A high-precision meteorological early warning system based on water vapor tomography inversion

Abstract

The present invention relates to the technical field of meteorological forecasting, and discloses a high-precision meteorological early warning system based on water vapor tomography inversion, including: a data acquisition module for acquiring multi-sensor meteorological observation data; a data preprocessing module for uniformly converting the multi-sensor meteorological observation data into a three-dimensional grid rectangular coordinate system and completing time synchronization and denoising processing; a wet refractive index field construction module for constructing an initial wet refractive index field based on GNSS data and geostationary satellite data; a dynamic optimization module for performing spatiotemporal dynamic optimization of the wet refractive index field based on partial differential dynamic equations. By adopting a partial differential dynamic equation model to dynamically optimize the wet refractive index field, a technical effect of capturing the law of change of the wet refractive index field over time and space is achieved. Compared with the problem in the prior art that the static optimization model cannot reflect the spatiotemporal evolution characteristics of water vapor under complex meteorological conditions, the shortcomings of poor adaptability to dynamic changes of the wet refractive index field and low prediction accuracy are solved.

Description

Technical Field

[0001] The invention relates to the technical field of weather forecasting, and in particular to a water vapor tomography inversion high-precision weather early warning system. Background Art

[0002] Although the content of water vapor in the atmosphere is relatively small, it is an important factor in the global energy balance. Due to the phase change effect of water vapor, it has various states such as rain, fog, cloud, and snow. It is an important factor causing weather changes and is also the main driving force for the formation and evolution of disastrous weather. In weather forecasting, the understanding of the water vapor content, changes in water vapor, and temporal and spatial distribution in the atmosphere determines the success or failure of short-term forecasting. The analysis of the humidity field in the atmosphere will directly affect the accuracy of numerical forecasting. The wet refractive index field is an important basic data in meteorology, which is used to describe the distribution and dynamic changes of water vapor in the atmosphere. By accurately inverting the wet refractive index field, it can provide support for weather forecasting, severe convective weather warnings, typhoon path prediction, and water vapor transport research.

[0003] The current construction of wet refractivity field mainly relies on a single data source, such as GNSS observation data or geostationary satellite data. However, although GNSS data has a high temporal resolution, its spatial coverage is limited, especially in complex terrain or data blind areas where it is difficult to obtain enough observation points. Although geostationary satellite data has wide-area coverage capabilities, its spatial resolution is low, and the observation accuracy is significantly affected by atmospheric conditions.

[0004] The temporal and spatial distribution of the wet refractive index field is affected by factors such as water vapor transport, convective activity and local evaporation, and has significant dynamic change characteristics. However, existing technologies mostly use static modeling methods to directly interpolate or invert the field based on observational data. This method ignores the temporal variation characteristics of the wet refractive index field, resulting in insufficient dynamic response capabilities of the wet refractive index field under rapidly changing weather systems such as severe convective weather and short-term heavy rainfall, and the inversion results lag behind the actual changes. Especially in the process of the formation or disappearance of the hygrocline, the static modeling method cannot capture key features in a timely manner, reducing the accuracy of early warning of extreme weather events.

[0005] In view of the deficiencies in the prior art, the present invention provides a high-precision meteorological early warning system based on water vapor tomography inversion. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a high-precision meteorological warning system for water vapor tomography inversion, which solves the problem that the current construction of the wet refractive index field mainly relies on a single data source, GNSS observation data or geostationary satellite data. Although GNSS data has a high temporal resolution, its spatial coverage is limited.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a water vapor tomography inversion high-precision meteorological early warning system, comprising: Data acquisition module, used to collect multi-sensor meteorological observation data; Data preprocessing module, used to uniformly convert multi-sensor meteorological observation data into a three-dimensional grid rectangular coordinate system and complete time synchronization and denoising processing; A wet refractive index field construction module is used to construct an initial wet refractive index field based on GNSS data and geostationary satellite data; Dynamic optimization module, used for spatiotemporal dynamic optimization of the wet refractive index field based on partial differential dynamics equations; A spatial optimization module for spatial interpolation and physical constraint correction of the wet refractive index field using a variational optimization method; The fusion correction module is used to adjust and correct the weights of multi-sensor data of the wet refractive index field based on Bayesian dynamic fusion; the output module is used to output the optimized wet refractive index field as the basic data for water vapor tomography inversion and meteorological warning.

[0008] Preferably, the data acquisition module collects GNSS observation data, geostationary satellite layered precipitable water data, sounding profile data and reanalysis data. The GNSS observation data collected by the data acquisition module include slant path wet refractive index observations and total delay path information, the geostationary satellite data include layered atmospheric precipitable water observations, the sounding profile data include hygrocline and vertical wet refractive index distribution, and the reanalysis data provides background wet refractive index distribution and three-dimensional water vapor flow rate information of the atmosphere.

[0009] Preferably, the wet refractive index field construction module fuses the GNSS data and the geostationary satellite data to construct an initial wet refractive index field in the following manner: Use GNSS observation data to perform three-dimensional grid interpolation, and the interpolation value is determined by weighting the distance from the GNSS observation point to the target grid point; The geostationary satellite layered atmospheric precipitable water data are converted into wet refractive index values. The geostationary satellite layered data are calculated using the gas constant, water density, gravitational acceleration and vertical temperature distribution; The errors of GNSS data and geostationary satellite data are evaluated based on the sounding profile data, and the initial wet refractivity field is generated after fusion after weights are assigned inversely proportional to the errors.

[0010] Preferably, the dynamic optimization module optimizes the initial wet refractive index field based on the partial differential dynamic equation, which describes the relationship between the wet refractive index and time and space, including the time variation term of the wet refractive index, the three-dimensional spatial flow term and the diffusion smoothing term, wherein the diffusion coefficient is dynamically calculated based on the error of GNSS and geostationary satellite data, and the water vapor source term is dynamically adjusted based on the hygroscopic layer characteristics of the sounding profile.

[0011] Preferably, the dynamic optimization module uses the finite difference method to numerically solve the partial differential dynamics equation, and the optimized wet refractive index field is iteratively updated by each time step and corrected in combination with the water vapor flow rate and the regional hygrocline characteristics.

[0012] Preferably, the spatial optimization module performs spatial interpolation on the wet refractive index field by a variational optimization method, with the goal of minimizing the gradient fluctuation of the wet refractive index field while ensuring the consistency of the interpolation and the observed value and satisfying the physical law of vertical decrease of the wet refractive index.

[0013] Preferably, the spatial optimization module uses an energy minimization model to optimize the wet refractive index field, the optimization goal is to smooth the wet refractive index gradient and satisfy the physical decreasing law of the wet refractive index in the vertical direction, and the constraint items of the energy model dynamically adjust the weights according to the reliability of the observed data.

[0014] Preferably, the fusion correction module uses a Bayesian update model to dynamically fuse the wet refractive index field, the fusion prior distribution is provided by the reanalysis data, the conditional probability distribution is constructed according to the error model of the multi-sensor observation data, and the posterior distribution is calculated by the Bayesian formula.

[0015] Preferably, the fusion correction module dynamically adjusts the weights of GNSS, geostationary satellite and sounding data, and the weights are dynamically allocated according to the real-time errors and coverage areas of each data source to improve the spatial consistency and spatiotemporal accuracy of the wet refractive index field.

[0016] Preferably, the output module uses the optimized wet refractive index field for water vapor tomography inversion and meteorological warning, and supports dynamic monitoring of atmospheric water vapor distribution and real-time analysis of abnormal changes.

[0017] The present invention provides a high-precision meteorological early warning system for water vapor tomography inversion. It has the following beneficial effects: 1. The present invention achieves the technical effect of capturing the time and space variation law of the wet refractive index field by using the partial differential dynamic equation model to dynamically optimize the wet refractive index field. Compared with the problem that the static optimization model used in the prior art cannot reflect the spatiotemporal evolution characteristics of water vapor under complex meteorological conditions, the present invention solves the problems of poor adaptability to dynamic changes of the wet refractive index field and low prediction accuracy.

[0018] 2. The present invention adopts the variational optimization method to realize the spatial optimization and physical constraint correction of the wet refractive index field, achieving the technical effect of smoothing the wet refractive index gradient and strictly following the vertical decrease law of the wet refractive index. Compared with the technical solution in the prior art that simply uses the interpolation method to cause large local deviations and insufficient consistency of physical laws, the problem of abnormal local data and spatial distribution of the wet refractive index field that does not conform to the actual atmospheric laws is solved.

[0019] 3. The present invention uses the Bayesian dynamic fusion model to perform weighted correction on multi-sensor data and calculate the posterior probability of the wet refractive index field, achieving the technical effect of integrating the advantages of multi-source data and improving the overall accuracy of the wet refractive index field. Compared with the simple weighting method in the prior art that does not consider the differences in multi-sensor error characteristics, it solves the technical problems of error accumulation in local areas and insufficient overall consistency of the wet refractive index field. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a system framework diagram of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] See also Figure 1 The embodiment of the present invention provides a high-precision meteorological early warning system for water vapor tomography inversion, comprising: Data acquisition module, used to collect multi-sensor meteorological observation data; Data preprocessing module, used to uniformly convert multi-sensor meteorological observation data into a three-dimensional grid rectangular coordinate system and complete time synchronization and denoising processing; A wet refractive index field construction module is used to construct an initial wet refractive index field based on GNSS data and geostationary satellite data; Dynamic optimization module, used for spatiotemporal dynamic optimization of the wet refractive index field based on partial differential dynamics equations; A spatial optimization module for spatial interpolation and physical constraint correction of the wet refractive index field using a variational optimization method; The fusion correction module is used to adjust and correct the weights of multi-sensor data of the wet refractive index field based on Bayesian dynamic fusion; the output module is used to output the optimized wet refractive index field as the basic data for water vapor tomography inversion and meteorological warning.

[0023] Step 1: Multi-sensor data acquisition and preprocessing Principle and function: Multi-sensor data provide comprehensive spatiotemporal distribution information for the construction of the wet refractivity field. GNSS data provide point-observed wet refractivity values ​​and total delay path information, geostationary satellites provide layered atmospheric water vapor data, sounding data provide constraints on the vertical distribution of the wet refractivity field, and reanalysis data provide three-dimensional background fields and water vapor velocity information.

[0024] Specific implementation: GNSS data collection and processing: Collect wet refractive index point values ​​and slant path delay (SWD) information of GNSS observation points, and ensure data reliability through denoising and outlier removal.

[0025] Geostationary satellite data processing: Collect stratified atmospheric precipitable water data from geostationary satellites (such as Fengyun-4A), including stratified water vapor information at low, middle and high levels, and convert its pressure altitude into geometric altitude.

[0026] Sounding profile data processing: Sounding data contain information such as hygrocline, air pressure and temperature, which provide physical constraints for the vertical distribution of wet refractive index.

[0027] Reanalysis data integration: Obtain the three-dimensional background wet refractivity field from ERA5 or ECMWF data to provide macroscopic characteristics of water vapor distribution and wind speed information.

[0028] Data preprocessing: Convert all data to a three-dimensional gridded rectangular coordinate system.

[0029] Through time synchronization, multi-source data are aligned to a unified time resolution.

[0030] Perform preliminary denoising and outlier removal.

[0031] Beneficial effects: Through the comprehensive collection and processing of multi-source data, multi-dimensional data support is provided for the construction of the wet refractive index field.

[0032] Uniform spatiotemporal resolution and denoising processing improve the spatiotemporal consistency and reliability of the data.

[0033] Step 2: Initial construction of wet refractive index field Principle and function: The initial distribution of the wet refractive index field is constructed by weighted fusion of GNSS point observation data and geostationary satellite layered data, combined with the sounding profile data to dynamically adjust the weight ratio. The initial field lays the foundation for subsequent optimization.

[0034] Specific implementation: GNSS wet refractivity field interpolation: The wet refractive index values ​​of the GNSS observation points are extended to the 3D grid by weighted interpolation, and the interpolation is calculated based on the distance weight. The weight of the observation point that is closer is larger, and the weight of the observation point that is farther away is smaller.

[0035] Conversion of wet refractive index field of geostationary satellite: The stratified atmospheric precipitable water data from geostationary satellites are converted into wet refractivity values ​​using the gas constant, water density and temperature gradient.

[0036] Multi-source data fusion: The weight ratio of GNSS and geostationary satellite data is dynamically adjusted according to the sounding data. The weight is inversely proportional to the error of the two types of data. Data with smaller errors have a greater weight to improve the reliability of the fusion results.

[0037] Initial wet refractive index field generation: The GNSS interpolation field and the geostationary satellite wet refractivity field are weighted averaged to generate the initial value of the wet refractivity field.

[0038] Beneficial effects: Through weighted fusion of multi-source data, the initial field has higher spatial consistency.

[0039] Dynamically adjusting weights reduces data inconsistency issues in the initial field.

[0040] Step 3: Dynamic optimization based on partial differential dynamics equations Principle and function: The dynamic variation of the wet refractive index field is described by partial differential dynamic equations, which include time variation, spatial diffusion and smoothness constraints. The wet refractive index field is iteratively updated by discretizing the equations to optimize the temporal consistency of the field.

[0041] Specific implementation: The dynamic variation of the wet refractive index field consists of a time variation term, a spatial gradient term and a diffusion term. The diffusion term ensures the spatial consistency of the wet refractive index field by smoothing the wet refractive index gradient.

[0042] The equation is discretized by the finite difference method and further iteratively updated time by combining with the initial wet refractive index field.

[0043] Dynamically adjust the distribution of water vapor source terms and correct the hygrocline or abnormal areas in combination with sounding data.

[0044] Beneficial effects: The optimized wet refractive index field can more realistically reflect the dynamic changes of water vapor distribution.

[0045] The introduction of the diffusion term significantly improves the smoothness and continuity of the wet refractive index field.

[0046] Step 4: Spatial optimization based on variational methods Principle and function: The spatial optimization of the wet refractive index field is solved by an energy minimization model, and the energy function includes gradient smoothness constraints, observation consistency constraints, and physical decreasing constraints.

[0047] Specific implementation: Gradient smoothness constraint: Optimize the spatial continuity of the wet refractive index gradient to avoid mutations caused by data sparsity.

[0048] Observation consistency constraints: The interpolated wet refractive index values ​​should be as close as possible to the observed data.

[0049] Physical decreasing constraints: The law of wet refractivity decreasing with height is constrained by fitting parameters of sounding data.

[0050] Beneficial effects: Improved vertical resolution and physical consistency of the wet refractive index field.

[0051] Spatial optimization reduces the error accumulation in local areas.

[0052] Step 5: Multi-sensor Bayesian dynamic fusion Principle and function: The wet refractive index field is corrected by fusing multi-sensor data through the Bayesian dynamic update formula, and the weights of different data sources are dynamically adjusted by combining prior background and real-time observation data.

[0053] Specific implementation: The prior distribution of the wet refractive index field is constructed based on the ERA5 background field as a preliminary estimate of the wet refractive index.

[0054] The conditional probability distribution is constructed based on the error distribution of the observed data, incorporating the uncertainty of different data sources into the calculation.

[0055] The prior distribution and observed data are combined using a Bayesian formula to generate the posterior distribution and correct the wet refractive index field.

[0056] Step S1: Multi-sensor data acquisition and preprocessing When implementing the water vapor tomography inversion high-precision meteorological early warning system of the present invention, it is first necessary to comprehensively collect and preprocess the multi-sensor data to ensure the reliability of the basic data for the subsequent wet refractive index field construction and optimization. In the present invention, the multi-sensor data collection includes GNSS observation data, geostationary satellite layered precipitable water data (PWV), sounding profile data and reanalysis data. These data have different spatial resolutions, temporal resolutions and error characteristics. Therefore, necessary coordinate unification, time synchronization, data alignment and noise removal processing are required to ensure the consistency and applicability of multi-source data. Data preprocessing is the basis of the entire system, and its results directly determine the accuracy and physical consistency of the wet refractive index field construction.

[0057] In this embodiment, the following technical contents are specifically included: GNSS data collection and preprocessing In the present invention, GNSS observation data is an important input for wet refractive index field construction, providing wet refractive index values ​​and slant path delay (SWD) information of point observations. Specifically, the main contents of GNSS data collection include: wet refractive index point value N of each GNSS observation station w,GNSS , the total delay path value (SWD) of the GNSS signal propagation path, as an important observation input for wet refractive index field inversion, generally, the collected GNSS data may contain noise or outliers, which may be caused by satellite signal multipath effects, receiver errors or ionospheric delay effects. In a possible implementation, this embodiment uses a statistical method to remove outliers and filter the GNSS data to remove random noise in the observation data.

[0058] In order to adapt the GNSS observation point values ​​to the subsequent three-dimensional wet refractive index field construction, the present invention adopts a coordinate conversion algorithm to convert the geographic coordinate system (longitude, latitude, elevation) into a gridded three-dimensional rectangular coordinate system (x, y, z). Specifically, the geocentric rectangular coordinate transformation formula is used: x=(R+h)·cosφ·cosλ y=(R+h)cosφ·sinλ z=(R(1-e 2 )+h)·sinφ Wherein, R is the equatorial radius of the earth, e is the first eccentricity of the earth, h is the elevation of the observation point, φ is the latitude, and λ is the longitude. In some embodiments, in order to further improve the accuracy of GNSS data, the GNSS point wet refractive index value can also be preliminarily corrected based on the sounding data to ensure that the GNSS data has a high consistency with other data sources.

[0059] The acquisition and processing of geostationary satellite data Geostationary satellite data (such as Fengyun-4A) provide stratified atmospheric precipitable water volume (PWV) data, covering water vapor distribution information in the low, middle and high layers. Geostationary satellite data have a high temporal resolution, generally up to once every 15 minutes, and a spatial resolution of several kilometers, which is suitable for reflecting the regional distribution characteristics of atmospheric water vapor.

[0060] Specifically, the processing of geostationary satellite data includes: According to the pressure layer distribution of the geostationary satellite, the PWV data of the low, middle and high layers are extracted.

[0061] The pressure coordinate system of satellite data is converted into geometric height coordinates so that it can be merged with GNSS data in the same coordinate system.

[0062] In a possible implementation, the present invention uses the following formula to calculate the geometric height: Wherein, H is the geometric height, Φ is the potential height, g is the gravitational acceleration, and the potential height is calculated by numerical integration based on the pressure layer data observed by the geostationary satellite. In some embodiments, the geostationary satellite data may have local observation blind spots or errors. In this case, it can be corrected in combination with the reanalysis data. The correction method includes filling the missing data based on the interpolation algorithm or establishing a correction model using the difference between the observation and reanalysis data.

[0063] Acquisition and processing of sounding profile data The sounding data provide the fine structure of the wet refractive index distribution in the vertical direction, especially in the wet cline region, where the sounding data can provide more accurate wet refractive index distribution characteristics.

[0064] In general, sounding profile data include the following information: Wet refractive index gradient in the vertical direction.

[0065] Height and intensity distribution of hygrocline.

[0066] Water vapor characteristics at the top of the layer and at the tropospheric boundary.

[0067] In a possible implementation, the present invention uses sounding data to fit the vertical decrease law of wet refractive index. The specific decrease law can be expressed in the following form: N w (h) = N 0 ·e -α·h Among them, N w (h) is the wet refractive index at height h, N 0 is the surface wet refractivity value, α is the decrease coefficient, which is obtained by fitting the sounding profile data.

[0068] Reanalysis data collection and integration Reanalysis data (such as ERA5) provide three-dimensional background field and water vapor flow field information for the construction of wet refractivity field. The spatial resolution of reanalysis data is generally tens of kilometers, and the temporal resolution is hourly.

[0069] Specifically, this embodiment uses reanalysis data to provide: Background distribution of wet refractive index over a large area, used as a benchmark for the initial wet refractive index field.

[0070] The velocity components of atmospheric water vapor, including horizontal and vertical wind speeds, are used for dynamic optimization of the wet refractive index field.

[0071] In some embodiments, the wet refractive index values ​​in the reanalysis data may also be used to correct the GNSS or geostationary satellite data to reduce bias that may occur during the data fusion process.

[0072] Data time synchronization and preprocessing Since the time resolution and time base of GNSS, geostationary satellite, sounding and reanalysis data may be different, in order to ensure the consistency of multi-source data, in this embodiment, all data need to be time aligned. The specific implementation method of time synchronization includes: interpolating the geostationary satellite, sounding and reanalysis data to the same time point according to the timestamp of the GNSS data.

[0073] Time completion is performed on data with low time resolution through linear interpolation or high-order interpolation algorithms.

[0074] Step S2: Initial construction of wet refractive index field After completing the multi-sensor data acquisition and preprocessing in step S1, the next step is to use multi-source data to construct the initial distribution of the wet refractive index field. This step generates the initial distribution of the wet refractive index field by fusing GNSS observation data, geostationary satellite layered precipitable water data, and sounding profile data, laying the foundation for subsequent dynamic optimization and spatial optimization. In general, the construction of the initial wet refractive index field needs to combine the characteristics of different data sources, adopt a weighted fusion strategy, and correct the inconsistency of the data and the difference in spatial coverage to ensure the rationality and accuracy of the initial field.

[0075] In the present invention, the core of the initial construction of the wet refractive index field is to interpolate and expand the GNSS point observation data, convert and process the geostationary satellite layered data, and dynamically adjust the errors of multi-source data in combination with sounding data. This step realizes the fusion of multi-source data through mathematical modeling and generates a wet refractive index field with preliminary physical consistency.

[0076] In this embodiment, the wet refractive index values ​​of the GNSS observation points are first spatially interpolated. Generally, the GNSS observation points provide discrete wet refractive index values, which need to be interpolated into a three-dimensional grid to generate a continuous wet refractive index field. As an option, the inverse distance weighted interpolation method is used in this embodiment to interpolate the GNSS data. The specific interpolation formula is as follows: Among them, N w,i is the wet refractive index value of the i-th GNSS observation point, d i is the distance from the observation point to the interpolation target point, and u is a weighted index parameter used to control the smoothness of the interpolation. In a possible implementation, the weighted index u can be dynamically adjusted according to the density of the observation point distribution. For example, when the observation points are sparsely distributed, the value of u can be appropriately reduced to enhance the smoothness of the interpolation.

[0077] Generally, there will be certain errors in GNSS data after interpolation. In order to reduce the interpolation error, the present embodiment further uses the sounding data to make preliminary corrections to the interpolation results. For example, the sounding data can provide vertical wet refractive index gradient information, which is used to correct the vertical variation characteristics of the GNSS interpolation field, thereby improving the physical consistency of the interpolation field.

[0078] In this embodiment, the geostationary satellite layered precipitable water volume (PWV) data is also converted. Generally, the geostationary satellite data provides layered PWV values ​​in the form of pressure coordinates, which need to be converted into wet refractive index values ​​and mapped to the geometric height coordinate system. As an option, the following formula is used in this embodiment to convert PWV data into wet refractive index: Among them, N w (h) is the wet refractive index at the geometric height h, ρ w is the density of water, g is the acceleration of gravity, PWV is the stratified precipitable water volume, R is the gas constant, and T(h) is the temperature at height h. Specifically, the temperature T(h) is extracted from the temperature profile data or reanalysis data provided by the geostationary satellite.

[0079] In this embodiment, in order to fuse GNSS data with geostationary satellite data, a weighted average fusion method is adopted. Generally, GNSS data has high temporal resolution and point observation accuracy, but limited spatial coverage, while geostationary satellite data has a wider coverage but relatively low accuracy. In order to fully utilize the advantages of the two types of data, in this embodiment, the weights are dynamically adjusted according to the errors of the two types of data. The specific weighting formula is as follows: N w,init (x,y,z)=w GNSS ·N w,GNSS (x,y,z)+w FY ·N w,FY (x,y,z) Among them, w GNSS and w FY are the weights of GNSS data and geostationary satellite data respectively, and the calculation formula is: σ GNSS and σ FY They are the mean square errors of GNSS data and geostationary satellite data, respectively. The error values ​​are determined by evaluating the two types of data using sounding data.

[0080] In some embodiments, the fused wet refractive index field is also preliminarily checked to ensure the consistency and rationality of the data. Specifically, the generated initial wet refractive index field can be compared and analyzed with the background wet refractive index field of the reanalysis data, and a local adjustment algorithm can be used to correct the area with large deviations. For example, in areas with large wet jump layer intensity, the vertical distribution of the wet refractive index can be further dynamically adjusted in combination with the sounding profile data to improve the accuracy of the regional field.

[0081] Step S3: Dynamic optimization based on partial differential dynamics equations After completing the initial construction of the wet refractive index field in step S2, the wet refractive index field already has a certain spatial distribution and physical consistency. However, due to the differences in spatiotemporal coverage and resolution of multi-source data, the dynamic change characteristics of the initial wet refractive index field have not been fully reflected, especially under complex meteorological conditions, its dynamic evolution characteristics may be biased. Therefore, this step dynamically optimizes the initial wet refractive index field based on the partial differential dynamics equation to obtain a more realistic spatiotemporal distribution.

[0082] Generally speaking, the change of the wet refractive index field is jointly affected by factors such as time change, spatial flow and diffusion smoothness. The present invention establishes a partial differential dynamic equation to describe the change law of the wet refractive index field in time and space, combines the numerical discrete method, iteratively solves the dynamic distribution of the wet refractive index field, and provides high-quality dynamic input for subsequent spatial optimization.

[0083] In this embodiment, the dynamic evolution of the wet refractive index field is described by the following partial differential dynamic equation: Among them, N w Represents the wet refractive index, in units of 10 -6 The parameters in the formula are defined as follows: The rate of change of wet refractivity over time is used to describe the dynamic evolution of atmospheric water vapor distribution in the time dimension. u, v, w are the velocity components of the water vapor flow field in the x, y, and z directions, respectively, in meters per second (m / s). These velocity components come from reanalysis data. D is the diffusion coefficient, which represents the smoothness of the wet refractivity field, in square meters per second (m 2 / s), the diffusion coefficient is dynamically adjusted according to the humidity gradient and data resolution in different regions. The second-order partial derivative of the wet refractive index in all directions of space is used to describe the spatial diffusion effect of the wet refractive index. S(x,y,z,t): Water vapor source term, which represents the increase or decrease of the wet refractive index in a specific area, in units of 10 -6 / s, and the water vapor source term is provided by the regional water vapor distribution in the geostationary satellite data and the hygrocline information in the sounding data.

[0084] In this embodiment, the diffusion coefficient D is dynamically adjusted according to the wet refractive index gradient and the multi-source data error. The value is determined by the following formula: Among them, α is an empirical constant, σ GNSS and σ FY are the errors of GNSS data and geostationary satellite data, respectively. In general, the diffusion effect of the wet refractive index field can effectively smooth noise and local outliers. However, in order to avoid loss of details due to excessive smoothing, in some embodiments, this embodiment dynamically constrains the diffusion coefficient of important areas such as the hygrocline. For example, when the intensity of the regional hygrocline is large, the diffusion coefficient is appropriately reduced to retain details.

[0085] In this embodiment, the water vapor source term S(x, y, z, t) is mainly provided by the stratified water vapor data and sounding data of the geostationary satellite. Specifically, the geostationary satellite data is used to describe the changes in water vapor distribution at different altitudes, while the sounding data provides dynamic change information of the hygrocline. For example, under severe convective weather conditions, the rapid changes in the hygrocline will significantly affect the distribution of the water vapor source term. As an option, in the calculation of the water vapor source term, this embodiment performs weighted processing on the hygrocline strength in the sounding data to improve the accuracy of the source term distribution. The weight of the hygrocline strength is determined by the following formula: Among them, ΔN w represents the variation of wet refractive index in the hygrocline, H layer Indicates the thickness of the hygrocline.

[0086] In the dynamic optimization process, the present invention updates the wet refractive index field in a time-by-time iterative manner. Generally, the wet refractive index field after each iteration will be regionally verified to ensure the rationality of its spatiotemporal distribution. For example, for areas where the water vapor flow field changes greatly, the dynamic consistency of the wet refractive index field can be further improved by increasing the calculation accuracy of the time step.

[0087] Through the above steps, this embodiment realizes the dynamic optimization of the wet refractive index field, generates a wet refractive index distribution that is more in line with actual meteorological conditions, and provides dynamic consistency input for subsequent spatial optimization. The dynamically optimized wet refractive index field not only accurately describes the time evolution characteristics of water vapor distribution, but also significantly improves the applicability of the wet refractive index field under complex meteorological conditions.

[0088] Step S4: Spatial optimization based on variational methods After completing the dynamic optimization of step S3, the wet refractive index field has certain dynamic evolution characteristics and consistency of overall spatiotemporal distribution. However, due to the discontinuity of the data and the irregularity of the wet refractive index field in local areas under complex meteorological conditions, there may still be deviations in the distribution of details. Therefore, this step is based on the variational optimization method to perform spatial optimization on the wet refractive index field to achieve smoothing of the wet refractive index gradient, consistency correction of the observed data, and physical constraints of the vertical decrease law.

[0089] In general, the spatial optimization of the wet refractive index field is achieved by constructing an energy minimization model. This model comprehensively considers the smoothness of the wet refractive index gradient, its consistency with the observed value, and the physical law of vertical decrease, and uses a numerical optimization method to solve it, which can significantly improve the spatial accuracy and physical consistency of the wet refractive index field.

[0090] In this embodiment, the spatial optimization of the wet refractive index field is modeled by the following energy minimization function: Among them, E(N w ) represents the total energy of the wet refractive index field, is a gradient smoothing term, which represents the sum of the squares of the wet refractive index gradient and is used to constrain the spatial continuity of the wet refractive index field and reduce local excessive fluctuations. (N w -N obs ) 2 is the observation consistency term, which represents the difference between the wet refractive index field and the observed data, and λ is the weight factor, which is used to control the proportion of observation consistency in the energy model. is a vertical decrease constraint term, which is used to constrain the decrease law of wet refractivity with height. α is the decrease coefficient, which is determined by fitting the sounding data. γ is the weight factor, which controls the influence of this constraint in the energy model.

[0091] In a possible implementation, α is obtained by exponential fitting of the vertical distribution of wet refractivity in historical sounding data. For example, in the troposphere, the wet refractivity usually decreases exponentially with height, and its decreasing law can be described by the following relationship: N w (h) = N 0 ·e -α·h Among them, N 0 is the surface wet refractivity, h is the height, and α is the decreasing parameter.

[0092] Specifically, in this embodiment, the solution of the energy function is iteratively optimized using the gradient descent method or the conjugate gradient method. The iterative formula of the gradient descent method is as follows: in, and are the wet refractive index fields of the nth and n+1th iterations, respectively. η is the learning rate, which is used to control the step size of each iteration and is usually adjusted dynamically according to the changing trend of the energy function.

[0093] Under hygrocline or severe convective weather conditions, the vertical decline law may change abnormally. In this embodiment, the vertical decline constraint in a specific area is adjusted by adding regional constraints. For example, in an area with a strong hygrocline, the weight of the decline parameter α can be dynamically increased to highlight the influence of the hygrocline.

[0094] In some embodiments, this embodiment also performs regional optimization on the weight of the gradient smoothing term. Generally, the gradient smoothing term is used to constrain the spatial continuity of the wet refractive index field, but in areas with large changes in water vapor gradients such as the hygrocline or front, too strong smoothing constraints may lead to loss of detail information. In order to solve this problem, this embodiment introduces regional sensitive smoothing weights, that is, appropriately reduce the weight of the gradient smoothing term in the hygrocline and front regions, and the specific implementation method is as follows: Among them, w 平滑 is the regional smoothing weight, which is dynamically adjusted according to the hygrocline strength or water vapor gradient. For example, when the hygrocline strength is large, w is appropriately reduced. 平滑的值 .

[0095] In order to improve the convergence speed of the energy function solution, a layer-by-layer optimization strategy is adopted in this embodiment, that is, first optimizing the low-layer distribution of the wet refractive index field, and then gradually expanding to the high-layer. This layer-by-layer optimization method can effectively reduce the numerical coupling problem in the vertical direction and improve the optimization calculation efficiency. As a possible implementation method, in this embodiment, the troposphere is divided into multiple vertical layers, and the wet refractive index field of each layer is calculated by the following formula: Where Δh is the thickness of the current layer, is the wet refractive index gradient in the vertical direction.

[0096] Through the above steps, this embodiment completes the spatial optimization of the wet refractive index field. The optimized wet refractive index field not only has higher spatial resolution and physical consistency, but also effectively reduces the error accumulation in the local area, providing high-quality input data for subsequent multi-sensor fusion correction.

[0097] Step S5: Multi-sensor Bayesian dynamic fusion correction After completing the spatial optimization of the wet refractive index field in step S4, the wet refractive index field has high spatiotemporal consistency and physical rationality. However, due to differences in observation characteristics, resolution, and error models of different data sources, the optimized wet refractive index field may still have local uncertainties and systematic deviations. Therefore, this step further corrects the wet refractive index field based on the Bayesian dynamic fusion model, and dynamically updates the wet refractive index field by integrating the prior information, observation condition probability, and error distribution of multi-sensor data to improve its overall accuracy and reliability.

[0098] Generally, Bayesian dynamic fusion correction is achieved by constructing a posterior probability distribution, that is, combining the prior distribution of the wet refractive index field and the conditional probability distribution of multi-sensor observation data to calculate the wet refractive index posterior value of each grid point. In the present invention, the fused sensors include GNSS, geostationary satellite and sounding data. The weight of each data source is dynamically adjusted according to its reliability and coverage to achieve multi-source information fusion of the wet refractive index field.

[0099] In this embodiment, the core model of multi-sensor Bayesian dynamic fusion correction is: P(N w |D)∝P(D|N w )·P(N w ) in: P(N w |D) is the posterior probability, which means that after combining the observed data D, the wet refractive index N w The optimal distribution of P(D|N w ) is the conditional probability, indicating that the wet refractive index is N w Under the assumption that w ) is the prior probability, representing the initial distribution of the wet refractive index field.

[0100] Specifically, in this embodiment, the prior distribution of the wet refractive index field P(N w ) is provided by reanalysis data. Reanalysis data has the characteristics of global coverage and high stability. Its wet refractivity distribution can be used as a preliminary reference benchmark. In one possible implementation, the wet refractivity value in the reanalysis data is mapped to the grid resolution of the current wet refractivity field by linear interpolation to ensure that the prior distribution has the same spatial resolution as the observed data. The conditional probability P(D|N w ) is the key to Bayesian dynamic fusion. Its construction needs to consider the error characteristics and distribution laws of multi-sensor data. As an option, in this embodiment, the conditional probability P(D|N w ) is modeled as a Gaussian distribution, which is specifically expressed as: Where D is the multi-sensor observation value, including the wet refractivity observation value of GNSS, geostationary satellite and sounding data, N w is the predicted value of the current wet refractive index field, σ 2 is the variance of the observation error, reflecting the uncertainty of different sensor data.

[0101] In this embodiment, the observation error variance σ 2 The calculation of needs to be combined with specific sensor characteristics. In general: the error of GNSS data mainly comes from multipath effect and ionospheric delay, and the error variance can be estimated by the standard deviation of historical observation data.

[0102] The errors in geostationary satellite data are related to the accuracy of stratified water vapor observations, and their error variance is usually large, especially in the boundary layer region.

[0103] The error of sounding data mainly comes from instrument accuracy, and its error variance is small, so it can usually be regarded as a high-precision reference.

[0104] As an option, in this embodiment, the errors of GNSS and geostationary satellite data are dynamically corrected by using sounding data. Specifically, when the sounding data covers a small area, the error ratio of its wet refractive index distribution can be used to weightedly adjust the error variance of other data sources to ensure the fusion accuracy of multi-sensor data. In the actual implementation of Bayesian dynamic fusion, this embodiment uses weighted posterior update to calculate the correction value of the wet refractive index field. The formula is as follows: in, is the corrected wet refractive index value, is the observed wet refractive index value of the ith sensor, w i is the weight of the i-th sensor, which is determined inversely by its observation error variance.

[0105] In some embodiments, in order to further improve the computational efficiency, this embodiment introduces a regional weighted Bayesian update strategy, that is, according to the uncertainty distribution of the wet refractive index field, the high uncertainty region is corrected first, and the specific implementation method is as follows: First, high uncertainty regions, such as near the hygrocline or front, are identified by calculating the rate of change of the gradient of the wet refractive index field.

[0106] Then, a higher weight is given to the observation data in the high uncertainty region to enhance the influence of the sensor data in this region.

[0107] In addition, under conditions of highly time-variable weather events (such as typhoons or severe convective weather), this embodiment dynamically adjusts the time variation trend of the prior distribution by introducing a time-correlated Bayesian update method. For example, the wet refractive index field at the previous moment is used as the prior distribution at the current moment, and the time continuity correction of the wet refractive index field is achieved through iterative updating of the time step.

[0108] Through the above steps, this embodiment completes the multi-sensor Bayesian dynamic fusion correction of the wet refractive index field and generates the final high-precision wet refractive index field. The corrected wet refractive index field has the comprehensive advantages of multi-source data in spatial distribution, and achieves dynamic consistency in time changes, providing reliable data support for the meteorological early warning system.

[0109] Step S6: Optimizing the output of wet refractive index field and supporting meteorological warning After completing the correction of the wet refractive index field based on multi-sensor Bayesian dynamic fusion in step S5, the system generates a wet refractive index field distribution with high precision, strong physical consistency and reasonable dynamic changes. However, in order to ensure that the wet refractive index field can be efficiently used in actual meteorological warnings, it needs to be standardized, dynamically analyzed and adapted to the warning system. This step aims to convert the optimized wet refractive index field into data output suitable for meteorological warning needs, while providing key dynamic water vapor information and real-time anomaly detection related to the warning system.

[0110] In general, the output of the wet refractive index field needs to consider spatial resolution, temporal resolution and format specifications to meet the direct call requirements of the meteorological system. The present invention also combines the results of optimizing the wet refractive index field to further develop dynamic water vapor distribution characteristics extraction and hygrocline anomaly analysis functions to improve the effectiveness of the meteorological early warning system.

[0111] In this embodiment, the optimization of the output and adaptation of the wet refractive index field is divided into the following main parts: Optimizing the Normalization and Formatting of the Wet Refractive Index Field In this embodiment, the optimized wet refractive index field is first standardized to a format and resolution suitable for a meteorological warning system. Generally, the wet refractive index field is output as three-dimensional spatiotemporal data, recording the wet refractive index value of each grid point and its temporal variation.

[0112] Specifically, in this embodiment, the wet refractive index field is subjected to the following normalization processing: The wet refractive index value unit is unified to 10 -6 , to meet meteorological standards.

[0113] The wet refractive index field is spatially interpolated to adjust its resolution to the requirements of the early warning system. For example, if the meteorological early warning system uses a horizontal resolution of 5 kilometers and a vertical resolution of 500 meters, the linear interpolation method is used in this embodiment to resample the resolution of the wet refractive index field.

[0114] The data in the temporal dimension are resampled to adjust the time step of the raw wet refractivity field to the real-time needs of the early warning system. As an option, the temporal resolution can be adjusted from every hour to every 15 minutes.

[0115] In this embodiment, the standardized wet refractive index field is stored in NetCDF (Network Common Data Format). The NetCDF format has efficient multi-dimensional data storage and fast reading capabilities, and is widely used in the storage and transmission of meteorological warning and environmental monitoring data.

[0116] Analysis of Dynamic Water Vapor Distribution in Wet Refractive Index Field In order to meet the analysis requirements of the water vapor distribution characteristics for meteorological warning, the optimized wet refractive index field is further processed in this embodiment to extract the dynamic distribution information of water vapor.

[0117] Generally, the dynamic characteristics of the wet refractive index field can be extracted by gradient analysis and integral calculation. In this embodiment, the dynamic analysis is performed by the following method: Calculate the horizontal and vertical gradients of the wet refractive index field to identify abnormal areas of water vapor distribution. For example, the horizontal gradient is used to locate the water vapor front, and the vertical gradient is used to identify the height and strength of the hygrocline. The calculation of the horizontal and vertical gradients is based on the discrete difference values ​​of the wet refractive index field at the grid points.

[0118] The precipitable water volume (PWV) is calculated by integration. The precipitable water volume is obtained by integrating the wet refractive index field in the vertical direction and is used to describe the total water vapor in the atmospheric column. In general, the precipitable water volume is calculated as: The wet refractivity field is divided according to the levels of the meteorological system, and the water vapor content of each layer is calculated.

[0119] The water vapor content of each layer is accumulated to obtain the total precipitable water, and key indicators are extracted by region.

[0120] In some embodiments, in order to improve the accuracy of dynamic water vapor distribution analysis, a local weighted method is introduced in this embodiment to weight the wet refractive index field in high gradient areas to highlight the impact of important weather systems on water vapor transport.

[0121] Real-time monitoring of abnormal changes in wet refractive index field In this embodiment, another key function of optimizing the wet refractive index field is real-time monitoring of abnormal changes. This function is mainly used to identify key change features in the wet refractive index field that may lead to severe convective weather, local rainstorms or other extreme meteorological events.

[0122] In general, anomaly monitoring includes temporal rate of change analysis and spatial outlier detection: Time rate analysis: By calculating the time derivative of the wet refractive index field, the hygrocline or the area of ​​rapid change of water vapor transport can be identified. The analysis method of time rate is usually based on the difference calculation of grid points, combined with dynamic threshold to filter out significant change points.

[0123] Spatial outlier detection: By statistically analyzing the distribution of the wet refractive index field in a certain area, the grid points with abnormal deviations are marked. For example, if the wet refractive index value of a grid point exceeds two standard deviations of the mean of its neighborhood, it is marked as an outlier.

[0124] As a possible implementation method, in this embodiment, the abnormal change points are further classified and marked, for example, the hygroscopic layer changes, frontal water vapor transport and local evaporation changes are classified and recorded separately, so that the meteorological warning system can generate corresponding warning information for different types of abnormal changes.

[0125] Interface development and system integration of wet refractive index field In order to facilitate the meteorological warning system to directly call the optimized wet refractive index field, a set of standardized interfaces is developed in this embodiment. Generally, the interface needs to support multi-dimensional data query and real-time push functions, and has the ability to directly call dynamic analysis results.

[0126] Specifically, the interface developed in this embodiment has the following functions: Data query function: supports querying historical and real-time data of wet refractive index field by time, spatial range and specific area.

[0127] Data push function: supports real-time data subscription of the meteorological warning system, and pushes the latest wet refractive index field data to the warning system at fixed time intervals (such as 15 minutes).

[0128] Dynamic analysis function: supports directly calling the dynamic analysis results of the wet refractive index field through the interface, such as obtaining the change trend of the hygroscopic layer height or the distribution of precipitable water in a certain area.

[0129] In one possible implementation, the interface is designed using the REST architecture, supports cross-platform calls, and can be seamlessly integrated with existing meteorological systems.

[0130] Application of Optimized Wet Refractive Index Field in Meteorological Warning In this embodiment, the optimized wet refractive index field and its dynamic analysis results can be applied to multiple meteorological warning scenarios. For example, in severe convective weather warnings, the areas of enhanced convective intensity can be identified by monitoring the temporal changes in the height of the hygrocline.

[0131] In local heavy rain warnings, the core location of water vapor accumulation and its possible precipitation intensity are predicted through dynamic analysis of water vapor transport.

[0132] In typhoon tracks and precipitation warnings, the impact range of typhoon periphery water vapor transport is evaluated through the dynamic changes of the wet refractive index field.

[0133] Through the above steps, this embodiment realizes the standardized output, dynamic characteristic analysis and real-time warning support functions of optimizing the wet refractive index field. The data output module ensures that the wet refractive index field can be directly applied to the meteorological warning system. At the same time, the accuracy and response efficiency of meteorological warnings are further improved through abnormal monitoring and dynamic analysis functions.

[0134] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-precision meteorological early warning system based on water vapor tomography inversion, characterized in that: The system comprises: Data acquisition module, used to collect multi-sensor meteorological observation data; Data preprocessing module, used to uniformly convert multi-sensor meteorological observation data into a three-dimensional grid rectangular coordinate system and complete time synchronization and denoising processing; A wet refractive index field construction module is used to construct an initial wet refractive index field based on GNSS data and geostationary satellite data; Dynamic optimization module, used for spatiotemporal dynamic optimization of the wet refractive index field based on partial differential dynamics equations; A spatial optimization module for spatial interpolation and physical constraint correction of the wet refractive index field using a variational optimization method; A fusion correction module is used to adjust and correct the weight of multi-sensor data of the wet refractive index field based on Bayesian dynamic fusion; The output module is used to output the optimized wet refractive index field as the basic data for water vapor tomography inversion and meteorological warning.

2. A high-precision weather warning system based on water vapor tomography inversion according to claim 1, characterized in that: The data acquisition module collects GNSS observation data, geostationary satellite layered precipitable water data, sounding profile data and reanalysis data. The GNSS observation data collected by the data acquisition module include slant path wet refractivity observation values ​​and total delay path information, the geostationary satellite data include layered atmospheric precipitable water observation values, the sounding profile data include hygrocline and vertical wet refractivity distribution, and the reanalysis data provides background wet refractivity distribution and three-dimensional water vapor velocity information of the atmosphere.

3. The high-precision meteorological early warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The wet refractivity field construction module fuses GNSS data and geostationary satellite data in the following way to construct the initial wet refractivity field: Use GNSS observation data to perform three-dimensional grid interpolation, and the interpolation value is determined by weighting the distance from the GNSS observation point to the target grid point; The geostationary satellite layered atmospheric precipitable water data are converted into wet refractive index values. The geostationary satellite layered data are calculated using the gas constant, water density, gravitational acceleration and vertical temperature distribution; The errors of GNSS data and geostationary satellite data are evaluated based on the sounding profile data, and the initial wet refractivity field is generated after fusion after weights are assigned inversely proportional to the errors.

4. The high-precision meteorological early warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The dynamic optimization module optimizes the initial wet refractive index field based on the partial differential dynamic equation, which describes the relationship between the wet refractive index and time and space, including the time variation term of the wet refractive index, the three-dimensional spatial flow term and the diffusion smoothing term. The diffusion coefficient is dynamically calculated according to the error of GNSS and geostationary satellite data, and the water vapor source term is dynamically adjusted according to the hygroscopic layer characteristics of the sounding profile.

5. The high-precision meteorological early warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The dynamic optimization module uses the finite difference method to numerically solve the partial differential dynamics equations. The optimized wet refractive index field is iteratively updated by each time step and corrected in combination with the water vapor flow rate and regional hygrocline characteristics.

6. The high-precision weather warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The spatial optimization module performs spatial interpolation on the wet refractive index field through a variational optimization method. The goal is to minimize the gradient fluctuation of the wet refractive index field while ensuring the consistency of the interpolated value with the observed value and satisfying the physical law of vertical decrease of the wet refractive index.

7. The high-precision meteorological early warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The spatial optimization module uses an energy minimization model to optimize the wet refractive index field. The optimization goal is to smooth the wet refractive index gradient and satisfy the physical decreasing law of the wet refractive index in the vertical direction. The constraints of the energy model dynamically adjust the weights according to the reliability of the observed data.

8. The high-precision meteorological early warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The fusion correction module uses the Bayesian update model to dynamically fuse the wet refractive index field. The fusion prior distribution is provided by the reanalysis data, the conditional probability distribution is constructed according to the error model of multi-sensor observation data, and the posterior distribution is calculated using the Bayesian formula.

9. The high-precision weather warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The fusion correction module dynamically adjusts the weights of GNSS, geostationary satellite and sounding data. The weights are dynamically allocated according to the real-time errors and coverage areas of each data source to improve the spatial consistency and spatiotemporal accuracy of the wet refractive index field.

10. The high-precision weather warning system based on water vapor tomography inversion according to claim 1 is characterized in that: The output module uses the optimized wet refractive index field for water vapor tomography inversion and meteorological warning, and supports dynamic monitoring of atmospheric water vapor distribution and real-time analysis of abnormal changes.

Images