A cross-border regional pm2.5 three-dimensional reconstruction method and device
By processing multi-source data and training the Transformer model, a three-dimensional reconstruction of PM2.5 in cross-border areas was achieved, which solved the problem of inconsistent vertical profile information of PM2.5 in cross-border areas and provided detailed and reliable data on pollutant transport and air quality assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 云南省生态环境监测中心
- Filing Date
- 2026-05-08
- Publication Date
- 2026-06-02
AI Technical Summary
Significant differences exist in the ability to acquire PM2.5 vertical profile information across cross-border regions, making it difficult to construct a unified and consistent description of vertical distribution, which affects the accurate identification of pollutant transport height and pollution layer thickness.
By acquiring multi-source data and performing time alignment, spatial alignment, and normalization, a three-dimensional PM2.5 reconstruction model was trained using the Transformer model combined with cross-border meteorological profile data to generate cross-border PM2.5 three-dimensional concentration field data.
In the absence of overseas vertical observation, a spatially continuous and structurally consistent three-dimensional PM2.5 concentration field was constructed, which improved the precision and reliability of cross-border pollutant transport analysis and air quality assessment.
Smart Images
Figure CN122134950A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of atmospheric environment monitoring technology, and specifically relates to a method and device for three-dimensional reconstruction of PM2.5 in cross-border areas. Background Technology
[0002] Transboundary regions are typically characterized by high altitudes, significant topographic relief, and complex meteorological conditions. Their atmospheric environment is highly sensitive to regional and transregional pollutant transport processes. Under the combined influence of large-scale circulation and monsoon systems, fine particulate matter (PM2.5) and other pollutants emitted from surrounding areas can be transported into transboundary regions via long-distance and multi-altitude pathways, significantly impacting regional air quality, ecological environment, and climate systems. Therefore, a refined characterization and quantitative analysis of the spatial distribution characteristics of PM2.5 and other pollutants in transboundary regions, as well as their transregional and trans-altitude transport processes, is an important technical requirement in current atmospheric environmental monitoring and pollution transport research.
[0003] However, vertical PM2.5 information in cross-border areas has long been discontinuous. Domestic areas can usually obtain PM2.5 vertical profile information with high spatiotemporal resolution by relying on remote sensing methods such as lidar, while overseas areas are limited by monitoring conditions and often only have near-surface concentration data, or even lack stable and continuous observation data. This significant difference in the ability to obtain vertical profile information between domestic and overseas areas makes it difficult to construct a unified and consistent description of PM2.5 vertical distribution across cross-border areas, thus affecting the accurate identification of pollutant transport height, pollution layer thickness and its temporal evolution characteristics. Summary of the Invention
[0004] Based on the above analysis, the embodiments of the present invention aim to provide a method and apparatus for three-dimensional reconstruction of PM2.5 in cross-border areas, in order to solve the technical problem that there are significant differences in the ability to obtain vertical profile information between domestic and foreign countries in the prior art.
[0005] The objective of this invention is achieved as follows: A first aspect of the present invention provides a method for three-dimensional reconstruction of PM2.5 in a cross-border area, comprising: Acquire multi-source data from cross-border regions, including domestic PM2.5 vertical profile data, overseas PM2.5 mass concentration data, cross-border meteorological profile data, and overseas PM2.5 simulated profile data; perform time alignment, spatial alignment, and normalization processing on the multi-source data; The vertical profile data of PM2.5 outside the country is calculated based on the simulated profile data and mass concentration data of PM2.5 outside the country; the vertical profile data of PM2.5 inside the country and the vertical profile data of PM2.5 outside the country are mapped to the same spatial grid system to obtain the vertical profile data of PM2.5 across the country. Using cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples, a Transformer model was trained to obtain a cross-border PM2.5 three-dimensional reconstruction model. Meteorological profile data of the cross-border area to be studied are input into the cross-border PM2.5 three-dimensional reconstruction model, and PM2.5 three-dimensional concentration field data of the cross-border area to be studied are output.
[0006] Furthermore, the domestic PM2.5 vertical profile data is obtained by inverting the aerosol backscattering signal collected by lidar and combining it with the conversion relationship between PM2.5 and aerosol optical parameters; the overseas PM2.5 mass concentration data is obtained by monitoring overseas ground air quality monitoring stations; the cross-border meteorological profile data is obtained by meteorological models; and the overseas PM2.5 simulated profile data is obtained by chemical transport models.
[0007] Furthermore, the process of performing time alignment, spatial alignment, and normalization on the multi-source data includes: mapping multi-source data with different time resolutions to a unified target time step node; mapping multi-source data with different spatial resolutions and vertical layers to a unified three-dimensional spatial grid through horizontal interpolation and vertical resampling; and normalizing the mapped data based on the sample mean and standard deviation.
[0008] Further, the calculation of overseas PM2.5 vertical profile data based on overseas PM2.5 simulated profile data and overseas PM2.5 mass concentration data includes: using overseas simulated profile data output by a chemical transport model as a basis; and using the overseas PM2.5 mass concentration data, performing deviation correction on the simulated profile data through a weighting function that varies with altitude to generate the overseas PM2.5 vertical profile data, expressed as: , in, This represents the vertical profile data of PM2.5 outside of China. This indicates simulated profile data from overseas. Represents the weighting function. This represents PM2.5 mass concentration data outside of China, where z represents altitude and t represents time.
[0009] Furthermore, the step of training the Transformer model using cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples includes: constructing the cross-border meteorological profile data into input feature vectors arranged by altitude layers; the Transformer model converting the input feature vectors into query, key, and value vectors through linear mapping, and using the Softmax function to calculate attention scores based on the query, key, and value vectors; and using the mean squared error between the cross-border PM2.5 vertical profile predicted by the Transformer model and the supervised samples as the loss function to optimize and train the model parameters.
[0010] Furthermore, the step of inputting meteorological profile data of the cross-border region to be studied into the cross-border PM2.5 three-dimensional reconstruction model and outputting PM2.5 three-dimensional concentration field data of the cross-border region to be studied includes: acquiring meteorological profile data corresponding to each spatial grid point in the cross-border region to be studied, wherein the meteorological profile data of the cross-border region to be studied is consistent with the model training stage in terms of vertical structure and feature dimensions; inputting the meteorological profile data corresponding to each spatial grid point into the cross-border PM2.5 three-dimensional reconstruction model respectively, and outputting PM2.5 vertical profile data of the corresponding spatial grid points at different height layers; and stitching together the PM2.5 vertical profile data of each spatial grid point to generate PM2.5 three-dimensional concentration field data of the cross-border region to be studied.
[0011] Furthermore, the PM2.5 three-dimensional concentration field data is used to characterize the spatial distribution characteristics of PM2.5 at different altitudes within the cross-border region, and is used for cross-border pollution transport analysis, pollution layer height identification, or regional air quality assessment.
[0012] A second aspect of the present invention provides a three-dimensional reconstruction device for PM2.5 in a cross-border area, comprising: The data acquisition and processing module is used to acquire multi-source data from cross-border areas, including domestic PM2.5 vertical profile data, overseas PM2.5 mass concentration data, cross-border meteorological profile data, and overseas PM2.5 simulated profile data; and to perform time alignment, spatial alignment, and normalization processing on the multi-source data. The cross-border PM2.5 vertical profile construction module is used to calculate the cross-border PM2.5 vertical profile data based on the simulated profile data and mass concentration data of PM2.5 outside the country; and to map the vertical profile data of PM2.5 inside the country and the vertical profile data of PM2.5 outside the country to the same spatial grid system to obtain the cross-border PM2.5 vertical profile data. The model training module is used to train the Transformer model with cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples to obtain a cross-border PM2.5 three-dimensional reconstruction model. The PM2.5 3D reconstruction module is used to input meteorological profile data of the cross-border area to be studied into the cross-border PM2.5 3D reconstruction model and output PM2.5 3D concentration field data of the cross-border area to be studied.
[0013] A third aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the three-dimensional reconstruction method for cross-border PM2.5 in any embodiment.
[0014] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the three-dimensional reconstruction method for PM2.5 in cross-border areas as described in any embodiment.
[0015] Compared with the prior art, the present invention can achieve at least the following beneficial effects: By combining PM2.5 vertical profile data obtained from domestic lidar inversion with simulation results from overseas chemical transport models and ground observation data, this invention effectively compensates for the lack of PM2.5 vertical information caused by inconsistencies in vertical observation capabilities between domestic and overseas regions. Simultaneously, based on a unified cross-border meteorological profile-driven deep learning model, the vertical structure of PM2.5 is learned and inferred, enabling the reconstruction of the vertical distribution of PM2.5 in cross-border regions. This invention can construct a spatially continuous and structurally consistent three-dimensional PM2.5 concentration field even in the absence of overseas vertical observation conditions, providing more refined and reliable data support for the identification of cross-border pollutant transport heights, pollution layer structure analysis, and cross-regional air quality assessment, significantly improving the three-dimensional monitoring and analysis capabilities of cross-border air pollution. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings.
[0017] Figure 1 A flowchart of the three-dimensional reconstruction method for PM2.5 in cross-border areas provided in Embodiment 1 of the present invention; Figure 2 A schematic diagram of a three-dimensional reconstruction device for PM2.5 in a cross-border area provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the electronic device architecture provided in Embodiment 3 of the present invention. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. It should be noted that, unless otherwise specified, the implementation methods and features in the implementation methods in this disclosure can be combined, separated, interchanged, and / or rearranged. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0019] Example 1 A specific embodiment of the present invention, such as Figure 1 As shown, a three-dimensional reconstruction method for PM2.5 in cross-border areas is disclosed, including the following steps: S1. Obtain multi-source data for the cross-border region, including domestic PM2.5 vertical profile data, overseas PM2.5 mass concentration data, cross-border meteorological profile data, and overseas PM2.5 simulated profile data; perform time alignment, spatial alignment, and normalization processing on the multi-source data.
[0020] In this embodiment, the domestic PM2.5 vertical profile data is obtained by inverting the aerosol backscattering signal collected by lidar and combining it with the conversion relationship between PM2.5 and aerosol optical parameters. The overseas PM2.5 mass concentration data is obtained by monitoring overseas ground air quality monitoring stations. The cross-border meteorological profile data is obtained by meteorological models. The overseas PM2.5 simulated profile data is obtained by chemical transport models.
[0021] Specifically, within the domestic region, aerosol echo signals at different altitudes are acquired using a deployed lidar system. Based on the lidar equations, the Fernald inversion algorithm is used to invert the aerosol backscattering signals, obtaining the aerosol extinction coefficient profiles for each altitude level. Furthermore, combined with ground-based PM2.5 mass concentration monitoring data, a conversion relationship between aerosol optical parameters and PM2.5 mass concentration is established, thereby obtaining the vertical PM2.5 profile data for the domestic region. Transboundary meteorological profile data are obtained through numerical meteorological model simulation, specifically based on reanalysis. Meteorological data or routine meteorological observation data are used to conduct numerical simulations of cross-border areas, outputting the vertical distribution results of meteorological elements such as temperature, wind speed, wind direction, and humidity at different altitude levels, and constructing cross-border meteorological profile data based on this. The simulated profile data of PM2.5 outside the border is obtained through a chemical transport model. Driven by the meteorological field provided by the meteorological model, the chemical transport model numerically simulates the emission, transport, diffusion, chemical transformation, and deposition processes of PM2.5 in the border area, outputting the simulated concentration distribution of PM2.5 at different altitude levels, forming the simulated profile data of PM2.5 outside the border.
[0022] In this embodiment, the time alignment, spatial alignment, and normalization processing of the multi-source data includes: Map multi-source data with different time resolutions to a unified target time step node; By using horizontal interpolation and vertical resampling, multi-source data with different spatial resolutions and vertical layers are mapped to a unified three-dimensional spatial grid. The mapped data are normalized based on the sample mean and standard deviation.
[0023] Specifically, by performing time alignment, spatial alignment, and normalization, multi-source data with different sources, resolutions, and units of measurement are kept consistent within a unified time scale, spatial grid, and numerical range. This eliminates the spatiotemporal mismatch between multi-source data, reduces the impact of scale differences of different variables on the subsequent model training process, and provides a stable and unified data foundation for the construction of cross-border PM2.5 vertical profiles and the training of deep learning models.
[0024] S2. Calculate the vertical profile data of PM2.5 outside the country based on the simulated profile data and mass concentration data of PM2.5 outside the country; map the vertical profile data of PM2.5 inside the country and the vertical profile data of PM2.5 outside the country to the same spatial grid system to obtain the vertical profile data of PM2.5 across the border.
[0025] In this embodiment, the calculation of overseas PM2.5 vertical profile data based on overseas PM2.5 simulated profile data and overseas PM2.5 mass concentration data includes: Based on the simulated profile data of the overseas region output by the chemical transport model; Using the aforementioned PM2.5 mass concentration data from overseas, the simulated profile data is corrected for deviation using a weighting function that varies with altitude, generating the overseas PM2.5 vertical profile data, which is represented as follows: , in, This represents the vertical profile data of PM2.5 outside of China. This indicates simulated profile data from overseas. Represents the weighting function. This represents PM2.5 mass concentration data outside of China, where z represents altitude and t represents time.
[0026] S3. Using cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples, train the Transformer model to obtain a cross-border PM2.5 three-dimensional reconstruction model.
[0027] In this embodiment, training the Transformer model using cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples includes: The cross-border meteorological profile data is constructed as an input feature vector arranged by altitude layer; The Transformer model transforms the input feature vector into query, key, and value vectors through a linear mapping, and uses the Softmax function to calculate attention scores based on the query, key, and value vectors. The mean squared error between the cross-border PM2.5 vertical profile predicted by the Transformer model and the supervised samples was used as the loss function to optimize the model parameters during training.
[0028] Specifically, firstly, under a unified spatial grid system, the meteorological vertical profile corresponding to the target spatial grid point at the target time is selected as the model input, and a one-dimensional sequence is formed according to the height order and input into the Transformer model. During the model inference process, the input meteorological profile sequence is first converted into a unified-dimensional feature representation through a linear mapping layer, and then fused with the position encoding of the corresponding height layer to introduce the relative position information between different height layers. Subsequently, the feature sequence is passed through the multi-head self-attention layer and the feedforward neural network layer in the Transformer encoding structure. The multi-head self-attention layer is used to model the correlation between meteorological features at different height layers, and the feedforward neural network layer is used to perform nonlinear transformation and feature enhancement on the correlated features. After the above multi-layer encoding processing, the model outputs a feature representation that corresponds one-to-one with the input height layer. The feature representation is further converted into the PM2.5 mass concentration prediction value at each height layer through the output mapping layer, thereby forming the PM2.5 vertical profile corresponding to the target spatial grid point.
[0029] S4. Input the meteorological profile data of the cross-border area to be studied into the cross-border PM2.5 three-dimensional reconstruction model, and output the PM2.5 three-dimensional concentration field data of the cross-border area to be studied.
[0030] In this embodiment, step S4 specifically includes: Meteorological profile data corresponding to each spatial grid point in the cross-border area to be studied are obtained, and the meteorological profile data of the cross-border area to be studied is consistent with the model training stage in terms of vertical structure and feature dimension. The meteorological profile data corresponding to each spatial grid point are input into the cross-border PM2.5 three-dimensional reconstruction model, and the vertical profile data of PM2.5 at different height layers of the corresponding spatial grid points are output. The PM2.5 vertical profile data of each spatial grid point are stitched together to generate three-dimensional PM2.5 concentration field data of the cross-border area to be studied.
[0031] Specifically, the meteorological profile data corresponding to each spatial grid point in the cross-border area to be studied is input into the cross-border PM2.5 three-dimensional reconstruction model to obtain the PM2.5 vertical profile results of the spatial grid point at each altitude layer; by uniformly stitching the PM2.5 vertical profile results of all spatial grid points in the area, a PM2.5 three-dimensional concentration field data covering the entire cross-border area and continuous in both horizontal and vertical directions can be formed, thereby realizing the three-dimensional reconstruction of the spatial distribution of PM2.5 in the cross-border area.
[0032] In some embodiments, the PM2.5 three-dimensional concentration field data is used to characterize the spatial distribution characteristics of PM2.5 at different altitudes within a cross-border region, and is used for cross-border pollution transport analysis, pollution layer height identification, or regional air quality assessment.
[0033] Compared with existing technologies, the three-dimensional reconstruction method for PM2.5 in cross-border regions provided in this embodiment effectively compensates for the lack of vertical PM2.5 information caused by the inconsistency in vertical observation capabilities between domestic and foreign regions. This is achieved by jointly constructing a method using vertical profile data of PM2.5 obtained from domestic lidar inversion, simulation results from overseas chemical transport models, and ground observation data. Simultaneously, based on a unified cross-border meteorological profile-driven deep learning model, the vertical structure of PM2.5 is learned and inferred, realizing the reconstruction of the vertical distribution of PM2.5 in cross-border regions. This invention can construct a spatially continuous and structurally consistent three-dimensional PM2.5 concentration field even in the absence of overseas vertical observation conditions, providing more refined and reliable data support for the identification of cross-border pollutant transport heights, analysis of pollution layer structure, and cross-regional air quality assessment, significantly improving the three-dimensional monitoring and analysis capabilities of cross-border air pollution.
[0034] Example 2 This embodiment provides a three-dimensional reconstruction device for PM2.5 in cross-border areas, such as... Figure 2 As shown, it includes: The data acquisition and processing module is used to acquire multi-source data from cross-border areas, including domestic PM2.5 vertical profile data, overseas PM2.5 mass concentration data, cross-border meteorological profile data, and overseas PM2.5 simulated profile data; and to perform time alignment, spatial alignment, and normalization processing on the multi-source data. The cross-border PM2.5 vertical profile construction module is used to calculate the cross-border PM2.5 vertical profile data based on the simulated profile data and mass concentration data of PM2.5 outside the country; and to map the vertical profile data of PM2.5 inside the country and the vertical profile data of PM2.5 outside the country to the same spatial grid system to obtain the cross-border PM2.5 vertical profile data. The model training module is used to train the Transformer model with cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples to obtain a cross-border PM2.5 three-dimensional reconstruction model. The PM2.5 3D reconstruction module is used to input meteorological profile data of the cross-border area to be studied into the cross-border PM2.5 3D reconstruction model and output PM2.5 3D concentration field data of the cross-border area to be studied.
[0035] Example 3 This embodiment provides an electronic device, such as... Figure 3 As shown, it includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the three-dimensional reconstruction method for PM2.5 in cross-border areas as described in any of the above embodiments.
[0036] Example 4 This embodiment provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it implements the three-dimensional reconstruction method for PM2.5 in cross-border areas as described in any of the above embodiments.
[0037] Computer-readable storage media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0038] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0039] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0040] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for three-dimensional reconstruction of PM2.5 in a cross-border region, characterized in that, include: Acquire multi-source data from cross-border regions, including domestic PM2.5 vertical profile data, overseas PM2.5 mass concentration data, cross-border meteorological profile data, and overseas PM2.5 simulated profile data; perform time alignment, spatial alignment, and normalization processing on the multi-source data; The vertical profile data of PM2.5 outside the country is calculated based on the simulated profile data and mass concentration data of PM2.5 outside the country; the vertical profile data of PM2.5 inside the country and the vertical profile data of PM2.5 outside the country are mapped to the same spatial grid system to obtain the vertical profile data of PM2.5 across the country. Using cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples, a Transformer model was trained to obtain a cross-border PM2.5 three-dimensional reconstruction model. Meteorological profile data of the cross-border area to be studied are input into the cross-border PM2.5 three-dimensional reconstruction model, and PM2.5 three-dimensional concentration field data of the cross-border area to be studied are output.
2. The method for three-dimensional reconstruction of PM2.5 in cross-border areas according to claim 1, characterized in that, The domestic PM2.5 vertical profile data is obtained by inverting the aerosol backscattering signal collected by lidar and combining it with the conversion relationship between PM2.5 and aerosol optical parameters. The overseas PM2.5 mass concentration data is obtained by monitoring overseas ground air quality monitoring stations. The cross-border meteorological profile data is obtained by meteorological models. The overseas PM2.5 simulated profile data is obtained by chemical transport models.
3. The method for three-dimensional reconstruction of PM2.5 in cross-border areas according to claim 1, characterized in that, The process of performing time alignment, spatial alignment, and normalization on the multi-source data includes: Map multi-source data with different time resolutions to a unified target time step node; By using horizontal interpolation and vertical resampling, multi-source data with different spatial resolutions and vertical layers are mapped to a unified three-dimensional spatial grid. The mapped data are normalized based on the sample mean and standard deviation.
4. The method for three-dimensional reconstruction of PM2.5 in cross-border areas according to claim 2, characterized in that, The calculation of overseas PM2.5 vertical profile data based on overseas PM2.5 simulated profile data and overseas PM2.5 mass concentration data includes: Based on the simulated profile data of the overseas region output by the chemical transport model; Using the aforementioned PM2.5 mass concentration data from overseas, the simulated profile data is corrected for deviation using a weighting function that varies with altitude, generating the overseas PM2.5 vertical profile data, which is represented as follows: ， in, This represents the vertical profile data of PM2.5 outside of China. This indicates simulated profile data from overseas. Represents the weighting function. This represents PM2.5 mass concentration data outside of China, where z represents altitude and t represents time.
5. The three-dimensional reconstruction method for PM2.5 in cross-border areas according to claim 2, characterized in that, The process of training a Transformer model using cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples includes: The cross-border meteorological profile data is constructed as an input feature vector arranged by altitude layer; The Transformer model transforms the input feature vector into query, key, and value vectors through a linear mapping, and uses the Softmax function to calculate attention scores based on the query, key, and value vectors. The mean squared error between the cross-border PM2.5 vertical profile predicted by the Transformer model and the supervised samples was used as the loss function to optimize the model parameters during training.
6. The three-dimensional reconstruction method for PM2.5 in cross-border areas according to claim 5, characterized in that, The process of inputting meteorological profile data of the cross-border region to be studied into the cross-border PM2.5 three-dimensional reconstruction model and outputting PM2.5 three-dimensional concentration field data of the cross-border region to be studied includes: Meteorological profile data corresponding to each spatial grid point in the cross-border area to be studied are obtained, and the meteorological profile data of the cross-border area to be studied is consistent with the model training stage in terms of vertical structure and feature dimension. The meteorological profile data corresponding to each spatial grid point are input into the cross-border PM2.5 three-dimensional reconstruction model, and the vertical profile data of PM2.5 at different height layers of the corresponding spatial grid points are output. The PM2.5 vertical profile data of each spatial grid point are stitched together to generate three-dimensional PM2.5 concentration field data of the cross-border area to be studied.
7. The method for three-dimensional reconstruction of PM2.5 in cross-border areas according to any one of claims 1-6, characterized in that, The PM2.5 three-dimensional concentration field data is used to characterize the spatial distribution characteristics of PM2.5 at different altitudes within a cross-border region, and is used for cross-border pollution transport analysis, pollution layer height identification, or regional air quality assessment.
8. A three-dimensional reconstruction device for PM2.5 in a cross-border area, characterized in that, The device includes: The data acquisition and processing module is used to acquire multi-source data from cross-border areas, including domestic PM2.5 vertical profile data, overseas PM2.5 mass concentration data, cross-border meteorological profile data, and overseas PM2.5 simulated profile data; and to perform time alignment, spatial alignment, and normalization processing on the multi-source data. The cross-border PM2.5 vertical profile construction module is used to calculate the cross-border PM2.5 vertical profile data based on the simulated profile data and mass concentration data of PM2.5 outside the country; and to map the vertical profile data of PM2.5 inside the country and the vertical profile data of PM2.5 outside the country to the same spatial grid system to obtain the cross-border PM2.5 vertical profile data. The model training module is used to train the Transformer model with cross-border meteorological profile data as input samples and cross-border PM2.5 vertical profile data as supervised samples to obtain a cross-border PM2.5 three-dimensional reconstruction model. The PM2.5 3D reconstruction module is used to input meteorological profile data of the cross-border area to be studied into the cross-border PM2.5 3D reconstruction model and output PM2.5 3D concentration field data of the cross-border area to be studied.
9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the three-dimensional reconstruction method for PM2.5 in cross-border areas as described in any one of claims 1-7.
10. A storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the three-dimensional reconstruction method for PM2.5 in cross-border areas as described in any one of claims 1-7.