Short-term imminent forecasting method based on multi-method fusion radar reflectivity of topographic correction

Abstract

The invention relates to the technical field of short-term near weather forecasting, and discloses a short-term near weather forecasting method based on multi-method fusion radar reflectivity of topographic correction. The method comprises the following steps: firstly, acquiring radar reflectivity echo data and target area topographic data; and then processing is carried out through four different technical paths: 1, an echo extrapolation forecast field is generated by adopting a traditional motion vector extrapolation technology; secondly, a machine learning echo forecast field is obtained by applying a machine learning model; 3, forming an ensemble disturbance echo forecast field by utilizing ensemble disturbance forecast; and 4, correcting the radar echo based on the topographic data to obtain a topographic correction echo forecast field. And finally, according to a preset target weight coefficient, carrying out weighted fusion on the four forecasting fields to obtain a fused echo forecasting field, and making short and imminent forecasting of the target area in 0-3 hours and 3-6 hours in the future based on the fused echo forecasting field. According to the method, the accuracy, timeliness and objectification level of mountain area short-term and temporary weather forecast are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of short-term nowcasting technology, specifically to a short-term forecasting method for complex terrain areas that uses multi-method fusion of radar reflectivity based on terrain correction. Background Technology

[0002] Short-term nowcasting (SMF) is crucial for disaster prevention and mitigation, tourism safety, power supply, and reservoir management. Currently, SMF typically employs traditional extrapolation methods, such as the Thunderstorm Identification, Tracking, Analysis, and Nowcasting (TITAN) algorithm. This algorithm is based on radar observations and uses radar echo data to monitor and predict severe convective weather. However, in areas like Huangshan with dramatic topographic relief and distinct local microclimates, the complex vertical upward movement and horizontal convergence and divergence of airflow caused by the terrain can severely interfere with local dual-polarization radar echoes, leading to significant signal distortion and making it difficult to accurately identify and track severe convective systems. This makes it challenging for traditional extrapolation methods to accurately extrapolate storm paths and intensity changes, resulting in significant deviations in storm path and intensity predictions and reduced forecast accuracy. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a short-term forecasting method based on terrain correction and multi-method fusion of radar reflectivity, which can solve the problem of low accuracy of short-term forecasts under complex terrain and ensure timeliness.

[0004] The specific technical solutions of this invention are as follows: This application provides a short-term forecasting method based on terrain correction and multi-method fusion of radar reflectivity, the method comprising: Acquire radar echo data and high-precision terrain data of the target area; Based on the radar echo data, the radar echo is extrapolated using a traditional motion vector extrapolation algorithm to obtain the echo extrapolation prediction field; The fusion weights are determined based on the historical echo sequences, and radar echo extrapolation is performed based on the fusion weights, the radar echo data, and the preset physical attenuation rate to obtain the machine learning echo prediction field. Based on the radar echo data and the preset disturbance coefficient, multiple disturbance members are generated, and the multiple disturbance members are ensemble averaged to obtain the ensemble disturbance echo prediction field. Based on the terrain data and the preset correction strategy, the radar echo data is dynamically corrected to obtain the terrain-corrected echo prediction field. Based on preset target weights, the echo extrapolation forecast field, the machine learning echo forecast field, the ensemble perturbation echo forecast field, and the terrain correction echo forecast field are weighted and fused to obtain a fused echo forecast field. Based on the fused echo forecast field, short-term forecasts for the target area in the next 0-3 hours and 3-6 hours are determined, wherein the short-term forecasts include short-term heavy precipitation, hail, and thunderstorms.

[0005] In some embodiments, the step of dynamically correcting the radar echo data based on the terrain data and a preset correction strategy to obtain a terrain-corrected echo prediction field includes: Based on the terrain data, the corresponding correction parameters are determined from the correction strategy; The radar echo data is multiplied by the correction parameters to obtain the corrected radar echo data, and the terrain-corrected echo prediction field is constructed based on the corrected radar echo data.

[0006] In some embodiments, the terrain data includes terrain height; the correction parameters include a first parameter; and the correction strategy includes at least: When the terrain height is greater than the first preset height, the first parameter is a first value; When the terrain height is less than the first preset height but greater than the second preset height, the first parameter is the second value; When the terrain height is less than the second preset height, the first parameter is a third value; the first value is greater than the second value, and the second value is greater than the third value.

[0007] In some embodiments, the terrain data includes terrain roughness; the correction parameter includes a second parameter; and the correction strategy includes at least: When the terrain roughness is greater than the first preset roughness, the second parameter is the fourth value; When the terrain roughness is less than the first preset roughness and greater than the second preset roughness, the second parameter is the fifth value; When the terrain roughness is less than the second preset roughness, the second parameter is the sixth value; the fourth value is less than the fifth value, and the fifth value is less than the sixth value.

[0008] In some embodiments, the radar echo data includes at least a first radar echo and a second radar echo, wherein the first radar echo and the second radar echo are at different times; the conventional motion vector extrapolation algorithm is a radar echo correlation tracking algorithm. The step of extrapolating radar echoes based on the radar echo data using a traditional motion vector extrapolation algorithm to obtain the echo extrapolation prediction field includes: Based on the first radar echo and the second radar echo, the motion vector of the radar echo data is determined by the radar echo correlation tracking algorithm. Based on the motion vector and the preset attenuation coefficient, the echo extrapolation prediction field is generated.

[0009] In some embodiments, the echo extrapolation prediction field is obtained using the following formula: in, This indicates the position of the radar echo at a future time t+Δt. This indicates the position of the radar echo at the current time t. Represents the motion vector of the radar echo. Indicates the time step of the forecast. This represents the reflectivity intensity of the radar echo at a future time t+Δt. This represents the reflectivity intensity of the radar echo at the current time t. Indicates the attenuation coefficient. This indicates the centroid position of the radar echo at the current time t. This indicates the centroid position of the radar echo at the previous time t-Δt.

[0010] In some embodiments, the machine learning echo prediction field is obtained using the following formula: in, This indicates the position of the radar echo at a future time t+Δt. This indicates the position of the radar echo at the current time t. This represents the east-west component of the radar echo's motion vector. This represents the north-south component of the radar echo's motion vector. This represents the fusion weight of the components of the motion vector in the east-west direction. This represents the fusion weight of the components of the motion vector in the north-south direction. This represents the reflectivity intensity of the radar echo at a future time t+Δt. This represents the reflectivity intensity of the radar echo at the current time t. Indicates the physical attenuation rate. Indicates the time step of the forecast. This represents the time normalization factor.

[0011] In some embodiments, each perturbation member is generated using the following formula: in, This represents the longitude of the k-th perturbation member. This represents the dimension of the k-th perturbation member. Indicates the longitude of the target region. This indicates the latitude of the target region. The disturbance coefficient represents the longitude of the k-th disturbance member. This represents the perturbation coefficient of the k-th perturbation member at the latitude. Indicates the forecast time interval. Indicates the observation time interval. This represents the radar echo of the k-th disturbance member. This refers to the radar echo data. This represents the disturbance coefficient of the radar echo of the k-th disturbance member.

[0012] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, radar echo data and topographic data of the target area are first acquired; then, based on the radar echo data, radar echo extrapolation is performed using a traditional extrapolation algorithm to obtain an echo extrapolation forecast field; fusion weights are determined based on historical echo sequences, and radar echo extrapolation is performed based on the fusion weights, radar echo data, and a preset physical attenuation rate to obtain an echo forecast field; multiple perturbation members are generated based on the radar echo data and preset perturbation coefficients, and the multiple perturbation forecast members are ensemble averaged to obtain an ensemble perturbation echo forecast field; the radar echo data is corrected based on topographic data and a preset correction strategy to obtain a corrected echo field; finally, based on preset target weights, the echo extrapolation forecast field, the echo forecast field, the ensemble perturbation echo forecast field, and the corrected echo field are weighted and fused to obtain a fused echo forecast field, and based on the fused echo forecast field, short-term forecasts for the target area for the next 0-3 hours and 3-6 hours are determined. Thus, the embodiments of this application can compensate for or suppress radar echo distortion caused by terrain through dynamic terrain correction; in addition, by integrating four methods—traditional extrapolation method, ensemble forecast method, machine learning method and terrain correction method—the advantages can be complemented to improve the accuracy of severe convective weather identification and tracking. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a flowchart illustrating the short-term forecasting method based on terrain correction and multi-method fusion of radar reflectivity provided in an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention. 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.

[0015] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0016] It should be noted that the terms "first, second, and third" used in the embodiments of the present invention are only used to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.

[0017] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which these embodiments of the invention pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0018] In the Huangshan area, the complex vertical upward motion and horizontal convergence and divergence of airflow caused by the terrain make it difficult for traditional extrapolation methods to accurately extrapolate the movement path and intensity changes of storms. Machine learning methods lack sufficient training data specific to the topographic features of Huangshan, resulting in poor forecast accuracy and difficulty in accurately capturing the complex relationship between topography and severe convective weather. Ensemble forecasts also present significant uncertainties regarding the development and evolution of small- and medium-scale systems under the complex terrain of Huangshan. Since 2025, the Anhui Provincial Meteorological Bureau has been conducting short-term nowcasting operational pilot projects in Huangshan City, focusing on research and operational practice on the difficulties of short-term forecasting of two types of hazardous weather: heavy rain and severe convective weather. However, the complex topography of Huangshan causes severe interference to local dual-polarization radar echoes, resulting in significant signal distortion and making it difficult to accurately identify and track severe convective systems.

[0019] Figure 1This is a flowchart illustrating a short-term forecasting method based on terrain correction and multi-method fusion of radar reflectivity, provided in an embodiment of the present invention. The method can be executed via a control device, which may include at least one of a personal computer, laptop computer, smartphone, tablet computer, and portable wearable device; however, this embodiment does not limit the specific device used.

[0020] like Figure 1 As shown, the short-term forecasting method based on terrain correction and multi-method fusion of radar reflectivity provided in this embodiment of the invention may include steps S101-S106.

[0021] S101. Acquire radar echo data and terrain data of the target area.

[0022] In some embodiments, the radar device emits radar waves and receives the reflected radar echoes, and then sends the radar echoes to the control device. The control device receives the radar echoes emitted by the radar device to obtain radar echo data. The control device can also obtain topographic data of the target area (such as the Huangshan area) from a Geographic Information System (GIS) database. The topographic data includes at least the topographic height, topographic roughness, and topographic slope of the target area.

[0023] For example, the control device acquires combined reflectivity data from the national weather radar mosaic V3.0 network at a time resolution of 6 minutes to obtain radar echo data. Simultaneously, it obtains 500-meter resolution digital elevation model data of the target area from a geographic information system database, and extracts terrain parameters such as terrain height and terrain roughness from the 25-meter resolution digital elevation model data of the target area.

[0024] S102. Based on radar echo data, radar echo extrapolation is performed using a traditional motion vector extrapolation algorithm to obtain the echo extrapolation prediction field.

[0025] In some embodiments, the radar echo data includes at least a first radar echo and a second radar echo, wherein the first radar echo and the second radar echo are different in time.

[0026] For example, the first radar echo and the second radar echo are the radar echoes from the two most recent times. In some application scenarios, the first radar echo and the second radar echo can be spaced 6 minutes apart. This application does not limit the time interval of the first radar echo and the second radar echo.

[0027] In some embodiments, the conventional motion vector extrapolation algorithm is the Tracking Radar Echoes by Correlation (TREC) algorithm. Based on radar echo data, the conventional motion vector extrapolation algorithm is used to extrapolate radar echoes to obtain the echo extrapolation prediction field, which includes: determining the motion vector of the radar echo data based on the first radar echo and the second radar echo using a cross-correlation algorithm; and generating the echo extrapolation prediction field based on the motion vector and a preset attenuation coefficient.

[0028] For example, the control device can take the radar echoes from the two most recent times (6 minutes apart) (i.e., the first radar echo and the second radar echo), calculate the motion vector field of the radar echoes using the TREC algorithm, then extrapolate the echo positions along the direction of the motion vector, and introduce an attenuation coefficient to simulate intensity reduction, ultimately generating an echo extrapolation prediction field. The echo extrapolation prediction field is used to represent the radar echo distribution of the target area in the future for a period of time after the current time (such as the next 0-3 hours and the next 3-6 hours).

[0029] It should be noted that the attenuation coefficient is a preset value, typically between 0.8 and 1.2. In practical applications, it can be set according to actual needs, and this embodiment does not limit this.

[0030] In some embodiments, the echo extrapolation prediction field is obtained using the following formula 1: (Formula 1) in, This indicates the position of the radar echo at the future time t+Δt. This indicates the position of the radar echo at the current time t. Represents the motion vector of the radar echo. Indicates the time step of the forecast. This represents the reflectivity intensity of the radar echo at a future time t+Δt. This represents the reflectivity intensity of the radar echo at the current time t. Indicates the attenuation coefficient. This indicates the centroid position of the radar echo at the current time t. This indicates the centroid position of the radar echo at the previous time t-Δt.

[0031] S103. Determine the fusion weights based on the historical echo sequence, and extrapolate the radar echoes based on the fusion weights, radar echo data, and preset physical attenuation rates to obtain the machine learning echo prediction field.

[0032] In some embodiments, a historical echo sequence refers to radar echo data from multiple past time intervals. Fusion weights are used to represent the importance of the motion vector components in different directions (e.g., east-west and north-south).

[0033] For example, the fusion weights can be set according to empirical values ​​based on historical echo sequences; alternatively, the historical echo sequences can be input into a trained neural convolutional network, which learns the temporal evolution patterns and spatial distribution characteristics of the historical echo sequences to determine and output the fusion weights. The neural convolutional network can employ structures such as U-Net or ResNet, and this embodiment of the application does not limit this approach.

[0034] In some embodiments, the physical attenuation rate is a preset value used to represent the rate attenuation of radar echo intensity during propagation due to atmospheric absorption, scattering, and other factors.

[0035] For example, fusion weights can be obtained from historical radar echo sequences (data from at least one year) using a trained neural convolutional network. , The control device then uses fusion weights. , Radar echo data, combined with a preset physical attenuation rate β3, is used to predict the intensity and obtain a machine learning echo prediction field.

[0036] In some embodiments, the machine learning echo prediction field is obtained using the following formula 2: (Formula 2) in, This indicates the position of the radar echo at the future time t+Δt. This indicates the position of the radar echo at the current time t. This represents the east-west component of the radar echo's motion vector. This represents the north-south component of the radar echo's motion vector. This represents the fusion weight of the components of the motion vector in the east-west direction. This represents the fusion weight of the components of the motion vector in the north-south direction. This represents the reflectivity intensity of the radar echo at a future time t+Δt. This represents the reflectivity intensity of the radar echo at the current time t. Indicates the physical attenuation rate. Indicates the time step of the forecast. This represents the time normalization factor, T = 360 min. In some application scenarios, It can be 0.5. It can be 0.3. It can be 0.2.

[0037] S104. Based on radar echo data and preset disturbance coefficients, generate multiple disturbance members, and perform ensemble averaging on the multiple disturbance forecast members to obtain the ensemble disturbance echo forecast field.

[0038] In some embodiments, different perturbation coefficients are applied to the motion vector, position, or echo intensity to represent different weather system evolution assumptions (such as eastward deceleration, system development, path shift, etc.) and to construct different perturbation members.

[0039] For example, multiple (e.g., 5) perturbation members can be generated according to the setup instructions for each perturbation member shown in Table 1.

[0040] Table 1. Setting Instructions for Each Disturbance Member In this table, member (K) represents the member number, member 1 is the control experiment, and members 2-5 apply eastward / northward velocity perturbations, intensity perturbations, and path-turning perturbations, respectively, and are assigned different weights. The adjustment in Table 1 represents the adjustment of the motion vector V. x and V y The perturbation method; for example: V represents the motion vector without perturbation; 0.8V x 1.1V y This means reducing the east-west component of the motion vector by 20% and increasing the north-south component by 10%; 1.2V x 0.9V y This means increasing the east-west component of the motion vector by 20% and decreasing the north-south component by 10%; 1.1V x 0.8V y This indicates that the east-west component of the motion vector will be increased by 10%, and the north-south component will be decreased by 20%. The weights represent the weights of the corresponding members in the ensemble average; for example, member 2 has a weight of 0.90, and member 3 has a weight of 0.95. The physical meaning is used to explain the weather system behavior assumptions corresponding to this disturbance; for example, "slowing eastward, deflecting northward" means that the radar echo moves more slowly eastward and deflects northward; "system development" means that the radar echo intensity increases; "northeastward turn, rapid" means that the radar echo path turns northeastward and moves faster.

[0041] In some embodiments, each perturbation member is generated using the following formula 3: (Formula 3) in, This represents the longitude of the k-th perturbation member. This represents the dimension of the k-th perturbation member. Indicates the longitude of the target region. This indicates the latitude of the target region. The disturbance coefficient represents the longitude of the k-th disturbance member. This represents the perturbation coefficient of the k-th perturbation member at the latitude. Indicates the forecast time interval. Indicates the observation time interval. This represents the radar echo of the k-th disturbance member. This refers to the radar echo data. This represents the disturbance coefficient of the radar echo of the k-th disturbance member.

[0042] S105. Based on terrain data and a preset correction strategy, the radar echo data is corrected to obtain the terrain-corrected echo prediction field.

[0043] In some embodiments, correcting radar echo data based on terrain data and a preset correction strategy to obtain a terrain-corrected echo forecast field includes: determining corresponding correction parameters from the correction strategy based on terrain data; multiplying the radar echo data with the correction parameters to obtain corrected radar echo data; and constructing a terrain-corrected echo forecast field based on the corrected radar echo data.

[0044] For example, the control device queries the correction parameters corresponding to the terrain data from the correction strategy, then multiplies the radar echo data with the correction parameters to obtain the corrected radar echo data, and constructs a terrain-corrected echo prediction field based on the corrected radar echo data.

[0045] In some embodiments, terrain data may include terrain height, in which case the correction parameter includes a first parameter; the correction strategy includes at least: when the terrain height is greater than a first preset height, the first parameter is a first value; when the terrain height is less than the first preset height but greater than a second preset height, the first parameter is a second value; and when the terrain height is less than the second preset height, the first parameter is a third value. Wherein, the first value is greater than the second value, and the second value is greater than the third value; that is, the higher the terrain height, the larger the correction parameter, in order to compensate for terrain occlusion and simulate the terrain lifting enhancement effect.

[0046] For example, when the terrain height is greater than 500m (first preset height), the first parameter is 1.3 (first value); when the terrain height is less than 500m but greater than 100m (second preset height), the first parameter is 1.0 (second value); and when the terrain height is less than 100m, the first parameter is 0.95 (third value). In this case, the terrain-corrected echo prediction field can be obtained using the following formula 4: (Formula 4) in, This indicates the topographically corrected echo prediction field. Represents radar echo data, Indicates terrain elevation.

[0047] In some embodiments, the terrain data may include terrain roughness, in which case the correction parameter includes a second parameter; the correction strategy includes at least: when the terrain roughness is greater than a first preset roughness, the second parameter is a fourth value; when the terrain roughness is less than the first preset roughness but greater than a second preset roughness, the second parameter is a fifth value; and when the terrain roughness is less than the second preset roughness, the second parameter is a sixth value. Wherein, the fourth value is less than the fifth value, and the fifth value is less than the sixth value, meaning that the higher the terrain roughness, the smaller the correction parameter, used to suppress signal distortion caused by beam scattering and terrain clutter.

[0048] In some embodiments, terrain data may include only terrain height, only terrain roughness, or both terrain height and terrain roughness. This application does not limit this aspect.

[0049] For example, when the terrain data includes both terrain height and terrain roughness, the correction parameter corresponding to the terrain height can be multiplied by the correction parameter corresponding to the terrain roughness to obtain the final correction parameter. That is, the terrain-corrected echo prediction field is determined using the following formula: . This indicates the topographically corrected echo prediction field. Represents radar echo data, This indicates the correction parameters corresponding to the terrain height. This represents the correction parameter corresponding to the terrain roughness.

[0050] It is understood that the embodiments of this application use terrain data to dynamically correct radar echoes. For example, the radar echoes can be dynamically modulated according to the terrain height (i.e., terrain height) to compensate for terrain occlusion or simulate the terrain uplift enhancement effect. In this way, the influence of terrain occlusion and distortion on radar echoes can be reduced, data distortion can be avoided, and the accuracy of forecasts can be improved.

[0051] S106. Based on the preset target weights, the echo extrapolation forecast field, the machine learning echo forecast field, the ensemble perturbation echo forecast field, and the terrain correction echo forecast field are weighted and fused to obtain the fused echo forecast field. Based on the fused echo forecast field, the short-term forecasts for the target area in the next 0-3 hours and 3-6 hours are determined.

[0052] In some embodiments, the target weights include the weights corresponding to the echo extrapolation prediction field, the machine learning echo prediction field, the ensemble perturbation echo prediction field, and the terrain-corrected echo prediction field, respectively. The target weights are preset values ​​and can be set according to actual needs; this application embodiment does not limit this setting.

[0053] In some application scenarios, the original weights of the echo extrapolation prediction field, the machine learning echo prediction field, the ensemble perturbation echo prediction field, and the terrain-corrected echo prediction field can be set according to actual needs. Then, the original weights are normalized using the following formula 5 to obtain the target weights: (Formula 5) in, This represents the normalized weights, i.e., the target weights. This represents the i-th original weight. This represents the sum of the original weights of the echo extrapolation prediction field, the machine learning echo prediction field, the ensemble perturbation echo prediction field, and the terrain-corrected echo prediction field. The original weights of the echo extrapolation forecast field, the machine learning echo forecast field, the ensemble perturbation echo forecast field, and the terrain correction echo forecast field are represented. The original weight of the echo extrapolation forecast field is 0.5, the original weight of the machine learning echo forecast field is 0.3, the original weight of the ensemble perturbation echo forecast field is 0.1, and the original weight of the terrain correction echo forecast field is 0.1.

[0054] After obtaining the target weights, the fused echo prediction field can be obtained using the following formula 6: (Formula 6) in, To integrate the echo prediction field, For the target weight, For the i-th data, It includes echo extrapolation prediction fields, machine learning echo prediction fields, ensemble perturbation echo prediction fields, and topographically corrected echo prediction fields.

[0055] In some embodiments, short-term forecasts include short-term heavy precipitation, hail, and thunderstorms, i.e., short-term forecasts include precipitation forecasts and forecasts of the areas affected by severe convective weather.

[0056] For example, after obtaining the fused echo forecast field, for precipitation forecasting, the reflectivity data in the fused echo forecast field can be converted into precipitation intensity based on the relationship between radar reflectivity factor and rainfall intensity (also known as the ZR relationship), and cumulative precipitation forecast fields for the next 0-3 hours and 3-6 hours can be generated by time accumulation. For severe convective weather location forecasting, areas with combined reflectivity greater than or equal to a preset reflectivity intensity threshold (e.g., 45 dBz) in the fused echo forecast field can be identified as severe convective potential areas. Then, the TREC algorithm or optical flow method can be used to track and analyze the evolution of these severe convective potential areas, thereby determining the location range of severe convective weather in the next 0-3 hours and 3-6 hours, and generating location forecasts for severe convective weather such as thunderstorms, strong winds, short-duration heavy precipitation, and hail.

[0057] In some embodiments, a timing module can be set within the control device. The timing module can be triggered at a preset time, and when the timing module is triggered, the control device will execute S101-S106 as described above. Specifically, after obtaining the short-term forecast, the short-term forecast can be automatically generated in the form of a Word report, a webpage written in Hyper Text Markup Language (HTML), etc., and distributed via an intranet or the Internet (e.g., GitHub Pages). The layout of the short-term forecast can use an administrative division map as the base map, displaying forecast information such as time-series precipitation and areas of severe convection, and indicating the issuing unit and time.

[0058] It is understood that the embodiments of this application construct short-term forecasts of 0-3 hours and 3-6 hours by fusing echo extrapolation forecast fields, machine learning echo forecast fields, ensemble perturbation echo forecast fields and terrain correction echo forecast fields. This can combine the advantages of the four methods of traditional extrapolation method, ensemble forecast method, machine learning method and terrain correction method, and further improve the accuracy and reliability of the forecast.

[0059] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, the phrase "in one embodiment" or "in an embodiment" appearing in every place throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in each embodiment of the invention, the sequence number of each process described above does not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the invention. The sequence numbers of the embodiments of the invention described above are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0060] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0061] In the several embodiments provided by this invention, it should be understood that the disclosed methods can be implemented in other ways. The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments. The features disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0062] The above description is merely an embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A short-term forecasting method based on terrain correction and multi-method fusion of radar reflectivity, characterized in that, The method includes: Acquire radar echo data and high-precision terrain data of the target area; Based on the radar echo data, the radar echo is extrapolated using a traditional motion vector extrapolation algorithm to obtain the echo extrapolation prediction field. The fusion weights are determined based on the historical echo sequences, and radar echo extrapolation is performed based on the fusion weights, the radar echo data, and the preset physical attenuation rate to obtain the machine learning echo prediction field. Based on the radar echo data and the preset disturbance coefficient, multiple disturbance members are generated, and the multiple disturbance members are ensemble averaged to obtain the ensemble disturbance echo prediction field. Based on the terrain data and the preset correction strategy, the radar echo data is dynamically corrected to obtain the terrain-corrected echo prediction field. Based on preset target weights, the echo extrapolation forecast field, the machine learning echo forecast field, the ensemble perturbation echo forecast field, and the terrain correction echo forecast field are weighted and fused to obtain a fused echo forecast field. Based on the fused echo forecast field, short-term forecasts for the target area in the next 0-3 hours and 3-6 hours are determined, wherein the short-term forecasts include short-term heavy precipitation, hail, and thunderstorms.

2. The method according to claim 1, characterized in that, The step of dynamically correcting the radar echo data based on the terrain data and a preset correction strategy to obtain a terrain-corrected echo prediction field includes: Based on the terrain data, the corresponding correction parameters are determined from the correction strategy; The radar echo data is multiplied by the correction parameters to obtain the corrected radar echo data, and the terrain-corrected echo prediction field is constructed based on the corrected radar echo data.

3. The method according to claim 2, characterized in that, The terrain data includes terrain height; the correction parameters include a first parameter; the correction strategy includes at least: When the terrain height is greater than the first preset height, the first parameter is a first value; When the terrain height is less than the first preset height but greater than the second preset height, the first parameter is the second value; When the terrain height is less than the second preset height, the first parameter is a third value; the first value is greater than the second value, and the second value is greater than the third value.

4. The method according to claim 2 or 3, characterized in that, The terrain data includes terrain roughness; the correction parameters include a second parameter; the correction strategy includes at least: When the terrain roughness is greater than the first preset roughness, the second parameter is the fourth value; When the terrain roughness is less than the first preset roughness and greater than the second preset roughness, the second parameter is the fifth value; When the terrain roughness is less than the second preset roughness, the second parameter is the sixth value; the fourth value is less than the fifth value, and the fifth value is less than the sixth value.

5. The method according to claim 1, characterized in that, The radar echo data includes at least a first radar echo and a second radar echo, wherein the first radar echo and the second radar echo are at different times; the conventional motion vector extrapolation algorithm is a radar echo correlation tracking algorithm. The step of extrapolating radar echoes based on the radar echo data using a traditional motion vector extrapolation algorithm to obtain the echo extrapolation prediction field includes: Based on the first radar echo and the second radar echo, the motion vector of the radar echo data is determined by the radar echo correlation tracking algorithm. Based on the motion vector and the preset attenuation coefficient, the echo extrapolation prediction field is generated.

6. The method according to claim 1 or 5, characterized in that, The echo extrapolation prediction field is obtained using the following formula: in, This indicates the position of the radar echo at a future time t+Δt. This indicates the position of the radar echo at the current time t. Represents the motion vector of the radar echo. Indicates the time step of the forecast. This represents the reflectivity intensity of the radar echo at a future time t+Δt. This represents the reflectivity intensity of the radar echo at the current time t. Indicates the attenuation coefficient. This indicates the centroid position of the radar echo at the current time t. This indicates the centroid position of the radar echo at the previous time t-Δt.

7. The method according to claim 1, characterized in that, The machine learning echo prediction field is obtained using the following formula: in, This indicates the position of the radar echo at a future time t+Δt. This indicates the position of the radar echo at the current time t. This represents the east-west component of the radar echo's motion vector. This represents the north-south component of the radar echo's motion vector. This represents the fusion weight of the components of the motion vector in the east-west direction. This represents the fusion weight of the components of the motion vector in the north-south direction. This represents the reflectivity intensity of the radar echo at a future time t+Δt. This represents the reflectivity intensity of the radar echo at the current time t. Indicates the physical attenuation rate. Indicates the time step of the forecast. This represents the time normalization factor.

8. The method according to claim 1, characterized in that, Each perturbation member is generated using the following formula: in, This represents the longitude of the k-th perturbation member. This represents the dimension of the k-th perturbation member. Indicates the longitude of the target region. This indicates the latitude of the target region. The disturbance coefficient represents the longitude of the k-th disturbance member. This represents the perturbation coefficient of the k-th perturbation member at the latitude. Indicates the forecast time interval. Indicates the observation time interval. This represents the radar echo of the k-th disturbance member. This refers to the radar echo data. This represents the disturbance coefficient of the radar echo of the k-th disturbance member.

Images