Three-dimensional wind field retrieval method based on dual-polarization radar radial velocity and polarization parameters

Abstract

The application provides a three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters, which comprises the following steps: through analyzing the multi-dimensional information of radar radial velocity and polarization parameters, a three-dimensional wind field containing high-precision vertical velocity W is output; based on column / hole and foot recognition of strong convective updraft / downflow area, a three-dimensional airflow type field is generated; a scene adaptive HVVP fitting horizontal wind field is adopted, and the median absolute deviation is used to remove abnormal points; a polarization correction model is introduced in the downflow area, and is fused with the radial velocity fitting result; a 26-neighbor connected domain marking supports maximum 55-degree tilt propagation; a pyramid interpolation is used to obtain a 1km resolution three-dimensional wind field. The application takes dual-polarization microphysical characteristics as a quantitative constraint of strong convective vertical velocity, and proposes a polarization correction model of "dragging + momentum downward" double mechanism in the downflow area.

Description

Technical Field

[0001] This invention relates to the fields of radar signal processing and wind field inversion technology, and in particular to a three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters. Background Technology

[0002] Vertical velocity W (unit: m / s) is a key dynamic parameter in the application of strong convection radar wind field inversion products. Updraft intensity directly indicates the potential for convection development, while downdraft intensity is closely related to surface thunderstorms, downbursts, and other hazards. However, vertical velocity cannot be directly measured by radar and must be obtained indirectly through inversion methods.

[0003] Limitations of existing 3D wind field products: Most commercially available 3D wind field products currently only provide horizontal wind speed components (u, v), excluding vertical velocity W. This is because vertical velocity is on the order of its magnitude (only cm / s in non-convective regions), making inversion difficult and resulting in low accuracy. However, in strong convective systems, the peak value of W can reach tens of m / s, serving as a direct indicator for identifying rising / sinking regions. Therefore, developing a high-precision vertical velocity inversion method to output a 3D wind field product that includes reliable W has significant commercial value.

[0004] Existing vertical velocity inversion methods are mainly divided into three categories: methods based on continuous equation integration (such as VAD, or Velocity Azimuth Display; and VVP, or Volume Velocity Processing), methods based on single-Doppler radar wind field inversion (such as HVVP, or Horizontal and Vertical Velocity Processing), and qualitative identification methods based on dual-polarization features. These methods suffer from the following technical drawbacks:

[0005] (1) The vertical velocity inversion accuracy is low, and the error is particularly significant in the sinking zone: radial velocity The sensitivity to W is much lower than that to horizontal winds, and the sinking zone lacks an independent physical model, with relative errors often exceeding 50%.

[0006] (2) The dual-polarization microphysics information is not fully utilized: the dual-polarization radar's A column (a columnar high-value area formed by updrafts lifting large raindrops above the 0°C layer). Holes (low-value areas formed by strong downdrafts or hail) and Features such as the "foot" (a foot-shaped high-value area formed by hail melting near the 0°C layer) are only used for qualitative identification and have not been converted into quantitative physical constraints.

[0007] (3) Fixed parameter configuration and poor scene adaptability: The fitting order, statistical radius and outlier removal threshold are fixed and cannot adapt to the changes from strong convection to stratiform clouds.

[0008] (4) The inclined propagation structure of airflow is not considered: In actual strong convective storms, the maximum inclination of rising / falling airflow can reach 55°, and the existing vertical propagation assumption leads to identification bias. Summary of the Invention

[0009] To address the problems existing in the prior art, this invention provides a three-dimensional wind field inversion method based on the radial velocity and polarization parameter of dual-polarization radar. By analyzing the radial velocity observation equation and combining it with the quantitative constraint of polarization parameter, a three-dimensional wind field product containing high-precision vertical velocity W is output.

[0010] This invention provides a three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters, comprising:

[0011] The dual-polarization echo signal data of each radar within the network area are projected onto a three-dimensional grid to obtain... column, donghe Foot test results;

[0012] based on column, donghe The detection results of the foot generate a three-dimensional airflow type field, and propagate along the vertical direction based on the three-dimensional airflow type field;

[0013] Select the corresponding HVVP parameters according to the airflow type and height layer after propagation, establish the radial velocity observation equation according to the HVVP parameters, and solve the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field.

[0014] The radial velocity observation equation is solved by weighted least squares fitting based on radial velocity and background horizontal wind field, the vertical velocity is inverted, and directional constraints are applied to the vertical velocity according to airflow type.

[0015] A multi-scale pyramid interpolation method was used to construct a pyramid grid of multiple scales. Interpolation was performed layer by layer from coarse to fine to complete the effective data grid points and obtain the complete background horizontal wind field.

[0016] Output a three-dimensional wind field, including three-dimensional wind field data containing horizontal wind speed components and vertical velocity.

[0017] According to the present invention, a three-dimensional wind field inversion method based on the radial velocity and polarization parameters of dual-polarization radar is provided, which projects the dual-polarization echo signal data of each radar in the network area onto a three-dimensional grid to obtain... column, donghe The test results for the feet include:

[0018] Identify differential reflectance Greater than Three-dimensional connected regions that meet the column detection threshold and whose volume is greater than the minimum connected component volume threshold are marked as... column;

[0019] Identify differential reflectance Less than Hole detection threshold and reflectivity factor Regions with echo intensity greater than the minimum echo intensity threshold are marked as Hole;

[0020] Identify the differential propagation phase shift rate near the 0°C layer Greater than Regions that meet the detection threshold and whose volume is greater than half of the minimum connected component volume threshold are marked as... foot.

[0021] This invention provides a three-dimensional wind field inversion method based on the radial velocity and polarization parameters of dual-polarization radar. column, donghe The detection results of the feet generate a three-dimensional airflow type field, including:

[0022] If any lattice point simultaneously satisfies the condition of being located at If a grid point is located within a convection identification area and simultaneously meets the conditions of having graupel or hail at temperatures above 0°C and the echo bottom is not grounded, then the grid point is determined to be an updraft area.

[0023] If any lattice point simultaneously satisfies the condition of being located at Inside the cave and located within the convection identification area, or simultaneously meeting the following conditions: If the grid point is located within the convection recognition area, or if hail or graupel exists below 0°C, then the grid point is determined to be a descending airflow area.

[0024] If the grid point is not in an updraft or downdraft region, then the grid point is determined to be in a neutral flow region.

[0025] According to the present invention, a three-dimensional wind field inversion method based on the radial velocity and polarization parameters of dual-polarization radar is provided, which is based on the propagation of the three-dimensional airflow type field along the vertical direction, including...

[0026] Using the identified updraft region in the three-dimensional airflow type field as the seed layer, the propagation extends downwards to the echo bottom;

[0027] Using the identified downdraft region in the three-dimensional airflow type field as the seed layer, the propagation extends upward to the top of the echo and downward to the bottom of the echo.

[0028] The propagation path proceeds along the inclined direction, and the inclined angle is determined by... Determined by the centroid method or radial wind method;

[0029] During propagation, only the positions originally marked as neutral airflow zones are filled, while existing markings for other airflow types remain unchanged. Propagation stops if an echo gap is encountered or the maximum tilt angle is exceeded on the propagation path.

[0030] According to the present invention, a three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters is provided, wherein the HVVP parameters include fitting order, statistical radius and minimum number of samples.

[0031] According to the present invention, a three-dimensional wind field inversion method based on radial velocity and polarization parameters of dual-polarization radar is provided, which obtains the background horizontal wind field by solving the radial velocity observation equation through weighted least squares fitting. The method further includes:

[0032] Outliers are removed using the absolute deviation of the median.

[0033] According to the present invention, a three-dimensional wind field inversion method based on radial velocity and polarization parameters of dual-polarization radar is provided, wherein the weighted least squares method uses distance weight and time weight to weight the fitted samples;

[0034] The distance weight The calculation formula is: ,in, The horizontal distance from the sampling point to the center of HVVP. For statistical radius;

[0035] The time weight The calculation formula is: Δt is the difference between the observation time and the reference time. It is a time decay constant, which depends on the radar scan time and the current network deployment time.

[0036] According to the present invention, a three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters, after solving the radial velocity observation equation using a weighted least squares method based on radial velocity and background horizontal wind field to invert the vertical velocity, the method further includes:

[0037] Calculate the drag velocity and momentum transfer velocity in the downdraft region;

[0038] Based on the towing speed and momentum transfer speed, the polarization-estimated sinking speed is calculated.

[0039] The polarization-estimated sinking velocity is fused with the radial velocity fitting results to obtain the final vertical velocity.

[0040] The present invention also provides a three-dimensional wind field inversion device for dual-polarization radar radial velocity and polarization parameters, comprising:

[0041] The detection module is used to project the dual-polarization echo signal data of each radar within the network area onto a three-dimensional grid to obtain... column, donghe Foot test results;

[0042] The propagation module is used for propagation based on column, donghe The detection results of the foot generate a three-dimensional airflow type field, and propagate along the vertical direction based on the three-dimensional airflow type field;

[0043] The first solution module is used to select the corresponding HVVP parameters according to the airflow type and height layer after propagation, establish the radial velocity observation equation according to the HVVP parameters, and solve the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field.

[0044] The second solution module is used to solve the radial velocity observation equation based on the radial velocity and the background horizontal wind field using the weighted least squares method, invert the vertical velocity, and apply directional constraints to the vertical velocity according to the airflow type.

[0045] The interpolation module is used to construct a pyramid grid of multiple scales using the multi-scale pyramid interpolation method, interpolating layer by layer from coarse to fine to complete the effective data grid points and obtain the complete background horizontal wind field.

[0046] The output module is used to output three-dimensional wind field data, including horizontal wind speed components and vertical velocity components.

[0047] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters as described above.

[0048] The present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters as described above.

[0049] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters as described above.

[0050] The three-dimensional wind field inversion method based on the radial velocity and polarization parameter of dual-polarization radar provided by this invention belongs to the field of radar signal processing technology. Its core lies in the mathematical inversion of radar echo signals (radial velocity and polarization parameter) to obtain the three-dimensional wind field, rather than the statistical prediction or numerical forecast of meteorological elements, which is fundamentally different from meteorological statistical methods.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] Dual polarization microphysical features ( column / Hole / The problem of systematically underestimating the vertical velocity in the sinking zone was effectively solved by converting the vertical velocity of the sinking zone into a quantitative physical constraint.

[0053] A dual-mechanism polarization correction model of "dragging + momentum transfer" in the sinking zone is proposed, and the dragging velocity of hydrogel and the momentum transfer velocity at high altitude are calculated based on the physical mechanism.

[0054] The use of 26-neighbor connectivity domain labeling supports 55° tilt propagation, effectively tracking the tilt structure of airflow in severe convective storms.

[0055] Scene-adaptive MAD removal and polarization correction are optimized together to form a complete closed loop of "data quality control → physical correction".

[0056] Achieving 1 km high-resolution three-dimensional wind field output through multi-scale pyramid interpolation. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0058] Figure 1 This is a flowchart illustrating the three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters provided by the present invention.

[0059] Figure 2 This is a schematic diagram illustrating the application effect of the three-dimensional wind field inversion method based on the radial velocity and polarization parameters of dual-polarization radar provided by the present invention in a specific case.

[0060] Figure 3 This is a schematic diagram of the structure of the three-dimensional wind field inversion device based on the radial velocity and polarization parameters of dual-polarization radar provided by the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0062] The following is combined Figure 1 This invention describes a three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters, comprising:

[0063] Step 101: Project the dual-polarization echo signal data of each radar within the network area onto a three-dimensional grid to obtain... column, donghe Foot test results;

[0064] Step 102, based on column, donghe The detection results of the foot generate a three-dimensional airflow type field, and propagate along the vertical direction based on the three-dimensional airflow type field;

[0065] Step 103: Select the corresponding HVVP parameters according to the airflow type and height layer after propagation, establish the radial velocity observation equation according to the HVVP (Hurricane Volume Velocity Processing) parameters, and solve the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field.

[0066] Step 104: Based on the radial velocity and the background horizontal wind field, the radial velocity observation equation is solved by weighted least squares fitting, the vertical velocity is inverted, and the vertical velocity is subject to directional constraints according to the airflow type.

[0067] Step 105, Multi-scale pyramid interpolation: Construct a pyramid grid with 9 scales from 10 km to 160 km, interpolate layer by layer from coarse to fine, complete the effective data grid points, and obtain the complete background horizontal wind field with a horizontal resolution of 10 km;

[0068] Step 106, 3D wind field output: Output 3D wind field data containing horizontal wind speed components (u,v) and vertical velocity W.

[0069] Data Acquisition and Preprocessing: Acquire dual-polarization echo signal data from each radar within the network area, including reflectivity factors. radial velocity Differential reflectivity Differential propagation phase shift rate Correlation coefficient And the particle phase state data derived externally from the aforementioned radar echo parameters. The resolution of each radar data is unified and projected onto a three-dimensional Cartesian coordinate grid with a horizontal resolution of 0.01°×0.01° and a vertical resolution of 500m.

[0070] Identification of strong convective rising / falling airflow regions: based on column, donghe The detection results of the feet are used to generate a three-dimensional airflow type field through three-dimensional connected component labeling and vertical propagation algorithm. .in For grid point coordinates, =1 indicates an updraft. =-1 indicates a descending airflow. =0 indicates the neutral region.

[0071] Columnar differential reflectivity A columnar region that is consistently higher than the surrounding environment in the vertical direction is formed by updrafts lifting large raindrops to above the 0°C layer, indicating a region of strong upward motion. Cave finger In relatively low spatial areas, strong downdrafts or hail fall indicate areas of strong downdraft movement. near the 0°C layer of the toes The abnormally enlarged foot-shaped area was formed by the hail melting process and is closely related to the strong subsidence area.

[0072] Scene-adaptive horizontal wind field inversion: based on airflow type Dynamic selection of HVVP fitting order for height layer (Fit order), statistical radius Based on the sample number threshold, a weighted least squares fitting method based on the median absolute deviation is used to obtain the background horizontal wind field (u,v) (horizontal wind speed component) and the root mean square error of fitting ε (root mean square error of fitting).

[0073] Vertical velocity refinement inversion: Based on radial velocity on a fine grid with a horizontal resolution of 1 km. The vertical velocity was inverted using the weighted least squares method, along with the background horizontal wind field (u, v). And according to airflow type Apply directional constraints; for the descending airflow region ( =-1) Enable the polarization correction model and calculate the polarization-estimated sinking velocity. and the results of fitting with radial velocity By merging, the final vertical velocity is obtained. .

[0074] Unlike step 103, where the horizontal wind field and vertical velocity are solved simultaneously, this step uses (u,v) obtained in step 103 as known quantities and substitutes them into the observation equation, solving only for W as a single parameter. The physical basis for this step-by-step solution strategy is that the radial velocity Vr is much more sensitive to the horizontal wind (u,v) than to the vertical velocity W (under the conventional scanning strategy, the elevation angle θ is small, and the coefficient of the sinθ term is much smaller than that of the cosθ term). If all three are solved simultaneously, W is easily overwhelmed by the horizontal wind error due to insufficient weight. By fixing (u,v) first and then inverting W in a step-by-step strategy, the coupled influence of horizontal wind and vertical velocity can be effectively separated, improving the stability of the vertical velocity inversion.

[0075] A directional constraint is applied based on the airflow type τ to ensure that the inversion results are consistent with the physical properties of the rising / falling airflow identified in step 102:

[0076] (1) If τ=1 (rising region): Force W raw =max(W raw ,0.5), upper limit W raw =50 m / s. This constraint ensures that the vertical velocity of the grid points in the rising region is positive. The lower limit of 0.5 m / s is used to filter out noise disturbances in the weak rising region, and the upper limit of 50 m / s is within the physically reasonable range (refer to the extreme value of the rising velocity in strong convection).

[0077] (2) If τ = -1 (sinking zone): Force W raw =min(W raw The constraint is set to -0.5), and polarization correction is enabled. This constraint ensures that the vertical velocity of the grid points in the sinking region is negative. The negative constraint triggers the polarization correction model, which physically corrects the sinking velocity using microphysical information.

[0078] (3) If τ=0 (neutral region): No direction is forced, W is output directly. raw .

[0079] The aforementioned directional constraint uses the airflow type identification result from step 102 as a priori physical constraint to ensure that the inversion result has the correct directionality in the rising and sinking regions, avoids sign errors caused by pure mathematical fitting, and provides triggering conditions for polarization correction in the sinking region.

[0080] The radial velocity observation equation is expressed as:

[0081] ;

[0082] in, Let θ be the radial velocity observation value (m / s) of the m-th sample, and θ be the radar elevation angle (rad). Let be the azimuth angle (rad) of the m-th sample. The fitting residual is (m / s).

[0083] The above observation equation establishes the radar radial velocity The quantitative relationship between the wind field components (u, v, W) and the three-dimensional wind field components. θ is the radial projection component of the horizontal wind, and Wsinθ is the radial projection component of the vertical velocity.

[0084] This equation is the core foundation of HVVP inversion: in step 103, the equation is solved by least-squares fitting to obtain the horizontal wind field (u,v) as the background field; in step 104, after fixing (u,v), the equation is used again to invert the vertical velocity W. This step-by-step solution strategy effectively avoids mutual interference between W and (u,v), improving the stability and accuracy of the vertical velocity inversion.

[0085] The inversion solution of the above observation equation is a typical secondary processing method for radar echo signals. Its technical essence lies in inverting the three-dimensional wind field from radar observations (radial velocity), which falls under the category of radar signal processing.

[0086] The horizontal wind field inversion in step 103 may have some grid points lacking valid data (such as areas with insufficient echo coverage), which needs to be filled in by spatial interpolation. This embodiment uses a multi-scale pyramid interpolation method to achieve complete wind field reconstruction through two stages: information aggregation from fine to coarse and information recovery from coarse to fine.

[0087] A multi-scale pyramid grid was constructed, with a scale set S={10,20,40,60,80,100,120,140,160} (unit: km). A total of 9 scales were established, covering the range from local to mesoscale.

[0088] Phase 1: From fine to coarse scale coarsening. Starting with the finest scale (10 km), for each coarse-scale grid point, the effective wind speed values ​​of all fine-scale grid points within its coverage area are statistically analyzed. If the number of effective values ​​is ≥3, then the wind speed is weighted according to distance: w = exp(-d 2 / σ 2 The weighted average is used as the wind speed value for the coarse-scale grid point (σ is half the grid spacing at this scale, and d is the horizontal distance (in km) from the center of the fine-scale grid point with effective wind speed to the center of the target coarse-scale grid point). If the number of effective values ​​is less than 3, it is marked as missing, and aggregation continues to a coarser scale. This process is repeated layer by layer until the coarsest scale (160 km), ensuring that each layer has enough data to support subsequent interpolation.

[0089] The second stage: coarse-to-fine recovery. Starting from the coarsest scale, missing grid points in the next finer-scale grid are filled using bilinear interpolation based on the current coarse-scale grid. This process is repeated layer by layer until a resolution of 10 km is restored. This process leverages the continuity of wind fields at large scales: wind field changes are gradual and data is complete at the coarse scale, and the large-scale information is gradually transferred to the missing locations at the finer scales through progressively refined interpolation.

[0090] Through the above two-stage pyramid interpolation, a complete background horizontal wind field (u,v) with a horizontal resolution of 10 km can be obtained, providing a spatially continuous horizontal wind field input for the vertical velocity inversion in step 104.

[0091] 3D wind field output: The final 3D wind field (u,v,W) is output as NetCDF (Network Common Data Form).

[0092] Based on the above embodiments, this embodiment projects the dual-polarization echo signal data of each radar within the network area onto a three-dimensional grid to obtain... column, donghe The test results for the feet include:

[0093] Column detection: identification > ( The column detection threshold is set to 1.0 to 1.5 dB based on climate statistics, and the volume is greater than [a certain value]. (Minimum connected component volume threshold, unit: number of grid points, taken as 20 to 50 based on climate statistics) 3D connected regions are labeled as column.

[0094] Hole detection: identification < ( Hole detection threshold, set to 0.5 dB) and with strong echo ( The overlapping area (>35dBZ) is marked as Hole.

[0095] Foot detection: Identifying the area near the 0°C layer > ( The detection threshold is met (based on climate statistics, 2.0° / km) and the volume is greater than [a certain value]. The area of ​​ / 2 is marked as foot.

[0096] A 26-neighborhood 3D connected component labeling algorithm is employed, supporting tilt propagation up to 55°. The 26-neighborhood includes all adjacent combinations of the current grid point in the row, column, and layer dimensions (a total of 3...). 3 (-1=26 directions), which can effectively connect the characteristic areas of the inclined distribution and adapt to the inclined structure of the airflow in strong convective storms.

[0097] Based on the above embodiments, this embodiment is based on column, donghe The detection results of the feet generate a three-dimensional airflow type field. ,include:

[0098] Rising region determination: If a grid point simultaneously satisfies "located in..." "within the column" and "located in the convection recognition region (i.e., Z)" H If a grid point is located within an echo region of ≥35 dBZ, or simultaneously meets the conditions of "graupel or hail phase present above 0°C" and "echo bottom not grounded (i.e., echo bottom height higher than 1 km)," then the grid point is determined to be an updraft region and marked. =1; This rule is based on the physical mechanism by which updrafts carry large raindrops or ice particles upwards.

[0099] Sinking zone determination: If a grid point simultaneously satisfies "located in "Inside the cave" and "located within the convection identification area", or simultaneously satisfying "located within the convection identification area". If a grid point is located "within the foot" and "within the convection recognition area", or "has hail or graupel phase below 0℃", then the grid point is determined to be a descending airflow area and marked. =-1; This rule is based on the physical mechanism of downdrafts accompanying hail or the dragging of precipitation particles.

[0100] Neutral region: Grid points that do not meet any of the above conditions are marked. =0.

[0101] Based on the above embodiments, this embodiment, after completing the initial classification of airflow types, propagates along the vertical direction to fill the discontinuous areas caused by missing echoes or identification thresholds, and corrects the spatial offset caused by the tilted structure:

[0102] Updraft propagation: Using the identified updraft region as the seed layer, it propagates downwards to the echo bottom (the lowest altitude layer with an echo intensity ≥ 35 dBZ). The propagation path proceeds along the tilt direction, passing through... The maximum supported tilt angle is determined by either the centroid method (calculating the horizontal offset of the echo centroid at different heights) or the radial wind method (using the radial wind horizontal gradient to estimate the airflow tilt angle).

[0103] Downdraft propagation: Using the identified downdraft zone as the seed layer, it propagates upwards to the top of the echo and downwards to the bottom of the echo. The propagation path is also along the inclined direction, and the method for determining the inclination angle is the same as that for updraft.

[0104] Propagation constraints: During propagation, only the positions originally marked as neutral points are filled, and existing airflow type markings remain unchanged; if an echo gap is encountered or the maximum tilt angle is exceeded on the propagation path, propagation stops.

[0105] Based on the above embodiments, the HVVP parameters in this embodiment include the fitting order, statistical radius, and minimum number of samples.

[0106] The HVVP parameters are dynamically selected based on the airflow type and altitude, as shown in Table 1.

[0107] Table 1

[0108] The above fitting order N order This refers to the order of expansion of the spatial variation term of the horizontal wind field in the HVVP observation equation. When N order When =6, the observation equation includes horizontal wind (u,v) and horizontal divergence ( ), horizontal deformation ( ), vertical shear ( It can analyze complex nonlinear wind field structures in strong convection regions by identifying six unknown parameters, including N. order When =4, the observation equation only includes horizontal wind (u,v) and horizontal divergence ( The two main parameters have relatively low computational cost and are suitable for scenarios with sparse upper-level echoes and relatively uniform wind fields. Through the above-mentioned scenario-specific adaptive configuration, a high-order fitting is used in the strong convection region to preserve details, while a low-order fitting is used in the upper-level sparse region to ensure computational stability.

[0109] The above statistical radius R stat The value of is set based on the difference in echo sampling density at different altitudes. In the lower layers (<0℃), sampling points are dense, so a smaller radius (5-15 km) is used to preserve wind field details; in the middle layers (0~20℃), sampling point density is moderate, so a medium radius (15-20 km) is used to ensure sufficient sample numbers; in the upper layers (>20℃), echoes are sparse and sampling points are few, so a larger radius (20-30 km) is used to ensure sufficient samples to support the solution of the equations. In the strong convective uplift / sinking zone, the wind field changes drastically, so a smaller radius of 5-10 km is used to avoid over-smoothing and preserve the mesoscale wind field structure characteristics.

[0110] The minimum sample size is set to match the fitting order: a 6th-order equation requires solving for 6 unknown parameters, necessitating at least 800 valid sampling points to ensure a stable solution; a 4th-order equation only requires 500 valid sampling points. Since high-altitude echoes are sparse and sampling points are few, the minimum sample size for high-altitude areas is reduced to 500.

[0111] Based on the above embodiments, this embodiment solves the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field, and also includes:

[0112] Outliers are removed using the absolute deviation of the median.

[0113] Traditional HVVP methods use a fixed threshold (such as 3 times RMSE, i.e., Root Mean Square Error) to remove outliers, and this threshold remains unchanged in all scenarios. However, in strong convective regions, radial velocity changes drastically, and the actual meteorological signal itself has large dispersion. A fixed threshold can easily misclassify valid extreme velocity values ​​as outliers and remove them, resulting in the loss of wind field details in the convective core area. On the other hand, in stratiform clouds or clear-sky areas, velocity changes are gradual, and a fixed threshold may miss some clutter points.

[0114] To address the aforementioned issues, this embodiment introduces the Median Absolute Deviation (MAD) robust statistical method and employs a scenario-adaptive thresholding strategy. The MAD method is insensitive to outliers and can more accurately estimate data dispersion in strong convective scenarios, thereby avoiding the erroneous rejection of real meteorological signals. Let the fitting residual be r. i ,but:

[0115] MAD=median i (∣r i -median(r)∣);

[0116] The outlier criterion is:

[0117] ∣r i -median(r)∣>τ MAD* MAD;

[0118] Where r represents the sum of all individual residuals r i The overall dataset consists of r = {r1, r2, ..., r} n} (where n is the total number of valid samples within the current statistical radius); median(r) refers to the median of the entire residual dataset (i.e., the median of all r). i (The median value after sorting). Threshold multiple τ MAD Adaptive adjustment based on scene type: τ is used for strong convection scenes. MAD=4.5 (retaining more extreme values ​​to reflect convection intensity), τ is taken for the ordinary middle layer. MAD =3.0, τ is taken for ordinary high-rise buildings. MAD =2.5.

[0119] The above adaptive strategy is based on the following physical principles: the velocity changes drastically in strong convection areas, requiring a wider threshold range to preserve the true convection signal; and the upper-level echoes are sparse and clutter interference is relatively prominent, requiring a narrower threshold range to ensure data quality.

[0120] Based on the above embodiments, the contribution of different sampling points to HVVP fitting should be different in this embodiment: the farther the sampling point is from the HVVP center, the weaker its representativeness; the earlier the observation time of the sampling point, the lower its correlation with the wind field at the current time. Therefore, this embodiment adopts distance weighting. and time weight Weighting the samples makes the fitting results more reasonable:

[0121] The distance weight The calculation formula is: ,in, The horizontal distance (km) from the sampling point to the center of HVVP. The statistical radius is (km).

[0122] The time weight The calculation formula is: Δt is the difference (s) between the observation time and the reference time. It is the time decay constant, which depends on the radar scan time and the time (s) between the current network formation and the current network formation time.

[0123] Distance weight The weighting decreases exponentially with increasing horizontal distance, giving higher weights to sampling points closer to the center and gradually decreasing weights to sampling points farther from the center. This ensures that the fitting results mainly reflect the wind field characteristics near the HVVP center.

[0124] Existing dual-Doppler radar inversion methods require strict synchronous observation from two radars, placing stringent demands on scanning strategies and spatiotemporal matching, making them difficult to implement operationally in large-scale networks. This embodiment is based on a single-Doppler radar network, where each radar scans independently, eliminating the need for time synchronization. However, different radars complete volume scans at different times. To fully utilize the sampling information from each radar, this invention introduces time weighting. The samples are weighted. This weighting assigns higher weights to sampling points whose observation time is closer to the reference time, while older data have gradually reduced weights but still participate in the fitting process. This allows for the effective fusion of observation information from multiple radars without the need for strict time synchronization. Through the combined strategy of distance and time weighting, the interference of distant and outdated samples on the fitting results can be effectively suppressed, improving the accuracy and robustness of horizontal wind field inversion.

[0125] Based on the above embodiments, this embodiment, after solving the radial velocity observation equation using a weighted least squares method based on the radial velocity and background horizontal wind field, and inverting the vertical velocity, further includes:

[0126] Calculate the drag velocity and momentum transfer velocity in the downdraft region;

[0127] Based on the towing speed and momentum transfer speed, the polarization-estimated sinking speed is calculated.

[0128] The polarization-estimated sinking velocity is fused with the radial velocity fitting results to obtain the final vertical velocity.

[0129] To address the issue of systematically underestimating the vertical velocity in the sinking zone, this embodiment proposes a polarization correction model based on dual-polarization microphysical characteristics. This model consists of two parts: drag velocity and momentum transfer velocity, corresponding to the gravitational drag effect of precipitation particles and the vertical transfer effect of horizontal momentum from the upper atmosphere, respectively.

[0130] (1) Dragging speed The calculation formula is:

[0131] ;

[0132] in, The load drag coefficient is a dimensionless parameter with a value range of 1.0-2.0, preferably 1.5. The vertical integral height is expressed in km. These are vertical coordinates, in km, typically taken up to the current altitude level; The differential propagation phase shift rate is (° / km). It is directly proportional to the liquid water content, and its vertical integral reflects the cumulative load of the hydrogel. The greater the load, the greater the sinking speed caused by dragging.

[0133] (2) Momentum transfer speed The calculation formula is:

[0134] ;

[0135] in, The base sinking velocity, expressed in m / s, is determined based on microphysical characteristics. The speed range is -6 m / s to -4 m / s in the cave area and -12 m / s to -8 m / s in the hail area; α is the momentum transfer enhancement coefficient, a dimensionless parameter with a value range of 0.003-0.005; The maximum horizontal wind speed below the 0°C level is expressed in m / s. This is the threshold for initiating momentum downslope; the default value is 5 m / s. Physical meaning: Horizontal momentum from the upper atmosphere is transported downwards through precipitation particles; the greater the horizontal wind speed, the stronger the downslope.

[0136] (3) Polarization estimation of sinking velocity .

[0137] (4) The fusion strategy is .

[0138] The more negative (stronger sinking) value is taken from the radial velocity fitting result and the polarization estimation result. The physical basis for this is: the radial velocity fitting result... Often, the polarization estimation results are systematically underweighted (underestimating the subsidence intensity) due to insufficient weight of W in the observation equation. Based on independent calculations using microphysical mechanisms, taking more negative values ​​for both can effectively compensate for the systematic underestimation of sinking velocity by traditional methods.

[0139] To verify the actual effect of the present invention, Figure 2 The application effect of this method in a case study is demonstrated, showing the vertical profile feature map of the wind field and water condensate phase.

[0140] Combination Figure 2 It is evident that this method successfully constructs joint constraints of microphysics and dynamics within strong convective cloud systems:

[0141] Qualitative and quantitative analysis of microphysical characteristics: Figure 2 The cross-section clearly reveals the complex phase evolution within the convective cloud. In the middle section of the cross-section (approximately 20-25 km), there is a sharp abrupt change from the lower-level "snow / rain" (blue / green) to the upper-level "ice crystals" (purple). At an altitude of 5-7 km, a dense "heavy precipitation" phase (red patches) appears. These mixed phase regions and phase abrupt changes provide a high-confidence a priori constraint for identifying the airflow type field (i.e., "microphysical source-sink terms") in this invention, effectively helping to distinguish between the condensation latent heat release zone and the precipitation drag deposition zone.

[0142] The physical manifestation of dynamic constraints: In the profile region corresponding to the heavy precipitation phase, there is a significant vertical wind shear—the lower layer (2-4km) has dense arrows with wind speeds reaching 10-15 m / s, while the wind speed in the middle and upper layers (6-10km) decreases significantly. The dynamic difference between the strong inflow at the lower layer and the weakened outflow at the upper layer constitutes the dynamic boundary conditions for scene adaptive HVVP fitting and vertical velocity direction constraints in this invention.

[0143] Precise Vertical Structure Reconstruction: Utilizing the aforementioned microphysical phase tracing and dynamic boundary constraints, this method successfully identified the boundary between the strong updraft on the front side (right inflow region) and the descending airflow region dominated by precipitation drag on the rear side (left side). Comparing the inversion results with observation data from automatic weather stations, the maximum vertical velocity of the descending region obtained by this method is approximately -40 m / s, with minimal error compared to the actual thunderstorm wind (43 m / s) measured by surface meteorological stations in the same area on that day. This invention effectively solves the problem of traditional methods systematically underestimating the vertical velocity of the descending region (errors often >50%), verifying the practicality and accuracy of this method in wind field inversion applications under severe convective weather scenarios.

[0144] The following describes the three-dimensional wind field inversion device based on the radial velocity and polarization parameters of dual-polarization radar provided by this invention. The three-dimensional wind field inversion device based on the radial velocity and polarization parameters of dual-polarization radar described below can be referred to in correspondence with the three-dimensional wind field inversion method based on the radial velocity and polarization parameters of dual-polarization radar described above.

[0145] like Figure 3 As shown, the device includes:

[0146] The detection module 301 is used to project the dual-polarization echo signal data of each radar within the network area onto a three-dimensional grid to obtain... column, donghe Foot test results;

[0147] Propagation module 302 is used for... column, donghe The detection results of the foot generate a three-dimensional airflow type field, and propagate along the vertical direction based on the three-dimensional airflow type field;

[0148] The first solution module 303 is used to select the corresponding HVVP parameters according to the airflow type and height layer after propagation, establish the radial velocity observation equation according to the HVVP parameters, and solve the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field.

[0149] The second solution module 304 is used to solve the radial velocity observation equation based on the radial velocity and the background horizontal wind field using the weighted least squares method, invert the vertical velocity, and apply directional constraints to the vertical velocity according to the airflow type.

[0150] The interpolation module 305 is used to construct a pyramid grid of multiple scales using the multi-scale pyramid interpolation method, interpolating layer by layer from coarse to fine to complete the effective data grid points and obtain the complete background horizontal wind field;

[0151] The output module 306 is used to output three-dimensional wind field data, which includes horizontal wind speed components and vertical velocity components.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A three-dimensional wind field inversion method based on the radial velocity and polarization parameters of dual-polarization radar, characterized in that, include: The dual-polarization echo signal data of each radar within the network area are projected onto a three-dimensional grid to obtain... column, donghe Foot test results; based on column, donghe The detection results of the foot generate a three-dimensional airflow type field, and propagate along the vertical direction based on the three-dimensional airflow type field; Select the corresponding HVVP parameters according to the airflow type and height layer after propagation, establish the radial velocity observation equation according to the HVVP parameters, and solve the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field; The radial velocity observation equation is solved by weighted least squares fitting based on radial velocity and background horizontal wind field, vertical velocity is inverted, and directional constraints are applied to the vertical velocity according to airflow type. A multi-scale pyramid interpolation method was used to construct a pyramid grid of multiple scales. Interpolation was performed layer by layer from coarse to fine to complete the effective data grid points and obtain the complete background horizontal wind field. Output a three-dimensional wind field, including three-dimensional wind field data with horizontal wind speed components and vertical velocity; After solving the radial velocity observation equation using a weighted least squares method based on the radial velocity and background horizontal wind field, and inverting the vertical velocity, the following steps are also included: Calculate the drag velocity and momentum transfer velocity in the downdraft region; Based on the towing speed and momentum transfer speed, the polarization-estimated sinking speed is calculated. The polarization-estimated sinking velocity is fused with the radial velocity fitting results to obtain the final vertical velocity.

2. The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters according to claim 1, characterized in that, The dual-polarization echo signal data of each radar within the network area are projected onto a three-dimensional grid to obtain... column, donghe The test results for the feet include: Identify differential reflectance Greater than Three-dimensional connected regions that meet the column detection threshold and whose volume is greater than the minimum connected component volume threshold are marked as... column; Identify differential reflectance Less than Hole detection threshold and reflectivity factor Regions with echo intensity greater than the minimum echo intensity threshold are marked as Hole; Identify the differential propagation phase shift rate near the 0°C layer Greater than Regions that meet the detection threshold and whose volume is greater than half of the minimum connected component volume threshold are marked as... foot.

3. The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters according to claim 1, characterized in that, based on column, donghe The detection results of the feet generate a three-dimensional airflow type field, including: If any lattice point simultaneously satisfies the condition of being located at If a grid point is located within a convection identification area and simultaneously meets the conditions of having graupel or hail at temperatures above 0°C and the echo bottom is not grounded, then the grid point is determined to be an updraft area. If any lattice point simultaneously satisfies the condition of being located at Inside the cave and located within the convection identification area, or simultaneously meeting the following conditions: If the grid point is located within the convection recognition area, or if hail or graupel exists below 0°C, then the grid point is determined to be a descending airflow area. If the grid point is not in an updraft or downdraft region, then the grid point is determined to be in a neutral flow region.

4. The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters according to claim 1, characterized in that, Based on the propagation of the three-dimensional airflow type field along the vertical direction, including Using the identified updraft region in the three-dimensional airflow type field as the seed layer, the propagation extends downwards to the echo bottom; Using the identified downdraft region in the three-dimensional airflow type field as the seed layer, the propagation extends upward to the top of the echo and downward to the bottom of the echo. The propagation path proceeds along the inclined direction, and the inclined angle is determined by... Determined by the centroid method or radial wind method; During propagation, only the positions originally marked as neutral airflow zones are filled, while existing markings for other airflow types remain unchanged. Propagation stops if an echo gap is encountered or the maximum tilt angle is exceeded on the propagation path.

5. The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters according to claim 1, characterized in that, The HVVP parameters include the fitting order, statistical radius, and minimum number of samples.

6. The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters according to claim 1, characterized in that, The radial velocity observation equation is solved by weighted least squares fitting to obtain the background horizontal wind field, which also includes: Outliers are removed using the absolute deviation of the median.

7. The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters according to claim 1, characterized in that, The weighted least squares method uses distance weight and time weight to weight the fitted samples; The distance weight The calculation formula is ,in, The horizontal distance from the sampling point to the center of HVVP. For statistical radius; The time weight The calculation formula is Δt is the difference between the observation time and the reference time. It is a time decay constant, which depends on the radar scan time and the current network deployment time.

8. A three-dimensional wind field inversion device based on the radial velocity and polarization parameters of a dual-polarization radar, characterized in that, The three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters, as described in any one of claims 1-7, includes: The detection module is used to project the dual-polarization echo signal data of each radar within the network area onto a three-dimensional grid to obtain... column, donghe Foot test results; The propagation module is used for propagation based on column, donghe The detection results of the foot generate a three-dimensional airflow type field, and propagate along the vertical direction based on the three-dimensional airflow type field; The first solution module is used to select the corresponding HVVP parameters according to the airflow type and height layer after propagation, establish the radial velocity observation equation according to the HVVP parameters, and solve the radial velocity observation equation by weighted least squares fitting to obtain the background horizontal wind field. The second solution module is used to solve the radial velocity observation equation based on the radial velocity and the background horizontal wind field using the weighted least squares method, invert the vertical velocity, and apply directional constraints to the vertical velocity according to the airflow type. The interpolation module is used to construct a pyramid grid of multiple scales using the multi-scale pyramid interpolation method, interpolating layer by layer from coarse to fine to complete the effective data grid points and obtain the complete background horizontal wind field. The output module is used to output three-dimensional wind field data, including horizontal wind speed components and vertical velocity components.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the three-dimensional wind field inversion method based on dual-polarization radar radial velocity and polarization parameters as described in any one of claims 1 to 7.

Images