Search

CN-121983183-A - Dynamic injection method and system for spherical aerosol polarized scattering data

CN121983183ACN 121983183 ACN121983183 ACN 121983183ACN-121983183-A

Abstract

The invention discloses a dynamic injection method and a system for spherical aerosol polarized scattering data, and relates to the technical field of atmospheric remote sensing. The method comprises the steps of receiving a continuously-changing microphysical parameter set of aerosol, substituting the continuously-changing microphysical parameter set into a particle size distribution model, calculating an original scattering matrix under a Stokes observation space in real time based on Mie scattering theory, mapping the original scattering matrix to a calculation space with diagonal characteristics through a substrate conversion protocol to extract independent characteristics of polarization components, carrying out structured encapsulation on scattering data subjected to substrate conversion treatment according to the memory layout of a target engine, and injecting the encapsulated data into a calculation kernel of the target engine in a forward simulation process by adopting a memory dynamic coverage technology. The invention aims to solve the problem that the stratosphere aerosol polarization scattering interference is difficult to accurately eliminate, and extracts polarization independent characteristics through the diagonalization conversion of a physical symmetry explicit substrate, thereby improving the simulation efficiency and eliminating the systematic errors introduced by model distortion.

Inventors

  • ZHAO XIN
  • HOU BIN
  • ZHOU HAIJIN

Assignees

  • 中国科学院合肥物质科学研究院

Dates

Publication Date
20260505
Application Date
20260408

Claims (12)

  1. 1. The dynamic injection method of spherical aerosol polarized scattering data is characterized by comprising the following steps: (1) Receiving a continuously varying set of microphysical parameters of the aerosol; (2) Substituting the micro-physical parameter set into a particle size distribution model, and calculating an original scattering matrix in Stokes observation space in real time based on Mie scattering theory ; (3) Converting the original scattering matrix through a substrate conversion protocol Mapping from the Stokes observation space to a computation space with diagonalized features to extract independent features of polarization components, wherein cross-coupling terms in the computation space satisfy decoupling constraints due to physical symmetry of aerosol particles; (4) Carrying out structured packaging on the scattered data subjected to substrate conversion treatment according to the memory layout of the target engine; (5) And injecting the packaged data into a computing kernel of the target engine in a forward simulation process of the target engine by adopting a memory dynamic coverage technology.
  2. 2. The method of dynamic injection of spherical aerosol polarized scattering data according to claim 1, wherein, the micro-physical parameters in the micro-physical parameter group comprise a first micro-physical parameter effective radius Geometric standard deviation of second micro-physical parameter Complex refractive index of third micro physical parameter 。
  3. 3. The method of dynamic injection of spherical aerosol polarized scattering data according to claim 2, wherein the first micro-physical parameter effective radius Calculated according to the following formula: Wherein, the Is the geometric mean radius of the aerosol.
  4. 4. A method of dynamic injection of spherical aerosol polarized scattering data according to claim 3, wherein step (2) is specifically: substituting the micro-physical parameter set into an aerosol unimodal lognormal distribution model in a particle size space Weighted integration is carried out on the obtained product, and the extinction coefficient of the aerosol is obtained by calculation Scattering coefficient The original scattering matrix The distribution formula of the aerosol unimodal lognormal distribution model is as follows: Wherein, the Is a vertical height of the steel plate, and is a vertical height, Is the concentration of particles as a function of height; Indicating the cumulative number concentration.
  5. 5. The method of dynamic injection of spherical aerosol polarized scattering data according to any one of claims 1 to 4, wherein the aerosol is a stratospheric aerosol.
  6. 6. The method of dynamic injection of spherical aerosol polarized scattering data according to claim 1, wherein the substrate conversion protocol in step (3) is implemented by a linear transformation operator And its inverse matrix The spatial mapping is performed and the conversion formula is as follows: Wherein, the Is a mapping matrix under Chandrasekhar space, and the linear transformation operator The method comprises the following steps: 。
  7. 7. The method of claim 6, wherein the decoupling constraints are expressed as a pair of mapped matrices Explicit mapping and zeroing processing is performed, and the polarization components are according to standard 6-channel vectors Wherein the processing rules for spherical particles include: mapped first channel component ; Mapped fifth channel component ; Mapped second channel cross-coupled term components Performing explicit zero forcing processing, i.e. ; Mapped sixth channel component Satisfy the following requirements To maintain physical consistency; Wherein, the For the original scattering matrix The 1 st row 1 st column element of (c), For the original scattering matrix Row 1, column 2 elements of (c).
  8. 8. The method of dynamic injection of spherical aerosol polarized scatter data of claim 1, further comprising the step of dynamically building a look-up table: and carrying out full-angle numerical integration on the mapped polarization components of each channel, and constructing a normalized cumulative scattering probability distribution CDF so as to support random sampling of photons in Monte Carlo simulation.
  9. 9. The method of claim 8, wherein the step of constructing the normalized cumulative scattering probability distribution CDF includes synchronously creating a resampling lookup table comprising a first resampling lookup table and a second resampling lookup table, wherein the first resampling lookup table is created based on the equal probability spacing and the second resampling lookup table is created based on the equal angle spacing.
  10. 10. The method of claim 1, wherein the memory layout of the target engine in the step (4) is And is also provided with Multi-dimensional tensor structure of (1), wherein Representing 6 reserved polarized memory slots, the reserved fixed memory slots ensure that the addressing logic of the radiation transmission engine is kept constant when the radiation transmission engine reads aerosol particle scattering data with different shapes.
  11. 11. The method for dynamically injecting spherical aerosol polarized scattering data according to claim 1, wherein the memory dynamic coverage technique in step (5) is specifically: Under the condition that the calculation state of the target engine is not interrupted and the disk I/O is not executed, the packaged data is replaced with the original static scattering data in the memory space of the target engine in real time through the redirection of the video memory handle by the video memory address mapping technology.
  12. 12. A dynamic injection system for spherical aerosol polarized scatter data, comprising: The parameter receiving module is used for receiving the continuously-changing microphysical parameter set of the aerosol; the matrix calculation module is used for substituting the micro-physical parameter set into a particle size distribution model and calculating an original scattering matrix in Stokes observation space in real time based on Mie scattering theory ; A substrate conversion module for converting the original scattering matrix by a substrate conversion protocol Mapping from the Stokes observation space to a computation space with diagonalized features to extract independent features of polarization components, wherein cross-coupling terms in the computation space satisfy decoupling constraints due to physical symmetry of aerosol particles; the structure packaging module is used for carrying out structural packaging on the scattered data subjected to substrate conversion treatment according to the memory layout of the target engine; And the dynamic injection module is used for injecting the packaged data into a calculation kernel of the target engine in the forward simulation process of the target engine by adopting a memory dynamic coverage technology.

Description

Dynamic injection method and system for spherical aerosol polarized scattering data Technical Field The invention relates to the technical field of atmosphere remote sensing, in particular to a dynamic injection method and a system for spherical aerosol polarized scattering data. Background In the process of carrying out ozone inversion work on satellite-borne loads, stratosphere aerosols are core interference components causing inversion errors. The conventional radiation transmission simulation engine generally adopts a preset static lookup table mode when processing the problem of stratosphere aerosol polarized scattering. According to the scheme, data such as extinction coefficients, scattering matrixes and the like corresponding to different aerosol parameters are calculated and stored in advance, and scattering data of the corresponding parameters are called in an interpolation mode in the simulation and inversion process. However, the parameter space of such a static model is fixed and cannot cover the scene that the micro-physical parameters of the stratospheric aerosol continuously change between the calm period and the volcanic activity period. Because the interpolation operation under the nonstandard event can introduce uncontrollable system errors, the ozone inversion accuracy is directly reduced. Meanwhile, the static lookup table shields the conversion process from the micro-physical parameters to the scattering substrate, and the real-time linkage of the micro-physical parameters and the phase function cannot be realized, so that the algorithm optimization lacks data support. In addition, the general full-polarization radiation transmission simulation engine directly adopts a Stokes matrix of 4 rows and 4 columns to calculate, physical symmetry of spherical liquid drops of a stratosphere is not combined to perform calculation path optimization, calculation redundancy is caused, a parallel calculation architecture of a graphic processor cannot be fully utilized, and simulation efficiency does not reach the upper limit of hardware. Therefore, how to realize accurate and efficient dynamic injection of stratospheric aerosol polarized scattering data becomes a technical problem to be solved urgently. Disclosure of Invention The invention mainly aims to provide a dynamic injection method and a system for spherical aerosol polarized scattering data, which aim to realize accurate and efficient dynamic injection of stratospheric aerosol polarized scattering data. In order to achieve the above object, the present invention provides a method for dynamically injecting spherical aerosol polarized scattering data, comprising the steps of: (1) Receiving a continuously varying set of microphysical parameters of the aerosol; (2) Substituting the micro-physical parameter set into a particle size distribution model, and calculating an original scattering matrix in Stokes observation space in real time based on Mie scattering theory ; (3) Converting the original scattering matrix through a substrate conversion protocolMapping from the Stokes observation space to a computation space with diagonalized features to extract independent features of polarization components, wherein cross-coupling terms in the computation space satisfy decoupling constraints due to physical symmetry of aerosol particles; (4) Carrying out structured packaging on the scattered data subjected to substrate conversion treatment according to the memory layout of the target engine; (5) And injecting the packaged data into a computing kernel of the target engine in a forward simulation process of the target engine by adopting a memory dynamic coverage technology. Preferably, the micro-physical parameters in the micro-physical parameter set include a first micro-physical parameter effective radiusGeometric standard deviation of second micro-physical parameterComplex refractive index of third micro physical parameter。 Preferably, the first micro-physical parameter effective radiusCalculated according to the following formula: Wherein, the Is the geometric mean radius of the aerosol. Preferably, the step (2) specifically includes: substituting the micro-physical parameter set into an aerosol unimodal lognormal distribution model in a particle size space Weighted integration is carried out on the obtained product, and the extinction coefficient of the aerosol is obtained by calculationScattering coefficientThe original scattering matrixThe distribution formula of the aerosol unimodal lognormal distribution model is as follows: Wherein, the Is a vertical height of the steel plate, and is a vertical height,Is the concentration of particles as a function of height; Indicating the cumulative number concentration. Preferably, the aerosol is a stratospheric aerosol. Preferably, the base conversion protocol in the step (3) is implemented by a linear transformation operatorAnd its inverse matrixThe spatial mapping is performed and the conversion formula