Approach rainfall forecasting method based on three-dimensional variation assimilation

Abstract

The invention discloses an adjacent rainfall forecasting method based on three-dimensional variation assimilation, and relates to the technical field of rainfall forecasting. Substituting the historical meteorological data into a preset model to obtain a first background error covariance matrix; determining a target water vapor value set according to the radar image set, and obtaining a correction error covariance matrix according to the water vapor pressure data set and the first background error covariance matrix; generating a second background error covariance matrix through a numerical weather forecast model, and fusing the correction error covariance matrix and the second background error covariance matrix through adaptive fusion to obtain a target error covariance matrix; and determining the rainfall condition according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix. The static background error covariance matrix based on historical data is adjusted through real-time observation data, and then the correction error covariance matrix and the second background error covariance matrix are fused, so that the accuracy and reliability of the prediction in the near rainfall prediction are improved more accurately.

Description

Technical Field

[0001] The invention belongs to the technical field of rainfall forecasting, and in particular relates to a rainfall forecasting method based on three-dimensional variational assimilation. Background Art

[0002] The accuracy of water vapor distribution in the troposphere of current weather forecast models still faces certain challenges, especially in the near-term rainfall forecast. Since water vapor is one of the key factors that determine the formation of precipitation, its temporal and spatial distribution is crucial to the accurate forecast of precipitation intensity and location. Therefore, conducting high-precision water vapor monitoring and effectively assimilating it into numerical weather forecast models has become an important way to improve forecast performance.

[0003] In numerical weather forecasting, the background error covariance matrix is ​​one of the indispensable key elements in the data assimilation process. It not only describes the error size in the forecast background field, but also reflects the intercorrelation of errors between different variables. It has a direct impact on how the observed data is assimilated into the background field. In practical applications, the background error covariance matrix is ​​used to adjust the difference between the observed data and the model background field, so that the assimilated analysis field can better represent the true state of the atmosphere.

[0004] At present, the construction of the background error covariance matrix is ​​usually based on the statistical analysis of historical meteorological data or the model perturbation method. For example, in the three-dimensional variational assimilation system, the construction of the background error covariance matrix relies on the static statistical estimation of historical data in advance. These methods perform well under stable weather systems, but because they are based on the average statistical characteristics of past data, the static background error covariance matrix may be difficult to fully capture the characteristics of real-time changes, resulting in a decrease in forecast accuracy. Summary of the invention

[0005] The purpose of the present invention is to solve the problem that the static background error covariance matrix may be difficult to fully capture the characteristics of real-time changes, thereby causing a decrease in forecast accuracy, and to propose a rainfall forecasting method based on three-dimensional variational assimilation.

[0006] The present invention proposes a method for nowcasting rainfall based on three-dimensional variational assimilation, the method comprising:

[0007] Acquire historical meteorological data of a target area, and substitute the historical meteorological data into a preset model to obtain a first background error covariance matrix;

[0008] Acquiring radar images and water vapor pressure data in the target area in real time to obtain a radar image set and a water vapor pressure data set, determining a target water vapor value set according to the radar image set, and obtaining a correction error covariance matrix according to the water vapor pressure data set and the first background error covariance matrix;

[0009] Generate a second background error covariance matrix through a numerical weather forecast model, and fuse the correction error covariance matrix and the second background error covariance matrix through adaptive fusion to obtain a target error covariance matrix;

[0010] Obtaining an initial cost function according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix, and minimizing the initial cost function to obtain an analysis field;

[0011] Substituting the analysis field into the numerical weather forecast model to obtain a target weather forecast model, and predicting the rainfall conditions in the target area according to the target weather forecast model.

[0012] Optionally, obtaining historical meteorological data of the target area, and substituting the historical meteorological data into a preset model to obtain a first background error covariance matrix includes:

[0013] Acquire meteorological data recorded in the target area in the past three years to obtain first meteorological data, and pre-process the first meteorological data to obtain second meteorological data;

[0014] Substituting the second meteorological data into the recurrent neural network model obtains a first background error covariance matrix.

[0015] Optionally, determining a target water vapor value set according to the radar image set includes:

[0016] For each radar image in the radar image set, correcting the radar image by using a digital elevation model to obtain a first radar image;

[0017] For each first radar image, a first radar image set is obtained by sorting them in time series, and for any adjacent first radar image in the first radar image set, a phase change analysis is performed to obtain a target phase change;

[0018] The environmental parameters in the target area are acquired, and a target water vapor value set is obtained according to the environmental parameters and the target phase change; the environmental parameters include water vapor partial pressure, temperature, and air pressure.

[0019] Optionally, obtaining a target water vapor value set according to the environmental parameter and the target phase change includes:

[0020] By formula Obtaining target water vapor value;

[0021] Wherein, e is the water vapor partial pressure, R v is the water vapor gas constant, T is the temperature, Δφ is the target phase change, λ is the radar wavelength, π is a constant, P d is the air pressure, P 0is the standard atmospheric pressure, W is the target water vapor value;

[0022] All target water vapor values ​​are acquired to obtain the target water vapor value set.

[0023] Optionally, obtaining a correction error covariance matrix according to the water vapor pressure data set and the first background error covariance matrix includes:

[0024] Generate an observation matrix according to the parameter type in the water vapor pressure data set, and obtain an observation residual value according to the observation matrix and the water vapor pressure data set;

[0025] The Kalman gain is calculated according to the first background error covariance matrix and the observation residual value;

[0026] The first background error covariance matrix is ​​updated according to the Kalman gain to obtain a correction error covariance matrix.

[0027] Optionally, fusing the correction error covariance matrix and the second background error covariance matrix by adaptive fusion to obtain a target error covariance matrix includes:

[0028] Obtained by the formula Get the target weight;

[0029] The target error covariance matrix is ​​obtained by formula C=α*A+(1-α)B;

[0030] Among them, A is the correction error covariance matrix, B is the second background error covariance matrix, α is the target weight, and C is the target error covariance matrix.

[0031] Optionally, obtaining an initial cost function according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix includes:

[0032] For the water vapor pressure data set and the target water vapor value set, calculating the covariance between the water vapor pressure data set and the target water vapor value set after time alignment to obtain a covariance matrix;

[0033] By formula

[0034] Get the initial cost function;

[0035] Where J(x) is the initial cost function, x is the analysis field, and x 0 is the target water vapor value set, is the inverse matrix of the target error covariance matrix, z is the water vapor pressure data set, H is the impact factor, R -1 is the inverse matrix of the covariance matrix.

[0036] Beneficial effects of the present invention:

[0037] The present invention proposes a nowcasting rainfall forecasting method based on three-dimensional variational assimilation. The method comprises the following steps: acquiring historical meteorological data of a target area, substituting the historical meteorological data into a preset model to obtain a first background error covariance matrix; collecting radar images and water vapor pressure data in the target area in real time to obtain a radar image set and a water vapor pressure data set, determining a target water vapor value set according to the radar image set, and obtaining a correction error covariance matrix according to the water vapor pressure data set and the first background error covariance matrix; generating a second background error covariance matrix through a numerical weather forecasting model, and fusing the correction error covariance matrix and the second background error covariance matrix through adaptive fusion to obtain a target error covariance matrix; obtaining an initial cost function according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix, and minimizing the initial cost function to obtain an analysis field; substituting the analysis field into the numerical weather forecasting model to obtain a target weather prediction model, and predicting rainfall conditions in the target area according to the target weather prediction model. By collecting radar images and water vapor pressure data sets of the target area in real time, more accurate water vapor distribution information related to the current atmospheric state can be obtained. These real-time data can better capture the temporal and spatial changes of water vapor in the atmosphere, and determine the target water vapor value set based on the radar image set, thereby providing dynamic data support for subsequent steps. By introducing real-time observation data to adjust the static background error covariance matrix based on historical data, it can better reflect the current atmospheric conditions, especially the changes in water vapor distribution. Then, through the adaptive fusion algorithm, the correction error covariance matrix and the second background error covariance matrix are fused. The target error covariance matrix finally generated not only contains the statistical characteristics of historical meteorological data, but also fully combines real-time water vapor observation information, so that it can respond to complex weather conditions more accurately, especially in the near-term rainfall forecast, which improves the accuracy and reliability of the forecast. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention will be further described below in conjunction with the accompanying drawings.

[0039] Figure 1 A flowchart of a method for nowcasting rainfall based on three-dimensional variational assimilation is provided for an embodiment of the present invention;

[0040] Figure 2 A flowchart of another method for nowcasting rainfall based on three-dimensional variational assimilation is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the present invention, the description of "first", "second", etc. is only used for descriptive purposes, and cannot be understood as indicating or implying its relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0042] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0043] The embodiment of the present invention provides a method for nowcasting rainfall based on three-dimensional variational assimilation. Figure 1 , Figure 1 A flowchart of a method for nowcasting rainfall based on three-dimensional variational assimilation is provided in an embodiment of the present invention. The method comprises the following steps:

[0044] S101, obtaining historical meteorological data of a target area, substituting the historical meteorological data into a preset model to obtain a first background error covariance matrix;

[0045] S102, collecting radar images and water vapor pressure data in the target area in real time to obtain a radar image set and a water vapor pressure data set, determining a target water vapor value set according to the radar image set, and obtaining a correction error covariance matrix according to the water vapor pressure data set and a first background error covariance matrix;

[0046] S103, generating a second background error covariance matrix through a numerical weather forecast model, and fusing the correction error covariance matrix and the second background error covariance matrix through adaptive fusion to obtain a target error covariance matrix;

[0047] S104, obtaining an initial cost function according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix, and minimizing the initial cost function to obtain an analysis field;

[0048] S105, substituting the analysis field into the numerical weather forecast model to obtain a target weather forecast model, and predicting the rainfall in the target area according to the target weather forecast model.

[0049] A nowcasting rainfall forecasting method based on three-dimensional variational assimilation provided by an embodiment of the present invention obtains more accurate water vapor distribution information related to the current atmospheric state by real-time acquisition of radar images and water vapor pressure data sets of the target area. These real-time data can better capture the spatiotemporal changes of water vapor in the atmosphere, and determine the target water vapor value set according to the radar image set, thereby providing dynamic data support for subsequent steps. The static background error covariance matrix based on historical data is adjusted by introducing real-time observation data to better reflect the current atmospheric conditions, especially the changes in water vapor distribution. The correction error covariance matrix and the second background error covariance matrix are fused through an adaptive fusion algorithm. The target error covariance matrix finally generated not only includes the statistical characteristics of the historical meteorological data, but also fully combines the real-time water vapor observation information, so that it can more accurately cope with complex weather conditions, especially in the nowcasting rainfall forecast. Improve the accuracy and reliability of the forecast.

[0050] In one implementation, historical meteorological data of the target area is collected and substituted into a preset model using the historical meteorological data to generate a first background error covariance matrix, thereby ensuring a basic covariance matrix based on long-term statistical characteristics, thereby describing the error characteristics of the target area under general weather conditions and forming a static background field to prepare for the follow-up.

[0051] In one implementation, by collecting radar images and water vapor pressure data sets of the target area in real time, more accurate water vapor distribution information related to the current atmospheric state can be obtained. These real-time data can better capture the temporal and spatial changes of water vapor in the atmosphere and determine the target water vapor value set based on the radar image set, thereby providing dynamic data support for subsequent steps.

[0052] In one implementation, the second background error covariance matrix is ​​generated by the numerical weather forecast model by selecting WRF as the weather forecast model, and inputting the first environmental data in the target area into the WRF model to obtain the second background error covariance matrix; the first environmental data can be temperature, humidity, wind speed, wind direction, air pressure, etc.

[0053] In one implementation, a correction error covariance matrix is ​​generated based on real-time water vapor pressure data and the first background error covariance matrix. Thus, real-time observation data can be used to adjust the static background error covariance matrix based on historical data, thereby better reflecting current atmospheric conditions, especially changes in water vapor distribution, and enhancing the system's adaptability to rapidly changing weather systems.

[0054] In one implementation, historical meteorological data may be temperature, humidity, wind speed, wind direction, air pressure, precipitation, etc.; the target area is the area for which weather forecast is to be conducted; the radar image is an image taken by an interferometric synthetic aperture radar; and the water vapor pressure data is an average value measured by ground meteorological stations, satellite remote sensing, radiosondes, etc.

[0055] In one implementation, a second background error covariance matrix is ​​generated through a numerical weather forecast model. It is based on the simulation field of the current numerical weather forecast and can capture the error characteristics in the model background field. Then, through an adaptive fusion algorithm, the correction error covariance matrix and the second background error covariance matrix are fused to obtain a target error covariance matrix. The fused matrix has both historical statistical characteristics and real-time observation information, and can be dynamically adjusted with atmospheric changes, thereby improving the accuracy of the description of complex weather systems.

[0056] In one implementation, the target water vapor value set, water vapor pressure data set, and target error covariance matrix are used to construct an initial cost function, and the analysis field is obtained by minimizing the function. This step ensures the accuracy of the analysis field in the forecast system, optimizes the assimilation process, and enables the model to better assimilate real-time observation data, thereby improving the accuracy of weather forecasts.

[0057] In one embodiment, historical meteorological data of the target area is obtained, and the historical meteorological data is substituted into a preset model to obtain a first background error covariance matrix, including:

[0058] Acquire the meteorological data recorded in the past three years in the target area to obtain first meteorological data, and pre-process the first meteorological data to obtain second meteorological data;

[0059] Substitute the second meteorological data into the recurrent neural network model to obtain the first background error covariance matrix.

[0060] In one implementation, the recurrent neural network can well capture the potential temporal relationships and patterns in historical meteorological data. By analyzing the meteorological data of the past three years, the recurrent neural network can learn the characteristics of meteorological variables changing over time, and use this information to generate a more reasonable background error covariance matrix. Compared with the traditional covariance matrix based on statistical average, it can better reflect the time dependence and dynamic change characteristics of the meteorological system, thereby improving the accuracy of the system.

[0061] In one implementation, preprocessing includes removing noise, standardizing data, etc. First, the collected historical data is preprocessed, and then the processed data is divided into three parts, of which the training set accounts for 70%, the validation set accounts for 20%, and the test set accounts for 10%. A recurrent neural network is selected as the learning model and the mean square error is used as the loss function. After the data is input into the model, the difference between the predicted value and the true value, that is, the current error of the model, is calculated by the loss function. The gradient of the error to the model parameters is calculated by the back propagation algorithm, and the model parameters are adjusted according to these gradients to reduce the prediction error. The Adam optimizer is used as the optimization algorithm, and the model will continue for multiple iterations until the error of the test set test result is less than the preset value, where the preset value is determined by the technician.

[0062] In one embodiment, determining a target water vapor value set according to a radar image set includes:

[0063] For each radar image in the radar image set, correcting the radar image by using a digital elevation model to obtain a first radar image;

[0064] For each first radar image, a first radar image set is obtained by sorting them in time series, and for any adjacent first radar image in the first radar image set, a phase change analysis is performed to obtain a target phase change;

[0065] The environmental parameters in the target area are obtained, and the target water vapor value set is obtained according to the environmental parameters and the target phase change; the environmental parameters include water vapor partial pressure, temperature, and air pressure.

[0066] In one implementation, terrain undulations can affect the propagation path of radar waves, causing errors in radar images. By using a digital elevation model for correction, it can be ensured that the radar image more accurately reflects the actual meteorological conditions of the target area, especially the water vapor distribution, thereby improving the accuracy of water vapor monitoring.

[0067] In one implementation, the water vapor content in the atmosphere is constantly changing. By analyzing radar images at different time points in chronological order, the spatiotemporal evolution of water vapor can be better reflected. The phase change analysis of adjacent radar images can reveal the changing trend of water vapor, providing a more dynamic and accurate basis for determining the target water vapor value set.

[0068] In one implementation, phase change analysis can accurately invert the change of water vapor in the atmosphere, especially when there is water vapor difference in the radar wave path, making the estimation of water vapor content more refined and helping to accurately obtain the water vapor value set.

[0069] In one embodiment, obtaining a target water vapor value set according to environmental parameters and target phase changes includes:

[0070] By formula Obtaining target water vapor value;

[0071] Where e is the partial pressure of water vapor, R v is the water vapor gas constant, T is the temperature, Δφ is the target phase change, λ is the radar wavelength, π is a constant, P d is the air pressure, P 0 is the standard atmospheric pressure, W is the target water vapor value;

[0072] Get all target water vapor values ​​to obtain a target water vapor value set.

[0073] In one implementation, R v is the water vapor gas constant, which is 461.5 J / (kg·K), λ is the radar wavelength set when taking the radar image, which is determined by the technicians, π is a constant of 3.1415, P 0 The standard atmospheric pressure here is 101324Pa.

[0074] In one implementation, the formula not only considers radar phase change data, but also integrates meteorological environmental parameters. The comprehensive use of multi-source data helps to overcome the limitations of a single data source and improves the comprehensiveness and reliability of water vapor estimation by effectively integrating multiple observation information.

[0075] In one embodiment, see Figure 2 , step S102 further includes:

[0076] S1021, generating an observation matrix according to the parameter type in the water vapor pressure data set, and obtaining an observation residual value according to the observation matrix and the water vapor pressure data set;

[0077] S1022, calculating a Kalman gain according to the first background error covariance matrix and the observed residual value;

[0078] S1023, updating the first background error covariance matrix according to the Kalman gain to obtain a correction error covariance matrix.

[0079] In one implementation, by using the observation data in the water vapor pressure dataset and combining it with the background error covariance matrix, the difference between the observation and the model prediction, namely the observation residual value, can be effectively captured. By calculating the Kalman gain, the model output can be reasonably adjusted to improve the accuracy of data assimilation, thereby ensuring that the update of the background field is more in line with the current actual atmospheric state.

[0080] In one implementation, for the water vapor pressure data set, the dimension of the observation matrix H is m×n, where m is the number of observations and n is the number of state variables. Assuming that the corresponding observation data extracted from the water vapor pressure data set include observations of water vapor partial pressure, temperature, and air pressure, and there are multiple observation time points, the observation matrix is ​​generated based on these three observation parameters: Where e is the water vapor partial pressure, T is the temperature, p is the observed value of the air pressure, and m is the number of observation time points. The sum of the observation matrix and the first background error covariance matrix is ​​calculated to obtain the target matrix, and the observation residual value is obtained by subtracting the target matrix from the water vapor pressure data set.

[0081] In one implementation, the Kalman gain is calculated based on the first background error covariance matrix and the observed residual value, and the correction error covariance matrix is ​​obtained by updating the first background error covariance matrix based on the Kalman gain:

[0082] By the formula K = p 0 H T (Hp 0 H T +R) -1 Get the Kalman gain;

[0083] By the formula A = (I-KH)p 0 Get the corrected error covariance matrix;

[0084] Where K is the Kalman gain, p 0 is the first background error covariance matrix, H is the observation matrix, R is the observation residual value, H T is the transpose of the observation matrix, A is the correction error covariance matrix, and I is the identity matrix.

[0085] In one implementation, the Kalman filter framework is used to update the background error covariance matrix so that the background error estimate can respond to changes in observation data in real time, thereby more flexibly adapting to the real-time characteristics of the atmospheric system and improving the performance of the model under complex meteorological conditions.

[0086] In one embodiment, the target error covariance matrix is ​​obtained by fusing the correction error covariance matrix and the second background error covariance matrix through adaptive fusion, including:

[0087] Obtained by the formula Get the target weight;

[0088] The target error covariance matrix is ​​obtained by the formula C=α*A+(1-α)B;

[0089] Among them, A is the correction error covariance matrix, B is the second background error covariance matrix, α is the target weight, and C is the target error covariance matrix.

[0090] In one implementation, the target error covariance matrix is ​​generated by adaptively fusing the correction error covariance matrix and the second background error covariance matrix, which not only improves the accuracy and flexibility of the model, but also effectively manages uncertainty and optimizes the effect of data assimilation, making the weather forecast model more adaptable and with higher forecast accuracy under complex meteorological conditions.

[0091] In one embodiment, obtaining an initial cost function according to a water vapor pressure data set, a target water vapor value set, and a target error covariance matrix includes:

[0092] For the water vapor pressure data set and the target water vapor value set, the covariance between the water vapor pressure data set and the target water vapor value set is calculated after time alignment to obtain a covariance matrix;

[0093] By formula

[0094] Get the initial cost function;

[0095] Where J(x) is the initial cost function, x is the analysis field, and x 0 is the target water vapor value set, is the inverse matrix of the target error covariance matrix, z is the water vapor pressure data set, H is the impact factor, R -1 is the inverse of the covariance matrix.

[0096] In one implementation, by combining the background field error term and the observation error term, an improved weather forecast result is obtained, making the weather forecast more accurate.

[0097] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for nowcasting rainfall based on three-dimensional variational assimilation, characterized in that: The method comprises: Acquire historical meteorological data of a target area, and substitute the historical meteorological data into a preset model to obtain a first background error covariance matrix; Acquiring radar images and water vapor pressure data in the target area in real time to obtain a radar image set and a water vapor pressure data set, determining a target water vapor value set according to the radar image set, and obtaining a correction error covariance matrix according to the water vapor pressure data set and the first background error covariance matrix; Generate a second background error covariance matrix through a numerical weather forecast model, and fuse the correction error covariance matrix and the second background error covariance matrix through adaptive fusion to obtain a target error covariance matrix; Obtaining an initial cost function according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix, and minimizing the initial cost function to obtain an analysis field; Substituting the analysis field into the numerical weather forecast model to obtain a target weather forecast model, and predicting the rainfall conditions in the target area according to the target weather forecast model.

2. The method for nowcasting rainfall based on three-dimensional variational assimilation according to claim 1, characterized in that: Acquiring historical meteorological data of the target area, and substituting the historical meteorological data into a preset model to obtain a first background error covariance matrix includes: Acquire meteorological data recorded in the target area in the past three years to obtain first meteorological data, and pre-process the first meteorological data to obtain second meteorological data; Substituting the second meteorological data into the recurrent neural network model obtains a first background error covariance matrix.

3. The method for nowcasting rainfall based on three-dimensional variational assimilation according to claim 1, characterized in that: Determining a target water vapor value set according to the radar image set includes: For each radar image in the radar image set, correcting the radar image by using a digital elevation model to obtain a first radar image; For each first radar image, a first radar image set is obtained by sorting them in time series, and for any adjacent first radar image in the first radar image set, a phase change analysis is performed to obtain a target phase change; The environmental parameters in the target area are acquired, and a target water vapor value set is obtained according to the environmental parameters and the target phase change; the environmental parameters include water vapor partial pressure, temperature, and air pressure.

4. The method for nowcasting rainfall based on three-dimensional variational assimilation according to claim 3 is characterized in that: Obtaining a target water vapor value set according to the environmental parameter and the target phase change includes: By formula Obtaining target water vapor value; Wherein, e is the water vapor partial pressure, R v is the water vapor gas constant, T is the temperature, Δφ is the target phase change, λ is the radar wavelength, π is a constant, P d is the air pressure, P0 is the standard atmospheric pressure, and W is the target water vapor value; All target water vapor values ​​are acquired to obtain the target water vapor value set.

5. The method for nowcasting rainfall based on three-dimensional variational assimilation according to claim 1, characterized in that: Obtaining a correction error covariance matrix according to the water vapor pressure data set and the first background error covariance matrix includes: Generate an observation matrix according to the parameter type in the water vapor pressure data set, and obtain an observation residual value according to the observation matrix and the water vapor pressure data set; The Kalman gain is calculated according to the first background error covariance matrix and the observation residual value; The first background error covariance matrix is ​​updated according to the Kalman gain to obtain a correction error covariance matrix.

6. The method for nowcasting rainfall based on three-dimensional variational assimilation according to claim 1, characterized in that: The target error covariance matrix is ​​obtained by fusing the correction error covariance matrix and the second background error covariance matrix through adaptive fusion, including: Obtained by the formula Get the target weight; The target error covariance matrix is ​​obtained by formula C=α*A+(1-α)B; Among them, A is the correction error covariance matrix, B is the second background error covariance matrix, α is the target weight, and C is the target error covariance matrix.

7. The method for nowcasting rainfall based on three-dimensional variational assimilation according to claim 1, characterized in that: Obtaining an initial cost function according to the water vapor pressure data set, the target water vapor value set and the target error covariance matrix includes: For the water vapor pressure data set and the target water vapor value set, calculating the covariance between the water vapor pressure data set and the target water vapor value set after time alignment to obtain a covariance matrix; By formula Get the initial cost function; Wherein, J(x) is the initial cost function, x is the analysis field, x0 is the target water vapor value set, is the inverse matrix of the target error covariance matrix, z is the water vapor pressure data set, H is the impact factor, R -1 is the inverse matrix of the covariance matrix.

Images