Assimilation method based on GNSS multi-dimensional water vapor cooperation WRF meteorological mode

Abstract

The invention discloses an assimilation method based on a GNSS multi-dimensional water vapor cooperation WRF meteorological mode. The assimilation method comprises the steps of 1, GNSS water vapor inversion; 2, assimilating the precipitable water quantity inversed by the GNSS for the first time; 3, carrying out GNSS three-dimensional chromatography inversion; 4, interpolating the water vapor density value of the three-dimensional chromatography grid into the forecast three-dimensional grid to obtain a water vapor density value of a three-dimensional interpolation grid which is the same as the forecast three-dimensional grid; 5, carrying out inversion on the water vapor density value of the three-dimensional interpolation grid and the temperature data of the forecast three-dimensional grid to obtain a relative humidity profile; and 6, assimilating the relative humidity profile again. The method is reasonable in design, first assimilation is carried out by using the precipitable water volume inverted by the GNSS, the relative humidity profile is obtained by using the three-dimensional water vapor density and the temperature data of the first assimilation, the relative humidity profile is assimilated again based on the WRF meteorological mode, assimilation is effectively carried out by using the three-dimensional water vapor density data, and the meteorological parameter forecasting precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorological data assimilation, and in particular relates to an assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model. Background Art

[0002] Water vapor is a greenhouse gas in the atmosphere and plays a major role in driving the occurrence and development of weather. Precipitable water vapor (PWV) is one of the manifestations of water vapor content in the atmosphere, and its changes are an important driving factor affecting related weather processes. The water vapor detection technology of the Global Navigation Satellite System (GNSS) is not affected by weather and has the advantages of high precision, low cost, and continuous operation. It can provide high-precision, high-resolution multi-dimensional water vapor information. Therefore, the assimilation of the GNSS-inverted precipitable water vapor PWV based on the WRF meteorological model came into being to improve the forecast accuracy of meteorological parameters.

[0003] However, the current assimilation uses PWV, which only provides information on the water vapor content per unit area in the zenith direction of the GNSS station, and cannot fully reflect the water vapor distribution in the surrounding space. In addition, the three-dimensional water vapor density data obtained based on GNSS tomography technology has not been fully applied to the WRF meteorological model data assimilation, limiting its potential in improving the accuracy of meteorological parameter forecasts.

[0004] Therefore, a reasonably designed GNSS multidimensional water vapor assimilation method for the WRF meteorological model is needed. The precipitable water inverted by GNSS is first assimilated for the first time, and the three-dimensional water vapor density and the first assimilated temperature data are used to obtain the relative humidity profile. The relative humidity profile is then assimilated again based on the WRF meteorological model, and the three-dimensional water vapor density data is effectively assimilated to improve the accuracy of the initial water vapor field of the WRF meteorological model and improve the forecast accuracy of meteorological parameters. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide an assimilation method based on GNSS multidimensional water vapor and WRF meteorological model in response to the deficiencies in the above-mentioned prior art. The method has simple steps and reasonable design. The precipitable water inverted by GNSS is first assimilated for the first time, and the three-dimensional water vapor density and the temperature data assimilated for the first time are used to obtain the relative humidity profile, and the relative humidity profile is assimilated again based on the WRF meteorological model. The three-dimensional water vapor density data is effectively assimilated to improve the accuracy of the initial water vapor field of the WRF meteorological model and improve the forecast accuracy of meteorological parameters.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: an assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model, characterized in that the method comprises the following steps:

[0007] Step 1: GNSS water vapor inversion:

[0008] Step 101, using a computer to obtain the precipitable water volume PWV inverted by GNSS in the area to be studied according to the formula PWV=Π·ZWD; wherein ZWD represents the zenith wet delay, and Π represents the atmospheric water vapor conversion coefficient;

[0009] Step 102, repeating step 101 multiple times to obtain the precipitable water inverted by multiple GNSS stations; wherein the precipitable water inverted by the j-th GNSS station is recorded as PWV(j) by a computer, and the longitude, latitude and altitude corresponding to the j-th GNSS station are obtained and recorded as lat1(j), lon1(j) and h1(j) respectively;

[0010] Step 2: First assimilation of GNSS-derived precipitable water:

[0011] Step 201: using a computer to convert the longitude lat1(j), latitude lon1(j), altitude h1(j), precipitable water volume PWV(j), and first error Δ1(j) corresponding to the j-th GNSS station into precipitable water volume data in Little_R format;

[0012] Step 202: Using a computer, the Little_R formatted precipitable water data is input into the observation value processor OBSPROC module of the WRF meteorological model and processed into an observation value file for WRFDA;

[0013] Step 203: using a computer to input the observation value file for WRFDA into the WRFDA module of the WRF meteorological model for three-dimensional variational assimilation, and outputting a meteorological data file in NetCDF format;

[0014] Step 204, extracting temperature data of the forecast three-dimensional grid from the meteorological data file in NetCDF format;

[0015] Step 3: GNSS 3D tomography inversion:

[0016] Step 301, gridding the study area along longitude, latitude and altitude directions to form a three-dimensional tomographic grid; and using the GNSS water vapor tomography model to obtain the water vapor density inversion value of the three-dimensional tomographic grid;

[0017] Step 302: using the trained LSTM neural network model, inputting the oblique path water vapor content and the oblique path wet delay to obtain the water vapor density prediction value of the three-dimensional tomographic grid;

[0018] Step 303, performing mean processing on the water vapor density inversion value of the three-dimensional tomographic grid and the water vapor density prediction value of the three-dimensional tomographic grid to obtain the water vapor density value of the three-dimensional tomographic grid;

[0019] Step 4: interpolate the water vapor density value of the three-dimensional tomography grid into the forecast three-dimensional grid to obtain the water vapor density value of the three-dimensional interpolation grid that is the same as the forecast three-dimensional grid;

[0020] Step 5: Invert the water vapor density value of the three-dimensional interpolation grid and the temperature data of the forecast three-dimensional grid to obtain the relative humidity profile:

[0021] Step 501: Obtain the water vapor density value ρ of the i′th grid from the water vapor density value of the three-dimensional interpolation grid. 0i′ At the same time, the temperature T of the i′th grid is obtained from the temperature data of the forecast three-dimensional grid i′ and saturated water vapor pressure e si′ ; Wherein, i′ is a positive integer;

[0022] Step 502: Using a computer according to the formula Get the water vapor pressure of the i′th grid; where R v is the water vapor gas constant; M w represents the molar mass of water vapor;

[0023] Step 503: According to Get the relative humidity RH of the i′th grid i′ ;

[0024] Step 504, repeating step 503 multiple times to obtain a relative humidity profile; wherein the relative humidity profile includes the longitude, latitude, altitude of each grid and the relative humidity corresponding to each grid;

[0025] Step 6: Re-assimilation of relative humidity profile:

[0026] Step 601: convert the longitude, latitude, altitude of each grid, the relative humidity corresponding to each grid, and the second error corresponding to each relative humidity into relative humidity data in Little_R format;

[0027] Step 602: According to the method of step 202 and step 203, the relative humidity data in Little_R format is input into the WRF meteorological model for three-dimensional variational assimilation again, and a meteorological data file in NetCDF format is output.

[0028] The above-mentioned assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model is characterized in that: the acquisition of zenith wet delay ZWD and atmospheric water vapor conversion coefficient Π in step 101 is specifically carried out as follows:

[0029] Step A, using a computer to solve the GNSS station data using GAMIT / GLOBK software to obtain the zenith total delay ZTD;

[0030] Step B: Using the Saastamoinen model using a computer The zenith dry delay ZHD is obtained; where P represents the surface pressure at the GNSS station location, represents the latitude of the GNSS station, and H represents the geodetic height of the GNSS station;

[0031] Step C, using a computer to obtain the zenith wet delay ZWD according to the formula ZWD=ZTD-ZHD;

[0032] Step D: Using a computer based on Get the weighted average temperature T m ; Where I represents the number of atmospheric layers, i is the layer number, i and I are both positive integers, and 1≤i≤I; T i represents the absolute temperature of the top of the ith atmospheric layer, e i represents the water vapor pressure at the top of the i-th atmospheric layer, Δh i represents the thickness of the i-th atmospheric layer;

[0033] Step E: Use a computer to calculate the The atmospheric water vapor conversion coefficient Π is obtained; where ρ w represents the density of liquid water, R v is the water vapor gas constant, T m represents the weighted average temperature, k2' represents the first constant of atmospheric refractive index, and k3 represents the second constant of atmospheric refractive index.

[0034] The above-mentioned assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model is characterized in that: the LSTM neural network model trained in step 302 has the following specific process:

[0035] Step 3021: using the GNSS water vapor tomography model, obtaining the water vapor density inversion value of the three-dimensional tomography grid at each time inside the tomography model;

[0036] Step 3022, repeating step 3021 multiple times to obtain water vapor density inversion values ​​of the three-dimensional tomographic grid at multiple times, and using the oblique path water vapor content and oblique path wet delay at multiple times and the water vapor density inversion values ​​of the three-dimensional tomographic grid corresponding to the multiple times as a training set;

[0037] Step 3023: construct an LSTM neural network model;

[0038] Step 3024: use the oblique path water vapor content and oblique path wet delay in the training set as input, and the water vapor density value of the three-dimensional tomography grid as the output layer, and input them into the LSTM neural network model for training to obtain a trained LSTM neural network model.

[0039] The above-mentioned assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model is characterized by: Step 4, the specific process is as follows:

[0040] Step 401: record any grid intersection point in the forecast three-dimensional grid as the target grid (lat t ,lon t ,z t ), (lat t ,lon t ,z t ) are the longitude, latitude and altitude of the target grid respectively;

[0041] Step 402: Obtain the longitude and latitude of four adjacent grid centers on the lower horizontal plane projected downward by the target grid in the three-dimensional tomographic grid, and record the longitude and latitude of the first grid center as (lat1, lon1), the longitude and latitude of the second grid center as (lat1, lon2), the longitude and latitude of the third grid center as (lat2, lon1), and the longitude and latitude of the fourth grid center as (lat2, lon2); wherein the height of the lower horizontal plane is z1;

[0042] The water vapor density of the four adjacent grid centers is recorded as the water vapor density value of the first grid center ρ 11 , the water vapor density value ρ at the center of the second grid 12 , the water vapor density value ρ at the center of the third grid 21 , the water vapor density value ρ at the center of the fourth grid 22 ;

[0043] Step 403: According to ρ(lat t ,lon t ,z1)=(1-α)(1-β)ρ 11 +α(1-β)ρ 21 +β(1-α)ρ 12 +αβρ 22 , and obtain the water vapor density value ρ (lat t ,lon t ,z1); where α represents the first interpolation weight, and β represents the second interpolation weight, and

[0044]

[0045] Step 404: According to the method of step 402 and step 403, the water vapor density value ρ (lat t ,lon t ,z2);

[0046] Step 405: According to Get the water vapor density value ρ(lat t ,lon t ,z t );

[0047] Step 405, repeating steps 401 to 404 multiple times, interpolating the water vapor density value of the three-dimensional tomography grid into the forecast three-dimensional grid, and obtaining the water vapor density value of the three-dimensional interpolation grid that is the same as the forecast three-dimensional grid.

[0048] The above-mentioned assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model is characterized in that: the acquisition of the first error Δ1(j) in step 201 is specifically performed as follows:

[0049] The precipitable water obtained by the radiosonde is taken as the true value, and the error between the precipitable water obtained by the jth GNSS station and the true value is obtained by computer and recorded as Δ1(j);

[0050] The specific process of obtaining the second error in step 601 is as follows:

[0051] The relative humidity obtained by the radiosonde is taken as the true value, and the error between each relative humidity and the true value is obtained by computer and recorded as the second error.

[0052] Compared with the prior art, the present invention has the following advantages:

[0053] 1. The present invention performs the first assimilation based on the precipitable water inversion from GNSS, and the second assimilation based on the relative humidity profile, and uses the GNSS multi-dimensional water vapor information to improve the relative humidity information of the numerical model, thereby improving the prediction accuracy of the model.

[0054] 2. The present invention uses bilinear interpolation and altitude layer interpolation techniques to interpolate the water vapor density value of the three-dimensional tomographic grid into the forecast three-dimensional grid, generates high-precision data that can be used for assimilation and combines it with temperature data to convert it into an assimilable relative humidity profile, thereby achieving effective assimilation of water vapor information.

[0055] 3. The present invention combines the water vapor density inversion value of the three-dimensional tomographic grid obtained by the GNSS water vapor tomography model and the water vapor density prediction value of the three-dimensional tomographic grid obtained by the LSTM neural network model to obtain high-resolution three-dimensional water vapor information, and converts it to generate a relative humidity profile that can be directly assimilated, and then assimilates the relative humidity profile again to improve the initial field humidity information of the numerical forecast model. Through the assimilation of multidimensional water vapor data, the accuracy of the initial field humidity information of the numerical forecast model has been improved, laying the foundation for the precise numerical forecast of subsequent meteorological parameters.

[0056] 4. This invention makes up for the deficiency of traditional two-dimensional assimilation in vertical spatial resolution, enabling the WRF meteorological model to more comprehensively capture the spatiotemporal distribution characteristics of water vapor in the atmosphere, especially the enhanced detail description in the vertical direction, thus enhancing the WRF meteorological model's ability to simulate and forecast complex weather processes.

[0057] 5. The present invention is compatible with the existing WRF meteorological model and GNSS site observations, has high universality and practicability, and can be widely used in different regions and various types of meteorological forecasting tasks.

[0058] In summary, the method of the present invention has simple steps and reasonable design. The precipitable water inverted by GNSS is first assimilated for the first time, and the three-dimensional water vapor density and the temperature data assimilated for the first time are used to obtain the relative humidity profile, and the relative humidity profile is assimilated again based on the WRF meteorological model. The three-dimensional water vapor density data is effectively assimilated to improve the accuracy of the initial water vapor field of the WRF meteorological model and improve the forecast accuracy of meteorological parameters.

[0059] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The figure is a flowchart of the method of the present invention. DETAILED DESCRIPTION

[0061] like Figure 1 As shown, the assimilation method of the present invention based on GNSS multi-dimensional water vapor and WRF meteorological model includes the following steps:

[0062] Step 1: GNSS water vapor inversion:

[0063] Step 101, using a computer to obtain the precipitable water volume PWV inverted by GNSS in the area to be studied according to the formula PWV=Π·ZWD; wherein ZWD represents the zenith wet delay, and Π represents the atmospheric water vapor conversion coefficient;

[0064] Step 102, repeating step 101 multiple times to obtain the precipitable water inverted by multiple GNSS stations; wherein the precipitable water inverted by the j-th GNSS station is recorded as PWV(j) by a computer, and the longitude, latitude and altitude corresponding to the j-th GNSS station are obtained and recorded as lat1(j), lon1(j) and h1(j) respectively;

[0065] Step 2: First assimilation of GNSS-derived precipitable water:

[0066] Step 201: using a computer to convert the longitude lat1(j), latitude lon1(j), altitude h1(j), precipitable water volume PWV(j), and first error Δ1(j) corresponding to the j-th GNSS station into precipitable water volume data in Little_R format;

[0067] Step 202: Using a computer, the Little_R formatted precipitable water data is input into the observation value processor OBSPROC module of the WRF meteorological model and processed into an observation value file for WRFDA;

[0068] Step 203: using a computer to input the observation value file for WRFDA into the WRFDA module of the WRF meteorological model for three-dimensional variational assimilation, and outputting a meteorological data file in NetCDF format;

[0069] Step 204, extracting temperature data of the forecast three-dimensional grid from the meteorological data file in NetCDF format;

[0070] Step 3: GNSS 3D tomography inversion:

[0071] Step 301, gridding the study area along longitude, latitude and altitude directions to form a three-dimensional tomographic grid; and using the GNSS water vapor tomography model to obtain the water vapor density inversion value of the three-dimensional tomographic grid;

[0072] Step 302: using the trained LSTM neural network model, inputting the oblique path water vapor content and the oblique path wet delay to obtain the water vapor density prediction value of the three-dimensional tomographic grid;

[0073] Step 303, performing mean processing on the water vapor density inversion value of the three-dimensional tomographic grid and the water vapor density prediction value of the three-dimensional tomographic grid to obtain the water vapor density value of the three-dimensional tomographic grid;

[0074] Step 4: interpolate the water vapor density value of the three-dimensional tomography grid into the forecast three-dimensional grid to obtain the water vapor density value of the three-dimensional interpolation grid that is the same as the forecast three-dimensional grid;

[0075] Step 5: Invert the water vapor density value of the three-dimensional interpolation grid and the temperature data of the forecast three-dimensional grid to obtain the relative humidity profile:

[0076] Step 501: Obtain the water vapor density value ρ of the i′th grid from the water vapor density value of the three-dimensional interpolation grid. 0i′ At the same time, the temperature T of the i′th grid is obtained from the temperature data of the forecast three-dimensional grid i′ and saturated water vapor pressure e si′ ; Wherein, i′ is a positive integer;

[0077] Step 502: Using a computer according to the formula Get the water vapor pressure of the i′th grid; where R v is the water vapor gas constant; M w represents the molar mass of water vapor;

[0078] Step 503: According to Get the relative humidity RH of the i′th grid i′ ;

[0079] Step 504, repeating step 503 multiple times to obtain a relative humidity profile; wherein the relative humidity profile includes the longitude, latitude, altitude of each grid and the relative humidity corresponding to each grid;

[0080] Step 6: Re-assimilation of relative humidity profile:

[0081] Step 601: convert the longitude, latitude, altitude of each grid, the relative humidity corresponding to each grid, and the second error corresponding to each relative humidity into relative humidity data in Little_R format;

[0082] Step 602: According to the method of step 202 and step 203, the relative humidity data in Little_R format is input into the WRF meteorological model for three-dimensional variational assimilation again, and a meteorological data file in NetCDF format is output.

[0083] In this embodiment, the specific process of obtaining the zenith wet delay ZWD and the atmospheric water vapor conversion coefficient Π in step 101 is as follows:

[0084] Step A, using a computer to solve the GNSS station data using GAMIT / GLOBK software to obtain the zenith total delay ZTD;

[0085] Step B: Using the Saastamoinen model using a computer The zenith dry delay ZHD is obtained; where P represents the surface pressure at the GNSS station location, represents the latitude of the GNSS station, and H represents the geodetic height of the GNSS station;

[0086] Step C, using a computer to obtain the zenith wet delay ZWD according to the formula ZWD=ZTD-ZHD;

[0087] Step D: Using a computer based on Get the weighted average temperature T m ; Where I represents the number of atmospheric layers, i is the layer number, i and I are both positive integers, and 1≤i≤I; T i represents the absolute temperature of the top of the ith atmospheric layer, e i represents the water vapor pressure at the top of the i-th atmospheric layer, Δh i represents the thickness of the i-th atmospheric layer;

[0088] Step E: Use a computer to calculate the The atmospheric water vapor conversion coefficient Π is obtained; where ρ w represents the density of liquid water, R v is the water vapor gas constant, T m represents the weighted average temperature, k2' represents the first constant of atmospheric refractive index, and k3 represents the second constant of atmospheric refractive index.

[0089] In this embodiment, the LSTM neural network model trained in step 302 has the following specific process:

[0090] Step 3021: using the GNSS water vapor tomography model, obtaining the water vapor density inversion value of the three-dimensional tomography grid at each time inside the tomography model;

[0091] Step 3022, repeating step 3021 multiple times to obtain water vapor density inversion values ​​of the three-dimensional tomographic grid at multiple times, and using the oblique path water vapor content and oblique path wet delay at multiple times and the water vapor density inversion values ​​of the three-dimensional tomographic grid corresponding to the multiple times as a training set;

[0092] Step 3023: construct an LSTM neural network model;

[0093] Step 3024: use the oblique path water vapor content and oblique path wet delay in the training set as input, and the water vapor density value of the three-dimensional tomography grid as the output layer, and input them into the LSTM neural network model for training to obtain a trained LSTM neural network model.

[0094] In this embodiment, step 4, the specific process is as follows:

[0095] Step 401: record any grid intersection point in the forecast three-dimensional grid as the target grid (lat t ,lon t ,z t ), (lat t ,lon t ,z t ) are the longitude, latitude and altitude of the target grid respectively;

[0096] Step 402: Obtain the longitude and latitude of four adjacent grid centers on the lower horizontal plane projected downward by the target grid in the three-dimensional tomographic grid, and record the longitude and latitude of the first grid center as (lat1, lon1), the longitude and latitude of the second grid center as (lat1, lon2), the longitude and latitude of the third grid center as (lat2, lon1), and the longitude and latitude of the fourth grid center as (lat2, lon2); wherein the height of the lower horizontal plane is z1;

[0097] The water vapor density of the four adjacent grid centers is recorded as the water vapor density value of the first grid center ρ 11 , the water vapor density value ρ at the center of the second grid 12 , the water vapor density value ρ at the center of the third grid 21 , the water vapor density value ρ at the center of the fourth grid 22 ;

[0098] Step 403: According to ρ(lat t ,lon t ,z1)=(1-α)(1-β)ρ 11 +α(1-β)ρ 21 +β(1-α)ρ 12 +αβρ 22 , and obtain the water vapor density value ρ (lat t ,lon t ,z1); where α represents the first interpolation weight, and β represents the second interpolation weight, and

[0099]

[0100] Step 404: According to the method of step 402 and step 403, the water vapor density value ρ (lat t ,lon t ,z2);

[0101] Step 405: According to Get the water vapor density value ρ(lat t ,lon t ,z t );

[0102] Step 405, repeating steps 401 to 404 multiple times, interpolating the water vapor density value of the three-dimensional tomography grid into the forecast three-dimensional grid, and obtaining the water vapor density value of the three-dimensional interpolation grid that is the same as the forecast three-dimensional grid.

[0103] In this embodiment, the specific process of obtaining the first error Δ1(j) in step 201 is as follows:

[0104] The precipitable water obtained by the radiosonde is taken as the true value, and the error between the precipitable water obtained by the jth GNSS station and the true value is obtained by computer and recorded as Δ1(j);

[0105] The specific process of obtaining the second error in step 601 is as follows:

[0106] The relative humidity obtained by the radiosonde is taken as the true value, and the error between each relative humidity and the true value is obtained by computer and recorded as the second error.

[0107] In this embodiment, during the specific implementation, the unit of the surface pressure P corresponding to the GNSS station position in step B is hPa, and the latitude of the GNSS station is The unit of is rad, and the unit of the GNSS station geodetic height H is km.

[0108] In this embodiment, during the specific implementation, the value of I in step D is 37, and the specific stratification refers to the 37 pressure layers of the ERA5 high-altitude ground data.

[0109] In this embodiment, during the specific implementation, the density of liquid water in step E is w , the value is 1.0×10 3 kg / m 3 , R v The value of T is 461.495 J / (kg·K), m The unit is K, and the value of k2′ is 17±10K / hPa, and the value of k3 is (3.776±0.04)×10 5 K 2 / hPa.

[0110] In this embodiment, during the specific implementation, a sigmoid activation function and a tanh activation function are set in the LSTM neural network model, the number of neurons in the input layer is 2, the number of neuron nodes in the hidden layer is 500, the number of neurons in the output layer is the same as the number of three-dimensional tomography grids, the learning rate is 0.0001 to 0.001, and the number of training iterations ranges from 100 to 200.

[0111] In summary, the method of the present invention has simple steps and reasonable design. The precipitable water inverted by GNSS is first assimilated for the first time, and the three-dimensional water vapor density and the temperature data assimilated for the first time are used to obtain the relative humidity profile, and the relative humidity profile is assimilated again based on the WRF meteorological model. The three-dimensional water vapor density data is effectively assimilated to improve the accuracy of the initial water vapor field of the WRF meteorological model and improve the forecast accuracy of meteorological parameters.

[0112] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for assimilation based on GNSS multidimensional water vapor and WRF meteorological model, characterized in that: The method comprises the following steps: Step 1: GNSS water vapor inversion: Step 101, using a computer to obtain the precipitable water volume PWV inverted by GNSS in the area to be studied according to the formula PWV=Π·ZWD; wherein ZWD represents the zenith wet delay, and Π represents the atmospheric water vapor conversion coefficient; Step 102, repeating step 101 multiple times to obtain the precipitable water inverted by multiple GNSS stations; wherein the precipitable water inverted by the j-th GNSS station is recorded as PWV(j) by a computer, and the longitude, latitude and altitude corresponding to the j-th GNSS station are obtained and recorded as lat1(j), lon1(j) and h1(j) respectively; Step 2: First assimilation of GNSS-derived precipitable water: Step 201: using a computer to convert the longitude lat1(j), latitude lon1(j), altitude h1(j), precipitable water volume PWV(j), and first error Δ1(j) corresponding to the j-th GNSS station into precipitable water volume data in Little_R format; Step 202: Using a computer, the Little_R formatted precipitable water data is input into the observation value processor OBSPROC module of the WRF meteorological model and processed into an observation value file for WRFDA; Step 203: using a computer to input the observation value file for WRFDA into the WRFDA module of the WRF meteorological model for three-dimensional variational assimilation, and outputting a meteorological data file in NetCDF format; Step 204, extracting temperature data of the forecast three-dimensional grid from the meteorological data file in NetCDF format; Step 3: GNSS 3D tomographic inversion: Step 301, gridding the study area along longitude, latitude and altitude directions to form a three-dimensional tomographic grid; and using the GNSS water vapor tomography model to obtain the water vapor density inversion value of the three-dimensional tomographic grid; Step 302: using the trained LSTM neural network model, inputting the oblique path water vapor content and the oblique path wet delay to obtain the water vapor density prediction value of the three-dimensional tomographic grid; Step 303, performing mean processing on the water vapor density inversion value of the three-dimensional tomographic grid and the water vapor density prediction value of the three-dimensional tomographic grid to obtain the water vapor density value of the three-dimensional tomographic grid; Step 4: interpolate the water vapor density value of the three-dimensional tomography grid into the forecast three-dimensional grid to obtain the water vapor density value of the three-dimensional interpolation grid that is the same as the forecast three-dimensional grid; Step 5: Invert the water vapor density value of the three-dimensional interpolation grid and the temperature data of the forecast three-dimensional grid to obtain the relative humidity profile: Step 501: Obtain the water vapor density value ρ of the i′th grid from the water vapor density value of the three-dimensional interpolation grid. 0i′ At the same time, the temperature T of the i′th grid is obtained from the temperature data of the forecast three-dimensional grid i′ and saturated water vapor pressure e si′ ; Wherein, i′ is a positive integer; Step 502: Using a computer according to the formula Get the water vapor pressure of the i′th grid; where R v is the water vapor gas constant; M w represents the molar mass of water vapor; Step 503: According to Get the relative humidity RH of the i′th grid i′ ; Step 504, repeating step 503 multiple times to obtain a relative humidity profile; wherein the relative humidity profile includes the longitude, latitude, altitude of each grid and the relative humidity corresponding to each grid; Step 6: Re-assimilation of relative humidity profile: Step 601: convert the longitude, latitude, altitude of each grid, the relative humidity corresponding to each grid, and the second error corresponding to each relative humidity into relative humidity data in Little_R format; Step 602: According to the method of step 202 and step 203, the relative humidity data in Little_R format is input into the WRF meteorological model for three-dimensional variational assimilation again, and a meteorological data file in NetCDF format is output.

2. The assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model according to claim 1, characterized in that: The specific process of obtaining the zenith wet delay ZWD and the atmospheric water vapor conversion coefficient Π in step 101 is as follows: Step A, using a computer to solve the GNSS station data using GAMIT / GLOBK software to obtain the zenith total delay ZTD; Step B: Using the Saastamoinen model using a computer The zenith dry delay ZHD is obtained; where P represents the surface pressure at the GNSS station location, represents the latitude of the GNSS station, and H represents the geodetic height of the GNSS station; Step C, using a computer to obtain the zenith wet delay ZWD according to the formula ZWD=ZTD-ZHD; Step D: Using a computer based on Get the weighted average temperature T m ; Where I represents the number of atmospheric layers, i is the layer number, i and I are both positive integers, and 1≤i≤I; T i represents the absolute temperature of the top of the ith atmospheric layer, e i represents the water vapor pressure at the top of the i-th atmospheric layer, Δh i represents the thickness of the i-th atmospheric layer; Step E: Use a computer to calculate the The atmospheric water vapor conversion coefficient Π is obtained; where ρ w represents the density of liquid water, R v is the water vapor gas constant, T m represents the weighted average temperature, k2' represents the first constant of atmospheric refractive index, and k3 represents the second constant of atmospheric refractive index.

3. The assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model according to claim 1, characterized in that: The specific process of the LSTM neural network model trained in step 302 is as follows: Step 3021: using the GNSS water vapor tomography model, obtaining the water vapor density inversion value of the three-dimensional tomography grid at each time inside the tomography model; Step 3022, repeating step 3021 multiple times to obtain water vapor density inversion values ​​of the three-dimensional tomographic grid at multiple times, and using the oblique path water vapor content and oblique path wet delay at multiple times and the water vapor density inversion values ​​of the three-dimensional tomographic grid corresponding to the multiple times as a training set; Step 3023: construct an LSTM neural network model; Step 3024: use the oblique path water vapor content and oblique path wet delay in the training set as input, and the water vapor density value of the three-dimensional tomography grid as the output layer, and input them into the LSTM neural network model for training to obtain a trained LSTM neural network model.

4. The assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model according to claim 1, characterized in that: Step 4: The specific process is as follows: Step 401: record any grid intersection point in the forecast three-dimensional grid as the target grid (lat t ,lon t ,z t ), (lat t ,lon t ,z t ) are the longitude, latitude and altitude of the target grid respectively; Step 402: Obtain the longitude and latitude of four adjacent grid centers on the lower horizontal plane projected downward by the target grid in the three-dimensional tomographic grid, and record the longitude and latitude of the first grid center as (lat1, lon1), the longitude and latitude of the second grid center as (lat1, lon2), the longitude and latitude of the third grid center as (lat2, lon1), and the longitude and latitude of the fourth grid center as (lat2, lon2); wherein the height of the lower horizontal plane is z1; The water vapor density of the four adjacent grid centers is recorded as the water vapor density value of the first grid center ρ 11 , the water vapor density value ρ at the center of the second grid 12 , the water vapor density value ρ at the center of the third grid 21 , the water vapor density value ρ at the center of the fourth grid 22 ; Step 403: According to ρ(lat t ,lon t ,z1)=(1-α)(1-β)ρ 11 +α(1-β)ρ 21 +β(1-α)ρ 12 +αβρ 22 , and obtain the water vapor density value ρ (lat t ,lon t ,z1); where α represents the first interpolation weight, and β represents the second interpolation weight, and Step 404: According to the method of step 402 and step 403, the water vapor density value ρ (lat t ,lon t ,z2); Step 405: According to Get the water vapor density value ρ(lat t ,lon t ,z t ); Step 405, repeating steps 401 to 404 multiple times, interpolating the water vapor density value of the three-dimensional tomography grid into the forecast three-dimensional grid, and obtaining the water vapor density value of the three-dimensional interpolation grid that is the same as the forecast three-dimensional grid.

5. The assimilation method based on GNSS multi-dimensional water vapor and WRF meteorological model according to claim 1, characterized in that: The specific process of obtaining the first error Δ1(j) in step 201 is as follows: The precipitable water obtained by the radiosonde is taken as the true value, and the error between the precipitable water obtained by the jth GNSS station and the true value is obtained by computer and recorded as Δ1(j); The specific process of obtaining the second error in step 601 is as follows: The relative humidity obtained by the radiosonde is taken as the true value, and the error between each relative humidity and the true value is obtained by computer and recorded as the second error.

Images