	US 12,461,028 B2
	Method and device for inverting ocean-atmosphere optical parameters and storage medium
Xianqiang He, Hangzhou (CN); Tianfeng Pan, Hangzhou (CN); Yan Bai, Hangzhou (CN); Fang Gong, Hangzhou (CN); Teng Li, Nanjing (CN); and Difeng Wang, Hangzhou (CN)
Assigned to SECOND INSTITUTE OF OCEANOGRAPHY, MNR, Hangzhou (CN)
Filed by SECOND INSTITUTE OF OCEANOGRAPHY, MNR, Zhejiang (CN)
Filed on Nov. 19, 2024, as Appl. No. 18/952,261.
Claims priority of application No. 202311624941.8 (CN), filed on Nov. 30, 2023.
Prior Publication US 2025/0180477 A1, Jun. 5, 2025
Int. Cl. G01N 21/53 (2006.01); G01N 21/17 (2006.01); G01N 33/18 (2006.01); G06F 17/16 (2006.01)

CPC G01N 21/53 (2013.01) [G01N 33/18 (2013.01); G06F 17/16 (2013.01); G01N 2021/1793 (2013.01)]

11 Claims

1. A method for inverting ocean-atmosphere optical parameters, comprising:

obtaining a vector apparent reflectance ρ_l, ρ_Q, ρ_Uof water-leaving radiance at top of atmosphere and observation geometry of a multi-angle polarized satellite, wherein the observation geometry comprises three angles: a solar zenith angle, a view zenith angle and a relative azimuth angle;

inputting the vector apparent reflectance ρ_l, ρ_Q, ρ_Uand the observation geometry into a trained ocean-atmosphere optical parameter model, to output a group of ocean-atmosphere optical parameters, wherein the output ocean-atmosphere optical parameters comprise: a vector apparent reflectance ρ_BOA^wof water-leaving radiance at bottom of atmosphere above sea surface of the multi-angle polarized satellite, a fine/coarse mode aerosol optical thickness, a fine/coarse mode aerosol particle size distribution radius, a sea surface wind speed, a chlorophyll concentration and a suspended particulate matter concentration;

wherein the ocean-atmosphere optical parameter model is an XGBoost machine training model,

wherein the method further comprises training the ocean-atmosphere optical parameter model, and the training comprises:

obtaining multiple groups of sample ocean-atmosphere optical parameters, wherein the sample ocean-atmosphere optical parameters in each group comprise a sample fine/coarse mode aerosol optical thickness, a sample aerosol complex refraction index real/imaginary part, a sample fine/coarse mode aerosol particle size distribution radius, a sample sea surface wind speed, a sample chlorophyll concentration, a sample yellow substance absorption coefficient, a sample yellow substance absorption spectrum slope, and a sample suspended particulate matter concentration;

obtaining a sample observation geometry of the multi-angle polarized satellite;

inputting the multiple groups of sample ocean-atmosphere optical parameters and the sample observation geometry into an Ocean Successive Orders with Atmosphere-Advanced radiative transfer model, to output a vector apparent reflectance ρ_TOA^wof water-leaving radiance at top of atmosphere, and obtaining a first vector apparent reflectance ρ_BOA^wof water-leaving radiance at bottom of atmosphere above sea surface; and

training the ocean-atmosphere optical parameter model by using the multiple groups of sample ocean-atmosphere optical parameters, the sample observation geometry, the vector apparent reflectance ρ_TOA^wof water-leaving radiance at top of atmosphere and the first vector apparent reflectance ρ_BOA^wof water-leaving radiance at bottom of atmosphere above sea surface; and

wherein the method further comprises optimizing the ocean-atmosphere optical parameter model by retrieving, from a plurality of global sea stations, measured values corresponding to actual atmospheric-ocean conditions, generating a plurality of output results from the ocean-atmosphere optical parameter model based on a portion of the actual atmospheric-ocean conditions, and comparing the plurality of output results to actual atmospheric-ocean conditions not included in the portion.