An adaptive weighted fusion land water storage anomaly inversion method

Abstract

The application provides a novel adaptive weighted fusion land water storage anomaly inversion method, and relates to the technical field of earth science, and the method comprises the following steps: acquiring global navigation satellite system (GNSS) monitored surface vertical deformation time series data and gravity field data of GRACE / GRACE-FO satellites, and performing pretreatment; constructing a Green function inversion model according to the pretreated GNSS data, and obtaining initial land water storage anomaly TWSAGNSS. The purpose of the application is to provide a novel adaptive weighted fusion land water storage anomaly inversion method, which improves the precision and reliability of land water storage anomaly inversion by fusing GNSS and GRACE / GRACE-FO data characteristics, adaptive weighting algorithm and high space-time resolution technical means.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of earth science, in particular to a self-adaptive weighted fusion land water storage anomaly inversion method. BACKGROUND

[0002] With the continuous development of earth observation technology, various technical means are applied to the monitoring of land water storage anomalies, among which global navigation satellite system and gravity recovery and climate experiment and its follow-up mission GRACE-FO satellite data have become the research focus in this field. Although GNSS technology can provide high time resolution TWSA data, its inversion accuracy is relatively low, mainly affected by factors such as atmospheric load, non-tidal ocean load and common mode error caused by observation means; the spatial resolution of GRACE / GRACE-FO satellite data is relatively low (about 330km), the time resolution is long (monthly scale), and there is a nearly one-year data blank period between GRACE and GRACE-FO, which limits its application in continuous and high-precision TWSA monitoring. SUMMARY

[0003] The technical problem to be solved by the application is to provide a self-adaptive weighted fusion land water storage anomaly inversion method. The self-adaptive weighted fusion land water storage anomaly inversion method and system improve the accuracy and reliability of land water storage anomaly inversion through multi-source data fusion, self-adaptive weighting algorithm and high spatio-temporal resolution technical means.

[0004] To solve the above technical problems, the technical scheme of the application is as follows:

[0005] In a first aspect, a self-adaptive weighted fusion land water storage anomaly inversion method is provided, the method comprising:

[0006] Obtaining global navigation satellite system (GNSS) monitored ground surface vertical deformation time series data and GRACE / GRACE-FO satellite gravity field data, and performing preprocessing;

[0007] According to the preprocessed GNSS data, a Green function inversion model is constructed to obtain an initial land water storage anomaly TWSAGNSS;

[0008] According to the preprocessed GRACE / GRACE-FO data, the data is divided into GRACE period and GRACE-FO period, and the weight proportion of each period is calculated;

[0009] According to the weight proportion of each period of TWSA GNSS and GRACE / GRACE-FO, through adaptive weighted fusion algorithm, iterative optimization is carried out with 0.01 as the step, the GNSS inversion result and the GRACE / GRACE-FO inversion result are linearly combined by weight, and the fused land water storage anomaly TWSAFUSION is obtained;

[0010] According to the adaptive weighted fusion method, the data blank period of GRACE / GRACE-FO is directly filled by TWSA GNSS, so as to obtain the final inversion result with high space-time resolution.

[0011] Further, the global navigation satellite system GNSS monitored ground vertical deformation time series data and the gravity field data of GRACE / GRACE-FO satellite are obtained and preprocessed, including:

[0012] According to the geological and geomorphic conditions of the research area, climate conditions and deformation monitoring requirements, the layout of GNSS monitoring station is planned;

[0013] Through the high-precision GNSS receiver, the satellite signal is continuously received according to the predetermined sampling frequency, and the original observation data including carrier phase and pseudorange observation value are recorded;

[0014] The GNSS original observation data are preprocessed, the baseline solution is carried out through GAMIT / GLOBK software, the baseline vector between adjacent stations is calculated, and the baseline solution result is obtained;

[0015] The baseline solution result is included in the overall adjustment network, and the adjustment calculation is carried out to obtain the accurate coordinates of each GNSS station;

[0016] According to the station coordinates obtained by adjustment calculation, the coordinate change information in the vertical direction of the ground is extracted, and time series analysis is carried out to obtain the ground vertical deformation time series data;

[0017] From the relevant data publishing agencies, the gravity field data products of GRACE / GRACE-FO satellite are obtained according to the research requirements, and preprocessed.

[0018] Further, according to the preprocessed GNSS data, a Green function inversion model is constructed to obtain the initial land water storage anomaly TWSA GNSS, including:

[0019] According to the preprocessed GNSS data, the research area is determined and grid division is carried out;

[0020] Select an earth model to calculate the Green function between each grid node and the GNSS station, and store the Green function;

[0021] The observation equation between GNSS observation data and Green function is established, and the regularization parameter is determined through double summation and integral operation, combination of spherical harmonic function and related parameters, so as to obtain the Green function inversion model;

[0022] According to the Green function inversion model, an inversion algorithm is used to solve the observation equation to obtain an initial land water storage anomaly vector.

[0023] Further, according to the preprocessed GRACE / GRACE-FO data, the data is divided into GRACE period and GRACE-FO period, and the weight proportion of each period is calculated, including:

[0024] According to the preprocessed GRACE / GRACE-FO data, the data is divided into GRACE period and GRACE-FO period;

[0025] Extract the spatial and temporal distribution characteristics and data quality characteristics of the data in each period, and use the GRACE data to calculate the mass change density of a certain point on the ground at a certain time, which is the product of the gravitational constant and the mass change amount and the square of the average distance, and the equivalent water height change amount is the proportion of the mass change density and the reference density and the gravitational acceleration coefficient;

[0026] According to the characteristics of the data in each period, the weight calculation method is selected, the weight value of each period is calculated, the reciprocal of the error index of each period is taken, the weight numerator of the period is obtained, the weight numerators of all periods are added, the normalized denominator is obtained, and the weight proportion is determined.

[0027] Further, according to the weight proportion of TWSA GNSS and GRACE / GRACE-FO in each period, through the adaptive weighted fusion algorithm, the step is 0.01 for iterative optimization, the GNSS inversion result and the GRACE / GRACE-FO inversion result are weighted and linearly combined, and the fused land water storage anomaly TWSAFUSION is obtained, including:

[0028] According to the weight proportion of TWSA GNSS and GRACE / GRACE-FO in each period, the weight proportion range for weighted fusion is set, and the step for iterative optimization is determined as 0.01;

[0029] Set the initial value of the iteration number j=0, and initialize the variable for storing the fusion result;

[0030] The weighted combination value under the current weight is calculated, and for each iteration number j, the weighted linear combination value of TWSA GNSS and TWSA in different time periods is calculated according to the current weight wj, the total water storage anomaly data of different sources are fused together by linear weighting to obtain the fused land water storage anomaly TWSA FUSION.

[0031] Further, according to the adaptive weighted fusion method, the TWSA GNSS is directly filled in the GRACE / GRACE-FO data blank period to obtain the final inversion result with high space-time resolution, including:

[0032] According to the adaptive weighted fusion method, the TWSA GNSS is directly filled in the GRACE data blank period by using the determined optimal weight to obtain the fused TWSA data sequence;

[0033] According to the fused TWSA data sequence, the linear correlation is evaluated by using the Pearson correlation coefficient, the covariance of the fused sequence and the reference sequence is calculated and divided by the product of their standard deviations to obtain the root mean square error, which is used to evaluate the error between the fused sequence and the reference sequence, the average value of the sum of squares of the difference of each corresponding element is calculated, and then the root mean square error is obtained;

[0034] The TWSA result obtained by the AWFTM inversion is compared with the GRACE, traditional GNSS and GLDAS results, a comprehensive evaluation value is obtained by calculating the ratio of the Pearson correlation coefficient and the root mean square error and taking the square root, and the accuracy evaluation result is obtained;

[0035] Through the accuracy evaluation result, the final inversion result with high space-time resolution based on the AWFTM is output.

[0036] The second aspect is an adaptive weighted fusion land water storage anomaly inversion system, comprising:

[0037] The acquisition module is configured to acquire global navigation satellite system (GNSS) monitored ground surface vertical deformation time series data and GRACE / GRACE-FO satellite gravity field data, and perform preprocessing;

[0038] The construction module is configured to construct a Green function inversion model according to the preprocessed GNSS data, and obtain an initial land water storage anomaly TWSA GNSS.

[0039] The weight calculation module is configured to divide the preprocessed GRACE / GRACE-FO data into GRACE period and GRACE-FO period, and calculate the weight proportion of each period respectively.

[0040] An adaptive weighted fusion algorithm module is configured to perform iterative optimization by adaptive weighted fusion algorithm with a step of 0.01 according to the weight proportion of each period of TWSA GNSS and GRACE / GRACE-FO, and to obtain the fused land water storage anomaly TWSAFUSION by weighted linear combination of GNSS inversion results and GRACE / GRACE-FO inversion results.

[0041] A processing module is configured to directly fill in the data blank period of GRACE / GRACE-FO by TWSA GNSS according to the adaptive weighted fusion method, so as to obtain the final inversion result with high spatio-temporal resolution.

[0042] In a third aspect, a computing device includes:

[0043] One or more processors;

[0044] A storage device is configured to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0045] In a fourth aspect, a computer readable storage medium stores a program, and when the program is executed by a processor, the method is implemented.

[0046] The above scheme of the present application at least has the following beneficial effects:

[0047] By combining the GNSS monitored ground vertical deformation data and the gravity field data of GRACE / GRACE-FO satellite, the advantages of different data sources are fully utilized, and the accuracy and reliability of the inversion result are improved; by adaptive weighted fusion algorithm with a step of 0.01 for iterative optimization, the weight of different data sources can be dynamically adjusted, so that the fusion result is more scientific and reasonable, and adapts to the change of different periods and data quality; by directly filling in the data blank period of GRACE / GRACE-FO using TWSA GNSS, the final inversion result with high spatio-temporal resolution can be obtained, which is of great significance for monitoring and evaluating the change of land water storage. The accuracy and reliability of the inversion result are ensured by precision evaluation of the fusion result through Pearson correlation coefficient and root mean square error evaluation indexes; the system is designed as multiple modules, including an acquisition module, a construction module, a weight calculation module, an adaptive weighted fusion algorithm module and a processing module, which has clear structure and is easy to implement and maintain. The adaptive weighted fusion land water storage anomaly inversion method and system improve the accuracy and reliability of land water storage anomaly inversion through multi-source data fusion, adaptive weighted algorithm and high spatio-temporal resolution technical means. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1A flowchart of a self-adaptive weighted fusion land water storage anomaly inversion method provided by an embodiment of the present application is shown.

[0049] Figure 2 A self-adaptive weighted fusion land water storage anomaly inversion system diagram provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0050] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0051] As shown in Figure 1 An embodiment of the present application proposes a self-adaptive weighted fusion land water storage anomaly inversion method, which comprises the following steps:

[0052] Step 11, obtaining global navigation satellite system (GNSS) monitored surface vertical deformation time series data and gravity field data of GRACE / GRACE-FO satellites, and performing preprocessing;

[0053] Step 12, constructing a Green function inversion model according to the preprocessed GNSS data to obtain an initial land water storage anomaly TWSAGNSS;

[0054] Step 13, dividing the preprocessed GRACE / GRACE-FO data into GRACE and GRACE-FO periods, and calculating the weight proportion of each period respectively;

[0055] Step 14, according to the weight proportion of each period of TWSAGNSS and GRACE / GRACE-FO, through a self-adaptive weighted fusion algorithm, iterative optimization is performed with a step of 0.01, the GNSS inversion result is weighted and linearly combined with the GRACE / GRACE-FO inversion result, to obtain a fused land water storage anomaly TWSAFUSION;

[0056] Step 15, according to the self-adaptive weighted fusion method, using TWSAGNSS to directly fill in the blank period of GRACE / GRACE-FO data, to obtain a final inversion result with high spatio-temporal resolution.

[0057] In the embodiment of the present application, the advantages of different data sources are integrated by obtaining the ground vertical deformation time series data monitored by the global navigation satellite system and the gravity field data of the GRACE / GRACE-FO satellite, and fusing the two; the discontinuity problem of the GRACE / GRACE-FO satellite data is effectively solved by filling the blank period of the GRACE / GRACE-FO data with TWSA GNSS, and the continuous land water storage anomaly data sequence is obtained by filling the blank period, which provides a more reliable data basis for in-depth research on the dynamic change of land water storage. The adaptive weighted fusion algorithm is adopted, and the GNSS inversion result and the GRACE / GRACE-FO inversion result are linearly combined with a step of 0.01 for iterative optimization, so that the weight proportion is automatically adjusted according to different data sources, and the fusion result can more accurately reflect the true situation of the land water storage. The initial land water storage anomaly TWSAGNSS is obtained by constructing the Green function inversion model according to the pretreated GNSS data, which can fully consider the complex relationship between the ground mass load and the crustal deformation; the final inversion result with high space-time resolution can more finely depict the space-time variation characteristics of the land water storage, in time, the high-resolution data can capture the short-term fluctuation and seasonal change of the land water storage, and provide timely and effective information for water resource management, flood warning and the like; in space, the high-resolution data can reveal the distribution difference of the land water storage in different regions, which is helpful to deeply understand the regional hydrological cycle process and the mechanism of ecological environment change.

[0058] In a preferred embodiment of the present application, the above step 11 can include:

[0059] Step 111, according to the geological and geomorphological conditions of the research area, climate conditions and deformation monitoring requirements, the layout of the GNSS monitoring station is planned;

[0060] Step 112, the original observation data including carrier phase and pseudorange observation values are recorded by continuously receiving satellite signals according to the predetermined sampling frequency through the high-precision GNSS receiver;

[0061] Step 113, the GNSS original observation data are preprocessed, the baseline vectors between adjacent stations are calculated by GAMIT / GLOBK software for baseline solution, and the baseline solution result is obtained;

[0062] Step 114, the baseline solution result is included in the overall adjustment network, and the adjustment calculation is performed to obtain the accurate coordinates of each GNSS station;

[0063] Step 115, according to the station coordinates obtained by adjustment calculation, the coordinate change information in the vertical direction of the ground is extracted, and time series analysis is performed to obtain the ground vertical deformation time series data;

[0064] Step 116, according to the research needs, obtain the gravity field data products of GRACE / GRACE-FO satellites from relevant data publishing agencies and perform preprocessing.

[0065] In the embodiment of the application, the GNSS monitoring station layout is planned according to the geological and geomorphic conditions, climate conditions and deformation monitoring requirements of the research area, which can make the monitoring stations more reasonably distributed in key areas. Through scientific planning, the limited resources are maximized under the premise of meeting the monitoring requirements, and the efficiency of monitoring work is improved; using high-precision GNSS receivers to continuously receive satellite signals at a predetermined sampling frequency and record raw observation data can obtain high-quality and high-precision raw data; through GAMIT / GLOBK software for baseline solution and overall adjustment calculation, the professional algorithms and powerful functions of the software can be fully utilized to effectively eliminate errors and interference factors in the observation data. Time series analysis of the station coordinates obtained by adjustment calculation can extract coordinate change information in the vertical direction of the earth's surface, and the vertical deformation time series data of the earth's surface can be obtained. The gravity field data products of GRACE / GRACE-FO satellites are obtained from relevant data publishing agencies, and these data products have been processed and verified by professional agencies, and have high authority and reliability. Preprocessing of the obtained GRACE / GRACE-FO satellite gravity field data products can eliminate noise and errors in the data and improve the quality and consistency of the data.

[0066] In the embodiment of the application, the specific steps include:

[0067] Step 111, collect geological structure maps, topographic maps, soil type distribution maps and other information of the research area, obtain meteorological data of the research area including long-term observation data of precipitation, temperature, wind speed, etc., analyze the seasonal and interannual variation rules of climate change, and determine the deformation monitoring focus and target of the research area. According to the geological and geomorphic conditions, climate conditions and deformation monitoring requirements of the research area, adjust and optimize the station layout, determine the final GNSS monitoring station layout scheme, and draw the station layout map, mark the position, number and other information of each station.

[0068] Step 112, select high-precision GNSS receivers according to the monitoring accuracy and sampling frequency requirements. According to the monitoring requirements and research purposes, set the sampling frequency, for example, for projects that require high-precision monitoring, set a higher sampling frequency, start the receiver, continuously receive satellite signals at a predetermined sampling frequency, and record raw observation data including carrier phase and pseudorange observation values.

[0069] Step 113, import the GNSS raw observation data into the GAMIT / GLOBK software, check the integrity and correctness of the data, check whether there are missing values, outliers and other problems in the data, set the parameters for baseline solution in the GAMIT / GLOBK software, such as satellite orbit model, atmospheric delay model and the like, select the parameters according to the characteristics and data of the research area, run the software, and perform baseline solution to calculate the baseline vectors between adjacent stations, and obtain the baseline solution results, including baseline length, azimuth angle and the like.

[0070] Step 114, incorporate the baseline solution results into the overall adjustment network, construct an adjustment model, run the adjustment calculation program, and obtain the accurate coordinates of each GNSS station.

[0071] Step 115, according to the accurate coordinate data of each GNSS station, extract the coordinate change information in the vertical direction of the ground surface, calculate the vertical displacement of each station in different time periods, and pre-process the extracted vertical displacement data, such as removing outliers, filling missing values and the like, analyze the vertical displacement data, and obtain the vertical deformation time series data of the ground surface.

[0072] Step 116, according to the research area and research purpose, determine the type and time range of the GRACE / GRACE-FO satellite gravity field data product required to be obtained, download the required GRACE / GRACE-FO satellite gravity field data product from the website of the relevant data publishing agency, such as NASA Earth Science Data System, convert the downloaded data into the format for subsequent processing, for example, convert the data into the format readable by the general scientific computing software, remove the noise and errors in the data, such as orbit error, instrument noise and the like, by using filtering and the like, if there are missing values in the data, fill them by using a suitable interpolation method.

[0073] In a preferred embodiment of the present application, the above-mentioned step 12 can comprise:

[0074] Step 121, according to the pre-processed GNSS data, determine the research area and perform grid division;

[0075] Step 122, select an earth model, calculate the Green function between each grid node and the GNSS station, and store the Green function;

[0076] Step 123, establish the observation equation between the GNSS observation data and the Green function, determine the regularization parameter by double summation and integral operation, combined with spherical harmonics and related parameters, to obtain the Green function inversion model;

[0077] Step 124, according to the Green function inversion model, solve the observation equation by using an inversion algorithm to obtain the initial land water storage anomaly vector.

[0078] In the embodiments of the present application, by determining the study area according to the pretreated GNSS data, the study range can be accurately defined, the study area is divided into grids, the continuous geographic space is discretized, and each grid node can be analyzed in detail. After selecting an earth model and calculating the Green function between each grid node and the GNSS station, the Green function is stored, which avoids repeated calculation of the Green function in the inversion process, greatly improves the calculation efficiency, and directly calls the stored Green function in the subsequent inversion calculation, saves the calculation time, and stores the Green function, which is helpful for data management and maintenance. According to the corresponding relationship between the grid nodes and the GNSS stations, a suitable data structure can be used for storage; the observation equation between the GNSS observation data and the Green function is established, the physical model is combined with the actual observation data, the inversion model is more scientific and reliable. The Green function reflects the physical relationship between the internal mass change of the earth and the surface deformation, which is connected with the GNSS observation data through the observation equation, and can more accurately describe the internal relationship between the land water storage anomaly and the surface deformation; the regularization parameter can balance the fitting accuracy of the observation data and the smoothness of the model, avoid overfitting or underfitting, and make the inversion result more reasonable. According to the Green function inversion model, the initial land water storage anomaly vector is obtained by solving the observation equation using the inversion algorithm, which can accurately reflect the initial abnormal distribution of the land water storage in the study area.

[0079] In the embodiments of the present application, the specific steps include:

[0080] Step 121, according to the distribution of the GNSS station, the geographical range covered by the data is determined. The coverage range of the data can be intuitively understood by drawing a station distribution map; combined with the specific research target, such as the land water storage change of a specific watershed, urban area or geological structure unit, the specific boundary of the study area is further determined on the basis of the data coverage range. According to the characteristics and requirements of the study area, the grid type is selected, such as regular grid (rectangular grid) or irregular grid (triangular grid), and for most land water storage inversion research, regular grid is more commonly used. According to the requirements of research accuracy and calculation resources, the interval of the grid is determined; according to the determined grid type and interval, the coordinates of the grid nodes in the study area are generated, and the generation of the grid nodes can be realized by using geographic information system (GIS) software or programming language (such as Python).

[0081] Step 122, according to the geological conditions of the study area, research purposes and other factors, select the earth model; select the Green function calculation method, such as analytical method, numerical method, etc., for complex earth model and geometric shape, numerical method is more commonly used; use programming language to write Green function calculation program, input earth model parameters, grid node coordinates and GNSS station coordinates, calculate the Green function between each grid node and GNSS station; according to the needs of subsequent use, select the storage format, such as text file, binary file, etc.

[0082] Step 123, according to the physical principle, analyze the mathematical relationship between GNSS observation data (such as surface vertical deformation) and Green function, for example, surface vertical deformation Can be expressed as the convolution of mass load And Green function, Wherein, And Respectively represent the average radius and mass of the earth, Indicates the load Love number, Indicates the order, Indicates the angular distance between GNSS station and mass load, Indicates the Legendre polynomial, Indicates the earth radius, Indicates the distance from the station to the center of mass load, Indicates the mass load change corresponding to the order and number. Establish the general form of observation equation, substitute the calculated Green function and preprocessed GNSS observation data into the observation equation, and obtain the specific observation equation set; The regularization parameter can be determined by using empirical method, cross validation method and other methods, and the determined observation equation and regularization parameter are integrated together to obtain the complete Green function inversion model.

[0083] Step 124, according to the characteristics and scale of Green function inversion model, select inversion algorithm, for example, for large-scale inversion problem, conjugate gradient method has high calculation efficiency; Use programming language (such as Python, Matlab, etc.) to write inversion algorithm program, in the program, need to call Green function inversion model and observation data. Set the iteration conditions of the inversion algorithm, such as convergence accuracy, maximum iteration number, etc. Input the Green function inversion model and observation data into the inversion program, run the program to solve the observation equation, and get the initial land water storage anomaly vector after the program runs.

[0084] In a preferred embodiment of the present application, the above step 13 can include:

[0085] Step 131, according to the preprocessed GRACE / GRACE-FO data, divide into GRACE period and GRACE-FO period;

[0086] Step 132, extract the spatio-temporal distribution characteristics and data quality characteristics of each period data, the mass change density of a certain point on the ground at a certain time is the product of the gravitational constant and the mass change amount divided by the square of the average distance, and the equivalent water height change amount is the ratio of the mass change density to the reference density and the gravity acceleration proportional coefficient;

[0087] Step 133, according to the characteristics of each period data, select the weight calculation method, calculate the weight value of each period, take the reciprocal of the error index of each period to obtain the weight numerator of the period, add the weight numerators of all periods to obtain the normalized denominator, and determine the weight proportion.

[0088] In the embodiment of the application, the preprocessed GRACE / GRACE-FO data is divided into GRACE period and GRACE-FO period, which can clearly distinguish the characteristics of different satellite observation data, and can provide more targeted data support for the research of land water storage change in different time periods; the spatio-temporal distribution characteristics of each period data can reveal the change rule of land water storage in different time and space; the data quality characteristics can evaluate the accuracy and reliability of the data. According to the characteristics of each period data, the weight calculation method is selected, and the weight value of each period is calculated for normalization processing, which can fully consider the difference of data in different periods; through the weight calculation, the important period data can be highlighted, and the influence of low quality period data on the fusion result can be reduced. Accurate land water storage data is the key to establish and improve the land water cycle model. Through the research on land water storage data in different periods, the interaction mechanism can be revealed, and for water resource management department, accurate land water storage information is an important basis for formulating water resource management and planning scheme.

[0089] In the embodiment of the application, the specific steps include:

[0090] Step 131, collect the gravity field data of the preprocessed GRACE and GRACE-FO satellites, check whether the data has missing, abnormal value and other problems. Determine the observation time range of GRACE satellite and GRACE-FO satellite, according to the launch, operation and end time of the satellite, determine the start and end time of GRACE period and GRACE-FO period; use data processing software (such as Python pandas library) or programming language, according to the determined time boundary, divide the preprocessed data into GRACE period data and GRACE-FO period data, and store them as different files or data structures.

[0091] Step 132: For each time period, calculate the time series statistical indicators of terrestrial water storage anomalies, such as mean, standard deviation, maximum, and minimum values. Use methods such as linear regression and moving average to analyze the changing trend of terrestrial water storage anomalies over time. GRACE data is used to estimate the mass changes in the surface and surrounding areas through complex geophysical model inversion calculations, and to determine the relationship between mass changes and changes in the gravitational field. ,in, Indicates geographical location and time At that location, the density of mass change corresponding to the gravitational field change observed by GRACE. Represents the gravitational constant. Indicates geographical location and time The change in mass of the Earth's surface and surrounding area. Represents the distance from the Earth's center of mass to its geographical location. The average distance at a given location varies depending on geographical location, as the Earth is not a perfect sphere. Changes in mass are converted into equivalent water height. ,in, Indicates geographical location and time The equivalent water level change at the location. This represents the density of mass change corresponding to the gravitational field changes observed by GRACE. It is a constant, representing a certain reference density. It is gravitational acceleration, a physical constant. Using Geographic Information System (GIS) software or data visualization tools, spatial distribution maps of land water storage anomalies for each time period are created, visually displaying the anomalies in land water storage across different regions. Based on error information provided by the data products, error indices for each time period are calculated, and the main sources of error, such as satellite orbit errors, instrument noise, and data processing errors, are analyzed to understand the extent of their impact on the data. The number and proportion of missing values ​​in the data for each time period are statistically analyzed, and their temporal and spatial distribution is examined.

[0092] Step 133: Based on the spatiotemporal distribution characteristics and data quality characteristics of the data for each time period, such as data accuracy, completeness, and trends, select a weighting calculation method. If there are significant differences in data quality, a weighting calculation method based on error indices can be used; for example, the smaller the error, the larger the weight. If data from certain time periods is more important to the research results (such as data during key climate events), different weights can be assigned according to the importance of the data. Determine the specific calculation formula based on the selected weighting calculation method. For example, if an error-based weighting calculation method is used, weights can be set... ,in For the first an error indicator for each time period, denotes the total number of time periods, denotes the sum of the reciprocals of all time period error indicators, used for normalizing weights to ensure that the sum of all weights is 1.

[0093] In a preferred embodiment of the present application, the above step 14 can include:

[0094] Step 141, according to the weight proportion of each time period of TWSAGNSS and GRACE / GRACE-FO, set the weight proportion range for weighted fusion, determine the step size of iterative optimization as 0.01;

[0095] Step 142, set the initial value of the iteration number j=0, and initialize the variable for storing the fusion result;

[0096] Step 143, calculate the weighted combination value under the current weight, for each iteration number j, calculate the weighted linear combination value of TWSAGNSS and TWSA_{GRACE / GRACE-FO} in different time periods according to the current weight wj, and through linear weighting, the land water storage anomaly data of different sources are fused together to obtain the fused land water storage anomaly TWSAFUSION.

[0097] In the embodiment of the present application, by setting the range of the weight proportion of each time period of TWSAGNSS and GRACE / GRACE-FO, and iterating optimization with a step size of 0.01, the contribution of different data sources in the fusion process can be finely adjusted, thereby improving the accuracy and reliability of the fusion result; at the same time, by initializing the iteration number and the storage variable, and calculating the weighted combination value under the current weight, the effective fusion of land water storage anomaly data of different sources is realized, and a more comprehensive and accurate land water storage anomaly TWSAFUSION is obtained. This method not only improves the precision of data fusion, but also provides more reliable data support for subsequent water resource management and environmental change research.

[0098] In the embodiment of the present application, the specific steps include:

[0099] Step 141, according to the weight proportion of each time period of TWSAGNSS and GRACE / GRACE-FO, determine the weight proportion range for weighted fusion, for example, the weight proportion range can be set from 0 to 1. Determine the step size of iterative optimization as 0.01.

[0100] Step 142, set the initial value of the iteration number j=0, initialize the variable for storing the fusion result, for example, an array or list can be created to store the TWSAFUSION value obtained in each iteration.

[0101] Step 143, for each iteration number j, the weighted linear combination values of TWSA GNSS and TWSA_{GRACE / GRACE-FO} in different time periods are calculated according to the current weight wj. The segmented weighted linear combination formula is , , , , , , , , , , , , , , , , ,

[0102] In a preferred embodiment of the present application, the above step 15 can comprise:

[0103] Step 151, according to the adaptive weighted fusion method, the determined optimal weight is used to directly fill in the blank period of GRACE data with TWSA GNSS to obtain the fused TWSA data sequence;

[0104] Step 152, according to the fused TWSA data sequence, the linear correlation is evaluated by using Pearson correlation coefficient, the covariance of the fused sequence and the reference sequence is calculated and divided by the product of their standard deviations, the root mean square error is used to evaluate the error between the fused sequence and the reference sequence, the average value of the square sum of each corresponding element difference is calculated, and then the root mean square error is obtained;

[0105] Step 153, the TWSA result obtained based on AWFTM inversion is compared with the GRACE, traditional GNSS and GLDAS results, a comprehensive evaluation value is obtained by calculating the ratio of the Pearson correlation coefficient and the root mean square error and taking the square root, and the accuracy evaluation result is obtained;

[0106] Step 154, according to the accuracy evaluation result, the final inversion result of high space-time resolution based on AWFTM is output.

[0107] In the embodiment of the application, the TWSA GNSS data is used to directly fill in the blank period of the GRACE data, ensuring the continuity and integrity of the data, and through the adaptive weighted fusion method, the advantages of different data sources are combined to obtain a higher quality fused TWSA data sequence; the Pearson correlation coefficient is used to evaluate the linear correlation between the fused sequence and the reference sequence, and a quantitative correlation index is provided; the root mean square error (RMSE) is used to evaluate the error between the fused sequence and the reference sequence, and a quantitative error index is provided. Combined with the Pearson correlation coefficient and the root mean square error, a comprehensive evaluation value (SEC) is obtained, which more comprehensively evaluates the accuracy of the fused data, can determine the best fusion proportion, optimizes the data fusion effect, and improves the accuracy of the inversion result. The TWSA result obtained based on AWFTM inversion is compared with the GRACE, traditional GNSS and GLDAS results, the reliability and accuracy of the inversion result are verified, through comparison and accuracy evaluation, the final inversion result of high space-time resolution based on AWFTM is finally output.

[0108] In the embodiment of the application, the specific steps include:

[0109] Step 151, the adaptive weighted fusion method (AWFTM) is used to determine the optimal weight of each data source (such as TWSA GNSS and GRACE), and the TWSA GNSS data is used to fill in the blank period of the GRACE data. According to the determined optimal weight, the TWSA GNSS data and the GRACE data are fused to obtain a fused TWSA data sequence.

[0110] Step 152, according to the fused TWSA data sequence, the linear correlation between the fused sequence and the reference sequence GLDAS is evaluated by using Pearson correlation coefficient, the weight ratio when the PCC value of GRACE / GRACE-FO two periods is maximum is judged, and the formula is as follows: wherein, () represents the covariance, represents the reference sequence of GLDAS corresponding to the GRACE / GRACE-FO period, and represents the standard deviation of the sequence and represents a specific time point. The root mean square error is used to evaluate the error between the fused sequence and the reference sequence, so as to judge the weight ratio when the RMSE value of GRACE / GRACE-FO two periods is minimum, and the formula is as follows: wherein, represents the root mean square error at time point , the length of the sequence (GRACE period is 2372, and GRACE-FO period is 1613), and represent the first element of the previous and later fused sequences, respectively, and represent the first element of the previous and later reference sequences, respectively, represents the index of data points, ranging from 1 to .

[0111] Step 153, the TWSA result obtained based on AWFTM inversion is compared with the results of GRACE, traditional GNSS and GLDAS, and the consistency of the compared data in time and space is ensured. The fused sequence is comprehensively evaluated, the best fusion proportion is more intuitively evaluated, a SEC index is defined, which combines RMSE and PCC, and the best fusion proportion is determined, and the formula is as follows: wherein represents the comprehensive evaluation value at time point , according to the size of the comprehensive evaluation value (SEC), the precision of the inversion result is evaluated, and the smaller the SEC value is, the more accurate the inversion result is.

[0112] Step 154, through the precision evaluation result, the reliability and accuracy of the inversion result are verified, and the final inversion result of high spatio-temporal resolution based on AWFTM is output.

[0113] As Figure 2 ​As shown, the embodiment of the present application also provides a self-adaptive weighted fusion land water storage anomaly inversion system 20, comprising:

[0114] An acquisition module 21 is configured to acquire global navigation satellite system (GNSS) monitored surface vertical deformation time series data and gravity field data of GRACE / GRACE-FO satellites, and perform preprocessing;

[0115] A construction module 22 is configured to construct a Green function inversion model according to the preprocessed GNSS data, to obtain an initial land water storage anomaly TWSAGNSS;

[0116] A weight calculation module 23 is configured to divide the preprocessed GRACE / GRACE-FO data into GRACE periods and GRACE-FO periods, and calculate the weight proportion of each period, respectively;

[0117] A self-adaptive weighted fusion algorithm module 24 is configured to obtain a fused land water storage anomaly TWSAFUSION by performing iterative optimization with a step of 0.01 through a self-adaptive weighted fusion algorithm, and performing weighted linear combination of the GNSS inversion result and the GRACE / GRACE-FO inversion result, according to the TWSAGNSS and the weight proportion of each period of GRACE / GRACE-FO;

[0118] A processing module 25 is configured to directly fill in the blank period of the GRACE / GRACE-FO data by using the TWSAGNSS according to the self-adaptive weighted fusion method, to obtain a final inversion result with high spatio-temporal resolution.

[0119] It should be noted that the system is a system corresponding to the above method, and all implementation manners in the above method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0120] The above is the preferred embodiment of the present application, and it should be noted that for ordinary skilled persons in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered as the protection scope of the present application.

Claims

1. An adaptive weighted fusion method for inverting terrestrial water storage anomalies, characterized in that, The method includes: This study acquires and preprocesses time-series data of vertical surface deformation from GNSS monitoring and gravity field data from the GRACE / GRACE-FO satellites. The preprocessing includes: planning the layout of GNSS monitoring stations based on the geological features, climate conditions, and deformation monitoring requirements of the study area; continuously receiving satellite signals at a predetermined sampling frequency using high-precision GNSS receivers to record raw observation data, including carrier phase and pseudorange observations; preprocessing the raw GNSS observation data by performing baseline calculations using GAMIT / GLOBK software to calculate baseline vectors between adjacent stations; incorporating the baseline calculation results into the overall adjustment network for adjustment calculations to obtain the precise coordinates of each GNSS station; extracting vertical coordinate change information from the adjusted station coordinates and performing time-series analysis to obtain time-series data of vertical surface deformation; and acquiring gravity field data products from the GRACE / GRACE-FO satellites from relevant data publishing agencies according to research needs, and preprocessing them. Based on the preprocessed GNSS data, a Green's function inversion model is constructed to obtain the initial land water storage anomaly (TWSAGNSS). This process includes: determining the study area and dividing it into grids based on the preprocessed GNSS data; selecting an Earth model, calculating and storing the Green's function between each grid node and the GNSS station; establishing the observation equation between the GNSS observation data and the Green's function, and determining the regularization parameter through double summation and integration operations, combined with spherical harmonics and related parameters, to obtain the Green's function inversion model; and using an inversion algorithm to solve the observation equation based on the Green's function inversion model to obtain the initial land water storage anomaly vector. Based on the preprocessed GRACE / GRACE-FO data, the data is divided into GRACE and GRACE-FO time periods. The weight ratios for each time period are calculated, including: dividing the data into GRACE and GRACE-FO time periods; extracting the spatiotemporal distribution and data quality characteristics of each time period; using GRACE data to calculate the mass change density at a certain point on the Earth's surface at a certain time as the ratio of the product of the gravitational constant and the mass change to the square of the average distance; and calculating the equivalent water height change as the mass change density proportional to the reference density and gravitational acceleration coefficient. Based on the characteristics of the data in each time period, a weight calculation method is selected to calculate the weight value for each time period. The reciprocal of the error index for each time period is taken to obtain the weight numerator for that time period. The weight numerators of all time periods are summed to obtain the normalized denominator, thus determining the weight ratio. Based on the weight ratios of TWSAGNSS and GRACE / GRACE-FO at different time periods, an adaptive weighted fusion algorithm is used to iteratively optimize the results with a step size of 0.

01. The GNSS inversion results are then combined with the GRACE / GRACE-FO inversion results in a weighted linear combination to obtain the fused terrestrial water storage anomaly TWSAFUSION. Based on the adaptive weighted fusion method, TWSAGNSS is used to directly fill the gaps in the GRACE / GRACE-FO data to obtain the final inversion results with high spatiotemporal resolution.

2. The adaptive weighted fusion method for inverting terrestrial water storage anomalies according to claim 1, characterized in that, Based on the weight ratios of TWSAGNSS and GRACE / GRACE-FO at different time periods, an adaptive weighted fusion algorithm is used to iteratively optimize the results with a step size of 0.

01. The GNSS inversion results are then combined linearly with the GRACE / GRACE-FO inversion results to obtain the fused terrestrial water storage anomaly TWSAFUSION, which includes: Based on the weight ratios of TWSAGNSS and GRACE / GRACE-FO at different time periods, the range of weight ratios used for weighted fusion is set, and the step size for iterative optimization is determined to be 0.

01. Set the initial number of iterations j=0, and initialize the variables used to store the fusion results; Calculate the weighted combination value under the current weight. For each iteration number j, calculate the weighted linear combination value of TWSAGNSS and TWSA_{GRACE / GRACE-FO} at different time periods based on the current weight wj. By using linear weighting, the terrestrial water storage anomaly data from different sources are fused together to obtain the fused terrestrial water storage anomaly TWSAFUSION.

3. The adaptive weighted fusion method for inverting terrestrial water storage anomalies according to claim 2, characterized in that, Based on the adaptive weighted fusion method, TWSAGNSS is used to directly fill the gaps in the GRACE / GRACE-FO data to obtain the final inversion results with high spatiotemporal resolution, including: According to the adaptive weighted fusion method, using the determined optimal weights, TWSAGNSS is used to directly fill the gaps in the GRACE data, resulting in the fused TWSA data sequence. Based on the fused TWSA data sequence, the linear correlation is assessed using the Pearson correlation coefficient, which is obtained by calculating the covariance of the fused sequence and the reference sequence and dividing it by the product of their standard deviations. The root mean square error is used to assess the error between the fused sequence and the reference sequence, which is obtained by calculating the average of the sum of squares of the differences between each corresponding element and then taking the square root. The TWSA results obtained based on AWFTM inversion are compared with the results of GRACE, traditional GNSS and GLDAS. By calculating the ratio of Pearson correlation coefficient to root mean square error and taking the square root, a comprehensive evaluation value is obtained, and the accuracy assessment result is obtained. Based on the accuracy evaluation results, the final inversion results with high spatiotemporal resolution obtained by AWFTM are output.

4. An adaptive weighted fusion land water storage anomaly inversion system, the system implementing the method as described in any one of claims 1 to 3, characterized in that, include: The acquisition module is used to acquire time series data of vertical deformation of the Earth's surface monitored by the Global Navigation Satellite System (GNSS) and gravity field data of the GRACE / GRACE-FO satellites, and to perform preprocessing. The module is used to construct a Green's function inversion model based on the preprocessed GNSS data to obtain the initial land water storage anomaly TWSAGNSS. The weight calculation module is used to divide the preprocessed GRACE / GRACE-FO data into GRACE time periods and GRACE-FO time periods, and calculate the weight ratio of each time period. The adaptive weighted fusion algorithm module is used to perform iterative optimization with a step size of 0.01 based on the weight ratio of TWSAGNSS and GRACE / GRACE-FO at each time period. It combines the GNSS inversion results with the GRACE / GRACE-FO inversion results in a weighted linear combination to obtain the fused terrestrial water storage anomaly TWSAFUSION. The processing module is used to directly fill the gaps in the GRACE / GRACE-FO data using the TWSAGNSS method based on the adaptive weighted fusion method, so as to obtain the final inversion result with high spatiotemporal resolution.

5. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 3.

Images