A method for inversion of urban high-resolution aerosol optical thickness based on a dual-satellite network
Through the method of Gaofen 1 and Gaofen 6 dual-star networking, combined with cloud/cloud shadow identification, surface reflectivity database construction and aerosol pattern refinement estimation, the optical thickness inversion of high-space resolution aerosol in complex urban contexts is solved. The problem of band restrictions and insufficient revisit cycles achieved high-precision aerosol monitoring.
Patent Information
- Application Number
- CN202310241912.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-14
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2043-03-14
AI Technical Summary
The existing technology is difficult to achieve high spatial resolution aerosol optical thickness inversion in complex urban contexts, especially due to the band limitations and insufficient revisit periods of traditional satellite sensors, which makes it difficult to meet the small-scale fine monitoring of urban aerosol emissions. need.
Using the method based on Gaofen 1 and Gaofen 6 dual-star networking, the identification of cloud/cloud shadows is combined with cross-radiation calibration, surface reflectivity database construction, aerosol mode refinement estimation and atmospheric radiation transmission model. Achieving high spatial resolution aerosol optical thickness inversion overcomes the problems of band restriction and insufficient revisit period in traditional methods.
Achieving high-precision optical thickness inversion of aerosols, breaking through the band limitations of traditional satellite sensors, improving the aerosol monitoring accuracy in urban areas, meeting the needs of small-scale refined monitoring, and shortening the revisit cycle through dual-star networking .
Smart Images

Figure CN116337701B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of atmospheric remote sensing, and in particular relates to a high spatial resolution aerosol optical thickness inversion method under a complex urban background based on a dual satellite network of GF-1 and GF-6. Background Art
[0002] Aerosols in the atmosphere come from various natural and anthropogenic sources, including the emission of primary particulate matter or the formation of secondary particulate matter from gaseous precursors. It is well known that aerosols can have negative impacts on human health, environmental quality and even climatology. Urban areas usually have high population density. Therefore, it is necessary to obtain high-resolution AOD inversion results at local scales over complex urban surfaces.
[0003] Most existing AOD data products use low- to medium-resolution remote sensing satellite data for inversion, and their spatial resolution is generally at the kilometer level, which is difficult to meet the needs of fine-scale monitoring of urban aerosol emissions. In order to obtain high-spatial-resolution AOD products in urban areas, domestic and foreign research institutions have conducted many studies.
[0004] Previous studies have proposed several algorithms for retrieving aerosol optical thickness over urban land with a resolution of 500 m using MODIS. The modified minimum reflectance technique (MRT) was used to determine the surface reflectance of MODIS data over Hong Kong and the Pearl River Delta in 2007 (Man Sing, W., Kwon-Ho, L., Nichol, J.E., & Zhanqing, L. (2010). Retrieval of Aerosol Optical Thickness Using MODIS 500×500 m 2, a Study in Hong Kong and the Pearl River Delta Region.IEEE Transactions on Geoscience and Remote Sensing,48,3318-3327). Based on the extended surface reflectance estimation scheme and dynamic aerosol model, an improved AOD inversion method for 500m MODIS data in Wuhan from 2016 to 2018 in winter is proposed (Jin,S.,Ma,Y.,Zhang,M.,Gong,W.,Dubovik,O.,Liu,B.,Shi,Y.,&Yang,C.(2019).Retrieval of 500m AerosolOptical Depths from MODIS Measurements over Urban Surfaces under HeavyAerosol Loading Conditions in Winter.Remote Sensing,11,2218).
[0005] However, all these AOD products with relatively coarse spatial resolution (≥500 meters) are difficult to fully meet the requirements of analyzing the structure of atmospheric pollution or its relationship with complex urban landscapes. Existing studies have also attempted to solve this defect by inverting high-resolution AOD using Landsat and Sentinel images (Lin, H., Li, S., Xing, J., Yang, J., Wang, Q., Dong, L., & Zeng, X. (2021). Fusing Retrievals of High Resolution Aerosol Optical Depth from Landsat-8 and Sentinel-2 Observations over Urban Areas. Remote Sensing, 13, 4140).
[0006] However, all these sensors can only obtain some effective observation images within a month, which greatly limits their application in the relevant research of urban air pollution, because urban air pollution has high-frequency changes. For Landsat data, the surface reflectance of densely vegetated areas and bright areas is estimated by combining the short-wave infrared (2.22μm) atmospheric top reflectance and the surface reflectance climate data records of Landsat 4-7. Wei J et al. inverted the AOD over Beijing from 2008 to 2016 (Wei, J., Huang, B., Sun, L., Zhang, Z., Wang, L., & Bilal, M. (2017). A Simple and Universal Aerosol Retrieval Algorithm for Landsat Series Images Over Complex Surfaces. Journal of Geophysical Research: Atmospheres, 122, 13, 338-313, 355). Multi-year MODIS bidirectional reflectivity distribution function (BRDF) / albedo model parameter products and RossThick-LiSparse kernel driver were selected as surface reflectance estimates, which are useful for retrieving AOD over Beijing and Madrid in 2013 and 2017 using Landsat8 data (Tian, X., Liu, Q., Song, Z., Dou, B., & Li, X. (2018). Aerosol Optical Depth Retrieval From Landsat 8 OLI Images Over Urban Areas Supported by MODIS BRDF / Albedo Data. IEEE Geoscience and Remote Sensing Letters, 15, 976-980). Based on machine learning algorithms, Landsat7 and 8 were used for AOD inversion in Beijing from 2013 to 2019 (Liang, T., Liang, S., Zou, L., Sun, L., Li, B., Lin, H., He, T., & Tian, F. (2022). Estimation of Aerosol Optical Depth at 30m Resolution Using Landsat Imagery and Machine Learning. Remote Sensing, 14, 1053).Sentinel-2 data are also used for high-resolution AOD inversion in urban areas. Yang et al. developed an aerosol inversion algorithm based on 60 meters (Yang, Y., Chen, Y., Yang, K., Cermak, J., & Chen, Y. (2021). High-resolution aerosol retrieval over urban areas using sentinel-2 data. Atmospheric Research, 264, 105829). AOD comes from the wide field of view (WFV) camera data on China's Gaofen-1 satellite, with a spatial resolution of 160 meters × 160 meters and a revisit period of 4 days (Sun, K., Chen, XL, Zhu, ZM, & Zhang, TH (2017). High Resolution Aerosol Optical Depth Retrieval Using Gaofen-1 WFV Camera Data. Remote Sensing, 9). This revisit period cannot meet the needs of urban AOD monitoring and application. Fortunately, with the launch of Gaofen-6 in June 2018, it will be networked with the Gaofen-1 satellite in orbit, greatly shortening the revisit interval. Their coordinated observations can achieve a revisit period of 1 to 2 days, which makes it possible to invert high-resolution AOD with an ideal revisit period.
[0007] Our previous study (Zhang, TH, Zhu, ZM, Gong, W., Zhu, ZR, Sun, K., Wang, LC, Huang, YS, Mao, FY, Shen, HF, Li, ZW, & Xu, K. (2018). Estimation of ultrahighresolution PM2.5 concentrations in urban areas using 160m Gaofen-1 AODretrievals. Remote Sensing of Environment, 216, 91-104) used the minimum reflectance technique to achieve AOD retrieval and application in three major cities in China. However, due to the huge changes in urban land surfaces, there are also some challenges in the urban high-resolution AOD retrieval algorithm. First, GaoFen WFV has only three visible light bands and bands (0.77-0.89μm), which will make cloud detection and AOD retrieval more difficult without SWIR (shortwave infrared), MIR (midwave infrared) and TIR (thermal infrared) bands. Limited bands make it impossible to apply the classic inversion algorithms MODIS DB and DT algorithms to high-resolution WFV data. The high resolution of 16 meters makes it difficult to decompose the top of the atmosphere reflectivity from the surface reflectivity and Rayleigh path reflectivity. At high spatial resolution, the BRDF effect in urban areas is also stronger. In addition, the radiation differences caused by multiple sensors also bring difficulties to the high-resolution AOD inversion. Summary of the invention
[0008] The purpose of the present invention is to overcome the deficiencies described in the background technology, and propose a high-resolution urban aerosol optical thickness inversion method based on a dual-satellite network, with "data preprocessing suitable for domestic satellites - ground-air decoupling suitable for urban surfaces - refined estimation of aerosol patterns" as the main research line. It is proposed to develop a high-spatial-resolution aerosol optical thickness (AOD) inversion method suitable for complex urban backgrounds. It is used to overcome the lack of spectral information observed by WFV / high-resolution satellite sensors, and improve the problem of inaccurate construction of the surface reflectance library due to the complex and bright urban surface. The implementation method includes the following steps:
[0009] Step 1: Identification and separation of clouds / cloud shadows, and cross-radiance calibration: Use multi-scale feature fusion based on spatial spectral adaptive correction to obtain high-precision cloud areas; and use Landsat8 OLI as the reference sensor to achieve cross-radiance calibration of the WFV sensors on the Gaofen-1 and Gaofen-6 satellites;
[0010] Step 2: Building a database of surface reflectance: Using a semi-empirical BRDF model that takes into account the significant characteristics of the urban back-hotspot effect, the surface reflectance results are obtained and input into the surface bidirectional reflectance parameter module corresponding to the solar spectrum vector (6SV) radiation transfer model;
[0011] Step 3, aerosol model refinement estimation: using the aerosol model refinement estimation results obtained based on ground-based remote sensing AERONET, input it into the regional custom aerosol model module corresponding to the 6SV atmospheric radiation transfer model;
[0012] Step 4: Improve the input parameter requirements of the 6SV atmospheric radiation transfer model, establish a lookup table and perform data matching based on TOA (Top Of Atmosphere, TOA) reflectivity to achieve AOD inversion;
[0013] Step 5: Verify and evaluate the spatial distribution characteristics of the high-resolution satellite aerosol optical depth by combining the AOD data from ground-based sites with other remote sensing aerosol products such as the MODIS aerosol optical depth product.
[0014] Moreover, the implementation method of step 1 is to overcome the disadvantage of the limited band of the high-resolution sensor, fully explore the spatial and spectral characteristics of the visible light and near-infrared bands, introduce the optical, geometric morphology, texture, and spatial and spectral correlation constraints of cloud / cloud shadow, non-cloud-bright surface and haze, mainly including considering the target texture and geometric characteristics, studying the multi-scale convolution feature fusion method, and realizing the recognition of cloud / cloud shadow and non-cloud-bright surface in urban areas; for cross-radiation calibration, Landsat8 OLI is used as the reference sensor to compare the simulated surface reflectance of Landsat-8OLI after spectral correction and BRDF correction and the reflectance of high-resolution satellite WFV after atmospheric correction. If the difference between them is small enough, that is, when the coefficient of variation of the gain coefficient in the radiation calibration coefficient obtained by 5 consecutive iterations is less than 0.01%, or when the number of iterations reaches 10,000 times, the convergence condition is met, then the optimal radiation calibration coefficient can be obtained. If the situation is the opposite, the radiation calibration coefficient is updated using the assimilation algorithm to calculate the new surface reflectance simulated by Landsat-8OLI. The same loop iteration is performed until the termination condition is reached. Finally, the surface reflectance measured on the ground from the RadCalNet website is used for verification.
[0015] Furthermore, step 2 is implemented by using the RPV (Rahman-Pinty-Verstraete) model given by equations (1)-(5) to decompose the observed angular surface reflectance into three independent components, representing intensity (amplitude), shape anisotropy, and asymmetry coefficient (relative proportion of forward and backscatter). Given the solar zenith angle θ s , observation zenith angle θv and relative azimuth The RPV model decomposes the observed angular surface reflectivity into three independent components, representing amplitude, shape anisotropy, and asymmetry coefficient. The RPV model finally calculates the surface bidirectional reflectivity Calculate according to the following formula.
[0016]
[0017] with
[0018]
[0019]
[0020]
[0021]
[0022] Among them, in formula (1) is the surface bidirectional reflectivity finally calculated by the RPV model. F(g) in formulas (1) and (2) is an intermediate variable (no actual physical meaning). H(G) in formulas (1) and (3) is the part describing the hotspot, which captures the backward hotspot effect that is very important for urban surfaces. cosg in formula (4) is the intermediate variable in formula 2 (no actual physical meaning). G in formula (5) is the intermediate variable in formula (3) (no actual physical meaning). The parameter ρ0∈[0,1[ in formulas (1) and (3) describes the intensity (amplitude) of the target, which is independent of the angle change. The parameter k∈[0,2[ in formula (1) represents the anisotropy of the target. A k value less than 1.0 represents a bowl-shaped anisotropic mode, and a k value greater than 1.0 represents a bell-shaped anisotropic mode. The ideal case of k=0 represents a Lambertian surface. Finally, the asymmetry coefficient in formula (2) Controlled forward scatter and backscatter relative proportion.
[0023] Moreover, step 3 is implemented by making a refined estimate of the aerosol pattern on a monthly time scale and a regional spatial scale; using AERONET ground-based observation data, comprehensively analyzing the spatiotemporal characteristics of aerosol optical and physical parameters (extinction coefficient, absorption coefficient, scattering coefficient, and particle size spectrum, etc.), and importing the above parameters into the Mie scattering model to calculate the single scattering albedo (SSA), absorption coefficient, and complex refraction index, etc.; based on monthly statistical analysis, extracting optical characteristics that can indicate the urban aerosol pattern from a large number of aerosol optical parameters such as single scattering albedo, absorption coefficient, and complex refraction index to obtain a refined aerosol pattern.
[0024] Moreover, the implementation method of step 4 is to improve the requirements of the input parameters of the 6SV atmospheric radiation transfer model, establish a lookup table and perform data matching according to the TOA reflectivity. According to the requirements of the input parameters of the 6SV atmospheric radiation transfer model, other modules of the input file are customized, including geometric parameters, atmospheric model, target altitude, and the calculation mainly includes the atmospheric self reflectivity ρ, atmospheric total transmittance T, atmospheric lower hemisphere albedo S, aerosol optical thickness τ, water vapor ozone content δ, altitude h and other parameters. For each high-resolution image to be inverted, the AOD inversion is achieved by matching its TOA reflectivity with the TOA reflectivity simulated by the lookup table.
[0025] Spectral reflectance of the top of the atmosphere as observed by satellite sensors under the assumptions of a plane-parallel atmosphere and a non-Lambertian ground surface The expression is,
[0026]
[0027]
[0028] Among them, θ s and θ v are the solar and observation zenith angles respectively; θ s and θ v are the sun and observation azimuth respectively; is the relative azimuth angle between the sun and the satellite; T g is the total gas transmittance; is the atmospheric path reflectivity directly contributed by aerosols and atmospheric molecules. T↓(23) and T↑(24) are the atmospheric downward transmittance through the sun and the atmospheric upward transmittance through the sensor viewing direction, respectively; is the bidirectional spectral reflectance of the surface; S is the hemispherical albedo of the atmosphere. As we all know, the essence of AOD inversion is the decoupling of the atmosphere and the surface, which mainly includes the intrinsic contribution of the atmosphere and the coupled contribution of the atmosphere and the surface. Except for the bidirectional reflectance of the surface, all parameters on the right side of formula (6) are functions of the atmospheric conditions and the sun-satellite geometry. For the corresponding spectral response of the sensor, these parameters can be set as inputs by assuming the aerosol model, and the solar spectral vector (6SV) radiation transfer model is used to perform a second simulation of the satellite signal under different geometric conditions.
[0029] Moreover, the implementation method of step 5 is as follows: Finally, the AOD data observed by the remote sensing AERONET is used to conduct multi-point verification at the urban area scale. According to the imaging time of each scene of remote sensing data, the aerosol optical thickness values of the ground-based stations within half an hour before and after the satellite transit are averaged to obtain the aerosol optical thickness of each station for evaluation and verification of the inversion results. In addition, this study will also use other remote sensing aerosol products such as MODIS aerosol optical thickness data to verify and evaluate the spatial distribution characteristics of high-resolution satellite aerosol optical thickness. The expression is:
[0030] EE=±(0.15×AOD (Aeronet) +0.05) (7)
[0031]
[0032] Among them, EE represents the expression of EE line, RMSE represents the root mean square error, AOD (Aeronet) Represents the AOD value observed at the AERONET station. (GaoFen) Represents the AOD value obtained by Gaofen satellite inversion.
[0033] The verification results show that the WFV AOD inverted by this method shows good consistency with the ground AERONET AOD, and the determination coefficient R 2 The match ratio reaches 0.8, and nearly 70% of the matches are within the expected error range (±0.05±0.15×AOD AERONET ), indicating that this method can achieve high-precision inversion of aerosol optical thickness in urban areas using high-resolution satellites.
[0034] The present invention has the following advantages and positive effects:
[0035] 1) Breaking through the traditional band limitations of Gaofen satellites, achieving cloud / cloud shadow separation, and implementing cross-radiation calibration steps for the networking of Gaofen-1 and Gaofen-6. These pre-processing solutions tailored for WFV help improve the accuracy of high spatial resolution AOD inversion in urban areas;
[0036] 2) The use of a semi-empirical BRDF model that takes into account the significant characteristics of the backward hotspot effect in urban areas can solve the problem of inaccuracy in the traditional surface reflectance library;
[0037] 3) By networking GF-1 and GF-6, high spatial resolution and short revisit period are achieved, breaking the contradiction between the two in traditional AOD products and fully meeting the small-scale and refined monitoring needs of urban aerosol emissions;
[0038] This method can be widely used in related industries such as environmental protection and environmental exposure health research. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 It is a flow chart of an embodiment of the present invention, which optimizes three modules of domestic satellite WFV data preprocessing, database construction considering the surface bidirectional reflection effect, and aerosol pattern estimation under the framework of atmospheric radiation transmission.
[0040] Figure 2 It is the surface reflectivity result of step 2 of the embodiment of the present invention.
[0041] Figure 3 is an inversion accuracy comparison image of an embodiment of the present invention, wherein Figure 3 (a) Comparison and verification of the high-resolution WFV AOD inversion results obtained by the method of the present invention with those of three AERONET stations in Beijing, Figure 3 (b) is the Beijing site among the three sites. Figure 3 (c) is the Beijing CAMS site among the three sites. Figure 3 (d) is the Xianghe site among the three sites.
[0042] Figure 4 is a comparison diagram of the effect images of the embodiment of the present invention, wherein Figure 4 (a) is the WFV AOD inverted using this method on May 21, 2014. Figure 4 (b) is the true color image of the day. Figure 4 (c) is the MODIS DB AOD product, Figure 4 (d) is MODIS DT AOD, Figure 4 (e) is MODIS MAIAC AOD. DETAILED DESCRIPTION
[0043] In order to facilitate ordinary technicians in the field to understand and implement the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the implementation examples described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0044] In order to overcome the limitations of the prior art, the present invention proposes a high-resolution urban aerosol optical thickness inversion method based on a dual-satellite network. This method overcomes the disadvantage of the limited band of the high-resolution sensor, realizes the recognition of cloud / cloud shadow in the urban area based on spectral features, geometric features and texture features, and uses Landsat8 OLI as a reference sensor to realize the cross-radiation calibration of the GF-1 and GF-6 WFV sensors; secondly, a surface reflectance library suitable for complex cities at high spatial resolution scales is constructed to achieve surface-atmosphere signal decoupling; thirdly, a refined estimation scheme of the urban aerosol model based on AErosol RObotic NETwork (AERONET) is adopted, and finally, the aerosol optical thickness (AOD) data product based on the domestic satellite GF-1 and GF-6 satellite network is obtained under the framework of the atmospheric radiation transfer model. Compared with traditional methods, this method overcomes the disadvantage of limited band of high-resolution sensors, uses the BRDF model suitable for complex urban surfaces to establish a surface reflectance library, solves the inaccuracy problem of traditional library construction methods, and uses a dual-satellite network to double the revisit period of 2-4 days in the past. While obtaining high spatial resolution observation data (16m), the revisit period is increased from 2-4 days to 1-2 days, solving the problem of the contradiction between high spatial resolution and short revisit period of previous AOD products.
[0045] See also Figure 1 The embodiment of the present invention is based on this method, such as Figure 1 The inversion process shown in the figure is to invert the aerosol optical thickness of a certain area on May 21, 2014 under the condition of few clouds and heavy pollution. The specific calculation steps are as follows.
[0046] Step 1: Identification and cross-radiance calibration of clouds / cloud shadows. This method can effectively improve the accuracy of AOD inversion of domestic satellite WFV.
[0047] Step 1.1: In view of the limited band characteristics of the WFV sensor of the high-resolution satellite, multi-scale feature fusion based on spatial spectrum adaptive correction is used to obtain high-precision cloud areas;
[0048] Step 1.2, and use Landsat8 OLI as the reference sensor to achieve cross-radiometric calibration of the WFV sensors onboard the GF-1 and GF-6 satellites.
[0049] Step 2: Input the surface reflectance result into the 6SV atmospheric radiation transfer model. Using the semi-empirical BRDF model that takes into account the significant characteristics of the urban back-hotspot effect, the ground-air decoupling of the complex urban surface is performed, and the following is obtained: Figure 2 The surface reflectivity results shown are input into the surface bidirectional reflectivity parameter module corresponding to the 6SV atmospheric radiation transfer model.
[0050] Step 3: Input the refined estimation results of the aerosol model. The refined estimation results of the monthly aerosol model are obtained by using the ground-based remote sensing AERONET as shown in Table 1, and then input into the regional custom aerosol model module corresponding to the 6SV atmospheric radiation transfer model.
[0051] Table 1
[0052]
[0053]
[0054] Step 4: Improve the requirements of the 6SV atmospheric radiation transfer model input parameters, establish a lookup table and perform data matching based on the TOA reflectivity. Customize other modules of the input file according to the requirements of the 6SV atmospheric radiation transfer model input parameters, including geometric parameters, atmospheric model, and target altitude. The calculation mainly includes the atmospheric reflectivity ρ, atmospheric total transmittance T, atmospheric lower hemisphere albedo S, aerosol optical thickness τ, water vapor ozone content δ, altitude h and other parameters. For the high-resolution image to be inverted, the AOD inversion is achieved by matching the TOA reflectivity of the top of the atmosphere with the TOA reflectivity simulated using the lookup table.
[0055] Step 5: Verify by combining site data with other remote sensing aerosol products. The specific implementation steps are as follows:
[0056] Step 5.1: Perform multi-point verification at the urban area scale. According to the imaging time of each scene of remote sensing data, the aerosol optical thickness values of the ground-based stations within half an hour before and after the satellite transit are averaged to obtain the aerosol optical thickness of each station for evaluation and verification of the inversion results. Figure 3 As shown, in this study, AERONET AOD was used to verify the high-resolution WFV AOD on 550 nm measurements in two seasons, as Figure 3 As shown in (a), the comparison of three AERONET stations shows that the retrieval accuracy of high-resolution AOD is comparable in both urban and rural areas. The comparison between seasons also reveals that high-resolution AOD has higher accuracy in different periods of the year.
[0057] from Figure 3 (b)-(d) show that the Xianghe station in the rural area ( Figure 3 (d)) and the urban areas of Beijing ( Figure 3 (b)) and Beijing CAMS station ( Figure 3 (c) There is almost no difference in accuracy, which indicates that the performance of the RPV model is almost unaffected by different land cover types. In order to ensure the reliability and fairness of the comparison strategy, the comparison of the validation results of four seasons was further conducted.
[0058] Step 5.2: In addition, this study will also use other remote sensing aerosol products such as MODIS aerosol optical depth products to verify and evaluate the spatial distribution characteristics of high-resolution satellite aerosol optical depth. Figure 3 The heavy pollution days shown are from Figure 4 As can be seen from (b), on that day, only the southeastern part of Beijing had clouds, while the central and southern parts had haze. DB AOD( Figure 4 (c)) The cloud area has not been removed, and there is a clear area in the west. DT AOD( Figure 4 (d)) and MAIAC AOD( Figure 4 (e)) also has many invalid values, which may be the result of mist being mistakenly identified as cloud and removed, while the high-score AOD ( Figure 4 (a) retains the complete distribution of AOD, especially in the central area of the city, and shows the spatial details of haze. The transport mode of pollutants may be exogenous transport pollution, which moves from southwest Beijing to the city center, and the AOD concentration changes from strong to weak.
[0059] In general, the comparison among the four AOD datasets shows that the 10km resolution MODIS DB AOD has a tendency to underestimate the high surface brightness aerosol load in the urban center of southeastern Beijing. The MODIS DT AOD at 3km resolution has obvious limitations in spatial coverage of urban areas. The 1km resolution MAIAC AOD performs better than the above two MODID AOD products in the experimental area, but performs poorly in capturing the spatial heterogeneity of small areas less than 1 square kilometer, and even the data is very smooth, making it difficult to truly reflect the spatial distribution of 1km resolution. Some missing retrieval values also appear in the MAIAC AOD, especially in urban areas. The 160m resolution high-resolution AOD shows continuous and clear spatial details in the study area, which can better illustrate the spatial resolution changes of aerosols in Beijing, especially in urban areas.
[0060] During specific implementation, the present invention can use computer software technology to realize automatic operation process.
[0061] It should be understood that the above description of the preferred embodiment is relatively detailed, and it cannot be regarded as limiting the scope of protection of the patent of the present invention. Under the enlightenment of the present invention, ordinary technicians in this field can also make substitutions and modifications without departing from the scope protected by the patent requirements of the present invention, all of which fall within the scope of protection of the present invention. The scope of protection requested by the present invention shall be based on the attached claims.
Claims
1. A method for inverting high-resolution urban aerosol optical thickness based on a dual-satellite network, characterized by: The method comprises the following steps: Step 1: Identification and separation of clouds / cloud shadows, and cross-radiance calibration: high-precision cloud areas are obtained by multi-scale feature fusion based on spatial spectral adaptive correction; Landsat-8OLI is used as the reference sensor to implement cross-radiance calibration of the WFV sensor carried on the Gaofen-1 and Gaofen-6 satellites; the identification and separation of clouds / cloud shadows is achieved by introducing the optical, geometric morphology, texture, and spatial and spectral correlation constraints of clouds / cloud shadows, non-cloud bright surfaces, and haze. The implementation methods include: (1) considering the target texture and geometric features, using a multi-scale convolution feature fusion method , realizing the recognition of clouds / cloud shadows and non-cloud bright surfaces in urban areas; (2) overcoming the disadvantage of limited bands of high-resolution sensors, fully exploring the spatial and spectral characteristics of visible light and near-infrared bands, developing a spatial-spectral adaptive cloud and haze recognition method, and realizing the separation of clouds and haze; (3) using the cloud and haze recognition data in step (2) to correct the cloud detection results in step (1), separate the haze pixels that were misjudged as clouds in step (1), and finally obtain an optimized cloud mask, solving the problem of AOD inversion failure after haze in urban areas is misjudged as clouds, and providing an accurate cloud mask for subsequent surface-atmosphere signal separation; Step 2: Build a database of surface reflectance: Using a semi-empirical BRDF model that takes into account the significant characteristics of the urban back-hotspot effect, the ground-air decoupling of the complex urban surface is performed to obtain the surface reflectance results, which are then input into the surface bidirectional reflectance parameter module corresponding to the solar spectrum vector 6SV radiation transfer model. Specifically, the RPV model is used to build a database of WFV surface reflectance, which includes three steps: (1) First, collect the WFV apparent reflectance corrected by clear air for a long time series; (2) Then, the WFV surface bidirectional reflectance data is obtained through atmospheric correction to fit the RPV model, and the MODIS BRDF parameters are set as initial values based on the simulated annealing algorithm; (3) Finally, similar to the quality evaluation of BRDF / albedo inversion in the MODIS business system, the root mean square error (RMSE) and decision weight (WoD) are used to evaluate the accuracy of WFV surface BRDF modeling; in addition, the threshold combination of these two indicators is used to filter out pixels that are not suitable for BRDF modeling in all seasons to avoid relatively large uncertainties in AOD inversion. Step 3, refined estimation of aerosol model: using the refined estimation results of aerosol model obtained based on ground-based remote sensing AERONET, input it into the regional custom aerosol model module corresponding to the 6SV radiation transfer model; Step 4: Improve the input parameter requirements of the 6SV radiation transfer model, establish a lookup table and perform data matching based on the top of atmosphere reflectivity TOA to achieve AOD inversion; Step 5: Verify and evaluate the spatial distribution characteristics of the high-resolution satellite aerosol optical depth by combining the AOD data from ground-based sites and the aerosol optical depth product of the remote sensing aerosol product MODIS.
2. The method for inverting urban high-resolution aerosol optical thickness based on a dual-satellite network as claimed in claim 1, characterized in that: For cross-radiance calibration, Landsat-8OLI is used as the reference sensor to compare the surface reflectance simulated by Landsat-8OLI after spectral correction and BRDF correction and the reflectance of the high-resolution satellite WFV after atmospheric correction. If the difference between them is small enough, that is, when the coefficient of variation of the gain coefficient in the radiation calibration coefficient obtained after 5 consecutive iterations is less than 0.01%, or when the number of iterations reaches 10,000, the convergence condition is met, then the optimal radiation calibration coefficient is obtained; if the opposite is true, the radiation calibration coefficient is updated using the assimilation algorithm, and the new Landsat-8OLI simulated surface reflectance is calculated; the same loop iteration is performed until the termination condition is reached; finally, the surface reflectance measured on the ground on the RadCalNet website is used for verification.
3. The method for inverting urban high-resolution aerosol optical thickness based on a dual-satellite network as claimed in claim 1, characterized in that: Step 2 is implemented as follows: The RPV model given by equations (1)-(5) decomposes the observed angular surface reflectance into three independent components, representing intensity, shape anisotropy, and asymmetry coefficient; given the solar zenith angle θ s , observation zenith angle θ v and relative azimuth The surface bidirectional reflectivity finally calculated by the RPV model Calculated according to the following formula: Among them, in formula (1) is the surface bidirectional reflectivity finally calculated by the RPV model. is an intermediate variable with no actual physical meaning. H(G,ρ0) in formulas (1) and (3) is the part describing the hotspot. What is captured here is the backward hotspot effect that is very important for urban surfaces. cosg in formula (4) is an intermediate variable in formula (2) with no actual physical meaning. G in formula (5) is an intermediate variable in formula (3) with no actual physical meaning. The parameter ρ0∈[0,1] in formulas (1) and (3) describes the intensity of the target, that is, the amplitude, which is independent of the angle change. The parameter k∈[0,2] in formula (1) represents the anisotropy of the target. A k value less than 1.0 represents a bowl-shaped anisotropic mode, a k value greater than 1.0 represents a bell-shaped anisotropic mode, and the ideal case of k=0 represents a Lambertian surface. Finally, the asymmetry coefficient in formula (2) The relative proportion of forward scattering and backscattering is controlled.
4. The method for inverting urban high-resolution aerosol optical thickness based on a dual-satellite network as claimed in claim 1, characterized in that: Step 3 is implemented by making a refined estimate of the aerosol pattern on a monthly time scale and a regional spatial scale; using AERONET ground-based observation data, comprehensively analyzing aerosol optical and physical parameters, including the spatiotemporal characteristics of the extinction coefficient, absorption coefficient, scattering coefficient, and particle size spectrum, and importing the above parameters into the Mie scattering model to calculate the single scattering albedo SSA, absorption coefficient, and complex refraction index; based on monthly statistical analysis, extracting optical features that can indicate urban aerosol patterns from a variety of aerosol optical parameters such as single scattering albedo, absorption coefficient, and complex refraction index, and obtaining a refined aerosol pattern for each month.
5. The method for inverting urban high-resolution aerosol optical thickness based on a dual-satellite network as claimed in claim 1, characterized in that: Step 4 is implemented by improving the requirements of the 6SV radiation transfer model input parameters, establishing a lookup table and matching data according to the TOA reflectivity; customizing other modules of the input file according to the requirements of the 6SV radiation transfer model input parameters, including geometric parameters, atmospheric model, target altitude, and calculating the parameters including the atmospheric self reflectivity ρ, atmospheric total transmittance T, atmospheric lower hemisphere albedo S, aerosol optical thickness τ, water vapor ozone content δ and altitude h; for each high-resolution image to be inverted, the TOA reflectivity is matched with the TOA reflectivity simulated by the lookup table to achieve AOD inversion; Spectral reflectance of the top of the atmosphere as observed by satellite sensors under the assumptions of a plane-parallel atmosphere and a non-Lambertian ground surface The expression is, Among them, θ s and θ v They are the solar zenith angle and the observation zenith angle; and are the solar azimuth and the observation azimuth respectively; is the relative azimuth angle between the sun and the satellite; T g is the total gas transmittance; is the atmospheric path reflectivity directly contributed by aerosols and atmospheric molecules, T ↓ (θ s ) and T ↑ (θ v ) are respectively the atmospheric downward transmittance through the sun direction and the atmospheric upward transmittance through the sensor viewing angle direction; is the bidirectional spectral reflectance of the surface; S is the hemispherical albedo of the atmosphere; Except for the surface bidirectional reflectance, all parameters on the right side of formula (6) are functions of atmospheric conditions and sun-satellite geometry; for the corresponding spectral response of the sensor, these parameters are set as inputs by assumptions about the aerosol model, and the solar spectrum vector 6SV radiation transfer model is used to perform a second simulation of the satellite signal under different geometric conditions.
6. The method for inverting urban high-resolution aerosol optical thickness based on a dual-satellite network as claimed in claim 1, characterized in that: Step 5 is implemented by using the observation data of ground-based remote sensing AERONET to conduct multi-point verification at the urban regional scale. According to the imaging time of each scene of remote sensing data, the aerosol optical thickness values of the ground-based stations within half an hour before and after the satellite transit are averaged to obtain the aerosol optical thickness of each station for evaluation and verification of the inversion results. In addition, the aerosol optical thickness data of the remote sensing aerosol product MODIS are used to verify and evaluate the spatial distribution characteristics of the aerosol optical thickness of the high-resolution satellite. The selected accuracy evaluation indicators are the root mean square error RMSE and EE line, which are expressed as follows: EE=±(0.15×AERONET_AOD+0.05) (7) Among them, EE represents the expression of EE line, RMSE represents the root mean square error, AOD (Aeronet) represents the AOD value observed at the AERONET station. (GaoFen) Represents the AOD value obtained by Gaofen satellite inversion.
Citation Information
Patent Citations
Retrieval method for aerosol optical thickness based on high resolution satellite image data
CN106407656A
Urban complex surface reflectance estimation method supporting high-resolution aerosol optical thickness inversion
CN113324915A
Methods for atmospheric correction of solar-wavelength Hyperspectral imagery over land
US20020096622A1
Atmospheric compensation in satellite imagery
US20140270332A1
Modeled atmospheric correction objects
US20140270502A1
