Coastal acoustic tomography data assimilation method based on stream function fitting

Abstract

The invention discloses a coastal sound tomography data assimilation method based on stream function fitting. The method comprises the following steps: arranging a coastal sound tomography system, obtaining underwater sound data containing observation domain hydrological information, carrying out correlation peak extraction, and calculating to obtain a sound ray mutual return transmission time difference; scanning topographic information of an observation domain and temperature and profile data between stations, inputting the topographic information and the temperature and profile data into a sound velocity model for sound ray simulation, determining an actual propagation path of sound rays, and inverting path average flow; constructing a three-dimensional unstructured grid of the observation domain, and determining a projection relation between an actual propagation path of sound rays and the grid; establishing a state transition matrix by taking a sound ray path average flow as an observation vector of extended Kalman filtering and taking a coefficient matrix of a flow function least square fitting ocean numerical pattern calculation result as a state vector; and assimilating and updating by using an extended Kalman filtering algorithm, and obtaining an optimized flow field. According to the invention, the observation precision, the resolution and the spatial range of the coastal flow field can be obviously improved.

Description

Technical Field

[0001] The invention belongs to the technical field of ocean physical observation and data assimilation, and in particular relates to a coastal acoustic tomography data assimilation method based on stream function fitting. Background Art

[0002] The ocean has a profound impact on human society. The coastal areas are the most active areas for human activities, so understanding their environmental characteristics is of great significance for carrying out production operations. However, due to technical limitations, it is often difficult to obtain long-term, three-dimensional flow field data.

[0003] The Chinese patent document with publication number CN117805433A discloses a flow measurement device and method based on float pulse technology. Through a set of velocity measuring floats distributed throughout the channel section, the arrival time of the velocity measuring float terminal position is determined based on the pulse signal transmission positioning technology, and the velocity of the water flow in different layers and different positions is measured. However, this method can only measure the flow velocity of a vertical section, and the deployment is cumbersome and not portable.

[0004] The Chinese patent document with publication number CN116559492A discloses a coastal acoustic tomography flow measurement method and system. Several data collection points are placed in the coastal sea area, and the cross-correlation data is collected by mutual transmission and reception of sound lines between two points; the flow field between the two points is established through the collected basin data and the reciprocal transmission of sound waves and the acoustic Doppler effect, and the two-dimensional flow field of the entire sea area is established through the flow field data between two points. However, this method does not consider the boundary conditions during inversion, and its applicability is poor in areas with complex coastlines.

[0005] At present, although the existing acoustic tomography technology can obtain the flow field information in the deployment area through the inversion method, it still has the problems of limited observation scale, traditional inversion method and too single consideration factors. The flow field information obtained cannot meet the needs of high spatial resolution monitoring. Summary of the invention

[0006] The present invention provides a coastal acoustic tomography data assimilation method based on stream function fitting, which can improve the accuracy, range and resolution of acoustic tomography data inversion. The assimilated flow field can be applied to tidal dynamics research, marine environment monitoring and resource development.

[0007] A coastal acoustic tomography data assimilation method based on stream function fitting comprises the following steps:

[0008] (1) Deploy CAT systems at more than three stations to obtain raw data containing hydrological information within the observation area;

[0009] (2) Perform data correlation on the original data, extract the arrival peak, and obtain the reciprocal transmission time of the sound rays between stations;

[0010] (3) Using the collected information on topography, temperature, salinity, etc. between CAT stations, the sound line simulation is performed to obtain the reference propagation time of different sound lines between the stations, and the reference propagation time is matched with the actual transmission time of the sound line to determine the actual propagation path of the sound line and invert the average flow velocity of the path;

[0011] (4) construct a three-dimensional unstructured grid in the observation domain and determine the projection relationship between the actual propagation path of the sound line and the grid;

[0012] (5) Establish an ocean model for the observation area and surrounding sea areas, add boundary drive, and use the conical least squares method to calculate the coefficient matrix of the output flow field of the stream function fitting model;

[0013] (6) The path average velocity obtained by CAT inversion is used as the observation vector, and the coefficient matrix of the flow field of the stream function fitting model is used as the state vector. The state transfer matrix of the two is established and assimilated using the extended Kalman filter; the high-resolution flow field results after assimilation are output and the flow field is visualized.

[0014] The present invention establishes the relationship between acoustic tomography observation data and background field through stream function fitting, and adopts ensemble Kalman filtering algorithm to assimilate observation data and pattern output, so as to improve the accuracy and spatial resolution of flow field observation, solve the problems of sparse ocean observation data and large errors in the prior art, and provide a more accurate and efficient method for flow field analysis in complex nearshore waters.

[0015] In step (1), the CAT system includes a buoy, a data acquisition system, an acoustic transducer, a GPS positioning system, a temperature-salinity-depth profiler, a buoy and a positioning anchor; the buoy is located on the water surface, the data acquisition system is located inside the buoy and connected to the acoustic transducer; the acoustic transducer is arranged at different depths underwater, one end of which is connected to the buoy and the other end is anchored to the bottom of the water.

[0016] The CAT system transmits a high-frequency carrier signal modulated by an M sequence through an acoustic transducer and adopts a simultaneous transmission and reception mode to ensure the acquisition of bidirectional reciprocal transmission sound lines.

[0017] In step (3), the topographic data of the observation area are obtained by cruise mapping using an acoustic Doppler current profiler; the reference propagation time of different sound lines between the stations is obtained by a sound line simulation program in combination with the GPS positioning of the station, the temperature profile and the water depth of the transducer deployment at each station.

[0018] In step (3), the reference propagation time of different sound lines between each station is simulated and matched with the actual transmission time of the sound line to determine the actual propagation path of the sound line and invert the average flow velocity of the path. The specific process is as follows:

[0019] (a-1) Calculate the reference propagation time of the sound line between two stations using the sound line model formula:

[0020]

[0021] Among them, t + and t - are the forward and reverse sound ray reference propagation times, C 0 is the reference sound velocity, ΔC is the reference sound velocity deviation, V c is the flow velocity along the direction of sound line propagation, Γ is the propagation path of the sound line between the two stations, and ds is the arc length of the sound line.

[0022] (a-2) Due to C 0 Much larger than ΔC and V c , expand the above formula using Taylor series, and discard the higher-order terms above the second order, and get the approximate solution of the reciprocal propagation time of the sound rays between the two stations:

[0023]

[0024] (a-3) Calculate the reciprocal propagation time difference of the sound rays between the two stations and derive its relationship with the average flow velocity of the path:

[0025]

[0026] Let V h is the velocity along the horizontal propagation direction, dl is the infinitesimal distance of the sound line horizontal propagation, and from the geometric relationship, we can get:

[0027]

[0028] in, is the angle between the microelement of the sound line arc and the horizontal direction. Therefore, the relationship between the reciprocal propagation time difference of the sound line between the two stations and the average flow velocity of the path is:

[0029]

[0030] Wherein, L is the horizontal propagation path of the sound.

[0031] (a-4) Calculate the average flow along the sound propagation path:

[0032]

[0033] The specific process of step (4) is as follows:

[0034] (b-1) The horizontal grid of the study area was drawn using SMS software. The Delauney partitioning method was used in the horizontal direction and the uniform layering method was used in the vertical direction. The water depth data was interpolated to the grid nodes using the inverse distance weighted method to establish a three-dimensional unstructured grid model.

[0035] (b-2) Determine the relationship between the path average flow and the mode state in the horizontal direction:

[0036] Projecting the actual sound line obtained in step (3) along the horizontal direction, it is easy to know that the average flow velocity of the propagation path of the sound line r in (a-4) can be represented by the combination of the mode states in the grid it passes through in the horizontal direction:

[0037]

[0038] Among them, u r 、v r are the east and north components of the path average velocity, θ r is the angle between the horizontal path of the sound line and the east direction, N is the total number of horizontal grids in the CAT deployment area, l r_n is the length of the sound line r in the nth grid, L r is the total length of the sound line r in the horizontal direction, u n 、v n Output the east and north components of the model velocity in the nth grid.

[0039] Therefore, for the R sound lines extracted in step (3), we have:

[0040]

[0041] In short, it can be written in matrix form:

[0042] V=E h Ψ z

[0043] Among them, E h ,Ψ z They represent the flow velocity matrix output by the horizontal projection operator and the z-th layer of the pattern respectively.

[0044]

[0045] (b-3) Determine the three-dimensional projection relationship between the path average flow velocity and the mode state:

[0046] Considering that the velocity and distribution of each layer are different, the path average velocity is further extended to the weighted combination of the velocity components of the vertical layers passed by the sound line. The length h of each layer passed by the calibration sound line in the vertical section r_z , set the scale factor of each layer to:

[0047]

[0048] Since the vertical distribution of sound rays has no effect on the horizontal projection, the horizontal projection operator E of the rth sound ray is h_rThe proportional coefficient of each vertical layer is brought into the horizontal projection to obtain the three-dimensional projection relationship between the mean flow of the path and the mode state:

[0049]

[0050] In short, it can be written in matrix form:

[0051] V=DΨ

[0052] Where D and Ψ represent the three-dimensional projection operator and the set of flow velocity matrices output by each layer of the model, respectively.

[0053]

[0054] In step (5), the ocean model uses an unstructured grid finite volume ocean model (FVCOM), the unstructured grid comes from the three-dimensional grid created in step (4), and the open boundary tide driving information comes from Tpxo9-atlas-v5, which is generated using TMD software forecasting.

[0055] The coefficient matrix of the flow field output by the stream function fitting mode is calculated using the conical least squares method. The specific process is as follows:

[0056] (c-1) Expand the stream function into a two-variable Taylor series form:

[0057]

[0058] Among them, a pq is the coefficient to be determined, x and y are the coordinates of the points in the region, R T Truncation error

[0059] (c-2) List the equation for the output flow rate of the stream function fitting model:

[0060] According to the properties of the stream function:

[0061]

[0062] Then, in each layer of the model output, the Taylor series expressions of the east and north components of the velocity are:

[0063]

[0064] In short, it can be written in matrix form:

[0065] Ψ=EA+N

[0066] Among them, Ψ, N, E, and A represent the velocity matrix, error noise matrix, position coordinate matrix, and fitting coefficient matrix of the z-th layer output of the model respectively:

[0067]

[0068] A=[a 00 a 01 …a pq …a T0 ]

[0069] Then the equation for the output flow velocity of all layers in the stream function fitting model is:

[0070] Ψ=EA+N

[0071] Among them, Ψ, N, E, and A represent the flow velocity matrix, error noise matrix, position coordinate matrix, and the set of fitting coefficient matrices to be calculated for each layer of the model output respectively:

[0072]

[0073] E=diag(E,E,…,E)

[0074] (c-3) Use the tapered least squares method to find the optimal solution for the fitting coefficient matrix:

[0075] Define the loss function:

[0076] J=N T N+α 2 A T A=(Ψ-EA) T (Ψ-EA)+α 2 A T A

[0077] Among them, α is a constant to be determined. Perform singular value decomposition on E, extract the main components, and further determine the value of α through the L-curve method. The expected optimal solution of A that minimizes the loss function can be obtained

[0078] The specific process of step (6) is as follows:

[0079] (d-1) The coefficient matrix of the flow field of the stream function fitting model in step (5) has a dimension M = 2N·Z. Perturb the model boundary and generate k different sets as the state vector X:

[0080]

[0081] Calculate the state error covariance matrix P e :

[0082]

[0083] (d-2) Add k disturbances to the average velocity of the R sound lines obtained by inverting the CAT data in step (3) as the observation vector Y:

[0084]

[0085] η=(σ 1 ,σ 2 ,……,σ k )

[0086] Calculate the observation error covariance matrix R e :

[0087]

[0088] (d-3) Establish the state transfer matrix:

[0089] V=DΨ=DEA=HA

[0090] H=DE

[0091] (d-4) Use the ensemble Kalman filter method to update the state error covariance matrix:

[0092] X a =X+K(Y-HX)

[0093]

[0094] Where K is the Kalman gain:

[0095] K=P e H T [HP e H T +R e ] -1

[0096] Set a judgment threshold. If the inversion error exceeds the set threshold, return to step (d-4) for iterative calculation until the error meets the requirements.

[0097] (d-5) Input the model to predict the state at the next moment:

[0098]

[0099] in, They represent the nonlinear calculation process and calculation error of the ocean model respectively.

[0100] Compared with the prior art, the present invention has the following beneficial effects:

[0101] 1. Improve accuracy. By combining stream function fitting and ensemble Kalman filtering, the simulation error of the ocean model is significantly reduced;

[0102] 2. Expanded scope. Effectively solve the sparsity problem of CAT observation data and expand the observation scope from the original deployment range to the grid coverage range;

[0103] 3. High resolution. Compared with traditional inversion, it can be used for refined observation of coastal tides, internal waves and other small and medium-scale dynamic processes.

[0104] 4. Strong applicability. The present invention is suitable for flow field analysis under complex terrain and sparse observation conditions, and provides an efficient and low-cost solution for tidal dynamics research and marine environment monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0106] Figure 1 A schematic flow chart of a coastal acoustic tomography data assimilation method based on stream function fitting provided by the present invention;

[0107] Figure 2 This is the horizontal grid distribution map of the FVCOM ocean model;

[0108] Figure 3 This is a comparison of the flow field before and after assimilation at 9:00 on June 2, 2016. DETAILED DESCRIPTION

[0109] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0110] like Figure 1 As shown, a coastal acoustic tomography data assimilation method based on stream function fitting includes: deploying a coastal acoustic tomography system, acquiring hydroacoustic data containing hydrological information of the observation domain, extracting correlation peaks, and calculating the time difference of mutual return transmission of sound lines; scanning the terrain information of the observation domain and the temperature and profile data between stations, inputting the sound velocity model to simulate the sound line, determining the actual propagation path of the sound line, and inverting the path average flow; constructing a three-dimensional unstructured grid in the observation domain, and determining the projection relationship between the actual propagation path of the sound line and the grid; using the average flow of the sound line path as the observation vector of the extended Kalman filter, using the coefficient matrix of the calculation result of the stream function least squares fitting ocean numerical model as the state vector, and establishing the state transfer matrix; using the extended Kalman filter algorithm to assimilate and update and obtain the optimized flow field.

[0111] The present invention is further described in detail below by taking the flow field assimilation of the Bali Strait as an example.

[0112] Step 1: Collect CAT experimental data in the Bali Strait of Indonesia. Set up 4 sets of CAT systems with a transducer frequency of 10kHz and place them on the east and west coasts of the Bali Strait respectively. Three transducers (N2, N3, N4) are fixed at a depth of 5m from the seabed, and one transducer (N1) is fixed in the middle layer of the water depth, 4m from the seabed. The transmitting signal adopts M10Q3R3, and each station sends it in turn at an interval of 1 minute. The two-way sound line propagation time between the four stations is obtained.

[0113] Step 2: The acquired CAT raw observation data is processed by 30-minute time averaging to improve the signal-to-noise ratio and reduce high-frequency tidal disturbances. The pre-processed signal is processed by orthogonal correlation, and the moving window peak search algorithm is used to find the sound line arrival peak and record the sound line transmission time.

[0114] Among them, the signal-to-noise ratio of the signal received by station N2 is relatively low, and the signal-to-noise ratio of its receiving peak is about 45dB, which is greater than the signal-to-noise ratio of about 30dB at other locations. At 10:00, due to the multipath effect of the sound line caused by surface reflection and bottom reflection, station N3 receives multiple peaks from station N2, and the first arrival peak has a higher signal-to-noise ratio because it passes through fewer bottom reflections and surface reflections.

[0115] Step 3: Use the terrain information between stations obtained by the ship-borne ADCP cruise scan and the reference sound speed and temperature profiles obtained by CTD acquisition to simulate the sound line and obtain all possible sound line paths. By comparing the simulation results with the transmission time recorded in step 2, the actual sound line propagation path is determined.

[0116] Step 4: Use the unstructured grid finite volume ocean model (FVCOM) to output the three-dimensional flow field results. The horizontal grid coastline data comes from the GSHHG dataset of the National Oceanic and Atmospheric Administration of the United States, including 9371 nodes and 18116 unstructured grids, such as Figure 2 As shown; the vertical direction uses bottom sigma coordinates and is evenly divided into 3 layers. The water depth data comes from the world ocean depth map GEBCO.

[0117] In the output results, the highest tide level is 0.73m during high tide and the lowest tide level is -1.31m during low tide; the maximum flow velocity is 2.16m / s and the maximum average flow velocity is 0.33m / s. The model results are numerically consistent with the previous observation results, but there are still some errors, which need data assimilation correction.

[0118] After the model calibration is completed, the least cone square method is used to fit the stream function. By comparison, it can be seen that due to the reduction of data dimension, some details are lost, but the main flow field information is retained, which is convenient for exploring and describing the overall flow field in the area.

[0119] Ψ=EA+N

[0120]

[0121] E=diag(E,E,E)

[0122]

[0123] Step 5: Establish the average flow field data of the acoustic tomography path extracted in step 3 and the state transfer matrix in step 4. Due to the land obstruction between N1-N2 and N3-N4, only the data can be received between the four station lines. The coverage area has 60 grids, and the angles between the lines of N1-N3, N1-N4, N2-N3, and N3-N4 and the east direction are: θ 1 =-2°,θ 2 =-31°,θ 3 =-45°,θ 4 =-51°, a total of 6 sound lines were observed.

[0124] The horizontal projection operator is:

[0125]

[0126] Sound lines with the same horizontal projection but passing through different vertical layers, i.e., multipath sound lines arriving successively between two stations, have the same angle with the east direction and the same horizontal projection operator.

[0127] The three-dimensional projection operator is:

[0128]

[0129] The state transfer matrix is:

[0130] H=DE

[0131] Step 6: Use the ensemble Kalman filter algorithm to assimilate the disturbed CAT observation data set and the disturbed stream function fitting coefficient data set:

[0132] X a =X+K(Y-HX)

[0133]

[0134] K=P e H T [HP e H T +R e ] -1

[0135]

[0136] The assimilation results are compared with CAT data and tide observation data, such as Figure 3 As shown in the figure, it is found that the assimilated flow field can more accurately and truly describe the ebb and flow conditions of the Bali Strait than the flow field directly fitted with the stream function of FVCOM. The assimilated result reduces the error of the background field by more than 30%.

[0137] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A coastal acoustic tomography data assimilation method based on stream function fitting, characterized in that: The following steps are involved: (1) Deploy CAT systems at more than three stations to obtain raw data containing hydrological information within the observation area; (2) Perform data correlation on the original data, extract the arrival peak, and obtain the reciprocal transmission time of the sound rays between stations; (3) Using the terrain, temperature, and salinity information collected between CAT stations, the sound line simulation is performed to obtain the reference propagation time of different sound lines between the stations, and the reference propagation time is matched with the actual transmission time of the sound line to determine the actual propagation path of the sound line and invert the average flow velocity of the path; (4) construct a three-dimensional unstructured grid in the observation domain and determine the projection relationship between the actual propagation path of the sound line and the grid; (5) Establish an ocean model for the observation area and surrounding sea areas, add boundary drive, and use the conical least squares method to calculate the coefficient matrix of the output flow field of the stream function fitting model; (6) The path average velocity obtained by CAT inversion is used as the observation vector, and the coefficient matrix of the stream function fitting model flow field is used as the state vector. The state transfer matrix of the two is established and assimilated using the extended Kalman filter; Output the assimilated high-resolution flow field results and perform flow field visualization.

2. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: In step (1), the CAT system includes a buoy, a data acquisition system, an acoustic transducer, a GPS positioning system, a temperature-salinity-depth profiler, a buoy and a positioning anchor; the buoy is located on the water surface, the data acquisition system is located inside the buoy and connected to the acoustic transducer; the acoustic transducer is arranged at different depths underwater, one end of which is connected to the buoy and the other end is anchored to the bottom of the water.

3. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: In step (1), the CAT system transmits a high-frequency carrier signal modulated by an M sequence through an acoustic transducer, and adopts a simultaneous transmission and reception mode to ensure the acquisition of bidirectional mutual return transmission sound lines.

4. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: In step (3), the topographic data of the observation area are obtained by cruise mapping using an acoustic Doppler current profiler; the reference propagation time of different sound lines between the stations is obtained by a sound line simulation program in combination with the GPS positioning of the station, the temperature profile and the water depth of the transducer deployment at each station.

5. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: In step (3), the reference propagation time of different sound lines between each station is simulated and matched with the actual transmission time of the sound line to determine the actual propagation path of the sound line and invert the average flow velocity of the path. The specific process is as follows: (a-1) Calculate the reference propagation time of the sound line between two stations using the sound line model formula: Among them, t + and t - They represent the forward and reverse sound propagation time, C0 is the reference sound velocity, ΔC is the reference sound velocity deviation, V c is the flow velocity along the direction of sound line propagation, Γ is the propagation path of the sound line between the two stations, and ds is the arc length of the sound line; (a-2) Since C0 is much larger than ΔC and V c , expand the above formula using Taylor series, and discard the higher-order terms above the second order, and get the approximate solution of the reciprocal propagation time of the sound rays between the two stations: (a-3) Calculate the reciprocal propagation time difference of the sound rays between the two stations and derive its relationship with the average flow velocity of the path: Let V h is the velocity along the horizontal propagation direction, dl is the infinitesimal distance of the sound line horizontal propagation, and from the geometric relationship, we can get: in, is the angle between the microelement of the sound line arc segment and the horizontal direction; therefore, the relationship between the reciprocal propagation time difference of the sound line between the two stations and the average flow velocity of the path is: Wherein, L is the horizontal propagation path of the sound line; (a-4) Calculate the average flow along the sound propagation path:

6. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: The specific process of step (4) is as follows: (b-1) The horizontal grid of the study area was drawn using SMS software, the Delauney partitioning method was used in the horizontal direction, the uniform layering method was used in the vertical direction, and the water depth data was interpolated to the grid nodes using the inverse distance weighted method to establish a three-dimensional unstructured grid model; (b-2) Determine the relationship between the path average flow and the mode state in the horizontal direction: Projecting the actual sound line obtained in step (3) along the horizontal direction, it is easy to know that the average flow velocity of the propagation path of the sound line r in (a-4) is represented by the combination of the mode states in the grid it passes through in the horizontal direction: Among them, u r 、v r are the east and north components of the path average velocity, θ r is the angle between the horizontal path of the sound line and the east direction, N is the total number of horizontal grids in the CAT layout area, l r_n is the length of the sound line r in the nth grid, L r is the total length of the sound line r in the horizontal direction, u n 、v n are the east and north components of the model output velocity in the nth grid; Therefore, for the R sound lines extracted in step (3), we have: In short, it can be written in matrix form: V=E h Ψ z Among them, E h ,Ψ z They represent the flow velocity matrix output by the horizontal projection operator and the z-th layer of the model respectively; (b-3) Determine the three-dimensional projection relationship between the path average flow velocity and the mode state: Considering that the velocity size and distribution of each layer are different, the path average velocity is further extended to the weighted combination form of the velocity components of the vertical layer passed by the sound line; the length h of each layer passed by the sound line in the vertical section is calibrated r_z , set the scale factor of each layer to: Since the vertical distribution of sound rays has no effect on the horizontal projection, the horizontal projection operator E of the rth sound ray is h_r unchanged; bring the scale factor of each vertical layer into the horizontal projection to obtain the three-dimensional projection relationship between the mean flow of the path and the mode state: In short, it can be written in matrix form: V=DΨ Where D and Ψ represent the set of flow velocity matrices output by the three-dimensional projection operator and each layer of the model, respectively; 7. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: In step (5), the ocean model uses an unstructured grid finite volume ocean model, the unstructured grid comes from the three-dimensional grid created in step (4), the open boundary tide driving information comes from Tpxo9-atlas-v5, and is generated using TMD software forecasting.

8. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: In step (5), the coefficient matrix of the flow field output by the stream function fitting mode is calculated using the conical least squares method. The specific process is as follows: (c-1) Expand the stream function into a two-variable Taylor series form: Among them, a pq is the coefficient to be determined, x and y are the coordinates of the points in the region, R T is the truncation error; (c-2) List the equation for the output flow rate of the stream function fitting model: According to the properties of the stream function: Then, in each layer of the model output, the Taylor series expressions of the east and north components of the velocity are: In short, it can be written in matrix form: Ψ=EA+N Among them, Ψ, N, E, and A represent the velocity matrix, error noise matrix, position coordinate matrix, and fitting coefficient matrix of the z-th layer output of the model respectively: A=[a 00 a 01 … a pq … a T0 ] Then the equation for the output flow velocity of all layers in the stream function fitting model is: Ψ=EA+N Among them, ψ, N, E, A represent the flow velocity matrix, error noise matrix, position coordinate matrix and the set of fitting coefficient matrix to be calculated output by each layer of the model respectively: E=diag(E,E,…,E) (c-3) Use the tapered least squares method to find the optimal solution for the fitting coefficient matrix: Define the loss function: J=N T N+α 2 A T A=(Ψ-EA) T (Ψ-EA)+α 2 A T A Among them, α is a constant to be determined. Perform singular value decomposition on E, extract the main components, and further determine the value of α through the L-curve method; obtain the expected optimal solution of A that minimizes the loss function 9. The coastal acoustic tomography data assimilation method based on stream function fitting according to claim 1 is characterized in that: The specific process of step (6) is as follows: (d-1) The coefficient matrix of the flow field of the stream function fitting model in step (5) has a dimension M = 2N·Z; perturb the model boundary and generate k different sets as the state vector X: Calculate the state error covariance matrix P e : (d-2) Add k disturbances to the average velocity of the R sound lines obtained by inverting the CAT data in step (3) as the observation vector Y: η=(σ1,σ2,……,σ k ) Calculate the observation error covariance matrix R e : (d-3) Establish the state transfer matrix: V=DΨ=DEA=HA H=DE (d-4) Use the ensemble Kalman filter method to update the state error covariance matrix: X a =X+K(Y-HX) Where K is the Kalman gain: K=P e H T [HP e H T +R e ] -1 Set a judgment threshold. If the inversion error exceeds the set threshold, return to step (d-4) to perform iterative calculation until the error meets the requirement; (d-5) Input the model to predict the state at the next moment: in, ω represents the nonlinear calculation process and calculation error of the ocean model respectively.

Images