A sea surface salinity inversion method based on Fengyun meteorological satellite

Abstract

The present invention discloses a sea surface salinity inversion algorithm based on Fengyun meteorological satellite, comprising the following steps: S1: training independent variable (X): increasing the number of hidden layers according to the needs of regression tasks, building a model for deep neural network to perform sea surface salinity satellite inversion algorithm, which is a total of 4 independent variables of 4-band satellite observation remote sensing reflectivity (Rrs), aCDOM, sea surface temperature (SST), and suspended matter concentration (TSM); S2: measuring sea surface salinity (Y): establishing the relationship between aCDOM, SST, TSM and sea surface salinity through a deep learning model, and using the FY-3D / MERSI global observation capability to achieve global sea surface salinity observation. The present invention establishes a sea surface salinity inversion algorithm for domestic Fengyun meteorological satellites (FY-3D / MERSI) based on a deep learning method, and realizes the observation of high spatiotemporal resolution of sea surface salinity. Due to the limitations of their low spatiotemporal resolution and data quality, SMOS and Aquarius / SAC-D satellites have low coverage of sea surface salinity data, which makes it impossible for sea surface salinity to obtain high spatiotemporal resolution observation.

Description

Technical Field

[0001] The invention relates to the technical field of sea surface salinity observation, and in particular to a sea surface salinity inversion method based on Fengyun meteorological satellites. Background Art

[0002] Sea surface salinity is an important physical and chemical parameter of the ocean, and is also one of the important parameters that determine the basic properties of seawater. Sea surface salinity is not only an essential environmental variable in exploring ocean phenomena such as thermohaline circulation and global sea level change, but also provides a forced reference for water mass analysis, climate research, and global ocean model research. The monitoring of sea surface salinity is also widely used in many fields such as petroleum, import and export trade, military waterway measurement, and offshore fisheries. However, the sea surface salinity measured by traditional ship aerial surveys and drifting buoys is discrete in time and space, and cannot accurately detect changes in the spatial and temporal distribution of sea surface salinity.

[0003] The Moderate Resolution Spectral Imager (MERSI-2) carried on my country's second-generation Fengyun polar-orbiting meteorological satellite FY-3D integrates the functions of the two imaging instruments (MERSI-1 and VIRR) on the original Fengyun-3 satellite. It is the world's first imaging instrument capable of acquiring global 250-meter resolution infrared split window area data. It can obtain global 250-meter resolution true color images seamlessly every day, and realize high-precision quantitative inversion of atmospheric, land and ocean parameters such as clouds, aerosols, water vapor, land surface characteristics and ocean color. The high temporal and spatial resolution of FY-3D / MERSI (250m, scanning the target area once a day) is used to form high-temporal and spatial resolution observations of the sea surface.

[0004] Artificial intelligence has become the core driving force of the fourth industrial revolution. Artificial Intelligence (AI) is a science. Machine learning is the most mainstream method for implementing artificial intelligence, and deep learning is a branch of machine learning. Deep learning is a method in machine learning based on data characterization learning. It was proposed by Hinton et al. in 2006. It is a machine learning method that can simulate the neural structure of the human brain, bringing hope for solving optimization problems related to deep structures. The present invention uses deep learning technology to develop a sea surface salinity inversion method that can be used for the medium-resolution spectroscopic imager (FY-3D / MERSI) of my country's second-generation Fengyun polar-orbiting meteorological satellite. Summary of the invention

[0005] The purpose of the present invention is to provide a sea surface salinity inversion method based on Fengyun meteorological satellites to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a sea surface salinity inversion method based on Fengyun meteorological satellite, comprising the following steps:

[0007] S1: Training independent variables (X): According to the needs of the regression task, the number of hidden layers is increased, and a deep neural network is used to build a satellite inversion algorithm model for sea surface salinity. There are four independent variables, namely, 4-band satellite observation remote sensing reflectance (Rrs), aCDOM, sea surface temperature (SST), and suspended matter concentration (TSM). The steps for obtaining the four independent variables are as follows:

[0008] S11: satellite observation remote sensing reflectivity (Rrs);

[0009] S12: aCDOM inversion;

[0010] S13: SST inversion;

[0011] S14: TSM inversion;

[0012] S2: Measure sea surface salinity (Y): Invert aCDOM, sea surface temperature (SST), and suspended matter concentration in FY-3D / MERSI

[0013] Based on the sea surface salinity (TSM), the relationship between aCDOM, SST, TSM and sea surface salinity was established through a deep learning model, and the global observation capability of FY-3D / MERSI was used to realize the observation of global sea surface salinity.

[0014] Preferably, the S11: satellite observed remote sensing reflectivity (Rrs): can be directly obtained from the FY-3D / MERSI satellite observed reflectivity after atmospheric correction.

[0015] Preferably, the S12: aCDOM inversion: In the process of obtaining aCDOM using remote sensing data, firstly the total absorption of non-aqueous water at 443 nm is used. nw Backscattering coefficient b at (443) and 555 nm bp (555) and remote sensing reflectance parameters to establish an empirical formula to derive the absorption coefficient a of non-phytoplankton particles d (443):

[0016] a d (443) = 0.60 × σ 0.90 ,

[0017] in, ,

[0018] a nw (443) and b bp (555) is obtained by the QAA algorithm. Secondly, after obtaining a d(443), using the relationship between the bands, any band a can be calculated d Values:

[0019]

[0020] In getting a d After that, using the total absorption of non-aqueous nw , calculate the total absorption value a of CDOM and phytoplankton phg :

[0021] a phg (λ) = a nw (λ)-a d (λ),

[0022] Finally, through the 412nm, 443nm and 490nm a phg Separate the absorption coefficient a of the yellow substance at 443nm CDOM (443): a CDOM (443) = a phg (443) / (1+9.56×10 4 ×e -11.13×ψ ).

[0023] Preferably, in S12:

[0024]

[0025] λ1=412, λ0=443, λ2=490.

[0026] Preferably, S13: SST inversion: Sea surface temperature is an important climate and weather parameter, which has a great influence on the distribution of sea surface salinity. The infrared band of the satellite sensor can be used to invert the sea surface temperature, and a multivariate nonlinear inversion algorithm based on atmospheric correction is used to invert SST:

[0027] SST=a+bT 11 +c(T 11 -T 12 )T sfc +d(T11-T12)(sec(θ)-1),

[0028] Among them, T 11 and T 12 are the brightness temperatures of the 11um and 12um bands directly observed by the satellite, θ is the zenith angle of satellite observation, T 12 is the initial SST field, a, b, c and d are constants, and the corresponding values ​​can be obtained by simulating the transmission of infrared radiation in the atmosphere based on the numerical model.

[0029] Preferably, the S14: TSM inversion: inverting the suspended matter concentration in the water body using an algorithm for inverting the suspended matter concentration in the water body based on the intrinsic optical properties (IOP) of the water body.

[0030] TSM=1.73*b p (443);

[0031] b p (443) = b bp (443) / 0.015.

[0032] Preferably, the S14: in the TSM inversion p (443) and b bp (443) are the scattering coefficient and backscattering coefficient of the particles, b p (443) and b bp (443) can be calculated using the QAA (Quasi-analytical algorithm) semi-analytical algorithm (see Table 1 for the algorithm flow).

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. Based on the deep learning method, the present invention establishes the sea surface salinity inversion algorithm of the domestic Fengyun meteorological satellite (FY-3D / MERSI), realizing the observation of sea surface salinity with high temporal and spatial resolution. The SMOS and Aquarius / SAC-D satellites have low temporal and spatial resolution (30-300km, scanning period of more than 3 days) and data quality limitations, resulting in low coverage of sea surface salinity data, making it impossible to observe sea surface salinity with high temporal and spatial resolution;

[0035] 2. The present invention solves the problem that SMOS and Aquarius satellites use the relationship between sea surface temperature, seawater dielectric constant and sea surface salinity detected by microwave band to invert sea surface salinity by establishing a sea surface salinity inversion algorithm for the domestic Fengyun meteorological satellite (FY-3D / MERSI). However, microwaves are affected by the atmosphere and background radiation during transmission, the sea surface radiation brightness temperature gain caused by sea surface roughness, and the influence of the white canopy on seawater brightness temperature all affect the inversion accuracy of sea surface salinity. In addition, the low spatiotemporal resolution (30-300km, scanning period of more than 3 days) of microwave sensor data also makes it difficult to meet the demand for high spatiotemporal resolution sea surface salinity monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is an overall flow chart of a method for inverting sea surface salinity based on Fengyun meteorological satellites according to the present invention;

[0037] Figure 2This is a neural network structure diagram of a sea surface salinity inversion method based on Fengyun meteorological satellites in the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] See also Figure 1-2 The present invention provides a technical solution: a sea surface salinity inversion method based on Fengyun meteorological satellite, comprising the following steps:

[0040] S1: Training independent variables X: According to the needs of the regression task, the number of hidden layers is increased, and a deep neural network is used to perform a satellite inversion algorithm model for sea surface salinity. There are four independent variables, namely, the 4-band satellite observation remote sensing reflectance Rrs, aCDOM, sea surface temperature SST, and suspended matter concentration TSM. The specific steps for obtaining the four independent variables are as follows:

[0041] S11: Satellite observation remote sensing reflectivity Rrs: can be directly obtained from the FY-3D / MERSI satellite observation reflectivity after atmospheric correction;

[0042] S12: aCDOM inversion: In the process of obtaining aCDOM using remote sensing data, the total non-water absorption a at 443 nm is first used. nw Backscattering coefficient b at (443) and 555 nm bp (555) and remote sensing reflectance parameters to establish an empirical formula to derive the absorption coefficient a of non-phytoplankton particles d (443):

[0043] a d (443) = 0.60 × σ 0.90 ,

[0044] in, ,

[0045] a nw (443) and b bp (555) is obtained by the QAA algorithm. Secondly, after obtaining a d (443), using the relationship between the bands, any band a can be calculated d Values:

[0046]

[0047] In getting ad After that, using the total absorption of non-aqueous nw , calculate the total absorption value a of CDOM and phytoplankton phg :

[0048] a phg (λ) = a nw (λ)-a d (λ),

[0049] Finally, through the 412nm, 443nm and 490nm a phg Separate the absorption coefficient a of the yellow substance at 443nm CDOM (443): a CDOM (443) = a phg (443) / (1+9.56×10 4 ×e -11.13×ψ ).

[0050] in,

[0051] λ1=412, λ0=443, λ2=490.

[0052] S13: SST inversion: Sea surface temperature is an important climate and weather parameter, which has a great influence on the distribution of sea surface salinity. The infrared band of satellite sensors can be used to invert sea surface temperature, and a multivariate nonlinear inversion algorithm based on atmospheric correction is used to invert SST:

[0053] SST=a+bT 11 +c(T 11 -T 12 )T sfc +d(T11-T12)(sec(θ)-1),

[0054] Among them, T 11 and T 12 are the brightness temperatures of the 11um and 12um bands directly observed by the satellite, θ is the zenith angle of satellite observation, T 12 is the initial SST field (the National Centers for Environmental Prediction (NCEP) reanalysis sea surface temperature dataset is used in this patent study), a, b, c and d are constants, and the corresponding values ​​can be obtained by simulating the transmission of infrared radiation in the atmosphere based on the numerical model.

[0055] S14: TSM inversion: The suspended matter concentration in water is inverted using an algorithm based on the intrinsic optical properties (IOP) of water.

[0056] TSM=1.73*b p (443);

[0057] bp (443) = b bp (443) / 0.015.

[0058] where b p (443) and b bp (443) are the scattering coefficient and backscattering coefficient of the particles, b p (443) and b bp (443) can be calculated using the QAA (Quasi-analytical algorithm) semi-analytical algorithm (see Table 1 for the algorithm flow).

[0059] Table 1 QAA algorithm calculation steps for absorption coefficient and backscattering coefficient:

[0060]

[0061] Among them, λ is the spectral wavelength, rrs(λ) is the remote sensing emissivity in the λ band, χ, μ(λ) and η are intermediate variables, g0 and g1 are model constants, and a w (λ) is the absorption coefficient of pure water in the λ band, a(λ) is the total absorption coefficient of water, b b (λ) and b bp (λ) are the water backscattering coefficient and the particle backscattering coefficient respectively;

[0062] S2: Measurement of sea surface salinity Y: Based on the inversion of aCDOM, sea surface temperature (SST), and suspended matter concentration (TSM) by FY-3D / MERSI, the relationship between aCDOM, SST, TSM and sea surface salinity was established through a deep learning model, and the global observation capability of FY-3D / MERSI was used to realize the observation of global sea surface salinity.

[0063] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0064] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A sea surface salinity inversion method based on Fengyun meteorological satellite, characterized in that: The following steps are involved: S1: Training independent variables X: According to the needs of the regression task, the number of hidden layers is increased, and a deep neural network is used to perform a satellite inversion algorithm model for sea surface salinity. There are four independent variables, namely, the 4-band satellite observation remote sensing reflectivity Rrs, aCDOM, sea surface temperature SST, and suspended matter concentration TSM. The steps for obtaining the four independent variables are as follows: S11: Satellite observation remote sensing reflectivity Rrs; S12: aCDOM inversion. In the process of obtaining aCDOM using remote sensing data, the total non-water absorption at 443 nm is first used. nw Backscattering coefficient b at (443) and 555 nm bp (555) and remote sensing reflectance parameters to establish an empirical formula to derive the absorption coefficient a of non-phytoplankton particles d (443): a d (443)=0.60×σ 0.90 , in, a nw (443) and b bp (555) is obtained by the QAA algorithm. Secondly, after obtaining a d (443), using the relationship between the bands, any band a can be calculated d Values: In getting a d After that, using the total absorption of non-aqueous nw , calculate the total absorption value a of CDOM and phytoplankton phg : a phg (λ)=a nw (l)-a d (l), Finally, through the 412nm, 443nm and 490nm a phg Separate the absorption coefficient a of the yellow substance at 443nm CDOM (443): a CDOM (443) = a phg (443) / (1+9.56×10 4 ×e -11.13×ψ ); S13: SST inversion. Sea surface temperature is an important climate and weather parameter, which has a great influence on the distribution of sea surface salinity. The infrared band of satellite sensors can be used to invert sea surface temperature. The multivariate nonlinear inversion algorithm based on atmospheric correction is used to invert SST: SST=a+bT 11 +c(T 11 -T 12 )T sfc +d(T11-T12)(sec(θ)-1), Among them, T 11 and T 12 are the brightness temperatures of the 11um and 12um bands directly observed by the satellite, θ is the zenith angle of satellite observation, T 12 is the initial SST field, a, b, c and d are constants, and the corresponding values ​​can be obtained by simulating the transmission of infrared radiation in the atmosphere based on numerical models; S14: TSM inversion, using the algorithm based on the inherent optical property of water IOP to invert the concentration of suspended matter in water; S2: Measure sea surface salinity Y.

2. The method for inverting sea surface salinity based on Fengyun meteorological satellite according to claim 1, characterized in that: The satellite observed remote sensing reflectivity Rrs is directly obtained from the FY-3D / MERSI satellite observed reflectivity after atmospheric correction.

3. The sea surface salinity inversion method based on Fengyun meteorological satellite according to claim 1 is characterized in that: In S12: λ1=412, λ0=443, λ2=490.

4. The sea surface salinity inversion method based on Fengyun meteorological satellite according to claim 1 is characterized in that: The formula for inverting the suspended matter concentration in water body using the algorithm for inverting the suspended matter concentration in water body based on the inherent optical property IOP of water body in the TSM inversion is: TSM=1.73*b p (443);b p (443)=b bp (443) / 0.015。 5. The method for inverting sea surface salinity based on Fengyun meteorological satellite according to claim 4, characterized in that: The TSM inversion p (443) and b bp (443) are the scattering coefficient and backscattering coefficient of the particles, b p (443) and b bp (443) can be calculated using the QAA semi-analytical algorithm.

Images