A stratiform cloud melting layer identification method and system based on a cloud radar

Abstract

This invention discloses a method and system for identifying layered cloud melting layers based on cloud radar, belonging to the field of cloud radar identification technology. The method includes: acquiring airborne cloud radar observation data; resampling the airborne cloud radar observation data to a unified geographic grid; for each sample, extracting continuous time periods with radar bright band characteristics and performing time averaging to obtain a comprehensive average reflectivity factor profile; calculating the vertical gradient of the comprehensive average reflectivity factor profile; and identifying the effective melting layer based on the vertical gradient. This invention automatically identifies the height of the top and bottom of the melting layer by analyzing the vertical gradient changes of reflectivity factor and vertical velocity, exhibiting high accuracy and strong adaptability.

Description

Technical Field

[0001] This invention relates to the field of cloud radar identification technology, and more specifically to a method and system for identifying layered cloud melting layers based on cloud radar. Background Technology

[0002] Over the past decade, the technical approaches for identifying melting layers using Ka / W band cloud radar have mainly fallen into two categories: (1) Single-parameter method: relying solely on the bright band peak value of the reflectivity factor Z, combined with a fixed dBZ threshold or an Otsu adaptive threshold, and then using the 0°C layer height for simple screening. (2) Multi-parameter method: simultaneously introducing three parameters: Z, radial velocity V, and spectral width SW, constructing a comprehensive index using manually set weights or empirical formulas, and then using Otsu or a fixed threshold to divide the bright band interval. Both of these schemes have been deployed in field operations and can be used as inputs for weather modification and numerical model verification.

[0003] However, existing technologies still have the following problems: (1) Fixed weights: The coefficients of Z, V and SW are given once based on experience. They need to be recalibrated across seasons and climate zones, resulting in poor regional adaptability.

[0004] (2) The false alarm rate is still high: the concept of "confidence" is lacking, weak peaks and side lobes are easily mistaken for real bright bands, and the false alarm rate in multi-layer scenarios is 8.4%.

[0005] (3) Insufficient thickness accuracy: The 0.27km error exceeds the hard target of <0.2km for artificial weather modification command, and cannot be directly used for rocket / anti-aircraft gun aiming.

[0006] (4) Heavy hardware burden: It must output a wide spectrum channel, the FPGA needs to perform three parallel FFTs, the power consumption is 5W, and the battery life in the field is limited.

[0007] (5) High maintenance cost: The fixed threshold drifts with the seasons and needs to be manually corrected regularly, making it difficult to achieve "zero knob" unattended operation.

[0008] Therefore, in view of the shortcomings of the existing technology, how to provide a method and system for identifying layered cloud melting layers based on cloud radar, reduce the thickness error of multiple melting layers to <0.2km and the false alarm rate to <3% without increasing hardware, and achieve full automation is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0009] In view of this, the present invention provides a method and system for identifying the melting layer of layered clouds based on cloud radar. By analyzing the vertical gradient changes of reflectivity factor and vertical velocity, the height of the top and bottom of the melting layer can be automatically identified, which has high accuracy and strong adaptability.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: a method for identifying layered cloud melting layers based on cloud radar, comprising: Acquire airborne cloud radar observation data; The airborne cloud radar observation data is resampled to a unified geographic grid to standardize spatial coordinates; For each sample, a continuous time period with radar bright band characteristics is extracted and averaged over time to obtain a comprehensive average reflectivity factor profile. Calculate the vertical gradient of the composite average reflectance factor profile; Based on the vertical gradient, an effective melting layer is identified.

[0011] Preferably, based on the vertical gradient, identifying the effective melting layer includes: starting from the 0°C layer and searching downwards to obtain the top of the melting layer based on a preset gradient; Below the top of the melt layer, the height at which the radar reflectivity reaches its maximum value is determined to be the bottom of the melt layer; A layer is considered valid when a pair of gradient-peak structures is detected and the thickness of the melt layer exceeds a preset value.

[0012] Preferably, the airborne cloud radar observation data includes vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T.

[0013] Preferably, the vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T are normalized to obtain normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T'.

[0014] Preferably, the normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T' are used to construct the fusion index MLI, as shown in the following formula: MLI(i)=α·Z'(i)+β·V'(i)+γ·T'(i); Where i is the height index; α, β, γ are determined adaptively using a Bayesian optimization algorithm on a training set containing 200,000 profiles.

[0015] Preferably, based on the vertical gradient, local peaks are detected in the MLI vertical profile, and layers with confidence Ck > 0.65 are retained; If the thickness of two adjacent layers is less than 0.3 km and the temperature difference is less than 1℃, they are merged into one layer. The layer with only the temperature between 0℃ and 2℃ is obtained, and the top height, bottom height and thickness of the effective fusion layer are output.

[0016] Preferably, a layered cloud melting layer identification system based on cloud radar includes: The data acquisition module is used to acquire observation data from the airborne cloud radar. The data resampling module is used to resample the airborne cloud radar observation data to a unified geographic grid; The data processing module is used to extract continuous time periods with radar bright band characteristics for each sample and perform time averaging to obtain a comprehensive average reflectivity factor profile. The calculation module is used to calculate the vertical gradient of the composite average reflectivity factor profile; The identification module is used to identify the effective melting layer based on the vertical gradient.

[0017] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a method and system for identifying layered cloud melting layers based on cloud radar. By analyzing the vertical gradient changes of reflectivity factor and vertical velocity, it automatically identifies the height of the top and bottom of the melting layer, which has high accuracy and strong adaptability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present 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 only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 A schematic diagram of the vertical velocity and reflectivity profiles after data resampling and averaging processing provided in an embodiment of the present invention.

[0020] Figure 2 This is a schematic flowchart of a method for identifying the melting layer of layered clouds based on cloud radar, provided in an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of a layered cloud melting layer identification system based on cloud radar, provided as an embodiment of the present invention. Detailed Implementation

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

[0023] This invention discloses a method for identifying layered cloud melting layers based on cloud radar, such as... Figure 1 As shown, it includes: Acquire airborne cloud radar observation data; The airborne cloud radar observation data is resampled to a unified geographic grid. For each sample, a continuous time period with radar bright band characteristics is extracted and averaged over time to obtain a comprehensive average reflectivity factor profile. Calculate the vertical gradient of the composite average reflectance factor profile; Based on the vertical gradient, an effective melting layer is identified.

[0024] This invention utilizes vertical reflectivity factor (dBZ) data acquired by airborne cloud radar (HCR) to precisely locate the melt layer (bright band) by capturing abrupt changes in the vertical reflectivity gradient caused by ice crystal melting near the 0°C isotherm height. First, the HCR data is resampled to a uniform geographic grid (2Hz temporal resolution, 19.1867m vertical resolution). Then, in each case study, time periods with distinct and continuous radar bright band characteristics are extracted, and the reflectivity within these periods is time-averaged to obtain a comprehensive average reflectivity factor profile. This data processing aims to suppress high-frequency noise interference while preserving key bright band gradient features.

[0025] like Figure 1 The diagram illustrates how to identify the location of the melt layer based on the reflectivity profile. This diagram corresponds to the vertical velocity and reflectivity profiles after data resampling and averaging from 01:15:00 to 01:17:00 on January 23, 2018. The shaded area in the diagram represents the identified melt layer.

[0026] After calculating the vertical gradient of the profile, starting from near the 0°C layer and searching downwards, the first gradient is ≤ -0.5dBZ (19.1867m). -1 The height of the top of the melt layer is determined as the height at which ice crystals melt and coalesce, resulting in a sharp increase in reflectivity; the height at which the maximum radar reflectivity occurs below this is determined as the height at which the bottom of the melt layer is determined as the height at which melting is basically complete, particles fall as raindrops, and reflectivity growth stops. Only when the above-mentioned paired gradient-peak structures are detected and the thickness of the melt layer exceeds 40m (twice the vertical grid spacing) is it considered a valid melt layer.

[0027] Specifically, based on the vertical gradient, the effective melting layer is identified, including: starting from the 0°C layer and searching downwards to obtain the top of the melting layer based on a preset gradient; Below the top of the melt layer, the height at which the radar reflectivity reaches its maximum value is determined to be the bottom of the melt layer; A layer is considered valid when a pair of gradient-peak structures is detected and the thickness of the melt layer exceeds a preset value.

[0028] Furthermore, since ice crystals fall at a speed of approximately 1 meter per second, while raindrops fall at approximately 4 meters per second, the melting of ice crystals in the melting layer will be directly reflected in the gradient change. Therefore, a mirror approach is adopted for the vertical velocity: from the same 0°C layer downwards, the first gradient ≥ 0.1 (19.1867 m). -1 The position is denoted as H1_Vel, and the height of the extreme velocity below it is denoted as H2_Vel.

[0029] The comparison revealed that H1_Vel is consistently lower than H1_Ze, corresponding to a higher temperature. This indicates that the velocity profile's response to the onset of melting lags slightly behind the microphysical changes captured by the reflectivity profile. In other words, the gradient jump in the dynamic response occurs after the ice crystals have clearly begun to melt. To ensure consistency with the definition of radar bright bands, this paper uniformly uses H1_Ze and H2_Ze identified by radar reflectivity as the final values ​​for the top and bottom heights of the melt layer.

[0030] Furthermore, since radar data is unavailable during the aircraft's descent through the clouds, this embodiment of the invention uses the nearest effective profile (usually within 1-2 minutes) before and after descent as the melt layer identification sample, and uses the obtained height as the representative value of the melt layer position during the cloud penetration process.

[0031] Specifically, the airborne cloud radar observation data includes vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T.

[0032] Specifically, the vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T are normalized to obtain normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T'.

[0033] Specifically, the normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T' are used to construct the fusion index MLI, as shown in the following formula: MLI(i)=α·Z'(i)+β·V'(i)+γ·T'(i); Wherein, α, β, γ are adaptive optimization parameters on a 200,000 profile training set optimized by Bayes.

[0034] Specifically, Bayesian optimization settings are used: Search space: α, β, γ∈[0,1] and α+β+γ=1; Objective function: min(thickness error MAE + 0.5 × false alarm rate); Optimal weights: α=0.50, β=0.32, γ=0.18.

[0035] Specifically, based on the vertical gradient, 1D-CMF detection is performed on the MLI vertical profile to detect local peaks, and layers with confidence Ck > 0.65 are retained; If the thickness of two adjacent layers is less than 0.3 km and the temperature difference is less than 1℃, they are merged into one layer. The final result is a layer that retains only the temperature at 0℃±2℃, and the top height, bottom height and thickness of the effective fusion layer are output.

[0036] The data source uses W-band pulse Doppler cloud radar with a wavelength of 3.2 mm, a range resolution of approximately 19 m, and a sampling interval of 0.5 s; the in-situ temperature measurement from the synchronous airborne temperature sensor is used as the true value.

[0037] Specifically, the vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T are normalized to obtain normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T', including: Z'=(Z-Zmin) / (Zmax-Zmin); V'=|ΔV| / 5ms-1; The upper limit is 1T' = 1 - |T| / 5℃, with an upper limit of 1.

[0038] The embodiments of this invention employ Bayesian triples, using Bayesian optimization to calculate α, β, and γ in one step, so that Z′, V′, and T′ are combined to form MLI, eliminating the need for further parameter tuning throughout the year and across all regions.

[0039] By using a confidence-merging double screening method, weak peaks with Ck≤0.65 are first removed, and then adjacent layers with a thickness <0.3km and a temperature difference <1°C are merged to prevent the same layer from being split or false bright bands.

[0040] A secondary temperature mask is used, and after the peak value is determined, a final sieve is made at 0℃±2℃ to ensure that each output layer has the significance of a physical melting layer.

[0041] The two-parameter architecture without spectral width uses only the most basic Z and V data from radar, eliminating the spectral width channel. This reduces FPGA resources and power consumption by about 25%, while still achieving a thickness error of ≤0.19km.

[0042] In one specific embodiment of the present invention, a layered cloud melting layer identification system based on cloud radar, such as... Figure 2 As shown, it includes: The data acquisition module is used to acquire observation data from the airborne cloud radar. The data resampling module is used to resample the airborne cloud radar observation data to a unified geographic grid; The data processing module is used to extract continuous time periods with radar bright band characteristics for each sample and perform time averaging to obtain a comprehensive average reflectivity factor profile. The calculation module is used to calculate the vertical gradient of the composite average reflectivity factor profile; The identification module is used to identify the effective melting layer based on the vertical gradient.

[0043] Compared with existing technical solutions, the embodiments of the present invention have the following obvious advantages: 1. Adaptive weighting, strong regional applicability In this embodiment of the invention, Bayesian optimization is used to automatically optimize α, β, and γ on a large dataset of 200,000 profiles. When the same set of coefficients was moved from Beijing to Guangzhou for testing, the thickness error increased by only 0.02 km, which is still lower than the 0.2 km command requirement, achieving "out-of-the-box usability".

[0044] 2. Multi-level false alarm rate decreased significantly.

[0045] The embodiments of the present invention introduce dual constraints of "peak confidence level Ck > 0.65" and "adjacent layer thickness < 0.3 km and temperature difference < 1℃ automatically merged". Bootstrap resampling shows that the false alarm rate is reduced to 3.3%, a reduction of 61%, and the 95% confidence intervals have no overlap, with statistical significance p < 0.01.

[0046] 3. Thickness error reaches "command-level" precision for the first time.

[0047] The embodiments of the present invention compress the error to 0.19±0.13km by using the ternary MLI index and merging correction, reducing the mean by 30% while keeping the maximum error at <0.3km, thus meeting the practical requirements of direct radar guidance for rocket anti-aircraft guns.

[0048] 4. Hardware resources and power consumption are reduced simultaneously.

[0049] Traditional methods require simultaneous output of three parameters: Z, V, and SW. This necessitates parallel three-channel FFT on the FPGA, consuming approximately 5W of power. The embodiment of this invention removes the spectral width variable, retaining only Z, V, and temperature. This reduces FPGA resource usage by 25%, and the measured power consumption drops to 3.7W while maintaining a latency of <0.2s. The battery life is extended by 35% with the same battery, making it suitable for unattended field sites.

[0050] 5. Fully automated "zero-knob" operation and maintenance

[0051] Existing fixed threshold schemes require manual adjustment of the dBZ threshold according to the season; the embodiment of the present invention deeply couples the Otsu adaptive threshold with the 0℃±2℃ temperature window. During eight consecutive months of field tests in 2023-2024, no manual intervention was required, the output was stable, and the maintenance cost was close to zero.

[0052] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0053] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying layered cloud melting layers based on cloud radar, characterized in that, include: Acquire airborne cloud radar observation data; The airborne cloud radar observation data is resampled to a unified geographic grid. For each sample, a continuous time period with radar bright band characteristics is extracted and averaged over time to obtain a comprehensive average reflectivity factor profile. Calculate the vertical gradient of the composite average reflectance factor profile; Based on the vertical gradient, an effective melting layer is identified.

2. The method for identifying layered cloud melting layers based on cloud radar according to claim 1, characterized in that, Based on the vertical gradient, the effective melting layer is identified, including: starting from the 0°C layer and searching downwards to obtain the top of the melting layer based on the preset gradient; Below the top of the melt layer, the height at which the radar reflectivity reaches its maximum value is determined to be the bottom of the melt layer; A layer is considered valid when a pair of gradient-peak structures is detected and the thickness of the melt layer exceeds a preset value.

3. The method for identifying layered cloud melting layers based on cloud radar according to claim 1, characterized in that, The airborne cloud radar observation data includes vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T.

4. The method for identifying layered cloud melting layers based on cloud radar according to claim 3, characterized in that, The vertical reflectivity factor data Z, vertical velocity data V, and in-situ measured temperature data T are normalized to obtain the normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T'.

5. The method for identifying layered cloud melting layers based on cloud radar according to claim 4, characterized in that, The fusion index MLI is constructed using the normalized vertical reflectivity factor data Z', vertical velocity data V', and in-situ measured temperature data T', as shown in the following formula: MLI(i)=α·Z'(i)+β·V'(i)+γ·T'(i); Where α, β, γ are adaptive optimization parameters on the profile training set by Bayesian optimization.

6. The method for identifying layered cloud melting layers based on cloud radar according to claim 5, characterized in that, Based on the vertical gradient, local peaks are detected in the MLI vertical profile, and layers with confidence Ck > 0.65 are retained; If the thickness of two adjacent layers is less than 0.3 km and the temperature difference is less than 1℃, they are merged into one layer. The layer with only the temperature between 0℃ and 2℃ is obtained, and the top height, bottom height and thickness of the effective fusion layer are output.

7. A layered cloud melting layer identification system based on cloud radar, characterized in that, include: The data acquisition module is used to acquire observation data from the airborne cloud radar. The data resampling module is used to resample the airborne cloud radar observation data to a unified geographic grid; The data processing module is used to extract continuous time periods with radar bright band characteristics for each sample and perform time averaging to obtain a comprehensive average reflectivity factor profile. The calculation module is used to calculate the vertical gradient of the composite average reflectivity factor profile; The identification module is used to identify the effective melting layer based on the vertical gradient.

Images