A method and system for retrieving near-surface NO2 concentrations over all time and with full coverage

Abstract

This invention discloses a method and system for inverting near-surface NO2 concentration with full coverage throughout the day and night, belonging to the field of atmospheric environmental information technology. The method includes acquiring satellite observation data and ground-based observation data, preprocessing them to obtain a matching dataset; constructing a first NO2 concentration inversion model and a second NO2 concentration inversion model based on the matching dataset, and using the two models to obtain the first and second NO2 concentration inversion results respectively; fusing the first and second NO2 concentration inversion results based on Poisson image adjustment theory to obtain a full-domain NO2 concentration inversion result; and correcting the inversion error of the full-domain NO2 concentration inversion result based on discrete cosine transform to obtain the final near-surface NO2 concentration. This invention applies Poisson adjustment and discrete cosine transform combined with satellite and ground-based observation data to obtain full-coverage near-surface NO2 concentration data with full coverage throughout the day and night.

Description

Technical Field

[0001] This invention belongs to the field of atmospheric environmental information technology, specifically relating to a method and system for retrieving near-surface NO2 concentrations with all-day, all-weather coverage. Background Technology

[0002] NO2 monitoring data can be categorized into ground-based station observations, airborne observations, and satellite remote sensing observations. While ground-based station observations can provide relatively accurate real-time near-surface or tropospheric data around the station, their observation range is limited, and the stations are sparse and unevenly distributed. Airborne observations offer high mobility and can obtain vertical column concentration profiles within the flight area; their results are often used to validate satellite remote sensing products, but their high cost prevents large-area observations and widespread use. In contrast, satellite remote sensing observations can achieve long-term, continuous, large-area observations. Therefore, utilizing high-resolution satellite data and other auxiliary variables to retrieve near-surface NO2 concentrations is the current development trend.

[0003] However, current NO2 monitoring still relies primarily on polar-orbiting satellites, which typically observe the same region only once every 1-3 days, making it difficult to capture short-lived pollution events and the complete diurnal variation process. Although the development of geostationary satellites has made continuous monitoring possible, geostationary satellite sensors used for NO2 monitoring are still in their infancy, and NO2 retrieval is mainly based on differential absorption spectroscopy in the ultraviolet spectrum, using a spectral window typically of 400-500 nm. This means that current technologies cannot acquire NO2 observations at night; however, the distribution of NO2 at night cannot be ignored. Furthermore, due to cloud cover and other factors, current instantaneous NO2 retrieval results suffer from large areas of data loss, limiting the operational application of the retrieval results. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for inverting near-surface NO2 concentrations throughout the day and night, thereby achieving full regional coverage of NO2 concentration acquisition.

[0005] This invention provides the following technical solution:

[0006] Firstly, a method for retrieving near-surface NO2 concentration with full-time and all-weather coverage is provided, comprising: acquiring satellite observation data and ground-based observation data, and preprocessing the satellite observation data and ground-based observation data to obtain a matching dataset; constructing a first NO2 concentration retrieval model and a second NO2 concentration retrieval model based on the matching dataset, and obtaining a first NO2 concentration retrieval result and a second NO2 concentration retrieval result respectively using the first NO2 concentration retrieval model and the second NO2 concentration retrieval model; fusing the first NO2 concentration retrieval result and the second NO2 concentration retrieval result based on Poisson image adjustment theory to obtain a full-domain NO2 concentration retrieval result; and correcting the retrieval error of the full-domain NO2 concentration retrieval result based on discrete cosine transform to obtain the final near-surface NO2 concentration.

[0007] As an optional technical solution of the present invention, the satellite observation data includes satellite brightness temperature data, meteorological parameters, normalized vegetation index and surface elevation data.

[0008] As an optional technical solution of the present invention, the satellite observation data and ground-based observation data are preprocessed, including: resampling the meteorological parameters, normalized vegetation index and surface elevation data; and matching the ground-based observation data with the satellite brightness temperature data and the resampled meteorological parameters, normalized vegetation index and surface elevation data.

[0009] As an optional technical solution of the present invention, the first NO2 concentration inversion model is expressed as follows:

[0010] ;

[0011] The second NO2 concentration inversion model is expressed as follows:

[0012] ;

[0013] in, This represents the first NO2 concentration inversion model. This represents the second NO2 concentration inversion model. Represents a function, This indicates the brightness temperature measured by the first channel of the satellite sensor. This indicates the brightness temperature measured by the second channel of the satellite sensor. This indicates the brightness temperature measured by the third channel of the satellite sensor. This indicates the brightness temperature measured by the fourth channel of the satellite sensor. Indicates the solar altitude angle. Indicates the relative humidity of the Earth's surface. This indicates the air temperature at a depth of 2 meters above the Earth's surface. Indicates surface pressure. This represents the meridional component of wind speed at a depth of ten meters above the Earth's surface. This represents the zonal component of wind speed at a depth of ten meters above the Earth's surface. Indicates the boundary layer height. It represents the amount of precipitable water in the atmosphere. This indicates the mass mixing ratio of near-surface ozone. This represents the total amount of ozone in the atmospheric column. Represents the normalized vegetation index. This represents surface elevation data.

[0014] As an optional technical solution of the present invention, the fusion of the first NO2 concentration inversion result and the second NO2 concentration inversion result based on Poisson image adjustment theory to obtain a full-domain NO2 concentration inversion result includes:

[0015] In the region that does not contain the first NO2 concentration inversion result but contains the second NO2 concentration inversion result Above, the objective function f to be solved satisfies:

[0016] ;

[0017] in, , represents the gradient operator, z represents the region Vector field on, Indicates the region The boundary of S, where S represents the global domain. Indicates the region The first NO2 concentration inversion result is a known function. Indicates that it is located at f on Indicates that it is located at On ;

[0018] For a grid p in a global field S, the four adjacent grids (upper, lower, left, and right) of grid p are defined as the neighborhood space N of grid p. p , is represented as:

[0019] ;

[0020] Where q represents the neighboring grids of grid p. ;

[0021] area boundary Represented as:

[0022] ;

[0023] The discretized form of the objective function f to be solved is expressed as:

[0024] ;

[0025] in, , This represents the value of the objective function f at grid point p. This represents the value of the objective function f at grid point q. This represents the gradient vector between grids p and q. , This represents the result of the second NO2 concentration inversion at grid point p before Poisson image adjustment. This represents the result of the second NO2 concentration inversion at grid q before Poisson image adjustment;

[0026] for , satisfy:

[0027] hour,

[0028] get ;

[0029] in, This represents the value of the first NO2 concentration inversion result at grid point q;

[0030] Now, solving for the objective function f in the region... The fused NO2 concentration inversion result is obtained, and the first NO2 concentration inversion result is combined with the fused NO2 concentration inversion result to form a full-domain NO2 concentration inversion result.

[0031] As an optional technical solution of the present invention, the step of correcting the inversion error of the full-domain NO2 concentration inversion result based on discrete cosine transform to obtain the final near-surface NO2 concentration includes:

[0032] The ground-based observation data XG station XG with the overall NO2 concentration inversion results domain The ratio is expressed as:

[0033] ;

[0034] in, The closer the absolute value is to 1, the smaller the inversion error;

[0035] right Spatial interpolation is performed to obtain the initial distribution of the full-domain error. , is represented as:

[0036] ;

[0037] Where i represents the i-th ground-based observation station, and n represents the total number of ground-based observation stations. , representing the ground observation data a of the i-th ground observation station. i The inversion results of the full-domain NO2 concentration at the same time and location (b) i The ratio, where B represents the total NO2 inversion concentration at grid (u, v), u represents the longitude coordinate, and v represents the latitude coordinate. , u i v represents the longitude coordinates of the i-th ground-based observation station. i Represents the latitude coordinates of the i-th ground-based observation station;

[0038] The discrete cosine transform includes a forward transform and an inverse transform, which applies to the initial distribution of errors across the entire domain. Perform iterative correction, expressed as:

[0039] ;

[0040] Where k represents the k-th iteration, Indicates the relaxation factor. Represents a 0-1 mask. , The constants represent the matrix type; STDCT represents the forward transform, and ISTDCT represents the inverse transform.

[0041] When the infinity norm of the difference between two consecutive iterations is less than 10 -3 If the number of iterations exceeds 200, the iteration stops, and the final iteration result is used as the inversion error distribution. The final near-surface NO2 concentration is expressed as:

[0042] .

[0043] As an optional technical solution of the present invention, the transformation formula of the positive transformation is expressed as:

[0044] ;

[0045] ;

[0046] The transformation formula for the inverse transform is expressed as follows:

[0047] ;

[0048] ;

[0049] in, The error distribution data before iteration is represented in three-dimensional matrix form. The error distribution data after one complete iteration is represented in three-dimensional matrix form; M, N, and P represent the total number of rows, total number of columns, and time series length, respectively, and m, j, and t represent the row position, column position, and observation time of the target grid, respectively. Indicates by After a positive transformation, the three-dimensional matrix h, l, and w represent the transformed frequency domain coordinates, and their range of variation is the same as that of m, j, and t.

[0050] Matrix type constants Represented as:

[0051] ;

[0052] in, The parameters are: h∈[0, M-1], l∈[0, N-1], w∈[0, P-1].

[0053] Secondly, a near-surface NO2 concentration inversion system with all-day, all-weather coverage is provided, including:

[0054] The data acquisition module is used to acquire satellite observation data and ground-based observation data, and to preprocess the satellite observation data and ground-based observation data to obtain a matching dataset;

[0055] The model building and prediction module is used to build a first NO2 concentration inversion model and a second NO2 concentration inversion model based on the matching dataset, and to obtain the first NO2 concentration inversion result and the second NO2 concentration inversion result by using the first NO2 concentration inversion model and the second NO2 concentration inversion model, respectively.

[0056] The data fusion module is used to fuse the first NO2 concentration inversion result and the second NO2 concentration inversion result based on the Poisson image adjustment theory to obtain the full-domain NO2 concentration inversion result;

[0057] The error correction module is used to correct the inversion error of the full-domain NO2 concentration inversion result based on discrete cosine transform, so as to obtain the final near-surface NO2 concentration.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] This invention provides a method for retrieving near-surface NO2 concentrations across all time periods. By matching satellite observation data and ground-based observation data, and applying Poisson adjustment theory and discrete cosine transform, it obtains NO2 concentration data with full coverage across all time periods, thereby improving the accuracy of NO2 concentration retrieval. At the same time, this method uses a geostationary satellite sensor that is not specifically designed for trace gas detection, leveraging its high time resolution advantage in infrared detection to achieve hourly-level NO2 retrieval near-surface concentrations. Attached Figure Description

[0060] Figure 1 This is a flowchart of the near-surface NO2 concentration inversion method with full coverage throughout the day, as described in this embodiment of the invention.

[0061] Figure 2 This is a schematic diagram of Poisson image adjustment in an embodiment of the present invention;

[0062] Figure 3 This is a cross-validation comparison chart of near-surface NO2 concentration at various times throughout the day in January and February 2020 and ground-based observations in an embodiment of the present invention.

[0063] Figure 4 This is a comparison diagram of the effect of superimposing the first NO2 concentration inversion result and the second NO2 concentration inversion result with the full-domain NO2 concentration inversion result in this embodiment of the invention;

[0064] Figure 5 This is a map showing the distribution of near-surface NO2 concentration at various times throughout the day on March 2, 2020, based on the first NO2 concentration inversion model, in an embodiment of the present invention.

[0065] Figure 6 This is the distribution of near-surface NO2 concentration at various times throughout the day on March 2, 2020, based on the second NO2 concentration inversion model, in an embodiment of the present invention.

[0066] Figure 7 This is the near-surface NO2 concentration distribution at various times throughout the day on March 2, 2020, based on discrete cosine transform, in an embodiment of the present invention. Detailed Implementation

[0067] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0068] Example 1

[0069] This embodiment provides a method for retrieving near-surface NO2 concentrations over all time and with full coverage, such as... Figure 1 As shown, the specific steps include the following:

[0070] Step 1: Acquire satellite observation data and ground-based observation data, and preprocess the satellite observation data and ground-based observation data to obtain a matching dataset.

[0071] 1.1 The satellite observation data includes satellite brightness temperature data, meteorological parameters, normalized difference vegetation index (NDVI), and land surface elevation data. Because satellite brightness temperature data is affected by cloud cover, all satellite observation data uses samples taken under clear, cloudless skies.

[0072] In this embodiment, the satellite used is the Japanese Himawari-8 / 9 satellite. The satellite brightness temperature data includes the brightness temperatures tbb7, tbb8, tbb9, and tbb10 measured by four channels of the AHI sensor on the Japanese Himawari-8 / 9 satellite: Band7 (λ=3.9μm), Band8 (λ=6.2μm), Band9 (λ=6.9μm), and Band10 (λ=7.3μm). The temporal sampling resolution of the satellite brightness temperature data is 1 hour.

[0073] In this embodiment, the meteorological parameters used are the 1-hour resolution ERA-5 meteorological parameters published by the European Centre for Mesoscale Weather Prediction; the normalized vegetation index (NWDI) is the MODIS normalized vegetation index published by NASA; and the surface elevation data is the SRTM surface elevation data published by NASA. The meteorological parameters include surface relative humidity (RH), air temperature (TEMP), surface pressure (SP), and the east-west meridional component of wind at 10 meters above the surface. The north-south latitudinal component of wind at 10 meters above the ground Boundary layer height (BLH), atmospheric precipitable water (PW), near-surface ozone mass mixing ratio (O3mr), and total atmospheric column ozone (O3all).

[0074] In this embodiment, the ground-based observation data uses data from the National Ambient Air Quality Monitoring Network operated by the China National Environmental Monitoring Center.

[0075] 1.2 Due to the significant structural heterogeneity and spatiotemporal resolution differences in the acquired data, preprocessing is required to obtain a spatiotemporally matched dataset as training samples. Specifically:

[0076] 1.2.1 The meteorological parameters, normalized vegetation index and surface elevation data were resampled with a sampling resolution of 0.05°.

[0077] 1.2.2 Match the ground-based observation data with the satellite brightness temperature data, as well as the resampled meteorological parameters, normalized vegetation index, and surface elevation data.

[0078] Step 2: Based on the matching dataset, construct a first NO2 concentration inversion model and a second NO2 concentration inversion model. Using the first NO2 concentration inversion model and the second NO2 concentration inversion model, obtain the first NO2 concentration inversion result and the second NO2 concentration inversion result, respectively.

[0079] 2.1 In this embodiment, based on the LightGBM machine learning algorithm, a first NO2 concentration inversion model and a second NO2 concentration inversion model are constructed according to the matching dataset.

[0080] For cloudless areas and time periods, all matched data are input as explanatory variables into the initialized LightGBM model to obtain the first NO2 concentration inversion model, which is expressed as:

[0081] .

[0082] Data other than satellite brightness temperature data from the matching data were input as explanatory variables into the initialized LightGBM model to obtain the second NO2 concentration inversion model, which is expressed as:

[0083] .

[0084] in, This represents the first NO2 concentration inversion model. This represents the second NO2 concentration inversion model. Represents a function, This represents the brightness temperature measured by the first channel of the satellite sensor (λ=3.9μm). This represents the brightness temperature measured by the second channel (λ=6.2μm) of the satellite sensor. This represents the brightness temperature measured by the third channel (λ=6.9μm) of the satellite sensor. This represents the brightness temperature measured by the fourth channel (λ=7.3μm) of the satellite sensor. Indicates the solar altitude angle. Indicates the relative humidity of the Earth's surface. This indicates the air temperature at a depth of 2 meters above the Earth's surface. Indicates surface pressure. This represents the meridional component of wind speed at a depth of ten meters above the Earth's surface. This represents the zonal component of wind speed at a depth of ten meters above the Earth's surface. Indicates the boundary layer height. It represents the amount of precipitable water in the atmosphere. This indicates the mass mixing ratio of near-surface ozone. This represents the total amount of ozone in the atmospheric column. Represents the normalized vegetation index. Representing surface elevation data

[0085] 2.2. The first NO2 concentration inversion model and the second NO2 concentration inversion model are used to obtain the first NO2 concentration inversion result and the second NO2 concentration inversion result, respectively.

[0086] In this embodiment, the first NO2 concentration inversion result has high accuracy, but it is not a full-domain result due to cloud cover and other factors. The second NO2 concentration inversion result does not rely on satellite observations, thus obtaining a seamless spatiotemporal NO2 inversion result between day and night. However, due to the lack of necessary satellite observation data, its inversion accuracy is lower than that of the first NO2 concentration inversion result.

[0087] Step 3: Based on Poisson image adjustment theory, fuse the first NO2 concentration inversion results and the second NO2 concentration inversion results to obtain the full-domain NO2 concentration inversion results. Specifically, this includes:

[0088] like Figure 2 As shown, S represents the total NO2 concentration region. This represents the region from the second NO2 concentration inversion result. This represents the region from which the first NO2 concentration inversion result was obtained. There is a known function It is necessary to calculate the definition of satisfying the fixed boundary conditions. The objective function f is as follows:

[0089] In the region that does not contain the first NO2 concentration inversion result but contains the second NO2 concentration inversion result Above, the objective function f to be solved satisfies:

[0090] ;

[0091] in, , represents the gradient operator, z represents the region Vector field on, Indicates the region The boundary of S, where S represents the global domain. Indicates the region The first NO2 concentration inversion result is a known function. Indicates that it is located at f on Indicates that it is located at On .

[0092] For a grid p in a global domain S, the four adjacent grids (upper, lower, left, and right) of grid p are defined as the neighborhood space Np of grid p, denoted as:

[0093] ;

[0094] Where q represents the neighboring grids of grid p. When the target grid is at the boundary, its number of adjacent grids is less than 4.

[0095] area boundary Represented as:

[0096] .

[0097] The discretized form of the objective function f to be solved is expressed as:

[0098] ;

[0099] in, , This represents the value of the objective function f at grid point p. This represents the value of the objective function f at grid point q. This represents the gradient vector between grids p and q. , This represents the result of the second NO2 concentration inversion at grid point p before Poisson image adjustment. This represents the result of the second NO2 concentration inversion at grid q before Poisson image adjustment.

[0100] From solving the functional form and related knowledge of multivariable quadratic functions, we know that for , satisfy:

[0101] hour,

[0102] get ;

[0103] in, This represents the value of the first NO2 concentration inversion result at grid point q.

[0104] Now, solving for the objective function f in the region... The fused NO2 concentration inversion result is obtained, and the first NO2 concentration inversion result is combined with the fused NO2 concentration inversion result to form a full-domain NO2 concentration inversion result.

[0105] like Figure 2 As shown, the region of the second NO2 concentration inversion result. The region of the first NO2 concentration inversion result This forms the global domain S.

[0106] Step 4: Correct the inversion error of the full-domain NO2 concentration inversion results based on discrete cosine transform to obtain the final near-surface NO2 concentration. Details are as follows:

[0107] 4.1 The aforementioned ground observation data XG station XG with the overall NO2 concentration inversion results domain The ratio is expressed as:

[0108] ;

[0109] The closer the absolute value of is to 1, the smaller the inversion error.

[0110] Due to XG station The distribution is limited, so This is not full-domain data. Spatial interpolation is performed to obtain the initial distribution of the full-domain error. , is represented as:

[0111] ;

[0112] Where i represents the i-th ground-based observation station, and n represents the total number of ground-based observation stations. , representing the ground observation data a of the i-th ground observation station. i The inversion results of the full-domain NO2 concentration at the same time and location (b) i The ratio, where B represents the total NO2 inversion concentration at grid (u, v), u represents the longitude coordinate, and v represents the latitude coordinate. , u i v represents the longitude coordinates of the i-th ground-based observation station. i This represents the latitude coordinates of the i-th ground-based observation station.

[0113] 4.2 The discrete cosine transform includes a forward transform and an inverse transform, which are applied to the initial error distribution across the entire domain. Perform iterative correction, expressed as:

[0114] ;

[0115] Where k represents the k-th iteration, Indicates the relaxation factor. Represents a 0-1 mask. , The constants represent the matrix type; STDCT represents the forward transform, and ISTDCT represents the inverse transform.

[0116] Furthermore, the transformation formula for the positive transformation is expressed as:

[0117] .

[0118] ;

[0119] The transformation formula for the inverse transform is expressed as follows:

[0120] ;

[0121] ;

[0122] in, The error distribution data before iteration is represented in three-dimensional matrix form. The error distribution data after one complete iteration is represented in three-dimensional matrix form; M, N, and P represent the total number of rows, total number of columns, and time series length, respectively, and m, j, and t represent the row position, column position, and observation time of the target grid, respectively. Indicates by The three-dimensional matrix after one positive transformation, h, l, and w represent the transformed frequency domain coordinates, and their range of variation is the same as that of m, j, and t.

[0123] Furthermore, the Discrete Cosine Transform (DCT) can use the cosine function to transform the signal from the spatial domain to the frequency domain, thereby representing the data as a weighted sum of different frequency components, effectively smoothing the observation error within the target spatiotemporal range.

[0124] Furthermore, constants of matrix type Represented as:

[0125] ;

[0126] in, The parameters are: h∈[0, M-1], l∈[0, N-1], w∈[0, P-1].

[0127] 4.3 When the infinity norm of the difference between two consecutive iterations is less than 10 -3 If the number of iterations exceeds 200, the iteration stops, and the final iteration result is used as the inversion error distribution. The final near-surface NO2 concentration is expressed as:

[0128] .

[0129] Example 2

[0130] Based on Example 1, this embodiment provides a specific application of the method, using the central and eastern parts of China (95°E ~ 130°E, 20°N ~ 45°N) in January and February 2020 as a demonstration study area.

[0131] like Figure 3As shown, based on the 10-fold cross-validation method, a time-by-time NO2 concentration inversion model for January-February 2020 is presented (MODEL). Ⅰ ), second NO2 concentration inversion model (MODEL) Ⅱ The comparison of four NO2 concentrations after Poisson adjustment and DCT (Discrete Cosine Transform) satellite-to-ground fusion with ground-based observations. The so-called ten-fold cross-validation method involves randomly dividing the dataset into 10 subsets of equal or approximately equal size, using 9 subsets as training data and 1 subset as test data in turn, and averaging the evaluation metrics over 10 iterations as a comprehensive evaluation of the model performance. For example... Figure 3 As shown, MODEL Ⅰ Its accuracy is higher than that of the MODEL Ⅱ Its R 2 The values ​​were 0.72 and 0.69 respectively. Although the overall accuracy did not change significantly after Poisson adjustment and RMSE and MAE decreased slightly, Poisson adjustment ensured the spatial continuity of NO2 distribution. Applying DCT satellite-to-ground fusion to the image after Poisson adjustment yielded high-precision NO2 inversion results, with R... 2 It is 0.86.

[0132] like Figure 4 As shown, (a) indicates that the MODEL Ⅰ and MODEL Ⅱ The results are then overlaid using simple layers, with (b) representing the result after Poisson adjustment. At 02:00 local time on January 5, 2020, when the MODEL... Ⅰ and MODEL Ⅱ When performing simple layer overlays on the results, i.e. using MODEL Ⅰ Effective inversion replacement MODEL Ⅱ The results show that there are jumps in the concentration distribution.

[0133] like Figure 5-7 As shown, the models are displayed hour by hour throughout the day on March 2, 2020. Ⅰ MODEL Ⅱ NO2 concentration distribution fused with satellite and ground-based observations. The results show that due to factors such as cloud cover, the MODEL... ⅠThe NO2 inversion results showed significant data gaps. However, the results from satellite-ground fusion achieved comprehensive coverage and were highly consistent with ground-based observations. Spatially, high NO2 concentrations were mainly concentrated in the Beijing-Tianjin-Hebei region and six surrounding provinces and municipalities, which is related to emissions from transportation, industry, and thermal power generation. In addition, high NO2 concentrations were also observed in parts of northern Gansu, Inner Mongolia, and Northeast China, consistent with other studies indicating that besides emissions from transportation, industry, and thermal power generation, farmland and grasslands also release NO2. Temporally, high NO2 concentrations were primarily observed at night, while daytime concentrations were generally lower, especially between 12:00 and 18:00. This is because strong sunlight during the day promotes the photolysis of NO2 in the atmosphere, while at night, the photolysis reaction significantly weakens, and the atmospheric boundary layer decreases significantly, leading to higher NO2 concentrations.

[0134] Example 3

[0135] This embodiment provides an all-weather, full-coverage near-surface NO2 concentration inversion system, including:

[0136] The data acquisition module is used to acquire satellite observation data and ground-based observation data, and to preprocess the satellite observation data and ground-based observation data to obtain a matching dataset.

[0137] The model building and prediction module is used to construct a first NO2 concentration inversion model and a second NO2 concentration inversion model based on the matching dataset, and to obtain the first NO2 concentration inversion result and the second NO2 concentration inversion result by using the first NO2 concentration inversion model and the second NO2 concentration inversion model, respectively.

[0138] The data fusion module is used to fuse the first NO2 concentration inversion result and the second NO2 concentration inversion result based on the Poisson image adjustment theory to obtain the full-domain NO2 concentration inversion result.

[0139] The error correction module is used to correct the inversion error of the full-domain NO2 concentration inversion result based on discrete cosine transform, so as to obtain the final near-surface NO2 concentration.

[0140] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0141] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0142] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0143] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0144] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for retrieving near-surface NO2 concentrations over all time and with full coverage, characterized in that, include: Acquire satellite observation data and ground-based observation data, and preprocess the satellite observation data and ground-based observation data to obtain a matching dataset; Based on the matching dataset, a first NO2 concentration inversion model and a second NO2 concentration inversion model are constructed. Using the first NO2 concentration inversion model and the second NO2 concentration inversion model, the first NO2 concentration inversion result and the second NO2 concentration inversion result are obtained respectively. The first NO2 concentration inversion result and the second NO2 concentration inversion result are fused based on the Poisson image adjustment theory to obtain the full-domain NO2 concentration inversion result; The inversion error of the full-domain NO2 concentration inversion result is corrected by discrete cosine transform, and the final near-surface NO2 concentration is obtained. The first NO2 concentration inversion model is expressed as follows: ； The second NO2 concentration inversion model is expressed as follows: ； in, This represents the first NO2 concentration inversion model. This represents the second NO2 concentration inversion model. Represents a function, This indicates the brightness temperature measured by the first channel of the satellite sensor. This indicates the brightness temperature measured by the second channel of the satellite sensor. This indicates the brightness temperature measured by the third channel of the satellite sensor. This indicates the brightness temperature measured by the fourth channel of the satellite sensor. Indicates the solar altitude angle. Indicates the relative humidity of the Earth's surface. This indicates the air temperature at a depth of 2 meters above the Earth's surface. Indicates surface pressure. This represents the meridional component of wind speed at a depth of ten meters above the Earth's surface. This represents the zonal component of wind speed at a depth of ten meters above the Earth's surface. Indicates the boundary layer height. It represents the amount of precipitable water in the atmosphere. This indicates the mass mixing ratio of near-surface ozone. This represents the total amount of ozone in the atmospheric column. Represents the normalized vegetation index. Represents surface elevation data; The method of fusing the first and second NO2 concentration inversion results based on Poisson image adjustment theory to obtain a full-domain NO2 concentration inversion result includes: In the region that does not contain the first NO2 concentration inversion result but contains the second NO2 concentration inversion result Above, the objective function f to be solved satisfies: ； in, , represents the gradient operator, z represents the region Vector field on, Indicates the region The boundary of S, where S represents the global domain. Indicates the region The first NO2 concentration inversion result is a known function. Indicates that it is located at f on Indicates that it is located at On ; For a grid p in a global field S, the four adjacent grids (upper, lower, left, and right) of grid p are defined as the neighborhood space N of grid p. p , is represented as: ； Where q represents the neighboring grids of grid p. ; area boundary Represented as: ； The discretized form of the objective function f to be solved is expressed as: ； in, , This represents the value of the objective function f at grid point p. This represents the value of the objective function f at grid point q. This represents the gradient vector between grids p and q. , This represents the result of the second NO2 concentration inversion at grid point p before Poisson image adjustment. This represents the result of the second NO2 concentration inversion at grid q before Poisson image adjustment; for , satisfy: hour, get ; in, This represents the value of the first NO2 concentration inversion result at grid point q; Now, solving for the objective function f in the region... The fused NO2 concentration inversion result is obtained, and the first NO2 concentration inversion result is combined with the fused NO2 concentration inversion result to form a full-domain NO2 concentration inversion result; The method of correcting the inversion error of the full-domain NO2 concentration inversion results based on discrete cosine transform to obtain the final near-surface NO2 concentration includes: The ground-based observation data XG station XG with the overall NO2 concentration inversion results domain The ratio is expressed as: ； in, The closer the absolute value is to 1, the smaller the inversion error; right Spatial interpolation is performed to obtain the initial distribution of the full-domain error. , is represented as: ； Where i represents the i-th ground-based observation station, and n represents the total number of ground-based observation stations. , representing the ground observation data a of the i-th ground observation station. i The inversion results of the full-domain NO2 concentration at the same time and location (b) i The ratio, where B represents the total NO2 inversion concentration at grid (u, v), u represents the longitude coordinate, and v represents the latitude coordinate. , u i v represents the longitude coordinates of the i-th ground-based observation station. i Represents the latitude coordinates of the i-th ground-based observation station; The discrete cosine transform includes a forward transform and an inverse transform, which applies to the initial distribution of errors across the entire domain. Perform iterative correction, expressed as: ； Where k represents the k-th iteration, Indicates the relaxation factor. Represents a 0-1 mask. , The constants represent the matrix type; STDCT represents the forward transform, and ISTDCT represents the inverse transform. When the infinity norm of the difference between two consecutive iterations is less than 10 -3 If the number of iterations exceeds 200, the iteration stops, and the final iteration result is used as the inversion error distribution. The final near-surface NO2 concentration is expressed as: 。 2. The all-weather, full-coverage near-surface NO2 concentration inversion method according to claim 1, characterized in that, The satellite observation data includes satellite brightness temperature data, meteorological parameters, normalized vegetation index, and surface elevation data.

3. The all-weather, full-coverage near-surface NO2 concentration inversion method according to claim 2, characterized in that: Preprocessing of the satellite observation data and ground-based observation data includes: The meteorological parameters, normalized vegetation index, and surface elevation data were resampled. The ground-based observation data was matched with satellite brightness temperature data, as well as the resampled meteorological parameters, normalized vegetation index, and surface elevation data.

4. The all-weather, full-coverage near-surface NO2 concentration inversion method according to claim 1, characterized in that: The transformation formula for the positive transformation is expressed as: ； ； The transformation formula for the inverse transform is expressed as follows: ； ； in, The error distribution data before iteration is represented in three-dimensional matrix form. The error distribution data after one complete iteration is represented in three-dimensional matrix form; M, N, and P represent the total number of rows, total number of columns, and time series length, respectively, and m, j, and t represent the row position, column position, and observation time of the target grid, respectively. Indicates by After a positive transformation, the three-dimensional matrix h, l, and w represent the transformed frequency domain coordinates, and their range of variation is the same as that of m, j, and t. Matrix type constants Represented as: ； in, The parameters are: h∈[0, M-1], l∈[0, N-1], w∈[0, P-1].

5. A near-surface NO2 concentration retrieval system with all-weather, all-time coverage, characterized in that, The method for inverting near-surface NO2 concentration with all-day, all-coverage coverage as described in any one of claims 1-4 includes: The data acquisition module is used to acquire satellite observation data and ground-based observation data, and to preprocess the satellite observation data and ground-based observation data to obtain a matching dataset; The model building and prediction module is used to build a first NO2 concentration inversion model and a second NO2 concentration inversion model based on the matching dataset, and to obtain the first NO2 concentration inversion result and the second NO2 concentration inversion result by using the first NO2 concentration inversion model and the second NO2 concentration inversion model, respectively. The data fusion module is used to fuse the first NO2 concentration inversion result and the second NO2 concentration inversion result based on the Poisson image adjustment theory to obtain the full-domain NO2 concentration inversion result; The error correction module is used to correct the inversion error of the full-domain NO2 concentration inversion result based on discrete cosine transform, so as to obtain the final near-surface NO2 concentration.

Images