A method and system for dynamic injection of spherical aerosol polarization scattering data

Abstract

This invention discloses a dynamic injection method and system for spherical aerosol polarization scattering data, relating to the field of atmospheric remote sensing technology. The method includes: receiving a continuously changing set of microphysical parameters of the aerosol; substituting these parameters into a particle size distribution model and calculating the original scattering matrix in the Stokes observation space in real time based on Mie scattering theory; mapping the original scattering matrix to a computational space with diagonalized features through a basis transformation protocol to extract independent features of the polarization components; structurally encapsulating the basis-transformed scattering data according to the target engine's memory layout; and injecting the encapsulated data into the target engine's computational kernel during the forward simulation process using dynamic memory overlay technology. This invention aims to solve the problem of accurately eliminating stratospheric aerosol polarization scattering interference by extracting independent polarization features through a physically symmetric explicit basis diagonalization transformation, thereby improving simulation efficiency and eliminating systematic errors introduced by model distortion.

Description

Technical Field

[0001] This invention relates to the field of atmospheric remote sensing technology, and in particular to a method and system for dynamically injecting spherical aerosol polarization scattering data. Background Technology

[0002] In ozone inversion operations conducted by spaceborne payloads, stratospheric aerosols are the core interference component causing inversion errors. Existing radiative transfer simulation engines typically use a pre-set static lookup table to handle stratospheric aerosol polarization scattering. This approach pre-calculates and stores extinction coefficients, scattering matrices, and other data corresponding to different aerosol parameters, and retrieves the scattering data for the corresponding parameters through interpolation during simulation and inversion.

[0003] However, the fixed parameter space of this static model cannot cover the scenario of continuous changes in microphysical parameters of stratospheric aerosols between quiescent and volcanic periods. Interpolation operations under non-standard events introduce uncontrollable systematic errors, directly reducing the accuracy of ozone inversion. Simultaneously, the static lookup table masks the conversion process from microphysical parameters to the scattering substrate, failing to achieve real-time linkage between microphysical parameters and phase functions, resulting in a lack of data support for algorithm optimization. Furthermore, the general-purpose fully polarized radiative transfer simulation engine directly uses a 4x4 Stokes matrix for calculation, without considering the physical symmetry of stratospheric spherical droplets for computational path optimization, leading to computational redundancy and failing to fully utilize the parallel computing architecture of the graphics processor, resulting in simulation efficiency falling below the hardware's limits.

[0004] Therefore, how to achieve accurate and efficient dynamic injection of stratospheric aerosol polarization scattering data has become an urgent technical challenge. Summary of the Invention

[0005] The main objective of this invention is to provide a method and system for the dynamic injection of spherical aerosol polarization scattering data, aiming to achieve accurate and efficient dynamic injection of stratospheric aerosol polarization scattering data.

[0006] To achieve the above objectives, this invention proposes a dynamic injection method for spherical aerosol polarization scattering data, comprising the following steps:

[0007] (1) Receive the continuously changing microphysical parameter set of aerosols;

[0008] (2) Substitute the set of microphysical parameters into the particle size distribution model, and calculate the original scattering matrix in the Stokes observation space in real time based on the Mie scattering theory. ;

[0009] (3) The original scattering matrix is ​​transformed using a basis transformation protocol. The Stokes observation space is mapped to a computational space with diagonalized features to extract independent features of the polarization components; wherein, the cross-coupling terms in the computational space satisfy the decoupling constraints due to the physical symmetry of aerosol particles;

[0010] (4) The scattering data after the basis transformation is structured and encapsulated according to the memory layout of the target engine;

[0011] (5) The memory dynamic overlay technology is used to inject the encapsulated data into the computing kernel of the target engine during the forward simulation process.

[0012] Preferably, the microphysical parameters in the microphysical parameter set include the effective radius of the first microphysical parameter. Geometric standard deviation of the second microphysical parameter and the third microphysical parameter, complex refractive index .

[0013] Preferably, the effective radius of the first microphysical parameter Calculated using the following formula:

[0014]

[0015] in, Let be the geometric mean radius of the aerosol.

[0016] Preferably, step (2) specifically includes:

[0017] Substituting the aforementioned microphysical parameter set into the aerosol unimodal log-normal distribution model, in the particle size space... The extinction coefficient of the aerosol is calculated by performing a weighted integral. Scattering coefficient and the original scattering matrix The distribution formula for the aerosol unimodal log-normal distribution model is as follows:

[0018]

[0019] in, Vertical height The particle number concentration varies with altitude; This indicates the cumulative concentration.

[0020] Preferably, the aerosol is a stratospheric spherical aerosol.

[0021] Preferably, the basis transformation protocol in step (3) is implemented using a linear transformation operator. and its inverse matrix Perform space mapping; the transformation formula is as follows:

[0022]

[0023] in, The mapping matrix in Chandrasekhar space; the linear transformation operator for:

[0024] .

[0025] Preferably, the decoupling constraint is expressed as an effect on the mapped matrix. Explicit mapping and zeroing are performed, and the polarization components are processed according to a standard 6-channel vector. The sequence is arranged; among which, the rules for processing spherical particles include:

[0026] The first channel component after mapping ;

[0027] The mapped fifth channel component ;

[0028] Mapped second channel cross-coupling term components Perform explicit forced zeroing, i.e. ;

[0029] The mapped sixth channel component satisfy To maintain physical consistency;

[0030] in, The original scattering matrix The element in the first row and first column, The original scattering matrix The element in the first row and second column.

[0031] Preferably, the method further includes the step of dynamically constructing a lookup table:

[0032] Numerical integration of the polarization components of each channel after mapping is performed over the entire angle to construct a normalized cumulative scattering probability distribution (CDF) to support random sampling of photons in Monte Carlo simulations.

[0033] Preferably, during the process of constructing the normalized cumulative scattering probability distribution (CDF), a resampling lookup table is established simultaneously; the resampling lookup table includes a first resampling lookup table and a second resampling lookup table, wherein the first resampling lookup table is established based on equal probability intervals, and the second resampling lookup table is established based on equal angular intervals.

[0034] Preferably, the memory layout of the target engine in step (4) is as follows: and The multidimensional tensor structure; in which Representing 6 reserved polarization memory slots, the fixed memory slots ensure that the addressing logic of the radiation transmission engine remains constant when reading scattering data of aerosol particles of different shapes.

[0035] Preferably, the memory dynamic overlay technique in step (5) specifically refers to:

[0036] By using video memory address mapping technology, without interrupting the target engine's computing state or performing disk I / O, the encapsulated data can replace the original static scattering data in the target engine's memory space in real time through video memory handle redirection.

[0037] This application also discloses a dynamic injection system for spherical aerosol polarization scattering data, including:

[0038] The parameter receiving module is used to receive the continuously changing microphysical parameter set of aerosols;

[0039] The matrix calculation module is used to substitute the microphysical parameter set into the particle size distribution model and calculate the original scattering matrix in the Stokes observation space in real time based on Mie scattering theory. ;

[0040] A basis transformation module is used to transform the original scattering matrix using a basis transformation protocol. The Stokes observation space is mapped to a computational space with diagonalized features to extract independent features of the polarization components; wherein, the cross-coupling terms in the computational space satisfy the decoupling constraints due to the physical symmetry of aerosol particles;

[0041] The structure encapsulation module is used to encapsulate the scattering data after basis transformation according to the memory layout of the target engine.

[0042] The dynamic injection module is used to inject encapsulated data into the target engine's computing kernel during the forward simulation process using memory dynamic overwrite technology.

[0043] The above technical solution has the following advantages:

[0044] This invention receives continuously changing microphysical parameters and substitutes them into a particle size distribution model to calculate the original scattering matrix in real time, ensuring the physical continuity of scattering characteristics as physical parameters change, thus eliminating interpolation errors in the parameter evolution process of traditional static lookup table schemes. By using a basis transformation protocol to map the scattering matrix from the Stokes observation space to a computational space with diagonalized characteristics, and utilizing the physical symmetry of aerosol particles, explicit decoupling of polarization components is achieved, reducing complex matrix operations to efficient parallel scalar operations. Combined with dynamic memory overlay technology, real-time injection of encapsulated data is achieved without interrupting computation. During the forward simulation of the target engine, hot data replacement is completed through memory handle redirection, enabling real-time response to the parameter adjustment needs of the inversion algorithm. This significantly improves the throughput of the full polarization simulation and the overall convergence accuracy of atmospheric composition inversion. Attached Figure Description

[0045] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings, wherein:

[0046] Figure 1 This is a schematic diagram illustrating the principle of a dynamic injection method for spherical aerosol polarization scattering data provided in an embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram of the input-output structure of a radiative transfer model containing a dynamic aerosol polarization scattering phase matrix, provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0049] Example 1

[0050] like Figures 1 to 2 As shown, this embodiment provides a dynamic injection method for spherical aerosol polarization scattering data. This method, through a dynamic polarization scattering data generation and injection system driven by microphysical parameters, solves the problem of accurately eliminating aerosol polarization scattering interference in stratospheric ozone retrieval. Stratospheric aerosols are a key interfering factor in spaceborne ozone retrieval, and their physical state dynamically evolves with the atmospheric environment. Stratospheric ozone retrieval depends on the scattering spectrum and polarization characteristics in the ultraviolet to visible light bands; aerosol polarization scattering directly alters the Stokes vector component incident on the satellite sensor. This embodiment reduces the interference of aerosols on ozone retrieval at its source through a dynamic injection mechanism. The specific implementation process is as follows:

[0051] First, step one is executed, where parameter receiving module 32 receives a continuously changing set of microphysical parameters of the aerosol. In this embodiment, the received microphysical parameters include the effective radius of the first microphysical parameter. Geometric standard deviation of the second microphysical parameter and the third microphysical parameter, complex refractive index These parameters have no fixed range restrictions and can fully cover the parameter evolution range of stratospheric aerosols from the quiescent period to the volcanic period, such as the effective radius. for Geometric standard deviation for Among them, the effective radius of the first microphysical parameter. Calculated using the following formula:

[0052]

[0053] In the formula, Let be the geometric mean radius of the aerosol.

[0054] Next, step two is executed, where matrix calculation module 33 substitutes the microphysical parameter set into the particle size distribution model. The stratospheric aerosol particle size distribution follows a unimodal log-normal distribution, and the distribution formula is as follows:

[0055]

[0056] In the formula, Vertical height The particle number concentration varies with altitude. This represents the cumulative number concentration. Matrix calculation module 33 is in the particle size space. The extinction coefficient of the aerosol is calculated in real time based on the Mie scattering theory by performing a weighted integral. Scattering coefficient and the original scattering matrix in the Stokes observation space Due to the physical symmetry of stratospheric spherical droplets, their original scattering matrix... satisfy Furthermore, terms involving the cross-coupling of linear and circular polarization are theoretically always equal to 0.

[0057] Then, step three is executed, where the basis transformation module 34 transforms the original scattering matrix using the basis transformation protocol. Mapping from the Stokes observation space to a computational space with diagonalized features. The standard 6-channel vector sequence defined in this embodiment is... This vector structure, as a full representation container, can simultaneously support the physical representation of spherical and non-spherical particles by reserving fixed memory slots, ensuring that the address logic of the radiation transfer engine remains constant for particles of different shapes.

[0058] Basis transformation protocol uses linear transformation operators and its inverse matrix Execution mapping:

[0059] .

[0060] The conversion formula is For spherical particles, this protocol establishes an explicit set of component mapping and "logical zero-filling" rules to extract parallel components. With vertical component Independent characteristics:

[0061] First channel (intensity gain slot): ;

[0062] Second channel (cross-coupling slot): For spherical particles, due to physical symmetry, the following conditions are met... and Theoretically, this value is zero. This scheme explicitly writes it to 0 using an algorithm, i.e. ;

[0063] Third and fourth channels (phase and interference slots): ;

[0064] Fifth Channel: ;

[0065] Sixth Channel: Due to ,make In order to maintain physical consistency.

[0066] This scheme achieves complete decoupling of the polarization energy path through explicit mapping. In the spherical particle example, by... Forced zero-filling of slots degenerates complex matrix operations into efficient parallel scalar operations, significantly reducing the computational overhead of forward simulation during high-frequency calls.

[0067] Next, in step four, the structure encapsulation module 35 structures and encapsulates the processed scattering data according to the memory layout of the target engine (such as SMART-G). The memory layout required by the target engine is... and The multidimensional tensor structure, in which This represents 6 reserved polarized memory slots. Representative at Within the range of degrees, the interval is The sampling points are set in degrees. This step also includes dynamically constructing a lookup table: performing full-angle numerical integration on each polarization channel component after mapping, constructing a normalized cumulative scattering probability distribution (CDF), and simultaneously establishing a first resampling lookup table based on "equal probability spacing" (to optimize random sampling efficiency) and a second resampling lookup table based on "equal angular spacing" (to optimize gradient interpolation calculation).

[0068] Finally, in step five, the dynamic injection module 36 employs dynamic memory overwriting technology to inject the encapsulated data into the target engine's computational kernel during the forward simulation process. Specifically, through memory address mapping technology, data is injected in real-time via memory handle redirection without interrupting the computation or performing disk I / O. When the target engine's computational kernel reads... When data is accessed from a slot, hardware-level branch optimization (computation pruning) is automatically triggered, skipping redundant multiplication and accumulation operations involving that item. This hot-swappable memory injection method eliminates the systematic errors caused by static lookup tables, significantly improving the overall convergence speed and accuracy of atmospheric composition inversion.

[0069] Example 2

[0070] This embodiment, based on the aforementioned Embodiment 1, further discloses the application details of this method in the evolution of atmospheric non-standard events. Specifically, the physical state of stratospheric aerosols is not static; there are significant differences in microphysical parameters between quiescent periods and volcanic activity periods.

[0071] When the stratosphere is in a quiescent period, the effective radius of the first microphysical parameter received by the parameter receiving module 32 It is usually at a low level, specifically in to Between. At this time, the scattering data packaged by the structural encapsulation module 35 reflects weak polarization scattering characteristics. However, after special atmospheric events such as volcanic eruptions, a large amount of sulfur dioxide gas is converted into sulfate aerosols, causing drastic and continuous changes in the microphysical parameters of stratospheric aerosols.

[0072] At this time, the parameter receiving module 32 can capture the effective radius in real time. from Rapidly evolved to The process, and the geometric standard deviation Fluctuations will also occur synchronously. The matrix calculation module 33, based on these continuously changing parameters, calculates the value through the particle size space. The weighted integral is applied to update the original scattering matrix in real time. Compared to the pre-defined static lookup table scheme used in traditional techniques, this embodiment does not require interpolation between discrete parameter points. Traditional schemes, when dealing with this leapfrog evolution from a quiescent to a volcanic period, often suffer from uncontrollable interpolation errors due to the fixed parameter space of the static lookup table, resulting in large ozone inversion residuals. This embodiment, however, ensures the physical continuity of scattering characteristics as physical parameters change through dynamic driving logic, eliminating systematic errors caused by interpolation at their source.

[0073] Example 3

[0074] This embodiment focuses on describing the detailed implementation of dynamically constructed lookup tables and resampling lookup tables to support the efficient operation of photons in Monte Carlo simulations.

[0075] After obtaining the 6-channel polarization components after basis transformation, the structure encapsulation module 35 needs to convert them into a random sampling structure adapted to the computational kernel. First, the structure encapsulation module 35 performs full-angle numerical integration on each mapped channel polarization component. Specifically, for each polarization calculation channel, in... to The normalized cumulative scattering probability distribution (CDF) is calculated within the scattering angle range. This CDF defines the probability distribution of a photon in the scattering angle direction after scattering.

[0076] To balance computational efficiency and accuracy, this embodiment simultaneously establishes two sets of resampling lookup tables. The first resampling lookup table is built based on equal probability intervals. Since in Monte Carlo simulations, random sampling of photon paths typically relies on... arrive Random numbers uniformly distributed between points can be directly mapped to scattering angles through resampling at equal probability intervals, greatly improving the sampling efficiency of photon tracking. The second resampling lookup table is built based on equal angular intervals. This table uses a fixed angular step size, for example... It records the scattering probability value and is mainly used for gradient interpolation calculation in the inversion iteration process.

[0077] Through this dual-index resampling technique, the structure encapsulation module 35 transforms the originally complex physical phase function into a data structure that can be quickly retrieved by the GPU computing core. For spherical particles, the structure encapsulation module 35 further utilizes symmetry to simplify redundant polarization path data, ensuring that the amount of data injected into memory is minimized, thus reducing the pressure on video memory bandwidth.

[0078] Example 4

[0079] This embodiment discloses in detail the implementation logic of dynamic memory overlay technology and hardware-level computational pruning.

[0080] The memory dynamic overwrite technique executed by the dynamic injection module 36 aims to achieve seamless hot replacement of scattering data. During the forward simulation of the target engine, the dynamic injection module 36 obtains the memory handle of the original static scattering data in the target engine's memory space through memory address mapping technology. Subsequently, the dynamic injection module 36 uses a memory handle redirection mechanism to directly write the encapsulated structured tensor data into the corresponding memory address.

[0081] This process does not interrupt the target engine's computational state, nor does it require cumbersome millisecond-level disk I / O operations, achieving a millisecond-level closed-loop response. When the target engine's computational kernel performs a full polarization radiative transfer simulation, the data it reads is already the dynamic phase function generated in real-time by this embodiment. After the radiative transfer engine completes the end-to-end calculation, it outputs the Stokes vector required by the onboard sensor. , , , Simulation results for radiance, polarization radiance, etc.

[0082] More importantly, since the basis conversion module 34 has already processed the cross-coupling term components during the preprocessing stage... Explicitly forcing zeroing triggers hardware-level branch optimization logic when the computation kernel reads these zero-value entries. In parallel computing architectures like CUDA, the computation unit can automatically skip over branches involving zero values. Redundant multiplication operations. In traditional methods... In the Stokes matrix approach, even if some physically terms are zero, the computational core still needs to perform full matrix multiplication. This embodiment, however, explicitly transforms physical symmetry features into diagonalized features of the computational basis, significantly improving the GPU's simulation throughput and achieving the hardware performance limit for full polarization simulation.

[0083] Example 5

[0084] This embodiment provides a dynamic injection system for spherical aerosol polarization scattering data, used to implement the methods described in the above embodiments.

[0085] The system includes a parameter receiving module 32, a matrix calculation module 33, a basis transformation module 34, a structure encapsulation module 35, and a dynamic injection module 36. The modules communicate with each other via a high-speed data bus, and their hardware can be a computing server containing a high-performance GPU.

[0086] The parameter receiving module 32 is responsible for acquiring atmospheric environment monitoring data in real time, such as observation inversion results from lidar. The matrix calculation module 33 is equipped with a Mie scattering calculation kernel, which can quickly complete the particle size distribution integration. The basis transformation module 34 performs matrix linear transformation, mapping the scattering matrix in physical space to a diagonal matrix in computational space. The structure encapsulation module 35 is responsible for tensor alignment of data and construction of the CDF resampling table. The dynamic injection module 36 interacts with the target engine's memory space through a driver interface.

[0087] Through the collaborative operation of this system, this embodiment not only supports efficient simulation of stratospheric spherical droplets but also possesses good physical universality. For example, for non-spherical aerosol particles, the substrate conversion module 34 can retain the mapped... By using component analysis instead of zeroing out, the system ensures its integrity when handling complex mixed aerosol scenarios, such as the co-inversion of ozone and aerosol particle size. As a forward simulation support for spaceborne ozone inversion, this system significantly improves the accuracy of ozone profile and column concentration inversions by achieving real-time linkage between aerosol scattering simulation and inversion algorithms.

[0088] In summary, the dynamic injection method and system for spherical aerosol polarization scattering data provided in this application achieves continuous and accurate simulation of the evolution of stratospheric aerosols from the quiescent to the volcanic phase through dynamic driving logic driven by microphysical parameters. This scheme utilizes the physical symmetry of spherical stratospheric aerosols for explicit basis diagonalization transformation, not only extracting independent features of the polarization components but also significantly improving the simulation efficiency of GPU hardware through computational path pruning. Simultaneously, the hot-swappable memory injection mechanism ensures that the forward simulation can respond in real-time to the inversion algorithm's call requirements, fundamentally eliminating the systematic errors caused by static lookup table models, and providing reliable technical support for high-precision spaceborne ozone inversion.

[0089] It should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. Where there is no conflict, the embodiments and features in the embodiments of this application can be combined with each other.

Claims

1. A method of dynamic injection of spherical aerosol polarized scattering data, characterized in that, Includes the following steps: (1) Receive the continuously changing microphysical parameter set of aerosols; (2) Substitute the set of microphysical parameters into the particle size distribution model, and calculate the original scattering matrix in the Stokes observation space in real time based on the Mie scattering theory. ; (3) The original scattering matrix is ​​transformed using a basis transformation protocol. The Stokes observation space is mapped to a computational space with diagonalized features to extract independent features of the polarization components; wherein, the cross-coupling terms in the computational space satisfy the decoupling constraints due to the physical symmetry of aerosol particles; (4) The scattering data after the basis transformation is structured and encapsulated according to the memory layout of the target engine; (5) Employing dynamic memory overlay technology, the encapsulated data is injected into the target engine's computational kernel during the forward simulation process; The memory dynamic overlay technique in step (5) specifically refers to: By using video memory address mapping technology, without interrupting the target engine's computing state or performing disk I / O, the encapsulated data can replace the original static scattering data in the target engine's memory space in real time through video memory handle redirection.

2. The method for dynamically injecting spherical aerosol polarization scattering data according to claim 1, characterized in that, The microphysical parameters in the microphysical parameter set include the effective radius of the first microphysical parameter. Geometric standard deviation of the second microphysical parameter and the third microphysical parameter, complex refractive index .

3. The method of dynamic injection of spheroid aerosol polarization scattering data according to claim 2, wherein, The first microphysical parameter effective radius The effective radius is calculated according to the following formula: wherein, is the geometric mean radius of the aerosol.

4. The method for dynamically injecting spherical aerosol polarization scattering data according to claim 3, characterized in that, Step (2) specifically involves: Substituting the aforementioned microphysical parameter set into the aerosol unimodal log-normal distribution model, in the particle size space... The extinction coefficient of the aerosol is calculated by performing a weighted integral. Scattering coefficient and the original scattering matrix The distribution formula for the aerosol unimodal log-normal distribution model is as follows: wherein is the vertical height, is the particle number concentration as a function of height; denotes the cumulative number concentration.

5. The method of dynamic injection of spherical aerosol polarization scattering data according to any one of claims 1 to 4, characterized in that, The aerosol is a stratospheric spherical aerosol.

6. The method for dynamically injecting spherical aerosol polarization scattering data according to claim 1, characterized in that, The basis transformation protocol in step (3) uses a linear transformation operator. and its inverse matrix Perform space mapping; the transformation formula is as follows: wherein is the mapping matrix under Chandrasekhar space; the linear transformation operator is: 。 7. The method for dynamically injecting spherical aerosol polarization scattering data according to claim 6, characterized in that, The decoupling constraint is represented as a constraint on the mapped matrix An explicit mapping and zeroing is performed, the polarization components are arranged in accordance with a sequence of standard 6-channel vectors The processing rule for the spherical particles comprises: mapped first channel component ; mapped fifth channel component ; Mapped second channel cross-coupling term components Perform explicit forced zeroing, i.e. ; mapped sixth channel component satisfies to maintain physical consistency; wherein is the first row, first column element of the original scattering matrix , is the first row, second column element of the original scattering matrix .

8. The method of dynamic injection of spherical aerosol polarization scattering data of claim 1, wherein, The method also includes the step of dynamically constructing a lookup table: Numerical integration of the polarization components of each channel after mapping is performed over the entire angle to construct a normalized cumulative scattering probability distribution (CDF) to support random sampling of photons in Monte Carlo simulations.

9. The method of dynamic injection of spherical aerosol polarization scattering data according to claim 8, wherein, During the process of constructing the normalized cumulative scattering probability distribution (CDF), a resampling lookup table is simultaneously established. The resampling lookup table includes a first resampling lookup table and a second resampling lookup table, wherein the first resampling lookup table is established based on equal probability intervals, and the second resampling lookup table is established based on equal angular intervals.

10. The method of dynamic injection of spherical aerosol polarization scattering data of claim 1, wherein, The memory layout of the target engine in step (4) is as follows: and The multidimensional tensor structure; in which Representing 6 reserved polarization memory slots, the fixed memory slots ensure that the addressing logic of the radiation transmission engine remains constant when reading scattering data of aerosol particles of different shapes.

11. A dynamic injection system for spherical aerosol polarization scattering data, characterized in that, include: The parameter receiving module is used to receive the continuously changing microphysical parameter set of aerosols; A matrix calculation module is configured to substitute the microphysical parameter group into a particle size distribution model and calculate a raw scattering matrix in a Stokes observation space in real time based on Mie scattering theory ; A basis transformation module is used to transform the original scattering matrix using a basis transformation protocol. The Stokes observation space is mapped to a computational space with diagonalized features to extract independent features of the polarization components; wherein, the cross-coupling terms in the computational space satisfy the decoupling constraints due to the physical symmetry of aerosol particles; The structure encapsulation module is used to encapsulate the scattering data after basis transformation according to the memory layout of the target engine. The dynamic injection module is used to inject the encapsulated data into the computing kernel of the target engine during the forward simulation process using memory dynamic overlay technology. Specifically, the memory dynamic overlay technology is as follows: through video memory address mapping technology, without interrupting the computing state of the target engine and without performing disk I / O, the encapsulated data replaces the original static scattering data in the memory space of the target engine in real time by redirecting the video memory handle.

Images