with the maximum peak value after column total constraint being 2.5, compared to 3 before constraint. However, the kernel function improved after column total constraint at other pressures. This improvement is due to the scaling factor after column total constraint, which enhanced the inversion accuracy of the entire tropospheric ozone profile.

[0048] 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 ozone profile inversion method based on thermal infrared remote sensing, characterized in that, The method comprises the following steps: S1, data collection and processing: obtaining satellite thermal infrared spectral data, cloud product data, temperature and humidity profile data of reanalysis ERA5 data, tropopause data, total column ozone data and tropospheric column ozone data, removing clouds from the thermal infrared spectral data using the cloud product data, and performing temporal and spatial matching of the reanalysis data, tropopause data, total column ozone and tropospheric column data based on the thermal infrared spectral data; S2, ozone prior profile model: updating the atmospheric background profile library based on ground station data, sounding data, meteorological data and ozone precursor data, obtaining the ozone prior profile and the ozone prior covariance matrix by using a deep learning model, and the covariance matrix B representing the background field covariance matrix or the error covariance of the initial profile, and the background covariance matrix being obtained by statistical analysis of the prior profile library: where X is a profile in the profile library, is the statistical mean of the profile library. S3, temperature and humidity profile error correction model: correcting the temperature and humidity profile by using the ground temperature and humidity profile data and the ERA5 data input into the radiation transfer model to obtain correction factors: analyzing according to the climate characteristics, statistically analyzing the input ERA5 temperature and humidity profile product and the IGRA ground data in different regions to obtain the temperature and humidity profile error at different heights, and then studying the influence of the temperature and humidity profile error on the radiation based on the forward radiation transfer model RTTOV to correct the radiation quantity and reduce the uncertainty of the temperature and humidity profile error; S4, thermal infrared ozone profile optimization inversion: using the absorption band of 9.6 μm of ozone in the thermal infrared absorption band to perform inversion, constraining the tropospheric and stratospheric ozone inversion results in the inversion process by introducing the total column ozone and tropospheric column data, and using the Levenberg-Marquardt model to invert the tropospheric ozone profile; S5, determination of the proportion factor: the proportion factors λ and ε are determined by using the existing ground, sounding and lidar ozone profile data as the benchmark, analyzing the root mean square error and root mean square relative error of the ozone profile obtained by different proportion factors in different regions, and then obtaining reasonable distribution proportion factors λ and ε.

2. The ozone profile retrieval method based on thermal infrared remote sensing according to claim 1, characterized in that, In the step S4, the cost function with the total column ozone is redefined as: where vector X=X(O3) is the n-layer ozone concentration profile, Y is the vector of brightness temperatures observed by m inversion channel satellites, γ is the introduced Lagrange multiplier, λ and ε are the proportional coefficients, TO3 and TO tropo are the total column amounts of ozone and tropospheric ozone, respectively, tropo represents the number of layers of the tropopause pressure, F(X) is the atmospheric forward radiative transfer model, and the norm is the Mahalanobis distance. The cost function is then written as: R is Y m The error covariance matrix contains the error caused by the instrument and system. B represents the initial error covariance matrix of the O3 profile. After defining the cost function, the iteration formula is derived and the constraints are strengthened according to the Levenberg-Marquardt model. In the Levenberg-Marquardt method, the positive definite matrix and the gamma factor are adaptively adjusted by the diagonal elements of the Hessian matrix and the nonlinearity. According to the steepest descent principle, let J'(X) = 0 to obtain the linear solution form of the problem, that is: Wherein, K is the first-order derivative of Y with respect to X, also known as the Jacobian matrix, since the inversion problem is nonlinear, the Gauss-Newton iteration method is used to improve the inversion accuracy and convergence speed of the problem, that is, the cost function is differentiated again to obtain: 0.5J" (X) = γB -1 +K T R -1 K (5); J”(X) is called the Hessian matrix, which is brought into formula (4) to obtain: Or: This method needs to give an accurate estimate of the background error covariance and set a suitable Lagrange multiplier γ, and the Lagrange multiplier is affected by human factors and the variance of the inversion error.

3. The ozone profile retrieval method based on thermal infrared remote sensing according to claim 2, characterized in that, In the Gauss-Newton iteration method, according to the method for solving the least square proposed by Levenberg-Marquardt, a positive definite matrix is constructed, and Wherein, H is a positive diagonal matrix, which is determined again in each iteration process, so that the cost function rapidly decreases.

4. The ozone profile retrieval method based on thermal infrared remote sensing according to claim 1, characterized in that, In step S1, thermal infrared hyperspectral data is obtained from the HIRAS infrared hyperspectral instrument carried on the FY-3D, FY-3E, and FY-3F satellites. The HIRAS instrument on FY3D is the first infrared hyperspectral instrument on a polar-orbiting meteorological satellite, possessing three wavelength ranges: long-wave infrared (8.80-15.39 μm), mid-wave infrared (5.71-8.26 μm), and short-wave infrared (3.92-4.64 μm), with a spectral resolution and spatial resolution of 0.625 cm⁻¹. -1 And 16km.

5. The ozone profile retrieval method based on thermal infrared remote sensing according to claim 1, characterized in that, In the step S1, the reanalysis data ERA5 is analyzed. The ERA5 is a new generation of reanalysis data, with a spatial resolution of 62 km and a time resolution of 1 hour, which provides accurate real-time atmospheric information for users. The meteorological parameters provided include temperature, dew point temperature, air pressure, potential height, tropopause, and ozone column concentration data, covering different pressure ranges from the ground to the lower stratosphere.

Images