Lightning forecasting method based on mesoscale mode polarization radar data assimilation

Abstract

The invention relates to the technical field of lightning forecasting, and discloses a lightning forecasting method based on mesoscale mode polarization radar data assimilation, which comprises the following steps: acquiring dual-polarization radar data, preprocessing the dual-polarization radar data, and generating an observation error covariance matrix; obtaining reanalysis data as an initial field and a boundary condition of a meteorological mode, constructing a background error covariance matrix as a background field, and performing cyclic assimilation through a reflectivity assimilation observation operator, a radial speed assimilation observation operator and a differential reflectivity assimilation observation operator by adopting a three-dimensional variational assimilation method; and taking the forecast field obtained after each assimilation as a background field of the next time, and finally generating a lightning forecast result. According to the invention, the influence of radar data errors on the accuracy of numerical mode forecasting is reduced, so that the accuracy of lightning forecasting under complex terrains can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of lightning prediction, and in particular to a lightning prediction method based on mesoscale model polarization radar data assimilation. Background Art

[0002] There are two main methods for forecasting lightning activity at present. The first type is the centroid tracking forecasting method, such as the TITAN and SCIT algorithms. The second lightning forecasting method is mainly based on numerical weather forecasting models. At present, in the field of numerical models, foreign researchers have achieved some good results in the research on the direction of forecasting lightning. The numerical forecasting methods of lightning by numerical models are mainly divided into two categories: lightning forecasting models can be divided into two categories. The first type of model is the numerical diagnostic forecast of lightning based on statistical relationships. This type of model analyzes historical data to identify weather patterns that may trigger lightning. The second type of model establishes a parameterization scheme for lightning rate based on the interaction between cloud dynamics, microphysical processes and electrification activities. This method allows scientists to predict the occurrence of lightning by simulating physical processes in clouds, thereby providing more direct and mechanistic predictions. These two methods have their own advantages. The former relies on statistical data to provide rapid predictions, while the latter focuses more on starting from basic physical processes to improve the accuracy and scientificity of predictions.

[0003] In recent years, Doppler radars have been gradually upgraded to polarization radars. Compared with conventional Doppler weather radars that only observe polarization in a single direction (usually horizontally), dual-polarization radars can detect additional polarization physical quantities such as differential reflectivity (ZDR), differential propagation phase shift rate (KDP), and zero-lag mutual correlation coefficient (ρHV) of precipitation systems by alternately or simultaneously observing horizontal and vertical polarization directions. Many studies have shown that these polarization quantities can provide stronger support for the near-warning and forecasting of disastrous weather.

[0004] At present, the research on polarimetric radar data at home and abroad mainly focuses on the fields of quality control, prediction of disastrous weather, and identification of precipitation particle phase. However, in the research on the assimilation of polarimetric radar data, the current involvement is still quite limited. It can be seen that the assimilation research of polarimetric radar data needs further in-depth exploration to give full play to its application potential in the field of meteorology. Polarimetric radar data can provide cloud microphysical information that is closely related to the development of severe convective systems. It is urgent to study a lightning forecasting method based on assimilated polarimetric radar data to further improve the accuracy of numerical forecasts, which is of great significance for optimizing weather forecast models and improving forecast effects. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a lightning forecasting method based on mesoscale model polarization radar data assimilation.

[0006] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0007] A lightning forecasting method based on mesoscale model polarization radar data assimilation comprises the following steps:

[0008] Acquire dual-polarization radar data, preprocess the dual-polarization radar data, and generate an observation error covariance matrix;

[0009] Reanalysis data are obtained as the initial field and boundary conditions of the meteorological model, and the background error covariance matrix is ​​constructed as the background field. The three-dimensional variational assimilation method is used to perform cyclic assimilation through the reflectivity assimilation observation operator, the radial velocity assimilation observation operator and the differential reflectivity assimilation observation operator. The forecast field obtained after each assimilation is used as the background field for the next time, and finally the lightning forecast result is generated.

[0010] Preferably, preprocessing the dual-polarization radar data includes:

[0011] The dual-polarization radar data is subjected to velocity fuzzy processing using a guided path algorithm;

[0012] The single radial nine-point median method is used to denoise the reflectivity of dual-polarization radar data.

[0013] Perform ground clutter filtering on dual-polarization radar data;

[0014] Perform format conversion on dual polarization radar data.

[0015] Preferably, performing format conversion processing on the dual-polarization radar data includes:

[0016] Eliminate the missing values ​​in the dual-polarization radar data;

[0017] Remove the data of the first elevation angle of the radar, and then use the bilinear interpolation method to interpolate the radar observation data to the model horizontal grid for coarsening, and set the resolution to be consistent with the model grid resolution so that the radar data is evenly distributed in the horizontal space;

[0018] Set radial velocity observation error and reflectivity observation error;

[0019] Convert the processed radar data into littler format.

[0020] Preferably, the objective function of the three-dimensional variational assimilation method is expressed as:

[0021]

[0022] Where J is the objective function, x′ is , x is the analysis variable, T is the transpose, B is the background error covariance matrix, H′ is , d is the update vector, and O is the observation error.

[0023] Preferably, the three-dimensional variational assimilation method includes:

[0024] Taking stream function, velocity potential, imbalance pressure and water vapor mixing ratio as control variables, the increments of water vapor mixing ratio, cloud water mixing ratio and rain water mixing ratio are obtained by adding warm rain process.

[0025] Preferably, the reflectivity assimilation observation operator is expressed as:

[0026] Z = 43.1 + 17.5log(ρq r )

[0027] Where Z is the radar reflectivity, ρ is the air density, and q r is the rainwater mixing ratio.

[0028] Preferably, the radial velocity assimilation observation operator is expressed as:

[0029]

[0030] Among them, V r is the radar radial wind speed, u is the radial wind speed, v is the zonal wind speed, w is the vertical speed, (x, y, z) is the coordinates of the observation station, (x i ,y i ,z i ) is the coordinate of the radar site, v r is the falling velocity of precipitation particles, r i is the distance from the observation site to the radar site.

[0031] Preferably, the differential reflectivity assimilation observation operator is expressed as:

[0032]

[0033] Among them, Z DR is the differential reflectivity, Z H is the horizontal reflectivity, Z V is the vertical reflectivity.

[0034] Preferably, the three-dimensional variational assimilation method further includes:

[0035] A double grid nesting method is used to obtain the optimal parameterization solution.

[0036] Preferably, the dual grid nesting method is used to obtain the optimal parameterization solution, including:

[0037] Setting the first horizontal resolution and the second horizontal resolution of the weather mode;

[0038] Setting the first time resolution and the second time resolution of the meteorological model;

[0039] In the vertical direction, the σ terrain following coordinate coefficient is used to set the vertical resolution, forming two nested areas with different resolutions of the meteorological model.

[0040] The present invention has the following beneficial effects:

[0041] The present invention controls the quality of dual-polarization radar data and uses a three-dimensional variational assimilation method to perform cyclic assimilation based on a reflectivity assimilation observation operator, a radial velocity assimilation observation operator and a differential reflectivity assimilation observation operator, thereby reducing the influence of radar data errors on the accuracy of numerical model prediction, thereby improving the accuracy of lightning prediction in complex terrain. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The figure is a flowchart of a lightning forecasting method based on mesoscale model polarimetric radar data assimilation. DETAILED DESCRIPTION

[0043] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0044] like Figure 1 As shown, a lightning prediction method based on mesoscale mode polarization radar data assimilation provided by an embodiment of the present invention includes the following steps S1 and S2:

[0045] S1, obtaining dual-polarization radar data, and preprocessing the dual-polarization radar data to generate an observation error covariance matrix;

[0046] In an optional embodiment of the present invention, step S1 uses a CC polarization radar, which can detect the occurrence, development and change process of clouds and rain, and can also simultaneously emit electromagnetic waves of two polarization modes, horizontal and vertical, to detect the state information of particles. By utilizing the measurement capability of the dual polarization radar, the amount of rainfall, the shape, size, direction, state and droplet spectrum distribution of precipitation particles can be derived in detail. The accurate acquisition of these data enables this embodiment to clearly distinguish different types of precipitation, such as hail, rain, snow and sleet. The application of this technology not only provides an important scientific basis for artificial weather modification and hail prevention operations, but also greatly improves the accuracy of quantitative precipitation estimation (QPE), thereby enhancing the effectiveness of weather forecasting and disastrous weather monitoring. This embodiment can achieve continuous observation 24 hours a day through the deployed dual polarization radar, and operates in accordance with the VCP21 mode with a high time resolution of 6 minutes. This high-frequency data recording can capture rapidly changing weather conditions and provide continuous and real-time observation data for meteorological analysis. This information is not only crucial for understanding local weather dynamics, but also provides strong technical support for preventing and preparing for possible meteorological disasters. With such advanced monitoring equipment, researchers and meteorological experts can analyze and predict weather patterns more accurately, making important contributions to protecting life and property and mitigating the impact of weather disasters. Previous experience has shown that during the data collection process, errors in observations may occur due to non-meteorological echoes and other reasons, which will affect the quantitative estimation of precipitation, the study of cloud microphysical processes, and data assimilation applications. Therefore, before the assimilation radar, strict quality control of the data is required to remove the influence of non-meteorological waves. The main technical indicators of the radar are shown in Table 1.

[0047] Table 1 Radar technical indicators

[0048]

[0049] In an optional embodiment of the present invention, step S1 of preprocessing the dual-polarization radar data includes:

[0050] The dual-polarization radar data is subjected to velocity fuzzy processing using a guided path algorithm;

[0051] The single radial nine-point median method is used to denoise the reflectivity of dual-polarization radar data.

[0052] Perform ground clutter filtering on dual-polarization radar data;

[0053] Perform format conversion on dual polarization radar data.

[0054] The format conversion process of dual polarization radar data includes:

[0055] Eliminate the missing values ​​in the dual-polarization radar data;

[0056] Remove the data of the first elevation angle of the radar, and then use the bilinear interpolation method to interpolate the radar observation data to the model horizontal grid for coarsening, and set the resolution to be consistent with the model grid resolution so that the radar data is evenly distributed in the horizontal space;

[0057] Set radial velocity observation error and reflectivity observation error;

[0058] Convert the processed radar data into littler format.

[0059] For example, the radial velocity field of the new generation of meteorological radars may have velocity ambiguity. This ambiguity hinders the application of radar velocity fields, complicates subjective analysis, and makes the results of quantitative precipitation estimation, data assimilation, and wind field inversion unreliable. Therefore, it is necessary to deal with velocity ambiguity.

[0060] This embodiment uses a quality-guided path method to deal with the speed ambiguity problem. This method prioritizes pixels with the highest quality scores, which show the highest reliability values, and gradually expands to pixels with lower reliability values ​​to reduce the possibility of error diffusion. The reliability of a pixel determines its position in the processing path, ensuring that the path determination process is optimized by reliability. Although the algorithm may not be able to completely avoid all expansion errors, especially those that rely on error propagation paths, it has been proven to be effective in practical applications. The method mainly sets the standard based on the gradient or difference between the pixel and the surrounding pixels. In particular, those pixels with the lowest 2π gradient difference with the surrounding are generally considered to have the highest processing priority and will be processed first. In addition, by introducing the secondary difference technique, the degree of convexity of the phase function can be measured more accurately, thereby effectively identifying inconsistent areas in the phase map. This comprehensive strategy enables the algorithm to have high accuracy and reliability in processing speed ambiguity.

[0061] The abnormal or irregular data points appearing in radar observation data may be caused by various interferences or errors, rather than real targets or reflections, which are noise points in radar observation. These noise points may interfere with the accurate identification and analysis of targets or phenomena, so they need to be removed or processed in radar data processing and analysis. In radar meteorological observations, those noises that are usually located outside the wide echo area, scattered and isolated are usually considered to be invalid observations. These noises may come from various factors, such as terrain interference, building reflections, equipment failures or reflections of atmospheric non-meteorological factors (such as flocks of birds or airplanes). Since these noises are unrelated to real meteorological phenomena, they may interfere with the accuracy of data analysis and weather forecasts. Since the resolution of the regional numerical model of the present embodiment is at least 1km, and the resolution of the radar range library used is 150m, filtering is required. The present embodiment adopts radial nine-point median filtering to make the reflectivity smoother. After filtering, the data not only maintains the trend of basic reflectivity changes, but also filters out isolated points.

[0062] In actual radar operation, ground clutter may be generated, which is unavoidable. Considering this problem, this embodiment filters out non-precipitation echoes with a reflectivity less than 10 dbz and removes radar data with an elevation angle less than 2° before radar assimilation.

[0063] When processing radar secondary data, the position of each range bin is described relative to the radar origin by three coordinate parameters: elevation (α), azimuth (β), and slant range (L). In order to effectively integrate radar data into the numerical prediction model, the latitude and longitude coordinates of each range bin and its vertical height relative to the earth's surface must be determined. This process involves the conversion from the radar's inherent coordinate system to the geographic coordinate system and height information. Although the resolution of the numerical model is still low compared to the density of the radar raw data, this limits the ability of the numerical model to capture more subtle motion patterns in radar observations. Therefore, the model often regards small motion patterns as noise, which produces observation errors in the radar data assimilation process, also known as representative errors. These errors need to be reduced as much as possible through precise coordinate transformation and data processing techniques to improve the accuracy of assimilated data and the reliability of model forecasts. If all radar observation information is directly assimilated into the numerical model, it will increase the amount of calculation greatly, and due to the correlation problem between the observation data, the assimilation results may be invalid or inaccurate. In order to avoid information loss and correlation problems, it is usually necessary to perform sparse processing on the original horizontal volume scan observation data of the radar before radar data assimilation.

[0064] Data coarsening, radar radial velocity and radial velocity processing, and radar format conversion are performed in the following four steps:

[0065] (1) In order to better utilize radar observation data for assimilation, it is usually necessary to remove the points that are displayed as missing values ​​(-999.00) in the original data. This is because these data points that lack effective information may interfere with the accuracy of data assimilation, resulting in invalid or inaccurate assimilation results.

[0066] (2) First, remove the data of the first elevation angle of the radar, and use the bilinear interpolation method to coarsen the radar observation data onto the model horizontal grid. The resolution is 1 km, which is consistent with the model grid resolution, so that the radar data is evenly distributed in the horizontal space. This can overcome the problem of insufficient internal data caused by the gradual decrease in data density from the center of the radar to the outside.

[0067] (3) The radial velocity observation error and reflectivity observation error are taken as empirical values ​​of 2 and 5 respectively.

[0068] (4) According to the assimilation format of the WRFDA official website, the processed radar data is converted into the littler format to complete the format conversion.

[0069] In an optional embodiment of the present invention, step S1 uses local mean square error statistics to calculate the observation error of the preprocessed dual-polarization radar data, thereby generating an observation error covariance matrix.

[0070] S2. Obtain reanalysis data as the initial field and boundary conditions of the meteorological model, and construct the background error covariance matrix as the background field. Use the three-dimensional variational assimilation method to perform cyclic assimilation through the reflectivity assimilation observation operator, radial velocity assimilation observation operator and differential reflectivity assimilation observation operator, and use the forecast field obtained after each assimilation as the background field for the next time, and finally generate the lightning forecast result.

[0071] In an optional embodiment of the present invention, this embodiment uses FNL (Final Analysis) reanalysis data provided by NCEP (National Centers for Environmental Prediction), which has a spatial resolution of 0.25°×0.25° and a temporal resolution of once every 6 hours. These reanalysis data are used as the initial field and boundary conditions of the meteorological model, providing the model with an accurate starting point and necessary boundary information to ensure the accuracy and practicality of the model prediction. By using such high-resolution data, researchers can capture the dynamic changes of the atmospheric state more finely, thereby improving the quality and accuracy of weather forecasts and climate simulations.

[0072] In an optional embodiment of the present invention, the background error (BE) covariance matrix is ​​an important tool used in data assimilation to describe the error characteristics of the initial conditions or background fields of the model. It is a matrix used to quantify the difference between model forecasts and observations, and to fuse the observed data with the model forecasts during the data assimilation process. The background error covariance matrix provides uncertainty information for the model forecast, usually expressed as an estimate of the uncertainty of the model state. This uncertainty can be determined based on the deviation statistics of historical model operations or the error characteristics of the model itself. In data assimilation, the background error covariance matrix is ​​used together with the observation error covariance matrix to calculate the Kalman gain, thereby combining the observed data with the model forecast to generate a more accurate analysis field. This helps to better initialize the model and improve the model's ability to predict future weather and climate change.

[0073] In order to accurately calculate this matrix, this embodiment uses the NMC (National Meteorological Center) method. This method uses the WRF (Weather Research and Forecasting) model background field data for one month from March 1 to March 31, 2023. By analyzing the data during this period, this embodiment can effectively estimate the statistical characteristics of the background error, thereby constructing an accurate background error covariance matrix.

[0074] In an optional embodiment of the present invention, the basic principle of the three-dimensional variational assimilation method is to obtain the best analysis field as close to the real atmospheric conditions as possible by solving the best guess value of the real atmospheric state of the objective function within the analysis time. In data assimilation, the core idea is to solve the extreme value problem of the objective functional J.

[0075] In the data assimilation process of numerical weather forecasting, the representative error refers to the error introduced when the model variable x is converted to the observed variable y through the observation operator H. This error originates from the limitation that the operator H cannot perfectly simulate the actual observation conditions and processes. Therefore, the representative error reflects the mismatch between the model and the observation. In the data assimilation framework, the role of the observation operator H is to map the predicted variable x in the pattern space to the observation space for comparison with the actual observation yo. If the representative error is taken into account, the equation y = H(x) can be further rewritten to include the influence of these errors, making the prediction of the observation more accurate. This embodiment includes the representative error E and the observation error O, and uses the incremental method to solve it, setting x = x-x b , x is the analysis increment, and let d = H(x b )-y o , d is the update vector, and the objective function of the three-dimensional variational assimilation method is expressed as:

[0076]

[0077] Where J is the objective function, x′ is the estimated value of the analysis variable, x is the analysis variable, T is the transpose, B is the background error covariance matrix, H′ is the tangent operator of the observation operator H, d is the update vector, and O is the observation error.

[0078] This embodiment has the following advantages after using the incremental algorithm. First, there is a unique minimum value. J(x) is a quadratic function. It can be seen that the objective function has a unique minimum value, which improves the accuracy of the algorithm. Furthermore, as an iterative algorithm, the incremental algorithm has a significant advantage in that it can significantly save time. In each iteration, it only needs to update some parameters without recalculating the entire objective function, thereby greatly reducing the computer's computing time and saving memory resources. In addition, during the calculation process, we use the method of converting control variables into pattern variables to further improve the efficiency of solution and analysis. These characteristics make the incremental algorithm more advantageous in data processing and pattern analysis.

[0079] In an optional embodiment of the present invention, due to radar reflectivity, radial velocity is not a model variable, which requires the radar observation to be linked to the model variable, and the mathematical relationship between them is called an observation operator. The observation operator is a mapping function used to convert the model state into the observation space during the data assimilation process. It describes the connection between the model forecast and the observation, and converts the model state into a predicted value corresponding to the observed data. The observation operator is usually represented by a mathematical function or algorithm, which maps the model state vector (usually multidimensional) to the observation space (usually one-dimensional). This mapping can be nonlinear because the relationship between the model forecast and the observation is often complex. In the data assimilation process, the role of the observation operator is to compare the model forecast with the observed data to evaluate the accuracy of the model forecast and adjust the model state according to the observed data. Through the observation operator, the difference between the model forecast and the observed data can be converted into an adjustment amount for the model state, thereby optimizing the initial conditions of the model to make it more consistent with the observed data.

[0080] Doppler radar data (including radial velocity and reflectivity) have significant characteristics compared to conventional observation data. Among them, radial velocity data is particularly noteworthy because it contains vertical motion information. This feature gives Doppler radar a significant advantage in capturing vertical motion patterns in the atmosphere, and plays a vital role in the occurrence and forecast of convective weather. Therefore, the background information of the mesoscale model can be corrected by assimilating radial velocity and radar reflectivity information. As mentioned above, due to the unique advantages of 3DVAR, many meteorological models have chosen the 3DVAR assimilation system, and WRF has also chosen 3DVAR assimilation. However, most 3DVAR systems do not support vertical velocity and hydrogeological analysis, so it is difficult to directly assimilate vertical velocity information and reflectivity information through the 3DVAR method. In order to effectively integrate Doppler radar data into the data assimilation process of the WRF-3DVAR algorithm, a linearized Richardson equation is specially designed in WRF-3DVAR. By expressing each variable in the equation as a combination of a basic state and a small increment, the reflectivity factor and vertical velocity component of the Doppler radar data can be assimilated at the same time. The linear sum of the Richardson equation is added to the 3DVAR system as a bridge between the vertical velocity component of the 3DVAR and Doppler radial velocity observations. In order to assimilate the reflectivity data, the present embodiment selects the water vapor mixing ratio (qt) as the control variable, and refines the increments of moisture and water flow in the physical conversion process (Up) by adding a warm rain process. The composition of the control variables includes multiple key elements: stream function, velocity potential, imbalance pressure, and various components of the water vapor mixing ratio. Such a design is intended to accurately simulate and analyze the moisture dynamics in the atmosphere, thereby improving the effect and accuracy of reflectivity data assimilation. Through this method, the precipitation process can be better understood and predicted, which is of great significance to the improvement and optimization of weather forecast models. When estimating the WRF-3DVAR background error statistics, since the cloud water variable does not follow a continuous distribution, it is not appropriate to select it in the control variables for background error statistics. Therefore, this embodiment introduces qt as a control variable and calculates the background error statistics based on qt. The warm rain process covers the key links such as water vapor condensation into clouds (PCON), cloud accretion by rain (PRA), cloud automatic conversion into rain (PRC) and rain evaporation into water vapor (PRE), which are effectively integrated into the 3DVAR system. In the 3DVAR analysis process, the increments of water vapor mixing ratio qt, cloud water mixing ratio qv and rain water mixing ratio qra are obtained through the partitioning mechanism. The effective application of the warm rain parameterization scheme enables the close relationship between cloud water, humidity, rain and temperature to be established. Once the three-dimensional variational system generates increments of qc and qr, the assimilation process of reflectivity is simplified, thereby improving the efficiency and accuracy of assimilation.

[0081] Currently, in WRF-3DVAR, there are direct assimilation and indirect assimilation methods for assimilating reflectivity data. The direct assimilation method can handle complex nonlinear relationships and directly use basic reflectivity factor data for analysis. In addition, it can effectively fuse observation data from different sources, different types and with different error characteristics, and can avoid serious errors caused by ill-posedness in the inversion process. Radar measurements depict the relationship between reflectivity and liquid or solid water content through an observation operator. Under the premise of assuming that the raindrop distribution follows the Marshal-Palmer distribution, this embodiment establishes an analytical form of the basic reflectivity factor observation operator expressed as:

[0082] Z = 43.1 + 17.5log(ρq r )

[0083] Where Z is the radar reflectivity, ρ is the air density, and q r is the rainwater mixing ratio. This expression provides a method to quantify the relationship between reflectivity and water content, thereby enhancing the efficiency and accuracy of data assimilation. This embodiment uses the reflectivity assimilation observation operator to directly implement the assimilation of reflectivity data in the WRF-3DVAR system; and the radar reflectivity is first inverted into the rainwater mixing ratio for assimilation, which can achieve better results than directly assimilating the radar reflectivity.

[0084] In WRF-3DVAR, this embodiment uses the three-dimensional wind field (u, v, w) and the final velocity v of falling raindrops in the model. r , the distance between the observed target and the radar antenna r i The radial velocity assimilation observation operator in WRFDA is constructed and expressed as:

[0085]

[0086] Among them, V r is the radar radial wind speed, u is the radial wind speed, v is the zonal wind speed, w is the vertical speed, (x, y, z) is the coordinates of the observation station, (x i ,y i ,z i ) is the coordinate of the radar site, v r is the falling velocity of precipitation particles, r i is the distance from the observation site to the radar site.

[0087] Doppler radars usually only make observations in a single horizontal polarization, while the innovation of dual-polarization radar technology is that it can simultaneously or alternately make observations in the horizontal and vertical polarization directions. This technology improves the detection capability of the radar, enabling it to not only capture conventional meteorological data, but also measure multiple key precipitation parameters. Specifically, dual-polarization radars can measure parameters such as differential reflectivity (ZDR), differential propagation phase shift (KDP), and zero-lag cross-correlation coefficients. The measurement of these parameters provides more comprehensive and accurate weather information for meteorological research and operational applications, meets the demand for advanced meteorological data, and thus greatly promotes the accuracy and efficiency of meteorological forecasting and analysis. One of the most important variables of dual-polarization radar is the differential reflectivity ZDR, which is defined as the ratio between the horizontal and vertical power echoes and is expressed as:

[0088]

[0089] Among them, Z DR is the differential reflectivity, Z H is the horizontal reflectivity, Z V is the vertical reflectivity.

[0090] In an optional embodiment of the present invention, in order to improve the prediction accuracy of the meteorological model, this embodiment adopts a double grid nesting design. Specifically, the two horizontal resolutions of the model are set to 3 kilometers and 1 kilometer, respectively, and the corresponding number of grid points is 130×130 and 140×140. This high-resolution grid design allows for more detailed capture of small-scale meteorological phenomena and improves the model's ability to simulate complex terrain and local weather systems. In terms of temporal resolution, the integration time step is set to 15 seconds and 3 seconds, respectively, which helps to deal with rapidly changing meteorological processes and ensure the stability of the calculation and the accuracy of the results. The center point of the forecast area is selected at the geographic coordinates of 104.49°E, 28.48°N, which is located in a significant terrain environment, which is particularly critical for capturing related meteorological activities. In addition, the forecast model uses a σ terrain following coordinate system in the vertical direction, with a total of 34 layers. Such a vertical resolution helps to better simulate the dynamic process of the atmosphere from the ground to higher levels.

[0091] The model's microphysical process parameterization scheme uses the Lin scheme. This scheme improves the accuracy of simulating cloud and precipitation processes by classifying ice-phase particles in detail. This scheme is particularly suitable for dealing with complex ice-phase processes, and therefore shows good performance in capturing cold cloud processes. In terms of radiation transfer, the model selects the RRTMG long-wave radiation scheme and the short-wave radiation scheme, which are both commonly used radiation parameterization schemes in current meteorological simulations and can effectively simulate the dynamic transmission of solar radiation and earth radiation. The near-surface layer scheme uses the improved MM5 scheme, which is particularly important for simulating the dynamic and thermal characteristics of the near-surface atmosphere. The land surface process uses the Noah scheme, which is widely used in multiple models around the world and can accurately simulate the interaction between soil moisture, evaporation and vegetation. In terms of boundary layer parameterization, the YSU (Yonsei University) scheme is used, which is a non-local closed scheme that effectively describes the turbulent exchange process in the atmospheric boundary layer.

[0092] This embodiment uses the WRF model and selects the NCEP Global Forecast System (GFS) reanalysis data as the initial and boundary conditions of the meteorology. These data have a spatial resolution of 0.25°×0.25° and a temporal resolution updated every 6 hours. The model is set to generate a model output file every 30 minutes, which is then used as the background field for assimilation analysis. In the process of cyclic assimilation, the forecast field obtained after each assimilation is used as the background field for the next simulation. In this way, the initial field of the WRF model is continuously updated. This cycle continues until the forecast field of the last simulation becomes the reanalysis forecast field after assimilation. Such a cyclic assimilation method effectively utilizes the model output updated hourly, thereby enhancing the consistency and accuracy of the model prediction.

[0093] In an optional embodiment of the present invention, the LPI index is calculated mainly based on the dynamic and microphysical processes in the weather system. Lynn et al. applied the Lightning Potential Index (LPI) and the data simulated by the WRF model in their research to predict the lightning density in Italy and Israel. The results show that the LPI value can effectively indicate the main active areas of lightning. In this way, the LPI value, as a key indicator, has been proven to be of great value in predicting the geographical location and intensity of lightning. This prediction tool provides a scientific basis for understanding and predicting lightning activity in a specific area, which helps relevant departments take preventive measures to reduce the risks that lightning may bring.

[0094] In the numerical model, the mixing ratios of snow, ice and graupel particles are represented by qs (snow particle mixing ratio), qi (ice particle mixing ratio) and qg (graupel particle mixing ratio), respectively, where the mixing ratio of each particle is in kg / kg. When the mixing ratio of the total liquid water in the cloud and the ice phase particles reaches equilibrium, the dimensionless coefficient ε can reach its maximum value. This value is an important parameter for evaluating the intensity of the electrochemical process in the cloud, indicating that the electrochemical phenomenon in the cloud depends on the interaction of multiple hydrometeor particles. Although the LPI does not directly consider the breakdown electric field in the calculation process, the index can still reflect the close connection between lightning activity and the microphysical and dynamic processes in the cloud to a certain extent. Therefore, LPI has become a valuable tool for predicting and analyzing lightning activity.

[0095] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0096] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0097] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0098] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0099] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A lightning forecasting method based on mesoscale model polarization radar data assimilation, characterized in that: The following steps are involved: Acquire dual-polarization radar data, preprocess the dual-polarization radar data, and generate an observation error covariance matrix; Reanalysis data are obtained as the initial field and boundary conditions of the meteorological model, and the background error covariance matrix is ​​constructed as the background field. The three-dimensional variational assimilation method is used to perform cyclic assimilation through the reflectivity assimilation observation operator, the radial velocity assimilation observation operator and the differential reflectivity assimilation observation operator. The forecast field obtained after each assimilation is used as the background field for the next time, and finally the lightning forecast result is generated.

2. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: Preprocessing of dual polarization radar data includes: The dual-polarization radar data is subjected to velocity fuzzy processing using a guided path algorithm; The single radial nine-point median method is used to denoise the reflectivity of dual-polarization radar data. Perform ground clutter filtering on dual-polarization radar data; Perform format conversion on dual polarization radar data.

3. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 2 is characterized in that: The format conversion process for dual polarization radar data includes: Eliminate the missing values ​​in the dual-polarization radar data; Remove the data of the first elevation angle of the radar, and then use the bilinear interpolation method to interpolate the radar observation data to the model horizontal grid for coarsening, and set the resolution to be consistent with the model grid resolution so that the radar data is evenly distributed in the horizontal space; Set radial velocity observation error and reflectivity observation error; Convert the processed radar data into littler format.

4. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: The objective function of the three-dimensional variational assimilation method is expressed as: Among them, J is the objective function, x ′ is the estimated value of the analysis variable, x is the analysis variable, T is the transpose, B is the background error covariance matrix, H ′ is the tangent operator of the observation operator H, d is the update vector, and O is the observation error.

5. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: Three-dimensional variational assimilation methods include: Taking stream function, velocity potential, imbalance pressure and water vapor mixing ratio as control variables, the increments of water vapor mixing ratio, cloud water mixing ratio and rain water mixing ratio are obtained by adding warm rain process.

6. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: The reflectivity assimilation observation operator is expressed as: Z=43.1+17.5log(ρq r ) Where Z is the radar reflectivity, ρ is the air density, and q r is the rainwater mixing ratio.

7. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: The radial velocity assimilation observation operator is expressed as: Among them, V r is the radar radial wind speed, u is the radial wind speed, v is the zonal wind speed, w is the vertical speed, (x, y, z) is the coordinates of the observation station, (x i ,y i ,z i ) is the coordinate of the radar site, v r is the falling velocity of precipitation particles, r i is the distance from the observation site to the radar site.

8. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: The differential reflectivity assimilation observation operator is expressed as: Among them, Z DR is the differential reflectivity, Z H is the horizontal reflectivity, Z V is the vertical reflectivity.

9. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 1 is characterized in that: The three-dimensional variational assimilation method also includes: A double grid nesting method is used to obtain the optimal parameterization solution.

10. The lightning prediction method based on mesoscale model polarization radar data assimilation according to claim 9, characterized in that: The double grid nesting method is used to obtain the optimal parameterization solution including: Setting the first horizontal resolution and the second horizontal resolution of the weather mode; Setting the first time resolution and the second time resolution of the meteorological model; In the vertical direction, the σ terrain following coordinate coefficient is used to set the vertical resolution, forming two nested areas with different resolutions of the meteorological model.

Images