CN114238847A - Surface layer accurate-rotation reconstruction method and system based on ocean measured data - Google Patents

Surface layer accurate-rotation reconstruction method and system based on ocean measured data Download PDF

Info

Publication number
CN114238847A
CN114238847A CN202111269295.9A CN202111269295A CN114238847A CN 114238847 A CN114238847 A CN 114238847A CN 202111269295 A CN202111269295 A CN 202111269295A CN 114238847 A CN114238847 A CN 114238847A
Authority
CN
China
Prior art keywords
data
salinity
sea surface
sea
satellite
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.)
Granted
Application number
CN202111269295.9A
Other languages
Chinese (zh)
Other versions
CN114238847B (en
Inventor
陈建
安玉柱
张伟涛
陈璇
闫恒乾
白成祖
沈晓晶
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.)
61540 Troops of PLA
Original Assignee
61540 Troops of PLA
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 61540 Troops of PLA filed Critical 61540 Troops of PLA
Priority to CN202111269295.9A priority Critical patent/CN114238847B/en
Publication of CN114238847A publication Critical patent/CN114238847A/en
Application granted granted Critical
Publication of CN114238847B publication Critical patent/CN114238847B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • G06F17/13—Differential equations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
    • G06F17/12—Simultaneous equations, e.g. systems of linear equations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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

Abstract

The invention relates to a surface layer accurate rotation reconstruction method and a system based on ocean measured data, wherein the method comprises the following steps: collecting data of various depths of the ocean as a training set; training SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm by taking the data of the sea table in the training set as input and the data of each depth below the sea table as output to obtain an ocean underwater data prediction model; obtaining sea table data observed by a satellite; correcting the sea surface data observed by the satellite by adopting a random forest regression model to obtain corrected sea surface data; replacing the monthly climate state background field of the daily sea surface data in the corrected sea surface data with a daily changing background field to obtain the sea surface data after replacing the background field; and inputting the sea surface data after the background field replacement into the marine underwater data prediction model to obtain corresponding marine data of each depth below the sea surface. The invention improves the accuracy and robustness of the reconstructed data.

Description

Surface layer accurate-rotation reconstruction method and system based on ocean measured data
Technical Field
The invention relates to the technical field of marine data prediction, in particular to a surface layer accurate-rotation reconstruction method and system based on marine measured data.
Background
The development of satellite observation technology in the last two decades plays an unprecedented promoting role in the research of the mesoscale phenomenon in the ocean. In recent years, quasi-global Sea Surface Temperature (SST) and Sea Surface Salinity (SSS) datasets have reached a grid spatial resolution of 0.05 °, such as OSTIA SST and SMOS-BEC-L4 SSS; the upcoming launch SWOT satellites will also provide quasi-global, mid-scale, sub-mid-Scale Sea Height (SSH) images. Nevertheless, the satellite can only observe the ocean surface. Argo buoys, on the other hand, can provide reliable profiles up to 2000 meters deep, but they are sporadic and sparse. Combining and supplementing satellite data and field observations with each other has become a recognized solution (e.g., vinogrova et al, 2019). The technique of estimating the interior of the ocean from satellite data is known as underwater inversion/reconstruction/estimation. This technology is seen as an extension of remote sensing, so-called mid-depth ocean remote sensing (Klemas & Yan, 2014). Due to the needs of practical application, statistical methods are mostly adopted, and two types of methods are mainly adopted.
One approach is to directly construct a (generalized) regression relationship between the surface field and the profile. The well-known Modular Ocean Data Assimilation System (MODAS) was first to reconstruct a three-dimensional field from surface temperature and sea level altitude using the multivariate linear regression Method (MLR) (Fox et al, 2002). Similar techniques based on MODAS, scientists estimated subsurface temperature/salinity fields (T/S) using various satellite data (e.g., Guineahut et al, 2004; buongino nardell, et al, 2012). Advances in Artificial Intelligence (AI) technology have inspired researchers to use artificial intelligence algorithms instead of the traditional linear methods. These algorithms include artificial neural networks (balabera-poy et al, 2009), self-organizing neural networks (Wu et al, 2012), support vector machines (Su et al, 2015), Random Forests (RF), generalized regression neural networks (FOAGRNN) (Bao et al, 2019) based on drosophila optimization algorithms, and the like. The extraction method based on EOF (empirical orthogonal function) mode can also derive the vertical projection mode of sea surface and underwater coupling. Such methods include single EOF reconstruction (seofs et al, 1990,1994), maximum empirical mode (GEM) (meien & american watts, 2000) and coupled mode reconstruction (CPR) (buongino nardeli & Santoleri,2004) and multivariate EOF reconstruction (meofs-R) methods (buongino nardeli & Santoleri, 2005; buongino nardeli et al, 2006; Wang et al, 2012; buongino nardeli et al, 2017; buongino nardeli et al, 2018).
In addition to the above-described purely statistical methods, another route of research is to use simplified kinetic models, such as the skin-aligned (SQG) method. As a reconstruction method, the SQG algorithm has long been used for inversion of tropospheric wind fields, the eddy current (PV) inversion (Held & Pierrehumbert, 1995). The first reconstruction of SQG method in the ocean utilized only the SQG modality (Isern-Fontanet et al, 2006). Lapeyre & Klein (2006) proposed an effective sqg (esqg) method of coupling the surface buoyancy to the subsurface PV. Wang et al (2013) proposed an inside-skin quasi-rotation (isQG, interior + SQG) method, which includes a SQG mode and an inside mode (positive pressure mode and first bias pressure mode). The isQG method has been applied to reanalyzed datasets (Liu et al, 2014) and preliminary validation was performed based on observed data (Liu et al, 2017).
Aiming at the problem that the SQG mode and the first bias pressure mode can only reflect the inherent defects of the PV structure (Asassi et al, 2016), SQG and the isQG method when a subsurface reinforced vortex (usually with the center between 200-1000 m) is simulated, Yan et al (2020) combines the SQG mode with the mEOF statistical mode to provide the SQG-mEOF-R method combining the advantages of the statistical method and the dynamic method, and the effectiveness of the new method in the northwest Pacific ocean (NWP; 30-38N, 150E-158E) dominated by surface enhanced vortices (SEs) and the southeast Pacific ocean (SEP; 16-24S, 96-104W) dominated by subsurface enhanced vortices (SSEs) is verified in the product output in OFES (ocean circulation mode for earth simulator) mode of vortex resolution through Observation System Simulation Experiment (OSSE).
In the SQG kinetic equation, T/S is not an explicit variable. Although T/S can be projected directly through the density profile shape (Pedlosky,1982) with the premise of linearizing the equation of state, the fitting of its amplitude is not robust. To this end, Yan et al (2021) first implemented a study of underwater temperature salt inversion using SQG density reconstruction (LS-mEOFs algorithm), projecting the density reconstruction field onto an empirical mode, and determining the amplitude of the mEOFs using least squares fitting. By comparing the T/S inversion of SQG and derivative algorithms (SQG, isQG, SQG-mEOF-R) with linear statistical algorithms (MLR, mEOF-R) and non-linear machine learning algorithms (FOAGRNN and RF) using vortex resolution mode data, it is shown that the combination of SQG-based density reconstruction (and any other density field) and LS-mEOFs algorithm-based T/S inversion is an effective power-statistics framework, and the combination of power signals can be much better than blind use of complex machine learning algorithms.
SQG and the derived algorithm have good effect in the observation system simulation experiment based on the model product, but the application of the algorithm to the measured data is not feasible at once. In fact, at present, only the work of Liu et al, 2017 involves the application of measured data of SQG method (hereinafter referred to as L17 scheme), but the reconstruction results with satellite data as input are quite poor. The L17 solution attributes the problem to large errors in the remotely sensed sea surface salinity data, but in fact produces large reconstruction errors even with a monthly climate state as the background field. In model product-based applications, it is not difficult to find that the background field varies from day to day. The low-pass filtered background field extracted from such day data is non-static, from which "true" outliers or perturbations can be extracted that are unbiased relative to the real field of the day. And the lunar climate state makes SQG difficult to be solved, and the obvious deviation of the solar data and the lunar data can be projected under water, thereby influencing the reconstruction effect.
Disclosure of Invention
The invention aims to provide a surface layer accurate rotation reconstruction method and system based on ocean measured data, and the accuracy and robustness of reconstructed data are improved.
In order to achieve the purpose, the invention provides the following scheme:
a surface layer accurate-rotation reconstruction method based on ocean measured data comprises the following steps:
collecting data of various depths of the ocean as a training set;
training SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm by taking the data of the sea table in the training set as input and the data of each depth below the sea table as output to obtain an ocean underwater data prediction model;
obtaining sea table data observed by a satellite;
correcting sea surface salinity observed by the satellite by adopting a random forest regression model to obtain corrected sea surface salinity;
replacing the monthly climate state background field of the corrected sea surface data with a background field changing day to obtain sea surface data after the background field is replaced; the corrected sea surface data comprise corrected sea surface salinity;
and inputting the sea surface data after the background field replacement into the marine underwater data prediction model to obtain corresponding marine data of each depth below the sea surface.
Optionally, the data for each depth in the training set includes temperature or salinity in a world oceanographic set, and temperature or salinity profile observations in an Argo global observation V3.0 dataset.
Optionally, the satellite observed sea surface data comprises sea surface temperature data of a Reynolds oist v2.1 data set, sea surface salinity data of a european space agency climate change initiative SSS v1.8 data set, and absolute power topology data of a combined data unification and altimeter system.
Optionally, the step of correcting sea surface salinity observed by the satellite by using a random forest regression model to obtain corrected sea surface salinity specifically includes:
training a random forest regression model by taking sea surface temperature data, sea surface salinity data, absolute power topology data, longitude and latitude observed by a satellite as input and corresponding offshore surface salinity as output to obtain a trained random forest regression model; the offshore surface salinity is the salinity of the shallowest layer in an Argo global observation V3.0 data set;
inputting the sea surface temperature, sea surface salinity, absolute power topology, longitude and latitude observed by the satellite into the trained random forest regression model to obtain the corrected sea surface salinity.
Optionally, the obtaining of the background field of daily variation further comprises:
training a multiple linear regression model by taking absolute power topological data observed from a satellite, offshore surface salinity, offshore surface temperature, longitude and latitude as inputs and salinity or temperature of each depth below a sea surface as outputs to obtain the trained multiple linear regression model; the offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 data set, and the offshore surface temperature is the temperature of the shallowest layer in the Argo global observation V3.0 data set;
inputting the sea surface salinity, the sea surface temperature, the absolute dynamic topological data, the longitude and the latitude which are observed by the satellite and are corrected every day into a multiple linear regression model, and performing low-pass filtering on the output of the multiple linear regression model to obtain a background field corresponding to daily change.
The invention also discloses a surface layer accurate-rotation reconstruction system based on the ocean measured data, which comprises the following steps:
the training set data acquisition module is used for acquiring data of each depth of the ocean as a training set;
the marine underwater data prediction model obtaining module is used for taking the data of the sea table in the training set as input, taking the data of each depth below the sea table as output to train SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm, and obtaining a marine underwater data prediction model;
the system comprises a satellite observation sea table data acquisition module, a satellite observation sea table data acquisition module and a satellite observation sea table data acquisition module, wherein the satellite observation sea table data acquisition module is used for acquiring the sea table data of the satellite observation;
the data correction module is used for correcting sea surface salinity observed by the satellite by adopting a random forest regression model to obtain corrected sea surface salinity;
the background field replacement module is used for replacing the monthly climate state background field of the corrected sea table data with a daily changing background field to obtain the sea table data after the background field replacement; the corrected sea surface data comprise corrected sea surface salinity;
and the ocean data acquisition module of each depth is used for inputting the sea surface data after the background field replacement into the ocean underwater data prediction model to acquire corresponding ocean data of each depth below the sea surface.
Optionally, the data for each depth in the training set includes temperature or salinity in a world oceanographic set, and temperature or salinity profile observations in an Argo global observation V3.0 dataset.
Optionally, the satellite observed sea surface data comprises sea surface temperature data of a Reynolds oist v2.1 data set, sea surface salinity data of a european space agency climate change initiative SSS v1.8 data set, and absolute power topology data of a combined data unification and altimeter system.
Optionally, the data correction module specifically includes:
the random forest regression model training unit is used for training a random forest regression model by taking sea surface temperature data, sea surface salinity data, absolute power topology data, longitude and latitude observed by a satellite as input and corresponding offshore surface salinity as output to obtain a trained random forest regression model; the offshore surface salinity is the salinity of the shallowest layer in an Argo global observation V3.0 data set;
and the data correction unit is used for inputting the sea surface temperature, the sea surface salinity, the absolute power topology, the longitude and the latitude observed by the satellite into the trained random forest regression model to obtain the corrected sea surface salinity.
Optionally, the ambient field replacement module further includes:
the multi-linear regression model training unit is used for training a multi-linear regression model by taking absolute power topological data observed from a satellite, offshore surface salinity, offshore surface temperature, longitude and latitude as input and salinity or temperature of each depth below a sea surface as output to obtain a trained multi-linear regression model; the offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 data set, and the offshore surface temperature is the temperature of the shallowest layer in the Argo global observation V3.0 data set;
and the daily change background field obtaining unit is used for inputting the sea surface salinity, the sea surface temperature, the absolute power topological data, the longitude and the latitude which are observed by the satellite and are corrected every day into the multiple linear regression model, and performing low-pass filtering on the output of the multiple linear regression model to obtain the corresponding daily change background field.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the method, the sea surface data observed by the satellite is corrected through the random forest regression model, the deviation of sea surface data input by the marine underwater data prediction model is reduced, the moon climate state background field of the sea surface data is replaced by the day-changing background field, the deviation of the sea surface data input by the marine underwater data prediction model is further reduced, and the reconstructed data precision and robustness are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a surface layer accurate rotation reconstruction method based on ocean measured data according to the present invention;
FIG. 2 is a detailed flow diagram of a surface layer accurate rotation reconstruction method based on ocean measured data according to the present invention;
fig. 3 is a schematic structural diagram of a surface layer quasi-rotation reconstruction system based on marine measured data according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a surface layer accurate rotation reconstruction method and system based on ocean measured data, and the accuracy and robustness of reconstructed data are improved.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a schematic flow chart of a surface layer quasi-rotation reconstruction method based on ocean measured data, and as shown in fig. 1, a surface layer quasi-rotation reconstruction method based on ocean measured data includes:
step 101: data of the ocean at various depths are collected as a training set.
The collected data includes Sea Surface Temperature (SST) data of a Reynolds OISSTv2.1 data set, Sea Surface Salinity (SSS) data of a European space agency climate change initiative (ESA-CCI) SSS V1.8 data set, Absolute Dynamic Topology (ADT) data of a Data Unification and Altimeter Combination System (DUACS), world oceanic map set (WOA) temperature/salinity (T/S) monthly climate state data, and T/S profile observation data of an Argo global observation V3.0 data set.
Data in 2010-2013 and 2015-2017 are used as training sets, and data in 2014 and 2018 are used as verification sets. Interpolating T/S of original profile data (Argo profile and WOA profile matched with the Argo profile) to 19 vertical layers of [10, 20, 30, 40, 50, 75, 100, 125, 150, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1000] dbar to obtain standard profile data; taking the T/S of the shallowest layer of the original profile data as near surface temperature/near surface salinity (NST/NSS); because one of ADT, DG and SH is taken as the Sea Surface Height (SSH) in different reconstruction schemes, the specific volume height (SH) and the Dynamic Height (DH) of the section data are calculated by taking 1000dbar as a reference layer, and the adjustment from ADT to DH or the adjustment from ADT to SH are carried out through regression between longitude, latitude, ADT and DH or SH, and the ADT or the adjusted DH/SH, SST and the like are taken as input data of the reconstruction scheme, namely sea surface data.
Step 102: and (3) training SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm by taking the data of the sea table in the training set as input and the data of each depth below the sea table as output to obtain the marine underwater data prediction model.
The surface quasi-inversion method (SQG) is based on the quasi-inversion equation, and can directly obtain the density of the ground-inverted flow and the seawater. SQG, the basic assumption is that, under a quasi-translational approximation, the source of the internal ocean current can be decomposed into two parts: the buoyancy/density of the boundary layer is abnormal, and the position vortex in the fluid is abnormal.
The original form of the quasi-geostationary vortex equation is:
Figure BDA0003328108550000071
wherein Ψ is a quasi-terrestrial transition function; f. of0Is the area-averaged coriolis force, β is the meridional average of the area-averaged coriolis force; q is quasi-ground displacement vortex (QGPV); h is the seafloor depth, x, y and z are the spatial locations, x represents longitude, y represents latitude, and z represents depth.
With a low-pass (LP) filter <, the background field should also satisfy equation (1):
L<Ψ>+f0+βy＝<Q> (2)
note that the background field in equation (2) is non-static, i.e., varying with time. The subtraction of equations (1) and (2) is the true QGPV equation after subtraction of the background field, i.e. the perturbation equation form for reconstruction:
Figure BDA0003328108550000081
wherein b iss＝-gρs/ρ0For sea surface buoyancy disturbances, ρsFor density disturbances at sea surface, p0Is the average density of the water column and g is the acceleration of gravity. Ψ and q are quasi-geostationary functions and perturbed versions of QGPV.
Surface-accurate translation (SQG) algorithm: since q is unknown, by setting q to 0, the solution in Fourier space (called the surface solution), which is either the SQG solution or the SQG mode Ψ, is solvedsThen according to
Figure BDA0003328108550000082
Figure BDA0003328108550000083
The density ρ and velocity V are found, which is the SQG method, which projects the skin density signal into the ocean.
Inside-skin quasi-rotation (isQG) algorithm: in an actual ocean, q is not zero inside the ocean, but only theoretical analysis can be performed because the vortex at the inner part cannot be directly observed. The iSQG (interior + SQG) method decomposes the equation into a surface solution and an internal solution. The internal solution is introduced by the sea level height, reflecting the influence of the sea level height signal on the interior of the sea. The positive pressure mode and the first oblique pressure mode are the first two modes of internal solution, and a positive pressure die (F) is superposed0) And a first inclined pressing die (F)1) To model SQG, amplitude A is determined by "sea level" (in fact dynamic altitude DH) anomaly η0，A1(formula (4)); by passing
Figure BDA0003328108550000084
The isQG reconstruction is implemented.
Figure BDA0003328108550000085
Wherein, the alphabet expression plus ^ above is the meaning of the estimated value,
Figure BDA0003328108550000086
an estimate of the value of psi is indicated,
Figure BDA0003328108550000087
to indicate psisIs determined by the estimated value of (c),
Figure BDA0003328108550000088
representing an estimate of η.
Surface quasi-rotation-multivariate EOF reconstruction (SQG-mEOF-R) algorithm: the statistical mEOF mode is used, but it is obtained by subtracting the SQG solution, using a first statistical modulus M1SH (dynamic altitude) abnormality SH of the sea surfaceaFitting amplitude a1(formula (5)); by rho ═ a1M1+ρSQGRealizing SQG-mEOF-R density reconstruction, rhoSQGIs the density of the SQG solution.
SHa(0)＝a1M1(0) (5)
Step 103: sea surface data of satellite observations are obtained.
Step 104: and correcting the sea surface salinity observed by the satellite by adopting a random forest regression model to obtain the corrected sea surface salinity.
Since satellite observations, and particularly SSS data, have large errors, it is unlikely to be perfectly consistent with the offshore table observations of Argo. From equation (3), the solution of SQG is determined by the buoyancy of the surface of the sea, i.e., the density of the surface of the sea. As in the Liu2017 study (L17 scenario), severe bias of satellite SSS is the leading cause of failure of reconstruction with satellite data.
Wherein, step 104 specifically includes:
training a random forest regression model by taking sea surface temperature data, sea surface salinity data, absolute power topology data, longitude and latitude observed by a satellite as input and corresponding offshore surface salinity as output to obtain a trained random forest regression model; offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 dataset.
Inputting sea surface temperature, sea surface salinity, absolute power topology, longitude and latitude observed by the satellite into the trained random forest regression model to obtain corrected sea surface salinity.
Step 105: replacing the monthly climate state background field of the corrected sea surface data with a daily change background field to obtain sea surface data after the background field is replaced; the corrected sea surface data includes corrected sea surface salinity.
The background field cannot simply be replaced by a monthly climatic background field. Equation (3) can be considered to be of course in modal production applications, since the low-pass (LP) filtered background field of the diurnal variation is always available. However, equation (2) does not hold for the monthly climate conditions used in the L17 scheme. Therefore, unless the extra residual of equation (2) is forced to be superimposed on the disturbance, equation (3) cannot be solved in fourier space. When the reconstruction is performed using equation (3), a systematic error will be added. The systematic errors include not only deviations between the monthly climate background and the daily changing background, but also deviations of sea surface disturbances inherited by the climate state background, which deviations will be projected further into the interior of the sea.
Wherein, step 105 specifically comprises:
training a multiple linear regression model by taking absolute power topological data observed by a satellite, offshore surface salinity, offshore surface temperature, longitude and latitude as input and salinity or temperature of each depth below a sea surface as output to obtain a trained multiple linear regression model; the offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 data set, and the offshore surface temperature is the temperature of the shallowest layer in the Argo global observation V3.0 data set.
Inputting the sea surface salinity, the sea surface temperature, the absolute dynamic topological data, the longitude and the latitude which are observed by the satellite and are corrected every day into a multiple linear regression model, and performing low-pass filtering on the output of the multiple linear regression model to obtain a background field corresponding to the daily change.
Step 106: and inputting the sea surface data after the background field replacement into the marine underwater data prediction model to obtain corresponding marine data of each depth below the sea surface.
Data for each depth in the training set includes temperature or salinity in the world ocean map set, and temperature or salinity profile observations in the Argo Global Observation V3.0 dataset.
The sea surface data observed by the satellite comprise sea surface temperature data of a Reynolds OISST v2.1 data set, sea surface salinity data of a European space agency climate change initiative SSS v1.8 data set and absolute power topological data of a data unification and altimeter combined system.
As shown in FIG. 2, the present invention utilizes the background field of the initial guess (the background field of the daily variation), the sea surface density abnormal field, and the average N of the Argo region2Profiles and adjusted DH or SH solutions for SQG, isQG and SQG-mEOF-R were calculated. The dashed line in fig. 2 represents the training node and the underline represents the tag quantity (output quantity).
The invention discloses a surface layer accurate-conversion reconstruction method based on ocean measured data, which is a primary guess (FG) scheme capable of enabling SQG and a derivative algorithm to be applied to the measured data. The scheme is robust, and reconstruction accuracy of the SQG, the isQG and the SQG-mEOF-R in the SEP region with relatively uniform weak layer junction and level can be better than that of an RF machine learning algorithm (the robustness is good, and the accuracy is improved). On one hand, compared with the L17 scheme, the FG scheme enables the equations (2) and (3) to be established by providing a more reliable background field of daily variation, thereby eliminating the deviation of the abnormal input of the sea surface and avoiding 0.97kg/m3(NWP waters) or-0.13 kg/m3The sea surface deviations of (SEP sea area) project into the sea interior. On the other hand, since sea surface salinity is also one of the sources of reconstruction errors, the FG scheme effectively reduces the uncertainty of sea surface density by using the CCI SSS product and performing random forest algorithm correction.
Fig. 3 is a schematic structural diagram of a surface layer quasi-rotation reconstruction system based on ocean measured data, and as shown in fig. 3, a surface layer quasi-rotation reconstruction system based on ocean measured data includes:
and the training set data acquisition module 201 is used for acquiring data of each depth of the ocean as a training set.
And the marine underwater data prediction model obtaining module 202 is used for taking the data of the sea table in the training set as input, taking the data of each depth below the sea table as output, and training SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm to obtain a marine underwater data prediction model.
And a sea table data acquisition module 203 for satellite observation, which is used for acquiring sea table data for satellite observation.
And the data correction module 204 is used for correcting the sea surface salinity observed by the satellite by adopting a random forest regression model to obtain the corrected sea surface salinity.
A background field replacement module 205, configured to replace the monthly climate state background field of the corrected sea table data with a background field that changes day, so as to obtain the sea table data after the background field replacement; the corrected sea surface data includes corrected sea surface salinity.
And the ocean data obtaining module 206 for each depth is used for inputting the sea surface data after the background field replacement into the ocean underwater data prediction model to obtain the corresponding ocean data of each depth below the sea surface.
Data for each depth in the training set includes temperature or salinity in the world ocean map set, and temperature or salinity profile observations in the Argo Global Observation V3.0 dataset.
The sea surface data observed by the satellite comprise sea surface temperature data of a Reynolds OISST v2.1 data set, sea surface salinity data of a European space agency climate change initiative SSS v1.8 data set and absolute power topological data of a data unification and altimeter combined system.
The data correction module 204 specifically includes:
the random forest regression model training unit is used for training a random forest regression model by taking sea surface temperature data, sea surface salinity data, absolute power topology data, longitude and latitude observed by a satellite as input and corresponding offshore surface salinity as output to obtain a trained random forest regression model; offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 dataset.
And the data correction unit is used for inputting the sea surface temperature, the sea surface salinity, the absolute power topology, the longitude and the latitude observed by the satellite into the trained random forest regression model to obtain the corrected sea surface salinity.
The background field replacement module 205 specifically includes:
the multi-linear regression model training unit is used for training a multi-linear regression model by taking absolute power topological data observed by a satellite and offshore surface salinity, offshore surface temperature, longitude and latitude as input and salinity or temperature of each depth below a sea surface as output to obtain a trained multi-linear regression model; the offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 data set, and the offshore surface temperature is the temperature of the shallowest layer in the Argo global observation V3.0 data set.
And the daily change background field obtaining unit is used for inputting the sea surface salinity, the sea surface temperature, the absolute power topological data, the longitude and the latitude which are observed by the satellite and are corrected every day into the multiple linear regression model, and performing low-pass filtering on the output of the multiple linear regression model to obtain the corresponding daily change background field.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (10)

1. A surface layer accurate-rotation reconstruction method based on ocean measured data is characterized by comprising the following steps:
collecting data of various depths of the ocean as a training set;
training SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm by taking the data of the sea table in the training set as input and the data of each depth below the sea table as output to obtain an ocean underwater data prediction model;
obtaining sea table data observed by a satellite;
correcting sea surface salinity observed by the satellite by adopting a random forest regression model to obtain corrected sea surface salinity;
replacing the monthly climate state background field of the corrected sea surface data with a background field changing day to obtain sea surface data after the background field is replaced; the corrected sea surface data comprise corrected sea surface salinity;
and inputting the sea surface data after the background field replacement into the marine underwater data prediction model to obtain corresponding marine data of each depth below the sea surface.
2. The method for reconstructing surface layer accurate-conversion based on marine measured data according to claim 1, wherein the data at each depth in the training set comprises temperature or salinity in a world marine map set and temperature or salinity profile observation data in an Argo global observation V3.0 data set.
3. The method for reconstructing surface quasi-terrestrial transitions based on marine measured data according to claim 1, wherein the sea surface data of the satellite observation comprises sea surface temperature data of a Reynolds OISST v2.1 data set, sea surface salinity data of a European space agency climate change initiative SSS v1.8 data set and absolute power topology data of a data unification and altimeter combined system.
4. The method for reconstructing the surface layer quasi-inversion based on the marine measured data according to claim 2, wherein the step of correcting the sea surface salinity observed by the satellite by using the random forest regression model to obtain the corrected sea surface salinity specifically comprises the steps of:
training a random forest regression model by taking sea surface temperature data, sea surface salinity data, absolute power topology data, longitude and latitude observed by a satellite as input and corresponding offshore surface salinity as output to obtain a trained random forest regression model; the offshore surface salinity is the salinity of the shallowest layer in an Argo global observation V3.0 data set;
inputting the sea surface temperature, sea surface salinity, absolute power topology, longitude and latitude observed by the satellite into the trained random forest regression model to obtain the corrected sea surface salinity.
5. The method for reconstructing surface layer accurate-translation based on marine measured data according to claim 1, wherein the obtaining of the background field of daily change further comprises:
training a multiple linear regression model by taking absolute power topological data observed from a satellite, offshore surface salinity, offshore surface temperature, longitude and latitude as inputs and salinity or temperature of each depth below a sea surface as outputs to obtain the trained multiple linear regression model; the offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 data set, and the offshore surface temperature is the temperature of the shallowest layer in the Argo global observation V3.0 data set;
inputting the sea surface salinity, the sea surface temperature, the absolute dynamic topological data, the longitude and the latitude which are observed by the satellite and are corrected every day into a multiple linear regression model, and performing low-pass filtering on the output of the multiple linear regression model to obtain a background field corresponding to daily change.
6. A surface accurate-rotation reconstruction system based on marine measured data is characterized by comprising:
the training set data acquisition module is used for acquiring data of each depth of the ocean as a training set;
the marine underwater data prediction model obtaining module is used for taking the data of the sea table in the training set as input, taking the data of each depth below the sea table as output to train SQG algorithm, isQG algorithm and SQG-mEOF-R algorithm, and obtaining a marine underwater data prediction model;
the system comprises a satellite observation sea table data acquisition module, a satellite observation sea table data acquisition module and a satellite observation sea table data acquisition module, wherein the satellite observation sea table data acquisition module is used for acquiring the sea table data of the satellite observation;
the data correction module is used for correcting sea surface salinity observed by the satellite by adopting a random forest regression model to obtain corrected sea surface salinity;
the background field replacement module is used for replacing the monthly climate state background field of the corrected sea table data with a daily changing background field to obtain the sea table data after the background field replacement; the corrected sea surface data comprise corrected sea surface salinity;
and the ocean data acquisition module of each depth is used for inputting the sea surface data after the background field replacement into the ocean underwater data prediction model to acquire corresponding ocean data of each depth below the sea surface.
7. The system of claim 6, wherein the data at each depth in the training set comprises temperature or salinity in a world ocean map set and temperature or salinity profile observation in an Argo Global Observation V3.0 dataset.
8. The marine survey data based surface quasi-rotating reconstruction system of claim 6 wherein the satellite observed surface data includes surface temperature data of the Reynolds OISST v2.1 data set, surface salinity data of the European space agency climate change initiative SSS v1.8 data set and absolute power topology data of the combined data unification and altimeter system.
9. The system according to claim 7, wherein the data correction module specifically comprises:
the random forest regression model training unit is used for training a random forest regression model by taking sea surface temperature data, sea surface salinity data, absolute power topology data, longitude and latitude observed by a satellite as input and corresponding offshore surface salinity as output to obtain a trained random forest regression model; the offshore surface salinity is the salinity of the shallowest layer in an Argo global observation V3.0 data set;
and the data correction unit is used for inputting the sea surface temperature, the sea surface salinity, the absolute power topology, the longitude and the latitude observed by the satellite into the trained random forest regression model to obtain the corrected sea surface salinity.
10. The system of claim 6, wherein the ambient field replacement module further comprises:
the multi-linear regression model training unit is used for training a multi-linear regression model by taking absolute power topological data observed from a satellite, offshore surface salinity, offshore surface temperature, longitude and latitude as input and salinity or temperature of each depth below a sea surface as output to obtain a trained multi-linear regression model; the offshore surface salinity is the salinity of the shallowest layer in the Argo global observation V3.0 data set, and the offshore surface temperature is the temperature of the shallowest layer in the Argo global observation V3.0 data set;
and the daily change background field obtaining unit is used for inputting the sea surface salinity, the sea surface temperature, the absolute power topological data, the longitude and the latitude which are observed by the satellite and are corrected every day into the multiple linear regression model, and performing low-pass filtering on the output of the multiple linear regression model to obtain the corresponding daily change background field.
CN202111269295.9A 2021-10-29 2021-10-29 Surface layer accurate-rotation reconstruction method and system based on ocean measured data Active CN114238847B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111269295.9A CN114238847B (en) 2021-10-29 2021-10-29 Surface layer accurate-rotation reconstruction method and system based on ocean measured data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111269295.9A CN114238847B (en) 2021-10-29 2021-10-29 Surface layer accurate-rotation reconstruction method and system based on ocean measured data

Publications (2)

Publication Number Publication Date
CN114238847A true CN114238847A (en) 2022-03-25
CN114238847B CN114238847B (en) 2023-02-10

Family

ID=80743315

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111269295.9A Active CN114238847B (en) 2021-10-29 2021-10-29 Surface layer accurate-rotation reconstruction method and system based on ocean measured data

Country Status (1)

Country Link
CN (1) CN114238847B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114781576A (en) * 2022-04-19 2022-07-22 广东海洋大学 Sound velocity profile estimation method and device based on random forest algorithm
CN115238569A (en) * 2022-06-21 2022-10-25 中国人民解放军61540部队 Temperature and salt reconstruction method and system based on information embedding depth FFNN model
CN115438523A (en) * 2022-11-09 2022-12-06 中关村睿宸卫星创新应用研究院 Method, device, equipment and medium for reconstructing ocean three-dimensional chlorophyll concentration data
CN115758865A (en) * 2022-11-01 2023-03-07 天津大学 Underwater three-dimensional temperature and salinity reconstruction method and system based on deep learning
CN116306038A (en) * 2023-05-19 2023-06-23 北京弘象科技有限公司 Vertical layer matching method and device for meteorological mode and assimilation system
CN117173477A (en) * 2023-09-05 2023-12-05 北京交通大学 Domain generalization pedestrian re-identification method based on multi-layer data disturbance strategy

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103292792A (en) * 2013-04-26 2013-09-11 国家海洋局第二海洋研究所 Actual measurement SVP reconstruction method suitable for submarine detection and pseudo-landform processing
CN109543356A (en) * 2019-01-07 2019-03-29 福州大学 Consider the ocean interior temperature-salinity structure remote sensing inversion method of Space atmosphere
CN109543768A (en) * 2018-11-30 2019-03-29 福州大学 Ocean interior thermohaline information intelligent extracting method based on multi-source satellite remote sensing
CN110378858A (en) * 2019-07-04 2019-10-25 浙江大学 A kind of static Sea-Viewing Satellites data reconstruction method based on empirical orthogonal function decomposition method
CN110909491A (en) * 2019-12-11 2020-03-24 新亚优华科技有限公司 Sea surface salinity inversion algorithm based on wind and cloud meteorological satellite
KR102248963B1 (en) * 2020-06-23 2021-05-06 세종대학교산학협력단 Method and apparatus for predicting sea fog
CN113063737A (en) * 2021-03-26 2021-07-02 福州大学 Ocean heat content remote sensing inversion method combining remote sensing and buoy data

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103292792A (en) * 2013-04-26 2013-09-11 国家海洋局第二海洋研究所 Actual measurement SVP reconstruction method suitable for submarine detection and pseudo-landform processing
CN109543768A (en) * 2018-11-30 2019-03-29 福州大学 Ocean interior thermohaline information intelligent extracting method based on multi-source satellite remote sensing
CN109543356A (en) * 2019-01-07 2019-03-29 福州大学 Consider the ocean interior temperature-salinity structure remote sensing inversion method of Space atmosphere
CN110378858A (en) * 2019-07-04 2019-10-25 浙江大学 A kind of static Sea-Viewing Satellites data reconstruction method based on empirical orthogonal function decomposition method
CN110909491A (en) * 2019-12-11 2020-03-24 新亚优华科技有限公司 Sea surface salinity inversion algorithm based on wind and cloud meteorological satellite
KR102248963B1 (en) * 2020-06-23 2021-05-06 세종대학교산학협력단 Method and apparatus for predicting sea fog
CN113063737A (en) * 2021-03-26 2021-07-02 福州大学 Ocean heat content remote sensing inversion method combining remote sensing and buoy data

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114781576A (en) * 2022-04-19 2022-07-22 广东海洋大学 Sound velocity profile estimation method and device based on random forest algorithm
CN115238569A (en) * 2022-06-21 2022-10-25 中国人民解放军61540部队 Temperature and salt reconstruction method and system based on information embedding depth FFNN model
CN115238569B (en) * 2022-06-21 2023-04-28 中国人民解放军61540部队 Warm salt reconstruction method and system based on information embedding depth FFNN model
CN115758865A (en) * 2022-11-01 2023-03-07 天津大学 Underwater three-dimensional temperature and salinity reconstruction method and system based on deep learning
CN115758865B (en) * 2022-11-01 2023-11-14 天津大学 Underwater three-dimensional warm salt reconstruction method and system based on deep learning
CN115438523A (en) * 2022-11-09 2022-12-06 中关村睿宸卫星创新应用研究院 Method, device, equipment and medium for reconstructing ocean three-dimensional chlorophyll concentration data
CN115438523B (en) * 2022-11-09 2023-03-07 中关村睿宸卫星创新应用研究院 Method, device, equipment and medium for reconstructing marine three-dimensional chlorophyll concentration data
CN116306038A (en) * 2023-05-19 2023-06-23 北京弘象科技有限公司 Vertical layer matching method and device for meteorological mode and assimilation system
CN117173477A (en) * 2023-09-05 2023-12-05 北京交通大学 Domain generalization pedestrian re-identification method based on multi-layer data disturbance strategy

Also Published As

Publication number Publication date
CN114238847B (en) 2023-02-10

Similar Documents

Publication Publication Date Title
CN114238847B (en) Surface layer accurate-rotation reconstruction method and system based on ocean measured data
O’dea et al. An operational ocean forecast system incorporating NEMO and SST data assimilation for the tidally driven European North-West shelf
Valsala et al. Simulation and assimilation of global ocean pCO2 and air–sea CO2 fluxes using ship observations of surface ocean pCO2 in a simplified biogeochemical offline model
Cummings et al. Ocean data assimilation systems for GODAE
Zhang et al. An inflated ensemble filter for ocean data assimilation with a biased coupled GCM
Briggs et al. A glacial systems model configured for large ensemble analysis of Antarctic deglaciation
McKeague et al. Statistical inversion of South Atlantic circulation in an abyssal neutral density layer
Haines et al. An ocean modelling and assimilation guide to using GOCE geoid products
Zhang et al. A near-global eddy-resolving OGCM for climate studies
Losch et al. Adjoint sensitivity of an ocean general circulation model to bottom topography
Jayne Circulation of the North Atlantic Ocean from altimetry and the Gravity Recovery and Climate Experiment geoid
CN104567802A (en) Survey line land-sea elevation transfer method employing integrated shipborne gravity and GNSS
Hirose et al. Sequential forecasting of the surface and subsurface conditions in the Japan Sea
Nechaev et al. A diagnostic stabilized finite-element ocean circulation model
Sivareddy A study on global ocean analysis from an ocean data assimilation system and its sensitivity to observations and forcing fields
Zhang et al. Assimilation of water level data into a coastal hydrodynamic model by an adjoint optimal technique
Zalesny et al. Numerical algorithm of data assimilation based on splitting and adjoint equation methods
Lu et al. Impact of coherent ocean stratification on AMOC reconstruction by coupled data assimilation with a biased model
Brüning et al. Operational ocean forecasting for German coastal waters
Xing et al. Fusing Experimental Measurements and ROM Predicted Data for Wind Load Prediction With Extended Kalman Filter
Auroux et al. Data assimilation methods for an oceanographic problem
CN115238569B (en) Warm salt reconstruction method and system based on information embedding depth FFNN model
Chen et al. Physics‐informed deep operator learning based on reduced‐order modeling for retrieving the ocean interior density from the surface
Styles et al. The sensitivity of an idealized Weddell Gyre to horizontal resolution
Sarkisyan A view of progress in the numerical modeling of physical characteristics of the World Ocean in the light of sixty-year experience

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
CB03 Change of inventor or designer information
CB03 Change of inventor or designer information

Inventor after: Chen Jian

Inventor after: Yan Hengqian

Inventor after: Zhang Ren

Inventor after: Wang Huizan

Inventor before: Chen Jian

Inventor before: An Yuzhu

Inventor before: Zhang Weitao

Inventor before: Chen Xuan

Inventor before: Yan Hengqian

Inventor before: Bai Chengzu

Inventor before: Shen Xiaojing

GR01 Patent grant
GR01 Patent grant