Cloud analysis and random perturbation combined method for regional convective scale ensemble forecast

Abstract

The invention discloses a cloud analysis and random perturbation combined method for regional convective scale ensemble forecast in the technical field of meteorological data processing, and the method comprises the steps: obtaining cloud analysis data, such as blackbody brightness temperature, total cloud amount and radar three-dimensional networking reflectivity, of a monitoring region and other data needed by analysis assimilation based on a regional WRF mode; aiming at the uncertainty source of a cloud analysis system, the method comprises the following steps of: randomly disturbing multi-source observation data input in the process of forming an initial field of each ensemble forecasting member cloud in ensemble forecasting; aiming at the uncertainty of a micro-physical parameterization scheme of the cloud analysis system, three-dimensional disturbance is carried out on key dynamics and thermodynamics sensitive parameters of cloud analysis, combined random disturbance of two processes of an initial value and the parameterization scheme is formed, and the initial value and cloud micro-physical parameterization uncertainty of the cloud analysis system and the interaction of the initial value and the cloud micro-physical parameterization uncertainty are further reflected. Convection triggering and rainfall evolution processes can be changed, and the cloud and rainfall forecasting accuracy of a mesoscale service ensemble forecasting system is improved.

Description

Technical Field

[0001] This invention relates to the field of meteorological data processing technology, specifically a method for random joint perturbation of cloud analysis in regional convective-scale ensemble forecasting. Background Technology

[0002] Severe convective weather is one of the major hazardous weather events. Meteorologically, severe convective weather refers to sudden, rapidly moving, violent, and extremely destructive convective weather events, including thunderstorms, hail, tornadoes, localized heavy rainfall, and squall lines. Due to its small spatial scale, short lifespan, sudden onset, and rapid development, severe convective weather presents a significant challenge for operational weather forecasting. However, its destructive power makes effective forecasting crucial for disaster prevention and mitigation, major social events, and refined weather services. Since its introduction into meteorology, ensemble forecasting has moved beyond the traditional numerical weather prediction approach, which treats forecasting as a deterministic initial value problem. It recognizes that the forecasting skill of atmospheric models depends not only on initial conditions, the forecast model itself, and its alignment with actual atmospheric conditions, but also on the stability of the atmospheric circulation itself. Ensemble forecasts, by incorporating existing achievements in traditional deterministic weather forecasting and taking into account the uncertainties of initial values ​​in numerical forecast models, forecast model errors, and the chaotic characteristics of atmospheric motion development, can not only provide higher forecasting skill than single deterministic forecasts, but also provide probabilistic forecasts of sudden and severe weather events. In practical operational forecasting, they have become an important component of weather forecasting systems.

[0003] Existing mainstream regional ensemble forecasting systems typically have a resolution of 2-10 km, focusing primarily on short-term weather forecasts of 1-3 days. They can predict some local details of the development process of weather systems and assess the predictability of severe weather events. Initial perturbation methods are a core component of constructing regional ensemble forecasting systems. Currently, there are three main methods for constructing initial perturbation fields for regional ensemble forecasts. The first is the dynamic downscaling method (Frogner et al., 2006; Li et al., 2008; Bowler and Mylne, 2009), which downscales global ensemble forecast initial perturbations to regional ensemble initial perturbations. This method is simple and easy to implement, but the resulting initial perturbations are difficult to accurately describe the small- and medium-scale information of atmospheric motion. The second method is to use the pattern itself to generate the initial perturbation value, such as the breeding method (BGM), the ensemble transform Kalman filter (ETKF), and the ensemble transform (ET) (Bishop et al., 2001, 2009; Bowler and Mylne, 2009; Du et al., 2003; Majumdar et al., 2002; Ma Xulin et al., 2006, 2008, 2021; Zhang Hanbin et al., 2014a, 2014b, 2017; Liu Kan et al., 2023).

[0004] While model-based perturbation methods can describe the uncertainties in the development of small- and medium-scale weather systems, the matching problem of lateral boundary perturbation scales can easily induce false gravity waves, affecting forecast performance (Caron, 2013). The third method combines the advantages of dynamic downscaling and model-based perturbation initial value generation methods with multi-scale blending (MSB). The blended perturbation fields generated using spectral decomposition and filtering methods have advantages in reflecting large- and small-scale motion development information (Zhang et al. 2015b; Zhang et al. 2016). However, filtering methods struggle to objectively and accurately determine large- and small-scale information, and the fusion of multi-scale perturbation information may lead to over- or under-fusion. Therefore, Wastl et al. (2021) and Ma et al. (2018) further developed a new scheme for blended-scale initial perturbation methods based on data assimilation. Numerical weather forecasting, as the core technology of weather forecasting, requires rapid response to precipitation and other weather events, and also requires it to have nowcasting capabilities. Generally speaking, nowcasting refers to weather forecasts for 0-2 hours, while the World Meteorological Organization (WMO) defines nowcasting as extending to 0-6 hours (Zheng Yongguang et al., 2009).

[0005] Currently, typical mesoscale models often lag behind cloud and precipitation responses by several hours, and their forecasts for high-impact weather events such as localized severe convection also frequently contain significant deviations. In most cases, at the initial moment of the model forecast, ground observation stations have already observed a considerable amount of precipitation or precipitation is in progress, and some areas have already observed or have a fairly thick cloud system above the precipitation area. However, the model needs to run for several hours for clouds to form and for rain to fall to the ground. This results in the predicted cloud system and precipitation significantly lagging behind the actual atmospheric conditions, with varying degrees of deviation between the cloud and rain area locations and the actual situation. Even the mesoscale thermal and dynamic structures may differ significantly, which has a considerable negative impact on short-term forecasts.

[0006] As the most critical technology in ensemble forecasting systems, existing initial value perturbation methods for convective-scale ensemble forecasting systems currently only consider the uncertainties of forecast variables (temperature, pressure, humidity, wind, etc.) at their initial moments, and the initial value perturbation variables are only for these forecast variables. Since most current ensemble forecasting systems do not include cloud analysis processes, perturbations of cloud and precipitation information in the initial field are almost nonexistent or minimal. Even in studies involving initial value perturbations of cloud and precipitation, only the uncertainties of observational data are considered; the uncertainties of key parameters in the cloud analysis system are not considered, or only a single parameter is considered, let alone joint perturbation methods for model initial variables (temperature, pressure, humidity, wind, etc.) and multiple key parameters of cloud analysis that address spatiotemporal uncertainties. Existing small- and medium-scale convective systems are highly sensitive to initial cloud structure; when the initial cloud fields of ensemble members are too similar, it will lead to insufficient forecast dispersion, making it difficult to capture sudden, strong convection. Summary of the Invention

[0007] The purpose of this invention is to provide a stochastic joint perturbation method for cloud analysis in regional convective-scale ensemble forecasts. Based on a regional model (e.g., WRF), it combines traditional ensemble forecast initial value perturbation methods with cloud analysis system perturbation methods. This method introduces random variable parameters to address the uncertainties in key parameters of radar, satellite, and ground observation data and cloud analysis systems. It also perturbs the range of values ​​for observed variables and parameters in microphysical parameterization schemes. This perturbation alters the uncertainties in microphysical quantities such as cloud water, rainwater, and ice crystals in the initial cloud field, as well as forecast variables such as temperature, pressure, humidity, and wind in the convective-scale ensemble forecast system. This changes the convection triggering and precipitation evolution process, improving the accuracy of mesoscale operational short-term nowcasting for cloud and precipitation forecasts. It is beneficial for enhancing disaster prevention and mitigation capabilities for severe convective weather such as heavy rain and squall lines, and addresses the problem of insufficient forecast dispersion when the initial cloud fields of ensemble members are too similar, thereby improving the ability to capture sudden severe convection.

[0008] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting, applicable to convective-scale models with a horizontal resolution of 2km to 4km, the method comprising: Based on the WRF model, the system acquires the analytical assimilation observation data, hourly average blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or the data required for analytical assimilation in the monitoring area to obtain ensemble forecast data. Among them, the data required for analytical assimilation includes aircraft report data, radiosonde data, and GPS water vapor inversion data, etc. Background extraction technology is used to extract background field information from the ensemble forecast data to obtain isobaric surface coarse grid file data. The isobaric surface coarse grid file data is then converted into the initial field and side boundary conditions on the model grid points of the model to obtain initial field data and side boundary condition field data. Based on a normal distribution random number generator and a cloud analysis system, the total cloud volume data of domestic Fengyun satellites, hourly average blackbody brightness temperature data, and radar three-dimensional network reflectivity dataset are input into the cloud analysis system. The lowest radar reflectivity of the cloud analysis system is collected, and a random perturbation of the standard deviation is performed based on the default mean to generate a random threshold for the lowest radar reflectivity that conforms to the default mean and standard deviation. Based on the error characteristics of blackbody brightness temperature products, hourly average blackbody brightness temperature data are extracted sparsely using a random sampling method to obtain the uncertain state data of hourly average blackbody brightness temperature data. The data of total cloud volume product of domestic Fengyun satellite is extracted sparsely using a random sampling method. Based on the error characteristics of the total cloud volume product, the total cloud volume data is extracted sparsely using a random sampling method. The extracted samples should reflect the spatiotemporal correlation of product error as much as possible to obtain the uncertainty data of total cloud volume. Based on different altitudes, the three-dimensional key sensitive parameter data of the microphysical parameterization scheme of the cloud analysis system are mutually converted. Using a normal distribution random perturbation generator, the following key parameters reflecting thermal and dynamic uncertainties in the cloud analysis process are randomly perturbed at different thresholds, constructing a sensitive parameter random perturbation module: the threshold for converting relative humidity to cloud cover (Rh0) + the minimum threshold for converting cloud cover to relative humidity (rh_thr1) + the maximum threshold for converting cloud cover to relative humidity (rh_thr2) + the minimum cloud cover threshold for converting cloud cover to relative humidity (cvr2rh_thr1) + the cloud cover threshold for reaching relative humidity saturation (cvr2rh_thr1). The conversion results are obtained by calculating key sensitive parameters of thermal and dynamic uncertainties during cloud analysis. Based on the hourly average blackbody brightness temperature data, total cloud data and radar three-dimensional network reflectivity dataset corresponding to the conversion results from domestic Fengyun satellites, joint random perturbation is performed, and the ensemble forecast product corresponding to the ensemble forecast data is obtained by using the control forecast integral method.

[0009] As a further aspect of the present invention: the acquisition of analytical assimilation observation data, radiosonde data, ship data, satellite cloud wind guidance and GPS water vapor inversion data, blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or data required for analysis and assimilation based on WRF mode for the monitoring area includes: Raw blackbody brightness temperature data is obtained from the Fengyun satellite data receiving station. The raw blackbody brightness temperature data is then decoded, decompressed, and time-averaged sequentially to obtain hourly average blackbody brightness temperature data. The raw data of total cloud volume is obtained from the Fengyun satellite data receiving station. The raw data of total cloud volume is then decoded and decompressed to obtain the total cloud volume data. Obtain the original weather radar mosaic data, query the legend reflectance RGB and annotation RGB of the original weather radar mosaic data, obtain the reflectance and annotation RGB mapping table, convert the annotation RGB of all grid points of the original weather radar mosaic data to reflectance, obtain the converted weather radar mosaic data, use the interpolation method to complete the missing grid points of the converted weather radar mosaic data, and obtain the radar three-dimensional network reflectance dataset. And / or collect raw data required for analysis and assimilation, convert the format of the raw data required for analysis and assimilation to obtain the data required for analysis and assimilation, and improve the quality of assimilation data through quality control schemes.

[0010] As a further aspect of the present invention: the background extraction technique is used to extract background field information from the ensemble forecast data to obtain isobaric surface coarse grid file data. This isobaric surface coarse grid file data is then converted into initial field and lateral boundary conditions at the model grid points of the model, resulting in initial field data and lateral boundary condition field data, including: The isobaric coarse mesh file data was processed sequentially using horizontal interpolation, vertical interpolation, and variable transformation to obtain the initial field and side boundary conditions; Each of the following data is assigned a corresponding initial model value set and a lateral boundary condition: observational data, radiosonde data, ship data, satellite cloud wind guidance and GPS water vapor inversion data, hourly average blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or data required for analysis and assimilation. This results in the initial model value set data and the lateral boundary condition set data. The initial model value set data is the initial field data, and the lateral boundary condition set data is the lateral boundary condition field data.

[0011] As a further aspect of the present invention: the cloud analysis system, based on a normal distribution random number generator, inputs the assimilated observation data, hourly average blackbody brightness temperature data, and radar three-dimensional network reflectivity dataset into the cloud analysis system to establish a cloud analysis observation data random perturbation module, including: When the cloud analysis system receives radar reflectivity data, it uses a normal distribution random number generator to randomly perturb the radar reflectivity around the default mean, generating the lowest random radar reflectivity result that conforms to the default mean and standard deviation, reflecting the uncertain random distribution data of radar reflectivity. When the cloud analytics system receives hourly average blackbody brightness temperature data, it uses a random sampling method to sparsify and extract the hourly average blackbody brightness temperature data based on the error characteristics of the blackbody brightness temperature product, thereby obtaining the uncertain state data of the hourly average blackbody brightness temperature data. When the cloud analysis system receives the total cloud volume data of the domestic Fengyun satellite, it uses a random sampling method to sparsely extract the total cloud volume data based on the error characteristics of the total cloud volume product. The extracted samples reflect the spatiotemporal correlation of the product error as much as possible, and obtain the uncertainty data of the total cloud volume. As a further aspect of the present invention: the key sensitive dynamic and thermodynamic parameters of the cloud analysis microphysical parameterization scheme are randomly perturbed to form a key sensitive parameter random perturbation module that reflects the uncertainty of the thermodynamic and dynamic processes of the cloud analysis system.

[0012] As a further aspect of the present invention: the conversion between relative humidity and total cloud volume data of the cloud analysis system based on different altitudes to obtain the conversion result includes: Based on the threshold of total cloud volume data corresponding to different altitudes, relative humidity with a standard deviation of 0.05 is converted into random total cloud volume data.

[0013] As a further aspect of the present invention: the conversion between relative humidity and total cloud volume data of the cloud analysis system based on different altitudes to obtain the conversion result includes: Based on a cloud analytics system, a threshold was set to convert relative humidity as it changes with altitude into cloud cover. When the altitude is less than 600m, the mean is 0.925 and the standard deviation is 0.025; when the altitude is 600m-1500m, the mean is 0.9 and the standard deviation is 0.025; when the altitude is 1500m-2500m, the mean is 0.85 and the standard deviation is 0.025; and when the altitude is greater than 2500m, the mean is 0.8 and the standard deviation is 0.075. Set a random perturbation value with a default mean of 0.5 and a standard deviation of 0.1 to generate the lowest random threshold for relative humidity corresponding to the total cloud volume data; Set a random perturbation value with a default mean of 1 and a standard deviation of 0.05 to generate the highest random threshold for relative humidity corresponding to the total cloud volume data; Set a random perturbation value with a default mean of 0.2 and a standard deviation of 0.1 to generate the minimum random threshold for the total cloud cover data corresponding to relative humidity; Set a random perturbation value with a default mean of 0.7 and a standard deviation of 0.1 to generate the corresponding total cloud volume data when the relative humidity reaches saturation.

[0014] As a further aspect of the present invention: the cloud analysis observation data and radar three-dimensional network reflectivity dataset corresponding to the transformation result are subjected to joint random perturbation, and the ensemble forecast product corresponding to the ensemble forecast data is obtained by using the control forecast integration method, including: A random generator is used to randomly perturb the external analytical assimilation data of the cloud analysis system. Based on the three-dimensional variational system and the set assimilation method, the initial values ​​of the regional set forecast perturbation are generated.

[0015] As a further aspect of this invention: A joint random perturbation is performed based on cloud analysis input data (analyzed and assimilated observation data, radar 3D network reflectivity, TBB dataset) and key sensitive parameters reflecting the uncertainty of cloud analysis microphysical dynamics and thermodynamic processes (the threshold for converting relative humidity to cloud cover (Rh0) + the minimum threshold for converting cloud cover to relative humidity (rh_thr1) + the maximum threshold for converting cloud cover to relative humidity (rh_thr2) + the minimum cloud cover threshold for converting cloud cover to relative humidity (cvr2rh_thr1) + the cloud cover threshold for reaching relative humidity saturation (cvr2rh_thr1)). This further reflects the initial values ​​of the cloud analysis system and the uncertainty of cloud microphysical parameterization, as well as their interaction. The ensemble forecast product corresponding to the ensemble forecast data is obtained using a control forecast integration method. The invention also includes: Initial field files and lateral boundary condition files for ensemble forecast data are generated based on other convective-scale ensemble forecasting methods; Using a 3km resolution WRF model and a dynamic downscaling method, the initial field file and the lateral boundary condition file are integrated to obtain the control forecast integral results; The perturbation member integrals are obtained by applying the perturbation mode in the WRF model to the ensemble forecast data, and the ensemble forecast products are obtained.

[0016] Secondly, the present invention provides a system comprising: The data acquisition module is configured to acquire, based on WRF mode, the analytical assimilation observation data, hourly average blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or the data required for analytical assimilation and cloud analysis of the monitoring area, to obtain ensemble forecast data. The information extraction module is configured to use background extraction technology to extract background field information from the ensemble forecast data to obtain isobaric surface coarse grid file data, and to convert the isobaric surface coarse grid file data into the initial field and side boundary conditions on the mode grid points of the mode to obtain initial field data and side boundary condition field data. The cloud analysis system's observation data module inputs the total cloud volume data from domestic Fengyun satellites, hourly average blackbody brightness temperature data, radar three-dimensional network reflectivity datasets, and initial field grid data after analysis and assimilation into the cloud analysis system. The cloud analysis observation data random perturbation module collects the lowest radar reflectivity of the cloud analysis system. It uses a normal distribution random number generator to randomly perturb the standard deviation based on the default mean, generating a random threshold for the lowest radar reflectivity that conforms to the default mean and standard deviation. It uses a random sampling method to sparsely extract hourly average blackbody brightness temperature data and total cloud volume data from the domestic Fengyun satellites, obtaining uncertainty data for hourly average blackbody brightness temperature data and total cloud volume. This controls the radar and satellite observation data entering the cloud analysis system, reflects the uncertainty of the observation data, and provides the ensemble forecasting system with multiple cloud analysis input observation datasets with different initial observation values. Three-dimensional random perturbation module for key sensitive parameters in cloud analysis system: Based on the default values ​​of the original parameters in the system, and considering the vertical changes of the key parameters in the cloud analysis system, a random perturbation module for sensitive parameters is constructed by randomly perturbing the following key parameters with different thresholds using a normal distribution random perturbation generator: threshold for converting relative humidity to cloud cover (Rh0) + minimum threshold for converting cloud cover to relative humidity (rh_thr1) + maximum threshold for converting cloud cover to relative humidity (rh_thr2) + minimum cloud cover threshold for converting cloud cover to relative humidity (cvr2rh_thr1) + cloud cover threshold for cloud cover to reach saturation state (cvr2rh_thr1). The cloud analysis system is subjected to joint random perturbation. Combining the above modules, the cloud analysis system observation data (S4) and key sensitive parameters (S5) are subjected to joint random perturbation at the same time, which reflects the uncertainty of the input data of the cloud analysis system and the uncertainty of the dynamic and thermodynamic processes of the cloud analysis system itself. The product acquisition module is configured to convert the relative humidity and total cloud volume data of the cloud analysis system to each other based on different altitudes, obtain the conversion results, analyze and assimilate the observation data and radar three-dimensional network reflectivity dataset corresponding to the conversion results, perform joint random perturbation, and use control forecasting and integration of different ensemble members to obtain the ensemble forecast product corresponding to the ensemble forecast data.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. In this invention, by randomly perturbing the multi-source observation data of the cloud initial field formation process of each ensemble forecast member in the ensemble forecast, as well as the key sensitive parameters in the cloud analysis microphysics process, it can simultaneously reflect the initial value of the cloud analysis system and the parameterization uncertainty of cloud microphysics, as well as the interaction between the two. This can change the convection triggering and precipitation evolution process, improve the accuracy of mesoscale operational short-term nowcasting numerical forecasts for cloud and precipitation forecasts, and help improve the disaster prevention and mitigation capabilities for severe convective weather such as rainstorms and squall lines.

[0018] 2. In this invention, by taking into account the uncertainties of cloud microphysical information and its equilibrium thermodynamic information in the initial field of the WRF model, and combining the ensemble forecasting method to perform three-dimensional random joint perturbation on the cloud analysis input observation data and key sensitivity parameters of the cloud analysis system during the initial field formation process of aircraft forecast data, ground forecast data, radiosonde data, ship data, satellite cloud wind guidance and GPS water vapor inversion data, hourly average blackbody brightness temperature data, total cloud volume data, radar three-dimensional network reflectivity dataset and / or data required for ensemble assimilation, it is possible to dynamically adjust the distribution of the three-dimensional condensate and humidity fields in the initial field of the model, reflect the uncertainties of microphysical processes such as automatic conversion (cloud water → rainwater), collision and coalescence, sublimation / evaporation, ice nucleation, etc., and better coordinate with the model analysis system, reducing initialization shock (SPIN-UP).

[0019] 3. In this invention, by performing joint random perturbation based on the blackbody brightness temperature data, total cloud volume data, and radar three-dimensional network reflectivity dataset corresponding to the conversion results, the systematic uncertainty characterization of the initial cloud field at the convective scale can be achieved, which can effectively enhance the dispersion of convective forecasts. By inputting the radar three-dimensional network reflectivity dataset into the cloud analysis system and performing random perturbation on the cloud analysis system, the uncertainty of the location of heavy precipitation can be reduced, thereby improving the accuracy of convective forecasts. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method steps of the present invention; Figure 2 This is a system module structure diagram of the present invention.

[0021] In the diagram: 1. Data acquisition module; 2. Information extraction module; 3. Cloud analysis system observation data module; 4. Cloud analysis observation data random perturbation module; 5. Cloud analysis system key sensitive parameter three-dimensional random perturbation module; 6. Cloud analysis system joint random perturbation; 7. Product acquisition module. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example: Please see Figure 1 In this embodiment of the invention, a cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting is applied to a convective-scale model with a horizontal resolution of 3 km. The method includes the following steps: S1: Based on the WRF model, acquire the analysis and assimilation data, hourly average blackbody brightness temperature data, total cloud volume data, radar three-dimensional network reflectivity dataset, and other data required for analysis and assimilation and cloud analysis of the monitoring area, and obtain the cloud analysis data required for ensemble forecasting. Among them, the data required for analysis and assimilation include aircraft report data, ground report data, radiosonde data, ship data, satellite cloud wind guide and GPS water vapor inversion data, etc. S2: Background extraction technology is used to extract background field information from the ensemble forecast data to obtain isobaric surface coarse grid file data. The isobaric surface coarse grid file data is then converted into the initial field and lateral boundary conditions on the model grid points to obtain the initial field data and lateral boundary condition field data. The format of the isobaric surface coarse grid file data is bckgdata. S3: Based on the cloud analysis system, the total cloud volume of domestic Fengyun satellites, hourly average blackbody brightness temperature data, radar three-dimensional network reflectivity dataset, and the initial field grid data after analysis and assimilation are input into the cloud analysis system. S4: Construct a random perturbation module for cloud analysis observation data: Collect the lowest radar reflectivity of the cloud analysis system, use a normal distribution random number generator, and perform random perturbation based on the default mean and standard deviation to generate a random threshold for the lowest radar reflectivity that conforms to the default mean and standard deviation; use random sampling methods to sparsely extract hourly average blackbody brightness temperature data and total cloud volume data from domestic Fengyun satellites to obtain uncertainty data for hourly average blackbody brightness temperature data and total cloud volume, thereby controlling the radar and satellite observation data entering the cloud analysis system, reflecting the uncertainty of the observation data, and providing the ensemble forecasting system with multiple cloud analysis input observation datasets with different initial observation values; S5: Three-dimensional random perturbation module for key sensitive parameters of the cloud analysis system: Based on the default values ​​of the original parameters in the system, and considering the vertical changes of the key parameters in the cloud analysis system, a random perturbation module for sensitive parameters is constructed by randomly perturbing the following key parameters with different thresholds using a normal distribution random perturbation generator: threshold for converting relative humidity to cloud cover (Rh0) + minimum threshold for converting cloud cover to relative humidity (rh_thr1) + maximum threshold for converting cloud cover to relative humidity (rh_thr2) + minimum cloud cover threshold for converting cloud cover to relative humidity (cvr2rh_thr1) + cloud cover threshold for reaching saturation when converting cloud cover to relative humidity (cvr2rh_thr1).

[0024] S6: Construction of Joint Random Perturbation Method for Cloud Analysis System: In the cloud analysis system, combined with the above perturbation modules, joint random perturbation is performed on the cloud analysis system observation data (S4) and key sensitive parameters (S5) simultaneously, which reflects the uncertainty of the input data of the cloud analysis system and the uncertainty of the dynamic and thermodynamic processes of the cloud analysis system itself.

[0025] S7: Based on the WRF model, model integration is performed on the control forecast (normal forecast, without any disturbance), multiple cloud analysis input observation datasets reflecting the initial uncertainty of the cloud analysis system (as input information for multiple ensemble members), and key sensitivity parameters reflecting the uncertainty of cloud analysis dynamic and thermodynamic processes, respectively, to obtain the ensemble forecast results of multiple members.

[0026] Based on different altitudes, the relative humidity and total cloud volume data of the cloud analysis system are converted to each other to obtain the conversion results. Based on the analysis and assimilation of the observation data and the radar three-dimensional network reflectivity dataset corresponding to the conversion results, joint random perturbation is performed, and the ensemble forecast product corresponding to the ensemble forecast data is obtained by using the control forecast integral method.

[0027] In this embodiment, the cloud analysis system lacks total cloud volume data and precipitation information.

[0028] In this embodiment, a normal distribution random number generator is used to randomly perturb the cloud analysis system to obtain a random distribution of the analyzed and assimilated observation data.

[0029] Preferably, step S1 includes collecting and analyzing assimilated observation data from ground meteorological observation stations; obtaining raw blackbody brightness temperature data from Fengyun satellite data receiving stations, and sequentially decoding, decompressing, and averaging the raw blackbody brightness temperature data to obtain hourly average blackbody brightness temperature data; acquiring satellite cloud images and radar data, and using cloud detection algorithms to calculate the total cloud volume data for the satellite cloud images and radar data respectively; acquiring raw weather radar mosaic data, querying the legend reflectance RGB and annotation RGB of the raw weather radar mosaic data to obtain a reflectance and annotation RGB mapping table, converting the RGB of annotations for all grid points in the raw weather radar mosaic data to obtain converted weather radar mosaic data, using interpolation to complete the missing grid points in the converted weather radar mosaic data to obtain a radar three-dimensional network reflectance dataset; and / or collecting raw data required for aggregation and assimilation, converting the format of the raw data required for aggregation and assimilation to obtain the data required for aggregation and assimilation.

[0030] Preferably, step S2 includes processing the isobaric coarse grid file data by sequentially using horizontal interpolation, vertical interpolation, and variable transformation to obtain the initial field and lateral boundary conditions; assigning a corresponding initial model value and a lateral boundary condition to the analysis and assimilation observation data, hourly average blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or dataset required for assimilation, respectively, to obtain the initial model value set data and the lateral boundary condition set data, wherein the initial model value set data is the initial field data, and the lateral boundary condition set data is the lateral boundary condition field data.

[0031] Preferably, step S3 includes the following steps: when the cloud analysis system receives the assimilated observation data, the normal distribution random number generator randomly perturbs the cloud analysis system to generate random distribution data of ground observation data, and uses a random sampling method to sparsify and extract hourly average blackbody brightness temperature data to obtain uncertain state data of hourly average blackbody brightness temperature data.

[0032] Preferably, step S5 includes a threshold based on the total cloud volume data corresponding to different altitudes, converting the relative humidity with a standard deviation of 0.05 into random total cloud volume data, and based on the cloud analysis system, setting a random perturbation value with a default mean of 0.5 and a standard deviation of 0.1 to generate the lowest random threshold for the relative humidity corresponding to the total cloud volume data. Set a random perturbation value with a default mean of 1 and a standard deviation of 0.05 to generate the highest random threshold for relative humidity corresponding to the total cloud volume data; Set a random perturbation value with a default mean of 0.2 and a standard deviation of 0.1 to generate the minimum random threshold for the total cloud cover data corresponding to relative humidity; Set a random perturbation value with a default mean of 0.7 and a standard deviation of 0.1 to generate the corresponding total cloud volume data when the relative humidity reaches saturation.

[0033] In this embodiment, the threshold values ​​for total cloud cover data corresponding to different altitudes are: When the altitude is less than 600 meters, the threshold for total cloud cover data is 0.95; When the altitude is 600-1500 meters, the threshold for total cloud cover data is 0.9; When the altitude is between 1501 and 2500 meters, the threshold for total cloud cover data is 0.85. When the altitude is greater than 2500 meters, the threshold for total cloud cover data is 0.8.

[0034] Preferably, step S5 includes using a random generator to randomly perturb the analytical assimilation observation data outside the cloud analysis system; generating initial values ​​for regional ensemble forecast perturbations based on a three-dimensional variational system and ensemble assimilation method; generating initial field files and lateral boundary condition files for ensemble forecast data based on convective-scale ensemble forecast initial value perturbation method and lateral boundary perturbation method; integrating the initial field files and lateral boundary condition files using a 3km resolution WRF model and dynamic downscaling method to obtain control forecast integration results; and obtaining perturbation member integrations for the ensemble forecast data using the perturbation model in the WRF model to obtain ensemble forecast products.

[0035] In this embodiment, the method of control forecast integration is to integrate the control forecast initial field and lateral boundary conditions obtained by the dynamic downscaling method to obtain the control forecast integration result; In this embodiment, the method of ensemble member integration utilizes the initial field of the model generated by ensemble integrator, the lateral boundary conditions from the global ensemble forecast, and the cloud analysis joint perturbation cloud analysis initial field obtained by the present invention. Finally, each ensemble member in the ensemble forecast data generates a cloud analysis initial field that reflects the uncertainty of the microphysical thermodynamic information of the cloud analysis process. For each ensemble forecast member, the uncertainty of the physical process is characterized by the SPPT stochastic perturbation scheme based on the first-order Markov process. The integration process is then run to obtain the perturbation member forecast results.

[0036] like Figure 2 As shown, this embodiment provides a system, including: Data acquisition module 1, configured based on WRF mode, has completely independent intellectual property rights, acquires the analysis and assimilation observation data of the monitoring area, hourly average blackbody brightness temperature data, total cloud volume data, radar three-dimensional network reflectivity dataset and / or data required for analysis and assimilation and cloud analysis, and obtains ensemble forecast data; Information extraction module 2 is configured to use background extraction technology to extract background field information from the ensemble forecast data, obtain isobaric surface coarse grid file data, and convert the isobaric surface coarse grid file data into the initial field and side boundary conditions on the mode grid points of the mode, thereby obtaining initial field data and side boundary condition field data. The cloud analysis system observation data module 3 inputs the total cloud data of domestic Fengyun satellites, hourly average blackbody brightness temperature data, radar three-dimensional network reflectivity dataset, and the initial field grid data after analysis and assimilation into the cloud analysis system. The cloud analysis observation data random perturbation module 4 aggregates the minimum radar reflectivity of the cloud analysis system. It uses a normal distribution random number generator to generate a random threshold for the minimum radar reflectivity that conforms to the default mean and standard deviation by randomly perturbing the standard deviation based on the default mean. It uses a random sampling method to sparsely extract the hourly average blackbody brightness temperature data and total cloud volume data of the domestic Fengyun satellite to obtain the uncertainty data of the hourly average blackbody brightness temperature data and total cloud volume. This controls the radar and satellite observation data entering the cloud analysis system, reflects the uncertainty of the observation data, and provides the ensemble forecast system with multiple cloud analysis input observation datasets with different initial observation values. Cloud Analysis System Key Sensitive Parameter Three-Dimensional Random Perturbation Module 5: For the following key sensitive parameters in the cloud analysis system, based on the default values ​​of the original parameters in the system, and considering the vertical changes of the key parameters in the cloud analysis system, a normal distribution random perturbation generator is used to randomly perturb the following key parameters at different thresholds to construct a sensitive parameter random perturbation module: threshold for converting relative humidity to cloud cover (Rh0) + minimum threshold for converting cloud cover to relative humidity (rh_thr1) + maximum threshold for converting cloud cover to relative humidity (rh_thr2) + minimum cloud cover threshold for converting cloud cover to relative humidity (cvr2rh_thr1) + cloud cover threshold for cloud cover to reach saturation state (cvr2rh_thr1). The cloud analysis system is subjected to joint random perturbation 6. In conjunction with the above modules, the cloud analysis system observation data (S4) and key sensitive parameters (S5) are subjected to joint random perturbation at the same time, which reflects the uncertainty of the input data of the cloud analysis system and the uncertainty of the dynamic and thermodynamic processes of the cloud analysis system itself. Product acquisition module 7 is configured to convert the relative humidity and total cloud volume data of the cloud analysis system to each other based on different altitudes, obtain conversion results, perform joint random perturbation on the analysis and assimilation observation data and radar three-dimensional network reflectivity dataset corresponding to the conversion results, and obtain the ensemble forecast product corresponding to the ensemble forecast data by using control forecast and integration of different ensemble members.

[0037] This invention utilizes the uncertainty of WRF models and the cloud microphysical information and its equilibrium thermodynamic information in the initial field of the model. By combining ensemble forecasting methods to jointly and randomly perturb the input data of the cloud analysis system and the key sensitivity parameters of the cloud analysis system itself during the initial field formation process of the analysis and assimilation data, hourly average blackbody brightness temperature data, total cloud volume data, radar three-dimensional network reflectivity dataset, and / or data required for ensemble assimilation, the distribution of the three-dimensional condensate and humidity fields in the initial field of the model can be dynamically adjusted.

[0038] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for cloud analysis with stochastic joint perturbation in regional convective-scale ensemble forecasting, characterized in that, The method, applicable to convective-scale models with a horizontal resolution of 2km to 4km, including WRF and CMA-MESO models, comprises: Based on the WRF model, the system acquires the analytical assimilation observation data, hourly average blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or the data required for analytical assimilation in the monitoring area to obtain ensemble forecast data. Among them, the data required for analytical assimilation includes aircraft report data, radiosonde data, and GPS water vapor inversion data. Background extraction technology is used to extract background field information from the ensemble forecast data to obtain isobaric surface coarse grid file data. The isobaric surface coarse grid file data is then converted into the initial field and side boundary conditions on the model grid points of the model to obtain initial field data and side boundary condition field data. Based on a normal distribution random number generator and a cloud analysis system, the total cloud volume data of domestic Fengyun satellites, hourly average blackbody brightness temperature data, and three-dimensional radar network reflectivity dataset are input into the cloud analysis system. Based on the statistical characteristics of observation errors, a set of observation data with the same probability in a statistical sense is generated, constituting two methods for representing the uncertainty of observation data in the cloud analysis system: The lowest radar reflectivity of the cloud analysis system is collected, and the random perturbation module of the observation data based on the default mean and standard deviation generates a random threshold for the lowest radar reflectivity that conforms to the default mean and standard deviation. To estimate the error characteristics of blackbody brightness temperature, a random sampling method is used to sparsify and extract hourly average blackbody brightness temperature data to obtain the uncertainty data of hourly average blackbody brightness temperature data. Based on the error characteristics of total cloud volume products, a random sampling method is used to sparsely extract data of total cloud volume products of domestic Fengyun satellites to obtain data on uncertain total satellite cloud volume. Based on different altitudes, the three-dimensional key sensitive parameter data of the cloud analysis system are converted to each other. On the basis of the default values ​​of the original parameters of the system, the following key parameters that reflect the thermal and dynamic uncertainties in the cloud analysis process are randomly perturbed at different thresholds through a normal distribution random perturbation generator, and a sensitive parameter random perturbation module is constructed: the threshold for converting relative humidity into cloud amount, the minimum threshold for converting cloud amount into relative humidity, the maximum threshold for converting cloud amount into relative humidity, the minimum cloud amount threshold for converting cloud amount into relative humidity, and the cloud amount threshold for converting cloud amount into relative humidity to reach saturation. The conversion results are obtained by calculating key sensitive parameters of thermal and dynamic uncertainties during cloud analysis. Based on the hourly average blackbody brightness temperature data, total cloud data and radar three-dimensional network reflectivity dataset corresponding to the conversion results from domestic Fengyun satellites, joint random perturbation is performed, and the ensemble forecast product corresponding to the ensemble forecast data is obtained by using the control forecast integral method.

2. The cloud analysis joint stochastic perturbation method for regional convective-scale ensemble forecasting according to claim 1, characterized in that: The WRF-based acquisition of the monitoring area's analytical assimilation observation data, hourly average blackbody brightness temperature data, total cloud cover data, radar 3D network reflectivity dataset, and / or data required for analysis and assimilation includes: Raw blackbody brightness temperature data is obtained from the Fengyun satellite data receiving station. The raw blackbody brightness temperature data is then decoded, decompressed, and time-averaged sequentially to obtain hourly average blackbody brightness temperature data. The raw data of total cloud volume is obtained from the Fengyun satellite data receiving station. The raw data of total cloud volume is then decoded and decompressed to obtain the total cloud volume data. Obtain the original weather radar mosaic data, query the legend reflectance RGB and annotation RGB of the original weather radar mosaic data, obtain the reflectance and annotation RGB mapping table, convert the annotation RGB of all grid points of the original weather radar mosaic data to reflectance, obtain the converted weather radar mosaic data, use the interpolation method to complete the missing grid points of the converted weather radar mosaic data, and obtain the radar three-dimensional network reflectance dataset. And / or collect raw data required for analysis and assimilation, convert the format of the raw data required for analysis and assimilation to obtain the data required for analysis and assimilation, and improve the quality of assimilation data through quality control schemes.

3. The cloud analysis joint stochastic perturbation method for regional convective-scale ensemble forecasting according to claim 2, characterized in that: The background extraction technique is used to extract background field information from the ensemble forecast data to obtain isobaric surface coarse grid file data. This isobaric surface coarse grid file data is then converted into initial field and lateral boundary conditions at the model grid points of the model, resulting in initial field data and lateral boundary condition field data, including: The isobaric coarse mesh file data was processed sequentially using horizontal interpolation, vertical interpolation, and variable transformation to obtain the initial field and side boundary conditions; Each of the following data is assigned a corresponding initial model value set and a lateral boundary condition: observational data, radiosonde data, ship data, satellite cloud wind guidance and GPS water vapor inversion data, hourly average blackbody brightness temperature data, total cloud cover data, radar three-dimensional network reflectivity dataset, and / or data required for analysis and assimilation. This results in the initial model value set data and the lateral boundary condition set data. The initial model value set data is the initial field data, and the lateral boundary condition set data is the lateral boundary condition field data.

4. The method for representing the uncertainty of input data in the cloud analysis system for regional convective-scale ensemble forecasting according to claim 3, characterized in that: The cloud analysis system, based on a normal distribution random number generator, inputs data on total cloud volume from domestic Fengyun satellites, hourly average blackbody brightness temperature data, and radar three-dimensional network reflectivity datasets into the cloud analysis system, including: When the cloud analysis system receives radar reflectivity data, it uses a normal distribution random number generator to randomly perturb the radar reflectivity around the default mean, generating the lowest random radar reflectivity result that conforms to the default mean and standard deviation, reflecting the uncertain random distribution data of radar reflectivity. When the cloud analysis system receives hourly average blackbody brightness temperature data, it uses a random sampling method to sparsely extract the hourly average blackbody brightness temperature data based on the error characteristics of the blackbody brightness temperature, thereby obtaining the uncertain state data of the hourly average blackbody brightness temperature data. When the cloud analysis system receives the total cloud volume data of the domestic Fengyun satellite, it uses a random sampling method to sparsely extract the total cloud volume data based on the error characteristics of the total cloud volume product. The extracted samples reflect the spatiotemporal correlation of the product error as much as possible, thus obtaining the uncertainty data of the total cloud volume.

5. The cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting according to claim 4 further includes: Random perturbations are applied to the key dynamic and thermodynamic sensitive parameters of the cloud analysis system's microphysical parameterization, forming a random perturbation module for key sensitive parameters that reflects the uncertainty of the cloud analysis microphysical parameterization scheme.

6. The cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting according to claim 5, characterized in that: Based on different altitudes, the relative humidity and total cloud volume data of the cloud analysis system are converted to each other to obtain the conversion results, including: Based on the threshold of total cloud volume data corresponding to different altitudes, relative humidity with a standard deviation of 0.05 is converted into random total cloud volume data.

7. The cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting according to claim 5, characterized in that: Based on different altitudes, the relative humidity and total cloud volume data of the cloud analysis system are converted to each other to obtain the conversion results, including: Based on a cloud analytics system, a threshold was set to convert relative humidity as it changes with altitude into cloud cover. When the altitude is less than 600m, the mean is 0.925 and the standard deviation is 0.025; when the altitude is 600m-1500m, the mean is 0.9 and the standard deviation is 0.025; when the altitude is 1500m-2500m, the mean is 0.85 and the standard deviation is 0.025; and when the altitude is greater than 2500m, the mean is 0.8 and the standard deviation is 0.

075. Set a random perturbation value with a default mean of 0.5 and a standard deviation of 0.1 to generate the lowest random threshold for relative humidity corresponding to the total cloud volume data; Set a random perturbation value with a default mean of 1 and a standard deviation of 0.05 to generate the highest random threshold for relative humidity corresponding to the total cloud volume data; Set a random perturbation value with a default mean of 0.2 and a standard deviation of 0.1 to generate the minimum random threshold for the total cloud cover data corresponding to relative humidity; Set a random perturbation value with a default mean of 0.7 and a standard deviation of 0.1 to generate the corresponding total cloud volume data when the relative humidity reaches saturation.

8. The cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting according to claim 7, characterized in that: The data, including hourly average blackbody brightness temperature data from domestic Fengyun satellites, total cloud cover data, and radar three-dimensional network reflectivity datasets, are subjected to joint random perturbation based on the transformation results. Then, a control forecast integration method is used to obtain the ensemble forecast products corresponding to the ensemble forecast data, including: A random generator is used to randomly perturb the external analytical assimilation data of the cloud analysis system. Based on the three-dimensional variational system and the set assimilation method, the initial values ​​of the regional set forecast perturbation are generated.

9. The cloud analysis stochastic joint perturbation method for regional convective-scale ensemble forecasting according to claim 8, characterized in that: To address the uncertainties in cloud analytics input data and key sensitive parameters reflecting the uncertainties in the cloud analytics microphysical parameterization process, joint random perturbations are applied. The cloud analytics input data includes assimilated observational data, radar 3D network reflectivity, and TBB datasets. Key sensitive parameters include the threshold for converting relative humidity to cloud cover, the minimum threshold for converting cloud cover to relative humidity, the maximum threshold for converting cloud cover to relative humidity, the minimum cloud cover threshold for converting cloud cover to relative humidity, and the cloud cover threshold at which relative humidity reaches saturation. The method of using control forecast integration to obtain the ensemble forecast product corresponding to the ensemble forecast data also includes: Initial field files and lateral boundary condition files for ensemble forecast data are generated based on the convective-scale ensemble forecast initial value perturbation method and the lateral boundary perturbation method. Using a 3km resolution WRF model and a dynamic downscaling method, the initial field file and the lateral boundary condition file are integrated to obtain the control forecast integral results; The perturbation member integrals are obtained by applying the perturbation mode in the WRF model to the ensemble forecast data, and the ensemble forecast products are obtained.

10. A system, characterized in that, include: The data acquisition module (1) is configured to acquire the analysis and assimilation observation data, hourly average blackbody brightness temperature data, total cloud volume data, radar three-dimensional network reflectivity dataset and / or data required for analysis and assimilation and cloud analysis in the monitoring area based on WRF mode, and obtain ensemble forecast data. Information extraction module (2), the information extraction module (2) is configured to use background extraction technology to extract background field information from the set forecast data, obtain isobaric surface coarse grid file data, convert the isobaric surface coarse grid file data into the initial field and side boundary conditions on the mode grid points of the mode, and obtain initial field data and side boundary condition field data; The cloud analysis system observation data module (3) inputs the total cloud data of domestic Fengyun satellites, hourly average blackbody brightness temperature data, radar three-dimensional network reflectivity dataset, and the initial field grid data after analysis and assimilation into the cloud analysis system; The cloud analysis observation data random perturbation module (4) collects the lowest radar reflectivity of the cloud analysis system. It uses a normal distribution random number generator to perform random perturbation of the standard deviation based on the default mean, and generates a random threshold for the lowest radar reflectivity that conforms to the default mean and standard deviation. It uses a random sampling method to sparsely extract the hourly average blackbody brightness temperature data and total cloud data of the domestic Fengyun satellite, and obtains the uncertainty data of the hourly average blackbody brightness temperature data and total cloud data, thereby controlling the radar and satellite observation data entering the cloud analysis system, reflecting the uncertainty of the observation data, and providing the ensemble forecast system with multiple cloud analysis input observation datasets with different initial observation values. Three-dimensional random perturbation module for key sensitive parameters of cloud analysis system (5): For the following key dynamic and thermal sensitive parameters in cloud analysis system, based on the default values ​​of the original parameters of the system, considering the vertical changes of the key parameters of cloud analysis system, the following key parameters are randomly perturbed at different thresholds through a normal distribution random perturbation generator to construct a random perturbation module for sensitive parameters: threshold for converting relative humidity into cloud amount (Rh0) + minimum threshold for converting cloud amount into relative humidity (rh_thr1) + maximum threshold for converting cloud amount into relative humidity (rh_thr2) + minimum cloud amount threshold for converting cloud amount into relative humidity (cvr2rh_thr1) + cloud amount threshold for converting cloud amount into relative humidity to reach saturation state (cvr2rh_thr1). The cloud analysis system is subjected to joint random perturbation (6). In conjunction with the above modules, the key sensitive parameters of dynamics and thermodynamics of the cloud analysis system observation data and microphysical parameterization scheme are subjected to joint random perturbation at the same time. This reflects the uncertainty of the input data of the cloud analysis system and the uncertainty of the dynamic and thermodynamic processes of the cloud analysis system itself. Product acquisition module (7) is configured to convert the relative humidity and total cloud volume data of the cloud analysis system to each other based on different altitudes, obtain conversion results, perform joint random perturbation based on the analysis and assimilation observation data and radar three-dimensional network reflectivity dataset corresponding to the conversion results, and obtain the ensemble forecast product corresponding to the ensemble forecast data by using control forecast and integration of different ensemble members.

Images