Reconstruction density clustering-based atmospheric pollutant propagation path discovery method

Abstract

Aiming at composite risks caused by atmospheric pollution cross-regional propagation, a traditional monitoring and model method has the limitations of insufficient data coverage, terrain simplification distortion, dynamic response lag and the like. The invention provides an atmospheric pollutant propagation path discovery technology based on reconstruction density clustering. Pollution space distribution is reconstructed through a density accumulation strategy, air quality information and marginal region recognition capability are enhanced, and meanwhile, a pollutant propagation process is reversely deduced to obtain a pollution source; introducing a relative distance parameter to simulate the influence of a complex terrain on diffusion; and dynamically adjusting the propagation path in combination with real-time meteorological data. According to the method, multi-source data fusion traceability is realized, limitation of a fixed monitoring station is broken through, an effect of discovering a dirt propagation path is relatively good, and a key technical support is provided for regional collaborative governance and emergency response.

Description

Technical Field

[0001] The present invention relates to the field of atmospheric pollutant propagation analysis, and in particular to a method for discovering atmospheric pollutant propagation paths based on reconstructed density clustering. Background Art

[0002] Air pollution has become a global environmental crisis. Its hazards are not only manifested in excessive concentrations of pollutants such as PM2.5 and O3, but also in the systemic risk of complex pollution exacerbated by the cross-regional spread of pollutants. For example, during periods of heavy winter pollution in the Beijing-Tianjin-Hebei region, pollutants can be transported through high altitudes and affect the Yangtze River Delta, leading to frequent regional pollution incidents. Accurately identifying pollution transmission pathways is crucial for scientific prevention and control. Model simulations can identify pollution transmission channels and key receptor areas, optimizing resource allocation. During sudden pollution incidents, path discovery can quickly pinpoint pollution sources and their spread, improving emergency response efficiency. Furthermore, quantifying the cross-regional contribution of pollution can provide a basis for coordinated regional governance and promote the establishment of ecological compensation mechanisms.

[0003] Traditional pollution path discovery methods rely on fixed monitoring stations and empirical formulas, and have obvious shortcomings: 1. Insufficient monitoring data coverage and difficulty in identifying pollution sources: Relying on only a few fixed monitoring stations makes it difficult to capture the diffusion trajectory of pollutants in rural and mountainous areas, resulting in missed reports or misjudgments of pollution events; 2. Model simplification and distortion: Simplified algorithms such as the Gaussian diffusion model are often used, assuming that pollutant concentrations are uniformly distributed. This cannot reflect the impact of complex terrain (such as valleys and urban canyons) on airflow and diffusion, resulting in large errors; 3. Dynamic response lag: Relying on historical meteorological data and manual analysis, it is impossible to capture sudden meteorological changes in real time (such as calm weather and sudden changes in wind direction), resulting in delayed or ineffective warnings. These limitations make it difficult for traditional methods to meet the needs of modern pollution prevention and control, and there is an urgent need to introduce new technologies to improve discovery capabilities.

[0004] Therefore, to address the above problems, a method for discovering the propagation paths of atmospheric pollutants based on reconstructed density clustering is proposed. This method considers the insufficient data coverage and the law of pollutant propagation and uses the density accumulation method to reconstruct the density of the polluted area; considers the problem of model simplification distortion and uses methods such as adding relative distance to simulate complex terrain; considers the dynamic response problem model and incorporates sudden meteorological changes into the modeling considerations, so that the method can more accurately identify the pollution source and discover its propagation path. Summary of the Invention

[0005] To solve the above problems, the present invention provides a method for discovering the propagation paths of atmospheric pollutants based on reconstructed density clustering. The method obtains data information from the monitored air quality data and discovers the propagation paths of pollutants from the data information. The method includes the following processing steps:

[0006] Step 1: Input the processed data set, which is the air quality index (AQI) of each city. The data format is {city, AQI}. Each data item represents the air quality index (AQI) of a city in a certain period of time. Input the required parameters: nearest neighbor k and threshold distance. .

[0007] Step 2: Consider complex geographical factors to obtain the information distance between air quality data, and further obtain the density information of the data set from the information distance. The closer the information distance, the more similar the two air quality data are. For the case where there is a complex geographical environment between the two data points, for example, the Qinling Mountains can effectively block the north-south diffusion of pollutants, making the concentration of pollutants on the northern slope significantly higher than that on the southern slope; the Sichuan Basin is blocked by the surrounding mountains, and pollutants tend to accumulate in the basin under the inversion layer in winter. Based on such geographical features, if there is an insurmountable terrain such as continuous mountains between the two cities, the straight-line distance is short, but the actual diffusion path of pollutants cannot be directly reached due to the terrain blocking. Therefore, the Euclidean distance is used to calculate the distance between all data points. For the case where there is a geographical space blocking the spread of pollutants, the distance is modified to the maximum distance so that the label cannot be propagated between data points with a large distance. Afterwards, each data point is separated according to the calculated distance between the data points. The pollutant information is converted into density information, and the calculation formula is as follows:

[0008]

[0009] in Indicates a point k nearest neighbors, Indicates a point and point distance.

[0010] Step 3: Use density accumulation to enhance air quality information and reverse engineer the pollutant propagation process to identify pollution sources. Insufficient information in suburban and mountainous areas has a profound impact on the spread of atmospheric pollutants. Furthermore, insufficient information can lead to delayed responses to sudden incidents at pollution sources, such as inactive warnings or low emergency response levels, exacerbating pollution levels. Long-term data loss can also hinder the formulation and implementation of environmental protection policies, making them less targeted and effective. Ultimately, these issues will weaken the effectiveness of regional joint prevention and control, exacerbate misjudgments of environmental risks, and pose a serious threat to overall improvements in regional air quality. Using a density accumulation strategy to reconstruct the density of each data point can enhance the density of the data point. In this method, density represents the concentration of the pollutant, thereby supplementing the air quality data. Simultaneously, the accumulation process reversely simulates the propagation of pollutants from high to low concentrations, effectively utilizing air quality information to simulate pollutant propagation. Therefore, through density accumulation, pollutant concentrations are enhanced and the pollutant source is reversed. This can address misjudgments caused by insufficient information in suburban and mountainous areas and effectively identify pollution sources. The specific implementation process is as follows: Starting from the lowest-density, unvisited point, Start traversing towards the highest density point and update the density condition as shown below. For each data point ,exist of Search The lowest density of unvisited points within the distance ,Will Density Add to Density superior, Merge 、 and Set to visited, Become new 、 Become new , and then continue to recursively add data points density, until there is no point that meets the conditions, then the final The point with the highest density is a pollution source. All points in are marked as visited, and then continue to traverse, looking for the next unvisited low-density point, and start a new round of density update. The density update formula is shown below, where k and The input parameters.

[0011]

[0012] Step 4: Correct the pollutant propagation path according to the wind direction. A sudden change in wind direction will cause the direction of pollutant propagation to change. This change may cause originally unaffected areas to suddenly face pollution risks, while the polluted areas that were originally downwind may temporarily get rid of the pollution impact due to the change in wind direction. If there is a problem of sudden meteorological events such as a change in wind direction, it is necessary to quickly adjust the pollution source of the relevant data points at that location, that is, trace back to the nearest high-density monitoring point to redefine the source and propagation path of the pollutants, so as to more effectively respond to pollution incidents. Therefore, when obtaining the nearest high-density point for each data point as its pollutant propagation source, the pollutant source of the relevant data points is adjusted at that location in response to sudden meteorological conditions and other issues. For sudden meteorological problems such as winds in the same direction of level 3 or above, the source of pollutant propagation in the sudden meteorological city is set to the city where the directional wind is located, and the density of the sudden meteorological city is set to the density of the city where the directional wind is located minus the minimum density difference.

[0013] Step 5: Obtain the final pollutant propagation path. After integrating the aforementioned information on geographical and wind direction factors, the point with the highest density obtained in Step 3 is used as the pollutant source. Labels are assigned using the nearest high-density method obtained in Step 4. The label-assigned path is the pollutant propagation path. The label assignment method is as follows: First, assign a label to the cluster center point. Then, in descending order of density, assign the label of each data point to the label of its nearest high-density point. The label-assigned path is the pollutant propagation path.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] (1) Use data modeling to simulate pollutant source tracing, efficiently integrate multi-source data and reveal underlying patterns, significantly improving data utilization efficiency and decision-making accuracy.

[0016] (2) A density accumulation strategy was proposed to enhance air quality data information and infer the source of pollutants, solving the problem of marginal areas or areas with low pollution concentration being ignored, and increasing the possibility of all regions participating in the discovery of pollutant transmission paths.

[0017] (3) Adding relative distances to simulate complex geographical terrains such as mountains that cannot be crossed makes the model take geographical factors into account and more realistic.

[0018] (4) Modify the nearest high-density city according to the sudden wind direction conditions, and then correct the propagation path to make the model more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of a method for discovering the propagation paths of atmospheric pollutants based on reconstructed density clustering according to the present invention.

[0020] Figure 2This is a density accumulation effect diagram of an atmospheric pollutant propagation path discovery method based on reconstructed density clustering according to the present invention.

[0021] Figure 3 Schematic diagram of clustering results of the Zigzag dataset under different algorithms of the present invention.

[0022] Figure 4 This is a schematic diagram of the clustering results of the Pathbased dataset of the present invention under different algorithms.

[0023] Figure 5 This is the urban air quality data of Gansu Province on May 2, 2025 in an implementation example of the atmospheric pollutant propagation path discovery method based on reconstructed density clustering of the present invention.

[0024] Figure 6 This is a pollutant flow path diagram of an implementation example of the method for discovering atmospheric pollutant propagation paths based on reconstructed density clustering of the present invention.

[0025] Figure 7 This is a map of urban pollutant propagation discovery in Gansu Province on May 2, 2025, based on a method for discovering the propagation path of atmospheric pollutants based on reconstructed density clustering according to the present invention. DETAILED DESCRIPTION

[0026] Method validation:

[0027] To validate the effectiveness of the algorithm, we compared our proposed method with nine other methods: DBSCAN, k-Means, OPTICS, Spectral Clustering (SC), DPC, MDPC, DN, Extreme, and DPC-CE, on 15 complex datasets and seven UCI datasets. The 15 complex datasets are categorized as follows: Jain, Mouse, Unbalance, and Varydensity are datasets containing clusters with widely varying densities; Toy, Circles, Pathbased, and Ring are datasets containing clusters with multiple centroids; Asymmetric, Skewed, Moon, S1, and Unbalance1 are datasets containing clusters with relatively even density; and Spiral and Zigzag are datasets containing clusters of arbitrary shapes. Furthermore, the ARI and NMI metrics were used to quantitatively compare the clustering results.

[0028] The experimental results are shown in Tables 1 to 4. We can observe from Tables 1 and 2 that our method performs better than other comparison algorithms on all four types of complex datasets. Furthermore, it can also better identify each cluster in complex datasets such as Pathbased, Toy, and Zigzag, which contain multiple types of clusters.

[0029] Table 1 Quantitative clustering results of complex data sets ARI

[0030]

[0031] As shown in Tables 3 and 4, our method achieved the highest evaluation index in all six UCI datasets and the second highest evaluation index in the remaining Seeds dataset.

[0032] Experimental results show that the clustering effect of the method is good and it can handle different types of complex data sets.

[0033] Table 2 Quantitative clustering results NMI of complex data sets

[0034]

[0035] Table 3 Quantitative clustering results of UCI dataset ARI

[0036]

[0037] Table 4 Quantitative clustering results NMI of UCI dataset

[0038]

[0039] Figure 2 Update the density distribution for the Zigzag dataset before and after. Figure 2 (a) It can be found that after the method updates the density distribution week, the density of the "Z"-shaped cluster in Zigzag is Figure 2 (b) The density of the cluster without density update can better show the characteristics of high density in the cluster center area and low density in the cluster edge area, so that the label of cluster center point 282 can be smoothly transferred to the entire "Z"-shaped cluster.

[0040] Figure 3 and Figure 4 Here are two examples of complex data sets. Figure 3 Clustering results of the Zigzag dataset under different algorithms. Figure 4 The following are the clustering results of the Pathbased dataset using different algorithms. As can be seen, for these two complex datasets, the method can identify clusters within the dataset and the clustering effect is good.

[0041] Example verification:

[0042] Obtain air quality data and then input the data into Figure 1In the pollutant transmission path discovery process shown, the pollutant transmission chain is discovered after the process is completed. The specific implementation process is as follows: To facilitate the understanding of the technical means, creative features, objectives and effects of the present invention, the present invention is further described below in conjunction with specific implementation methods.

[0043] 1. Obtain and enter air quality data into the method:

[0044] 1.1 Obtain air quality data. The data set includes city names and air quality index (AQI). We obtained air quality data for Gansu Province from relevant websites, filtered out prefecture-level cities in Gansu Province, and deleted missing data. Finally, we obtained the full-day air quality index data for Lanzhou, Jiayuguan, Jinchang, Baiyin, Wuwei, and Zhangye on May 2, 2025. Figure 5 shown.

[0045] 1.2 Input the dataset into this method, where each data point represents a city. Set the algorithm parameters to the default parameters: number of neighbors k = 2, threshold distance =10.

[0046] 2. Calculate the distance between data points in the dataset after considering geographical factors, and convert air quality information into density information based on the distance information:

[0047] 2.1 Calculate the Euclidean distance between all data points in the dataset , where n is the number of points in the data set;

[0048] 2.2 If there is complex terrain, set the distance between complex terrains to the maximum distance between all data points;

[0049] 2.3 If there is no complex terrain, calculate the k nearest neighbors (kNN) of all data points based on distance;

[0050] 2.4 At this point, the study found that there was no complex terrain between the data locations that would affect the spread of pollutants, so 2.5 was directly executed;

[0051] 2.5 Calculate all data points according to formula (1) Density , obtain the information contained in the air quality data;

[0052] 3. Use density accumulation strategy to enhance density information to enhance air quality information and obtain pollutant sources:

[0053] 3.1 Sort all data points by density from low to high, and set all data points to unvisited state;

[0054] 3.2 If the data point x iIf it has not been visited, then determine whether there is a point that meets the conditions of formula (2): If so, execute ρ j =ρ i +ρ j , kNN xj =kNN xj ∪kNN xi , set x i and x j For visited, x j Becomes the new x i , kNN xj Become the new kNN xi If not, then x i As the highest density point, that is, the source of pollutants, kNN xi Mark the points in as visited and repeat 3.2;

[0055] 3.3 If all data points are visited, assign labels to all the highest density points;

[0056] 4 Considering the problem of sudden wind direction, for each data point except the highest density point, obtain its nearest high-density point as the source of its pollutant transmission:

[0057] 4.1 If there is a sudden wind direction, the pollutant transmission source for the sudden meteorological city is set to the city with the directional wind, and the density of the sudden meteorological city is set to the density of the city with the directional wind minus the minimum density difference. All wind direction data from May 2, 2025, to May 3, 2025, are shown in Table 5. The meteorological data is sourced from https: / / datashareclub.com. Only the wind direction on May 2, 2025, was considered to correct the source of urban pollutant transmission. On May 2, 2025, the cities with wind directions greater than level 3 were Jinchang, Jiayuguan, and Wuwei. The cities where Jinchang experienced a level 7 north wind and Wuwei experienced a level 4-5 northeast wind are not located in Gansu Province, so no correction was made for their transmission sources. The level 5 east wind in Jiayuguan occurred in Zhangye, the nearest high-density point, so no correction was made.

[0058] Table 5 Meteorological data

[0059]

[0060] 6 Obtain the final pollutant transmission path:

[0061] 6.1 Obtain the label propagation path. Traverse the data points from high-density points to low-density points, and assign each unlabeled point the label of the nearest high-density point. After density reconstruction of the six cities in Gansu Province on May 2, 2025, the output label propagation path is the air pollution flow path as follows: Figure 6 shown.

[0062] 6.2 Put data points with the same label into a cluster;

[0063] 6.3 Obtain the pollutant propagation path according to the label propagation order. Figure 6 As shown in Figure 2, the resulting air pollution paths are 5-2, 5-1-4, and 5-0-3, corresponding to the cities of Zhangye-Jinchang, Zhangye-Jiayuguan-Wuwei, and Zhangye-Lanzhou-Baiyin. The specific pollutant transmission paths in Gansu Province on May 2, 2025 (① Zhangye-Jinchang, ② Zhangye-Jiayuguan-Wuwei, ③ Zhangye-Lanzhou-Baiyin) are as follows: Figure 7 shown.

[0064] 6.4 The specific analysis is as follows:

[0065] Zhangye-Jinchang route:

[0066] On May 2, the northwest wind (level 3) in Zhangye and the north wind (level 7) in Jinchang were in similar directions, which may have formed a wind field superposition effect and pushed pollutants to spread southeast to Jinchang.

[0067] On May 3, northwest winds (level 4) continued in Zhangye, while Jinchang turned to southeast winds (level 3), which may have formed a local vortex and caused pollutants to be trapped or backflow near Jinchang.

[0068] Zhangye-Jiayuguan-Wuwei route:

[0069] On May 2, the northwest wind in Zhangye was perpendicular to the east wind (level 5) in Jiayuguan, which may have formed a pollutant intersection area, and then was carried towards Wuwei by the east wind in Jiayuguan.

[0070] On May 3, the wind in Jiayuguan turned to northwest (level 8) at night, which may blow pollutants directly to Wuwei, while the southwest wind (level 3) in Wuwei may further push pollutants to spread to the southeast.

[0071] Zhangye-Lanzhou-Baiyin Route:

[0072] On May 2, the northwest wind from Zhangye and the northeast wind from Lanzhou (level 3) formed a convergence zone near Lanzhou, which may cause pollutants to accumulate in Lanzhou.

[0073] On May 3, the wind in Lanzhou weakened (<Level 3), but the Baiyin area may still be affected by sandstorms, and pollutants may slowly spread toward Baiyin with the weak wind field.

[0074] These analyses show that the pollutant propagation along the three paths is significantly affected by changes in wind fields, which also confirms the effectiveness of the method in discovering pollutant propagation paths.

[0075] In summary, the experimental process of applying this method to discover the propagation path of atmospheric pollution is completed.

Claims

1. A method for discovering atmospheric pollutant propagation paths based on reconstruction density clustering, characterized by: Different from chemical and physical tracing, pollution source tracing is done directly using air quality monitoring data from various locations, which includes the following algorithmic processing steps: Step 1: Input the AQI data of each city. Each data represents the AQI information of a city in a period of time. Input the required parameters of nearest neighbor k and threshold distance. ; Step 2: Calculate the distance between data points in the dataset after considering geographical factors, and assign each data point to the The air quality information is converted into density information; Step 3: Considering the urban air quality data information and the transmission of pollution sources from high concentration to low concentration, the density of data points is reconstructed using the density accumulation strategy to enhance the air quality data information, and the nearest neighbor k and threshold distance are used. Continuously increase the density of data points, simulate the spread of high-concentration pollution sources, reverse the pollutant propagation path, and provide information for the final discovery of the pollutant propagation path; Step 4: Considering the problem of sudden wind direction, for each data point except the highest density point, obtain its nearest high-density point as the source of its pollutant transmission; Step 5: After incorporating the influence of geographical factors and wind direction factors into the distance and the nearest high-density point, assign labels according to the nearest high-density method obtained in Step 4. The assigned label path is the pollutant propagation path; the assignment method is as follows: first assign a label to the point with the highest density, and then assign the label of the data point to the label of its nearest high-density point in order of density from high to low. The constructed point with the highest density is the source of the pollutant, and the path of the assigned labels is the pollutant propagation path.

2. The method for discovering air pollutant propagation paths based on reconstructed density clustering according to claim 1, wherein the method comprises the following steps: The core method for matching steps 1 to 5 is as follows: ① Based on the input data set, calculate the Euclidean distance between all data points in the data set , where n is the number of points in the data set; ②According to , get the k nearest neighbors of all points ; ③Calculate all data points according to formula (1) Density ; in Indicates a point k nearest neighbors, Indicates a point and point distance, Representative data points density; ④ From the lowest density and unvisited point First, according to the rule of formula (2) Harmony The distance is less than The minimum density point As the next point of reconstructed density, Density Add to Density superior, Merge 、 and Set to visited, Become new 、 Become new , and then continue to select points according to the rule of formula (2) until there is no point that meets the conditions, and record the last point that meets the conditions as a cluster center point. All points in the are marked as visited, and then continue to run ④ until all points are visited; in Indicates a point k nearest neighbors, Indicates a point and point distance, Representative data points The density, Representative data points The density, is the input distance threshold; ⑤All cluster center points obtained from ④ are all cluster center points of the data set; ⑥The density reconstructed in ④ is the new density of all points in the data set ; ⑦ Assign a separate label to all cluster center points in the data set, and then traverse from high-density points to low-density points, assigning each data point the label of its nearest high-density point. The process of assigning labels is recorded as the pollutant propagation process; ⑧Finally, the data points with the same label are placed into a cluster to complete the clustering of the data set.

3. The method for discovering air pollutant propagation paths based on reconstructed density clustering according to claim 1, characterized in that: In step 2, the geographical environment factors between the two places and the air quality information in the dataset are considered to obtain data information and convert it into density information. The specific process is as follows; S1: Calculate the distance matrix between all polluted city data points; S2: Set the distance between data points with special terrain. Natural barriers such as mountains in the terrain have a significant blocking effect on the transmission of air pollutants between cities. For such geographical environments that affect the transmission of pollutants, set the distance between two data points to the maximum distance in the distance matrix to simulate the transmission blocking situation. S3: According to the updated distance matrix, each data The air quality data is converted into density information.

4. The method for discovering air pollutant propagation paths based on reconstructed density clustering according to claim 1, characterized in that: In step 3, the density accumulation strategy is used to enhance the site data information and reverse the pollutant propagation process to find the pollution source. The specific process is as follows: S1: Use density accumulation to enhance air quality information and simulate the spread of pollutants from high concentration to low concentration to infer the pollution source; the specific steps of the density accumulation strategy are: starting from the lowest density and unvisited point Start traversing towards the highest density point and update the density condition as shown below. For each data point ,exist of Search The lowest density of unvisited points within the distance ,Will Density Add to Density superior, Merge 、 and Set to visited, Become new 、 Become new , and then continue to recursively add data points The density of , until there is no point that meets the conditions, then the final The point with the highest density is a pollutant source. All points in are marked as visited; the density update formula is shown in formula (2), where k and is the input parameter; S2: Continue traversing from the lowest density point, looking for the next unvisited low-density point, and start a new round of S1 density enhancement steps; all the highest density points found are simulated pollution sources with high concentrations of pollutants.

5. The method for discovering air pollutant propagation paths based on reconstructed density clustering according to claim 1, characterized in that: In step 4, the pollutant propagation path is corrected by considering sudden wind direction changes. The specific process is as follows: S1: Sort the data points by density from high to low to get the data points The nearest high-density point ; S2: Consider the wind direction factor to modify the pollutant propagation path; the change of wind direction will cause the propagation direction of atmospheric pollutants to change significantly, because wind is the main driving force for the spread of pollutants; therefore, if there is a sudden weather problem such as wind direction change, the pollution source of the relevant data point (i.e. the nearest high-density point) will be adjusted at that location; if there is a sudden wind direction change level of level 3 or above, set the sudden weather city The source of pollutants is the city where the wind is blowing , the density of cities with sudden meteorological events Set it to the density of the city where the wind is oriented minus the minimum density difference, that is The nearest high-density point is 、 Density .

Images