A three-dimensional atmospheric particulate matter high-precision inversion method and device of a multi-beam laser radar

Abstract

The application discloses a three-dimensional atmospheric particulate matter high-precision inversion method and equipment of a multi-beam laser radar, and belongs to the field of atmospheric detection and remote sensing. A multi-beam observation system with high center configuration and low edge configuration is faced, a multi-source data set containing a multi-beam laser radar vertical observation profile of a center beam and an outer beam and auxiliary data is established, atmospheric particulate matter inversion containing hierarchical identification, optical parameter inversion, layer parameter acquisition and cloud and aerosol classification is performed; meanwhile, a transferable condition of a center beam core parameter is calculated in a along-track direction, and a center beam transfer level matching is performed in a cross-track direction; core parameter extrapolation is performed according to the transfer level matching result, and atmospheric particulate matter inversion of the outer beam is performed again by using the extrapolated parameter. The application solves the problem that the existing laser radar cannot realize regional three-dimensional atmospheric particulate matter high-precision inversion, and reduces the extinction coefficient inversion error of the outer beam to 10%-25%.

Description

Technical Field

[0001] This invention belongs to the field of atmospheric detection and remote sensing technology, and particularly relates to a method and equipment for high-precision inversion of three-dimensional atmospheric particulate matter using a multi-beam lidar. Background Technology

[0002] The vertical optical and microphysical properties of atmospheric particulate matter, such as clouds and aerosols, are crucial parameters for quantifying their meteorological and climatic impacts and hold immense scientific and applied value. However, atmospheric particulate matter is widely and unevenly distributed globally, exhibiting different properties at different altitudes and potentially existing in complex states such as multi-layered cloud superposition or mixing with aerosols. This places high demands on the temporal and spatial resolution, coverage, and vertical resolution of modern cloud detection methods. Spaceborne remote sensing technology provides a global perspective for cloud detection. However, traditional passive remote sensing payloads, such as imaging spectrometers, lack vertical resolution and can only extract cloud signals from total radiation, making them susceptible to interference from surface signals and sub-cloud radiation. Active remote sensing, represented by lidar, can measure the vertical structure of the atmosphere with high precision, but due to limitations in laser emission angle, a single laser beam cannot meet the needs of regional three-dimensional atmospheric detection in terms of spatial coverage and observation frequency.

[0003] To address this problem, multi-beam lidar, which has achieved success in spaceborne ranging, offers a solution. It involves splitting a single laser beam into multiple beams using diffractive optical elements (DOEs) and distributing them across a swath of several kilometers to tens of kilometers, creating a regional three-dimensional detection capability. However, since ranging lidar only needs to measure the highly reflective surface of the Earth's surface, its design often employs a single-wavelength, high-energy pulsed laser, resulting in relatively simple echo signal processing. In contrast, lidar for measuring atmospheric particulate matter typically uses multiple wavelengths to invert information such as cloud and aerosol layer, type, and particle size. For high-precision inversion, it also requires the addition of high-spectral-resolution channels and Raman channels. Because atmospheric signals are much weaker than surface signals, in addition to the aforementioned high-precision detectors, atmospheric detection lidar also needs to be equipped with large-aperture receivers, multi-channel parallel processing optical and electronic systems to receive and independently process signals of different wavelengths and polarization states, and other components to assist in extremely high stability and environmental control. Therefore, based on current engineering conditions and cost calculations, it is not feasible to simply configure multiple high-performance atmospheric detection beams to achieve high-precision regional three-dimensional atmospheric particulate matter inversion.

[0004] To address the aforementioned issues, reference 1 (Design and Simulation of a New Generation Multibeam Cloud Measurement Lidar for Spaceborne Applications [J], Acta Optica Sinica, 2024, 44, 18) proposes a design with a high-configuration central beam (hyperspectral + multi-wavelength + polarization) and single-wavelength outer beams. By arranging multiple laser beams along the track crossing direction and combining their performance, the overall high-precision cloud inversion is achieved while significantly expanding the observation range, providing a new technical approach for acquiring the three-dimensional structure of atmospheric particulate matter. The paper mentions that the hyperspectral detection data of the core beam provides the lidar ratio for cloudless areas above 15 km altitude for the edge beams, indirectly achieving the hyperspectral detection effect of a nine-beam system. The lidar ratio, i.e., the ratio of extinction coefficient to backscattering coefficient, is a key parameter for high-precision cloud and aerosol inversion. However, the lidar ratios of different atmospheric particulate matter vary greatly; using only the lidar in cloudless areas as a reference cannot help the outer beams achieve high-precision inversion.

[0005] Reference 2 (Lidar Ratio Regional Transfer Method for Extinction Coefficient Accuracy Improvement in Lidar Networks [J], Remote Sens, 2022, 14) proposes a regional transfer method for lidar ratio, aiming to establish a nonlinear regression model (LR-AFNR) of lidar ratio and aerosol absorption ratio. This model transfers the lidar ratio retrieved from advanced lidar systems (such as Raman lidar or hyperspectral resolution lidar) to Mie scattering lidar sites within a certain range, eliminating the need for the latter to assume a lidar ratio during inversion, thus improving accuracy. This method is applicable to ground-based lidar networks, but the observation area in the vertical direction differs significantly from that of spaceborne lidar. Furthermore, this model relies on the light absorption properties of aerosols and is only applicable to a few types of aerosols such as dust and smoke, and cannot be used for clouds and non-absorbing aerosols.

[0006] Reference 3 (Analysis of global three-dimensional aerosol structure with spectral radiance matching [J], 2019, AMT, 12, 6541–6556) proposes a method for calculating the three-dimensional distribution of regional clouds and aerosols by fusing active and passive remote sensing data. This method establishes a one-to-one correspondence between atmospheric vertical profiles and multi-channel radiometric data by spatiotemporally matching lidar footprints with passive remote sensing pixels. However, this method, which uses passive remote sensing radiometric data as the mapping condition, faces problems such as limited correlation of vertical features and insufficient constraints in the extrapolation process. It is prone to masking weak-level signals during matching, leading to incorrect generation or omission of thin cloud and aerosol layers during extrapolation, resulting in deviations in key parameters such as cloud top height and cloud phase. Furthermore, since it relies solely on the central lidar profile as the vertical data source, the accuracy inevitably decreases gradually as it expands outwards. Summary of the Invention

[0007] To address the problem that existing lidar systems cannot achieve high-precision 3D atmospheric particulate matter inversion in a regional context, this invention provides a method and apparatus for high-precision 3D atmospheric particulate matter inversion using a multi-beam lidar. This method involves performing atmospheric particulate matter inversion on each beam of a multi-source dataset established using a multi-beam lidar, including hierarchical identification, optical parameter inversion, layer parameter and cloud and aerosol classification. Simultaneously, it calculates the transferability conditions of core parameters along the track direction and performs hierarchical matching along the track-crossing direction. Based on the matching results, the core parameters are extrapolated for re-inversion of the outer beams, resulting in complete high-precision 3D atmospheric particulate matter optical parameters and hierarchical information.

[0008] To achieve the above-mentioned objectives, an embodiment provides a high-precision inversion method for three-dimensional atmospheric particulate matter using a multi-beam lidar, comprising the following steps: Acquire vertical observation profiles and auxiliary data from a multi-beam lidar system, including the center beam and outer beams, and establish a multi-source dataset. Atmospheric particulate matter inversion is performed on each beam in the multi-source dataset, including hierarchical identification, optical parameter inversion, layer parameter acquisition, and cloud and aerosol classification. At the same time, the transferability conditions of the core parameters of the center beam are calculated in the track-along direction, and the transfer hierarchy matching of the center beam is performed in the track-crossing direction. Based on the transmission hierarchy matching results, the core parameters are extrapolated, and the extrapolated parameters are used to re-invert atmospheric particulate matter on the outer beam to obtain three-dimensional atmospheric particulate matter optical parameters and hierarchy information.

[0009] In one embodiment, the central beam is a high-spectral-resolution lidar with multi-wavelength and polarization detection capabilities; the outer beam is a single-wavelength Mie scattering lidar.

[0010] In one embodiment, the vertical observation profile is L1 level observation data of lidar, including attenuated backscattering coefficients and signal-to-noise ratios for different wavelengths and channels; The auxiliary data includes the latitude and longitude of the lidar beam and the land surface type.

[0011] In one embodiment, the hierarchical recognition is performed using a gradient method, including: Based on each beam in the multi-source dataset, the vertical gradient is calculated according to the abrupt change in the intensity of the attenuated backscattering coefficient in the vertical direction. ,in Distance Attenuated backscattering coefficient at the location; The vertical gradient is greater than the positive threshold. The location is marked as the bottom of the layer, and the vertical gradient is less than the negative threshold. The location is marked as the top of the layer, thus identifying the layers of clouds or aerosols.

[0012] Furthermore, before performing hierarchical recognition, the beam profiles are smoothed. The aim is to ensure that the signal-to-noise ratio (SNR) of the center beam is no less than 10 and the SNR of the outer beams is no less than 3, thus guaranteeing that the SNR of the outer low-performance channels meets the requirements for hierarchical recognition.

[0013] Furthermore, the vertical gradient is between 2 and 5 times.

[0014] In one embodiment, the optical parameter inversion is based on lidar equations. In the formula, The distance received by the lidar The echo signal power consists of four parts: These are system parameters for lidar that are independent of distance. For lidar geometric factors, Distance The backscattering coefficient at that location, For the laser from the lidar transmitter to the distance The loss during the process of returning to the lidar; Among them, the backscattering coefficient and extinction coefficient of the center beam are directly solved using the high spectral resolution lidar inversion process. For the outer beam, the Klett method or Fernald method is used, which assumes that the lidar ratio is used to invert the scattering coefficient and extinction coefficient.

[0015] In one embodiment, the layer parameters are obtained by performing layer integration or averaging on the optical parameters; wherein, the optical thickness of the layer is obtained by summing the extinction coefficient between the heights of the top and bottom layers; and the layer parameters, including the backscattering coefficient, depolarization ratio, lidar ratio, and color ratio, are obtained by calculating the average value between the heights of the top and bottom layers.

[0016] In one embodiment, the cloud and aerosol classification includes: Based on the retrieved optical parameters, the identified layers are divided into clouds and aerosols. Further, combined with the LiDAR configuration, they are further subdivided into different phases and types of subclasses. The classification index is represented by a probability density function established using feature parameters from the LiDAR and auxiliary information; the calculation formula is as follows: , in, This represents the probability that the layer is an aerosol. This represents the probability that the layer is a cloud. The value of is between -1 and 1. A positive value indicates that the layer is classified as an aerosol, while a negative value indicates that the layer is classified as a cloud.

[0017] Furthermore, the characteristic parameters include longitude, latitude, backscattering coefficient, and extinction coefficient; the center beam also includes at least one of depolarization ratio, lidar ratio, and color ratio, to further classify clouds into ice clouds, water clouds, and mixed clouds, and / or classify aerosols into marine type, smoke type, clean continental type, polluted continental type, dust type, and polluted dust type.

[0018] In one embodiment, the transitivity condition for calculating the core parameters of the center beam along the track direction includes: Let the threshold values ​​for the transmission distance in the cross-track direction of clouds and aerosols be respectively... and For each cloud or aerosol layer on the central beam, the distance along the track direction is statistically calculated. times Hierarchical data within the range, of which The distance from the center beam footprint to the edge beam footprint. ; For each level, the maximum distance satisfying all of the following conditions is used as the transmission distance threshold: the level type remains unchanged along the track direction; the level top height along the track direction is constant. and floor height The variation does not exceed 3 km; the variation of the average lidar ratio along the track direction is less than 15%.

[0019] Furthermore, the conditions for maintaining the unchanged hierarchical type along the track direction also include: ideally, the hierarchical subclass type along the track direction remains unchanged; when the surface type changes, data within the same surface type range is used, or twice the data is used. Data within the range.

[0020] In one embodiment, the transmission hierarchy matching of the center beam in the track crossing direction includes: Based on the layer type, find the layer with the same type and the closest average layer height on the center beam; and require that the difference between the layer thickness in the track crossing direction and the thickness of the selected layer does not exceed 3 km.

[0021] In one embodiment, the extrapolation of core parameters based on the transmission layer matching result includes: assigning the lidar ratio parameter of the center beam layer to the specific profile in the corresponding layer of the outer beam; the specific profile is the profile with the smallest sum of the height difference between the top and bottom layers and the height of the selected layer within the transmittable distance threshold range of the matching layer.

[0022] In one embodiment, the method of re-inverting atmospheric particulate matter using extrapolated parameters for the outer beam includes: re-inverting the L1 level observation data of the outer beam using the same high-precision inversion method as the central beam, namely optical parameter inversion, layer parameter acquisition, and cloud and aerosol classification, to obtain high-precision three-dimensional atmospheric particulate matter optical parameters and layer information.

[0023] The present invention also provides a computing device, including a memory and one or more processors, wherein the memory stores executable code, and the one or more processors execute the executable code to implement the three-dimensional high-precision inversion method for atmospheric particulate matter from the multi-beam lidar.

[0024] Compared with the prior art, the beneficial effects of the present invention include at least the following: The present invention provides a high-precision inversion method and device for three-dimensional atmospheric particulate matter using a multi-beam lidar. Compared with existing technologies, this invention is geared towards multi-beam observation systems with high center-mounted and low-mounted edge-mounted configurations. It establishes a multi-source dataset including the vertical observation profiles of the central and outer beams of the multi-beam lidar, along with auxiliary data. It performs atmospheric particulate matter inversion, including hierarchical identification, optical parameter inversion, layer parameter acquisition, and cloud and aerosol classification. Simultaneously, it calculates the transferability conditions of the core parameters of the central beam along the track direction and performs transfer hierarchy matching of the central beam in the track-crossing direction. Based on the transfer hierarchy matching results, core parameters are extrapolated, and atmospheric particulate matter is re-inverted for the outer beams using the extrapolated parameters. In simulated dataset tests, it achieved a 10%-25% reduction in the inversion error of the extinction coefficient of the outer beams.

[0025] This method overcomes the key technical bottlenecks of limited observation swath of single-beam lidar and lack of constraints and limited accuracy of traditional active-passive fusion extrapolation methods, providing a theoretical basis and technical method support for the inversion application of next-generation and future spaceborne multi-beam lidar missions with wider swaths. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0027] Figure 1 This is a flowchart illustrating the high-precision inversion method for three-dimensional atmospheric particulate matter using a multi-beam lidar provided in this embodiment.

[0028] Figure 2 This is a schematic diagram of the inversion process using a high-spectral-resolution lidar in this embodiment.

[0029] Figure 3 The embodiment provides a probability density function for cloud and aerosol classification based on hyperspectral resolution lidar, where Lat is latitude, Alt is the layer average height, Dep is the layer average depolarization ratio, CR is the layer integral color ratio, ATB is the layer integral attenuated backscattering coefficient, PD is the layer average particle depolarization ratio, and BC is the layer average backscattering coefficient.

[0030] Figure 4 This is a schematic diagram illustrating the method for calculating the transmission distance threshold in the track crossing direction, as provided in the embodiment.

[0031] Figure 5 This is an example of high-precision re-inversion of the outer beam of a multi-beam lidar observation system provided in the implementation example. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of this invention.

[0033] To address the limitations of existing lidar systems in achieving high-precision 3D atmospheric particulate matter inversion over a regional area—specifically due to the following issues: single-beam lidar has a limited swath width, making it difficult to cover large areas and thus unable to obtain regional 3D distribution information; traditional active-passive fusion extrapolation methods lack effective constraints, resulting in limited accuracy when extrapolating high-precision inversion parameters from high-configuration beams to low-configuration beams, failing to meet the requirements for high-precision inversion; and the core parameters required for high-precision inversion cannot be directly obtained from low-configuration beams, while using only high-configuration beams would significantly increase system cost and engineering complexity—this invention utilizes a multi-beam lidar system comprising a central high-performance beam and outer basic configuration beams, combined with algorithm-level parameter extrapolation. This system rationally transfers the parameters required for high-precision inversion to the outer beams along the track crossing direction, thereby achieving high-precision 3D atmospheric particulate matter inversion over a regional area without significantly increasing engineering complexity.

[0034] like Figure 1 As shown in the embodiment, a high-precision inversion method for three-dimensional atmospheric particulate matter using a multi-beam lidar includes the following steps: S1. Obtain the vertical observation profile and auxiliary data of the multi-beam lidar containing the center beam and outer beams, and establish a multi-source dataset.

[0035] In this embodiment, the attenuation backscattering coefficient profile and signal-to-noise ratio of the Atmospheric-1 ACDL lidar (DQ-1 / ACDL) and the attenuation backscattering coefficient profile and signal-to-noise ratio of the US CALIPSO satellite cloud-aerosol lidar (CALIPSO / CALIOP) were acquired from June 1, 2022 to May 31, 2023, along with their respective auxiliary data, including the latitude and longitude of the lidar beam and the surface type.

[0036] Both Atmospheric-1 and CALIPSO are polar-orbiting satellites in afternoon orbits. Therefore, during their periodic cycles, there will be observation segments with nearly overlapping orbits, which can be used as approximate simulations of a multi-beam lidar observation system. Specifically, the ACDL lidar is a three-wavelength (532-1064-1572 nm) hyperspectral resolution lidar that also includes a polarization channel; in this embodiment, it is used as the center beam simulation data for the multi-beam lidar system. The CALIOP lidar is a dual-wavelength Mie scattering lidar that also includes a polarization channel; in this embodiment, its 532 nm single beam is used as the outer beam simulation data for the multi-beam lidar system. The actual distance between their orbits is the simulation of the cross-track distance of the multi-beam observation system.

[0037] S2. Perform atmospheric particulate matter inversion on each beam in the multi-source dataset, including hierarchical identification, optical parameter inversion, layer parameter acquisition, and cloud and aerosol classification; at the same time, calculate the transferability conditions of the core parameters of the center beam in the track-along direction, and perform the transfer hierarchy matching of the center beam in the track-crossing direction.

[0038] S2.1 Layer Identification: In this embodiment, the gradient method is used for both the center beam and outer beam simulation data. Based on the abrupt change in the intensity of the attenuated backscattering coefficient in the vertical direction, layers that may be clouds or aerosols are identified. Specifically, the vertical gradient is calculated based on the abrupt change in the intensity of the attenuated backscattering coefficient in the vertical direction. ,in Distance Attenuated backscattering coefficient at the location; vertical gradient Greater than the positive threshold The location is marked as the bottom of the layer, and the vertical gradient is less than the negative threshold. The location is marked as the top of the layer, thus identifying the layers of clouds or aerosols; vertical gradient The empirical threshold is typically between 2 and 5 times. In this embodiment... The value is 2.

[0039] Before performing this step, you can selectively smooth each beam profile with a 5 km window to ensure that the signal-to-noise ratio (SNR) of the outer low-performance channels meets the requirements for hierarchical identification. It is recommended that the SNR of the center beam be no less than 10 and the SNR of the outer beams be no less than 3.

[0040] S2.2 Optical Parameter Inversion: In this embodiment, the optical parameter inversion is based on the lidar equation, which is: , In the formula, The distance received by the lidar The echo signal power generally consists of four parts: These are system parameters for lidar that are independent of distance. These are the geometric factors of the lidar, and these two parameters are basically determined by the structural parameters of the lidar device itself. Distance Backscattering coefficient at; For the laser from the lidar transmitter to the distance The loss during the return process to the lidar. Extinction coefficient. Included In the middle, the formula is .

[0041] Since different lidar configurations provide different parameters, the inversion method is not limited here, but the inversion accuracy of the center beam must theoretically be higher than that of the outer beams. This assumption applies when the center beam is a hyperspectral resolution lidar and the outer beams are Mie scattering lidars. In this case, the backscattering coefficient and extinction coefficient need to be solved in the lidar equations. Hyperspectral resolution lidar can be solved directly without the lidar ratio (LR) assumption, while Mie scattering lidar requires inversion using the Klett or Fernald method with the lidar ratio assumption; the former has higher accuracy. In this embodiment, the hyperspectral resolution lidar inversion process is used for the center beam (…). Figure 2 Using L2A multi-channel attenuated backscattered signals as input, after signal preprocessing, layer identification is performed according to the gradient method. Then, the optical parameters are independently inverted using the hyperspectral resolution lidar equations (which can be directly solved based on the hyperspectral resolution lidar equations without assuming lidar ratios). Scene classification is performed based on the obtained depolarization and derived optical parameters, and layer geometry and cylinder integral products are further obtained. The outer beam is inverted using the Fernald method, which assumes that the lidar ratios of cloud layers and aerosol layers are fixed values ​​during the inversion process. Specifically, the lidar ratios of aerosols are between 30-60 sr depending on the surface type, water clouds are between 20-30 sr, and ice clouds are between 20-40 sr. The distinction between ice clouds and water clouds is determined by latitude and layer center height.

[0042] During this step, other optical coefficients can be inverted based on the configuration of the lidar beam. For example, the depolarization ratio is defined as the ratio of the vertically polarized backscattered signal to the parallel polarized backscattered signal received by the lidar; the particle depolarization ratio is defined as the ratio of the vertically polarized component to the horizontally polarized component in the lidar signal; and the color ratio is defined as the ratio of the backscattering coefficient in the 532 nm band to the backscattering coefficient in the 1064 nm band. These are used to better classify cloud and aerosol scenes in the next step.

[0043] S2.3 Layer Parameter Acquisition: In this embodiment, the acquired optical parameters are integrated or averaged layer by layer; usually, the extinction coefficient is summed from the height of the top layer to the bottom layer, and the result is called the optical thickness (OD) of the layer; for other optical parameters, including backscattering coefficient, depolarization ratio, lidar ratio, color ratio, etc., the average value from the height of the top layer to the bottom layer is calculated for use in subsequent steps.

[0044] S2.4 Cloud and Aerosol Classification: In this embodiment, the identified layers are classified into clouds and aerosols based on the retrieved optical parameters, and can be further divided into different phases and types of subclasses according to the configuration of the lidar. The core of this step is to use the available lidar and auxiliary information related to the layer type as feature parameters to form a classification probability density function as an evaluation index, defined as: , in, This represents the probability that the layer is an aerosol. This represents the probability that the layer is a cloud. The value of is between -1 and 1. A positive value indicates that the layer is classified as an aerosol, while a negative value indicates that the layer is classified as a cloud.

[0045] In this embodiment, the characteristic parameters provided by each beam include at least longitude, latitude, backscattering coefficient, and extinction coefficient; wherein, the characteristic parameters provided by the center beam include latitude (Lat), layer average height (Alt), layer average depolarization ratio (Dep), layer integral chromaticity ratio (CR), layer integral attenuation backscattering coefficient (ATB), layer average particle depolarization ratio (PD), and layer average backscattering coefficient (BC). Figure 3 (This is used to further classify cloud phases, including ice clouds, water clouds, and mixed clouds, as well as aerosol types, including oceanic, soot, clean continental, polluted continental, dust, and polluted dust types.)

[0046] The characteristic parameters provided by the outer beam include longitude, latitude, layer average height, layer average depolarization ratio, layer integral attenuation backscattering coefficient, and layer average backscattering coefficient, which divide the layers into two categories: clouds and aerosols.

[0047] S2.5 Calculation of the transferability conditions for the center beam core parameters along the track direction: In the embodiment, the threshold for the transfer distance of clouds and aerosols along the track direction is set as follows: and If there are multiple cloud or aerosol layers on the center beam, they can be denoted from highest to lowest according to the height of their center. , , … , , For each level, the distance of its center beam along the track direction is statistically calculated. times Hierarchical data, This is the distance from the center beam footprint to the edge beam footprint, which is theoretically the farthest distance from which the inversion parameters need to be transmitted; ideally... In practice, reaching the edge of a hierarchy can result in less actual usable data than expected. When land surface types change, it is recommended to use data within the same land surface type range, or use data at twice the normal value. Data within the range.

[0048] For each level, or The calculation method is to solve for the maximum distance that meets all of the following conditions: 1. The hierarchical type remains unchanged along the track direction; ideally, the subtype remains unchanged; 2. The changes in the height of the top and bottom of the layer along the track direction. , No more than 3 km; 3. Variation of the average lidar ratio along the track direction <15%, the above conditions ensure the uniformity of the cloud field structure and the stability of optical parameters, and have the conditions for the transferability of core parameters.

[0049] In this embodiment, observation data from the Atmospheric-1 ACDL lidar and the US CALIPSO satellite cloud-aerosol lidar, with actual observation distances ranging from ±0 to 400 km, were selected. Different beams were simulated at 10 km intervals (i.e., observation data with actual distances between 15 and 25 km were considered as simulated data for the outer beam with a cross-track distance of 20 km). The near-coincidence of the observation orbits of the two lidars during January 7, 2023, was used as an example. Figure 4 The selected layer is limited by structural variations along the track direction. It is 213 km.

[0050] S2.6 Perform center beam transmission layer matching in the track crossing direction: In this embodiment, firstly, based on the layer type, find the layer with the same type and the closest average layer height on the center beam; then, require the difference between the layer thickness in the track crossing direction and the thickness of the selected layer. No more than 3 km.

[0051] In this embodiment, taking the near-coincidence of observation tracks of two lidars on January 7, 2023 as an example, the transmittable distance of the observation profile in the track crossing direction is 213 km, and the actual distance is 37 km, which meets the transmission requirements.

[0052] S3. Based on the transmission hierarchy matching results, extrapolate the core parameters and use the extrapolated parameters to re-invert atmospheric particulate matter on the outer beam to obtain three-dimensional atmospheric particulate matter optical parameters and hierarchy information.

[0053] In this embodiment, core parameters are extrapolated based on hierarchical matching. This involves assigning the parameters required for high-precision inversion, such as the lidar ratio at the center beam level, to the specific profiles at the corresponding levels of the outer beams. The specific profiles are determined by using the transmittable distance threshold at the matching level. or Within the range, the sum of the height differences between the top and bottom floors and the selected floor level. + The smallest profile.

[0054] Using the extrapolated inversion parameters, the L1 data of the outer beam is re-inverted with high precision. That is, the L1 level observation data of the outer beam is re-inverted using the same high-precision inversion method as the center beam, and complete high-precision three-dimensional atmospheric particulate matter optical parameters and hierarchical information are obtained.

[0055] In this embodiment, taking the near-coincidence orbits observed by two lidars on January 7, 2023 as an example, the observation results and high-precision inversion results are as follows: Figure 5 As shown, after high-precision re-inversion of the outer beam, the accuracy of hierarchical identification and parameter inversion are significantly improved.

[0056] During the experiment, extrapolation of core inversion parameters from a simulated multi-beam lidar and high-precision 3D atmospheric particulate matter inversion were performed on 369 near-track segments with close proximity and good data quality. The success rate of cloud extrapolation was 99.3% at a distance of 100 km in the track crossing direction, and the success rate of aerosol extrapolation was 74.1%. Based on the re-inversion results at actual distances less than ±5 km (i.e., when the track crossing distance is 0 km), the inversion results of the simulated center beam can be considered true values. Comparing the results before and after high-precision re-inversion, the extinction coefficient of the simulated outer beam and the average error of the center beam decreased across different levels, with a reduction of 10%-25%, consistent with the expected accuracy improvement of hyperspectral resolution lidar compared to Mie scattering lidar.

[0057] Based on the same inventive concept, the embodiment also provides a computing device, including a memory and one or more processors. The memory stores executable code, and when the one or more processors execute the executable code, it is used to implement the above-mentioned high-precision inversion method for three-dimensional atmospheric particulate matter from a multi-beam lidar, specifically including the following steps: Step 1: Acquire the vertical observation profile and auxiliary data of the multi-beam lidar containing the center beam and outer beams, and establish a multi-source dataset; Step 2: Perform atmospheric particulate matter inversion on each beam in the multi-source dataset, including hierarchical identification, optical parameter inversion, layer parameter acquisition, and cloud and aerosol classification; at the same time, calculate the transferability conditions of the core parameters of the center beam in the track-along direction, and perform the transfer hierarchy matching of the center beam in the track-crossing direction. Step 3: Extrapolate the core parameters based on the transmission hierarchy matching results, and use the extrapolated parameters to re-invert atmospheric particulate matter on the outer beam to obtain three-dimensional atmospheric particulate matter optical parameters and hierarchy information.

[0058] The computing device provided in this embodiment, at the hardware level, includes not only a processor and memory, but also internal buses, network interfaces, memory, and other hardware required for business operations. The memory is non-volatile memory. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the three-dimensional high-precision inversion method for atmospheric particulate matter from multi-beam lidar described in steps S1-S3 above. Of course, besides software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0059] The specific embodiments described above illustrate the technical solution and beneficial effects of the present invention in detail. It should be understood that the above description is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A high-precision inversion method for three-dimensional atmospheric particulate matter from a multi-beam lidar, characterized in that, Includes the following steps: Acquire vertical observation profiles and auxiliary data from a multi-beam lidar system, including the center beam and outer beams, and establish a multi-source dataset. Atmospheric particulate matter inversion is performed on each beam in a multi-source dataset, including hierarchical identification, optical parameter inversion, layer parameter acquisition, and cloud and aerosol classification. Layer parameter acquisition is achieved by hierarchical integration or averaging of optical parameters. Specifically, the optical thickness of the layer is obtained by summing the extinction coefficient between the layer top and bottom heights. Layer parameters, including backscattering coefficient, depolarization ratio, lidar ratio, and color ratio, are obtained by calculating the average value between the layer top and bottom heights. Simultaneously, the transferability conditions of the center beam core parameters are calculated along the track direction, including setting the track-crossing distance thresholds for clouds and aerosols as follows: and For each cloud or aerosol layer on the central beam, the distance along the track direction is statistically calculated. times Hierarchical data within the range, of which The distance from the center beam footprint to the edge beam footprint. For each level, the maximum distance that meets all of the following conditions is used as the transmission distance threshold: the level type remains unchanged along the track direction; the change in the height of the top and bottom of the level along the track direction does not exceed 3 km; the change in the average lidar ratio of the level along the track direction is less than 15%; and the transmission level matching of the center beam is performed in the track crossing direction. Based on the transmission hierarchy matching results, the core parameters are extrapolated, and the extrapolated parameters are used to re-invert atmospheric particulate matter on the outer beam to obtain three-dimensional atmospheric particulate matter optical parameters and hierarchy information.

2. The method for high-precision inversion of three-dimensional atmospheric particulate matter using multi-beam lidar according to claim 1, characterized in that, The central beam is a high-spectral-resolution lidar with multi-wavelength and polarization detection capabilities; the outer beam is a single-wavelength Mie scattering lidar.

3. The high-precision inversion method for three-dimensional atmospheric particulate matter using multi-beam lidar according to claim 2, characterized in that, The vertical observation profile of the multi-beam lidar is the L1 level observation data of the lidar, including the attenuation backscattering coefficient and signal-to-noise ratio of different wavelengths and channels; The auxiliary data includes the latitude and longitude of the lidar beam and the land surface type.

4. The method for high-precision inversion of three-dimensional atmospheric particulate matter using multi-beam lidar according to claim 1, characterized in that, The hierarchical recognition is performed using the gradient method, including: Based on each beam in the multi-source dataset, the vertical gradient is calculated according to the abrupt change in the intensity of the attenuated backscattering coefficient in the vertical direction. ,in Distance Attenuated backscattering coefficient at the location; The vertical gradient is greater than the positive threshold. The location is marked as the bottom of the layer, and the vertical gradient is less than the negative threshold. The location is marked as the top of the layer, thus identifying the layers of clouds or aerosols.

5. The method for high-precision inversion of three-dimensional atmospheric particulate matter using multi-beam lidar according to claim 1, characterized in that, The optical parameter inversion is based on the lidar equation. In the formula, The distance received by the lidar The echo signal power consists of four parts: These are system parameters for lidar that are independent of distance. For lidar geometric factors, Distance The backscattering coefficient at that location, For the laser from the lidar transmitter to the distance The loss during the process of returning to the lidar; The backscattering coefficient and extinction coefficient are directly solved for the center beam using a high-spectral-resolution lidar inversion process; for the outer beams, the Klett method or Fernald method is used to invert the backscattering coefficient and extinction coefficient by assuming a lidar ratio.

6. The method for high-precision inversion of three-dimensional atmospheric particulate matter using multi-beam lidar according to claim 1, characterized in that, The aforementioned cloud and aerosol classification includes: Based on the retrieved optical parameters, the identified layers are divided into clouds and aerosols. Further, combined with the LiDAR configuration, they are further subdivided into different phases and types of subclasses. The classification index is represented by a probability density function established using feature parameters from the LiDAR and auxiliary information; the calculation formula is as follows: ， in, This represents the probability that the layer is an aerosol. This represents the probability that the layer is a cloud. The value of is between -1 and 1. A positive value indicates that the layer is classified as an aerosol, while a negative value indicates that the layer is classified as a cloud.

7. The method for high-precision inversion of three-dimensional atmospheric particulate matter using multi-beam lidar according to claim 1, characterized in that, The aforementioned center beam transmission layer matching in the track crossing direction includes: finding layers of the same type and with the closest average layer height on the center beam according to the layer type; and requiring that the difference between the layer thickness in the track crossing direction and the thickness of the selected layer does not exceed 3 km. The extrapolation of core parameters based on the transmission layer matching results includes: assigning the lidar ratio parameter of the center beam layer to the specific profile in the corresponding layer of the outer beam; the specific profile is the profile with the smallest sum of the height difference between the top and bottom layers and the height of the selected layer within the transmission distance threshold range of the matching layer.

8. A computing device comprising a memory and one or more processors, wherein the memory stores executable code, characterized in that, When the one or more processors execute the executable code, they are used to implement the three-dimensional high-precision inversion method for atmospheric particulate matter from a multi-beam lidar as described in any one of claims 1-7.

Images