Radar altimeter sea condition deviation value estimation method, device, equipment, medium and product

Abstract

The invention discloses a method, a device, equipment, a medium and a product for estimating a sea condition deviation value of a radar altimeter, and relates to the field of ocean observation, and the method comprises the steps: screening L2-level GDR data of the radar altimeter based on a data identification bit and a sea surface parameter editing criterion of the radar altimeter; based on a threshold screening condition, constructing an SNSSB model according to the self-intersection point data set; according to different sea surface parameter combinations, training the SNSSB model, and determining a trained SNSSB model; and estimating a sea condition deviation value according to the trained SNSSB model, and according to the method, the sea condition deviation value of the radar altimeter can be accurately evaluated.

Description

Technical Field

[0001] The present application relates to the field of ocean observation, and particularly to a method, device, equipment, medium and product for estimating the sea state bias value of a radar altimeter. Background Art

[0002] The severe environmental challenges currently faced by humanity are mainly climate change. To assess the impact of climate change, especially the part caused by human activities, it is necessary to continuously, long-term and accurately monitor all hydrological, oceanic, atmospheric and geophysical changes caused by climate change globally. The global average sea level is one of the most important climate change indicators because it directly threatens the lives of residents living in coastal areas and around small islands (such as low-lying areas with an altitude below 10m).

[0003] The global monitoring and assessment of sea level rise are mainly carried out through satellite altimetry. A series of satellite altimeter missions led by the National Aeronautics and Space Administration (NASA) of the United States and the French Space Agency (CNES) have continuously measured the ocean height, thus generating a continuous time series of sea surface height data. In this process, other partners, such as the National Oceanic and Atmospheric Administration (NOAA) of the United States, the European Space Agency (ESA) and the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT), are also involved.

[0004] In previous studies, satellite orbit determination errors have always been the main error sources of altimetry data. However, in recent years, with the development of precise orbit determination technology, significant progress has been made in improving the orbit quality of altimetry satellites. The radial orbit error of altimetry satellites has been improved from 48 cm of the Seasat satellite launched in 1985 to the 1 cm level of contemporary altimetry missions in 2020. Existing research shows that the uncertainty of sea state bias is very large, reaching 2 cm. Therefore, sea state bias has replaced orbit error as one of the most important error sources in radar altimeter altimetry. Summary of the Invention

[0005] The purpose of the present application is to provide a method, device, equipment, medium and product for estimating the sea state bias value of a radar altimeter, which can accurately evaluate the sea state bias value of the radar altimeter.

[0006] To achieve the above purpose, the present application provides the following solutions:

[0007] In the first aspect, the present application provides a method for estimating the sea state bias value of a radar altimeter, including:

[0008] Based on the data identification bits of the radar altimeter and the sea surface parameter editing criteria, filter the L2-level GDR data of the radar altimeter; the data flag bits include the ground type at the sub-satellite point in the radar altimeter and the current operating state of the radar altimeter; the sea surface parameters include sea surface wind speed, significant wave height, mean wave period, backscattering coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, RMS of Ku-band backscattering coefficient, and RMS of Ku-band significant wave height, etc.;

[0009] Establish a self-crossing point data set according to the filtered L2-level GDR data;

[0010] Based on the threshold screening conditions, construct the SNSSB model according to the self-crossing point data set; the SNSSB model includes two independent and identically structured BP neural networks; each BP neural network includes an input layer, three hidden layers, and an output layer; the transfer relationship between adjacent layers is backpropagation; the threshold screening conditions include latitude limit, water depth limit, time limit, and the mismatch value limit of the sea surface height at the crossing point;

[0011] Train the SNSSB model according to different combinations of sea surface parameters to determine the trained SNSSB model;

[0012] Estimate the sea condition deviation value according to the trained SNSSB model.

[0013] In a second aspect, the present application provides a device for estimating the sea condition deviation value of a radar altimeter, including:

[0014] A screening module, configured to filter the L2-level GDR data of the radar altimeter based on the data identification bits of the radar altimeter and the sea surface parameter editing criteria; the data flag bits include the ground type at the sub-satellite point in the radar altimeter and the current operating state of the radar altimeter; the sea surface parameters include sea surface wind speed, significant wave height, mean wave period, backscattering coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, RMS of Ku-band backscattering coefficient, and RMS of Ku-band significant wave height, etc.;

[0015] A self-crossing point data set establishment module, configured to establish a self-crossing point data set according to the filtered L2-level GDR data;

[0016] An SNSSB model construction module, configured to construct the SNSSB model according to the self-crossing point data set based on the threshold screening conditions; the SNSSB model includes two independent and identically structured BP neural networks; each BP neural network includes an input layer, three hidden layers, and an output layer; the transfer relationship between adjacent layers is backpropagation; the threshold screening conditions include latitude limit, water depth limit, time limit, and the mismatch value limit of the sea surface height at the crossing point;

[0017] A training module, configured to train the SNSSB model according to different combinations of sea surface parameters and determine the trained SNSSB model;

[0018] A sea condition deviation value estimation module, configured to estimate the sea condition deviation value according to the trained SNSSB model.

[0019] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the radar altimeter sea condition deviation value estimation method described in any one of the above.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the radar altimeter sea condition deviation value estimation method described in any one of the above is implemented.

[0021] In a fifth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the radar altimeter sea condition deviation value estimation method described in any one of the above is implemented.

[0022] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0023] Based on the sea surface parameters related to the sea condition deviation value in the radar altimeter product parameters, the present application deeply explores the relationship between the above sea surface parameters and the sea condition deviation value, and conducts accuracy verification through the explained variance at the intersection point and the sea surface height variance difference index. The results prove that when the sea surface parameters are the same, the accuracy of the traditional 2D non-parametric sea state bias model (NPSSB) is lower than that of the 2D siamese network-based sea state bias model (SNSSB). And as the input dimension of the relevant sea surface parameters continues to increase, the accuracy of the SNSSB model becomes higher and higher. At the same time, the computational complexity of the traditional method for constructing the sea condition deviation model will increase significantly after increasing the sea surface parameter dimension, making it difficult to simply increase the dimension of the sea surface parameters. Therefore, the present application is also superior to the traditional method in terms of simplicity.

[0024] Based on the threshold screening conditions, this application constructs an SNSSB model according to the self-crossing point dataset, and trains the SNSSB model according to different combinations of sea surface parameters to obtain a more accurate estimated sea state deviation value, so as to improve the accuracy of the observed sea surface height. This is of great significance not only for monitoring climate change, but also for subsequent inversion of the Earth's gravity field and so on. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0026] Figure 1 It is a flowchart of a method for estimating the sea state deviation value of a radar altimeter provided by the present application;

[0027] Figure 2 It is a flowchart of another method for estimating the sea state deviation value of a radar altimeter provided by the present application;

[0028] Figure 3 It is a schematic diagram of solving the intersection point position provided by the present application;

[0029] Figure 4 It is a result diagram of the evaluation index (explained variance) provided by the present application;

[0030] Figure 5 It is a result diagram of the evaluation index (sea surface height variance difference index) provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0032] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0033] The embodiment of the present application provides a method for estimating the sea state deviation value of a radar altimeter. This method is executed by a computer device, which can be specifically executed by a computer device such as a terminal or a server alone, or jointly executed by a terminal and a server. In the embodiment of the present application, as Figure 1 shown, the method includes the following steps.

[0034] S1: Based on the data identification bits of the radar altimeter and the sea surface parameter editing criteria, screen the L2-level Geophysical Data Record (GDR data) of the radar altimeter; the data flag bits include the ground type at the sub-satellite point in the radar altimeter and the current operating state of the radar altimeter; the sea surface parameters include sea surface wind speed, significant wave height, mean wave period, backscatter coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, Root Mean Square (RMS) of the Ku-band backscatter coefficient, and RMS of the Ku-band significant wave height, etc.

[0035] S2: Establish a self-crossing point data set according to the screened L2-level GDR data; the self-crossing point data set includes: the latitude and longitude, wind speed, significant wave height, RMS of the significant wave height, RMS of the backscatter coefficient, dry atmospheric correction value, wet atmospheric correction value, sea surface height, and water depth, etc. of the data included in the altimeter products at the crossing points.

[0036] S3: Based on the threshold screening conditions, construct the SNSSB model according to the self-crossing point data set; the SNSSB model includes two independent and identically structured BP neural networks; each BP neural network includes an input layer, three hidden layers, and an output layer; the transfer relationship between adjacent layers is backpropagation; the threshold screening conditions include latitude limit, water depth limit, time limit, and the sea surface height mismatch value limit at the crossing points. In practical applications, backpropagation is used to update the parameters in the hidden layer.

[0037] S4: Train the SNSSB model according to different combinations of sea surface parameters to determine the trained SNSSB model.

[0038] S5: Estimate the sea condition deviation value according to the trained SNSSB model.

[0039] In an exemplary embodiment, the data identification bit in S1 refers to information such as the ground type at the sub-satellite point and the current operating state of the radar altimeter in the radar altimeter product. For example, data that cannot be used subsequently, such as the ground type being non-ocean and the operating state of the altimeter being abnormal, can be eliminated through this operation. During the observation process of the radar altimeter, due to the complex movement of the sea surface or abnormal fluctuations in its own state, some observed values at the sub-satellite point may be incorrect. Therefore, after screening according to the data identification bit, the remaining data needs to be further screened according to the threshold requirements. For example, the threshold range for dry tropospheric atmospheric correction is from -2500 mm to -1900 mm, and the threshold range for wet tropospheric atmospheric correction is from -500 mm to -1 mm. Any data outside the normal threshold range needs to be eliminated, that is, the entire trajectory is eliminated.

[0040] In practical applications, the sea surface parameter editing criterion in S1 is to judge whether each sea surface parameter value is normal according to the parameter threshold. That is, it is judged whether the parameter values of sea surface parameters including sea surface wind speed, significant wave height, mean wave period, backscattering coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, RMS of Ku-band backscattering coefficient, and RMS of Ku-band significant wave height are within a reasonable range.

[0041] In an exemplary embodiment, S2 can be replaced by the following steps.

[0042] S21: According to the filtered L2-level GDR data, determine the intersection point position of the radar altimeter through a double-loop operation.

[0043] S22: Based on the intersection point position, use the radial basis function interpolation method to determine each sea surface parameter at the intersection point.

[0044] S23: According to the sea surface parameters, establish a self-intersection point data set.

[0045] In an exemplary embodiment, S21 can be replaced by the following steps.

[0046] S211: Use the fast rejection method to determine whether there is an intersection between the closed rectangles formed by the ascending and descending orbits as diagonals. If so, execute S212; if not, execute S214.

[0047] S212: Use the straddle experiment method to judge whether there is an intersection relationship between the ascending and descending orbits. If so, execute S213; if not, execute S214.

[0048] S213: Determine the intersection point position between the ascending and descending orbits of the radar altimeter.

[0049] S214: Enter the next loop, return to S211, and start over to determine whether there is an intersection between the two new ascending and descending trajectories.

[0050] In practical applications, the fast rejection method: Suppose AB and CD are the diagonals of a rectangular closed area, and E and F represent the corresponding rectangular closed areas. If there is an intersection between AB and CD, then it will definitely result in the appearance of a rectangular closed area G.

[0051] Straddle experiment: Use the method of vector cross product to determine whether there is an intersection relationship between two line segments. Vector cross product contains a theorem: For vector A cross product vector B, if the result is less than 0, it means that vector B is in the clockwise direction of vector A; if the result is greater than 0, it means that vector B is in the counterclockwise direction of vector A; if the result is equal to 0, it means that vector B and vector A are parallel. Generally speaking, first determine whether there is an intersection point between the two trajectories through fast rejection, then verify the straddle relationship between them through the straddle experiment, and finally the position of the intersection point can be obtained through interpolation. Among them, vector A and vector B represent the satellite trajectories that meet the data requirements after screening. The longitude and latitude of the self-intersection point are determined through fast rejection and straddle experiment, and then the GDR data is interpolated to the self-intersection point according to the longitude and latitude.

[0052] In practical applications, after obtaining the accurate position information of the intersection point in S22, since the intersection point is almost never the sub-satellite point, there will be no various information. Therefore, the required observation items need to be interpolated to the obtained intersection point. The interpolation method is selected as radial basis function interpolation. Its principle is to use a combination of a group of basis functions to interpolate a group of known data points. Among them, the known data points represent the data of four points before and after the self-intersection point respectively, and each point contains the above-mentioned sea surface parameters. The sea surface parameters are interpolated separately, and the basis function for interpolating the known data points is a Gaussian function, and the expression is:

[0053]

[0054] Among them, φ(x) is the interpolated value of each sea surface parameter, x is the known sea surface parameter value, c is the center of the kernel function, and σ is the width parameter of the function.

[0055] In an exemplary embodiment, S3 can be replaced by the following steps.

[0056] S31: Use the required sea surface parameters and the sea surface height mismatch value at the intersection point in the self-intersection point data as the input value and the label value. Among them, the required sea surface parameters are the sea surface parameters related to the sea condition deviation.

[0057] S32: Construct a twin neural network model based on the radar altimeter product parameters according to the input value and the label value; the radar altimeter product parameters are sea surface parameters related to the sea condition deviation value, including sea surface wind speed, significant wave height, mean wave period, RMS of Ku-band significant wave height, and RMS of Ku-band backscatter coefficient in the sea surface parameters; the twin neural network model is the SNSSB model.

[0058] In an exemplary embodiment, the transfer relationship between adjacent layers in the SNSSB model is as follows:

[0059] a 0 = E

[0060] n m+1 = W m+1 a m + b m+1

[0061] a m+1 = f m+1 (n m+1 )

[0062] where a 0 is the input obtained by the initial neuron; E is the input vector, E = [E1, E2,..., E Q , Q represents the number of input vectors; n m+1 represents the input of the (m + 1)-th layer; a m and a m+1 are the outputs of the m-th layer and the (m + 1)-th layer respectively; W m+1 is the weight matrix of the (m + 1)-th layer; b m+1 is the bias vector of the (m + 1)-th layer; f m+1 is the activation function of the (m + 1)-th layer; m = 0, 1,..., M - 1, and M is the number of layers of the SNSSB model.

[0063] In an exemplary embodiment, in order to help the neural network better learn complex relationships and at the same time simplify the complexity of subsequent gradient calculations, the present application uses Sigmoid as the activation function, and its specific expression form is:

[0064]

[0065] S4 can be replaced by the following steps.

[0066] S41: Based on different combinations of sea surface parameters, use the gradient descent method to adjust the weights and biases of the SNSSB model to determine the trained SNSSB model; where the weights are: is the influence weight vector of the $i$-th neuron in the $(m - 1)$-th layer on the $j$-th neuron in the $m$-th layer at the $k$-th iteration; is the influence weight vector of the $i$-th neuron in the $(m - 1)$-th layer on the $j$-th neuron in the $m$-th layer at the $(k + 1)$-th iteration; $\alpha$ is the learning rate; is the influence weight vector of the $i$-th neuron in the $(m - 1)$-th layer on the $j$-th neuron in the $m$-th layer; $F$ is the loss function; the deviation is and are the bias vectors of the $i$-th neuron in the $m$-th layer at the $k$-th and $(k + 1)$-th iterations respectively; is the bias vector of the $i$-th neuron in the $m$-th layer.

[0067] In practical applications, the sea surface parameter combinations of 2D SNSSB are sea surface wind speed and significant wave height, those of 3D SNSSB are sea surface wind speed, significant wave height and mean wave period, and those of 5D SNSSB are sea surface wind speed, significant wave height, mean wave period, RMS of Ku-band backscattering coefficient and RMS of Ku-band significant wave height.

[0068] In practical applications, in order to save the SNSSB model with the best effect during multiple iterative runs, a model saving function is added. Usually, there are two ways to save the model. One way is to save the complete model, which includes not only the network structure of the model but also the model parameters; the other way is to save the model parameters. The advantage of this way is that it occupies less memory and is also a way recommended by the official. When the value obtained by the loss function remains almost unchanged within 10 steps, SNSSB will stop the iterative process. At this time, the saved model parameters are the best parameters after training.

[0069] In practical applications, before the formal model training, the dataset is divided into a training set, a validation set and a test set, and the data volumes among them are 6:2:2. At the same time, in order to reduce the dimensional difference between different input parameters and improve the training efficiency of the neural network during the training process, the input parameters are standardized, and the specific formula is as follows:

[0070]

[0071] where $X$ represents the original input data, which includes sea surface wind speed, significant wave height, mean wave period, RMS of Ku-band significant wave height and RMS of Ku-band backscattering coefficient; $\mu$ and $\sigma$ represent the mean and standard deviation of the original input data respectively, and $X'$ represents the data after standardization;

[0072] With the standardized data X′ as input data and the sea state deviation value as output data, a twin neural network based on the sea surface parameters of the radar altimeter is constructed. The twin neural network is the SNSSB model. Various hyperparameters of the twin neural network, such as the number of hidden layers, the number of neurons in each hidden layer, and the optimizer, need to be continuously tested. In the process of testing and adjustment, what needs to be done is to visualize the loss value, and continuously modify the hyperparameters through the loss curve, and finally find the optimal hyperparameters suitable for the model.

[0073] Furthermore, the sea state deviation value is estimated based on the SNSSB model. Based on the optimal hyperparameter settings that have been found, the data in the validation set is input into the SNSSB model to obtain the corresponding sea state deviation value.

[0074] Climate change affects all aspects of human life. The most direct impact of climate change on humans is the frequent occurrence of various abnormal situations, which affect people's normal production activities. In addition, climate change may lead to extreme weather events, which will also affect the well-being and safety of residents living far from the coast. As one of the most important climate change indicators, the average sea level height improves the accuracy of the observed sea level height by estimating a more accurate sea state deviation value, which is of great significance not only for monitoring climate change, but also for the subsequent inversion of the earth's gravity field.

[0075] This application constructs an SNSSB model to accurately evaluate the sea state deviation value, which can improve the accuracy of satellite altimetry, and is of great significance for establishing sea surface models, determining ocean geoid and bathymetry.

[0076] The following uses the L2-level GDR data product of the Jason-3 radar altimeter from 2016 to 2022 as an example to illustrate the technical solution of this application. Figure 2 shown.

[0077] This L2-level GDR data product is obtained using POE orbit determination data and waveform reconstruction methods. The data mainly includes dry tropospheric atmosphere correction, wet tropospheric atmosphere correction, ionospheric atmosphere correction, ocean tides and other parameter information, and has been fully verified. At the same time, the data delay time is: Jason-3 requires the data delay to be within 90 days.

[0078] (1) Obtaining intersection location and extracting data.

[0079] like Figure 3As shown in the figure, first, if we want to find the intersection points between satellite trajectories, we need to use the combination of the fast rejection test and the straddle test. First, we use the fast rejection test to screen out the trajectories that meet the conditions, and then use the straddle test to further judge. By the above operations, we can find two intersecting trajectory segments. Through the two ends of the segments, we can obtain the positions of four sub-satellite points. By interpolating the longitude and latitude of the four sub-satellite points, we can obtain the longitude and latitude information at the intersection point. Then, we take the data of the first four and the last four sub-satellite points at the intersection point respectively, and use the longitude and latitude information and the radial basis interpolation method to interpolate the sea surface parameters such as sea surface wind speed and significant wave height to the intersection point.

[0080] (2) Dataset construction.

[0081] After obtaining the position and parameter information at the intersection point, we need to screen the data to meet the subsequent usage requirements. The content includes latitude limit, water depth limit, sea surface height anomaly limit, and time limit.

[0082] Among them, the latitude limit is mainly to prevent the influence of sea ice on altimeter measurements. Therefore, the latitude limit range is between 66 degrees north and south latitudes.

[0083] The water depth limit is to avoid the influence of the near shore and islands. Therefore, it is set that the water depth at the intersection point should be greater than 1000 meters.

[0084] The sea surface height anomaly limit is to prevent large sea surface height fluctuations caused by sea surface dynamic changes. Therefore, the requirement is not greater than 25 cm.

[0085] The time limit is because the sea surface height does not change significantly within a few days. Therefore, the time range is limited within three days.

[0086] Through the above process, the data of the Jason-3 radar altimeter from 2016 to 2022 was extracted. The data is saved in such a way that the data for each year is saved separately instead of saving all the data as a whole. The reason for choosing this method is that when the model construction does not require too much data, it can avoid excessive computational workload for the subsequent work.

[0087] (3) Build the SNSSB model.

[0088] The SNSSB model of this application includes two independent and identically structured BP neural networks. Each BP neural network includes an input layer, three hidden layers, and an output layer. The transfer relationship between adjacent layers in SNSSB is set as follows:

[0089] a 0 =E

[0090] n m+1 =W m+1 am +b m+1

[0091] a m+1 = f m+1 (n m+1 )

[0092] where m = 0, 1, …, M - 1, and M is the number of layers of the neural network; n m+1 represents the input to the (m + 1)-th layer of the network; a m and a m +1 are the outputs of the m-th layer and the (m + 1)-th layer of the network respectively; W m+1 is the weight matrix of the (m + 1)-th layer of the network; b m+1 is the bias vector of the (m + 1)-th layer of the network; f m+1 is the activation function of the (m + 1)-th layer of the network; where the input vector E = [E1, E2, …, E Q , and Q represents the number of input vectors. To help the neural network better learn complex relationships and simplify the complexity of subsequent gradient calculations, this application uses Sigmoid as the activation function, and its specific expression form is:

[0093]

[0094] In the SNSSB model, the gradient descent rule is used to adjust the weights and biases in the network:

[0095]

[0096]

[0097] where α ∈ (0, 1) represents the learning rate; represents the influence weight vector of the i-th neuron in the (m - 1)-th layer on the j-th neuron in the m-th layer during the k-th iteration of the network, represents the influence weight vector of the i-th neuron in the (m - 1)-th layer on the j-th neuron in the m-th layer during the (k + 1)-th iteration of the network; and represent the bias vectors of the i-th neuron in the m-th layer during the k-th and (k + 1)-th iterations respectively.

[0098] (4) Model accuracy evaluation.

[0099] In the study of sea condition deviation, since there is a lack of accurate true values of sea condition deviation for verification, it is a common practice to analyze the sea surface height. To evaluate the accuracy of the SNSSB model, the following two parameters are used to assess the accuracy of the SNSSB model: the explained variance D and the sea surface height variance difference index (SVDI).

[0100] D = σ 2 (ΔSSH u ) - σ 2 (ΔSSH i )

[0101] where ΔSSH u represents the difference in sea surface height without sea condition deviation correction at the self-crossing point, and ΔSSH i represents the difference in sea surface height after sea condition deviation correction at the self-crossing point.

[0102]

[0103] where the dataset SSH D1 represents the difference in sea surface height obtained using the sea condition deviation calculated by the traditional sea condition deviation model. The dataset SSH D2 represents the difference in sea surface height obtained using the sea condition deviation calculated by SNSSB.

[0104] (5) Analysis of the results of the sea condition deviation model based on SNSSB.

[0105] Standardize the dataset obtained in (2) and divide it into a training set, a test set, and a validation set according to a ratio of 6:2:2. According to the above model evaluation indicators, the larger the value of the explained variance, the higher the accuracy of SNSSB compared to the traditional model (NPSSB), and the same is true for the sea surface height variance difference index. At the same time, when analyzing the results, select 10 degrees as the interval for statistical analysis to more clearly show the performance of the SNSSB model in different regions. The results of the model accuracy evaluation indicators D and SVDI are as shown in Figure 4 and Figure 5As shown. Through observation, it is found that at low, medium, and high latitudes, compared with the traditional model 2D NPSSB, the 2D SNSSB model has higher accuracy under the same input combination. Especially in the region from 70°S to 60°S, the improvement effect reaches nearly 30%. And as the dimension of the input combination increases, the accuracy of the SNSSB model is continuously increasing. When comparing the 2D NPSSB model and the 5D SNSSB model, within most of the globe, the accuracy improvement effect can reach 20%. Especially in the region from 10°S to 0°, the accuracy improvement effect of the SNSSB model is the highest, reaching approximately 40% at most, and in the region from 70°S to 60°S, the improvement effect of the SNSSB model also reaches 35%. This shows that the SNSSB model of this application is not only feasible for estimating sea state bias globally, but also has the characteristic of high accuracy.

[0106] In summary, obtaining a more accurate sea state bias value is crucial for the accurate acquisition of sea surface height. This application aims to deeply explore the relationship between the sea surface parameters observed by altimeters and sea state bias. Using the L2-level GDR data of the Jason-3 radar altimeter for 7 years (2016 - 2022), based on a method of twin neural networks, the sea state bias value at the corresponding sub-satellite point is estimated globally according to the sea surface parameters (wind speed, significant wave height, mean wave period, RMS of Ku-band significant wave height, and RMS of Ku-band backscatter coefficient).

[0107] This application uses the Jason-3 altimeter data with Ku-band and C-band for research, deeply explores the relationship between sea surface parameters and sea state bias values, and conducts accuracy verification through the explained variance at the intersection point and the sea surface height variance difference index. The results prove that under the same input parameters, the accuracy of the traditional sea state bias model (2D NPSSB) is lower than that of the 2D SNSSB model, and as the latitude of the input parameters increases, the accuracy of the SNSSB model becomes higher and higher. At the same time, the calculation amount of the traditional method for constructing the sea state bias model will increase greatly after increasing the dimension of the input parameters, making it difficult to simply increase the dimension of the input parameters. Therefore, the construction method proposed in this invention is also superior to the traditional method in terms of simplicity.

[0108] Based on the SNSSB model, the sea state bias values at the sub-satellite points in different sea areas globally can be accurately estimated, thus providing technical support for achieving long-term high-precision sea surface height observation.

[0109] Based on the same inventive concept, an embodiment of the present application further provides a radar altimeter sea state deviation value estimation device for implementing the radar altimeter sea state deviation value estimation method involved above. The implementation solutions provided by this device to solve problems are similar to the implementation solutions recorded in the above method. Therefore, the specific limitations in one or more embodiments of the radar altimeter sea state deviation value estimation device provided below can refer to the limitations on the radar altimeter sea state deviation value estimation method in the foregoing text, and will not be repeated here.

[0110] In an exemplary embodiment, a radar altimeter sea state deviation value estimation device is provided, including:

[0111] A screening module, configured to screen the L2-level GDR data of the radar altimeter based on the data identification bit of the radar altimeter and the sea surface parameter editing criterion; the data flag bit includes the ground type of the sub-satellite point in the radar altimeter and the current operating state of the radar altimeter; the sea surface parameters include sea surface wind speed, significant wave height, mean wave period, backscattering coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, RMS of Ku-band backscattering coefficient, and RMS of Ku-band significant wave height, etc.

[0112] A self-crossing point dataset establishing module, configured to establish a self-crossing point dataset according to the screened L2-level GDR data.

[0113] An SNSSB model constructing module, configured to construct an SNSSB model based on the threshold screening condition according to the self-crossing point dataset; the SNSSB model includes two independent and identically structured BP neural networks; each BP neural network includes an input layer, three hidden layers, and an output layer; the transfer relationship between adjacent layers is backpropagation; the threshold screening condition includes latitude limit, water depth limit, time limit, and crossing point sea surface height mismatch limit.

[0114] A training module, configured to train the SNSSB model according to different combinations of sea surface parameters to determine the trained SNSSB model.

[0115] A sea state deviation value estimating module, configured to estimate the sea state deviation value according to the trained SNSSB model.

[0116] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store radar altimeter sea state deviation value estimation data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for estimating radar altimeter sea state deviation values.

[0117] In an exemplary embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the above method is implemented.

[0118] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by the processor, the above method is implemented.

[0119] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by the processor, the above method is implemented.

[0120] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the various embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random-access memories (ReRAMs), magnetoresistive random-access memories (MRAMs), ferroelectric random-access memories (FRAMs), phase change memories (PCMs), graphene memories, and the like. Volatile memories can include random-access memories (RAMs) or external caches, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random-access memory (SRAM) or dynamic random-access memory (DRAM), etc.

[0121] In this application, all actions of obtaining signals, information, or data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where the location is located and obtaining authorization from the owner of the corresponding device.

[0122] The databases involved in the various embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., and are not limited thereto. The processors involved in the various embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., and are not limited thereto.

[0123] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0124] In this article, specific examples are used to elaborate on the principles and implementation manners of the present application. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for estimating the sea state deviation value of a radar altimeter, characterized in that, The method for estimating the sea state deviation value of the radar altimeter includes: Based on the data identification bit of the radar altimeter and the sea surface parameter editing criterion, screening the L2-level GDR data of the radar altimeter; the data flag bit includes the ground type of the sub-satellite point in the radar altimeter and the current operating state of the radar altimeter; the sea surface parameters include sea surface wind speed, significant wave height, mean wave period, backscattering coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, RMS of Ku-band backscattering coefficient, and RMS of Ku-band significant wave height; Establishing a self-crossing point data set according to the screened L2-level GDR data; Based on the threshold screening condition, constructing an SNSSB model according to the self-crossing point data set; the SNSSB model includes two independent and identically structured BP neural networks; each BP neural network includes an input layer, three hidden layers, and an output layer; the transfer relationship between adjacent layers is backpropagation; the threshold screening condition includes latitude limit, water depth limit, time limit, and cross-point sea surface height mismatch value limit; Training the SNSSB model according to different combinations of sea surface parameters to determine the trained SNSSB model; Estimating the sea state deviation value according to the trained SNSSB model.

2. The method for estimating the sea state deviation value of a radar altimeter according to claim 1, wherein Establishing a self-crossing point data set according to the screened L2-level GDR data, specifically including: Determining the crossing point position of the radar altimeter through a double-loop operation according to the screened L2-level GDR data; Based on the crossing point position, using the radial basis function interpolation method to determine the sea surface parameters at the crossing point; Establishing a self-crossing point data set according to the sea surface parameters.

3. The method for estimating the sea state deviation value of a radar altimeter according to claim 2, characterized in that, Determining the crossing point position of the radar altimeter through a double-loop operation according to the screened L2-level GDR data, specifically including: Using the fast rejection method to determine whether there is an intersection between the closed rectangles formed by the ascending and descending orbits as the diagonals, and obtaining a first judgment result; If the first judgment result is yes, determining that the necessary conditions for the existence of a crossing point are met between the two orbits, and using the straddle experiment method to judge whether there is an intersection relationship between the two orbits, and obtaining a second judgment result; If the second judgment result is yes, determining the self-crossing point position of the radar altimeter; If the first judgment result is no, entering the next loop and re-judging whether there is a crossing point between the new ascending and descending orbits; If the second judgment result is no, entering the next loop and re-judging whether there is a crossing point between the new ascending and descending orbits.

4. The method for estimating the sea state deviation value of a radar altimeter according to claim 1, characterized in that, Based on the threshold screening condition, constructing an SNSSB model according to the self-crossing point data set, specifically including: Taking the required sea surface parameters and sea surface height mismatch values at the crossing points in the self-crossing point data as input values and label values; Construct a twin neural network model based on the radar altimeter product parameters according to the input value and the label value; the radar altimeter product parameters are sea surface parameters related to the sea condition deviation value, including sea surface wind speed, significant wave height, mean wave period, RMS of Ku-band significant wave height, and RMS of Ku-band backscattering coefficient in the sea surface parameters; the twin neural network model is the SNSSB model.

5. The method for estimating the sea state deviation value of a radar altimeter according to claim 4, wherein The transfer relationship between adjacent layers in the SNSSB model is as follows: a 0 = E n m+1 = W m+1 a m + b m+1 a m+1 = f m+1 (n m+1 ) where a 0 is the input obtained by the initial neuron; E is the input vector, which is the sea surface parameter related to the sea condition deviation value; n m+1 represents the input of the (m + 1)-th layer; a m and a m+1 are the outputs of the m-th layer and the (m + 1)-th layer respectively; W m+1 is the weight matrix of the (m + 1)-th layer; b m+1 is the bias vector of the (m + 1)-th layer; f m+1 is the activation function of the (m + 1)-th layer.

6. The method for estimating the sea state deviation value of a radar altimeter according to claim 1, characterized in that Train the SNSSB model according to different combinations of sea surface parameters to determine the trained SNSSB model, specifically including: Based on different combinations of sea surface parameters, the weights and biases of the SNSSB model are adjusted using the gradient descent method to determine the trained SNSSB model; where the weights are: is the influence weight vector of the $i$-th neuron in the $(m - 1)$-th layer on the $j$-th neuron in the $m$-th layer at the $k$-th iteration; is the influence weight vector of the $i$-th neuron in the $(m - 1)$-th layer on the $j$-th neuron in the $m$-th layer at the $(k + 1)$-th iteration; $\alpha$ is the learning rate; is the influence weight vector of the $i$-th neuron in the $(m - 1)$-th layer on the $j$-th neuron in the $m$-th layer; $F$ is the loss function; The deviation is and respectively the bias vectors of the i-th neuron in the m-th layer at the k-th and (k + 1)-th iterations; is the bias vector of the i-th neuron in the m-th layer.

7. A device for estimating the sea state deviation value of a radar altimeter, characterized in that The radar altimeter sea condition deviation value estimation device includes: A screening module for screening the L2-level GDR data of the radar altimeter based on the data identification bit of the radar altimeter and the sea surface parameter editing criterion; the data flag bit includes the ground type of the sub-satellite point in the radar altimeter and the current operating state of the radar altimeter; the sea surface parameters include sea surface wind speed, significant wave height, mean wave period, backscattering coefficient, dry tropospheric atmospheric correction, wet tropospheric atmospheric correction, ocean tide, RMS of Ku-band backscattering coefficient, and RMS of Ku-band significant wave height; A self-crossing point dataset establishment module for establishing a self-crossing point dataset according to the screened L2-level GDR data; An SNSSB model construction module for constructing an SNSSB model based on the threshold screening condition according to the self-crossing point dataset; the SNSSB model includes two independent and identically structured BP neural networks; each BP neural network includes an input layer, three hidden layers, and an output layer; the transfer relationship between adjacent layers is backpropagation; the threshold screening condition includes latitude limit, water depth limit, time limit, and cross-point sea surface height mismatch value limit; A training module for training the SNSSB model according to different combinations of sea surface parameters to determine the trained SNSSB model; A sea condition deviation value estimation module for estimating the sea condition deviation value according to the trained SNSSB model.

8. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the radar altimeter sea condition deviation value estimation method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the radar altimeter sea condition deviation value estimation method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the radar altimeter sea condition deviation value estimation method according to any one of claims 1-6.

Images