Aerosol model dynamic matching method in radiation transfer calculation

Abstract

The invention relates to an aerosol model dynamic matching method in radiation transfer calculation. The method comprises the following steps: preprocessing aerosol data acquired by a satellite; calculating the ratio AROD of the optical thicknesses of the aerosols in two different wave bands of 470 nm and 660 nm; classifying the aerosol to obtain aerosol types; according to the aerosol type and a random forest algorithm, statistics is carried out, and a corresponding aerosol model is dynamically matched. According to the method, the types of the aerosols can be accurately classified, the aerosols are carefully classified by using the aerosol optical thickness ratio AROD of two different wavebands of 470 nm and 660 nm, the pertinence and applicability of aerosol model selection are ensured, the accuracy of model matching is greatly improved, the calculation complexity and time consumption are reduced, and the method is suitable for being applied to the field of aerosol model selection. The method greatly improves the model selection efficiency, enables the aerosol model to be faster and more effective in practical application, can respond to the change of a complex environment in real time, and guarantees the high precision of radiation transmission calculation.

Description

Technical Field

[0001] The invention relates to the technical field of atmospheric radiation transfer calculation, in particular to a method for dynamic matching of an aerosol model in radiation transfer calculation. Background Art

[0002] Atmospheric aerosol particles refer to various solid, liquid and solid-liquid mixed particles suspended in the atmosphere, with a particle size generally ranging from 0.001 to 100 μm. These particles have physical and chemical properties independent of the atmosphere. When the aerosol content in the atmosphere reaches a certain level, aerosol particles will have a greater impact on radiation transmission. At the same time, the composition of aerosols varies significantly over time and space, and is related to many factors such as aerosol source distribution, underlying surface composition, seasons and meteorological conditions. In order to facilitate calculation and research, traditional methods use systematic observation experiments, taking into account the two premises of aerosol composition and sampling area, and give several typical aerosol models.

[0003] Typical aerosol models are mainly based on observational experiments, combined with regional sampling and underlying surface types. However, these models are usually static and cannot effectively reflect the dynamic changes of aerosols in time and space. Especially under the influence of multiple factors such as meteorological conditions and topography, their diffusion and transmission characteristics are often closely related to the surrounding areas. Therefore, models that rely solely on underlying surface factors have limitations and are difficult to accurately describe the actual situation. In order to improve the accuracy of atmospheric radiation transfer calculations, it is necessary to comprehensively consider the dynamic changes of meteorological elements and the complex spatial and temporal distribution of aerosol sources, realize dynamic adjustment of the model, and thus overcome the shortcomings of existing technologies. Summary of the invention

[0004] In order to solve the problem that the existing aerosol model cannot fully reflect its dynamic changes and simply relying on the underlying surface factors cannot accurately capture the actual diffusion and transmission characteristics of aerosols, the purpose of the present invention is to provide a dynamic matching method of aerosol models in radiation transfer calculations that comprehensively considers the time series changes of meteorological elements and the complex spatiotemporal distribution of aerosol sources, and dynamically adjusts the aerosol model to achieve an accurate description of aerosol concentrations and radiation transfer characteristics under different regions and environmental conditions, thereby improving the accuracy of atmospheric radiation transfer calculations.

[0005] To achieve the above object, the present invention adopts the following technical solution: a method for dynamic matching of aerosol models in radiation transfer calculation, the method comprising the following steps in order:

[0006] (1) preprocessing aerosol data collected by the satellite to obtain preprocessed aerosol data, wherein the preprocessing includes removing outliers and data clipping;

[0007] (2) Select the preprocessed aerosol data of two different bands, 470 nm and 660 nm, input them into the AOD-AROD aerosol classification model, and calculate the ratio of the aerosol optical thickness (AROD) of the two different bands, 470 nm and 660 nm;

[0008] (3) Aerosols are classified according to the ratio of aerosol optical thickness (AROD) at two different wavelengths (470 nm and 660 nm) to obtain the aerosol type. i ;

[0009] (4) Aerosol Type i And random forest algorithm, perform statistics, and dynamically match the corresponding aerosol model.

[0010] The step (1) specifically refers to: preprocessing the aerosol data collected by the satellite in Matlab, opening the aerosol optical thickness data of all bands according to the bands, screening and removing all numerical points less than or equal to zero; using GIS processing software to calibrate the application area, and cutting the study area according to the administrative divisions or the longitude and latitude of the application area.

[0011] In step (2), the calculation formula of the ratio of aerosol optical thickness AROD is:

[0012] AROD=AOD 470 / AOD 660

[0013] Among them, AOD 470 、AOD 660 Represent the aerosol optical thickness values ​​at 470nm and 660nm respectively.

[0014] The step (3) specifically refers to:

[0015] If AROD>0.91, the aerosol type is judged to be dust type;

[0016] If AROD≤0.91 and AOD 470 ≤0.45, the aerosol type is judged to be continental;

[0017] If 0.68≤AROD≤0.91 and AOD 470 ≥0.45, the aerosol type is judged to be subcontinental;

[0018] If AOD 470 ≥0.45 and 0.58≤AROD≤0.68, the aerosol type is judged to be urban industrial type;

[0019] If AROD≤0.58 and AOD470 ≥0.45, the aerosol type is judged to be biomass burning type;

[0020] Based on AROD and AOD 470 The value of Aerosol Type i It is divided into five types: dust, continental, subcontinental, biomass burning and urban industrial. 470 represents the aerosol optical density at 470 nm.

[0021] The step (4) specifically refers to: for months in which the monthly average proportion of a certain aerosol type exceeds 90% for at least three consecutive years, directly matching the aerosol model corresponding to the aerosol type; otherwise, using the random forest algorithm to dynamically match the aerosol model:

[0022] Aerosol Model i ~Wind Direction i +Source Direction i +Wind Speed i +AerosolType i

[0023] Among them, Aerosol Model i Represents the dynamically matched aerosol model at the i-th moment, Wind Direction i Indicates the wind direction at the i-th moment, Source Direction i represents the location of the aerosol source at the i-th moment, Wind Speed i represents the wind speed at the ith moment.

[0024] It can be seen from the above technical scheme that the beneficial effects of the present invention are as follows: first, the present invention can accurately classify the types of aerosols. By using the aerosol optical thickness ratio AROD of two different bands of 470nm and 660nm, the aerosols are carefully classified, ensuring the pertinence and applicability of the aerosol model selection, and greatly improving the accuracy of model matching; second, based on the analysis of long-term aerosol type data, the model matching process is simplified. For areas where the monthly average proportion of a certain aerosol type exceeds 90% for more than three consecutive years, the present invention can directly match the corresponding aerosol model, reducing the calculation complexity and time consumption. Through this data-driven simplified operation, the efficiency of model selection is greatly improved, making the aerosol model faster and more effective in practical applications; third, the accuracy of radiation transfer calculation is improved by dynamically adjusting the model. The present invention uses a random forest algorithm to dynamically match aerosol types and regional data, avoiding the limitations of traditional static models that ignore spatiotemporal changes. Through dynamic adjustment of different meteorological conditions and aerosol sources, it can respond to changes in complex environments in real time, ensuring the high accuracy of radiation transfer calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a flow chart of the method of the present invention;

[0026] Figure 2 Outlier removal plot for satellite data;

[0027] Figure 3 This is a cropped image of satellite data;

[0028] Figure 4 It is an information map of land feature types;

[0029] Figure 5 Schematic diagram of the comparison between the predicted aerosol type and the actual aerosol type. DETAILED DESCRIPTION

[0030] like Figure 1 As shown, a method for dynamic matching of aerosol models in radiation transfer calculations includes the following steps in order:

[0031] (1) preprocessing aerosol data collected by the satellite to obtain preprocessed aerosol data, wherein the preprocessing includes removing outliers and data clipping;

[0032] (2) Select the preprocessed aerosol data of two different bands, 470 nm and 660 nm, input them into the AOD-AROD aerosol classification model, and calculate the ratio of the aerosol optical thickness (AROD) of the two different bands, 470 nm and 660 nm;

[0033] (3) Aerosols are classified according to the ratio of aerosol optical thickness (AROD) at two different wavelengths (470 nm and 660 nm) to obtain the aerosol type. i ;

[0034] (4) Aerosol Type i And random forest algorithm, perform statistics, and dynamically match the corresponding aerosol model.

[0035] The step (1) specifically refers to: preprocessing the aerosol data collected by the satellite in Matlab, opening the aerosol optical thickness data of all bands according to the bands, screening and removing all numerical points less than or equal to zero, such as Figure 2 As shown, in Figure 2 In the figure, (a) is a schematic diagram before removing outliers, and (b) is a schematic diagram after removing outliers. Use GIS processing software to calibrate the application area and cut the study area according to the longitude and latitude of the administrative division or application area, such as Figure 3 As shown, in Figure 3 In the figure, (a) is a schematic diagram before cutting, and (b) is a schematic diagram after cutting.

[0036] In step (2), the calculation formula of the ratio of aerosol optical thickness AROD is:

[0037] AROD=AOD 470 / AOD 660

[0038] Among them, AOD 470 、AOD 660 Represent the aerosol optical thickness values ​​at 470nm and 660nm respectively.

[0039] The step (3) specifically refers to:

[0040] If AROD>0.91, the aerosol type is judged to be dust type;

[0041] If AROD≤0.91 and AOD 470 ≤0.45, the aerosol type is judged to be continental;

[0042] If 0.68≤AROD≤0.91 and AOD 470 ≥0.45, the aerosol type is judged to be subcontinental;

[0043] If AOD 470 ≥0.45 and 0.58≤AROD≤0.68, the aerosol type is judged to be urban industrial type;

[0044] If AROD≤0.58 and AOD470 ≥0.45, the aerosol type is judged to be biomass burning type;

[0045] Based on AROD and AOD 470 The value of Aerosol Type i It is divided into five types: dust, continental, subcontinental, biomass burning and urban industrial. 470 represents the aerosol optical density at 470 nm.

[0046] The step (4) specifically refers to: for months in which the monthly average proportion of a certain aerosol type exceeds 90% for at least three consecutive years, directly matching the aerosol model corresponding to the aerosol type; otherwise, using the random forest algorithm to dynamically match the aerosol model:

[0047] Aerosol Model i ~Wind Direction i +Source Direction i +Wind Speed i +AerosolType i

[0048] Among them, Aerosol Model i Represents the dynamically matched aerosol model at the i-th moment, Wind Direction i Indicates the wind direction at the i-th moment, Source Direction i represents the location of the aerosol source at the i-th moment, Wind Speed i represents the wind speed at the ith moment.

[0049] Based on AROD and AOD 470 The value of is used to divide aerosols into five types: dust, continental, subcontinental, biomass burning and urban industrial. The proportion of various types of aerosols in each month is calculated, and the aerosol model classification of a certain study area from 2019 to 2022 is obtained. On a larger time scale, the aerosol types in this area have a relatively obvious stability result. The aerosols in the four months of January, February, November, and December are mainly continental and subcontinental aerosols, and they are very stable for many years from 2019 to 2022, and the proportion is basically above 90%. Therefore, as a guide, in January, February, November, and December in this region, the corresponding aerosol model in the atmospheric radiation transfer calculation model can be used to accurately calculate the radiation transfer. In other months, the judgment conditions for the stability of the aerosol model are not met, and the changes in aerosol types are more complex. Therefore, it is necessary to use meteorological elements to make a more detailed analysis of the specific aerosol model in the area for dynamic matching.

[0050] When building a random forest model, a larger training data set is required. Consider the wind speed, wind direction and location of the aerosol source in the application area. Select the aerosol type data calculated by the AOD-AROD model for many years and the ERA-5 meteorological data of the corresponding date. Since the transit time of the sensor Terra on the MODIS satellite is 10:30 am, trace back 24 hours and count the wind speed and wind direction data from 10:00 am on the current day to 11:00 am the previous day. The training data set shown below is constructed:

[0051] Table 1 Wind direction input parameters, namely Wind Direction i

[0052]

[0053] Table 2 Source Direction Input Parameters i

[0054]

[0055] Table 3 Wind speed and type parameters, namely Wind Speed i

[0056]

[0057] The entire data set has a total of 34 columns, 16 of which are used to characterize the wind direction in the study area as shown in Table 1. The numbers in the table represent the frequency of occurrence of wind in that direction.

[0058] The 16 columns used to characterize the direction of the aerosol source are shown in Table 2. According to the terrain elements around the study area, if there are terrain elements such as cities or industrial parks in this direction, it is set to 1, and if there are other terrain elements in this direction, it is set to 0. Figure 4 As shown, the terrain types in the south and west directions are cities, so the S and W columns are set to 1 and the remaining columns are set to 0.

[0059] Column 1 is used to illustrate the wind speed, and column 1 is used to illustrate the aerosol type as shown in Table 3. Among them, 1 represents dust type, 2 represents continental type, 3 represents subcontinental type, 4 represents urban industrial type, and 5 represents biomass burning type.

[0060] For the entire data set, 75% of the data is used as training data and 25% of the data is used as verification data. The predicted aerosol type is compared with the actual aerosol type. The comparison process is shown in the figure below. Figure 5As shown, Urbanindustry is urban industrial type, Biomass buming is biomass burning type, Subcontinent is subcontinent type, Continent is continental type, Dust is sandstorm type, red represents the actual aerosol type, and gray represents the aerosol type matched by the present invention. It can be seen that the types matched by the present invention are mostly consistent with the actual types, which improves the accuracy of model matching.

[0061] Taking the study area as an example, it is more appropriate to use the cleaner continental aerosol pattern in January, February, November, and December. In other months, the aerosol pattern judgment model established by meteorological elements and aerosol source locations can be used to make accurate choices, that is, dynamic matching.

[0062] In summary, the present invention can accurately classify the types of aerosols. By using the aerosol optical thickness ratio AROD of two different bands of 470nm and 660nm, the aerosols are carefully classified, ensuring the pertinence and applicability of the aerosol model selection, and greatly improving the accuracy of model matching. Based on the analysis of long-term aerosol type data, the model matching process is simplified. For areas where the monthly average proportion of a certain aerosol type exceeds 90% for more than three consecutive years, the present invention can directly match the corresponding aerosol model, reducing the calculation complexity and time consumption. Through this data-driven simplified operation, the efficiency of model selection is greatly improved, making the aerosol model faster and more effective in practical applications. The accuracy of radiation transfer calculation is improved by dynamically adjusting the model. The present invention uses a random forest algorithm to dynamically match aerosol types and regional data, avoiding the limitation of traditional static models that ignore temporal and spatial changes. Through dynamic adjustment of different meteorological conditions and aerosol sources, it can respond to changes in complex environments in real time, ensuring the high accuracy of radiation transfer calculation.

[0063] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any ideal transformation and modification that can be obtained by anyone familiar with the technology within the technical scope disclosed by the present invention should also be deemed to be included in the protection scope of the present invention.

Claims

1. A dynamic matching method for aerosol model in radiation transfer calculation, characterized by: The method comprises the following steps in the following order: (1) preprocessing aerosol data collected by the satellite to obtain preprocessed aerosol data, wherein the preprocessing includes removing outliers and data clipping; (2) Select the preprocessed aerosol data of two different bands, 470 nm and 660 nm, input them into the AOD-AROD aerosol classification model, and calculate the ratio of the aerosol optical thickness (AROD) of the two different bands, 470 nm and 660 nm; (3) Aerosols are classified according to the ratio of aerosol optical thickness (AROD) at two different wavelengths (470 nm and 660 nm) to obtain the aerosol type. i ; (4) Aerosol Type i And random forest algorithm, perform statistics, and dynamically match the corresponding aerosol model.

2. The aerosol model dynamic matching method in radiation transfer calculation according to claim 1, characterized in that: The step (1) specifically refers to: preprocessing the aerosol data collected by the satellite in Matlab, opening the aerosol optical thickness data of all bands according to the bands, and screening and removing all numerical points less than or equal to zero; Use GIS processing software to calibrate the application area and cut the study area according to the administrative divisions or the longitude and latitude of the application area.

3. The aerosol model dynamic matching method in radiation transfer calculation according to claim 1, characterized in that: In step (2), the calculation formula of the ratio of aerosol optical thickness AROD is: AROD=AOD 470 / AOD 660 Among them, AOD 470 、AOD 660 Represent the aerosol optical thickness values ​​at 470nm and 660nm respectively.

4. The aerosol model dynamic matching method in radiation transfer calculation according to claim 1, characterized in that: The step (3) specifically refers to: If AROD>0.91, the aerosol type is judged to be dust type; If AROD≤0.91 and AOD 470 ≤0.45, the aerosol type is judged to be continental; If 0.68≤AROD≤0.91 and AOD 470 ≥0.45, the aerosol type is judged to be subcontinental; If AOD 470 ≥0.45 and 0.58≤AROD≤0.68, the aerosol type is judged to be urban industrial type; If AROD≤0.58 and AOD 470 ≥0.45, the aerosol type is judged to be biomass burning type; Based on AROD and AOD 470 The value of Aerosol Type i It is divided into five types: dust, continental, subcontinental, biomass burning and urban industrial. 470 represents the aerosol optical density at 470 nm.

5. The aerosol model dynamic matching method in radiation transfer calculation according to claim 1, characterized in that: The step (4) specifically refers to: for months in which the monthly average proportion of a certain aerosol type exceeds 90% for at least three consecutive years, directly matching the aerosol model corresponding to the aerosol type; otherwise, using the random forest algorithm to dynamically match the aerosol model: Aerosol Model i ~Wind Direction i +Source Direction i +Wind Speed i +Aerosol Type i Among them, Aerosol Model i Represents the dynamically matched aerosol model at the i-th moment, Wind Direction i Indicates the wind direction at the i-th moment, Source Direction i represents the location of the aerosol source at the i-th moment, Wind Speed i represents the wind speed at the ith moment.

Images