Pollution risk identification and early warning method based on multistage scale

Abstract

The invention discloses a pollution risk identification and early warning method based on a multi-level scale. The method comprises the following steps: dividing county level grids, town level grids, village level grids and extensive plot grids; selecting a pressure index, a state index and a response index as risk identification indexes, collecting corresponding remote sensing data, ground sensor network data and agricultural data, and then performing space-time matching by adopting a Bayesian fusion model to obtain a standardized data set of each level of grid unit; a Delphi method is adopted to carry out multiple times of assignment to determine the weights of the three indexes, then a risk identification model is constructed, then a risk value is calculated, and a total quantity Quantile grading method is adopted to divide risk levels to identify the risk of each grid unit; a quantile and exponential smoothing method is adopted to dynamically adjust a threshold value, then early warning grades are divided according to the threshold value, and finally early warning is carried out according to the early warning grades. The agricultural non-point source pollution multi-level scale high-precision identification and real-time early warning are realized, and the precision and timeliness of pollution management are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of agricultural environmental protection, and particularly relates to a pollution risk identification and early warning method based on multiple scales. BACKGROUND

[0002] In the field of agricultural environmental protection, agricultural non-point source pollution has become one of the key problems restricting sustainable development. At present, the agricultural non-point source pollution risk identification technology has a significant scale fragmentation problem:

[0003] 1. Data source and scale disconnection: Large scale (county and above): mainly rely on statistical annual data (such as county chemical fertilizer application amount) and low resolution remote sensing image (such as Landsat 30 meters), which can only reflect the overall pollution load of the region, and cannot identify the differences between towns or the location of specific pollution sources. Medium scale (township-village): generally use hydrological models such as SWAT and HSPF, but due to insufficient precision of input data (such as soil property interpolation error > 20%), it is difficult to accurately simulate the migration path of pollutants in small watersheds (such as the runoff conduction process between villages). Small scale (pan-plot level): rely on high-cost field monitoring (such as water quality sensor layout density <1 / km 2 ), although accurate data can be obtained at the point, but the coverage is limited and cannot be extended to regional application.

[0004] 2. Incompatible model and evaluation system: County statistical model (such as risk index based on total chemical fertilizer application) and pan-plot level process model (such as risk assessment based on runoff TN concentration) cannot be compatible, resulting in that the large-scale results cannot be decomposed to micro units (for example: county chemical fertilizer application amount cannot be accurately allocated to specific field plots). At the same time, different scales use independent index system (such as "unit area chemical fertilizer load" for county and "soil nitrogen and phosphorus accumulation" for plot), the output results have different dimensions and grading standards, which hinder the cross-scale risk comparison and management coordination.

[0005] 3. Management decision gap: County early warning can only be located to administrative area (such as prompting "high risk in Fuling District"), but cannot be traced to responsible town (such as Nantuo Town contribution rate) or key pollution plot; plot-level monitoring data (such as 0.5-meter resolution crop growth obtained by unmanned aerial vehicle) are not included in county model calibration, resulting in that macro decision lacks micro data support. At the same time, due to unclear transmission mechanism between scales, prevention and control measures often appear "county planning and plot management misalignment" (such as excessive investment in management resources in low-risk areas, while the real hotspots are ignored). SUMMARY

[0006] In view of the above-mentioned shortcomings of the existing technology, the purpose of this invention is to provide a pollution risk identification and early warning method based on multi-level scale. This invention realizes high-precision identification and real-time early warning of agricultural non-point source pollution at multiple levels, which is conducive to improving the accuracy and timeliness of pollution management.

[0007] The technical solution of this invention is implemented as follows:

[0008] A pollution risk identification and early warning method based on multi-scale methods includes the following steps:

[0009] S1: Based on administrative boundaries and natural geographical units, divide the grid into county-level grids, township-level grids and village-level grids. At the same time, based on farmland patch boundaries, further divide the grid into pan-plot grids within the village-level grids, thus obtaining multiple grid units at different scales of county-level, township-level, village-level and pan-plot level.

[0010] S2: Select pressure indicators, state indicators and response indicators as risk identification indicators, and collect corresponding remote sensing data, ground sensor network data, agricultural statistics data and agricultural survey data. Then, use a Bayesian fusion model to perform spatiotemporal matching to obtain standardized datasets for each level of grid unit.

[0011] S3: Based on the standardized dataset, the weights of the three indicators are determined by multiple assignments using the Delphi method. Then, a risk identification model is constructed, and the risk values ​​of each grid unit are calculated. The model is shown in equation (1).

[0012] I = q1I1 + q2I2 + q3I3 (1)

[0013] In the formula: I is the risk value; q1 is the weight of the stress indicator; I1 is the score of the stress indicator; q2 is the weight of the state indicator; I2 is the score of the state indicator; q3 is the weight of the response indicator; I3 is the score of the response indicator.

[0014] S4: Based on the calculated risk value, the risk level is divided using the total Quantile classification method to identify the risk of each grid cell;

[0015] S5: The threshold is dynamically adjusted using quantile and exponential smoothing methods, then the warning level is divided according to the threshold, and finally the warning is issued according to the warning level.

[0016] Furthermore, in step S1, coding is performed simultaneously with dividing the grid units. The county-level grid code is the province code + city code + county code; the township-level grid code is the county-level grid code + township code; the village-level network code is the township-level grid code + village code; and the pan-plot-level network code is the village-level network code + sequence code.

[0017] Furthermore, the remote sensing data includes satellite remote sensing data and UAV remote sensing data, specifically including land use data, soil type and texture data, digital elevation model topographic data, slope data, river data, and road data.

[0018] Furthermore, the agricultural statistics include fertilizer usage, pesticide usage, livestock and poultry breeding, aquaculture, and crop planting area.

[0019] Furthermore, the agricultural survey data is obtained through questionnaires at the broad plot scale.

[0020] Furthermore, the stress indicators include fertilizer use intensity index, pesticide use intensity index, livestock and poultry farming intensity index, aquaculture intensity index, and residents' living intensity index, then:

[0021] I1=q 11 I 11 +q 12 I 12 +q 13 I 13 +q 14 I 14 +q 15 I 15 (2)

[0022] In the formula: q 11 and I 11 These represent the weights and scores of the fertilizer use intensity index, respectively; q 12 and I 12 These represent the weights and scores of the pesticide use intensity index, respectively; q 13 and I 13 These represent the weights and scores of the livestock and poultry intensity index, respectively; q 14 and I 14 These represent the weights and scores of the aquaculture intensity index, respectively; q 15 and I 15 These are the weights and scores of the residents' living intensity index, respectively; and each weight is determined by multiple assignments using the Delphi method.

[0023] Furthermore, the state indicators include rainfall erosion index, slope length and gradient index, sloping farmland index, soil erodibility index, and water distance index; then

[0024] I2 = q 21 I 21 +q 22 I 22 +q 23 I 23 +q 24 I 24 +q 25 I 25(3)

[0025] In the formula: q 21 and I 21 These represent the weights and scores of the rainfall erosion index, respectively; q 22 and I 22 These represent the weights and scores of the slope length and slope index, respectively; q 23 and I 23 These represent the weights and scores of the sloping farmland index, respectively; q 24 and I 24 These represent the weights and scores of the soil erodibility index; q 25 and I 25 These are the weights and scores of the water distance index, respectively; and each weight is determined by multiple assignments using the Delphi method.

[0026] Furthermore, the response indicators include forest land retention index, grassland retention index, and water conservation index; then

[0027] I3 = q 31 I 31 +q 32 I 32 +q 33 I 33 (4)

[0028] In the formula: q 31 and I 31 These represent the weights and scores for delayed payment of forest land; q 32 and I 32 These represent the weights and scores of the grassland retention index; q 23 and I 23 These are the weights and scores of the water conservation index; and each weight is determined by multiple assignments using the Delphi method.

[0029] Furthermore, in step S5, the threshold T t The result is obtained by calculation using formula (5):

[0030]

[0031] In the formula: T t The threshold for period t; O t Let be the risk value for period t; α is the smoothing coefficient, ranging from 0.3 to 0.7, with 0.7 for pollution-sensitive areas.

[0032] Furthermore, if the risk value exceeds the threshold by 10-20%, a blue warning is issued; if the risk value exceeds the threshold by 20-50%, a yellow warning is issued; if the risk value exceeds the threshold by 50-100%, an orange warning is issued; and if the risk value exceeds the threshold by more than 100%, a red warning is issued.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. This invention innovatively solves the problem of mesoscale discontinuity in traditional monitoring by constructing a four-level grid system of "county-township-village-pane". It achieves seamless connection from macro (county level) to micro (pane level) and solves the management discontinuity problem of "county-level early warning cannot locate polluted plots" in traditional technology. It realizes high-precision identification and real-time early warning of agricultural non-point source pollution at multiple scales, which is conducive to improving the accuracy and timeliness of pollution management.

[0035] 2. This invention constructs a unified risk identification model based on four scales: county, township, village, and general land parcel, which increases the universality of the model and helps to improve the comparability of risk assessments at different scales. Attached Figure Description

[0036] Figure 1 - A diagram illustrating the results of a county-level risk assessment.

[0037] Figure 2 - A diagram illustrating the results of a township-level risk assessment.

[0038] Figure 3 - A diagram illustrating the results of a village-level risk assessment.

[0039] Figure 4 - A schematic diagram of the results of the pan-plot-level risk assessment. Detailed Implementation

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0041] A pollution risk identification and early warning method based on multi-scale methods includes the following steps:

[0042] S1: Based on administrative boundaries and natural geographical units, divide the area into county-level grids, township-level grids, village-level grids, and pan-plot-level grids, and encode them. The county-level grid is coded as province code + city code + county code; the township-level grid is coded as county-level grid code + township code; the village-level network is coded as township-level grid code + village code; and the pan-plot-level network is coded as village-level network code + sequence code.

[0043] The multi-scale grid spatial partitioning method is as follows:

[0044] 1. County-level grid division

[0045] (1) Basis for division: Based on the boundaries of county-level administrative divisions, adjustments are made in conjunction with watershed watersheds (such as the first-level tributaries of the Yangtze River) to ensure that each county grid is a complete hydrological unit.

[0046] (2) Technical steps:

[0047] Import county-level administrative boundary vector data (accuracy 1:100,000), and use ArcGIS's topology inspection tool to ensure that the boundaries are free of overlap and gaps.

[0048] By overlaying watershed data, county-level grids that cross watersheds are re-divided according to watersheds. For example, Fuling District in Chongqing is divided into two sub-grids, Jiangnan and Jiangbei, with the Yangtze River as the boundary, because the Yangtze River runs through it.

[0049] (3) Grid coding rules: The three-segment coding of “province code + city code + county code” is adopted. For example, the code of Fuling District, Chongqing is 500102, where 50 is the province code, 01 is the city code, and 02 is the county code.

[0050] 2. Township-level grid division

[0051] (1) Division basis: Based on the administrative boundaries of townships, optimization is carried out by combining topographic units (such as ridge lines and valleys) and land use types (such as concentrated cultivated land areas and forest distribution areas).

[0052] (2) Technical steps:

[0053] Import township administrative boundary data (accuracy 1:50,000), and use DEM data (30-meter resolution) to extract terrain feature lines as the basis for grid adjustment.

[0054] For areas greater than 50km 2 Townships are divided into 2-3 sub-grids based on the density of cultivated land distribution. For example, Nantuo Town in Fuling District is divided into two sub-grids: the plain area along the river and the hilly area, due to differences in terrain.

[0055] (3) Grid coding rules: Add township code 01 after county code. For example, Nantuo Town is coded as 50010201.

[0056] 3. Village-level grid division

[0057] (1) Basis for division: Based on the boundaries of administrative villages, and further subdivided according to the distribution of natural villages, farmland patches and water system network.

[0058] (2) Technical steps:

[0059] Import administrative village boundary data (accuracy 1:10,000), overlay high-resolution remote sensing imagery (0.5 meters), and identify natural village settlements and farmland patches.

[0060] For areas greater than 1km 2 The villages are divided into 3-5 grids according to the distribution of farmland. For example, Muhe Village is divided into 6 village-level grids, each corresponding to farmland in different terrain areas.

[0061] (3) Grid coding rules: Add village code 01 after township code. Muhe Village code is 5001020101.

[0062] 4. Pan-plot-level grid division

[0063] (1) Division basis: Based on farmland patch boundaries, land type differences and pollution risk characteristics, village-level grids are further subdivided into pan-plot units.

[0064] (2) Technical steps:

[0065] Land cover classification was performed using UAV multispectral imagery (0.5-meter resolution) and a random forest algorithm was adopted. Twelve vegetation indices, including NDVI, NDRE, and LCI, were input, and the classification accuracy reached over 85%.

[0066] Combining topographic factors (slope, aspect) with farmland management units (such as field ridges and irrigation canals), continuous farmland is divided into independent, pan-plot grids with an average area of ​​0.05–0.2 hm². 2 .

[0067] (3) Grid coding rules: Add the sequence code 01 after the village code, such as the general plot code in the grid of Muhe Village No. 01 is 5001020101001.

[0068] S2: Pressure indicators, state indicators, and response indicators are selected as risk identification indicators, and corresponding remote sensing data, ground sensor network data, agricultural statistics data, and agricultural survey data are collected. Then, a Bayesian fusion model is used for spatiotemporal matching to obtain standardized datasets for each level of grid unit. The specific formula for fusion is shown in equation (1):

[0069]

[0070] In the formula, F represents the fused data, and D... i For the i-th type of data source at the corresponding scale, w i This represents the weight corresponding to the i-th type of data source at the corresponding scale.

[0071] Satellite remote sensing data includes data acquired by Landsat 8 (30m x 30m accuracy) and Sentinel-2 (10m x 10m accuracy), providing macroscopic information such as vegetation cover, land use, soil type and texture, digital elevation model (DEM) topographic data, slope data, river data, and road data. Low-altitude drone remote sensing data, acquired by the DJI Phantom 4 multispectral drone, provides mesoscale information such as crop type and growth status in farmland. Before use, the remote sensing data undergoes radiometric calibration, atmospheric correction, and geometric fine correction to control point errors to less than 0.5 pixels. Finally, the images are mosaicked and cropped.

[0072] The ground sensor network data includes data obtained from a 500m × 500m ground monitoring network consisting of soil nitrogen and phosphorus sensors (accuracy ±5%), weather stations (rainfall accuracy ±1mm), and water quality monitors (TN accuracy ±0.1mg / L). Outliers were removed using the 3σ principle, missing values ​​were filled using linear interpolation, and then the time was synchronized to UTC time.

[0073] Agricultural statistics: fertilizer application, pesticide use, livestock and poultry breeding, etc. (annual data at the county / district level, quarterly data at the township level); farmer survey data: information on farming methods and input usage at the plot level obtained through questionnaires (sampling ratio not less than 10%). The data format is standardized into Excel spreadsheets. Then, Gauss-Schlicken interpolation is used to interpolate point data (such as sensor data) into area data, with resolution consistent with the grid scale to achieve spatial matching. Finally, a timeline is established, and resampling techniques are used for data from different periods. Based on a time series model, satellite data is resampled into daily data.

[0074] S3: Based on the standardized dataset, the weights of the three indicators are determined by multiple assignments using the Delphi method. Then, a risk identification model is constructed, and the risk values ​​of each grid unit are calculated. The model is shown in Equation (2).

[0075] I = q1I1 + q2I2 + q3I3 (2)

[0076] In the formula: I is the risk value; q1 is the pressure indicator weight; I1 is the pressure indicator score; q2 is the state indicator weight; I2 is the state indicator score; q3 is the response indicator weight; I3 is the response indicator score.

[0077] The pressure indicators include fertilizer use intensity index, pesticide use intensity index, livestock and poultry farming intensity index, aquaculture intensity index, and residents' living intensity index. Therefore:

[0078] I1=q 11 I 11 +q 12 I 12 +q 13 I 13 +q 14 I 14 +q 15 I 15 (3)

[0079] In the formula: q 11 and I 11 These represent the weights and scores of the fertilizer use intensity index, respectively; q 12 and I 12These represent the weights and scores of the pesticide use intensity index, respectively; q 13 and I 13 These represent the weights and scores of the livestock and poultry intensity index, respectively; q 14 and I 14 These represent the weights and scores of the aquaculture intensity index, respectively; q 15 and I 15 These are the weights and scores of the residents' living intensity index, respectively; and each weight is determined by multiple assignments using the Delphi method.

[0080] The state indicators include rainfall erosion index, slope length and gradient index, sloping farmland index, soil erodibility index, and water distance index; therefore:

[0081] I2 = q 21 I 21 +q 22 I 22 +q 23 I 23 +q 24 I 24 +q 25 I 25 (4)

[0082] In the formula: q 21 and I 21 These represent the weights and scores of the rainfall erosion index, respectively; q 22 and I 22 These represent the weights and scores of the slope length and slope index, respectively; q 23 and I 23 These represent the weights and scores of the sloping farmland index, respectively; q 24 and I 24 These represent the weights and scores of the soil erodibility index; q 25 and I 25 These are the weights and scores of the water distance index, respectively; and each weight is determined by multiple assignments using the Delphi method.

[0083] The response indicators include forest land retention index, grassland retention index, and water conservation index; then

[0084] I3 = q 31 I 31 +q 32 I 32 +q 33 I 33 (5)

[0085] In the formula: q 31 and I 31 These represent the weights and scores for delayed payment of forest land; q 32 and I 32 These represent the weights and scores of the grassland retention index; q23 and I 23 These are the weights and scores of the water conservation index; and each weight is determined by multiple assignments using the Delphi method.

[0086] S4: Based on the calculated risk value, the risk level is divided into categories using the total Quantile classification method to identify the risk of each grid cell.

[0087] The risk values ​​at each scale, representing 20%, 40%, 60%, and 80% of the total, are categorized into five levels: no risk, low risk, medium risk, high risk, and extremely high risk, and assigned values ​​of 1, 2, 3, 4, and 5. Specifically, if I ∈ [0, 1], it is considered no risk; if I ∈ (1, 2], it is considered low risk; if I ∈ (2, 3], it is considered medium risk; if I ∈ (3, 4], it is considered high risk; and if I ∈ (4, 5], it is considered extremely high risk.

[0088] S5: The threshold is dynamically adjusted using quantile and exponential smoothing methods. Then, warning levels are assigned based on the thresholds, and finally, warnings are issued according to the warning levels. Threshold T t The result is obtained by formula (6):

[0089]

[0090] In the formula: T t The threshold for period t; O t Let be the risk value for period t; α is the smoothing coefficient, ranging from 0.3 to 0.7, with 0.7 for pollution-sensitive areas.

[0091] If the risk value exceeds the threshold by 10-20%, a blue warning is issued; if the risk value exceeds the threshold by 20-50%, a yellow warning is issued; if the risk value exceeds the threshold by 50-100%, an orange warning is issued; if the risk value exceeds the threshold by more than 100%, a red warning is issued.

[0092] Figure 1 Based on the risk identification results at the county level (provincial / municipal level), further analysis can be conducted at the township level (…) using the methods described above. Figure 2 Risk identification (within the district / county area) can clearly define the risk level of a specific township within the county. Furthermore, the same method can be used to conduct risk identification at the village level (…). Figure 3 Risk identification within townships and villages (township-level) clarifies the risk level of specific villages within the township. Risk identification at the broader plot level can also be conducted using the methods described above. Figure 4 Risk identification (within small watersheds) clarifies the risk level of specific plots within villages. Figures 1-4 Dark green indicates no risk, light green indicates low risk, yellow indicates medium risk, orange indicates high risk, and red indicates extremely high risk.

[0093] Finally, it should be noted that the above embodiments of the present invention are merely examples for illustrating the present invention and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations and modifications based on the above description. It is impossible to exhaustively list all possible implementations here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A pollution risk identification and early warning method based on multi-scale, characterized in that, Specifically, the following steps are included: S1: Based on administrative boundaries and natural geographical units, divide the grid into county-level grids, township-level grids and village-level grids. At the same time, based on farmland patch boundaries, further divide the grid into pan-plot grids within the village-level grids, thus obtaining multiple grid units at different scales of county-level, township-level, village-level and pan-plot level. S2: Select pressure indicators, state indicators and response indicators as risk identification indicators, and collect corresponding remote sensing data, ground sensor network data, agricultural statistics data and agricultural survey data. Then, use a Bayesian fusion model to perform spatiotemporal matching to obtain standardized datasets for each level of grid unit. S3: Based on the standardized dataset, the weights of the three indicators are determined by multiple assignments using the Delphi method. Then, a risk identification model is constructed, and the risk values ​​of each grid unit are calculated. The model is shown in Equation (1). I = q1I1 + q2I2 + q3I3 (1) In the formula: I is the risk value; q1 is the weight of the stress indicator; I1 is the score of the stress indicator; q2 is the weight of the state indicator; I2 is the score of the state indicator; q3 is the weight of the response indicator; I3 is the score of the response indicator. S4: Based on the calculated risk value, the risk level is divided using the total Quantile classification method to identify the risk of each grid cell; S5: The threshold is dynamically adjusted using quantile and exponential smoothing methods, then the warning level is divided according to the threshold, and finally the warning is issued according to the warning level.

2. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, In step S1, the grid cells are divided and coded simultaneously. The county-level grid is coded as province code + city code + county code; the township-level grid is coded as county-level grid code + township code; the village-level network is coded as township-level grid code + village code; and the pan-plot-level network is coded as village-level network code + sequence code.

3. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, The remote sensing data includes satellite remote sensing data and UAV remote sensing data, specifically including land use data, soil type and texture data, digital elevation model topographic data, slope data, river data, and road data.

4. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, The agricultural statistics include fertilizer usage, pesticide usage, livestock and poultry breeding, aquaculture, and crop planting area.

5. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, The agricultural survey data was obtained through questionnaires at the broad plot scale.

6. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, The pressure indicators include fertilizer use intensity index, pesticide use intensity index, livestock and poultry farming intensity index, aquaculture intensity index, and residents' living intensity index. Therefore: I1=q 11 I 11 +q 12 I 12 +q 13 I 13 +q 14 I 14 +q 15 I 15 (2) In the formula: q 11 and I 11 These represent the weights and scores of the fertilizer use intensity index, respectively; q 12 and I 12 These represent the weights and scores of the pesticide use intensity index, respectively; q 13 and I 13 These represent the weights and scores of the livestock and poultry intensity index, respectively; q 14 and I 14 These represent the weights and scores of the aquaculture intensity index, respectively; q 15 and I 15 These are the weights and scores of the residents' living intensity index, respectively; and each weight is determined by multiple assignments using the Delphi method.

7. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, The state indicators include rainfall erosion index, slope length and gradient index, sloping farmland index, soil erodibility index, and water distance index; then I2=q 21 I 21 +q 22 I 22 +q 23 I 23 +q 24 I 24 +q 25 I 25 (3) In the formula: q 21 and I 21 These represent the weights and scores of the rainfall erosion index, respectively; q 22 and I 22 These represent the weights and scores of the slope length and slope index, respectively; q 23 and I 23 These represent the weights and scores of the sloping farmland index, respectively; q 24 and I 24 These represent the weights and scores of the soil erodibility index; q 25 and I 25 These are the weights and scores of the water distance index, respectively; and each weight is determined by multiple assignments using the Delphi method.

8. The pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, The response indicators include forest land retention index, grassland retention index, and water conservation index; then I3=q 31 I 31 +q 32 I 32 +q 33 I 33 (4) In the formula: q 31 and I 31 These represent the weights and scores for delayed payment of forest land; q 32 and I 32 These represent the weights and scores of the grassland retention index; q 23 and I 23 These are the weights and scores of the water conservation index; and each weight is determined by multiple assignments using the Delphi method.

9. A pollution risk identification and early warning method based on multi-scale according to claim 1, characterized in that, In step S5, the threshold T t The result is obtained by calculation using formula (5): In the formula: T t The threshold for period t; O t Let be the risk value for period t; α is the smoothing coefficient, ranging from 0.3 to 0.7, with 0.7 for pollution-sensitive areas.

10. A pollution risk identification and early warning method based on multi-scale according to claim 9, characterized in that, If the risk value exceeds the threshold by 10-20%, a blue alert will be issued. If the risk value exceeds the threshold by 20-50%, a yellow alert is issued; if the risk value exceeds the threshold by 50-100%, an orange alert is issued. If the risk value exceeds the threshold by more than 100%, a red alert will be issued.

Images