Three-dimensional grid method for weather radar based on layer data iterative updating

Abstract

The application discloses a weather radar three-dimensional grid point method based on layer data iterative updating, and relates to the technical field of meteorological radar data processing. The method comprises the following steps: acquiring radial data blocks of each weather radar at a current sampling moment; mapping each radial data block to a preset three-dimensional grid to obtain a grid point data block, and calculating a data quality score of each grid point; loading each grid point data block into a data queue, and eliminating old data blocks with time attributes exceeding a preset time range to obtain an updated data queue; splicing and fusing all grid point data blocks in the updated data queue in a preset three-dimensional grid space to generate a three-dimensional grid point data body at the current sampling moment, and entering a next sampling moment. The method can realize fast updating of weather radar three-dimensional grid points by using layer data, and improves the timeliness, data quality and system robustness of data updating.

Description

Technical Field

[0001] This application relates to the field of meteorological radar data processing technology, and in particular to a three-dimensional gridding method for weather radar based on layer data iterative updates. Background Technology

[0002] Weather radar is an important means of weather monitoring. It detects particles by actively emitting electromagnetic waves and receiving the scattered waves. During detection, weather radar uses servo control to achieve detection at different azimuth and elevation angles. With social development, advancements in meteorological technology, and increasing demands for forecasting services, the requirements for severe convective weather forecasting and early warning capabilities are becoming increasingly stringent. Furthermore, nowcasting relies heavily on the timeliness of weather radar data. Therefore, shortening the processing time of weather radar data and enabling rapid visualization of three-dimensional results is crucial for weather forecasting and early warning.

[0003] However, existing weather radar scan intervals are six minutes. Weather radar data for a given area typically reaches the data processing center after a single volume scan, requiring further processing to obtain the three-dimensional gridded data volume. This results in an update cycle for the three-dimensional results exceeding six minutes, compromising the integrity of the weather radar data and making it difficult to detect short-lived severe convective weather systems. Summary of the Invention

[0004] To address the aforementioned problems and technical requirements, this application proposes a three-dimensional gridding method for weather radar based on iterative updates of layer data. The technical solution of this application is as follows: A method for three-dimensional gridding of weather radar based on iterative updating of layer data includes the following steps: Obtain radial data blocks for each weather radar at the current sampling time; each radial data block contains time attributes. Each radial data block is mapped onto a preset 3D grid to obtain gridded data blocks. The radial distance between each grid point on the preset 3D grid and the corresponding weather radar is combined to calculate the data quality score for each grid point. The data quality score indicates the data quality at the grid point; the higher the data quality score, the better the data quality at the grid point. Each gridded data block is loaded into the data queue, and old data blocks whose time attributes exceed the preset time range are removed based on the time attributes of the existing gridded data blocks in the data queue, resulting in an updated data queue. Based on the data quality scores of each grid point in each gridded data block, all gridded data blocks in the updated data queue are spliced ​​and merged in a preset three-dimensional grid space to generate a three-dimensional gridded data volume at the current sampling time, and then proceed to the next sampling time.

[0005] A further technical solution is that the preset time range is a dynamic time window, and determining the length T of the dynamic time window includes: Real-time acquisition of meteorological data, which is based on raw radial data blocks obtained from weather radar scans or other meteorological monitoring equipment; The Convection Intensity Index (CI) is calculated based on meteorological data. The CI indicates the degree of strong convection in the weather. The larger the CI, the stronger the convection in the weather. The dynamic time window length T is determined based on the convection intensity index CI.

[0006] The further technical solution is that the meteorological data includes the maximum reflectivity of the radar combination, the area proportion of the region where the maximum reflectivity exceeds the preset reflectivity threshold, the area proportion of the region where the echo top height exceeds the preset height threshold, the maximum vertical integral liquid water content, the mesocyclone detection results, and the lightning frequency. The fuzzy logic comprehensive index method is used to calculate the convection intensity index (CI) for each meteorological data.

[0007] A further technical solution involves determining the dynamic time window length T based on the convection intensity index CI, including: When the convection intensity index CI is lower than the first threshold, the dynamic time window length T is determined to be the preset maximum duration; when the convection intensity index CI is higher than the second threshold, the dynamic time window length T is determined to be the preset minimum duration; when the convection intensity index CI is not lower than the first threshold and not higher than the second threshold, the dynamic time window length T is calculated by substituting the convection intensity index CI into the mapping function. The mapping function represents that the dynamic time window length T decreases as the convection intensity index CI increases.

[0008] A further technical solution is that the gridded data blocks include confidence levels. The gridded data block obtained from the current sampling of any weather radar includes: The arrival time of the radial data blocks generated by the weather radar at the current sampling time is monitored; If no radial data block is received within a preset timeout threshold, it is determined that the original radial data block generated by the weather radar at the current sampling time is missing. An extrapolation algorithm is then used to map the received gridded data blocks from the previous two sampling times onto a preset 3D grid, generating the gridded data block for the weather radar at the current sampling time. The confidence level of the gridded data block at the current sampling time is then determined. Assigning a value of 0< <1; otherwise, it is determined that the original radial data block generated by the weather radar at the current sampling time is not missing, and the original radial data block is mapped onto a preset three-dimensional grid using an interpolation algorithm to generate the gridded data block of the weather radar at the current sampling time, and the confidence level of the gridded data block at the current sampling time is set. Assigned value .

[0009] A further technical solution involves gridded data blocks that also include echo intensity and noise; the calculated data quality score Q for any grid point includes: Determine the signal-to-noise ratio of grid point i based on the echo intensity and noise of any grid point i in the gridded data block. ; Based on the radial distance between grid point i and the weather radar Determine the distance decay factor for grid point i. The distance attenuation factor reflects the impact of radial distance on data quality. Larger distance attenuation factor The smaller the value, the worse the data quality. The data quality score corresponding to grid point i in the gridded data block is calculated. , It is the signal-to-noise ratio normalization function. , It is a weighting coefficient and , It's the confidence level.

[0010] A further technical solution involves determining the distance attenuation factor for grid point i. , It is the characteristic distance constant.

[0011] A further technical solution involves, when gridded data blocks in the updated data queue overlap spatially, stitching and fusing all overlapping gridded data blocks in a preset three-dimensional grid space, including: The data quality scores corresponding to the grid points of each gridded data block corresponding to the overlapping grid point m are used as weights for weighted fusion to obtain the value of the overlapping grid point m. , It is the data quality score of the k-th gridded data block that overlaps spatially, at the overlapping grid point m. It is the value at the overlapping grid point m of the k-th gridded data block that overlaps in space, and N is the number of gridded data blocks that overlap in space.

[0012] A further technical solution is that the gridded data block also includes elevation angle information. The extrapolation algorithm adopts an optical flow-based extrapolation algorithm. When the original radial data block generated at the sampling time is missing, the gridded data block generated by the weather radar at the current sampling time includes: Based on the gridded data blocks of the weather radar at the previous two sampling times at the current sampling time, the motion vector fields of the echo in the vertical and horizontal directions are calculated using the optical flow method. Based on the motion vector field and the height difference between the current sampling time and the corresponding elevation angle of the previous two sampling times, the displacement of the echo structure is calculated; the gridded data blocks of the previous two sampling times are translated along the motion vector direction to the height position of the corresponding elevation angle of the current sampling time, and the gridded data blocks of the weather radar at the current sampling time are generated by interpolation.

[0013] The beneficial technical effects of this application are: This application discloses a three-dimensional gridding method for weather radar based on layer data iterative updates, solving the problems of existing technologies that require waiting for complete volume scan data before three-dimensional gridding and whose update cycle is limited by the volume scan cycle. This method iteratively updates data in units of radial data blocks acquired at a single sampling time, controlling the update cycle through a preset time range to ensure that an update is triggered every time a radial data block is received, effectively reducing end-to-end latency in data processing. It achieves rapid three-dimensional gridding with a layer scan time cycle, significantly improving the timeliness of data updates and gaining valuable time for nowcasting and early warning, which is particularly important for monitoring severe convective weather. Furthermore, fusing all gridded data blocks based on data quality scores reduces the interference of low-quality data on the final three-dimensional gridded data volume, improving data quality and system robustness.

[0014] A dynamic time window mechanism is employed to dynamically adjust the time window based on real-time meteorological data. During severe convective weather, the time window is automatically shortened to ensure rapid capture of the three-dimensional structural evolution of the convective system; during stable weather, the time window is extended to reduce computational resource consumption and achieve intelligent resource allocation. The development stage of the weather system is comprehensively determined through the convection intensity index, avoiding the limitations of fixed time windows in dealing with complex weather and improving the scientific rigor and adaptability of the data update strategy. While ensuring monitoring accuracy, unnecessary computation and storage overhead is effectively reduced, lowering system operating costs and improving overall system efficiency.

[0015] By monitoring the temporal attributes of raw radial data blocks acquired by weather radar, a data missing anomaly handling mechanism is introduced. When the radial data block at the current sampling time is delayed or lost, the missing data is reconstructed using radial data blocks from the two preceding sampling times, and a low-confidence flag is added. This flag automatically reduces the weight of the missing data during subsequent fusion processes, effectively controlling the impact of interpolation uncertainty on the final result and balancing data integrity and result accuracy. Furthermore, an extrapolation algorithm based on optical flow is used to rapidly reconstruct the missing data, avoiding data queue interruptions, ensuring the continuity and integrity of the 3D gridded data volume, enhancing the system's robustness in complex environments and under equipment failure conditions, ensuring the stable operation of the business system, and reducing monitoring blind spots and forecast errors caused by data missing data.

[0016] Considering the differences in detection range and signal strength among different radars, a data quality score is calculated by combining signal-to-noise ratio, range attenuation factor, and confidence value to quantify the data quality of different data sources. This ensures that physical laws and signal characteristics are fully considered in the scoring, enhancing system stability in complex scenarios. The data quality score is automatically weighted during data fusion, ensuring that high-quality data dominates the final 3D gridded data volume, effectively improving the accuracy and reliability of the data fusion results. This provides a scientific basis for weight allocation in multi-radar network data fusion, effectively solving the fusion challenges caused by differences in the quality of data from different weather radars, and improving the consistency and reliability of the 3D gridded results of networked radars. Attached Figure Description

[0017] Figure 1 This is a flowchart of the three-dimensional gridding method for weather radar.

[0018] Figure 2 This is a flowchart for acquiring radial data blocks from a weather radar at the current sampling time.

[0019] Figure 3 This is a schematic diagram of radial data block interpolation mapping in an example.

[0020] Figure 4 This is a schematic diagram of a data queue loading a new gridded data block in an instance.

[0021] Figure 5 This is a schematic diagram illustrating the removal of old data blocks within a preset time range in an example.

[0022] Figure 6 This is a flowchart for determining the dynamic time window length T.

[0023] Figure 7 This is a schematic diagram of a three-dimensional gridded data volume obtained by fusing all gridded data blocks of a data queue in an instance. Detailed Implementation

[0024] The specific embodiments of this application will be further described below with reference to the accompanying drawings.

[0025] This application discloses a three-dimensional gridding method for weather radar based on layer data iterative updates. Please refer to [reference needed]. Figure 1 The flowchart shown illustrates the specific steps of this method: Step 1: Obtain radial data blocks for each weather radar at the current sampling time. Each radial data block contains a time attribute and records the sampling time corresponding to the radial data block.

[0026] When observing important weather events, weather radar generates a binary radial data stream containing information such as time and station names after each layer of data scanning, based on user-defined parameters. The data format of the file adopts the standard format of the National Weather Service, containing radial data information such as station, time, and elevation angle. When the networked weather radar is a phased array weather radar, each layer of data is a radial data stream generated after the phased array weather radar completes a sector scan. Each sampling time corresponds to one sampling process of the weather radar. During each sampling, the weather radar completes one layer scan or the phased array weather radar completes one sector scan, thus generating a raw radial data block containing time attributes.

[0027] Step 2: Map each radial data block onto a preset 3D grid to obtain gridded data blocks. Combine the radial distance between each grid point on the preset 3D grid and the corresponding weather radar to calculate the data quality score for each grid point. The data quality score indicates the data quality at the grid point. The higher the data quality score, the better the data quality at the grid point.

[0028] The data points in each radial data block are distributed rather sparsely, and are not a regular grid structure, such as... Figure 3 As shown, it typically covers a fan-shaped area, as indicated by the diagonal lines in the figure. It should be noted that, for ease of observation, the figure shows a two-dimensional planar schematic diagram of the data point distribution ignoring the depth direction.

[0029] In actual operation, the transmission of weather radar data may experience issues such as network latency, equipment failure, etc., resulting in missing or delayed arrival of the original radial data block at a certain sampling moment. This affects subsequent data processing, causes data queue interruptions, and severely impacts the continuity and integrity of the final generated 3D gridded data volume. To avoid these problems, in one embodiment, the gridded data block includes confidence levels. A data missing anomaly handling mechanism is introduced, and a confidence level is set for each gridded data block. This is to record the data source of the gridded data, and the data source reflects the reliability of the gridded data block. For details, please refer to... Figure 2The flowchart shown illustrates how to obtain a gridded data block from any weather radar at the current sampling time, including: The arrival time of the radial data blocks generated by the weather radar at the current sampling time is monitored; If no radial data block is received within a preset timeout threshold, it is determined that the original radial data block generated by the weather radar at the current sampling time is missing. An extrapolation algorithm is then used to map the gridded data blocks from the previous two sampling times onto a preset 3D grid, generating the gridded data block for the weather radar at the current sampling time. The confidence level of the gridded data block at the current sampling time is then determined. Assigning a value of 0< <1, Typical values ​​are any values ​​between 0.5 and 0.8; otherwise, it is determined that the original radial data block generated by the weather radar at the current sampling time is not missing, and the original radial data block is mapped onto a preset three-dimensional grid using an interpolation algorithm to generate the gridded data block of the weather radar at the current sampling time, and the confidence level of the radial data block at the current sampling time is set. Assigned value The specific value of the preset timeout threshold can be customized according to the actual application, and is usually 1.5 to 2 times the sampling time required for the weather radar to complete one layer of scanning.

[0030] If the original radial data block generated at the current sampling time is not missing, an interpolation algorithm is needed to map the original radial data block onto a preset 3D grid to generate a gridded data block for the current sampling time. The data points in the original radial data block are relatively scattered and do not have a regular grid structure, such as... Figure 3 As shown, it typically covers a fan-shaped area with an angle, as indicated by the diagonal lines in the figure. It should be noted that, for ease of observation, the figure shows a two-dimensional planar schematic diagram of the data point distribution, ignoring the depth direction. Figure 3 Taking the radial data block in the fan-shaped angle region as an example, regular gridded data blocks are formed as shown in the blue filled area in the figure. The preset 3D grid size and shape, as well as the specific interpolation algorithm used, are customized according to actual application requirements. For example, in the horizontal direction, the nearest neighbor interpolation algorithm is used, directly finding and assigning the value of the nearest radar radial data point to each target grid point; in the vertical direction, the interpolation fills the corresponding grid points according to the actual beam coverage height of the radar transmission, while grid points at heights not actually covered by the beam remain undetected. It should be noted that "undetected" here refers to data points without any observation data.

[0031] When the original radial data block generated by the current sampling moment is missing, the gridded data blocks from the previous two sampling moments are extrapolated to generate a replacement data block, thus preventing the data queue from not being updated for a long time due to waiting for missing data.

[0032] In one embodiment, the gridded data block also contains elevation angle information. The extrapolation algorithm uses an optical flow-based extrapolation algorithm to generate the gridded data block of the weather radar at the current sampling time, including: (1) Based on the gridded data blocks of the weather radar at the two sampling times before the current sampling time, the motion vector fields of the echo in the vertical and horizontal directions are calculated using the optical flow method. The specific optical flow method can be selected according to the actual application requirements. For example, for the overall motion trend analysis of large-scale weather systems, the Horn-Schunck optical flow method is used, and a global smoothing constraint is introduced to obtain a stable motion vector estimate. For the rapid motion feature capture of small-scale convective cells, the Lucas-Kanade optical flow method is used, and a fine motion vector is obtained based on a local window. According to the scale characteristics of the actual weather process, an appropriate algorithm is selected to obtain the horizontal and vertical motion components of the echo structure at each grid point.

[0033] (2) Calculate the displacement of the echo structure based on the motion vector field and the height difference between the current sampling time and the corresponding elevation angles of the previous two sampling times.

[0034] (3) According to the displacement of the echo structure, the gridded data blocks of the two previous sampling times are translated along the direction of the motion vector to the height position of the corresponding elevation angle at the current sampling time, and the gridded data blocks of the weather radar at the current sampling time are generated by interpolation. The interpolation method can be selected according to the actual application requirements. This application adopts bilinear interpolation.

[0035] The optical flow-based extrapolation algorithm, based on physical constraints (the continuity of echo motion), can more accurately reconstruct the three-dimensional echo structure of missing layers compared to simple linear interpolation or directly replacing it with gridded data blocks from adjacent sampling times.

[0036] Taking an X-band phased array weather radar sector scanning scenario as an example: When the X-band phased array weather radar is scanning sectors at an elevation angle of 5°, the data for the third sector fails to arrive on time due to a momentary interruption in the communication link, exceeding the preset timeout threshold. After detecting the missing data, the system uses the normally received data layers at elevation angles of 1° and 3° to calculate the echo motion vector field using the Lucas-Kanade optical flow method, obtaining that the echo structure is developing upwards in the vertical direction at a speed of approximately 2.5 m / s. Based on the height difference (approximately 500 m), the displacement is calculated, and the echo structure of the 1° and 3° elevation layers is extrapolated to the corresponding height at an elevation angle of 5°. The gridded data block of the missing sector is reconstructed through bilinear interpolation, and a low-confidence marker is added. =0.6). This method maintains the continuity of the data queue and the integrity of the three-dimensional gridding while ensuring the reliability of subsequent fusion results through low-confidence identification.

[0037] For each gridded data block, a data quality score corresponding to each grid point also needs to be calculated. In one embodiment, each gridded data block also includes echo intensity and noise; the data quality score comprehensively considers the signal-to-noise ratio of the data source and distance attenuation factors. The data quality score Q corresponding to any grid point is calculated as follows: Determine the signal-to-noise ratio of grid point i based on the echo intensity and noise of any grid point i in the gridded data block. ; Based on the radial distance between grid point i and the weather radar Determine the distance decay factor for grid point i. The distance attenuation factor reflects the impact of radial distance on data quality. Larger distance attenuation factor The smaller the value, the worse the data quality. The data quality score corresponding to grid point i in the gridded data block is calculated. , It is the signal-to-noise ratio normalization function, which will Mapping to the [0,1] interval can be achieved using any normalization function of the prior art, which will not be elaborated in this application; , It is a weighting coefficient and Typical value , ; It's the confidence level. =1 indicates that the data source of the gridded data block is the raw radial data block generated by the actual scan of the weather radar, 0 < <1 indicates that the original radial data block is missing, and the data source of the gridded data block is the radial data block generated by interpolation.

[0038] Distance attenuation compensation factor The distance attenuation factor decreases with increasing distance, reflecting the reduced reliability of long-distance echoes due to beam widening and electromagnetic wave attenuation. In one embodiment, the distance attenuation factor for grid point i is determined. , It is the characteristic distance constant, with typical values ​​for: S-band =100km, C-band =60km, X-band =30km.

[0039] Step 3: Load each gridded data block into the data queue, and based on the time attributes of the existing gridded data blocks in the data queue, remove old data blocks whose time attributes exceed the preset time range to obtain the updated data queue.

[0040] The data queue is a time-ordered collection of gridded data blocks containing data from multiple layers within the current time window, corresponding to radial data blocks acquired at multiple sampling times. Each radial data block in the data queue records echo intensity values, along with time attributes, elevation angle information, confidence level, and data quality score. Rapidly updating the data queue involves both loading new data and removing old data. The gridded data blocks acquired at the current sampling time are loaded into the data queue. For any given gridded data block, loading it into the data queue is as follows: Figure 4 As shown in the figure, the original data queue contains 10 gridded data blocks, which are gridded data blocks from the same weather radar at 10 sampling times. After loading new gridded data blocks, the data queue contains 11 gridded data blocks, including network data from the previous sampling time and newly received radial data blocks. Whenever a new gridded data block is loaded, an old data removal operation is performed. The core criterion for old data removal is the time window; when the difference between its time attribute and the current sampling time exceeds a preset time range, it is deleted from the data queue.

[0041] In one embodiment, the preset time range is a fixed value, the specific value of which can be customized according to the actual application. Figure 4 The preset time range in the example is the length of time (6 minutes) for a typical weather radar to complete one volume scan, corresponding to 10 sampling times and containing 10 gridded data blocks.

[0042] When a new gridded data block 0 is enqueued, if the time of the old gridded data block 0 exceeds the preset time range, the old gridded data block 0 needs to be deleted from the data queue. Figure 5 As shown, the data in the data queue is iteratively updated as sampling continues.

[0043] In another embodiment, the preset time range is a dynamic time window. Using a dynamic time window can improve the timeliness of data updates and the efficiency of resource utilization. Please refer to... Figure 6 The flowchart shown illustrates the specific steps for determining the dynamic time window length T, including: (1) Real-time acquisition of meteorological data, which is based on raw radial data blocks obtained by weather radar scanning or other meteorological monitoring equipment; (2) The convection intensity index CI is calculated based on meteorological data. The convection intensity index CI indicates the degree of strong convection in the weather. The larger the convection intensity index CI, the higher the degree of strong convection in the weather.

[0044] The system dynamically adjusts the time window length based on real-time meteorological conditions. When the weather radar detects severe convective weather events (such as hail, tornadoes, downbursts, etc.), the system automatically shortens the time window, allowing the 3D gridded results to be updated more frequently, thus enabling timely capture of the rapidly evolving 3D structure of severe convective systems. When the weather system is in a stable state (such as large-scale stratiform cloud precipitation), the system automatically extends the window length to reduce computational resource consumption.

[0045] In one embodiment, the meteorological data includes the maximum combined radar reflectivity, the area percentage of regions with reflectivity exceeding a preset reflectivity threshold, the area percentage of regions with echo top height exceeding a preset height threshold, the maximum vertically integrated liquid water content, mesocyclone detection results, and lightning frequency. The maximum combined radar reflectivity (CR_max), the area percentage of regions with reflectivity exceeding the preset reflectivity threshold (S_CR), the area percentage of regions with echo top height exceeding the preset height threshold (S_ET), the maximum vertically integrated liquid water content (VIL_max), and the mesocyclone detection results are all derived from basic data such as reflectivity and radial velocity collected by radar volume scanning, calculated secondaryly using a proprietary algorithm, and corresponding to specific calculation products: the maximum value (dBZ) is read from the combined reflectivity product CR; the area percentage of regions with reflectivity exceeding the reflectivity threshold is read from the CR product; the area percentage of regions with echo top height exceeding the height threshold is read from the ET product; the maximum vertically integrated liquid water content (kg / m²) is read from the VIL product; and the I_MDA flag is read from the output of the mesocyclone detection algorithm MDA. The specific calculation process for each calculation product is detailed in existing technologies and will not be elaborated upon here. Lightning frequency (F_LTG) is usually provided by lightning location systems or lightning imagers. As an important supplementary indicator for judging the intensity of convection, it can avoid the limitations of radar data caused by scanning strategies or beam obstruction, complement radar data, and improve the early warning capability of severe convective weather.

[0046] The fuzzy logic comprehensive index method is used to calculate the convection intensity index (CI) for each meteorological data.

[0047] The Convection Intensity Index (CI) is a comprehensive indicator used in meteorological operations to quantitatively assess the strength of initial or developing convection. Currently, the most mature and widely used calculation method is the fuzzy logic comprehensive index method, often used as the core algorithm in operational systems such as WSR-88D, directly for calculating the convection index. This embodiment uses the fuzzy logic comprehensive index method to calculate the Convection Intensity Index (CI). The specific calculation formula is as follows: , , , , , , These are the weighting coefficients, with typical values ​​as follows: =0.25, =0.20, =0.15, =0.15, =0.15, =0.10; , , , , This is a normalization function used to map all the original parameters (such as 55 dBZ, 5 kg / m², 40%, etc.) with different physical meanings, units, and numerical ranges in the calculation formula to a unified interval, usually between 0 and 1, to ensure that all parameters can be fairly compared and calculated under the same rule system. The I_MDA (Mesocyclone Detection Algorithm) flag is set to 1 (or True) whenever a mesocyclone is detected, and 0 otherwise.

[0048] Normalization functions typically employ a trapezoidal membership function, which includes four key points: a) Lower bound: Below this value, the function output is 0 (extremely weak); b) Left shoulder: From this point, the function value linearly increases from 0; c) Right shoulder: At this point, the function value reaches 1 (very strong); d) Upper bound: Above this value, the function output remains 1 (extremely strong). Typical values ​​are shown in the table below: Table 1 Typical values ​​of the normalization function at key points

[0049] (3) Determine the dynamic time window length T based on the convection intensity index CI.

[0050] In one embodiment, determining the dynamic time window length T based on the convection intensity index CI includes: When the convective intensity index (CI) is below the first threshold, the weather system is in a stable state. The dynamic time window length T is set to the preset maximum duration to reduce computational resource consumption. When the convective intensity index (CI) is above the second threshold, it indicates the presence of severe convective weather events such as hail, tornadoes, and downbursts. The dynamic time window length T is set to the preset minimum duration to ensure that the three-dimensional gridded data is updated at the highest frequency. The specific values ​​of the convective intensity index threshold, the preset maximum duration, and the preset minimum duration are set based on actual engineering application experience. In this application, the CI value is normalized, and the first threshold is set to 0.2, the preset maximum duration is 6 minutes, the second threshold is set to 0.6, and the preset minimum duration is 1 minute. When the convection intensity index CI is not lower than a first threshold and not higher than a second threshold, the dynamic time window length T is calculated by substituting the convection intensity index CI into a mapping function. The mapping function represents that the dynamic time window length T decreases as the convection intensity index CI increases. The mapping function can be a linear decreasing function or a nonlinear function. Through analysis of historical data, this application found that the calculation accuracy of the linear decreasing function is slightly lower than that of the nonlinear function, but the calculation speed is higher. Therefore, the specific application can be designed according to actual needs. For example, to ensure real-time performance, a linear decreasing function is selected. The slope and intercept of the linear decreasing function are determined based on the convection intensity index threshold and the preset maximum and minimum durations.

[0051] Taking a hailstorm as an example: weather radar detected a combined reflectivity maximum of 65 dBZ, an echo top height exceeding 12 km, MDA algorithm identified mesocyclone characteristics, and lightning frequency exceeded 15 times per minute. At this time, the CI calculation value was 0.85, exceeding the second threshold, and the dynamic time window was automatically shortened from the usual 6 minutes to 1 minute. Within this hour, the number of 3D gridded data updates increased from 10 to 60, and the system successfully captured the continuous evolution of the three-dimensional structure of the hail cloud in its early, mature, and dissipation stages.

[0052] Step 4: Based on the data quality score of each grid point in each gridded data block, stitch and merge all gridded data blocks in the updated data queue in the preset three-dimensional grid space to generate the three-dimensional gridded data volume at the current sampling time, and proceed to the next sampling time.

[0053] When multiple gridded data blocks obtained from weather radars are stitched and fused in a preset three-dimensional grid space, there may be spatial overlap between different gridded data blocks. In this case, numerical fusion is required based on the data quality scores of each grid point in each gridded data block. In one embodiment, when gridded data blocks in the updated data queue overlap spatially, stitching and fusing all overlapping gridded data blocks in the preset three-dimensional grid space includes: The data quality scores corresponding to the grid points of each gridded data block corresponding to the overlapping grid point m are used as weights for weighted fusion to obtain the value of the overlapping grid point m. , It is the data quality score of the k-th gridded data block that overlaps spatially, at the overlapping grid point m. This represents the value at overlapping grid point m of the k-th spatially overlapping gridded data block, where N is the number of spatially overlapping gridded data blocks. This weighted fusion method ensures that high-quality data contributes more to the final 3D gridded data volume, while the impact of low-quality data is weakened.

[0054] Taking a dual-radar network scenario as an example: Grid point data approximately 30km from radar A has a high SNR, small beamwidth, and a Q value of approximately 0.9; grid point data approximately 150km from radar B has a lower SNR, severe beamwidth, and a Q value of approximately 0.4. During fusion in the overlapping area, the data contribution weight of radar A is 0.9 / (0.9+0.4)≈0.69, while the data contribution weight of radar B is only 0.31. High-quality, near-range data dominates the fusion result.

[0055] The final 3D grid data volume obtained by stitching and fusion is as follows Figure 7 As shown, after completing the update of the 3D grid data volume at the current sampling time, the process of updating the 3D grid data volume at the next sampling time begins.

[0056] It should be noted that, Figure 7 This demonstration showcases the fusion result of gridded data blocks from a weather radar. Due to the discontinuity of weather radar data sampling in the vertical direction, the radial data blocks from all weather radars may not be able to cover the entire preset 3D grid space. Therefore, the fused 3D grid data volume will contain missing grid points. In this case, interpolation is required based on the radial data blocks with available observation data. The specific interpolation algorithm used is customized according to the actual application requirements. For example, the 8-point interpolation method first locates and locks the 8 nearest radial data points in the 3D space surrounding the target grid point, forming a minimum hexahedral grid. Then, based on the spatial distance from each vertex of the minimum hexahedral grid to the target grid point, a linear weighted average is calculated in three dimensions: distance, azimuth, and elevation. This average is then used to assign the interpolated result to the target grid after comprehensive calculation.

[0057] This application's method uses the scan time of one layer as the basic update cycle, triggering a 3D gridding update every time a data block of one layer or sector is received, significantly shortening the time delay from data acquisition to 3D product generation. Building upon this, an adaptive time window mechanism, a data quality score weighted fusion mechanism, and a data missing anomaly handling mechanism are further introduced to improve the overall performance of 3D gridding from three dimensions: update timeliness, data quality, and system robustness.

[0058] The above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations that can be directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.

Claims

1. A three-dimensional gridding method for weather radar based on layer data iterative updates, characterized in that, The three-dimensional gridding method for weather radar includes: Obtain radial data blocks for each weather radar at the current sampling time; each radial data block contains time attributes. Each radial data block is mapped onto a preset three-dimensional grid to obtain gridded data blocks. The radial distance between each grid point on the preset three-dimensional grid and the corresponding weather radar is combined to calculate the data quality score corresponding to each grid point. The data quality score indicates the data quality at the grid point. The higher the data quality score, the better the data quality at the grid point. Each gridded data block is loaded into the data queue, and old data blocks whose time attributes exceed the preset time range are removed based on the time attributes of the existing gridded data blocks in the data queue, resulting in an updated data queue. Based on the data quality scores of each grid point in each gridded data block, all gridded data blocks in the updated data queue are spliced ​​and merged in a preset three-dimensional grid space to generate a three-dimensional gridded data volume at the current sampling time, and then proceed to the next sampling time.

2. The three-dimensional gridding method for weather radar according to claim 1, characterized in that, The preset time range is a dynamic time window, and determining the length T of the dynamic time window includes: Real-time acquisition of meteorological data, which is based on raw radial data blocks obtained from weather radar scans or other meteorological monitoring equipment; The Convection Intensity Index (CI) is calculated based on the meteorological data. The CI indicates the degree of strong convection in the weather. The larger the CI, the stronger the convection intensity index. The dynamic time window length T is determined based on the convection intensity index CI.

3. The three-dimensional gridding method for weather radar according to claim 2, characterized in that, The meteorological data includes the maximum reflectivity of radar combined data, the area percentage of regions where the maximum reflectivity exceeds the preset reflectivity threshold, the area percentage of regions where the echo top height exceeds the preset height threshold, the maximum vertical integral liquid water content, mesocyclone detection results, and lightning frequency. The fuzzy logic comprehensive index method is used to calculate the convection intensity index (CI) for each meteorological data.

4. The three-dimensional gridding method for weather radar according to claim 2, characterized in that, Determining the dynamic time window length T based on the convection intensity index CI includes: When the convection intensity index CI is lower than the first threshold, the dynamic time window length T is determined to be a preset maximum duration; when the convection intensity index CI is higher than the second threshold, the dynamic time window length T is determined to be a preset minimum duration; when the convection intensity index CI is not lower than the first threshold and not higher than the second threshold, the convection intensity index CI is substituted into the mapping function to calculate the dynamic time window length T, and the mapping function represents that the dynamic time window length T decreases as the convection intensity index CI increases.

5. The three-dimensional gridding method for weather radar according to claim 1, characterized in that, Gridded data blocks contain confidence levels ; The gridded data block obtained from any weather radar at the current sampling time includes: The arrival time of the radial data blocks generated by the weather radar at the current sampling time is monitored; If no radial data block is received within a preset timeout threshold, it is determined that the original radial data block generated by the weather radar at the current sampling time is missing. An extrapolation algorithm is then used to map the received gridded data blocks from the previous two sampling times onto a preset 3D grid, generating the gridded data block for the weather radar at the current sampling time. The confidence level of the gridded data block at the current sampling time is then determined. Assigning a value of 0< <1; otherwise, it is determined that the original radial data block generated by the weather radar at the current sampling time is not missing, and the original radial data block is mapped onto a preset three-dimensional grid using an interpolation algorithm to generate the gridded data block of the weather radar at the current sampling time, and the confidence level of the gridded data block at the current sampling time is set. Assigned value .

6. The three-dimensional gridding method for weather radar according to claim 5, characterized in that, Gridded data blocks also include echo intensity and noise; The data quality score Q for any grid point is calculated as follows: Based on the echo intensity and noise of any grid point i in the gridded data block, determine the signal-to-noise ratio of grid point i. ; Based on the radial distance between grid point i and the weather radar Determine the distance decay factor for grid point i. The distance attenuation factor reflects the impact of radial distance on data quality. Larger distance attenuation factor The smaller the value, the worse the data quality. The data quality score corresponding to grid point i in the gridded data block is calculated. , It is the signal-to-noise ratio normalization function. , It is a weighting coefficient and , It's the confidence level.

7. The three-dimensional gridding method for weather radar according to claim 6, characterized in that, Determine the distance decay factor for grid point i , It is the characteristic distance constant.

8. The three-dimensional gridding method for weather radar according to claim 1, characterized in that, When gridded data blocks in the updated data queue overlap spatially, all overlapping gridded data blocks are stitched and merged in a preset 3D grid space, including: The data quality scores corresponding to the grid points of each gridded data block corresponding to the overlapping grid point m are used as weights for weighted fusion to obtain the value of the overlapping grid point m. , It is the data quality score of the k-th gridded data block that overlaps spatially, at the overlapping grid point m. It is the value at the overlapping grid point m of the k-th gridded data block that overlaps in space, and N is the number of gridded data blocks that overlap in space.

9. The three-dimensional gridding method for weather radar according to claim 5, characterized in that, The gridded data block also contains elevation angle information. The extrapolation algorithm uses an optical flow-based extrapolation algorithm. When the original radial data block generated at the current sampling time is missing, the gridded data block generated by the weather radar at the current sampling time includes: Based on the gridded data blocks of the weather radar at the previous two sampling times at the current sampling time, the motion vector fields of the echo in the vertical and horizontal directions are calculated using the optical flow method. Based on the motion vector field and the height difference between the current sampling time and the corresponding elevation angle of the previous two sampling times, the displacement of the echo structure is calculated; the gridded data blocks of the previous two sampling times are translated along the motion vector direction to the height position of the corresponding elevation angle of the current sampling time, and the gridded data blocks of the weather radar at the current sampling time are generated by interpolation.

Images