CN114488198A - Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest - Google Patents

Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest Download PDF

Info

Publication number
CN114488198A
CN114488198A CN202111623849.0A CN202111623849A CN114488198A CN 114488198 A CN114488198 A CN 114488198A CN 202111623849 A CN202111623849 A CN 202111623849A CN 114488198 A CN114488198 A CN 114488198A
Authority
CN
China
Prior art keywords
aerosol
look
properties
normalized
microphysical
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111623849.0A
Other languages
Chinese (zh)
Inventor
肖达
王南朝
童奕澄
张凯
刘崇
刘�东
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiaxing Research Institute of Zhejiang University
Original Assignee
Jiaxing Research Institute of Zhejiang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jiaxing Research Institute of Zhejiang University filed Critical Jiaxing Research Institute of Zhejiang University
Priority to CN202111623849.0A priority Critical patent/CN114488198A/en
Publication of CN114488198A publication Critical patent/CN114488198A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/95—Lidar systems specially adapted for specific applications for meteorological use
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Remote Sensing (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Electromagnetism (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

本发明公开了一种基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性方法，包括：构建气溶胶光学‑微物理特性查找表；归一化输入光学特性和查找表获得

Figure DDA0003439247470000011
和Gnorm；计算
Figure DDA0003439247470000012
与Gnorm对应个体的偏差并缩减查找表；对归一化光学特性全排列获得搜索序列，依次计算
Figure DDA0003439247470000013
与Gnorm对应的归一化光学特性的相对误差，按误差大小缩减查找表可行解，将保留的查找表可行解平均获得备选微物理特性；基于随机森林理论，生成NRF个搜索序列并获得NRF个备选解，平均备选解获得反演的微物理特性和光学特性；并进一步计算其余的微物理特性。本发明结合KNN‑最近邻和随机森林原理，基于多波长激光雷达实现细模态气溶胶复折射率精确反演并将其应用于更少激光雷达通道。

Figure 202111623849

The invention discloses a multi-wavelength laser radar aerosol micro-physical characteristic method based on multivariate full-array random forest, comprising: constructing a look-up table of aerosol optical-micro-physical characteristics; normalizing input optical characteristics and obtaining the look-up table

Figure DDA0003439247470000011
and G norm ; compute
Figure DDA0003439247470000012
The deviation of the individual corresponding to G norm is reduced and the look-up table is reduced; the search sequence is obtained by arranging the normalized optical properties, and then calculate
Figure DDA0003439247470000013
The relative error of the normalized optical characteristics corresponding to G norm , the feasible solutions of the look-up table are reduced according to the size of the error, and the remaining feasible solutions of the look-up table are averaged to obtain alternative microphysical properties; based on the random forest theory, N RF search sequences are generated and Obtain N RF candidate solutions, average the candidate solutions to obtain the inverted microphysical properties and optical properties; and further calculate the remaining microphysical properties. The invention combines the KNN-nearest neighbor and random forest principles, realizes the precise inversion of the complex refractive index of the fine-mode aerosol based on the multi-wavelength laser radar, and applies it to fewer laser radar channels.

Figure 202111623849

Description

Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest
Technical Field
The invention belongs to the technical field of atmospheric aerosol remote sensing laser radars, and particularly relates to a multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on a multi-element full-array random forest.
Background
Haze not only harms human health, promotes diseases such as respiratory diseases and cardiovascular diseases to be highly developed, but also causes regional and global climate change, so that weather such as extreme high temperature and strong rainfall is frequent, and the normal production and life of human are also seriously influenced. Aerosol particles suspended in the atmosphere are not only the main culprit of causing atmospheric pollution (haze) but also key factors of causing extreme weather phenomena. Atmospheric aerosols can be simply classified into different types of urban pollution, smog, sea salt, mineral dust, and the like. The first two aerosols with small spherical particles are usually dominated by the fine mode fraction, often referred to as fine mode aerosols, in terms of Particle Size Distribution (PSD). The fine mode is an important aspect of atmospheric research and is one of important sources of uncertainty of climate prediction, and is closely related to human activities.
In order to better quantify the impact of fine mode aerosols on earth energy budget, air quality and human health, their fundamental micro-physical properties, such as Complex Refractive Index (CRI), PSD, bulk properties (effective radius, number concentration, surface area concentration, volume concentration) and Single Scattering Albedo (SSA), need to be probed and studied.
YAG laser-based advanced multi-wavelength lidar, such as Raman lidar and High Spectral Resolution Lidar (HSRL), have been shown to describe the micro-physical properties of the fine-mode aerosol, which provides a unique opportunity for studying the fundamental properties of the fine-mode aerosol in distance resolution. The backscattering coefficients (β) at 355nm, 532nm and 1064nm and the extinction coefficients (α) at 355nm and 532nm, commonly referred to as 3 β +2 α structures, measured by a three-wavelength lidar, can be used to invert the aerosol micro-physical properties. However, limited to limited input parameters, inversion of aerosol microphysical properties from 3 β +2 α lidar data requires solving ill-posed inversion problems.
At present, regularization, linear evaluation method, permutation search average method and other methods have been developed to solve the inverse problem of inversion of micro physical properties. The regularization algorithm represents the PSD by a set of triangular basis functions, but it is very time consuming and insensitive to the inversion of the complex refractive index; the linear evaluation method utilizes the linear combination of the optical characteristics of the aerosol to represent the PSD, and can invert the bulk characteristics of the aerosol under certain precision; the permutation search averaging method extracts alternative solutions in the lookup table through different permutation orders to obtain complex refractive index and bulk characteristics, and the process is time-consuming and serious.
The above methods are difficult to obtain high-precision aerosol complex refractive index and single scattering albedo, and often rely on the optical characteristic input of 3 β +2 α to obtain satisfactory micro-physical characteristic inversion results (especially regularization algorithm), which is difficult to implement in most of the current laser radar systems. Particularly in airborne and space-borne lidar system applications.
Disclosure of Invention
Aiming at the defects of inversion of aerosol micro-physical characteristics of the multi-wavelength laser radar, the invention provides a multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on a multi-element full-array random forest, which is based on an improved unsupervised random forest algorithm (RF) and a K-nearest neighbor (KNN) algorithm, aims to solve the problems of high-precision inversion of a CRI of a fine-mode aerosol and inversion of micro-physical characteristics of the aerosol with few laser radar channels (such as a 3 beta +1 alpha structure and a 2 beta +1 alpha structure), and provides reliable input guarantee for SSA inversion of the laser radar.
A multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on a multi-element full-array random forest comprises the following steps:
(1) based on aerosol microReal part of complex refractive index m in physical propertiesrComplex imaginary part m of refractive indexiMode radius rmedConstructing a lookup table G according to the geometric variance sigma; the size of the lookup table G is the aerosol micro-physical property (m)r,mi,rmedσ) number of groups NLUTEach group comprising aerosol micro-physical properties (m)r,mi,rmedσ) and corresponding three-wavelength backscattering coefficient βλAnd extinction coefficient alphaλ；
(2) Input multi-wavelength aerosol optical properties ginputFor the optical characteristics ginputAnd a look-up table G for obtaining normalized optical properties
Figure BDA0003439247450000031
And normalized look-up table Gnorm(ii) a Wherein the optical characteristics are normalized
Figure BDA0003439247450000032
Is Nnorm；
(3) Based on K-nearest neighbor theory, calculating normalized lookup table GnormOf the sets of normalized optical properties and the input normalized optical properties
Figure BDA0003439247450000033
A deviation D of (A); sorting the obtained deviations D from small to large, and selecting the deviation D with the smallest error
Figure BDA0003439247450000034
Group elements as initial look-up tables, ω0To reduce the coefficient, record the new lookup table as G(0)；
(4) To pair
Figure BDA0003439247450000035
Performing a full permutation operation to obtain a product having NnormA search sequence of individual elements; sequentially calculating according to the sequence of each element of the search sequence
Figure BDA0003439247450000036
And searchingThe error d of the aerosol normalization optical characteristic individual corresponding to the table;
based on a K-nearest neighbor theory, utilizing a branch shearing coefficient omega to reduce a feasible solution range in a lookup table, wherein omega belongs to (0, 1); in the search sequence, the feasible solutions of the lookup table obtained after the last element is searched are averaged to obtain the alternative solution (m) of the micro-physical characteristicsr,mi,rmed,σ)retAnd inverted optical properties gret；
(5) Based on random forest theory, randomly generating NRFSearch sequence to obtain NRFIndividual micro-physical property alternative solution
Figure BDA0003439247450000037
And corresponding non-normalized optical properties
Figure BDA0003439247450000038
θ＝1,2,...NRF(ii) a Averaging the alternative solutions to obtain final inverted micro-physical property results
Figure BDA0003439247450000039
And non-normalized optical properties
Figure BDA00034392474500000310
(6) Based on the obtained
Figure BDA00034392474500000311
And
Figure BDA00034392474500000312
calculating to obtain other micro-physical characteristics including effective radius
Figure BDA00034392474500000313
Number concentration
Figure BDA00034392474500000314
Surface area concentration
Figure BDA00034392474500000315
Volume concentration
Figure BDA00034392474500000316
And mode radius under number concentration distribution
Figure BDA00034392474500000317
The final obtained micro physical characteristic set is
Figure BDA00034392474500000318
In the step (1), the three-wavelength backscattering coefficient betaλAnd extinction coefficient alphaλIn (b), λ is a wavelength corresponding to wavelengths of 355nm, 532nm and 1064 nm. The specific process of constructing the lookup table G is as follows:
the correlation between the micro-physical and optical properties of the aerosol is expressed by the Fredholm integral equation:
Figure BDA00034392474500000319
wherein g is the optical characteristics of backscattering coefficient or extinction coefficient, lambda is the wavelength, Kg(r, λ, m; p) is a volume kernel function, the value of which depends on the complex refractive index m ═ m of the aerosol particlesr+miAnd a particle size range r, p being the shape factor; r ismaxAnd rminRespectively the upper and lower limits of the particle size distribution; v (r) is the volume concentration particle size distribution, expressed as a log normal distribution for fine mode aerosols, as follows:
Figure BDA0003439247450000041
wherein, VtAs total volume concentration, rmedIs the mode radius, σ is the geometric variance;
the lookup table is built for volume concentration normalization (V)t＝1μm3cm-3) Thus by a set of determined fine mode aerosol micro physical characteristic parameters (m)r,mi,rmedσ) the optical properties of the aerosol at the corresponding wavelength λ can be obtained; by using NLUTConstructing a lookup table G by aerosol micro-physical characteristics of group global traversal, wherein the size of G is the group number N of the micro-physical characteristicsLUTRepresenting, each set comprising input aerosol micro-physical properties (m)r,mi,rmedσ) and corresponding three-wavelength backscattering coefficient βλAnd extinction coefficient alphaλ。
In the step (2), the optical characteristics ginputAnd lookup table G, normalized backscattering coefficient and extinction coefficient are expressed as:
Figure BDA0003439247450000042
Figure BDA0003439247450000043
wherein,
Figure BDA0003439247450000044
and
Figure BDA0003439247450000045
is the second order norm of the backscattering coefficient and extinction coefficient of different wavelengths; meanwhile, the ratio of extinction coefficient to backscattering of the aerosol is also independent of aerosol number concentration, which is expressed as:
Figure BDA0003439247450000046
for the 3 β +2 α structure, normalized optical properties were obtained as:
Figure BDA0003439247450000047
recording the number of the carbon atoms as Nnorm＝11。
For the 3 β +1 α structure (β:355,532,1064nm, α:532nm), normalized optical properties were obtained as:
Figure BDA0003439247450000051
normalized number of optical characteristics Nnorm＝6；
For the 2 β +1 α structure (β:355,532nm, α:532nm), normalized optical properties were obtained as:
Figure BDA0003439247450000052
normalized number of optical characteristics Nnorm＝4。
Respectively to optical characteristics ginputAnd look-up table G performs the above process to obtain normalized optical properties
Figure BDA0003439247450000053
And look-up table GnormAnd the normalized optical characteristics of both are arranged in the same order.
In the step (3), the deviation D is obtained by calculating the Mahalanobis distance, and the calculation formula is as follows:
Figure BDA0003439247450000054
wherein,
Figure BDA0003439247450000055
for look-up table GnormAnd the kth group element, S is the corresponding covariance matrix.
In step (4), a compound having N is obtainednormSearch sequence of elements, NnormThe elements are randomly generated and may be repeated.
Sequentially calculating according to the sequence of each element of the search sequence
Figure BDA0003439247450000056
The error d of the aerosol normalized optical characteristic individual corresponding to the lookup table is calculated by the following formula:
Figure BDA0003439247450000057
wherein, # denotes the number of searches,
Figure BDA0003439247450000058
an ith normalized optical property for a kth group element in the look-up table after the # th normalized optical property is reduced; reduced # normalized optical Properties lookup Table G(#)Number of feasible solutions of
Figure BDA0003439247450000059
In the step (6), the formula for obtaining other micro physical characteristics by calculation is as follows:
Figure BDA0003439247450000061
Figure BDA0003439247450000062
Figure BDA0003439247450000063
Figure BDA0003439247450000064
Figure BDA0003439247450000065
in the formula,
Figure BDA0003439247450000066
in terms of the volume concentration, the concentration of the active ingredient,
Figure BDA0003439247450000067
is the mode radius in a number concentration distribution,
Figure BDA0003439247450000068
in terms of a number of concentrations,
Figure BDA0003439247450000069
in terms of the surface area concentration,
Figure BDA00034392474500000610
is the effective radius.
Compared with the existing aerosol micro-physical characteristic inversion algorithm, the method provided by the invention can realize the high-efficiency and accurate inversion of the complex refractive index of the fine-mode aerosol and provide high-precision input for the inversion of the SSA. Meanwhile, the inversion result of the aerosol micro-physical property under fewer laser radar channels (such as 3 beta +1 alpha, 2 beta +1 alpha structures) can be optimized, so that the method has important significance for reducing the hardware cost and expanding the application range of the multi-wavelength laser radar for inverting the aerosol micro-physical property, and provides powerful support for more comprehensively understanding the aerosol property and the effect thereof in climate change.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a diagram of inversion micro-physics results based on noiseless data in an embodiment of the present invention;
FIG. 3 is a diagram of inversion microphysical property results based on Gaussian noise data in an embodiment of the invention
Detailed Description
The invention will be described in further detail below with reference to the drawings and examples, which are intended to facilitate the understanding of the invention without limiting it in any way.
As shown in fig. 1, a multi-wavelength lidar aerosol micro-physical property inversion method based on a multivariate full-array random forest comprises the following steps:
step S1: complex refractive index real part m based on micro physical propertyrComplex imaginary part of refractive index miMode radius rmedGeometric variance σ, etc. of micro-physical characteristicsAnd constructing a lookup table G.
The correlation between the micro-physical and optical properties of the aerosol can be expressed by the Fredholm integral equation:
Figure BDA0003439247450000071
wherein g is the optical characteristics such as backscattering coefficient or extinction coefficient, λ is the wavelength, and the common wave bands are 355,532 and 1064 nm. Kg(r, λ, m; p) is a volume kernel function, the value of which depends on the complex refractive index m ═ m of the aerosol particlesr+miAnd the particle size range r, p is the shape factor, the invention takes spherical particles as an example, and can be obtained by Mie scattering theory. r ismaxAnd rminThe upper and lower limits of the particle size distribution, and v (r) the volume concentration Particle Size Distribution (PSD). The particle size distribution of the fine mode aerosol is usually expressed by a lognormal distribution, and the formula is as follows:
Figure BDA0003439247450000072
wherein, VtAs total volume concentration, rmedIs the mode radius, σ is the geometric variance. The invention establishes volume concentration normalization (V)t＝1μm3cm-3) Thus by a set of determined fine mode aerosol micro physical characteristic parameters (m)r,mi,rmedσ) the optical properties of the aerosol at the corresponding wavelength λ are obtained. By using NLUTAnd constructing a lookup table G by the aerosol micro-physical characteristics of the global traversal. Number of groups N of micro physical characteristics for GLUTRepresenting, each set comprising input aerosol micro-physical properties (m)r,mi,rmed,σ,Vt) And corresponding three-wavelength backscattering coefficient betaλAnd extinction coefficient alphaλWhere λ is 355,532,1064 nm.
In this embodiment, the lookup table parameters are: m isr＝1.34:0.02:1.66，mi＝0:0.001:0.05， rmed＝100:10:300nm, ln σ 0.38:0.02:0.5, so NLUT127449. In rice scattering theory, p is 1 and the particle size range r is 0:0.001:50 μm, i.e. rmax＝50μm，rmin＝0。
Step S2: input multi-wavelength aerosol optical properties ginputFor the optical characteristics ginputAnd a look-up table G to obtain normalized optical properties
Figure BDA0003439247450000073
And look-up table GnormIn which the optical properties are normalized
Figure BDA0003439247450000074
Is Nnorm。
Taking the 3 β +2 α structure as an example (β:355,532,1064nm, α:355,532nm), the normalized backscattering coefficient and extinction coefficient are expressed as:
Figure BDA0003439247450000081
Figure BDA0003439247450000082
wherein,
Figure BDA0003439247450000083
and
Figure BDA0003439247450000084
is the second order norm of the backscattering coefficient and extinction coefficient at different wavelengths. Meanwhile, the ratio of extinction coefficient to backscattering of the aerosol is also independent of aerosol number concentration, which is expressed as:
Figure BDA0003439247450000085
for the 3 β +2 α structure, the normalized optical properties are:
Figure BDA0003439247450000086
recording the number of the carbon atoms as Nnorm＝11。
For the 3 beta +1 alpha (beta: 355,532,1064nm, alpha: 532nm) structure, the corresponding optical characteristics in the lookup table are selected, and the normalized optical characteristics with the normalized optical characteristics of
Figure BDA0003439247450000087
Normalized number of optical characteristics Nnorm＝6。
For a 2 β +1 α (β:355,532nm, α:532nm) structure, the normalized optical properties are
Figure BDA0003439247450000088
Normalized number of optical characteristics Nnorm＝4。
Respectively to optical characteristics ginputAnd look-up table G performs the above process to obtain normalized optical properties
Figure BDA0003439247450000089
And look-up table GnormAnd the normalized optical characteristics of both are arranged in the same order.
In this embodiment, to verify the universality of the method of the present invention, the optical characteristics generated by the simulation of a plurality of groups of micro physical characteristics are used as input, and the simulation parameters are as follows: m isr＝1.35,1.45,1.55,1.65， mi＝0.001,0.005,0.01,0.015,0.02,0.025,0.05，rmed100,140,180,240,300nm, ln σ 0.4, and V0.1, yielding 192 sets of multi-wavelength (λ 355,532,1064nm) optical properties. In the present embodiment, there are two sets of test data, one set of test data is noise-free, and the other set of test data is randomly added with gaussian noise, where the noise level is 20%, and in the present embodiment, it represents that the noise level reaches 20% at a position with three times of standard deviation of gaussian distribution. The two groups of observation data are respectively input into a 3 beta +2 alpha structure, a 3 beta +1 alpha structure (beta: 355,532,1064nm, alpha: 532nm) and a 2 beta +1 alpha structure (beta: 355,532nm, alpha: 532nm) for method test.
Step S3: based on K-nearest neighbor theory, calculatingNormalized optical properties
Figure BDA0003439247450000091
And GnormAnd the deviation D of each group of normalized optical characteristics is obtained by calculating the Mahalanobis distance, and the calculation formula is as follows:
Figure BDA0003439247450000092
wherein,
Figure BDA0003439247450000093
for look-up table GnormThe kth group element. And S is a corresponding covariance matrix. Sorting the obtained deviations D from small to large, and selecting the error with the minimum
Figure BDA0003439247450000094
Group elements as initial look-up tables, ω0To reduce the coefficient, record the new lookup table as G(0)(ii) a In this embodiment, ω0＝0.1。
Step S4: to pair
Figure BDA0003439247450000095
Performing a full permutation operation to obtain a product having NnormSearch sequence of elements, NnormThe elements are randomly generated and may be repeated. Sequentially calculating the correspondence according to the sequence of each element in the search sequence
Figure BDA0003439247450000096
The error d of the aerosol normalized optical characteristic individual corresponding to the lookup table is calculated by the following formula:
Figure BDA0003439247450000097
wherein, # denotes the number of searches,
Figure BDA0003439247450000098
for # normalized lightThe ith normalized optical property of the kth group element in the look-up table after the reduction of the optical property. Based on the K-nearest neighbor theory, the feasible solution range in the lookup table is reduced by utilizing a branch-cut coefficient omega, and omega belongs to (0, 1). Thus, the # normalized optical property reduced look-up table G(#)Number of feasible solutions of
Figure BDA0003439247450000099
Averaging the reduced lookup table obtained after searching the last element of the search sequence to obtain the alternative solution (m) of the micro-physical characteristicsr,mi,rmed,σ)retAnd inverted optical properties gret. In this embodiment, ω is 0.3.
Step S5: based on random forest theory, randomly generating NRFSearch sequence to obtain NRFIndividual micro-physical property alternative solution
Figure BDA00034392474500000910
And corresponding non-normalized optical properties
Figure BDA00034392474500000911
θ＝1,2,...NRF. Averaging the alternative solutions to obtain final inverted micro-physical property results
Figure BDA00034392474500000912
And non-normalized optical properties
Figure BDA00034392474500000913
In this example, NRF＝500。
Step S6: based on the obtained
Figure BDA00034392474500000914
And
Figure BDA00034392474500000915
calculating to obtain the effective radius
Figure BDA00034392474500000916
Number concentration
Figure BDA00034392474500000917
Surface area concentration
Figure BDA00034392474500000918
Volume concentration
Figure BDA00034392474500000919
And other micro-physical characteristics. The specific calculation process is as follows:
Figure BDA0003439247450000101
Figure BDA0003439247450000102
Figure BDA0003439247450000103
Figure BDA0003439247450000104
Figure BDA0003439247450000105
wherein,
Figure BDA0003439247450000106
is the mode radius in the number concentration distribution. The final set of available microphysical properties is therefore
Figure BDA0003439247450000107
As shown in fig. 2, (a) - (d) respectively represent absolute error of real part of complex refractive index, absolute error of imaginary part of complex refractive index, relative error of volume concentration inversion, and relative error of effective radius inversion when three optical characteristic structures of 3 β +2 α,3 β +1 α, and 2 β +1 α are input. As the number of input optical characteristics decreases, the inversion errors of the real part of the complex refractive index, the imaginary part, the volume concentration, and the effective radius gradually increase. Under a 3 beta +2 alpha structure, the average value of the absolute errors of the real part of the complex refractive index obtained by the method is 0.016 (<0.05), the average value of the absolute errors of the imaginary part is 0.0019, and the corresponding relative errors are 0.36(<0.5), so that the errors are obviously reduced compared with the error of the conventional algorithm, and the method can obtain a more accurate complex refractive index inversion result. Meanwhile, under three input optical characteristic structures, the maximum values of inversion relative errors of the volume concentration and the effective radius are respectively 26% and 18%, which shows that the method can be applied to fewer optical characteristic channels and has important significance for reducing hardware cost and expanding application range of the multi-wavelength laser radar for inverting the micro physical characteristics of the aerosol.
As shown in fig. 3, (a) - (d) respectively represent absolute error of real part of complex refractive index, absolute error of imaginary part of complex refractive index, relative error of volume concentration inversion, and relative error of effective radius inversion when three optical characteristic structures of 3 β +2 α,3 β +1 α, and 2 β +1 α are input. Under the condition of introducing 20% of Gaussian noise, under a 3 beta +2 alpha structure, the absolute error of the real part of the complex refractive index obtained by the invention can still be less than 0.05, and the relative errors of the volume concentration and the effective radius are respectively 16% and 8%. The method is applied to other input optical characteristic structures, and shows better anti-noise capability on inversion of several micro-physical characteristics, particularly on volume concentration and effective radius, and the maximum relative errors of the method are respectively less than 34% and 29%, so that the method has the potential of inverting the micro-physical characteristics of the aerosol under fewer laser radar channels and has better anti-noise capability.
The embodiments described above are intended to illustrate the technical solutions and advantages of the present invention, and it should be understood that the above-mentioned embodiments are only specific embodiments of the present invention, and are not intended to limit the present invention, and any modifications, additions and equivalents made within the scope of the principles of the present invention should be included in the scope of the present invention.

Claims (8)

1.一种基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，包括以下步骤：1. a multi-wavelength laser radar aerosol microphysical property inversion method based on multivariate full arrangement random forest, is characterized in that, comprises the following steps: (1)基于气溶胶微物理特性中的复折射率实部mr、复折射率虚部mi、模式半径rmed和几何方差σ构建查找表G；查找表G的大小为气溶胶微物理特性(mr,mi,rmed,σ)的组数NLUT，每组包含气溶胶微物理特性(mr,mi,rmed,σ)及对应的三波长后向散射系数βλ以及消光系数αλ；(1) Build a look-up table G based on the real part m r of the complex refractive index, the imaginary part of the complex refractive index m i , the mode radius r med and the geometric variance σ in the microphysical properties of the aerosol; the size of the look-up table G is the size of the aerosol microphysics The number of groups N LUT of properties (m r ,m i ,r med ,σ), each group contains aerosol microphysical properties (m r ,m i ,r med ,σ) and the corresponding three-wavelength backscattering coefficient β λ and the extinction coefficient α λ ; (2)输入多波长气溶胶光学特性ginput，对光学特性ginput和查找表G进行归一化以获得归一化光学特性
Figure FDA0003439247440000011
和归一化查找表Gnorm；其中，归一化光学特性
Figure FDA0003439247440000012
的数量为Nnorm；(2) Input the multi-wavelength aerosol optical properties g input , and normalize the optical properties g input and the look-up table G to obtain the normalized optical properties
Figure FDA0003439247440000011
and the normalized look-up table G norm ; where the normalized optical properties
Figure FDA0003439247440000012
The number is N norm ; (3)基于K-最近邻理论，计算归一化查找表Gnorm中各组归一化光学特性和输入的归一化光学特性
Figure FDA0003439247440000013
的偏差D；将获得的偏差D进行从小到大排序，选择误差最小的
Figure FDA0003439247440000014
组元素作为初始查找表，ω0为缩减系数，记新的查找表为G(0)；(3) Based on the K-nearest neighbor theory, calculate the normalized optical properties of each group in the normalized look-up table G norm and the normalized optical properties of the input
Figure FDA0003439247440000013
the deviation D; sort the obtained deviation D from small to large, and select the one with the smallest error
Figure FDA0003439247440000014
The group element is used as the initial look-up table, ω 0 is the reduction coefficient, and the new look-up table is recorded as G (0) ; (4)对
Figure FDA0003439247440000015
进行全排列操作，获得具有Nnorm个元素的搜索序列；按照搜索序列各元素顺序，依次计算
Figure FDA0003439247440000016
与查找表对应的气溶胶归一化光学特性个体的误差d；(4) Right
Figure FDA0003439247440000015
Perform a full permutation operation to obtain a search sequence with N norm elements; according to the order of each element of the search sequence, calculate in turn
Figure FDA0003439247440000016
The individual error d of the aerosol normalized optical properties corresponding to the look-up table; 基于K-最近邻理论，利用枝剪系数ω缩减查找表中可行解范围，ω∈(0,1)；在搜索序列中，对最后一个元素搜索后获得的查找表可行解进行平均，获得微物理特性备选解(mr,mi,rmed,σ)ret及反演的光学特性gret；Based on the K-nearest neighbor theory, the range of feasible solutions in the lookup table is reduced by the branch and shear coefficient ω, ω∈(0,1); in the search sequence, the feasible solutions of the lookup table obtained after the last element search are averaged to obtain the micro Alternative solutions for physical properties (m r , m i , r med , σ) ret and inverse optical properties g ret ; (5)基于随机森林理论，随机生成NRF个搜索序列，获得NRF个微物理特性备选解
Figure FDA0003439247440000017
和对应的非归一化光学特性
Figure FDA0003439247440000018
将备选解平均以获得最终反演的微物理特性结果
Figure FDA0003439247440000019
和非归一化光学特性
Figure FDA00034392474400000110
(5) Based on the random forest theory, randomly generate N RF search sequences, and obtain N RF candidate solutions for microphysical properties
Figure FDA0003439247440000017
and the corresponding unnormalized optical properties
Figure FDA0003439247440000018
Averaging alternative solutions to obtain final inverted microphysics results
Figure FDA0003439247440000019
and unnormalized optical properties
Figure FDA00034392474400000110
 (6)基于获得的
Figure FDA0003439247440000021
和
Figure FDA0003439247440000022
计算获得其他微物理特性，包括有效半径
Figure FDA0003439247440000023
数浓度
Figure FDA0003439247440000024
表面积浓度
Figure FDA0003439247440000025
体积浓度
Figure FDA0003439247440000026
以及数浓度分布下的模式半径
Figure FDA0003439247440000027
最终获得的微物理特性集合为
Figure FDA0003439247440000028
(6) Based on the obtained
Figure FDA0003439247440000021
and
Figure FDA0003439247440000022
Calculate other microphysical properties, including effective radius
Figure FDA0003439247440000023
number concentration
Figure FDA0003439247440000024
Surface area concentration
Figure FDA0003439247440000025
Volume concentration
Figure FDA0003439247440000026
and the mode radius under the number concentration distribution
Figure FDA0003439247440000027
The final set of microphysical properties obtained is
Figure FDA0003439247440000028
 2.根据权利要求1所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(1)中，三波长后向散射系数βλ以及消光系数αλ中，λ为波长，对应波长为355nm、532nm和1064nm。2. the multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full arrangement random forest according to claim 1, is characterized in that, in step (1), three wavelength backscattering coefficient β λ and extinction coefficient In α λ , λ is the wavelength, and the corresponding wavelengths are 355nm, 532nm and 1064nm. 3.根据权利要求2所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(1)中，构建查找表G的具体过程为：3. the multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full arrangement random forest according to claim 2, is characterized in that, in step (1), the concrete process of constructing look-up table G is: 气溶胶的微物理特性和光学特性的联系通过Fredholm积分方程表示：The relationship between the microphysical properties and optical properties of aerosols is expressed by the Fredholm integral equation:
Figure FDA0003439247440000029
Figure FDA0003439247440000029
 其中，g为后向散射系数或消光系数等光学特性，λ为波长，Kg(r,λ,m；p)为体积核函数，其数值取决于气溶胶粒子的复折射率m＝mr+mi和粒径范围r，p为形状因子，针对球形颗粒，通过Mie散射理论获得Kg(r,λ,m；p)；rmax和rmin分别为粒径分布的上下限；v(r)为体积浓度粒径分布，对细模态气溶胶用对数正态分布表示，公式如下：Among them, g is the optical properties such as backscattering coefficient or extinction coefficient, λ is the wavelength, and K g (r,λ, m; p) is the volume kernel function, the value of which depends on the complex refractive index of the aerosol particle m=m r +m i and particle size range r, p is the shape factor, for spherical particles, K g (r, λ, m; p) is obtained by Mie scattering theory; r max and r min are the upper and lower limits of the particle size distribution, respectively; v (r) is the volume concentration particle size distribution, which is represented by lognormal distribution for fine mode aerosols, and the formula is as follows:
Figure FDA00034392474400000210
Figure FDA00034392474400000210
 其中，Vt为总的体积浓度，rmed为模式半径，σ为几何方差；where V t is the total volume concentration, r med is the mode radius, and σ is the geometric variance; 建立的查找表为体积浓度归一化(Vt＝1μm3cm-3)的查找表，因此通过一组确定的细模态气溶胶微物理特性参数(mr,mi,rmed,σ)即可获得对应波长λ下的气溶胶光学特性；利用NLUT组全局遍历的气溶胶微物理特性构建查找表G，G的大小用微物理特性的组数NLUT表示，每组包括输入的气溶胶微物理特性(mr,mi,rmed,σ)以及对应的三波长后向散射系数βλ以及消光系数αλ。The established look-up table is a volume-concentration-normalized (V t =1 μm 3 cm -3 ) look-up table, so a set of fine-mode aerosol microphysical property parameters (m r ,m i ,r med ,σ ) to obtain the aerosol optical properties at the corresponding wavelength λ; a look-up table G is constructed by using the aerosol microphysical properties traversed globally by the N LUT group, and the size of G is represented by the number of groups of microphysical properties N LUTs , each group including the input Aerosol microphysical properties (m r , mi , r med , σ) and the corresponding three-wavelength backscattering coefficient β λ and extinction coefficient α λ . 4.根据权利要求1所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(2)中，对光学特性ginput和查找表G进行归一化时，对于3β+2α结构，归一化后向散射系数和消光系数表示为：4. the multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full arrangement random forest according to claim 1, is characterized in that, in step (2), optical characteristic g input and look-up table G are normalized. When normalized, for the 3β+2α structure, the normalized backscattering coefficient and extinction coefficient are expressed as:
Figure FDA0003439247440000031
Figure FDA0003439247440000031
Figure FDA0003439247440000032
Figure FDA0003439247440000032
 其中，
Figure FDA0003439247440000033
和
Figure FDA0003439247440000034
是不同波长后向散射系数和消光系数的二阶范数；同时，气溶胶的消光系数与后向散射之比也与气溶胶数浓度无关，其表示为：in,
Figure FDA0003439247440000033
and
Figure FDA0003439247440000034
is the second-order norm of the backscattering coefficient and extinction coefficient at different wavelengths; at the same time, the ratio of the aerosol extinction coefficient to backscattering is also independent of the aerosol number concentration, which is expressed as:
Figure FDA0003439247440000035
Figure FDA0003439247440000035
 对于3β+2α结构，获得归一化光学特性为：For the 3β+2α structure, the normalized optical properties are obtained as:
Figure FDA0003439247440000036
记其数量为Nnorm＝11；
Figure FDA0003439247440000036
Denote its number as N norm =11; 对于3β+1α结构，获得归一化光学特性为：For the 3β+1α structure, the normalized optical properties are obtained as:
Figure FDA0003439247440000037
归一化光学特性数量Nnorm＝6；
Figure FDA0003439247440000037
the number of normalized optical properties N norm =6; 对于2β+1α结构，获得归一化光学特性为
Figure FDA0003439247440000038
归一化光学特性数量Nnorm＝4；For the 2β+1α structure, the normalized optical properties are obtained as
Figure FDA0003439247440000038
the number of normalized optical properties N norm =4; 分别对光学特性ginput以及查找表G进行以上过程以获得归一化的光学特性
Figure FDA0003439247440000039
以及查找表Gnorm，并将两者的归一化光学特性以相同的顺序排列。Perform the above process on the optical characteristic g input and the lookup table G respectively to obtain the normalized optical characteristic
Figure FDA0003439247440000039
and the look-up table G norm , and arrange the normalized optical properties of both in the same order. 5.根据权利要求1所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(3)中，偏差D通过马氏距离计算获得，计算公式为：5. The multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full arrangement random forest according to claim 1, is characterized in that, in step (3), deviation D is obtained by Mahalanobis distance calculation, and the calculation formula for:
Figure FDA0003439247440000041
Figure FDA0003439247440000041
 其中，
Figure FDA0003439247440000042
为查找表Gnorm中第k组元素，S为对应的协方差矩阵。in,
Figure FDA0003439247440000042
is the kth element in the lookup table G norm , and S is the corresponding covariance matrix. 6.根据权利要求1所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(4)中，获得具有Nnorm个元素的搜索序列时，Nnorm个元素随机产生，可以重复。6. The multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full arrangement random forest according to claim 1, is characterized in that, in step (4), when obtaining the search sequence with N norm elements, N norm elements are randomly generated and can be repeated. 7.根据权利要求1所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(4)中，按照搜索序列各元素顺序，依次计算
Figure FDA0003439247440000043
与查找表对应的气溶胶归一化光学特性个体的误差d，计算公式如下：7. The multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full array random forest according to claim 1, is characterized in that, in step (4), according to the order of each element of the search sequence, calculate sequentially
Figure FDA0003439247440000043
The error d of the individual aerosol normalized optical properties corresponding to the look-up table is calculated as follows:
Figure FDA0003439247440000044
Figure FDA0003439247440000044
 式中，#表示搜索次数，
Figure FDA0003439247440000045
为第#个归一化光学特性缩减后的查找表中第k组元素第i个归一化光学特性；第#个归一化光学特性缩减后的查找表G(#)的可行解数量为
Figure FDA0003439247440000046
In the formula, # represents the number of searches,
Figure FDA0003439247440000045
is the ith normalized optical property of the k-th group element in the reduced look-up table for the #th normalized optical property; the number of feasible solutions of the reduced look-up table G (#) for the #th normalized optical property is
Figure FDA0003439247440000046
 8.根据权利要求1所述的基于多元全排列随机森林的多波长激光雷达气溶胶微物理特性反演方法，其特征在于，步骤(6)中，计算获得其他微物理特性的公式如下：8. the multi-wavelength lidar aerosol microphysical property inversion method based on multivariate full arrangement random forest according to claim 1, is characterized in that, in step (6), the formula that calculates and obtains other microphysical properties is as follows:
Figure FDA0003439247440000047
Figure FDA0003439247440000047
Figure FDA0003439247440000048
Figure FDA0003439247440000048
Figure FDA0003439247440000049
Figure FDA0003439247440000049
Figure FDA00034392474400000410
Figure FDA00034392474400000410
Figure FDA00034392474400000411
Figure FDA00034392474400000411
 式中，
Figure FDA00034392474400000412
为体积浓度，
Figure FDA00034392474400000413
为数浓度分布下的模式半径，
Figure FDA00034392474400000414
为数浓度，
Figure FDA00034392474400000415
为表面积浓度。In the formula,
Figure FDA00034392474400000412
is the volume concentration,
Figure FDA00034392474400000413
is the mode radius under the number concentration distribution,
Figure FDA00034392474400000414
is the number concentration,
Figure FDA00034392474400000415
is the surface area concentration.
CN202111623849.0A 2021-12-28 2021-12-28 Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest Pending CN114488198A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111623849.0A CN114488198A (en) 2021-12-28 2021-12-28 Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111623849.0A CN114488198A (en) 2021-12-28 2021-12-28 Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest

Publications (1)

Publication Number Publication Date
CN114488198A true CN114488198A (en) 2022-05-13

Family

ID=81496556

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111623849.0A Pending CN114488198A (en) 2021-12-28 2021-12-28 Multi-wavelength laser radar aerosol micro-physical characteristic inversion method based on multi-element full-array random forest

Country Status (1)

Country Link
CN (1) CN114488198A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118584504A (en) * 2024-05-27 2024-09-03 武汉大学 A multi-wavelength lidar aerosol microphysical characteristics inversion method and system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102008003037B3 (en) * 2008-01-02 2009-09-24 Leibniz-Institut für Troposphärenforschung e. V. Method for determining physical characteristics of atmospheric particles, involves preparing vertical profile of optical characteristic of particles and determining vertical profile of physical characteristic of particle by inversion method
CN109086801A (en) * 2018-07-06 2018-12-25 湖北工业大学 A kind of image classification method based on improvement LBP feature extraction
CN109884664A (en) * 2019-01-14 2019-06-14 武汉大学 A method and system for optical microwave synergistic inversion of urban above-ground biomass
CN110161532A (en) * 2019-05-30 2019-08-23 浙江大学 A method of based on multi-wavelength laser radar inverting microfluidic aerosol physical characteristic
CN110488252A (en) * 2019-08-08 2019-11-22 浙江大学 Overlap factor calibration device and calibration method for a ground-based aerosol lidar system
CA3146697A1 (en) * 2019-08-06 2021-02-11 Amgen Inc. Systems and methods for determining protein concentrations of unknown protein samples based on automated multi-wavelength calibration

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102008003037B3 (en) * 2008-01-02 2009-09-24 Leibniz-Institut für Troposphärenforschung e. V. Method for determining physical characteristics of atmospheric particles, involves preparing vertical profile of optical characteristic of particles and determining vertical profile of physical characteristic of particle by inversion method
CN109086801A (en) * 2018-07-06 2018-12-25 湖北工业大学 A kind of image classification method based on improvement LBP feature extraction
CN109884664A (en) * 2019-01-14 2019-06-14 武汉大学 A method and system for optical microwave synergistic inversion of urban above-ground biomass
CN110161532A (en) * 2019-05-30 2019-08-23 浙江大学 A method of based on multi-wavelength laser radar inverting microfluidic aerosol physical characteristic
CA3146697A1 (en) * 2019-08-06 2021-02-11 Amgen Inc. Systems and methods for determining protein concentrations of unknown protein samples based on automated multi-wavelength calibration
CN110488252A (en) * 2019-08-08 2019-11-22 浙江大学 Overlap factor calibration device and calibration method for a ground-based aerosol lidar system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
李晓涛: "大气气溶胶粒径分布的多波长激光雷达反演", 《光学学报》, vol. 44, no. 06, 8 April 2024 (2024-04-08), pages 165 - 172 *
赵双;陈曙晖;: "基于机器学习的流量识别技术综述与展望", 计算机工程与科学, no. 10, 15 October 2018 (2018-10-15), pages 34 - 44 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118584504A (en) * 2024-05-27 2024-09-03 武汉大学 A multi-wavelength lidar aerosol microphysical characteristics inversion method and system

Similar Documents

Publication Publication Date Title
CN110161532B (en) Method for inverting micro-physical characteristics of aerosol based on multi-wavelength laser radar
Hobbs et al. Simulation-based uncertainty quantification for estimating atmospheric co _2 from satellite data
CN105784556B (en) A kind of air fine particles PM based on Self-organized Fuzzy Neural Network2.5Flexible measurement method
CN120123697B (en) A PM2.5 chemical component vertical profile inversion system and method driven by multimodal reinforcement learning
CN106920007A (en) PM based on second order Self-organized Fuzzy Neural Network2.5Intelligent Forecasting
CN116881798A (en) Conditional gracile causal analysis method based on variable selection and reverse time lag feature selection for complex systems such as weather
CN119442019A (en) Bounded uncertainty analysis method of interior sound field based on physical information neural network
Chowdhury et al. Probabilistic stability assessment of slopes using high dimensional model representation
Doodson Relation of the mode, median and mean in frequency curves
Kaiser et al. Fast weighting functions for retrievals from limb scattering measurements
He et al. Application of the LSQR algorithm in non-parametric estimation of aerosol size distribution
CN116067338B (en) Surface subsidence monitoring method and system based on laser point cloud data
Jourdan et al. Statistical analysis of cloud light scattering and microphysical properties obtained from airborne measurements
Fussen et al. Evolution of stratospheric aerosols in the post-Pinatubo period measured by solar occultation
CN119148107A (en) Calibration method for temperature and humidity Raman laser radar in high aerosol concentration scene
Zawada et al. Two-dimensional analytic weighting functions for limb scattering
AdilabdAlkareem et al. Comparison different estimation methods for the parameters of non-linear regression
CN118584504B (en) Inversion method and system for micro-physical characteristics of multi-wavelength laser radar aerosol
Nakabayashi et al. An extension of the ensemble kalman filter for estimating the observation error covariance matrix based on the variational Bayes’s method
Chang et al. Predicting aerosol extinction coefficient with LiDAR data based on deep belief network
Kayser et al. Modelling the uncertainties of wind farm noise predictions
Li et al. Retrieval of aerosol microphysical parameters based on a few optical parameters from lidar
Thornton The use of an agglomerative numerical technique in physical evidence comparisons
CN119026026A (en) A method for identifying aerosol particle types based on optimal estimation
Zhangirov et al. Forward Propagation Neural Network Weighting Analysis as a Model Estimation Method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
AD01 Patent right deemed abandoned
AD01 Patent right deemed abandoned

Effective date of abandoning: 20251028