High temporal and spatial resolution water level inversion method integrating satellite altimetry data and GNSS-R

Abstract

The present invention discloses a high-temporal-spatial-resolution water level inversion method that integrates satellite altimetry data and GNSS-R. First, the target area is determined, and the GNSS data, satellite altimetry data, and water body mask data of the target area are obtained; the satellite altimetry data is processed to obtain the observed water level in the study area, and the system offset is eliminated by comparing the average water level deviation during the overlapping period of different altimetry satellites to obtain the multi-source altimetry satellite fusion water level; the GNSS data is processed to construct a mathematical model between the signal-to-noise ratio data and the water surface height, and the GNSS-R water level data is obtained using a nonlinear fitting method; finally, the satellite altimetry data and the GNSS-R data are combined to construct a high-temporal-spatial-resolution water level inversion model based on XGBoost. By combining satellite altimetry data and GNSS-R data, the present invention can achieve water level monitoring with higher temporal-spatial-resolution, significantly improve the accuracy and reliability of water level inversion, and has important applications in disaster warning and water resources management.

Description

Technical Field

[0001] The present invention relates to the technical field of GNSS-R water level inversion, and in particular to a high temporal and spatial resolution water level inversion method that integrates satellite altimetry data and GNSS-R. Background Art

[0002] Water level measurement is a crucial component of hydrological monitoring, playing a vital role in flood prevention and control, river infrastructure development, and other areas. Accurately and efficiently observing water level changes across large river basins has become a research hotspot in recent years. Traditional water level monitoring methods (such as water gauges and water level gauges) can generally meet these requirements, but these methods face challenges in remote areas, such as high installation costs, difficult maintenance, and poor equipment stability.

[0003] In recent years, the development of GNSS-R technology has provided a new technical path for water level monitoring. GNSS-R is a new remote sensing technology that uses GNSS reflected signals to invert geophysical parameters. Unlike traditional GNSS positioning, GNSS-R focuses on the characteristics of GNSS signals after reflection from the Earth's surface (such as water, soil, ice, etc.). By analyzing the time delay, phase change, power attenuation and other characteristics of these reflected signals, a series of geophysical parameters can be inverted. GNSS-R technology has shown significant advantages in water level monitoring. Compared with traditional water level measurement methods, it has the characteristics of low cost, global coverage, passive measurement and low energy consumption.

[0004] While ground-based GNSS-R technology enables dynamic, all-weather water-level monitoring, its coverage is limited, making it difficult to fully reflect water-level changes over large areas. This limitation can lead to blind spots in monitoring, compromising the comprehensiveness and reliability of hydrological information, resulting in delayed warnings and false alarms. These issues, particularly in complex terrain and environments with multiple water bodies, can reduce the effectiveness of disaster prevention and mitigation, increasing the risk of loss of life and property.

[0005] Satellite altimetry technology, with its all-day, all-weather capabilities and wide coverage, can reach areas difficult to reach with ground-based observations. Even in extreme weather conditions, it can provide accurate water level data, providing crucial support for flood prevention. However, the temporal resolution of satellite altimetry data is low, making it difficult to meet the high-speed water level data requirements for flood prevention and control. Summary of the Invention

[0006] Therefore, the present invention addresses the problem of limited coverage of ground-based GNSS-R. By combining ground-based GNSS-R technology with satellite altimetry technology to achieve complementary advantages, a high-temporal and spatial resolution water level inversion method integrating satellite altimetry data and GNSS-R is proposed to effectively improve the comprehensiveness and reliability of hydrological information, thereby better achieving the goal of disaster prevention and mitigation. The purpose of the present invention is achieved through the following technical solutions:

[0007] A high temporal and spatial resolution water level inversion method integrating satellite altimetry data and GNSS-R includes the following steps:

[0008] Step 1: Collect basic data: determine the target area and obtain GNSS observation data, satellite altimetry data, and water mask data of the target area;

[0009] Step 2: Obtain multi-source altimetry satellite fusion water level data: Extract the vector boundary of the target area based on the water body mask data. Then, obtain multiple satellite altimetry data within the target area based on the vector boundary. Error correction and outlier removal are performed on the satellite altimetry data to finally obtain the along-track observed water level. The observation data of different altimetry satellites are converted to a unified elevation benchmark, and the system offset is eliminated by comparing the average water level deviation during the overlap period of different altimetry satellites to obtain the multi-source altimetry satellite fusion water level.

[0010] Step 3: Obtain GNSS-R water level data: Extract the signal-to-noise ratio data of low satellite elevation angles from the GNSS observation data and process it using a low-order polynomial to remove interference from the direct signal; construct a mathematical model between the reflected signal and the water surface height, and select equally spaced sampling points to preliminarily determine the unknown parameters of the model. Then, use the nonlinear least squares method to fit and solve the parameters, and use the established mathematical model to calculate the water level at the sampling points;

[0011] Step 4: Convert GNSS-R water level data and satellite altimetry data to the same elevation benchmark, and then build an XGBoost-based model by combining the wide coverage of satellite altimetry data and the high temporal resolution of GNSS-R water level data. The multi-source altimetry satellite fusion water level in the study area, GNSS-R water level data from different stations, time, and longitude and latitude are used as sample data, and the XGBoost machine learning algorithm is used to build a high spatiotemporal resolution water level inversion model. The model is as follows:

[0012] W alt =f(t,lon,lat,W gnss )

[0013] Among them, W alt is the fusion water level of multi-source altimetry satellites; t is time; lon is longitude; lat is latitude; W gnss is the GNSS-R water level.

[0014] For further optimization, the specific operations of step 1 are as follows: First, based on research needs, define the target area and determine the scope of the water body of interest (such as lakes, rivers, or reservoirs), and clarify its geographic scope and spatial coordinates. Then, obtain GNSS observation data from GNSS receiving equipment or relevant databases within the target area. Simultaneously, extract satellite altimetry data (water level observation data) covering the target area from altimetry satellites (such as ICESat-2 and SWOT), ensuring sufficient temporal and spatial coverage. Extract water body distribution information in the target area through remote sensing imagery or public water body databases (such as GLCF or MODIS water body data), and generate water body mask data to limit the data's applicable scope and remove interference from non-water body areas.

[0015] Furthermore, in step 2: the satellite altimetry data is corrected for errors and the water surface height is calculated. The specific formula is as follows:

[0016] H=H alt -R range -(iono+wet+dry+solid+pole)-Giodt

[0017] Where, H represents the water level; H alt Indicates the distance from the satellite to the reference ellipsoid; R range Represents the distance from the satellite to the Earth's surface; iono represents the ionospheric correction (using the GIM global ionospheric model); wet represents the wet tropospheric correction; dry represents the dry tropospheric correction; solid represents the solid Earth tide correction; pole represents the polar tide correction; geoid is used to convert the vertical reference from the reference ellipsoid to the geoid (using the EGM2008 geoid model).

[0018] Furthermore, in step 2, eliminating the system offset by comparing the average water level deviation during the overlapping period of different altimetry satellites specifically includes: comparing the data during the overlapping observation period, calculating the mean and standard deviation of the deviation, and correcting the deviation using linear regression or other statistical methods.

[0019] Furthermore, the mathematical model constructed in step 3 is:

[0020]

[0021] Where δSNR is the signal-to-noise ratio after removing the direct signal; λ is the satellite signal wavelength; h is the reflection height; θ is the satellite elevation angle; k is the wave number; s is the roughness parameter of the reflecting surface; C1 and C2 are the in-phase and out-of-phase components, which are used to replace the amplitude A and phase

[0022]

[0023] After the mathematical model is established, m sampling points with equal intervals are selected and the initial values ​​of the undetermined parameters are determined. The undetermined parameters are set as C1, C2, h, s 2 ,The best fitting parameters are solved by the nonlinear least square method, and the water level height of the sampling point can be calculated through the mathematical model.

[0024] The application of the technical solution of the present invention has the following beneficial effects:

[0025] (1) Improve the spatiotemporal resolution of water level monitoring

[0026] Satellite altimetry data offers wide coverage and high spatial accuracy, while GNSS-R data boasts high temporal resolution. Fusion of the two overcomes the shortcomings of a single data source, generating high-resolution water level data in both time and space, enabling a more refined description of dynamic hydrological processes.

[0027] (2) Enhance the accuracy and reliability of water level monitoring

[0028] By fusing different data sources and leveraging their respective strengths, such as the millimeter-level accuracy of satellite altimetry and the temporal continuity of GNSS-R, the impact of errors or missing data in a single data source is reduced, thereby improving the accuracy and stability of overall water level monitoring. Through innovative data fusion and machine learning optimization, this method surpasses traditional single-source technologies in flexibility, adaptability, and accuracy, and exhibits strong real-time and reliability.

[0029] (3) Improve adaptability in complex hydrological environments

[0030] This method can provide reliable data support in areas with complex hydrological environments that are difficult to cover with traditional monitoring methods (such as plateau lakes and remote reservoirs). It is suitable for areas with complex terrain and limited monitoring conditions, and can achieve accurate water level inversion. The present invention will be further described in detail below with reference to the figures. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a flow chart of the high spatiotemporal resolution water level inversion method that integrates satellite altimetry data and GNSS-R in Example 1;

[0032] Figure 2 For example 1, the system deviations of different altimeters (1) and the changing trends of different height-measured water levels in the same period; (2) the correlation equation constructed by linear fitting;

[0033] Figure 3 This is a schematic diagram of the GNSS-R water level inversion principle;

[0034] Figure 4The process of constructing a high spatiotemporal resolution water level inversion model in Example 1;

[0035] Figure 5 Comparison between the simulated water levels and the measured water levels at four stations in Lake Huron in Example 1. (a), (b), (c), and (d) represent the four stations respectively. DETAILED DESCRIPTION

[0036] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. However, the present invention can be implemented in many different ways as defined and covered by the claims.

[0037] Example 1:

[0038] A high temporal and spatial resolution water level inversion method integrating satellite altimetry data and GNSS-R includes the following steps:

[0039] Step 1: Determine the target area and obtain GNSS data, satellite altimetry data, and water mask data for the target area;

[0040] According to the research needs, the target area is first delineated, the scope of the water body of interest (such as lakes, rivers or reservoirs) is determined, and its geographical scope and spatial coordinates are clarified. Subsequently, GNSS observation data are obtained from GNSS receiving equipment or related databases in the target area through the Canadian Active Control System and the United States National Geodetic Survey. At the same time, data from three radar altimetry satellites (Jason-3, Sentinel-3A and Sentinel-3B) and one laser altimeter (ICESat-2) in the Lake Huron region from January 2021 to December 2023 are selected to extract water level observation data covering the target area to ensure sufficient temporal and spatial coverage. In addition, water body distribution information of the target area is extracted through remote sensing images or public water body databases (such as GLCF or MODIS water body data), and a water body mask is generated to limit the applicable scope of the data and remove interference from non-water body areas.

[0041] Step 2: Based on the water mask data, the transit data of different altimeter satellites within the lake are extracted. These data are then corrected for errors and outliers are removed to obtain the along-track water level. The observations from these different altimeter satellites are then converted to a unified elevation datum. The systematic offset is eliminated by comparing the average water level deviation during the overlap period of the different altimeter satellites, resulting in a fused water level from multiple altimeter satellites.

[0042] Specifically, the water mask data is used to extract the vector boundary of the target area, and then the satellite altimeter data inside the target area is obtained based on the vector boundary. The satellite altimeter data is then corrected for errors and the water surface height is calculated using the following formula:

[0043] H=Halt -R range -(iono+wet+dry+solid+pole)-Geod (1)

[0044] Where, H represents the water level; H alt Indicates the distance from the satellite to the reference ellipsoid. m; R range Represents the distance from the satellite to the Earth's surface in meters; iono represents the ionospheric correction (using the GIM global ionospheric model); wet represents the wet tropospheric correction; dry represents the dry tropospheric correction; solid represents the solid Earth tide correction; pole represents the polar tide correction; geoid is used to convert the vertical datum from the reference ellipsoid to the geoid (using the EGM2008 geoid model).

[0045] Before integrating multi-source satellite altimetry data, relative calibration is required to eliminate systematic biases. This involves comparing data from overlapping observation periods, calculating the mean and standard deviation of the biases, and correcting for them using linear regression or other statistical methods.

[0046] For example, the data obtained by ICESat-2 and Sentinel-3B satellites in a certain area have 14 observation dates with the same observation date, and the trend of change is highly consistent, such as Figure 2 As shown in , the ICESat-2 satellite data are unified to the Sentinel-3B measurement benchmark through linear fitting to eliminate the systematic bias between the two.

[0047] Step 3: Extract the signal-to-noise ratio data of low satellite elevation angles from the GNSS data, use a low-order polynomial to remove the direct signal portion of the signal-to-noise ratio data, and establish a mathematical model between it and the water surface height. The principle is as follows: Figure 3 This is a schematic diagram of the water level inversion using GNSS-R technology. As can be seen from the figure, the GNSS antenna receives the direct signal from the GNSS satellite as well as the reflected signal from the water surface. Compared with the direct signal, the reflected signal has an additional path D when it reaches the antenna, which can be expressed as:

[0048] D=2hsinθ (2)

[0049] Where h is the vertical distance from the receiver antenna phase center to the water surface; θ is the satellite elevation angle. The path delay can be converted to phase delay using the following formula:

[0050]

[0051] Where λ is the wavelength of the satellite signal and θ is the satellite elevation angle. From equation (3), we can deduce that there is a relationship between frequency and reflection height:

[0052]

[0053] Then, the signal-to-noise ratio data after detrending is modeled as:

[0054]

[0055] Where A is the amplitude; is the phase; λ is the wavelength of the satellite signal; θ is the satellite altitude angle.

[0056] The nonlinear fitting method introduces an attenuation factor into formula (5), and in order to ensure a stable numerical solution, it is transformed to obtain formula (6):

[0057]

[0058] Where, δSNR is the signal-to-noise ratio data after removing the direct signal; is the attenuation factor; λ is the satellite signal wavelength; h is the reflection height; θ is the satellite altitude angle; k is the wave number; s is the roughness parameter of the reflecting surface; C1 and C2 are the in-phase and out-of-phase components, which are used to replace the amplitude A and phase

[0059]

[0060] After the mathematical model is established, m sampling points with equal intervals are selected and the initial values ​​of the undetermined parameters are determined. The undetermined parameters are set as C1, C2, h, s 2 ,The best fitting parameters are solved by the nonlinear least square method, and the water level height of the sampling point can be calculated through the mathematical model.

[0061] Step 4: The multi-source altimetry satellite fusion water level data, GNSS-R water level data from different stations, time, and longitude and latitude in the study area are used as sample data. Then, a high spatiotemporal resolution water level inversion model is constructed by combining the XGBoost machine learning algorithm. The model is as follows:

[0062] W alt =f(t,lon,lat,W gnss ) (8)

[0063] Among them, W alt is the fusion water level of multi-source altimetry satellites; t is time; lon is longitude; lat is latitude; W gnss is the GNSS-R water level.

[0064] Water level data in multiple target areas were randomly selected according to longitude and latitude, and the correlation and accuracy between the inverted water level and the measured water level were statistically analyzed to complete the accuracy verification of the high temporal and spatial resolution water level inversion model.

[0065] The input parameters of the high spatiotemporal resolution water level inversion model built with the XGBoost machine learning algorithm are GNSS-IR water level data, time, longitude and latitude, temperature and rainfall at different sites, and the output parameters are satellite altimetry water level data. The process of building a high spatiotemporal resolution water level inversion model is as follows: Figure 4 shown.

[0066] To construct a high-temporal-resolution water level inversion model, the entire dataset must first be divided into a training set and a test set in a ratio of 7:3. During the training process, a 5-fold cross-validation was implemented to enhance the generalization ability of the model. At the same time, the Bayesian optimization method was used to adjust the three key hyperparameters in the model, and the coefficient of determination R obtained by cross-validation was used to determine the best performance. 2 The average value of is used as the optimization target. After setting an appropriate number of iterations, the optimization process will produce an optimal set of hyperparameter configurations. Using this set of parameters, a high-temporal-resolution water level inversion model is trained on the training set. Finally, the model's predictive performance is validated on the test set to evaluate its effectiveness in practical applications.

[0067] Measured data from four stations within Lake Huron, numbered 11690, 11070, 9075014, and 9075099, were selected to evaluate the model's effectiveness. Due to Lake Huron's vast area, water levels can vary significantly even in different areas of the lake (comparing measured data from different stations revealed an average deviation of 2 to 5 cm between stations within the same lake). Therefore, fused data was collected within a 0.4-degree radius around each hydrological station. This radius not only ensures sufficient elevation observations but also keeps lake surface fluctuations within a relatively small range, allowing each station to be treated as an independent, small lake for study.

[0068] By fusing multi-source data, this model effectively overcomes the limitations of a single data source, enabling dynamic water level monitoring over a large area and reducing the risk of false alarms and missed reports caused by local data anomalies, thereby significantly improving the accuracy and real-time nature of water regime monitoring. To verify the effectiveness of the model, the simulated water levels at stations 11070, 11690, 9075014, and 9075099 with a time resolution of 1 hour from January 1, 2021, to December 31, 2023, were compared with the measured water levels. The statistical results are shown in the table below:

[0069] Table 1

[0070]

[0071] Figure 5The simulated water levels (blue line), measured data (red line), and satellite altimetry data (yellow scattered dots) from January 1, 2021, to December 31, 2023, are plotted together. Combining the graph and table, we can see that the simulated and measured water levels at each station show consistent trends. The RMSEs for the simulated and measured water levels at each station are 7.46 cm, 5.15 cm, 6.10 cm, and 7.34 cm, respectively, indicating that the simulated water levels obtained by the model are highly accurate.

[0072] The working principle and technical effect of the above technical solution are as follows: By matching satellite altimetry data with GNSS-R water level data from the same time and region, and using their shared temporal and spatial characteristics as input data to train a model, a high-temporal-resolution water level inversion model is constructed that integrates satellite altimetry data and GNSS-R. This model combines the advantages of satellite altimetry data and GNSS-R to improve the temporal and spatial resolution of water level monitoring.

[0073] Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A high temporal and spatial resolution water level inversion method integrating satellite altimetry data and GNSS-R, characterized in that: The following steps are involved: Step 1: Collect basic data: determine the target area and obtain GNSS observation data, satellite altimetry data, and water mask data of the target area; Step 2: Obtain multi-source altimetry satellite fusion water level data: Extract the vector boundary of the target area based on the water body mask data, obtain multiple satellite altimetry data within the target area based on the vector boundary, and perform error correction and outlier removal on the satellite altimetry data; convert the observation data of different altimetry satellites to a unified elevation benchmark, and eliminate the system offset by comparing the average water level deviation during the overlap period of different altimetry satellites to obtain the multi-source altimetry satellite fusion water level; Step 3: Obtain GNSS-R water level data: Extract the signal-to-noise ratio data from the GNSS observation data and process it to remove interference from the direct signal; construct a mathematical model between the reflected signal and the water surface height, and select equally spaced sampling points to preliminarily determine the model's unknown parameters. Then, use the nonlinear least squares method to fit and solve the parameters, and use the established mathematical model to calculate the water level at the sampling points; Step 4: The multi-source altimetry satellite fusion water level data, GNSS-R water level data from different stations, time, and longitude and latitude in the study area are used as sample data, and the XGBoost machine learning algorithm is combined to build a high spatiotemporal resolution water level inversion model. The model is as follows: W alt =f(t,lon,lat,W gnss ) Among them, W alt is the fusion water level of multi-source altimetry satellites; t is time; lon is longitude; lat is latitude; W gnss is the GNSS-R water level.

2. The high temporal and spatial resolution water level inversion method of integrating satellite altimetry data and GNSS-R according to claim 1 is characterized in that: The specific operations of step one are as follows: based on research needs, delineate the target area, determine the scope of the water body of interest, and clarify its geographical scope and spatial coordinates; obtain GNSS observation data from GNSS receiving equipment or related databases within the target area; extract satellite altimetry data covering the target area from altimetry satellites; extract water body distribution information in the target area through remote sensing images or public water body databases to generate water body mask data.

3. The high temporal and spatial resolution water level inversion method of integrating satellite altimetry data and GNSS-R according to claim 1 is characterized in that: In step 2: perform error correction on the satellite altimetry data and calculate the water surface height. The specific formula is as follows: H=H alt -R range -(iono+wet+dry+solid+pole)-Geiod Where, H represents the water level; H alt Indicates the distance from the satellite to the reference ellipsoid; R range Represents the distance from the satellite to the Earth's surface; iono represents the ionospheric correction; wet represents the wet tropospheric correction; dry represents the dry tropospheric correction; solid represents the solid Earth tide correction; pole represents the polar tide correction; Geoid is used to convert the vertical reference from the reference ellipsoid to the geoid.

4. The high temporal and spatial resolution water level inversion method integrating satellite altimetry data and GNSS-R according to claim 1, characterized in that: In step 2, eliminating the system offset by comparing the average water level deviations during the overlapping period of different altimetry satellites specifically includes: comparing the data during the overlapping observation period, calculating the average value and standard deviation of the deviations, and correcting the deviations using linear regression.

5. The high temporal and spatial resolution water level inversion method of integrating satellite altimetry data and GNSS-R according to claim 1, characterized in that: The mathematical model constructed in step 3 is: Where δSNR is the signal-to-noise ratio after removing the direct signal; λ is the satellite signal wavelength; h is the reflection height; θ is the satellite elevation angle; k is the wave number; s is the roughness parameter of the reflecting surface; C1 and C2 are the in-phase and out-of-phase components, which are used to replace the amplitude A and phase After the mathematical model is established, m sampling points with equal intervals are selected and the initial values ​​of the undetermined parameters are determined. The undetermined parameters are set as C1, C2, h, s 2 ,The best fitting parameters are solved by the nonlinear least square method, and the water level height of the sampling point can be calculated through the mathematical model.

Images