Regional sand and dust weather process identification method and system influencing air quality

Abstract

The invention belongs to the technical field of environment monitoring, and relates to a method and system for identifying a regional sand and dust weather process influencing air quality, and the method comprises the steps: 1) collecting data which comprise monitoring data, meteorological data and satellite remote sensing image hour data of national control / district and county air quality stations; 2) calculating the sand and dust coverage area based on the satellite remote sensing influence hour data, and judging whether the sand and dust source area has sand or not based on the sand and dust coverage area; 3) based on the collected data, utilizing a plurality of machine learning models to judge whether each national control / district / county air quality station is influenced by sand and dust, and based on a clustering algorithm, clustering the national control / district / county air quality stations influenced by sand and dust to form a station group influenced by sand and dust; and 4) identifying the regional sand and dust weather process influencing the air quality based on the number of the air quality stations of the China Control / District County in the station group influenced by the sand and dust, so that the sand and dust weather process can be dynamically identified, and the manual identification cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental monitoring, and relates to a method and system for identifying sandstorm weather processes, and in particular to a method and system for identifying regional sandstorm weather processes that affect air quality. Background Art

[0002] The meteorological department, in its national standards for sandstorm severity, links visibility, wind speed, and PM10 concentration to identify sandstorm processes for earthquake disaster prevention and mitigation purposes through single-station observations and regional process analysis. Environmental protection authorities, through the Satellite Environment Application Center, conduct daily satellite remote sensing monitoring to determine the extent of sandstorms and understand their spatial distribution and movement. However, identifying sandstorm processes that impact air quality remains a mystery.

[0003] Therefore, there is an urgent need for a high-precision, adaptive automatic identification technology for sandstorm weather processes to improve the credibility of environmental data and provide a scientific basis for precise pollution control. Summary of the Invention

[0004] In response to the technical gap in the existing field of sandstorm weather process identification, the present invention aims to provide a method for identifying regional sandstorm weather processes that affect air quality. Based on multi-dimensional information and machine learning models, the method can dynamically identify sandstorm weather processes and reduce manual maintenance costs.

[0005] In order to achieve the above object, the present invention provides the following technical solutions: A method for identifying regional sandstorm weather processes that affect air quality, characterized by comprising the following steps: 1) Collect data, including monitoring data from national / district / county air quality stations, meteorological data, and hourly satellite remote sensing imagery data; 2) Based on the hourly satellite remote sensing impact data, the dust cover area is calculated using the statistical pixel method and dust source area is determined to be dusty when the dust cover area is greater than a specific threshold; 3) Based on the collected data, multiple machine learning models are used to determine whether national / county air quality stations are affected by the dust storm. Clustering algorithms are used to cluster national / county air quality stations affected by the dust storm to form groups of stations affected by the dust storm. 4) Identify regional dust weather processes that affect air quality based on the number of national controlled / county air quality stations in the station group affected by dust.

[0006] Preferably, in step 1), the monitoring data of the national / district and county air quality stations include hourly SO2 concentration, hourly O3 concentration, hourly CO concentration, hourly NO2 concentration, hourly PM10 concentration and hourly PM2.5 concentration, and the meteorological data include temperature, humidity, wind speed and wind direction.

[0007] Preferably, in step 1), after collecting the data, the collected data is cleaned, including data normalization and data filtering.

[0008] Preferably, step 2) specifically comprises: using a combination of brightness temperature difference values ​​of the three channels of 8.6 μm, 10.4 μm and 12.3 μm in the satellite remote sensing impact hourly data to distinguish between non-dust and dust, thereby identifying the hourly dust coverage range, and then calculating the dust coverage area based on the hourly dust coverage range using the statistical pixel method, and determining that dust is rising from the dust source area when the dust coverage area is greater than a specific threshold.

[0009] Preferably, in step 3), the multiple machine learning models include a support vector machine model, a principal factor analysis model, a logistic regression model, an Adaboost model, a KNN model, a perceptron model, a ridge regression model and a random forest model, and when the judgment results of the multiple machine learning models are inconsistent, the judgment results of the multiple machine learning models are voted based on voting rules to determine the final judgment result.

[0010] Preferably, in step 3), the national / district / county air quality stations affected by sandstorms are clustered based on the DBSCAN clustering algorithm to form a station group affected by sandstorms, which specifically includes: 31) Select a national / county air quality station that has not been visited and is affected by dust; 32) Check the number of national / district / county air quality stations affected by dust and sandstorms within a certain distance from the unvisited national / district / county air quality station affected by dust and sandstorms. If the number of stations is ≥ 2, mark the unvisited national / district / county air quality station affected by dust and sandstorms as a core station. Otherwise, return to step 31); 33) Based on the core station, a station group affected by the dust is formed. Starting from the core station, all national / district / county air quality stations affected by the dust within a certain distance from the core station are found and added to the station group affected by the dust; 34) Traversing the dust-affected site group, for each newly added national / county air quality site affected by dust, find a national / county air quality site affected by dust that is not already added to the dust-affected site group within a certain distance, and add the found national / county air quality site affected by dust to the dust-affected site group; 35) Repeat step 34) until there are no new national / county air quality stations affected by the dust, and the group of stations affected by the dust is established.

[0011] Preferably, the step 4) is specifically as follows: 41) When a site group affected by sand and dust appears and the number of sites in the site group affected by sand and dust is greater than or equal to a certain value, the process of determining regional sand and dust impact begins; 42) When the number of stations in the station group affected by the sandstorm is greater than or equal to a certain value and partially overlaps with the stations in the station group affected by the sandstorm in the previous hour, it is determined that the regional sandstorm impact process continues; 43) When the number of sites in the site group affected by the dust is less than a certain value for two consecutive days, the process of determining that the regional dust impact has ended.

[0012] In addition, the present invention also provides a system for identifying regional sandstorm weather processes that affect air quality, which is characterized by comprising: The data collection module is used to collect monitoring data from national / district / county air quality stations, meteorological data, and hourly satellite remote sensing image data; a dust source area dust emission identification module, which is used to calculate the dust coverage area using a statistical pixel method based on the satellite remote sensing impact hourly data and determine that dust emission is in the dust source area when the dust coverage area is greater than a specific threshold; An impact area determination module, which uses multiple machine learning models based on collected data to determine whether national / county air quality stations are affected by the dust storm and clusters the national / county air quality stations affected by the dust storm based on a clustering algorithm to form groups of stations affected by the dust storm; A module for identifying processes affecting regional air quality due to sandstorms is used to identify regional sandstorm weather processes affecting air quality based on the number of national / district / county air quality stations in the group of stations affected by sandstorms.

[0013] Furthermore, the present invention also provides a device for identifying regional sandstorm weather processes that affect air quality, which is characterized by comprising: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for identifying regional sandstorm weather processes that affect air quality as described above. Finally, the present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the method for identifying regional sandstorm weather processes that affect air quality as described above are implemented.

[0014] Compared with the prior art, the method and system for identifying regional sandstorm weather processes that affect air quality of the present invention have one or more of the following beneficial technical effects: 1. The present invention can identify regional sandstorm weather processes that affect air quality, filling a gap in the industry.

[0015] 2. The present invention performs recognition based on multiple data and machine learning models, with high recognition accuracy and a process recognition rate close to 100%. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the method for identifying regional sandstorm weather processes that affect air quality according to the present invention.

[0017] Figure 2 It is a schematic diagram of the composition of the regional sandstorm weather process identification system that affects air quality of the present invention. DETAILED DESCRIPTION

[0018] Before describing in detail any embodiment of the present invention, it should be understood that the present invention is not limited in its application to the construction and arrangement details of the components set forth in the following description or illustrated in the following figures. The present invention is capable of other embodiments and can be practiced or carried out in various ways. In addition, it should be understood that the words and terms used herein are for descriptive purposes and should not be considered restrictive. The use of "including" or "having" and their variations herein is intended to cover the items and their equivalents set forth below and additional items. Unless otherwise specified or limited, the terms "mounted", "connected", "supported" and "coupled" and their variations are used broadly and cover direct mounting and indirect mounting, connection, support and coupling. In addition, "connected" and "coupled" are not limited to physical or mechanical connections or couplings. Furthermore, on the first hand, in the disclosure of the present invention, the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like to indicate orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore the above terms cannot be understood as limitations on the present invention; on the second hand, the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" cannot be understood as a limitation on the quantity.

[0019] Figure 1 The flow chart of the method for identifying regional sandstorm weather processes that affect air quality according to the present invention is shown. Figure 1 As shown, the method for identifying regional sandstorm weather processes that affect air quality of the present invention includes the following steps: 1. Data collection.

[0020] When identifying regional dust storms that affect air quality, the first step is to collect relevant data. In this invention, the collected data includes monitoring data from national / district / county air quality stations, meteorological data, and hourly satellite remote sensing image data.

[0021] Among them, the monitoring data of the national / district and county air quality stations include hourly SO2 concentration, hourly O3 concentration, hourly CO concentration, hourly NO2 concentration, hourly PM10 concentration and hourly PM2.5 concentration, and the meteorological data include temperature, humidity, wind speed and wind direction.

[0022] Furthermore, in the present invention, the collected raw data can be cleaned, including data normalization (normalization of data type, decimal places, units, etc.) and data filtering (filtering out obviously erroneous data), etc., so as to ensure the standardization and accuracy of the data.

[0023] 2. Identification of sandstorm source areas.

[0024] In the present invention, the dust coverage area is calculated based on the satellite remote sensing impact hourly data, and whether the dust source area is dusty is determined based on the dust coverage area.

[0025] Dust is typically composed of fine mineral particles (such as quartz and clay). These particles exhibit distinct absorption and scattering properties across the electromagnetic spectrum, unlike other atmospheric components such as clouds, water vapor, and aerosols. Dust exhibits specific absorption and scattering characteristics from the visible to infrared wavelengths, particularly in the mid-infrared (3-14 microns). Dust particles typically have strong absorption in the 8-13 micron band. By analyzing the differences in radiation across these bands, it is possible to effectively distinguish dust from other atmospheric components, enabling dust monitoring and identification.

[0026] The satellite data from the Sunflower-8 satellite features high temporal resolution, high image quality, and rich spectral information. The satellite has 16 channels covering visible light, near infrared, and thermal infrared. Its Advanced Hierarchical Imager (AHI) can scan five regions: Full Disk (global), Japan Area (two regions of Japan), Specific Area (two regions), and Landmark Area (two regions). Therefore, the present invention calculates dust cover area based on hourly remote sensing impact data from the Sunflower-8 satellite.

[0027] The basic principle for extracting dust ranges based on hourly remote sensing data from the Sunflower-8 satellite is to distinguish between non-dust and non-sand areas using a combination of brightness temperature differences across the 8.6μm, 10.4μm, and 12.3μm channels within the 8-13μm thermal infrared atmospheric window. Based on this theory, pseudo-color composite images generated from these bands can effectively extract the dust range.

[0028] 1. Transmit data from the official Sunflower satellite server. The data file name should be something like "NC_H08_20240301_0000_R21_FLDK.02401_02401.nc." This example shows the satellite image's transit time as 00:00 UTC on March 1, 2024 (8:00 Beijing Time). Select R21_FLDK for full coverage and 02401 for low resolution (06001 for high resolution).

[0029] 2. For the three loads sensitive to sand and dust, the design bands tbb_11, tbb_13, and tbb_15 (where tbb_11 corresponds to an 8.6μm channel, tbb_13 corresponds to a 10.4μm channel, and tbb_15 corresponds to a 12.3μm channel) are combined: B3 = tbb_15-tbb_13; B2 = tbb_13-tbb_11; B1 = tbb_13, forming the three bands B3, B2, and B1.

[0030] 3. Normalize the three bands B3, B2, and B1 to facilitate adjustment of the optimal threshold in subsequent experiments. The formula for normalization is: ,in For three combined bands, the nanmin and nanamx functions are used to ignore the influence of null values ​​in the image.

[0031] B3 = (B3-nanmin(B3(:))). / (nanmax(B3(:))-nanmin(B3(:))); B2= (B2-nanmin(B2(:))). / (nanmax(B2(:))-nanmin(B2(:))); B1 = (B1-nanmin(B1(:))). / (nanmax(B1(:))-nanmin(B1(:))).

[0032] 4. Connect the three bands B3, B2, and B1, and crop the specified area based on the number of rows and columns. In the R21_FLDK full-coverage and 02401 low-resolution sunflower image, the position of the specified area in the two-dimensional array is roughly 100-900 rows and 1-1200 columns. Based on this, crop the observation data within the specified range: rgb=cat(3,B3(100:900,1:1200),B2(100:900,1:1200),B1(100:900,1:1200) ) can generate a pseudo-color image showing the characteristics of sand and dust.

[0033] 5. Using multi-threshold joint identification, after multiple rounds of testing, the threshold for sand and dust extraction is obtained. The part with 0.65<=B3<=1&0.01<=B2<=0.4&0.4<=B1<=1 is retained, and the part that does not meet the conditions is set to white (i.e. rgb=[1,1,1]).

[0034] Based on the hourly dust coverage range identified by the above method, the dust coverage area is calculated using the statistical pixel method based on the hourly dust coverage range.

[0035] Dust coverage area = .

[0036] Where m is the row resolution of the satellite remote sensing impact hourly data; n is the column resolution of the satellite remote sensing impact hourly data; pixel(i,j) is whether the (i,j)th pixel in the satellite remote sensing impact hourly data is determined to be sand and dust. If it is determined to be sand and dust, it is 1, otherwise it is 0. It is the actual area represented by each pixel in the satellite remote sensing impact hourly data.

[0037] Then, it is determined whether the dust coverage area is larger than a specific threshold (for example, 100,000 square kilometers). If it is larger than the specific threshold, it is determined that the dust source area has dust emission; otherwise, it is determined that the dust source area has not dust emission.

[0038] 3. Determination of the affected area.

[0039] After determining the dust source area, the next step is to determine the affected area. In this paper, based on collected data, multiple machine learning models are used to determine whether national / county air quality stations are affected by the dust. A clustering algorithm is then used to cluster affected national / county air quality stations to form groups of stations affected by the dust.

[0040] Among them, the multiple machine learning models include support vector machine model, principal factor analysis model, logistic regression model, Adaboost model, KNN model, perceptron model, ridge regression model and random forest model.

[0041] 1. Support vector machine model.

[0042] The support vector machine model maps the high-dimensional space through the kernel function and finds the optimal hyperplane to separate dust / non-dust samples.

[0043] In the present invention, the parameters of the support vector machine model are set as: (1) Kernel function: Use the radial basis function (RBF) kernel to map samples to a high-dimensional space, making linearly inseparable data separable; (2) Penalty coefficient C: set to 1.0.

[0044] 2. Principal factor analysis model.

[0045] The principal factor analysis model extracts the main factors through dimensionality reduction and combines it with logistic regression to build a classification model.

[0046] In the present invention, the parameters of the main factor analysis model are set as: (1) Number of principal components: 8 dimensions.

[0047] 3. Logistic regression model.

[0048] The logistic regression model is based on the linear probability model and outputs the probability of occurrence of sandstorm impact moments.

[0049] In the present invention, the parameters of the logistic regression model are set as: (1) Regularization: L2 regularization (λ=1); (2) Classification threshold: 0.5.

[0050] 4. Adaboost model.

[0051] The Adaboost model improves the overall classification performance by iteratively weighting weak classifiers (decision stumps).

[0052] In the present invention, the parameters of the Adaboost model are set as: (1) Base learner: a decision tree with a maximum depth of 1; (2) Number of iterations: 200 iterations.

[0053] 5. KNN model.

[0054] The KNN model determines the category based on the sample distance (Euclidean distance) voting.

[0055] In the present invention, the parameters of the KNN model are set as: (1) K value: K=15; (2) Distance weight: The weight is a constant.

[0056] 6. Perceptron model.

[0057] The perceptron model is a linear binary classification model that drives weight updates through misclassification.

[0058] In the present invention, the parameters of the perceptron model are set as: (1) Initial learning rate: 1.0; (2) Maximum iterations: uniformly set to 1000, using the early stopping method (tolerating no improvement after 10 iterations).

[0059] 7. Ridge regression model.

[0060] The ridge regression model is linear regression + L2 regularization, which outputs continuous values ​​and then classifies them through thresholds.

[0061] In the present invention, the parameters of the ridge regression model are set as: (1) Regularization coefficient α: 1.0; (2) Optimization accuracy: 10 -2 ; (3) Solver: lsqr, which uses a dedicated regularized least squares algorithm for solving.

[0062] 8. Random forest model.

[0063] The random forest model integrates multiple decision trees and outputs the final result through a voting mechanism.

[0064] In the present invention, the parameters of the random forest model are set as: (1) Number of trees: set to 100; (2) Maximum depth: limited to 5 layers (to prevent overfitting).

[0065] In the present invention, only the above-mentioned machine learning model is used, rather than being improved, and therefore, a more detailed description thereof is omitted. Of course, prior to use, the support vector machine model, principal factor analysis model, logistic regression model, Adaboost model, KNN model, perceptron model, ridge regression model, and random forest model can be trained.

[0066] During training, historical data is first collected, and hourly city-level data is calculated based on the hourly data from nationally controlled air quality stations in the historical data. The historical data includes monitoring data from nationally controlled air quality stations, meteorological data, and data manually identified as sand and dust impact. The monitoring data from the nationally controlled air quality stations includes hourly SO2 concentration, hourly O3 concentration, hourly CO concentration, hourly NO2 concentration, hourly PM10 concentration, and hourly PM2.5 concentration. The hourly city-level data includes the average hourly SO2 concentration, hourly O3 concentration, hourly CO concentration, hourly NO2 concentration, hourly PM10 concentration, and hourly PM2.5 concentration for nationally controlled air quality stations in each city. The meteorological data includes temperature, humidity, wind speed, and wind direction. The data manually identified as sand and dust impact includes manual identification results (such as "dust impact time" or "non-dust impact time").

[0067] After collecting historical data (hourly city-level data, meteorological data, and data on manual identification of dust impact), the support vector machine model, principal factor analysis model, logistic regression model, Adaboost model, KNN model, perceptron model, ridge regression model, and random forest model are trained based on the historical data. Specifically, the support vector machine model, principal factor analysis model, logistic regression model, Adaboost model, KNN model, perceptron model, ridge regression model, and random forest model are trained using the hourly city-level data (hourly average SO2 concentration, hourly average O3 concentration, hourly average CO concentration, hourly average NO2 concentration, hourly average PM10 concentration, and hourly average PM2.5 concentration at each city's nationally monitored air quality station) and the city's temperature, humidity, wind speed, and wind direction as input, and the dust impact manual identification data as output to train the support vector machine model, principal factor analysis model, logistic regression model, Adaboost model, KNN model, perceptron model, ridge regression model, and random forest model, respectively.

[0068] After training, the collected data (i.e., hourly SO2 concentrations, hourly O3 concentrations, hourly CO concentrations, hourly NO2 concentrations, hourly PM10 concentrations, hourly PM2.5 concentrations, and the temperature, humidity, wind speed, and wind direction of the national / county air quality stations) is input into the trained support vector machine model, principal factor analysis model, logistic regression model, Adaboost model, KNN model, perceptron model, ridge regression model, and random forest model. These models then output "dust-affected time" or "non-dust-affected time" to determine whether the national / county air quality station is affected by dust. Specifically, if the output is "dust-affected time," the national / county air quality station is determined to be affected by dust; if the output is "non-dust-affected time," the national / county air quality station is determined to be unaffected by dust.

[0069] Of course, for the same national / district / county air quality station, the determination results of the multiple machine learning models may be inconsistent. When the determination results of the multiple machine learning models are inconsistent, the determination results of the multiple machine learning models are voted on based on the voting rules to determine the final determination result.

[0070] In the present invention, the voting rule is that the minority follows the majority, and when the votes are unanimous, it is determined to be a "time not affected by sandstorms." For example, if four models determine it to be a "time affected by sandstorms" and another four models determine it to be "not affected by sandstorms," ​​then it is determined to be a "time not affected by sandstorms." If five models determine it to be a "time affected by sandstorms" and another three models determine it to be "not affected by sandstorms," ​​then it is determined to be a "time affected by sandstorms." If three models determine it to be a "time affected by sandstorms" and another five models determine it to be "not affected by sandstorms," ​​then it is determined to be a "time not affected by sandstorms."

[0071] After obtaining the national / district / county air quality stations affected by the dust storm, the present invention further clusters the national / district / county air quality stations affected by the dust storm using a clustering algorithm to form a group of stations affected by the dust storm. The area where the group of stations affected by the dust storm is located is the dust storm-affected area.

[0072] Specifically, the national / county air quality stations affected by the dust storm were clustered based on the DBSCAN clustering algorithm to form a group of stations affected by the dust storm, which specifically includes: 1. Select an unvisited national / district / county air quality station affected by dust. When executing step 1 for the first time, a national / district / county air quality station affected by dust can be randomly selected as the unvisited national / district / county air quality station affected by dust.

[0073] 2. Check the number of national / district / county air quality stations affected by dust and sandstorms within a certain distance (e.g., 10 km) from the unvisited national / district / county air quality station. If the number of stations is ≥ 2, mark the unvisited national / district / county air quality station affected by dust and sandstorms as a core station. Otherwise, return to step 1 and continue searching for a core station until a core station is found or all national / district / county air quality stations affected by dust and sandstorms are traversed. If no core station is found after traversing all national / district / county air quality stations affected by dust and sandstorms, it indicates that there is no regional dust and sandstorm weather process affecting air quality.

[0074] 3. Based on the core site, a site group affected by sandstorms is formed. Starting from the core site, all national / district / county air quality sites affected by sandstorms within a certain distance (for example, 10 km) from the core site are found, and all the national / district / county air quality sites affected by sandstorms found are added to the site group affected by sandstorms.

[0075] 4. Traverse the group of sites affected by sandstorms. For each newly added national / county air quality site affected by sandstorms, find the national / county air quality sites affected by sandstorms that are within a certain distance (for example, 10 km) but are not added to the group of sites affected by sandstorms, and add the found national / county air quality sites affected by sandstorms to the group of sites affected by sandstorms.

[0076] 5. Repeat step 4 until there are no new national / county air quality stations affected by sandstorms, and the group of stations affected by sandstorms is established.

[0077] 4. Identification of processes affecting regional air quality by sandstorms.

[0078] Identifying regional dust weather processes that affect air quality based on the number of national controlled / county air quality stations in the dust-affected station group specifically includes: 1. When a group of sites affected by sandstorms appears and the number of national air quality control / county air quality sites in the group of sites affected by sandstorms is greater than or equal to a certain value (for example, 5), the process of determining the regional sandstorm impact begins.

[0079] 2. When the number of national / county air quality stations in the group of stations affected by sandstorms is greater than or equal to a certain value (for example, 5), and partially overlaps with the stations in the group of stations affected by sandstorms in the previous hour (for example, there are 3 overlapping national / county air quality stations, that is, there are 3 national / county air quality stations in the group of stations affected by sandstorms at the current moment and in the previous hour), it is determined that the regional sandstorm impact process continues.

[0080] 3. When the number of national air quality control stations / county air quality stations in the group of stations affected by sandstorms is less than a certain value (for example, 5) for two consecutive days, the process of determining that the regional sandstorm impact has ended.

[0081] For example, when a site group affected by sandstorms appears in the territory, and the number of sites in the site group affected by sandstorms is greater than or equal to 5, it is recorded as the beginning of sandstorms; when the number of sites in the site group affected by sandstorms at a certain moment is greater than or equal to 5, and the number of sites overlapping with the site group affected by sandstorms in the previous hour is 3, it is recorded as the continuation of sandstorms; when the number of sites in the site group affected by sandstorms is less than 5 for two consecutive days, it is recorded as the end of the sandstorm impact process.

[0082] In the present invention, after identifying the process of regional air quality being affected by dust, the effect of the process identification can be further evaluated, that is, the accuracy of the identification of the process of regional air quality being affected by dust is evaluated.

[0083] In the present invention, the accuracy of the identification of the process of the air quality in the area being affected by sand and dust is evaluated by the process identification rate. .

[0084] After evaluating the sandstorm weather identification process in a certain region in the autumn and winter of 2024 and the spring of 2025, it was found that the accuracy of the regional sandstorm weather process identification method affecting air quality of the present invention was close to 100%.

[0085] Figure 2 The schematic diagram of the system for identifying regional sandstorm weather processes that affect air quality according to the present invention is shown. Figure 2 As shown, the regional sandstorm weather process identification system affecting air quality of the present invention includes: 1. Data collection module.

[0086] The data collection module is used to collect monitoring data from national / district and county air quality stations, meteorological data, and hourly satellite remote sensing image data.

[0087] 2. Sand and dust source area sand emission identification module.

[0088] The dust source area dust emission identification module is used to calculate the dust coverage area using the statistical pixel method based on the satellite remote sensing impact hourly data and determine that dust emission is in the dust source area when the dust coverage area is greater than a specific threshold; Specifically, a combination of the brightness temperature difference values ​​of the three channels of 8.6μm, 10.4μm and 12.3μm in the satellite remote sensing impact hourly data is used to distinguish between non-dust and dust, thereby identifying the hourly dust coverage range. Then, based on the hourly dust coverage range, the dust coverage area is calculated using the statistical pixel method, and when the dust coverage area is greater than a specific threshold, it is determined that sand is rising in the dust source area.

[0089] 3. Impact area determination module.

[0090] The impact area determination module is used to determine whether each national / district / county air quality station is affected by sandstorms based on the collected data using multiple machine learning models, and to cluster the national / district / county air quality stations affected by sandstorms based on a clustering algorithm to form a group of stations affected by sandstorms.

[0091] 4. Module for identifying processes affecting regional air quality by sand and dust.

[0092] The module for identifying processes affecting regional air quality due to sandstorms is used to identify regional sandstorm weather processes affecting air quality based on the number of national / county air quality stations in the group of stations affected by sandstorms.

[0093] In addition, the present invention also provides a device for identifying regional sandstorm weather processes that affect air quality, which includes: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method for identifying regional sandstorm weather processes that affect air quality as described above. Finally, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for identifying regional sandstorm weather processes that affect air quality as described above.

[0094] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art may, based on the principles of the present invention, modify or replace the technical solutions of the present invention with equivalents without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for identifying regional sandstorm weather processes that affect air quality, characterized in that: The following steps are involved: 1) Collect data, including monitoring data from national / district / county air quality stations, meteorological data, and hourly satellite remote sensing image data; 2) calculating the dust cover area based on the satellite remote sensing impact hourly data and determining whether the dust source area is causing dust emission based on the dust cover area; 3) Based on the collected data, multiple machine learning models are used to determine whether national / county air quality stations are affected by the dust storm. Clustering algorithms are used to cluster national / county air quality stations affected by the dust storm to form groups of stations affected by the dust storm. 4) Identify regional dust weather processes that affect air quality based on the number of national controlled / county air quality stations in the station group affected by dust.

2. The method for identifying regional sandstorm weather processes that affect air quality according to claim 1, characterized in that: In step 1), the monitoring data of the national / district and county air quality stations include hourly SO2 concentration, hourly O3 concentration, hourly CO concentration, hourly NO2 concentration, hourly PM10 concentration and hourly PM2.5 concentration, and the meteorological data include temperature, humidity, wind speed and wind direction.

3. The method for identifying regional sandstorm weather processes that affect air quality according to claim 2, characterized in that: In the step 1), after collecting the data, the collected data is cleaned, including data normalization and data filtering.

4. The method for identifying regional sandstorm weather processes that affect air quality according to claim 1, characterized in that: The step 2) specifically comprises: using a combination of brightness temperature differences of the three channels (8.6μm, 10.4μm, and 12.3μm) in the satellite remote sensing impact hourly data to distinguish between non-dust and dust, thereby identifying the hourly dust coverage range; then, based on the hourly dust coverage range, using the statistical pixel method to calculate the dust coverage area; and determining that dust is rising from a dust source area when the dust coverage area is greater than a specific threshold.

5. The method for identifying regional sandstorm weather processes that affect air quality according to claim 1, characterized in that: In step 3), the multiple machine learning models include a support vector machine model, a principal factor analysis model, a logistic regression model, an Adaboost model, a KNN model, a perceptron model, a ridge regression model and a random forest model, and when the judgment results of the multiple machine learning models are inconsistent, the judgment results of the multiple machine learning models are voted based on voting rules to determine the final judgment result.

6. The method for identifying regional sandstorm weather processes that affect air quality according to claim 1, characterized in that: In step 3), the national / district / county air quality stations affected by the dust storm are clustered based on the DBSCAN clustering algorithm to form a station group affected by the dust storm, which specifically includes: 31) Select a national / county air quality station that has not been visited and is affected by the dust storm; 32) Check the number of national / district / county air quality stations affected by dust and sandstorms within a certain distance from the unvisited national / district / county air quality station affected by dust and sandstorms. If the number of stations is ≥ 2, mark the unvisited national / district / county air quality station affected by dust and sandstorms as a core station. Otherwise, return to step 31); 33) Based on the core station, a station group affected by the dust is formed. Starting from the core station, all national / district / county air quality stations affected by the dust within a certain distance from the core station are found and added to the station group affected by the dust; 34) Traversing the dust-affected site group, for each newly added national / county air quality site affected by dust, finding a national / county air quality site affected by dust that is not already in the dust-affected site group and adding the found national / county air quality site affected by dust to the dust-affected site group; 35) Repeat step 34) until there are no new national / county air quality stations affected by the dust, and the group of stations affected by the dust is established.

7. The method for identifying regional sandstorm weather processes that affect air quality according to claim 1, characterized in that: The step 4) is specifically as follows: 41) When a site group affected by sand and dust appears and the number of sites in the site group affected by sand and dust is greater than or equal to a certain value, the process of determining regional sand and dust impact begins; 42) When the number of stations in the station group affected by the sandstorm is greater than or equal to a certain value and partially overlaps with the stations in the station group affected by the sandstorm in the previous hour, it is determined that the regional sandstorm impact process continues; 43) When the number of sites in the site group affected by the dust is less than a certain value for two consecutive days, the process of determining that the regional dust impact has ended.

8. A regional sandstorm weather process identification system that affects air quality, characterized by: include: The data collection module is used to collect monitoring data from national / district / county air quality stations, meteorological data, and hourly satellite remote sensing image data; a dust source area dust emission identification module, configured to calculate the dust coverage area based on the satellite remote sensing impact hourly data and determine whether the dust source area is dust emission based on the dust coverage area; An impact area determination module, which uses multiple machine learning models based on collected data to determine whether national / county air quality stations are affected by the dust storm and clusters the national / county air quality stations affected by the dust storm based on a clustering algorithm to form groups of stations affected by the dust storm; A module for identifying processes affecting regional air quality due to sandstorms is used to identify regional sandstorm weather processes affecting air quality based on the number of national / district / county air quality stations in the group of stations affected by sandstorms.

9. A device for identifying regional sandstorm weather processes that affect air quality, characterized in that: include: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for identifying regional sandstorm weather processes that affect air quality as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for identifying a regional sandstorm weather process affecting air quality as described in any one of claims 1 to 7 are implemented.

Images