Collaborative inversion method for Beidou atmospheric water content based on multi-source heterogeneous data fusion

Abstract

The invention relates to the technical field of data processing, in particular to a multi-source heterogeneous data fusion Beidou atmospheric water content collaborative inversion method, which comprises the following steps: acquiring Beidou observation data, forecast mode data and ground meteorological data of a target monitoring station and an adjacent monitoring station; height integration is carried out according to the vertical gradient deviation of the temperature changing along with the height in the numerical forecasting mode, and a thermal power vertical decoupling index is determined; the water vapor advection intensity is calculated based on the spatial distribution difference of zenith total delay of a target monitoring station and an adjacent monitoring station, a self-adaptive collaborative weight is generated in combination with the water vapor advection intensity and a thermal power vertical decoupling index, dynamic weighted fusion is performed on ground measured data and forecast mode data, and the atmospheric water content is inverted in combination with atmospheric parameters. According to the method, through adaptive fusion driven by a physical mechanism, interference of a thermal inversion layer and turbulence is effectively inhibited, and the accuracy of water vapor inversion is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. Specifically, it relates to a method for collaborative inversion of BeiDou atmospheric water content based on the fusion of multi-source heterogeneous data. Background Technology

[0002] Atmospheric water content monitoring is a crucial link in short-term weather forecasting, extreme disaster early warning, and weather modification operations. Water vapor inversion using Global Navigation Satellite Systems (GNSS), especially the BeiDou Navigation Satellite System, has become an important means of meteorological monitoring due to its advantages of high temporal resolution, all-weather operation, and low cost. Its core principle is to use the delay generated when the signal passes through the troposphere and combine it with weighted average temperature to invert water vapor content. Typical applications include urban flooding early warning systems, airport meteorological support services, and agricultural drought monitoring networks.

[0003] However, existing methods for retrieving atmospheric water content mainly rely on linear regression models based on surface air temperature (such as the Bevis formula). This method implicitly assumes that surface temperature changes can linearly reflect the temperature structure of the upper atmosphere. However, in the actual complex physical atmospheric environment, inversion layers or strong vertical turbulence often occur. At this time, surface temperature and upper atmosphere temperature become decoupled, causing the linear model to fail.

[0004] The aforementioned issues can lead to deviations in the weighted average temperature of linear regression models based on surface air temperature during severe convective weather, frontal passages, or temperature inversions. These deviations reduce the accuracy of atmospheric water content retrieval results and consequently lower the accuracy of weather forecasts.

[0005] Therefore, there is an urgent need for a multi-source collaborative inversion method that can sense the stability of the vertical structure of the atmosphere in real time, quantify the degree of horizontal turbulence and advection, and adaptively adjust the multi-source data fusion strategy accordingly, in order to solve the problem of decreased inversion accuracy of existing technologies under complex atmospheric conditions. Summary of the Invention

[0006] To address the problem of inaccurate atmospheric water content retrieval results caused by linear regression models relying solely on surface air temperature under complex atmospheric conditions such as temperature inversion, turbulence, or strong advection in existing technologies, this invention proposes a collaborative inversion method for atmospheric water content retrieval using BeiDou data fusion from multiple sources, including: Acquire BeiDou satellite observation data, forecast model data, and ground measurement data from the target monitoring station and neighboring monitoring stations; extract carrier phase post-verification residual and zenith total delay from the BeiDou satellite observation data; extract zenith static delay from the ground measurement data; Based on the temperature in the forecast model data, the vertical gradient deviation of temperature with altitude is calculated, and the vertical gradient deviation is integrated with altitude to determine the thermodynamic vertical decoupling index; the water vapor advection intensity is calculated based on the spatial distribution difference of the total zenith delay between the target monitoring station and the neighboring monitoring stations. The water vapor non-uniformity factor is determined by combining the carrier phase post-verification residual, water vapor advection intensity, and thermodynamic vertical decoupling index. The historical statistical characteristics of the water vapor non-uniformity factor are obtained, and a distribution probability model based on statistical moments is constructed based on the historical statistical characteristics. The water vapor non-uniformity factor is mapped to an adaptive collaborative weight for adjusting the fusion ratio of multi-source data. Adaptive collaborative weighting is used to weight and fuse the temperature data in ground-based measured data and the temperature data in forecast model data, and atmospheric water content is inverted by combining zenith static delay and zenith total delay.

[0007] This technical solution quantifies the deviation between the actual atmospheric vertical temperature lapse rate and the ideal dry adiabatic state from a physical mechanism by calculating the thermodynamic vertical decoupling index. This allows for the precise identification of thermal decoupling phenomena between the ground and the atmosphere caused by inversion layers or strong convection. The vertical decoupling index is used as a physical gate in the analysis of residuals (microscopic turbulence) and network-level horizontal gradients (macroscopic advection) following satellite observations. This constructs a water vapor heterogeneity factor that integrates microscopic and macroscopic characteristics. When vertical stratification is unstable, both horizontal residual fluctuations and inter-regional water vapor transport gradients are more likely to be caused by severe weather processes. Instead of measurement noise, an adaptive collaborative confidence weight was generated by constructing a distribution probability model based on historical statistical characteristics. This weight can automatically adjust the confidence level of ground observation models and numerical prediction models according to the current atmospheric disturbance level. Under stable weather conditions, more reliance is placed on high-frequency ground observations; under abnormal weather conditions, more numerical models containing vertical information are introduced. Finally, the dynamically weighted temperature obtained through dynamic fusion retains the high real-time performance of ground measurement data and has the vertical structure anti-interference capability of numerical prediction models, thus improving the accuracy and robustness of atmospheric water content inversion under complex meteorological conditions.

[0008] Preferably, the thermodynamic vertical decoupling index is determined based on the following relationship:

[0009] Among them, The thermodynamic vertical decoupling index of the target monitoring station. The ground elevation of the target monitoring station. The height of the top of the troposphere. Let be the integral variable representing the vertical height. For forecast model data in Temperature value at that location, This represents the vertical gradient of temperature as a function of altitude. The preset dry adiabatic lapse rate constant, For forecast model data in The air pressure value at that location, These are air pressure values ​​measured on the ground. It is the absolute value symbol. For an integral infinitesimal element, This refers to the standard atmospheric temperature.

[0010] This technical solution accumulates the energy of temperature gradient anomalies across the entire troposphere through integral calculations. The absolute value of the temperature gradient difference directly reflects the degree of distortion of the local atmospheric stratification relative to the ideal state, while the pressure ratio term, as a density weight, reflects the physical fact that the lower atmosphere contributes more to the overall water vapor inversion. Through this refined integral calculation, discrete, local inversion points or unstable stratifications can be transformed into a macroscopic scalar index that can characterize the thermodynamic state of the entire atmosphere, providing a solid physical criterion for subsequent judgment on whether to introduce a numerical model.

[0011] Preferably, the water vapor advection intensity is determined based on the vector synthesis method, which includes: firstly, calculating the difference between the total zenith delay of the target monitoring station and the total zenith delay between each neighboring monitoring station, as well as the distance between the target monitoring station and each neighboring monitoring station, and using the ratio of the difference to the distance as the discrete gradient in the direction from the target monitoring station to the neighboring monitoring station; projecting the discrete gradient in the direction from the target monitoring station to all neighboring monitoring stations onto two orthogonal coordinate axes, east-west and north-south, using sine and cosine functions, and accumulating them to obtain two orthogonal components; and using the Euclidean norm to calculate the combined vector length of the two orthogonal components as the water vapor advection intensity of the target monitoring station.

[0012] This technical solution, through vector synthesis, can sense whether water vapor in the vicinity exhibits a steep slope with one side high and the other low, i.e., strong advection characteristics. The calculated water vapor advection intensity can be used as a key dynamic operator and input into the subsequent water vapor non-uniformity factor, which can transform microscopic differences between stations into macroscopic atmospheric dynamic characteristics. This enables the inversion system to identify exogenous water vapor interference events, thus providing a solid dynamic criterion for adaptively increasing the weight of numerical models under complex weather conditions dominated by the stratosphere.

[0013] Preferably, the determination of the water vapor non-uniformity factor based on the combined carrier phase verification residual, water vapor advection intensity, and thermodynamic vertical decoupling index is performed according to the following relationship:

[0014] in, The water vapor heterogeneity factor for the target monitoring station, The thermodynamic vertical decoupling index of the target monitoring station. This refers to the total number of BeiDou satellites that the target monitoring station can receive. The serial number of the BeiDou satellite. The first one received by the target monitoring station Carrier phase verification residuals in BeiDou satellite observation data corresponding to each BeiDou satellite For the first The elevation angle of each BeiDou satellite, The water vapor advection intensity at the target monitoring station, It is a sine function. This is the preset standard residual.

[0015] This technical solution achieves deep coupling of vertical, horizontal micro, and network macro information by multiplying the thermodynamic vertical decoupling index with a combination of the residual weighted root mean square and the horizontal advection intensity. The residuals of satellite observation data usually reflect local micro turbulence, while the introduced water vapor advection intensity can keenly capture the macroscopic transport of passing water vapor clouds. Using the vertical instability index as an amplifier, when the atmospheric vertical structure is chaotic, the micro residual fluctuations and the macroscopic advection gradient are simultaneously amplified and identified as effective water vapor anomaly signals. Conversely, when the atmosphere is stable, these fluctuations are suppressed and regarded as background noise. This multi-dimensional anisotropy compensation mechanism effectively solves the limitations of traditional methods that only use single-station information and improves the ability to perceive mobile heavy precipitation.

[0016] Preferably, mapping the water vapor heterogeneity factor to an adaptive collaborative weight for adjusting the multi-source data fusion ratio includes: obtaining the mean and standard deviation of the water vapor heterogeneity factor of the target monitoring station within a preset historical time period at the current moment, using it as historical statistical features, and constructing a probability distribution model based on statistical moments based on the historical statistical features. ;in, It is an adaptive collaborative weight. It is the water vapor heterogeneity factor of the target monitoring station. and These are the mean and the standard deviation, respectively.

[0017] This technical solution constructs a nonlinear response curve similar to a Butterworth filter. By adaptively defining the boundary of anomalies using the statistical characteristics of historical data and following the three-standard-deviation criterion in statistics, the algorithm can adapt to different seasons and regional climate backgrounds without the need for manually setting fixed empirical thresholds. The exponent term is 4, which makes the weight curve exhibit a flat-topped and steep-edge characteristic: when the anisotropy factor does not exceed the statistical anomaly threshold, the weight is kept around 1, ensuring that the system prioritizes the use of high real-time ground observation models. Once the threshold is exceeded, the weight rapidly decays to 0, forcibly switching to numerical mode dominance. This maximizes the system's response speed and stability while ensuring inversion accuracy.

[0018] Preferably, the temperature in the ground-measured data and the temperature in the forecast model data are weighted and fused using adaptive collaborative weights, including: calculating a first temperature based on the temperature in the ground-measured data using the Bevis formula; calculating a second temperature based on the temperature in the forecast model data using an integral method; weighting the first temperature using adaptive collaborative weights to obtain a first weighted temperature; weighting the second temperature by subtracting the adaptive collaborative weights from 1 to obtain a second weighted temperature; and using the sum of the first weighted temperature and the second weighted temperature as the dynamic weighted temperature.

[0019] This technical solution implements a soft-switching fusion strategy based on adaptive collaborative weights. This weighted fusion method avoids abrupt changes in the inversion results under critical conditions and ensures the temporal continuity of meteorological parameters. Through the dynamic adjustment of adaptive collaborative weights, this method actually constructs an adaptive transition physical model: during periods of stable weather, it is a high temporal resolution empirical regression model; during periods of severe convection, it smoothly transitions into a numerical physical model containing fine vertical structures. This mechanism solves the problem that a single model cannot simultaneously take into account both high temporal resolution and adaptability to complex vertical structures.

[0020] Preferably, the method for retrieving atmospheric water content is as follows: calculate the difference between the total zenith delay and the static zenith delay of the target monitoring station as the wet zenith delay of the target monitoring station; construct a water vapor conversion coefficient model based on the wet zenith delay and the pre-acquired atmospheric physical parameters; replace the temperature parameters in the water vapor conversion coefficient model with dynamic weighted temperature to correct the water vapor conversion coefficient model; and use the corrected water vapor conversion coefficient model to invert and calculate the atmospheric water content.

[0021] This technical solution addresses the issue that traditional inversion methods based on the Bevis formula typically use fixed regression formulas based on surface temperature to estimate weighted average temperature. This can introduce significant systematic biases under conditions of temperature inversion or severe convective weather. By injecting dynamically weighted temperature, which integrates vertical structure information from numerical models and horizontal anisotropy characteristics from satellite observations, into the core parameter positions of the water vapor conversion coefficient model, the solution achieves physical calibration of the water vapor conversion coefficient model. This eliminates conversion errors caused by atmospheric stratification instability, ensures the physical authenticity of the conversion process from zenith wet delay to atmospheric water content, and enables the inversion results to not only have high temporal resolution but also robustness against extreme weather interference, thereby improving the accuracy of water vapor inversion under all-weather conditions.

[0022] Preferably, after acquiring the BeiDou satellite observation data, forecast model data, and ground measurement data of the target monitoring station and neighboring monitoring stations, the following alignment operation is performed: for any one of the target monitoring station or neighboring monitoring stations, the time reference of the forecast model data and ground measurement data of that monitoring station is kept consistent with the time reference of the BeiDou satellite observation data; based on the geographical location of the monitoring station, spatial matching processing is performed on the forecast model data; and temporal interpolation processing is performed on the forecast model data to make the temporal resolution of the forecast model data consistent with the temporal resolution of the BeiDou satellite observation data.

[0023] Preferably, the method for extracting carrier phase verification residuals and zenith total delay from BeiDou satellite observation data is as follows: For any one of the target monitoring station or neighboring monitoring stations, acquire BeiDou satellite observation data corresponding to all BeiDou satellites that the monitoring station can receive. Construct a positioning observation model for the monitoring station based on the BeiDou satellite observation data. The zenith total delay is set as the parameter to be solved in the positioning observation model. Solve the positioning observation model using a parameter estimation algorithm to obtain the zenith total delay of the monitoring station. Substitute the zenith total delay into the positioning observation model to obtain the theoretical value of the BeiDou satellite observation data corresponding to each BeiDou satellite that the monitoring station can receive. Determine the difference between the actual value and the theoretical value of the BeiDou satellite observation data as the carrier phase verification residual of the BeiDou satellite observation data corresponding to that BeiDou satellite.

[0024] Preferably, the method for extracting zenith static delay from ground-based measured data is as follows: For any one of the target monitoring station or neighboring monitoring stations, obtain the air pressure, latitude, and altitude of the monitoring station from the ground-based measured data, and calculate the zenith static delay of the monitoring station based on the air pressure, latitude, and altitude using the Saastamoinen model.

[0025] The present invention has the following effects: This invention analyzes the thermodynamic vertical structure, dynamic horizontal turbulence, and regional water vapor advection characteristics of the atmosphere in real time. It constructs a thermodynamic vertical decoupling index and a water vapor heterogeneity factor to adaptively adjust the weights of the ground observation model and the numerical prediction model in the weighted average temperature calculation. This effectively suppresses the interference of inversion layers, strong turbulence, and exogenous advection on the inversion accuracy, and improves the accuracy and reliability of atmospheric water content inversion under extreme weather conditions. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram showing the trend of the thermodynamic vertical decoupling index and water vapor advection intensity of the target monitoring station of the present invention changing over time. Figure 3 This is a schematic diagram comparing the atmospheric water content obtained by the collaborative inversion of this invention with the atmospheric water content obtained by inversion using existing technologies. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0028] refer to Figure 1 This invention provides a method for collaborative inversion of BeiDou atmospheric water content based on multi-source heterogeneous data fusion, comprising: S1: Acquire BeiDou satellite observation data, forecast model data and ground measurement data from the target monitoring station and neighboring monitoring stations, and extract carrier phase post-verification residual, total zenith delay and zenith static delay.

[0029] A monitoring station equipped with a BeiDou satellite navigation system receiver and ground meteorological sensors (air pressure, temperature, humidity) is selected as the target monitoring station. Other monitoring stations within a pre-defined radius of 50km-100km are designated as all neighboring monitoring stations of the target monitoring station to construct a regional gradient observation network.

[0030] First, for the target monitoring station or any of the neighboring monitoring stations, the GNSS (Global Navigation Satellite System) receiver (or BeiDou satellite navigation receiver) of that monitoring station is used to acquire the BeiDou satellites that the monitoring station can receive, as well as the BeiDou satellite observation data corresponding to each BeiDou satellite. The BeiDou satellite observation data includes observation files and navigation files, specifically containing carrier phase observations, pseudorange observations, and broadcast ephemeris, used to reflect the satellite signal transmission time, reception time, geometric distance along the propagation path, and the delay caused by signal propagation in the atmosphere. Forecast model data for that monitoring station is acquired through a network communication interface (such as an FTP server). Forecast model data includes gridded data such as stratified air temperature, stratified air pressure, stratified specific humidity, and geopotential height, used to reflect the vertical distribution of the physical properties of the atmosphere above the monitoring station. Ground-based measured data, including surface air pressure, surface air temperature, and relative humidity, is acquired through the AWS (Automatic Weather Station) sensors of the monitoring station, used to reflect the thermodynamic boundary conditions at the bottom of the atmospheric boundary layer.

[0031] Next, for any one of the target monitoring station or neighboring monitoring stations, the carrier phase verification residual and zenith total delay are extracted from the BeiDou satellite observation data. The specific method is as follows: In GNSS meteorology and geodesy, the total zenith delay is the sum of the delays caused by a satellite signal passing through the entire atmosphere in the zenith direction (vertical direction). For a specific monitoring station, at a specific moment, it corresponds to only a specific geographical location and an air column perpendicular to the ground surface. Therefore, a monitoring station has only one total zenith delay and one zenith static delay. Although there are multiple satellites, the delay projections in each direction will be attributed to the zenith direction during the calculation.

[0032] Therefore, we acquire all the BeiDou satellites that each monitoring station can receive, as well as the BeiDou satellite observation data corresponding to each BeiDou satellite. Based on the BeiDou satellite observation data, we construct a positioning observation model for the monitoring station, specifically an Ionosphere-Free Combination (IF) observation equation in GNSS that eliminates the first-order influence of the ionosphere. The total zenith delay is the parameter to be solved in this observation equation. Based on the carrier phase observation values ​​of all the BeiDou satellites that the monitoring station can receive from the observation equation, we solve the observation equation simultaneously and use the least squares method to estimate the parameters of the observation equation, calculate the total zenith delay, substitute the total zenith delay into the positioning observation model to obtain the theoretical value of the carrier phase observation value in the BeiDou satellite observation data corresponding to each BeiDou satellite, and determine the difference between the actual value and the theoretical value of the carrier phase observation value as the carrier phase post-hoc residual of the BeiDou satellite observation data corresponding to that BeiDou satellite. The significance of carrier phase post-verification residuals lies in the fact that they contain information on atmospheric high-frequency jitter, multipath effects, and microscale water vapor turbulence that the model failed to absorb, and are used to reflect the non-uniform changes and micro-turbulence characteristics of the atmosphere in the line-of-sight direction.

[0033] Subsequently, the zenith static delay was extracted from the ground-based measured data, including: The Saastamoinen model is a classic meteorological model for estimating dry delay, used to accurately calculate the zenith static delay caused by dry air mass based on surface pressure. By obtaining the surface pressure, latitude, and altitude of the monitoring station from its measured surface data, the zenith static delay of the station is calculated using the Saastamoinen model based on these parameters.

[0034] Finally, spatiotemporal alignment is performed: For any monitoring station, either the target station or a neighboring station, the time reference of its forecast model data and ground-based measured data is aligned with the time reference of the BeiDou satellite observation data. Specifically, this is achieved by converting all data time labels to Coordinated Universal Time (UTC). Based on the station's geographical location, spatial matching is performed on the forecast model data. This is done using bilinear interpolation to map the data from the four nearest grid points of the forecast model data to the target station's location, based on the station's latitude and longitude. Temporal interpolation is then performed on the forecast model data to ensure its temporal resolution matches that of the BeiDou satellite observation data. This is achieved using cubic spline interpolation to match the lower temporal resolution of the forecast model data with that of the BeiDou observation data.

[0035] S2: Based on the temperature in the forecast model data, calculate the vertical gradient deviation of temperature with altitude, and integrate the vertical gradient deviation with altitude to determine the thermodynamic vertical decoupling index.

[0036] To address the problem that existing technologies cannot detect upper-level temperature inversions or complex stratifications by relying solely on ground data, this step aims to establish a vertical probe mechanism that utilizes temperature and pressure data from multiple isobaric surfaces (e.g., 37 layers from 1000 hPa to 1 hPa) or geometric height layers provided by forecast model data.

[0037] By utilizing standard isobaric surface stratification information (including temperature and pressure data at different altitudes) provided by forecast model data, the deviation of the current atmospheric vertical structure from the ideal physical model is reflected. This step, which quantifies the deviation of the current atmospheric vertical structure from the ideal physical model, is crucial for identifying atmospheric thermodynamic anomalies. Its purpose is to generate a macroscopic decoupling index to indicate whether the representativeness of ground-based measured data in the vertical dimension has failed.

[0038] Specifically, at the current moment, the thermodynamic vertical decoupling index of the target monitoring station is determined based on the following relationship:

[0039] in, The thermodynamic vertical decoupling index of the target monitoring station is a dimensionless scalar value. The ground elevation of the target monitoring station. This represents the top altitude of the troposphere, obtained through radiosonde data or standard atmospheric models. Let be the height integral variable, representing the vertical height, where , and The unit for all values ​​is kilometers. For forecast model data in Temperature value at that location, Represents the partial differential symbol. This represents the vertical gradient of temperature as a function of altitude. The preset dry adiabatic lapse rate constant is set to 9.8 K / km. For forecast model data in The air pressure value at that location, These are air pressure values ​​measured on the ground. It is the absolute value symbol. For an integral infinitesimal element, The value is 288.15 Kelvin, representing standard atmospheric temperature.

[0040] In this relation, Used to eliminate dimensional differences, For dimensionless scalar values, the first term This represents the absolute value of the deviation between the actual vertical lapse rate of atmospheric temperature and the ideal dry adiabatic lapse rate. The larger the value of the first term, the greater the deviation between the actual vertical lapse rate of atmospheric temperature and the ideal dry adiabatic state, indicating a more unstable atmospheric stratification and a higher likelihood of temperature inversion or strong convection. Conversely, a smaller value indicates a more stable atmospheric stratification that conforms to the ideal thermodynamic model. This represents the real-time rate of change of temperature with altitude given by the forecast model data; that is, the actual atmospheric temperature gradient measured by the target monitoring station at the current moment. This reflects the rate of temperature change of a dry air mass during vertical movement, solely due to expansion or compression. (Second term) This reflects the physical concept of atmospheric mass density weighting. The lower troposphere has a high atmospheric density and extremely rich water vapor content, accounting for the vast majority (over 90%) of the total delay, while the upper atmosphere is thin and dry. If temperature anomalies occur in the upper atmosphere, their impact on water vapor inversion is actually very small; however, if temperature anomalies occur in the lower atmosphere, the impact is huge. Therefore, using the characteristic of air pressure decreasing exponentially with altitude as a weight can highlight the contribution of lower atmospheric thermal anomalies to the inversion results, suppress upper-level noise, and is consistent with the physical facts of atmospheric refraction. By multiplying the first and second terms, the physical weighting of temperature gradient anomalies is achieved using the characteristics of atmospheric density distribution.

[0041] At each altitude, the first and second terms are combined by multiplication to represent the thermal anomaly energy density at that altitude. Finally, through integration, the thermal anomalies of each micro-layer in the vertical direction above the target monitoring station are multiplied by density weights and accumulated, thus achieving macroscopic quantification of the instability of the entire atmospheric layer above the target monitoring station.

[0042] The thermodynamic vertical decoupling index of a target monitoring station reflects the degree of deviation of the entire atmospheric vertical structure from the ideal dry adiabatic state. A smaller thermodynamic vertical decoupling index, approaching 0, indicates a stable atmospheric vertical structure, with temperature changes with altitude close to the ideal adiabatic state, suggesting the atmosphere conforms to the standard model. A larger thermodynamic vertical decoupling index, approaching 1, indicates the presence of a strong temperature inversion layer or convective disturbance within the atmosphere, causing the actual temperature gradient to deviate significantly from the theoretical value. This is because the thermodynamic vertical decoupling index is essentially a physical quantity that quantifies the disorder of atmospheric stratification.

[0043] S3: Calculate the water vapor advection intensity based on the spatial distribution difference of the total zenith delay between the target monitoring station and neighboring monitoring stations.

[0044] After obtaining the thermodynamic vertical decoupling index of the target monitoring station, given that atmospheric water vapor distribution is not only stratified vertically but also exhibits anisotropy in the horizontal direction due to turbulence (inconsistent water vapor content in different directions) and regional transport due to advection, the carrier phase a priori residuals of BeiDou satellite observation data, while containing local turbulence information (local refers to small-scale atmospheric changes only in the vertical line of sight of a single monitoring station), are insufficient to reflect macroscopic horizontal advection. The residuals of a single monitoring station can only show the fluctuations in the vertical observation data of its environment, but it is unclear whether these fluctuations are caused by local thermal convection or by advection caused by a large typhoon rain belt drifting laterally from a nearby region. Therefore, it is necessary to introduce the gradients of neighboring monitoring stations for judgment. Thus, this step introduces prior information in the vertical dimension as a gating mechanism and integrates the horizontal gradient information of neighboring monitoring stations to perform a full-dimensional physical identification of the atmospheric state at the current moment.

[0045] Specifically, at the current moment, the spatial distribution difference of the total zenith delay between the target monitoring station and neighboring monitoring stations is analyzed, and the water vapor advection intensity is determined using the vector synthesis method: First, calculate the difference between the total zenith delay of the target monitoring station and the total zenith delay between each neighboring monitoring station, as well as the distance between the target monitoring station and each neighboring monitoring station. Then, use the ratio of the difference to the distance as the discrete gradient in the direction from the target monitoring station to that neighboring monitoring station. The discrete gradients from the target monitoring station to all neighboring monitoring stations are projected onto two orthogonal coordinate axes, east-west and north-south, using sine and cosine functions. The gradients are then accumulated to obtain two orthogonal components. The combined vector length of the two orthogonal components is calculated using the Euclidean norm and is taken as the water vapor advection intensity of the target monitoring station.

[0046] Specifically, the vector synthesis process of water vapor advection intensity satisfies the following relationship:

[0047] in, The water vapor advection intensity at the target monitoring station reflects the macroscopic intensity of water vapor transport in the environment where the station is located. A higher water vapor advection intensity indicates the presence of a strong exogenous weather system (such as a front or typhoon) passing through, and that the water vapor distribution in the target monitoring station's environment is extremely uneven. Conversely, a lower water vapor advection intensity indicates a relatively uniform water vapor distribution in the target monitoring station's environment, and less influence from exogenous transport. The number of neighboring monitoring stations of the target monitoring station. The total zenith delay of the target monitoring station, For the target monitoring station Total zenith delay of neighboring monitoring stations, For the target monitoring station and the first The distance to the nearest monitoring station For the first The azimuth angle of each neighboring monitoring station relative to the target monitoring station is obtained by using the latitude and longitude coordinates of the target monitoring station and neighboring monitoring stations and calculating the true north azimuth angle using inverse trigonometric functions.

[0048] In this relation, the first term This represents the water vapor gradient component along the north-south direction. A larger value indicates a more significant difference in water vapor distribution along the north-south direction, while a smaller value indicates that the water vapor distribution along the north-south direction tends to be more uniform. This represents the discrete gradient in the direction from the target monitoring station to the nearest monitoring station. This indicates that the gradient is projected onto the north-south direction. Multiplying the two components represents the projected component of the discrete gradient in the north-south direction, and summing them represents the total water vapor gradient component in the north-south direction. The second term... This represents the water vapor gradient component along the east-west direction. The larger the value, the more significant the difference in water vapor density along the east-west direction. This indicates that the gradient is projected onto the east-west direction, and... Multiplication represents the projected components of the discrete gradient in the east-west direction, and summation represents the total water vapor gradient components in the east-west direction. Taking the square root of the sum of the squares of the first and second terms is a standard vector composition operation, ultimately yielding a scalar value that characterizes the overall intensity and direction of water vapor transport in the region.

[0049] Thus, the water vapor advection intensity of the target monitoring station is obtained through vector synthesis. This intensity indicates whether the water vapor in the vicinity of the target monitoring station exhibits a steep slope with one side higher than the other, i.e., whether it possesses strong advection characteristics. A higher water vapor advection intensity indicates stronger macroscopic transport of the passing water vapor cloud, signifying stronger advection characteristics. This implies the presence of a strong exogenous weather system (such as a front or typhoon) passing through, resulting in extremely uneven water vapor distribution within the region and a large amount of external water vapor being transported into the target area, and vice versa. It should be noted that when calculating the water vapor advection intensity of the target monitoring station, the unit of distance is converted to be consistent with the unit of total zenith delay.

[0050] S4: Determine the water vapor non-uniformity factor by combining the post-carrier phase verification residual, water vapor advection intensity, and thermodynamic vertical decoupling index.

[0051] After obtaining the water vapor advection intensity of the target monitoring station, considering that a single data source cannot simultaneously take into account both microscopic turbulence and macroscopic advection characteristics, this step aims to construct a comprehensive index that can fully characterize the anisotropic characteristics of the atmosphere. This is achieved by using the thermodynamic vertical decoupling index as a physical gate and adaptively weighting the superposition results of microscopic turbulence and macroscopic advection.

[0052] Specifically, the water vapor heterogeneity factor satisfies the following relationship:

[0053] In this relation, The water vapor non-uniformity factor of the target monitoring station is used to comprehensively characterize the complexity and non-uniformity of the spatial distribution of water vapor at the current moment. The larger the water vapor non-uniformity factor, the more unstable the atmosphere is and the stronger the spatial non-uniformity. The smaller the water vapor non-uniformity factor, the more stable the atmosphere is and the more uniform the spatial distribution. The thermodynamic vertical decoupling index of the target monitoring station. This refers to the total number of BeiDou satellites that the target monitoring station can receive. The serial number of the BeiDou satellite. The first one received by the target monitoring station The carrier phase residual in the BeiDou satellite observation data corresponding to each BeiDou satellite, in millimeters. For the first The elevation angle of each BeiDou satellite, The standard residual is a preset value of 1, in millimeters. The water vapor advection intensity at the target monitoring station, It is a sine function.

[0054] In this relation, the first term is This plays a role in nonlinear physical gating; when the vertical atmospheric stratification is extremely unstable, it is usually accompanied by strong convective motion. When the magnitude is large, the observed horizontal residual fluctuations or advection gradients are highly likely to be effective signals caused by severe weather (such as thunderstorms or fronts). Therefore, through a larger... Amplifying this instability leads to a significant increase in the final water vapor heterogeneity factor; conversely, when the atmospheric vertical stratification is stable, Approaching zero, small horizontal fluctuations are more likely to be measurement noise or terrain multipath effects, therefore, through smaller... This unstable characteristic should be suppressed to avoid false alarms.

[0055] In this relation, the second term is the weighted root mean square obtained by dimensionless calculation. and dimensionless water vapor advection intensity By superimposing the micro-turbulence and macro-advection intensities, a physical superposition combination of micro-turbulence and macro-advection intensity was achieved, which comprehensively characterized the horizontal non-uniformity of the target monitoring station after the superposition of micro-turbulence and macro-advection. This forms a weighted root mean square (RMS) form. The larger the weighted RMS, the greater the jitter of the BeiDou satellite signal as it passes through the atmosphere along the direction from the target monitoring station to the nearest monitoring station. This reflects the microscopic turbulence dispersion in that direction. The sine value of the elevation angle of the Beidou satellite is represented by... right The significance of weighting is to reduce the multipath noise weight introduced by low-elevation-angle satellites due to their long propagation paths. The concept of network collaboration is introduced, forming a physical coupling term that combines macroscopic and microscopic characteristics. This reflects the horizontal gradient vector constructed using the difference in total zenith delay and distance between the target monitoring station and its neighboring monitoring stations, reflecting the macroscopic advection intensity of the target monitoring station and its neighborhood. The dimensional anomaly was eliminated. It is a dimensionless value, and then combined with a dimensionless value. Superposition eliminates the inconsistency in dimensions between microscopic turbulence characteristics and macroscopic advection intensity during calculation.

[0056] when A large value indicates severe weather / atmospheric instability, suggesting a disconnect between the ground and upper atmosphere, rendering ground-based measurements unreliable. Therefore, the water vapor conversion coefficient model should place greater trust in forecast model data at this point, as forecast models contain vertical layer information and can detect temperature inversions. When the value is very low, it indicates that the weather is stable, the ground and upper atmosphere are in good agreement, and the ground measurement data is very accurate and has high real-time performance. Therefore, the ground measurement data should be trusted more at this time.

[0057] Thus, the first and second terms are coupled through multiplication. This is due to the thermodynamic correlation in the physical scenario: when the atmospheric vertical stratification is extremely unstable, it is usually accompanied by strong convection, resulting in a large thermodynamic vertical decoupling index. In this case, both the horizontal residual and the advection gradient are highly likely to be caused by severe weather. The amplification of the anomalous signal through a large thermodynamic vertical decoupling index indicates a high probability of encountering severe weather. The thermodynamic vertical decoupling index amplifies and confirms this, resulting in a very large final water vapor heterogeneity factor. Therefore, in the subsequent process of using the water vapor conversion coefficient model to retrieve atmospheric water content, the water vapor conversion coefficient model should place greater trust in the forecast model data. The forecast model data includes vertical stratification information, revealing temperature inversions, while ground-based data is distorted at this point. Conversely, if the atmospheric vertical stratification is very stable, the thermodynamic vertical decoupling index is small. Even if there is high-frequency non-systematic noise in the horizontal direction or a difference in water vapor background due to topographic differences between adjacent stations, it is more likely caused by noise or topographic differences. Therefore, it is suppressed by a small thermodynamic vertical decoupling index. In this case, the thermodynamic vertical decoupling index plays a role in suppression and filtering. In the subsequent process of using the water vapor conversion coefficient model to retrieve atmospheric water content, the water vapor conversion coefficient model should rely more on the ground-based measured data, because when the atmosphere is stable, the ground-based data has high real-time performance and can represent the state of the atmosphere above.

[0058] refer to Figure 2 As shown in the figure, this diagram illustrates the temporal evolution of the thermodynamic vertical decoupling index and water vapor advection intensity at the target monitoring station during a typical severe convective weather event. During periods of stable weather, the thermodynamic vertical decoupling index remains at a low oscillation level, indicating stable atmospheric vertical stratification. During the duration of convection, the curve shows a significant upward trend, with the thermodynamic vertical decoupling index increasing sharply, accompanied by high-frequency and large-amplitude fluctuations in water vapor advection intensity. This synchronous change verifies that the physical index constructed in this invention can sensitively capture the instability of the atmospheric thermodynamic structure, causing the water vapor non-uniformity factor to exceed the historical background noise level. This provides a reliable data trigger signal for adaptively reducing the weight of ground-based measured data and switching to forecast model data dominance in subsequent steps.

[0059] S5: Based on the historical statistical characteristics of the water vapor non-uniformity factor, construct a distribution probability model based on statistical moments, and determine an adaptive collaborative weight for adjusting the fusion ratio of multi-source data.

[0060] After obtaining the water vapor non-uniformity factor of the target monitoring station, this step aims to determine an adaptive collaborative weight for adjusting the multi-source data fusion ratio in order to achieve atmospheric water content inversion, rather than relying solely on the linear regression model of surface temperature as in the traditional method.

[0061] From a time perspective, considering the significant regional and seasonal differences in the fluctuation range of water vapor heterogeneity factors under different meteorological conditions, a relative decision mechanism based on historical data is established. By analyzing the statistical distribution of water vapor heterogeneity factors over a period of time, an adaptive probability model is constructed. This process essentially allows the system to learn the meteorological background noise level of the target monitoring station's location, with the aim of generating a standardized adaptive collaborative weight.

[0062] The reliability of data from different sources varies under different atmospheric conditions. When the atmosphere is in a non-stationary state (such as strong convection or temperature inversion), data from numerical weather prediction models are more accurate and more suitable for reliable inversion operations because they contain vertical stratification information. When the atmosphere is in a stationary state, ground-based measured data are more accurate because they are more real-time and closer to the true ground values, making them more suitable for reliable inversion operations.

[0063] First, the past 24 hours are taken as the historical time period. The water vapor heterogeneity factor of the target monitoring station at all historical moments within this historical time period is obtained, and the average value of the water vapor heterogeneity factor is calculated. and standard deviation As historical statistical features, a probability distribution model based on statistical moments is constructed based on these historical statistical features:

[0064] In this relation, For adaptive collaborative weights, It is the water vapor heterogeneity factor of the target monitoring station. and These are the mean and standard deviation of the water vapor non-uniformity factor over the historical time period, respectively. Setting the exponent to 4 constructs a high-order nonlinear decay function, similar to the amplitude-frequency response characteristics of a Butterworth low-pass filter.

[0065] In this relation, It is the core regulatory mechanism. Based on the three-standard-deviation criterion in statistics, the upper bound of the normal fluctuation of the water vapor heterogeneity factor under the meteorological background of the target monitoring area is defined. This indicates that the atmosphere is stable. Less than 1, becomes extremely small after being raised to the fourth power, making If it approaches 1, then This means that under normal or slightly disturbed weather conditions, the system determines that the atmosphere is in a stable state, trusts ground-based measured data, and prioritizes the use of highly real-time ground-based measured data; when This indicates the occurrence of a sudden, strong convection. Greater than 1, amplified significantly by the fourth power, making A sharp increase, leading to Rapid decay means that the system has detected extremely uneven spatial distribution of atmospheric water vapor, such as when a thunderstorm cell is generated, the local area is extremely wet while the surrounding area is relatively dry. This rapidly reduces the confidence in ground-based measured data and switches to forecast model data as the dominant factor.

[0066] Thus, through adaptive constraints of statistical moments and nonlinear mapping of higher-order power functions, the inversion accuracy is not lost during the atmospheric steady period, and the accurate switching between the atmospheric steady state and the atmospheric anomalous disturbance state can be achieved through adaptive collaborative weighting when the atmosphere changes drastically. This achieves the effect of automatically adapting to the best multi-source data fusion strategy under different meteorological backgrounds.

[0067] S6: Adaptive collaborative weighting is used to weight and fuse the temperature in the ground-based measured data and the temperature in the forecast model data, and combined with the zenith static delay and the total zenith delay to inversely derive the atmospheric water content.

[0068] After generating adaptive collaborative weights, the final inversion stage is initiated. Considering that ground-based measured data has extremely high temporal resolution but lacks vertical information, while forecast model data (such as ERA5, the fifth-generation atmospheric reanalysis data released by the European Centre for Medium-Range Weather Forecasts (ECMWF), which is currently recognized as one of the most accurate gridded meteorological data in the world) has a complete vertical physical structure but is time-lagging, this step aims to perform a dynamic fusion operation that complements each other's strengths.

[0069] The temperature calculated based on ground-based measured data and the temperature based on forecast model data are dynamically weighted by adaptive collaborative weighting. The aim is to obtain a dynamically weighted temperature that can quickly respond to temperature changes and resist temperature inversion interference. Finally, combined with GNSS delay data, a high-precision atmospheric water content can be obtained.

[0070] First, based on the temperature data from ground measurements, the first temperature is calculated using the Bevis formula:

[0071] Wherein, T1 is the first temperature, reflecting the thermal state near the ground. It is 70.2. It is 0.72. The temperature data is based on actual ground measurements (unit: Kelvin). If it is in Celsius, it needs to be converted first. and It is a classic global regression coefficient derived by the Bevis team through analysis of data from 8,718 radiosonde stations in North America.

[0072] Based on the temperature in the forecast model data, the second temperature T2 is accurately calculated using the integral method, which reflects the true thermal structure of the entire atmosphere. Specifically, the second temperature is obtained by dividing the integral of the ratio of water vapor pressure to temperature over vertical height by the integral of the ratio of the square of water vapor pressure to temperature over vertical height.

[0073] Then, T1 and T2 are weighted and fused using adaptive collaborative weights:

[0074] in, For dynamically weighted temperature, For adaptive collaborative weights, T1 is the first temperature, and T2 is the second temperature.

[0075] When the atmosphere is stable , Mainly composed of The decision retains the advantage of minute-level updates from ground-based measured data, meaning that ground weather stations output data every minute, enabling them to keenly capture rapid changes in temperature; while forecast model data is typically updated only once an hour. Under stable weather conditions, using ground-based data allows for more real-time inversion results; when atmospheric anomalies occur, such as temperature inversions (cold at the surface and warm at the upper atmosphere) or severe convective weather, the advantage of ground-based data can be further enhanced. The correction (calculated by the forecast model) is because It is calculated through integration, which accurately reflects that the upper atmosphere is warm, thus avoiding the error of the Bevis formula, which mistakenly assumes that the upper atmosphere is also cold simply because the ground is cold. , Mainly composed of It was decided to use vertical information from forecast model data to correct the deviations caused by temperature inversion.

[0076] In this way, dynamic correction of the temperature model based on atmospheric physical state perception is achieved, ensuring the physical authenticity of the weighted average temperature under different weather conditions.

[0077] Finally, the atmospheric water content is retrieved and calculated, including: Calculated total zenith delay of the target monitoring station With zenith static delay The difference is used to obtain the zenith wet delay. The zenith wet delay reflects the amount of refraction delay of radio signals caused solely by the polar moments of water vapor molecules in the atmosphere.

[0078] The water vapor conversion coefficient model is corrected using dynamic weighted temperature. Specifically, the temperature parameter in the water vapor conversion coefficient model is replaced with a dynamically weighted temperature. The final atmospheric water content is then obtained by inversion using the corrected water vapor conversion coefficient model and the zenith wet delay, based on the following relationship:

[0079] In this relation, The atmospheric water content obtained from the inversion (unit: millimeters). Let be the density of liquid water, and take a value of . ; Let be the water vapor gas constant, and take the value of . ; and The refractive index is the empirical constant of the atmosphere, preferably... Values , Values , The temperature is dynamically weighted (unit: Kelvin). Zenith wet delay (in millimeters) for the target monitoring station.

[0080] In this relation, Part of this is a modified water vapor conversion coefficient model, which is used to convert observed zenith wet delay into atmospheric water content. The classic water vapor conversion coefficient model is... The core parameter is calculated using the classical Bevis formula based on temperature from measured ground data. By multiplying the modified water vapor conversion coefficient model with the zenith wet delay, the atmospheric water content was inverted.

[0081] Traditional atmospheric water content retrieval operations are based on temperature parameters calculated using linear regression models of surface air temperature (such as the Bevis formula) and zenith wet delay. Atmospheric water content and temperature parameters show a positive correlation. This invention introduces a dynamically weighted temperature obtained after thermodynamic decoupling and network anisotropy compensation correction. A physical correction factor is embedded in the denominator of the water vapor conversion coefficient model. Under complex weather conditions such as temperature inversion or strong convection, it can automatically identify representative errors in surface temperature and use vertical stratification information to correct temperature parameters to the true atmospheric thermodynamic state. This eliminates systematic biases caused by estimation errors in the water vapor conversion coefficient model, ensuring that the retrieved atmospheric water content is not only numerically accurate but also truly reflects the physical evolution of the weather.

[0082] refer to Figure 3As shown, the accuracy differences between the proposed collaborative inversion method and the traditional Bevis formula-based inversion method for atmospheric water content inversion are compared, with radiosonde data as the true reference. Comparative analysis shows that during stable weather periods, the inversion results of the two methods differ little and both closely match the true values. However, during periods of atmospheric thermodynamic instability (i.e., when temperature inversions or strong advection occur), the inversion results of existing technologies exhibit significant systematic biases, deviating considerably from the true values. This is because surface temperature cannot represent the upper-air thermal state. In contrast, the inversion curve of this invention, under the adjustment of adaptive collaborative weights, dynamically incorporates vertical stratification information from forecast model data, successfully correcting temperature conversion errors and maintaining a high degree of consistency with the true values. This fully demonstrates that under complex meteorological conditions, this invention, through physical-driven adaptive fusion of multi-source heterogeneous data, effectively suppresses environmental interference and significantly improves the accuracy and robustness of atmospheric water content inversion.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for collaborative inversion of BeiDou atmospheric water content based on multi-source heterogeneous data fusion, characterized in that, include: Acquire BeiDou satellite observation data, forecast model data, and ground measurement data from the target monitoring station and neighboring monitoring stations; Extract carrier phase verification residuals and total zenith delay from BeiDou satellite observation data; extract zenith static delay from ground-measured data; Based on the temperature in the forecast model data, the vertical gradient deviation of temperature with altitude is calculated, and the vertical gradient deviation is integrated with altitude to determine the thermodynamic vertical decoupling index; the water vapor advection intensity is calculated based on the spatial distribution difference of the total zenith delay between the target monitoring station and the neighboring monitoring stations. The water vapor non-uniformity factor is determined by combining the carrier phase post-verification residual, water vapor advection intensity, and thermodynamic vertical decoupling index. The historical statistical characteristics of the water vapor non-uniformity factor are obtained, and a distribution probability model based on statistical moments is constructed based on the historical statistical characteristics. The water vapor non-uniformity factor is mapped to an adaptive collaborative weight for adjusting the fusion ratio of multi-source data. Adaptive collaborative weighting is used to weight and fuse the temperature data in ground-based measured data and the temperature data in forecast model data, and atmospheric water content is inverted by combining zenith static delay and zenith total delay.

2. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The thermodynamic vertical decoupling index is determined based on the following relationship: ； in, The thermodynamic vertical decoupling index of the target monitoring station. The ground elevation of the target monitoring station. The height of the top of the troposphere. Let be the integral variable representing the vertical height. For forecast model data in Temperature value at that location, This represents the vertical gradient of temperature as a function of altitude. The preset dry adiabatic lapse rate constant, For forecast model data in The air pressure value at that location, These are air pressure values ​​measured on the ground. It is the absolute value symbol. For an integral infinitesimal element, This refers to the standard atmospheric temperature.

3. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The water vapor advection intensity is determined based on the vector composition method, including: First, calculate the difference between the total zenith delay of the target monitoring station and the total zenith delay between each neighboring monitoring station, as well as the distance between the target monitoring station and each neighboring monitoring station. Then, use the ratio of the difference to the distance as the discrete gradient in the direction from the target monitoring station to that neighboring monitoring station. The discrete gradients from the target monitoring station to all neighboring monitoring stations are projected onto two orthogonal coordinate axes, east-west and north-south, using sine and cosine functions. The gradients are then accumulated to obtain two orthogonal components. The combined vector length of the two orthogonal components is calculated using the Euclidean norm and is taken as the water vapor advection intensity of the target monitoring station.

4. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The determination of the water vapor non-uniformity factor is based on the following relationship: (The equation is incomplete and requires further context to be fully translated.) ； in, The water vapor heterogeneity factor for the target monitoring station, The thermodynamic vertical decoupling index of the target monitoring station. This refers to the total number of BeiDou satellites that the target monitoring station can receive. The serial number of the BeiDou satellite. The first one received by the target monitoring station Carrier phase verification residuals in BeiDou satellite observation data corresponding to each BeiDou satellite For the first The elevation angle of each BeiDou satellite, The water vapor advection intensity at the target monitoring station, It is a sine function. This is the preset standard residual.

5. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The water vapor non-uniformity factor is mapped to an adaptive collaborative weight for adjusting the multi-source data fusion ratio, including: The mean and standard deviation of the water vapor heterogeneity factor of the target monitoring station within a preset historical time period at the current moment are obtained as historical statistical features. Based on these historical statistical features, a probability distribution model based on statistical moments is constructed. ;in, It is an adaptive collaborative weight. It is the water vapor heterogeneity factor of the target monitoring station. and These are the mean and the standard deviation, respectively.

6. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The temperature data from ground-based measured data and the temperature data from forecast models are weighted and fused using adaptive collaborative weighting, including: The first temperature is calculated using the Bevis formula based on the temperature from ground-measured data; the second temperature is calculated using the integral method based on the temperature from forecast model data. The first temperature is weighted using adaptive collaborative weights to obtain the first weighted temperature. The second temperature is weighted using 1 minus the adaptive collaborative weights to obtain the second weighted temperature. The sum of the first weighted temperature and the second weighted temperature is used as the dynamic weighted temperature.

7. The BeiDou atmospheric water content collaborative inversion method according to claim 6, characterized in that, The method for retrieving atmospheric water content is as follows: The difference between the total zenith delay and the static zenith delay of the target monitoring station is calculated as the wet zenith delay of the target monitoring station. Based on the wet zenith delay and the pre-acquired atmospheric physical parameters, a water vapor conversion coefficient model is constructed. The temperature parameter in the water vapor conversion coefficient model is replaced with a dynamically weighted temperature to correct the water vapor conversion coefficient model. The atmospheric water content is then calculated by inversion using the corrected water vapor conversion coefficient model.

8. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, After acquiring the BeiDou satellite observation data, forecast model data, and ground measurement data from the target monitoring station and neighboring monitoring stations, the following alignment operation is also performed: For any one of the target monitoring stations or neighboring monitoring stations, the time reference of the forecast model data and ground-measured data of that monitoring station is kept consistent with the time reference of the BeiDou satellite observation data; based on the geographical location of the monitoring station, spatial matching processing is performed on the forecast model data; Time interpolation is performed on the forecast model data to ensure that the time resolution of the forecast model data is consistent with that of the BeiDou satellite observation data.

9. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The method for extracting carrier phase post-hoc residuals and total zenith delay from BeiDou satellite observation data is as follows: For any one of the target monitoring station or neighboring monitoring stations, acquire the BeiDou satellite observation data corresponding to all BeiDou satellites that the monitoring station can receive. Construct a positioning observation model for the monitoring station based on the BeiDou satellite observation data. The total zenith delay is set as the parameter to be solved in the positioning observation model. Solve the positioning observation model using a parameter estimation algorithm to obtain the total zenith delay of the monitoring station. Substitute the total zenith delay into the positioning observation model to obtain the theoretical value of the BeiDou satellite observation data corresponding to each BeiDou satellite that the monitoring station can receive. Determine the difference between the actual value and the theoretical value of the BeiDou satellite observation data as the carrier phase verification residual of the BeiDou satellite observation data corresponding to that BeiDou satellite.

10. The BeiDou atmospheric water content collaborative inversion method according to claim 1, characterized in that, The method for extracting zenith static delay from ground-based measured data is as follows: For any one of the target monitoring station or neighboring monitoring stations, obtain the air pressure, latitude, and altitude from the ground-measured data of that monitoring station, and calculate the zenith static delay of that monitoring station based on the air pressure, latitude, and altitude using the Saastamoinen model.

Images