Multi-stage reconstruction method for sea surface height information in SWOT swath observation blank area

Abstract

The application discloses a kind of sea surface elevation information multistage reconstruction methods for SWOT swath observation blank area.The method comprises: the satellite data of surface water and marine topography of target area are preprocessed and constructed as initial grid field with swath observation blank area;Multi-level interpolation and expansion grid processing are carried out until the final grid field of target area is obtained by complete filling, and the multistage reconstruction of sea surface elevation information of swath observation blank area is realized.The application can realize the high-resolution and high-precision reconstruction of sea surface elevation information, improve the spatial coverage and signal integrity of SWOT satellite observation;Through the way of multilayer circulation, gradually expand the grid, can make full use of observation data, to maximize the insertion of information, realize the generation of high-precision data;Global step-by-step interpolation of target area is carried out, effectively reduces the measurement error, reduces the risk of overfitting;It can continuously and completely fill in blank data, form spatially continuous high-resolution sea surface height field observation data, and expand the application prospect.

Description

Technical Field

[0001] This invention relates to a multi-level reconstruction method for sea surface elevation information, belonging to the field of marine information technology, and specifically to a multi-level reconstruction method for sea surface elevation information in SWOT swath observation blank areas. Background Technology

[0002] As the first wide-swath radar interferometry satellite capable of providing centimeter-level accuracy, SWOT can acquire high-resolution elevation information covering the surface of water bodies such as oceans, rivers, and lakes worldwide. However, due to SWOT's use of two-way side-looking radar signal interferometry, blind spots inevitably appear in its ground-based observation area. Even with synchronous measurements using traditional altimeters along its orbit, it is impossible to form continuous, high-resolution sea surface elevation field observation data in space. This will directly affect the study of small-scale dynamic structures in the ocean and limit the application prospects of SWOT satellite's high-resolution observations.

[0003] In this context, interpolation algorithms remain the primary method for reconstructing information in observation blind areas. Currently, various information reconstruction algorithms exist, including Kriging, geographic weighted regression, and machine learning algorithms. However, each algorithm has its advantages and disadvantages: Kriging has a solid theoretical foundation but is sensitive to data distribution assumptions; geographic weighted regression can improve estimation accuracy but requires a large amount of observational data and is not suitable for sparse data regions; machine learning algorithms can handle large-scale data but require a large amount of labeled data to train the model, and the model is difficult to understand and adjust. How to effectively utilize SWOT satellite altimeter observation data to perform multi-level reconstruction of sea surface elevation information in observational blind areas is a particularly important topic. Summary of the Invention

[0004] To address the problems existing in the background technology, the present invention provides a multi-level reconstruction method for sea surface elevation information in SWOT swath observation blank areas.

[0005] The technical solution adopted in this invention is:

[0006] The present invention provides a multi-level reconstruction method for sea surface elevation information in SWOT swath observation blank areas, comprising:

[0007] Step 1) Preprocess the SWOT (Surface Water and Ocean Topography) satellite data of the target area and construct an initial grid field with swath observation blank areas.

[0008] Step 2) The blank areas of the initial grid field are interpolated and expanded in a loop until the blank areas of the initial grid field are completely filled to obtain the final grid field of the target area, thereby realizing the multi-level reconstruction of the sea surface elevation information of the blank areas of the initial grid field.

[0009] In step 1), the SWOT satellite data of the target area for surface water and ocean topography includes several parallel, spaced regions with elevation data obtained by SWOT satellite scanning of surface water and ocean topography. Between each two adjacent regions with elevation data is a swath observation blank area. The regions with elevation data consist of several parallel, continuous bar regions with elevation information, while the swath observation blank area consists of several parallel, continuous bar regions without data.

[0010] The SWOT satellite data of surface water and ocean topography in the target area are preprocessed. Specifically, for each blank area of ​​the swath observation, the two outermost bar regions without data in the blank area are used as interpolation target regions. The bar regions with elevation information closest to the blank area in the two adjacent regions with elevation data are used to construct the original sampling point region. The elevation data of land points in each original sampling point region are set to 0. Then, the elevation data of the missing land points in the middle are filled in according to the elevation data of the ocean region in the original sampling point region using the difference algorithm. Finally, an initial grid field with blank areas of the swath observation is constructed to reduce the negative impact of errors in the subsequent algorithm process.

[0011] In step 2), interpolation and grid expansion are performed cyclically on the blank areas of the initial grid field's ripped area observations. Specifically, for each blank area, the process is as follows:

[0012] The two interpolation target regions in the blank area of ​​the ripped area observation are interpolated based on the original sampling point region and then filled with elevation data. The two filled interpolation target regions are merged into the original sampling point region and the grid is expanded to construct the current sampling point region. The two outermost bar regions without data in the remaining area of ​​the ripped area observation blank area are taken as the current interpolation target regions. The sampling point region and the interpolation target region are used to continue to perform interpolation and grid expansion in a loop until the elevation information in the ripped area observation blank area is completely filled.

[0013] The final grid field of the target area is obtained by completely filling the blank area of ​​each ripped observation.

[0014] In step 2), the interpolation is specifically performed using a multi-level spectral interpolation method based on trigonometric basis function expansion.

[0015] This invention specifically utilizes a high-resolution multi-level interpolation method based on trigonometric basis function expansion. The coefficients of the trigonometric basis functions determine their highest order in each dimension, thus determining the highest resolution achievable by the function. First, a lower-order trigonometric basis function is constructed to achieve a rough approximation across the entire field, obtaining large-scale estimates of sea surface height structure information in blank areas, thus initially obtaining a spatially continuous sea surface height field with low resolution. As the initial grid fills in some blank areas, the order of the trigonometric basis functions can be increased to improve the interpolation resolution, resulting in higher-resolution estimates. This ensures that estimation errors are not introduced in areas with low data density, while in areas with high data density, high-frequency components are fully captured to obtain estimates at a finer scale. By repeatedly increasing the order of the trigonometric basis functions for interpolation until the function corresponding to the highest order resolution is obtained, the observation information is continuously enriched. This process achieves spatially continuous sea surface height field observation data while maintaining resolution and accuracy.

[0016] The beneficial effects of this invention are:

[0017] 1) This invention can achieve high-resolution and high-precision reconstruction of sea surface elevation information in the blank areas of SWOT satellite altimeter observation, thereby improving the coverage and signal integrity of satellite observation of sea surface information.

[0018] 2) This invention expands the grid gradually through a multi-layered loop. Each expansion can make full use of the information at the corresponding scale in the observation data to maximize the insertion of information and achieve high-precision reconstruction of multi-level signals.

[0019] 3) The high-resolution, high-precision environmental field reconstructed by this invention can be widely applied in multiple marine engineering fields such as path planning, marine monitoring, and verification of meteorological prediction models.

[0020] 4) This invention does not only focus on local interpolation results, but also performs global stepwise interpolation of the target region through a Fourier framework, which can effectively reduce high-frequency noise caused by measurement errors and reduce the risk of overfitting to a certain extent.

[0021] 5) This invention can continuously and completely fill in the blank areas of data, forming spatially continuous high-resolution sea surface height field observation data, which can be used for the study of small-scale dynamic structures in the ocean, expanding the application prospects of high-resolution observations from SWOT satellites. Attached Figure Description

[0022] Figure 1 This is a flowchart of the multi-level spectral interpolation elevation information reconstruction process of the present invention;

[0023] Figure 2 This is a schematic diagram of the region before and after reconstruction using the multi-level elevation information reconstruction algorithm of this invention. Figure 2 (a) is a schematic diagram of the region before reconstruction by the multi-level elevation information reconstruction algorithm of the present invention. Figure 2 (b) is a schematic diagram of the region after reconstruction by the multi-level elevation information reconstruction algorithm of the present invention;

[0024] Figure 3 This is an error diagram of the multi-level elevation information reconstruction algorithm results of the present invention, wherein, Figure 3 (a) is a schematic diagram of the error values ​​for this region obtained by comparing the reconstructed high-resolution environmental field with the actual environmental field data. Figure 3 (b) is an error distribution map of the region obtained by comparing the reconstructed high-resolution environmental field with the real environmental field data. Detailed Implementation

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] like Figure 1 As shown, the multi-level reconstruction method for sea surface elevation information in SWOT swath observation blank areas of the present invention includes:

[0027] Step 1) Preprocess the SWOT (Surface Water and Ocean Topography) satellite data of the target area and construct an initial grid field with swath observation blank areas.

[0028] In step 1), the SWOT satellite data of the target area for surface water and ocean topography includes several parallel and spaced areas with elevation data obtained by SWOT satellite scanning of surface water and ocean topography. Between each two adjacent areas with elevation data is a swath observation blank area. The areas with elevation data consist of several parallel and continuous bar-shaped areas with elevation information, while the swath observation blank area consists of several parallel and continuous bar-shaped areas without data.

[0029] The SWOT satellite altimetry data of the target area is preprocessed. Specifically, for each blank area of ​​the swath observation, the two outermost bar regions without data in the blank area are used as interpolation target regions. The bar regions with elevation information closest to the blank area in the two adjacent regions with elevation data are used to construct the original sampling point region. The elevation data of land points in each original sampling point region are set to 0. Then, the elevation data of the missing land points in the middle are filled in according to the elevation data of the ocean region in the original sampling point region using the difference algorithm. Finally, an initial grid field with blank areas of the swath observation is constructed to reduce the negative impact of errors in the subsequent algorithm process.

[0030] Step 2) The blank areas of the initial grid field are interpolated and expanded in a loop until the blank areas of the initial grid field are completely filled to obtain the final grid field of the target area, thereby realizing the multi-level reconstruction of the sea surface elevation information of the blank areas of the initial grid field.

[0031] In step 2), interpolation and grid expansion are performed cyclically on the blank areas of the initial grid field's ripped observations. Specifically, for each blank area, the process is as follows:

[0032] The two interpolation target regions in the blank area of ​​the ripped area observation are interpolated based on the original sampling point region and then filled with elevation data. The two filled interpolation target regions are merged into the original sampling point region and the grid is expanded to construct the current sampling point region. The two outermost bar regions without data in the remaining area of ​​the ripped area observation blank area are taken as the current interpolation target regions. The sampling point region and the interpolation target region are used to continue to perform interpolation and grid expansion in a loop until the elevation information in the ripped area observation blank area is completely filled.

[0033] The final grid field of the target area is obtained by completely filling the blank area of ​​each ripped observation.

[0034] In step 2), the interpolation is specifically performed using a multi-level spectral interpolation method based on trigonometric basis function expansion.

[0035] This invention specifically utilizes a high-resolution multi-level spectral interpolation method based on trigonometric basis function expansion. The coefficients of the trigonometric basis functions determine their highest order in each dimension, thus determining the highest resolution achievable by the function. First, a lower-order trigonometric basis function is constructed to achieve a rough approximation across the entire field, obtaining large-scale estimates of the sea surface height structure information in blank areas, thus initially obtaining a low-resolution spatially continuous sea surface height field. As the initial grid fills some local blank areas, the order of the trigonometric basis functions can be increased to improve the interpolation resolution, thereby obtaining higher-resolution estimation results. This ensures that estimation errors are not introduced in areas with low data density, while in areas with high data density, high-frequency components are fully captured to obtain estimates at a finer scale. By repeatedly increasing the order of the trigonometric basis functions for interpolation until the function corresponding to the highest order resolution is obtained, the observation information is continuously enriched. This process achieves spatially continuous sea surface height field observation data while maintaining resolution and accuracy.

[0036] Specific embodiments of the present invention are as follows:

[0037] This embodiment describes the overall process of the present invention in practical application and performs simulation analysis.

[0038] like Figure 2As shown in (a), the data matrix contains SWOT satellite observation data points of size 71*9866. The blank area in the middle represents the SWOT sub-satellite observation blind zone, consisting of 8 blank strips. Eight strips on each side of the blank area (a total of 16 strips) are selected as the initial grid, forming a 16*9866 low-resolution initial grid. The initial grid contains some land points, and the presence of these land points significantly affects subsequent interpolation algorithms. Therefore, the land points in the initial grid are processed. Land points in the grid are marked as 0, and then these points are filled with sea surface height data surrounding them. This process is accomplished using the interpolation method. The interpolation algorithm is based on trigonometric interpolation.

[0039] Since the blank area contains 8 bands, interpolation is performed 4 times, interpolating 2 bands each time. This process is repeated multiple times until the desired result is obtained. Figure 2 As shown in (b), to meet the resolution requirements of the grid field, interpolation yields 8 strips, obtaining continuous data. At this point, the sea surface height information reconstruction is complete. Furthermore, Figure 3 3(a) is the error value for this region obtained by comparing the reconstructed high-resolution environmental field with the actual environmental field data. Figure 3 (b) is the error distribution diagram, which can be used to analyze the accuracy of the reconstructed field.

[0040] To verify the effectiveness of multi-level reconstruction of sea surface elevation information for the blank areas of SWOT satellite altimeter observations, a two-dimensional simulation analysis of the marine environmental field is conducted below. After four levels of iterative grid expansion, the number of interpolation lines in the grid field gradually increased from the initial 16 to 18, 20, 22, and 24 lines, resulting in the final sea surface elevation reconstruction field as shown below. Figure 2 As shown in (b). By Figure 3 (a) and Figure 3 As shown in error plot (b), the data error is relatively small in most areas, mainly concentrated between -0.2 and 0.2. The smaller areas with larger errors may be due to two reasons: firstly, errors inherent in the coastal data itself; and secondly, larger errors occur when data differences are significant within certain grid areas. Overall, the multi-level sea surface elevation information reconstruction algorithm of this invention performs well in this example.

Claims

1. A multi-level reconstruction method for sea surface elevation information in SWOT swath observation blank areas, characterized in that, include: Step 1) Preprocess the SWOT satellite data of surface water and ocean topography of the target area and construct an initial grid field with swath observation blank areas; Step 2) The blank areas of the initial grid field are interpolated and expanded in a loop until the blank areas of the initial grid field are completely filled to obtain the final grid field of the target area, thereby realizing the multi-level reconstruction of the sea surface elevation information of the blank areas of the initial grid field. In step 1), the SWOT satellite data of the target area for surface water and ocean topography includes several parallel and spaced areas with elevation data obtained by scanning the SWOT satellite for surface water and ocean topography. Between each two adjacent areas with elevation data is a cut-off observation blank area. The areas with elevation data consist of several parallel and continuous bar areas with elevation information, and the cut-off observation blank area consists of several parallel and continuous bar areas without data. The SWOT satellite data of surface water and ocean topography in the target area are preprocessed. Specifically, for each blank area of ​​the ripped observation, the two outermost bar regions without data in the blank area are used as interpolation target regions. The bar regions with elevation information that are closest to the blank area of ​​the ripped observation are combined with the two adjacent regions with elevation data to form the original sampling point region. The elevation data of land points in each original sampling point region are set to 0. Then, the elevation data of the missing land points in the middle are filled into the values ​​according to the elevation data of the ocean region in the original sampling point region using the difference algorithm. Finally, the initial grid field with the blank area of ​​the ripped observation is constructed. In step 2), interpolation and grid expansion are performed cyclically on the blank areas of the initial grid field's ripped area observations. Specifically, for each blank area, the process is as follows: The two interpolation target regions in the blank area of ​​the ripped area observation are interpolated based on the original sampling point region and then filled with elevation data. The two filled interpolation target regions are merged into the original sampling point region and the grid is expanded to construct the current sampling point region. The two outermost bar regions without data in the remaining area of ​​the ripped area observation blank area are taken as the current interpolation target regions. The sampling point region and the interpolation target region are used to continue to perform interpolation and grid expansion in a loop until the elevation information in the ripped area observation blank area is completely filled. The final grid field of the target area is obtained by completely filling the blank area of ​​each ripped observation.

2. The multi-level reconstruction method for sea surface elevation information in SWOT swath observation blank areas according to claim 1, characterized in that: In step 2), the interpolation is specifically performed using a multi-level spectral interpolation method based on trigonometric basis function expansion.

Images