TROPOMI on-orbit calibration and performance tracking method based on background value statistics

Abstract

The invention discloses a TROPOMI on-orbit calibration and performance tracking method based on background value statistics, which comprises the following steps: constructing a pure ocean background value data set, eliminating interference pixels by utilizing multiple physical thresholds such as cloud screening, solar flare screening and absorptive aerosol screening, and reducing errors in combination with geometric constraint conditions to obtain a TROPOMI background value; an empirical model of background values and observation geometry is established, angle dependence of the background values is eliminated through geometric normalization processing, then a long-term background value sequence is established, and a time sequence decomposition technology (STL) is adopted to separate seasonal fluctuation of a time sequence and a long-term attenuation trend of an instrument; finally, dynamic prediction and correction of a full-view-field background value of the sensor can be realized based on the constructed integrated space-time model, and a time consistency correction coefficient is calculated so as to eliminate long-term deviation introduced by instrument performance attenuation. According to the invention, the long-term consistency and reliability of atmospheric component products such as absorptive aerosol index and the like are obviously improved.

Description

Technical Field

[0001] This invention belongs to the field of satellite remote sensing technology, specifically relating to a method for on-orbit radiometric performance monitoring, calibration, and data consistency correction of spaceborne ultraviolet hyperspectral instruments, which is particularly suitable for monitoring the performance degradation of optical sensors used for atmospheric composition detection, such as TROPOMI. Background Technology

[0002] In recent years, the on-orbit performance tracking and calibration of spaceborne optical sensors has attracted widespread attention from researchers. TROPOMI, as a core payload of the Sentinel-5P satellite, has already provided crucial data support for global air quality monitoring and climate change research.

[0003] The Absorbing Aerosol Index (AAI) product generated by TROPOMI is an important indicator for monitoring absorbing aerosols such as urban haze, smoke, and desert dust. Its long-term data can also provide more comprehensive aerosol optical parameters for atmospheric composition monitoring. However, during long-term operation in orbit, TROPOMI's radiation performance and calibration accuracy will decline due to factors such as mechanical vibration, space radiation (such as ultraviolet radiation, atomic oxygen, and proton radiation), and degradation of optical components (such as optical fibers and solar diffuse reflectors). Therefore, it is necessary to conduct full life-cycle on-orbit performance tracking to ensure the reliability of AAI data.

[0004] Current mainstream on-orbit calibration methods each have their limitations: some satellites (such as SPOT) carry onboard calibration equipment, but this equipment is prone to degradation and cannot cover the entire field of view due to limitations in conditions. Furthermore, ground calibration before launch is also limited by laboratory conditions. Ground-based pseudo-invariant calibration field (PICS) methods (such as those using desert areas) rely on ground measurement data; Vankempen et al. used this method to verify the stability of the Tropomi shortwave infrared module, but this method is highly dependent on ground references. Cross-calibration methods are widely used; for example, Wang et al. combined synchronous nadir transit (SNO) and double difference (DD) methods to perform UV-Vis spectral radiometric cross-calibration of EMI and Tropomi, but this requires a reference payload with well-defined performance. Ocean flare methods are greatly affected by atmospheric variability; cloud cover, sea surface wind speed, etc., can alter flare reflection characteristics, leading to uncertainties in calibration accuracy. In summary, existing technical solutions are either limited by the stability and coverage of the calibration equipment itself, or rely on external (ground or other satellite) reference references and specific observation conditions. These limitations make it difficult to establish an on-orbit calibration method that does not rely on dedicated on-board equipment, requires no external stable reference target, and can achieve continuous, autonomous, and full-field-of-view performance tracking. Developing a novel autonomous on-orbit performance tracking and calibration technology for long-term operational atmospheric monitoring payloads such as Tropomi is of great significance for ensuring the long-term quality and application value of their data products. Summary of the Invention

[0005] To overcome the shortcomings of existing on-orbit calibration methods that rely on external equipment or reference targets and struggle to achieve continuous autonomous full-field-of-view performance tracking, this invention provides an on-orbit calibration and performance tracking method for TROPOMI based on background value statistics. This method addresses the systematic bias in the Absorbing Aerosol Index (AAI) data caused by the attenuation of radiation performance during the long-term on-orbit operation of the TROPOMI Sentinel-5P satellite core payload. Simultaneously, it overcomes the limitations of existing calibration methods that rely on onboard equipment, external references, or ground measurements. This method enables autonomous tracking, attenuation quantification, and AAI data correction of TROPOMI's full-field-of-view on-orbit performance, ultimately ensuring the long-term accuracy and consistency of its L2-level AAI products. Furthermore, it provides a technical reference for the stability assessment of similar atmospheric composition detection payloads, ultimately guaranteeing the long-term consistency and reliability of remote sensing data products.

[0006] The proposed technical solution to address the above-mentioned technical problems is as follows:

[0007] A method for on-orbit calibration and performance tracking of TROPOMI based on background value statistics includes the following steps: Step 1, constructing a clean ocean background value data level: acquiring L1b radiation data and L2-level AAI products and auxiliary parameters of the target sensor (TROPOMI) over a preset Pacific region, and filtering pixel by pixel by setting multiple physical thresholds to eliminate non-ideal observations;

[0008] Step 2: From a spatial perspective, based on the spatial distribution pattern of background values ​​and combined with the observation geometry, an empirical model is established between background values ​​and the observation geometry in the track crossing direction to quantitatively describe the influence of observation geometry on background values; In the time dimension, background values ​​can be used as a qualitative indicator reflecting the on-orbit operating status of the instrument; Based on its time series characteristics, a seasonal trend decomposition method based on local weighted regression is adopted to extract the seasonal variation component and long-term trend component of background values, thereby revealing the variation characteristics at different time scales caused by changes in the external environment and the evolution of instrument performance.

[0009] Step 3: In the spatial dimension, by correcting the observation geometric effects, the abnormally high AAI in the right orbital region of TROPOMI can be effectively eliminated; in the temporal dimension, the performance degradation effect of the TROPOMI instrument over time can be corrected, thereby improving the long-term stability and consistency of the observation data.

[0010] Furthermore, the process of step 1 is as follows:

[0011] 1.1 Cloud Screening: The cloud score threshold was set to 0.03, and the measured reflectance R of the sensor in the 340nm or 380nm band was used. measClear-sky aerosol-free reflectivity R compared with the SCIATRAN radiative transfer model model The ratio is used to assist in the judgment and filter out cloud pixels that meet the conditions;

[0012] 1.2 Solar flare screening: Based on the calculated flare angle threshold, according to formula (1):

[0013] cos(ΔΩ glint )=cosθ0·cosθ+sinθ0·sinθ·cos(φ-φ0) (1);

[0014] Where θ0, θ, φ-φ0 are the solar zenith angle, the observed zenith angle, the observed azimuth angle, and the solar azimuth angle, respectively, and ΔΩ glint For the flare angle that needs to be calculated, pixels with a flare angle of less than 18° over the ocean are identified as solar flare pixels;

[0015] 1.3 Screening of Absorbing Aerosols: Based on the AAI product threshold, pixels with an Absorbing Aerosol Index (AAI) greater than 1 are considered to be absorbing aerosols.

[0016] By comprehensively applying the above screening and identification conditions, only pure ocean pixels that were not identified as clouds, solar flares, or absorbing aerosols were retained. Their observation values ​​were statistically averaged monthly to construct a monthly background value time series. 1.4 Geometric constraints: The latitude range of the pixels was limited to 60°S~60°N, and the solar zenith angle θ0 was limited to less than 85° to reduce geometric errors and data spatial non-uniformity caused by extreme observation angles. In order to suppress short-term fluctuations and minimize the impact of seasonal changes, a monthly background value dataset was constructed and used for subsequent analysis.

[0017] The process of step 2 is as follows:

[0018] 2.1 Establishment of Space Empirical Model: Background value data were selected during the initial stage of sensor operation and the period when performance degradation was negligible (such as the first year after launch). A polynomial empirical model of background value and observation geometric parameters was constructed to quantify the spatial distribution law of background value in the sensor scanning direction. According to formula (2):

[0019]

[0020] Among them, bg geoThe background values ​​are defined as those related to the observed geometry, representing the degree of influence of geometric factors on the background values. a0, a1, a2, and a3 are model fitting coefficients, obtained by fitting monthly background value data using the least squares method. These coefficients describe the functional relationship between background values ​​and pixel positions (i.e., observed geometry). The coefficient values ​​for different months reflect subtle changes in the spatial pattern. The core purpose of this model is to describe the variation of background values ​​with observed geometry, providing a foundation for subsequent elimination of viewpoint dependence and correction of spatial distribution biases.

[0021] 2.2 Geometric Normalization Processing: According to formula (2), based on the geometric dependency obtained from the fitting, bg geo To eliminate the influence of changes in observation angle on background values, the original background value of each pixel is normalized to a set of unified reference observation geometry conditions: solar zenith angle θ0 = 30°, observation zenith angle θ = 0°, relative azimuth angle φ - φ0 = 90°, and latitude Lat = 0°. The normalization calculation formula is as follows:

[0022] bg′=bg true -bg geo +bg * (3);

[0023] Where bg' is the background value after removing geometric effects; bg ture bg* represents the background value measured by the sensor, while bg* represents the background value under reference geometric conditions.

[0024] The representative background value for each month is obtained by spatial averaging the geometrically corrected background value bg', as shown in the following formula:

[0025]

[0026] Among them, bg″ ave (t m ) represents the tth m Monthly mean of geometrically normalized background values ​​for the past month; t m N is a time variable, representing a specific month; N is the t-th month. m The number of clean ocean pixels used in the calculation over a month; bg i ′(t m ) is the t-th m The geometrically normalized background value (bg″) of the i-th pixel in a month is obtained by averaging the bg' calculated from all clean ocean pixels in that month. ave (t m ), forming a time series;

[0027] 2.3 Time Series Decomposition: The monthly time series after geometric normalization is still affected by seasonal variations and instrument degradation. Therefore, it is necessary to further decompose the time series to separate seasonal fluctuations from the long-term instrument degradation effect and achieve time consistency correction of background values. Using a time series decomposition method, it is decomposed into three components. The monthly mean series is expressed as follows:

[0028] bg″ ave (t m )=G(t m )+S(t m )+ε(t m (5);

[0029] Wherein, S(t) m ), G(t) m ), ε(t) m The components are denoted as seasonal variation component, instrument gain term, and random disturbance term, respectively. The seasonal variation component S(t) is represented by... m ) and instrument gain term G(t m Decoupling can effectively identify seasonal effects and extract long-term trends, thereby achieving time correction of instrument gain changes; the seasonal trend decomposition (STL) method is used to perform periodic analysis on the monthly mean series to obtain S(t m The seasonal variation component is then removed from the monthly mean series, allowing the derivation of the long-term instrument gain term G(t). m Its linear fitting expression is as follows:

[0030]

[0031] Where a0 is the linear fitting constant term, c, c·t m c·t represents the linear rate of change over time and the long-term trend term. m The magnitude and sign of the value directly characterize the decay rate and direction of the instrument's radiation response; the annual average decay rate of the sensor in a specific band (such as the 340nm ultraviolet band) is calculated based on the slope c, so as to achieve quantitative assessment and tracking of the degree of instrument performance decay.

[0032] The process of step 3 is as follows:

[0033] 3.1 Based on the spatial empirical model established in step 2.1 and the trend components extracted in step 2.3, a spatiotemporal integrated model is constructed to predict the background value at any target time. Combined with the initial background value of the load, the formula is as follows:

[0034]

[0035]

[0036] in, G(t) is the gain factor. initial ) and G(t terminal The first two digits represent the trend components for the initial period (2018) and the target forecast period, respectively. and Then, S(t) represents the predicted background value in the initial period and the target time frame, respectively. terminal ) is the seasonal component of the target time frame;

[0037] 3.2 Time Consistency Correction: To eliminate long-term systematic biases in data introduced by the degradation of sensor performance over time, and to restore and ensure the consistency and comparability of the entire observation data archive over long time scales, a relative correction coefficient can be calculated at each time point based on the initial trend value at the beginning of the observation period.

[0038]

[0039] The coefficient c λ,t It is applied to deseasonalize sequences to correct temporal inconsistencies.

[0040] This invention fundamentally overcomes the dependence of traditional methods on external calibration equipment or specific reference targets, and realizes sustainable and full-field-of-view instrument performance degradation monitoring and quantification that can be completed using only the sensor's own operational observation data, providing a new technical path for ensuring the long-term consistency of remote sensing AAI data products.

[0041] The technical concept of this invention is to use the global ocean, a stable and homogeneous natural target, as a radiation reference source. By constructing and analyzing the spatiotemporal variation patterns of background value indicators in sensor observation data, an empirical model is established that is consistent with the observation geometry and time. The core innovation of this method lies in the decoupling of instrument attenuation signals from geographical / seasonal signals, thereby enabling the direct extraction of trend components reflecting changes in sensor performance and generating spatial and temporal parameters for correcting systematic biases in product-level data.

[0042] The effective effects of this invention are: significantly improving the accuracy and long-term comparability of atmospheric composition products. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the method.

[0044] Figure 2 This is a distribution map of background values ​​as a function of latitude zone and scanned pixels in the embodiment;

[0045] Figure 3 This example shows a comparison and regression analysis of the background values ​​predicted by the model with the actual statistical values.

[0046] Figure 4The time series decomposition (STL) results shown in the example illustrate the trend and seasonal components.

[0047] Figure 5 This is a comparison chart of the changes in the absorbable aerosol index along the track crossing direction before and after background value correction in the example.

[0048] Figure 6 This is a correlation comparison chart between the corrected TROPOMI AAI and GOME-2AAI in the examples;

[0049] Figure 7 This is a comparison chart of the linear fitting and time consistency correction effects of the background value trend term in the embodiment. Detailed Implementation

[0050] The invention will now be further described with reference to the accompanying drawings.

[0051] Reference Figures 1-7 A method for on-orbit calibration and performance tracking of TROPOMI based on background value statistics includes the following steps:

[0052] Step 1, constructing pure ocean background data level: acquire L1b radiation data and L2-level AAI products and auxiliary parameters of the target sensor (TROPOMI) over the preset Pacific region, and filter pixel by pixel by setting multiple physical thresholds to eliminate non-ideal observations.

[0053] The process of step 1 in this embodiment is as follows:

[0054] 1.1 Cloud Screening: The cloud score threshold is set to <0.03, and the measured reflectance R of the sensor in the 340nm or 380nm band is used. meas Clear-sky aerosol-free reflectivity R compared with the SCIATRAN radiative transfer model model The ratio is used to assist in the judgment and filter out cloud pixels that meet the conditions;

[0055] 1.2 Solar flare screening: Based on the calculated flare angle threshold, according to Formula 1:

[0056] cos(ΔΩ glint )=cosθ0·cosθ+sinθ0·sinθcos(φ-φ0) (1);

[0057] Where θ0, θ, φ-φ0 are the solar zenith angle, the observed zenith angle, the observed azimuth angle, and the solar azimuth angle, respectively, and ΔΩ glint For the flare angle that needs to be calculated, pixels with a flare angle of less than 18° over the ocean are identified as solar flare pixels;

[0058] 1.3 Screening of Absorbing Aerosols: Based on the AAI product threshold, pixels with an Absorbing Aerosol Index (AAI) greater than 1 are considered to be absorbing aerosols.

[0059] By comprehensively applying the above screening and identification conditions, only pure ocean pixels that were not identified as clouds, solar flares, or absorbing aerosols were retained. Their observation values ​​were statistically averaged monthly to construct a monthly background value time series. 1.4 Geometric constraints: The latitude range of the pixels was limited to 60°S~60°N, and the solar zenith angle θ0 was limited to less than 85° to reduce geometric errors and data spatial non-uniformity caused by extreme observation angles. In addition, in order to suppress short-term fluctuations and minimize the impact of seasonal changes, this study constructed and used a monthly background value dataset for subsequent analysis.

[0060] Step 2, modeling the spatiotemporal features of the background values, the process is as follows:

[0061] 2.1 Establishment of Space Empirical Model: Background value data were selected during the initial stage of sensor operation and the period when performance degradation was negligible (such as the first year after launch). A polynomial empirical model of background value and observation geometric parameters was constructed to quantify the spatial distribution law of background value in the sensor scanning direction. According to formula (2):

[0062]

[0063] Among them, bg geo The background values ​​are defined as geometrically relevant, representing the degree of influence of geometric factors on the background values. a0, a1, a2, and a3 are model fitting coefficients, obtained by fitting monthly background value data using the least squares method. These coefficients describe the functional relationship between background values ​​and pixel positions (i.e., observation geometry). The coefficient values ​​for different months reflect subtle changes in the spatial pattern. The core purpose of this model is to describe the variation of background values ​​with observation geometry, providing a foundation for subsequent elimination of viewpoint dependence and correction of spatial distribution biases.

[0064] 2.2 Geometric Normalization Processing: According to formula (2), based on the geometric dependency obtained from the fitting, bg geo To eliminate the influence of changes in observation angle on background values, the original background value of each pixel is normalized to a set of unified reference observation geometry conditions: solar zenith angle θ0 = 30°, observation zenith angle θ = 0°, relative azimuth angle φ - φ0 = 90°, and latitude Lat = 0°. The normalization calculation formula is as follows:

[0065] bg′=bg true -bg geo +bg * (3);

[0066] Where bg' is the background value after removing geometric effects; bgture bg* represents the background value measured by the sensor, while bg* represents the background value under reference geometric conditions.

[0067] The representative background value for each month is obtained by spatial averaging the geometrically corrected background value bg', as shown in the following formula:

[0068]

[0069] Among them, bg″ ave (t m ) represents the tth m Monthly mean of geometrically normalized background values ​​for the past month; t m N is a time variable, representing a specific month; N is the t-th month. m The number of clean ocean pixels used in the calculation over a month; bg i ′ (t m ) is the t-th m The geometrically normalized background value of the i-th pixel in a given month. The geometrically normalized background value bg″ for that month is obtained by averaging the bg' calculated from all clean ocean pixels for that month. ave (t m ), forming a time series.

[0070] 2.3 Time Series Decomposition: The monthly time series after geometric normalization is still affected by seasonal variations and instrument degradation. Therefore, it is necessary to further decompose the time series to separate seasonal fluctuations from the long-term instrument degradation effect and achieve time consistency correction of background values. Using a time series decomposition method, it is decomposed into three components. The monthly mean series is expressed as follows:

[0071] bg″ ave (t m )=G(t m )+S(t m )+ε(t m (5);

[0072] Where S(t) m ), G(t) m ), ε(t) m The terms S(t) are represented as seasonal variation component, instrument gain term, and random disturbance term, respectively. In this study, the seasonal variation component S(t) is used to represent the seasonal variation component. m ) and instrument gain term G(t m Decoupling can effectively identify seasonal effects and extract long-term trends, thereby enabling time correction of instrument gain changes. Therefore, the Seasonal Trend Decomposition (STL) method is used to perform periodic analysis on the monthly mean series to obtain S(t). m The seasonal variation component is then removed from the monthly mean series, allowing the derivation of the long-term instrument gain term G(t).m Its linear fitting expression is as follows:

[0073]

[0074] Where a0 is the linear fitting constant term, c, c·t m c·t represents the linear rate of change over time and the long-term trend term. m The magnitude and sign of the value directly characterize the decay rate and direction of the instrument's radiation response. Based on the slope c, the average annual decay rate of the sensor in a specific wavelength band (such as the 340nm ultraviolet band) can be calculated, enabling quantitative assessment and tracking of the degree of instrument performance degradation.

[0075] Step 3, correct the observation data, the process is as follows:

[0076] 3.1 Based on the spatial empirical model established in step 2.1 and the trend components extracted in step 2.3, a spatiotemporal integrated model is constructed to predict the background value at any target time. Combined with the initial background value of the load, the formula is as follows:

[0077]

[0078] in, G(t) is the gain factor. initial ) and G(t terminal The two numbers represent the trend components for the initial period (2018) and the target forecast period, respectively. and Then, S(t) represents the predicted background value in the initial period and the target time frame, respectively. terminal ) is the seasonal component of the target time frame.

[0079] 3.2 Time Consistency Correction: To eliminate long-term systematic biases in data introduced by the degradation of sensor performance over time, and to restore and ensure the consistency and comparability of the entire observation data archive over long time scales, a relative correction coefficient can be calculated at each time point based on the initial trend value at the beginning of the observation period.

[0080]

[0081] The coefficient c λ,t It is applied to deseasonalize sequences to correct temporal inconsistencies.

[0082] Based on the validation results of the proposed on-orbit calibration and performance tracking method for TROPOMI background values, Figures 2 to 6 The results of this method's practical application on the TROPOMI dataset are presented. Figure 2 It demonstrates the relationship between background values ​​and latitude, and ground pixels; Figure 3The results of the background value prediction based on the integrated spatiotemporal model are shown, which are highly consistent with the actual statistical values ​​(R = 0.9988). Figure 4 The STL decomposition method was used to separate the trend and seasonal components from the time series, and the model showed excellent fitting performance (R²). 2 =0.95, RMSE=0.02); Figure 5 The results show that, after correction for actual statistics and predicted background values, the high values ​​of orbital edge anomalies caused by observation geometry in the TROPOMI AAI data can be significantly suppressed. Figure 6 This further verified the consistency between the TROPOMI AAI and GOME-2 AAI data after background value correction (R increased from 0.81 to 0.83); Figure 7 By linearly fitting the background trend component, the annual attenuation rate of the instrument in the ultraviolet band (approximately 14.51%) was quantified, and the effectiveness of time consistency correction in suppressing long-term drift was demonstrated.

[0083] The background-based on-orbit calibration and performance tracking method in this embodiment can achieve long-term stable monitoring and accurate quantification of the performance of the TROPOMI sensor, effectively correct data anomalies caused by instrument attenuation and observation geometric coupling, and significantly improve the spatial consistency and temporal comparability of data products.

[0084] The effectiveness of this method has been verified by actual data, but the application of this invention is not limited to the above embodiments. Adaptive adjustments and extensions can be made to this method without departing from the core ideas and technical scope of this invention. The technical framework proposed in this invention is applicable to various Earth observation optical sensors, especially atmospheric composition detection payloads with high requirements for radiation stability, and can provide reliable technical support for on-orbit performance tracking and data quality control of similar payloads.

[0085] The embodiments described in this specification are merely examples of implementations of the inventive concept and are for illustrative purposes only. The scope of protection of this invention should not be considered limited to the specific forms described in these embodiments; rather, it extends to equivalent technical means conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for on-orbit calibration and performance tracking of TROPOMI based on background value statistics, characterized in that, The method includes the following steps: Step 1, construct pure ocean background data level: acquire L1b radiation data and L2-level AAI products and auxiliary parameters of the target sensor TROPOMI over the preset Pacific region, and filter pixel by pixel by setting multiple physical thresholds to eliminate non-ideal observations; Step 2: From a spatial perspective, based on the spatial distribution pattern of background values ​​and combined with the observation geometry, an empirical model is established between background values ​​and the observation geometry in the track crossing direction to quantitatively describe the influence of observation geometry on background values. In the time dimension, background values ​​serve as a qualitative indicator reflecting the on-orbit operating status of the instrument. Based on its time series characteristics, a seasonal trend decomposition method based on local weighted regression is adopted to extract the seasonal variation component and long-term trend component of background values, thereby revealing the variation characteristics at different time scales caused by changes in the external environment and the evolution of instrument performance. Step 3: In the spatial dimension, by correcting the observation geometric effects, the abnormally high AAI in the right orbital region of TROPOMI can be effectively eliminated; in the temporal dimension, the performance degradation effect of the TROPOMI instrument over time can be corrected, thereby improving the long-term stability and consistency of the observation data.

2. The TROPOMI on-orbit calibration and performance tracking method based on background value statistics as described in claim 1, characterized in that, The process of step 1 is as follows: 1.1 Cloud Screening: The cloud score threshold was set to 0.03, and the measured reflectance R of the sensor in the 340nm or 380nm band was used. meas Clear-sky aerosol-free reflectivity R compared with the SCIATRAN radiative transfer model model The ratio is used to assist in the judgment and filter out cloud pixels that meet the conditions; 1.2 Solar flare screening: Fibers with a flare angle of less than 18° over the ocean are identified as solar flare pixels; 1.3 Screening of Absorbing Aerosols: Based on the AAI product threshold, pixels with an Absorbing Aerosol Index (AAI) greater than 1 are considered to be absorbing aerosols. By comprehensively applying the above screening and identification conditions, only pure ocean pixels that were not identified as clouds, solar flares, or absorbing aerosols were retained, and their observation values ​​were statistically averaged monthly to construct a monthly background value time series. 1.4 Geometric Constraints: The latitude range of the pixels is limited to 60°S~60°N, and the solar zenith angle θ0 is limited to less than 85°. A background value dataset at a monthly scale is constructed and used for subsequent analysis.

3. The TROPOMI on-orbit calibration and performance tracking method based on background value statistics as described in claim 1 or 2, characterized in that, The process of step 2 is as follows: 2.1 Establishment of Space Empirical Model: Background value data were selected during the initial stage of sensor operation and the period when performance degradation was negligible. A polynomial empirical model of background value and observation geometric parameters was constructed to quantify the spatial distribution law of background value in the sensor scanning direction. 2.2 Geometric Normalization Processing: Based on the geometric dependency relationship bg obtained from the fitting. geo To eliminate the influence of changes in the observation angle on the background value, the original background value of each pixel is normalized to a set of unified reference observation geometry conditions, with solar zenith angle θ0 = 30°, observation zenith angle θ = 0°, relative azimuth angle φ-φ0 = 90°, and lat = 0°. The background value bg after geometric correction each month ’ Spatial averaging is performed to obtain the representative background value for the month. The average value of bg' calculated for all clean ocean pixels each month is then taken to obtain the geometrically normalized background value bg″ for that month. ave (t m ), forming a time series; 2.3 Time Series Decomposition: The time series decomposition method is used to decompose it into three components. A seasonal trend decomposition method is then used to perform periodic analysis on the monthly mean series, yielding S(t). m The seasonal variation component is then removed from the monthly mean series, allowing the derivation of the long-term instrument gain term G(t). m The average annual attenuation rate of the sensor in a specific band is calculated based on the linear fitting slope c, thereby enabling quantitative assessment and tracking of the degree of instrument performance degradation.

4. The TROPOMI on-orbit calibration and performance tracking method based on background value statistics as described in claim 3, characterized in that, The process of step 3 is as follows: 3.1 Based on the spatial empirical model established in step 2.1 and the trend components extracted in step 2.3, a spatiotemporal integrated model is constructed to predict the background value at any target time; 3.2 Time Consistency Correction: In order to eliminate the long-term systematic bias of data introduced by the degradation of sensor performance over time, and to restore and ensure the consistency and comparability of the entire observation data archive on a long time scale, a relative correction coefficient is calculated at each time point based on the initial trend value at the beginning of the observation period. This coefficient is applied to the deseasoned sequence to correct for time inconsistency.

Images