A precipitation particle phase state identification method of an X-band dual-polarization micro weather radar

Abstract

The present application relates to the field of radar technology, and more particularly to a precipitation particle phase state recognition method for an X-band dual-polarization micro-weather radar, which comprises the following steps: Step 1, scanning at a set pitch angle between horizontal and vertical, obtaining spectrum reflectivity, spectrum differential reflectivity and spectrum correlation coefficient along Doppler velocity distribution for each range bin; Step 2, determining a separation boundary according to a valley value formed by first decreasing and then increasing the spectrum correlation coefficient, dividing the range bin along Doppler velocity into a plurality of spectrum components corresponding to single phase states respectively; Step 3, obtaining component differential reflectivity, component correlation coefficient and relative order along fall velocity for each spectrum component, and obtaining phase state spatial distribution by summarizing and splicing. The present application can separate and recognize multiple coexisting phase states in a single resolution volume, and improves the accuracy of phase state recognition.

Description

Technical Field

[0001] This invention belongs to the field of radar technology, specifically relating to a method for identifying the phase state of precipitation particles in an X-band dual-polarization micro-weather radar. Background Technology

[0002] Dual-polarization weather radar obtains polarization quantities such as differential reflectivity and correlation coefficient from the echoes of horizontal and vertical polarization channels, and classifies the phase states of precipitation particles using methods such as fuzzy logic. This is a common method used by operational weather radar to identify water condensates such as rain, snow, sleet, and hail. The conventional approach is to calculate a single differential reflectivity and correlation coefficient for each range library and consult fuzzy logic criteria to obtain a phase state conclusion for that range library. X-band micro-weather radar is small in size and easy to maneuver and network, but its antenna aperture is small and the beam is wide, resulting in a relatively large single resolution element in space. When the beam crosses regions such as the melting layer, multiple precipitation particles with different falling velocities and phase states often fall into the same resolution element simultaneously. In this case, the correlation coefficient calculated for the entire range library is only a suppressed mixed value, which can only indicate the phase mixing at this point but cannot indicate where they separate, let alone provide information on multiple coexisting phase states within the same resolution element. Furthermore, X-band miniature radars have low transmit power, resulting in a low signal-to-noise ratio for long-range Coulomb echoes, causing fluctuations in correlation coefficient estimation and making misclassification based on fixed thresholds easy. X-band electromagnetic waves also experience two-way attenuation when passing through rain areas, systematically suppressing the absolute magnitude of reflectivity. Classification methods based primarily on absolute reflectivity levels are prone to overall misclassification after heavy precipitation at long distances. Therefore, existing technologies struggle to separate and accurately identify multiple coexisting precipitation particle phases within a single resolution cell. Summary of the Invention

[0003] The main objective of this invention is to provide a method for identifying the phase state of precipitation particles in an X-band dual-polarization micro-weather radar, in order to solve the problem that existing micro-weather radars, due to their small antenna aperture, wide beamwidth, and low transmission power, have difficulty separating and identifying multiple coexisting precipitation particles with different falling velocities and phase states within a single resolution element.

[0004] To solve the above problems, the technical solution of the present invention is implemented as follows: A method for identifying the phase state of precipitation particles in an X-band dual-polarization miniature weather radar. This method is based on an X-band dual-polarization miniature weather radar that detects precipitation by simultaneously transmitting and receiving horizontally and vertically polarized signals. The method includes the following steps: Step 1, Spectral polarization demixing: Scan the detection range at a set pitch angle between the horizontal and vertical directions. For each range library within the detection range, obtain the spectral reflectivity, spectral differential reflectivity, and spectral correlation coefficient distributed along the Doppler velocity channel from the echoes of that range library in the horizontal and vertical polarization channels. Step 2, Separation of mixed phases in the spectral domain: Based on the valley value of the spectral correlation coefficient formed by the first decrease and then increase of the spectral correlation coefficient along the direction of Doppler velocity, the separation boundary is determined. The distance library is divided into multiple spectral components corresponding to a single phase along the direction of Doppler velocity using the separation boundary as the boundary. Step 3, fuzzy discrimination of unmixed components: For each spectral component, obtain the component differential reflectance and component correlation coefficient, and obtain the relative order of each spectral component along the falling velocity. Find the membership degree of each spectral component to each candidate phase state on the membership function and obtain the membership degree. Take the candidate phase state with the largest membership degree as the phase state of the corresponding spectral component. Summarize the phase states of all spectral components in the same distance library into the coexisting phase state combination of the distance library. Piece together the coexisting phase state combinations of all distance libraries according to azimuth, distance and set elevation angle to obtain the spatial distribution of precipitation particle phase states.

[0005] Furthermore, the beamwidth corresponding to the antenna aperture of the X-band dual-polarization micro-weather radar allows a single resolution element to accommodate precipitation particles of various phases. The transmit power of the X-band dual-polarization micro-weather radar ensures that the signal-to-noise ratio of the single-channel echo from the long-range radar is at a level that causes fluctuations in the estimation of the spectral correlation coefficient.

[0006] Furthermore, the process of obtaining spectral reflectance, spectral differential reflectance, and spectral correlation coefficient in step 1 includes: continuously acquiring multiple sets of horizontally polarized time series and multiple sets of vertically polarized time series from the same distance library within a dwell time; applying window functions to each set of horizontally polarized time series and each set of vertically polarized time series to suppress spectral leakage; and then performing Doppler spectral transformation on each set; averaging the Doppler power spectrum obtained from each set of horizontally polarized time series along the Doppler velocity channel to obtain the average power spectrum of the horizontal polarization channel; and averaging the Doppler power spectrum obtained from each set of vertically polarized time series along the Doppler velocity channel. The average power spectrum of the vertical polarization channel is obtained by averaging the Doppler velocity channels. The cross spectrum of the echo complex signals of each set of horizontally polarized time series and the same set of vertically polarized time series is calculated, and the cross spectra of each set are averaged along the Doppler velocity channels to obtain the average cross spectrum. The spectral reflectivity is obtained from the average power spectrum of the horizontal polarization channel, and the spectral differential reflectivity is obtained from the average power spectra of the horizontal and vertical polarization channels. The spectral correlation coefficient is obtained from the average cross spectrum together with the average power spectra of the horizontal and vertical polarization channels. The noise level of the same distance library is estimated from the Doppler velocity channels containing only background noise.

[0007] Furthermore, the Doppler power spectrum uses the Doppler velocity channel as the horizontal axis and the echo power as the vertical axis. Each Doppler velocity channel corresponds to a Doppler velocity value determined by the radar wavelength and pulse repetition mode. When the Doppler velocity of precipitation particles exceeds the measurable velocity range of the radar, the Doppler power spectrum folds at the velocity boundary. A folded spectrum segment, which is attached to one velocity boundary and separated from the main spectrum segment near the other velocity boundary by a region containing only noise level, is shifted along the Doppler velocity axis by the span of a measurable velocity range, so that the folded spectrum segment is continuously connected to the main spectrum segment.

[0008] Further, step 2 includes: within each range library, connecting Doppler velocity channels with spectral reflectance higher than the noise level into a continuous region, which serves as the Doppler velocity range occupied by the precipitation particle swarm within that range library; within the Doppler velocity range, identifying a local minimum formed by the spectral correlation coefficient first decreasing and then increasing along the Doppler velocity direction; when the decrease in this local minimum relative to the adjacent local maximum exceeds the estimated fluctuation of the spectral correlation coefficient itself, and the spectral differential reflectance takes different values ​​on both sides of this local minimum, or the spectral correlation coefficient takes different values ​​on both sides of this local minimum, then confirming this local minimum. The valley of the spectral correlation coefficient is considered, and the Doppler velocity channel corresponding to the valley of the spectral correlation coefficient is determined as the separation boundary. The valley of the spectral correlation coefficient corresponds to the boundary between two types of precipitation particles with similar falling velocities but different phases. When the spectral differential reflectance takes similar values ​​on both sides of the local minimum, and the spectral correlation coefficient takes similar values ​​on both sides of the local minimum, the local minimum is retained within the same spectral component. The endpoint of the Doppler velocity range and all the obtained separation boundaries are used as the boundary. The distance library is divided into multiple adjacent velocity intervals along the Doppler velocity, and each velocity interval is a spectral component.

[0009] Furthermore, under the condition that the horizontal wind is approximately uniform within the resolution corresponding to the distance library, the horizontal wind applies the same Doppler velocity shift to all precipitation particles within the resolution, causing the spectrum of Doppler velocity variation within the distance library to be translated as a whole along the Doppler velocity axis, while the relative order and relative spacing of all separation boundaries and multiple spectral components along the Doppler velocity axis remain unchanged.

[0010] Further, step 3 includes: for each spectral component, summing the echo power of the average power spectrum of the horizontal polarization channel in each Doppler velocity channel of the velocity interval occupied by the spectral component along the Doppler velocity to obtain the component reflectivity of the spectral component, the component reflectivity characterizing the echo strength of the precipitation particle swarm corresponding to the corresponding spectral component; summing the echo power of the average power spectrum of the horizontal polarization channel and the vertical polarization channel in each Doppler velocity channel of the velocity interval along the Doppler velocity respectively, and obtaining the component differential reflectivity of the spectral component from the summation result; merging the values ​​of the average cross spectrum in each Doppler velocity channel of the velocity interval along the Doppler velocity, and obtaining the component correlation coefficient of the spectral component from the merged result together with the summation result of the two.

[0011] Furthermore, under a set pitch angle, the falling precipitation particles generate Doppler velocity components on the Doppler velocity axis. Precipitation particles falling faster tend to be biased towards the same end on the Doppler velocity axis, and the direction of the biased end is taken as the definite direction. Within the same distance, the spectral component corresponding to the faster falling precipitation particles is closer to the end of the definite direction, while the spectral component corresponding to the slower falling precipitation particles is closer to the end opposite to the definite direction. Under the condition that the horizontal wind within the resolution corresponding to the distance is approximately uniform, the horizontal wind applies the same Doppler velocity offset to each spectral component within the same distance, and the relative order of each spectral component along the definite direction remains unchanged. Thus, the relative order of each spectral component along the falling velocity within the same distance is obtained.

[0012] Furthermore, membership functions for six candidate phases—rain, wet snow, dry snow, ice crystals, graupel, and hail—are established based on the relative order of component differential reflectivity, component correlation coefficient, and velocity along the fall, respectively, to accommodate the influence of X-band two-way propagation through rain zones on the magnitude of reflectivity. Regarding component differential reflectivity, the membership degree of rain and ice crystals increases as the component differential reflectivity is greater than 0, while the membership degree of wet snow, dry snow, graupel, and hail increases as the component differential reflectivity approaches 0. Regarding component correlation coefficient, the membership degree of rain, dry snow, ice crystals, and graupel increases as the component correlation coefficient approaches 1. The membership degree of wet snow and hail increases as the component correlation coefficient decreases relative to spectral components close to 1 in the same distance library; along the relative order of falling velocity, the membership degree of rain, hail and graupel increases as the corresponding spectral component approaches the end of a defined direction, while the membership degree of dry snow, wet snow and ice crystals increases as the corresponding spectral component approaches the end opposite to the defined direction; the candidate phases of rain and ice crystals, dry snow and graupel, and wet snow and hail are similar to each other in terms of component differential reflectance and component correlation coefficient, and the two candidate phases in each pair are distinguished by the relative order of falling velocity.

[0013] Furthermore, for each spectral component, the membership degree to which the spectral component belongs to each candidate phase is obtained from the membership function. For the same candidate phase, the membership degrees obtained from the component differential reflectivity, component correlation coefficient and relative order along the falling velocity are multiplied to obtain the degree of belonging. The candidate phase with the highest degree of belonging is taken as the phase of the precipitation particle swarm corresponding to the spectral component, and the velocity range occupied by the spectral component is taken as the falling velocity range of the precipitation particles corresponding to the spectral component. Large particles that fall faster and whose polarization characteristics change when entering the resonance region in the X-band are separated into independent spectral components near one end of a certain direction, so that the polarization characteristics of the large particles can be distinguished within the velocity range occupied by the independent spectral components. When one spectral component is obtained in the range library, the coexisting phase combination in the range library contains one type of phase.

[0014] The precipitation particle phase identification method of the X-band dual-polarization micro weather radar of the present invention has the following beneficial effects: This invention detects precipitation particles at a set elevation angle between the horizontal and vertical directions, projecting the particle velocity onto the Doppler velocity axis while retaining considerable spectral differential reflectivity. This allows precipitation particles of various phases falling into the same resolution element to be distributed to different positions on the Doppler velocity axis according to their falling velocity. Furthermore, the separation boundary is determined based on the valley value formed by the spectral correlation coefficient decreasing and then increasing along the Doppler velocity. This separates the range library, which previously only provided mixed conclusions, into multiple spectral components corresponding to individual phases, solving the problem of overlapping and difficult-to-separate phases within a single resolution element of a wide-beam micro-radar. This invention sets an adaptive fluctuation threshold for the valley value based on echo quality and distinguishes between the valley value and the false dips caused by turbulence, wind shear, and estimated fluctuations, maintaining robustness even under conditions of low transmit power and low signal-to-noise ratio. This invention extracts the component differential reflectance and component correlation coefficient for each spectral component, and performs fuzzy discrimination by combining the relative order of each spectral component along the falling velocity. This makes the discrimination focus fall on quantities that are not sensitive to attenuation, avoiding the influence of X-band attenuation through rain areas on the absolute order of reflectance. Moreover, the relative order along the falling velocity is not affected by unknown horizontal winds. Finally, it gives multiple coexisting phase states within the same distance database and stitches them together to form the spatial distribution of phase states, thus improving the accuracy of phase state identification for micro-sized weather radar. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of scanning geometry at a set pitch angle provided in an embodiment of the present invention; Figure 2 A schematic diagram illustrating the projection relationship of Doppler velocities provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the spectral reflectance distribution along the Doppler velocity, provided in an embodiment of the present invention. Figure 4 A schematic diagram illustrating the determination of separation boundaries and division of spectral components based on the valley values ​​of the spectral correlation coefficient provided in an embodiment of the present invention; Figure 5 A schematic diagram of the valley depth and adaptive fluctuation threshold of the spectral correlation coefficient valley value provided in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the difference in spectral reflectance on both sides of the valley value of the spectral correlation coefficient provided in an embodiment of the present invention. Figure 7 This is a schematic diagram illustrating how spectral components distinguish similar phase pairs along the relative order of falling velocity, as provided in an embodiment of the present invention. Figure 8 This is a schematic diagram of the membership functions used for spectral component fuzzy discrimination provided in an embodiment of the present invention, wherein (a) is the membership function on the component differential reflectivity, and (b) is the membership function on the component correlation coefficient. Detailed Implementation

[0016] A method for identifying the phase state of precipitation particles in an X-band dual-polarization miniature weather radar. This method is based on an X-band dual-polarization miniature weather radar that detects precipitation by simultaneously transmitting and receiving horizontally and vertically polarized signals. The method includes the following steps: Step 1, Spectral polarization demixing: Scan the detection range at a set pitch angle between the horizontal and vertical directions. For each range library within the detection range, obtain the spectral reflectivity, spectral differential reflectivity, and spectral correlation coefficient distributed along the Doppler velocity channel from the echoes of that range library in the horizontal and vertical polarization channels. Step 2, Separation of mixed phases in the spectral domain: Based on the valley value of the spectral correlation coefficient formed by the first decrease and then increase of the spectral correlation coefficient along the direction of Doppler velocity, the separation boundary is determined. The distance library is divided into multiple spectral components corresponding to a single phase along the direction of Doppler velocity using the separation boundary as the boundary. Step 3, fuzzy discrimination of unmixed components: For each spectral component, obtain the component differential reflectance and component correlation coefficient, and obtain the relative order of each spectral component along the falling velocity. Find the membership degree of each spectral component to each candidate phase state on the membership function and obtain the membership degree. Take the candidate phase state with the largest membership degree as the phase state of the corresponding spectral component. Summarize the phase states of all spectral components in the same distance library into the coexisting phase state combination of the distance library. Piece together the coexisting phase state combinations of all distance libraries according to azimuth, distance and set elevation angle to obtain the spatial distribution of precipitation particle phase states.

[0017] In practical implementation, the X-band dual-polarization miniature weather radar operates by simultaneously transmitting and receiving horizontally and vertically polarized signals. For example, its center frequency is 9.4 GHz, corresponding to a wavelength of... Approximately 0.032m; pulse repetition frequency Use 2000Hz; the distance from the library is 30m. These values ​​are not unique; alternative ranges are given below at key steps.

[0018] Spectral polarization demixing first requires determining the elevation angle of the scan. Setting the elevation angle between horizontal and vertical, rather than the low elevation angle of conventional weather radar for horizontal scanning, is determined by two mutually constraining requirements. For example... Figure 1 As shown, the radar scans obliquely at a set elevation angle. Due to the small antenna aperture and wide beam, each individual resolution element is relatively large. When the beam crosses the melt layer, the upper part is snow and the lower part is rain. Multiple precipitation particles with different falling velocities and phases simultaneously fall onto the same resolution element. The Doppler velocities of precipitation particles within the reservoir at the same distance, as measured by the radar, are as follows: Figure 2 The figure can be written as ,in Radial velocity, This refers to the falling speed (vertically downward). To determine the component of the horizontal wind along the radar azimuth at the location of the target. To set the pitch angle. It can be seen that the contribution of the fall velocity to the Doppler velocity is... The contribution of coefficients and horizontal wind. Let be the coefficient. If Approaching 0 degrees Celsius When the velocity approaches zero, particles with different falling velocities have almost the same radial velocity, crowding together and unable to be separated according to their falling velocity, thus losing the prerequisite for spectral separation; if Approaching 90 degrees, while the falling velocity can be fully projected, when the beam is incident perpendicularly on an ellipsoidal raindrop, its polarization scales tend to be consistent, the echo power of the two channels is comparable, and the spectral differential reflectivity tends to be 0, resulting in the loss of differential information. By taking the elevation angle between these two extremes, the falling velocity can be sufficiently projected for differentiation, while the beam is incident obliquely on the particle, thus preserving a considerable spectral differential reflectivity. This embodiment takes... At 45 degrees, and Both are approximately 0.707; the projection of a raindrop with a falling velocity of 6 m / s is approximately 4.2 m / s, and the projection of dry snow with a falling velocity of 1 m / s is approximately 0.7 m / s, a difference of approximately 3.5 m / s. (Alternative locations are listed.) The angle can be selected between 30 and 60 degrees. A smaller angle is more conducive to preserving the spectral differential reflectance and weakening the velocity projection, while a larger angle has the opposite effect. The angle can be determined by combining the velocity distribution of the precipitation type of interest with the local wind field. Alternatively, it can be performed separately at several adjacent pitch angles to verify each other.

[0019] After determining the elevation angle, the radar stays at that elevation angle for a certain period of time. The radar scans the detection range. During its dwell time, the radar operates at a pulse repetition frequency. Continuous launch Each transmitted pulse is received; during reception, the echo of each transmitted pulse is sampled at the time delay corresponding to each range library. Thus, the sampling points of the same range library on consecutive pulses are arranged along the pulse direction, forming a time series of that range library. This... The sampling points are divided into equal groups according to the order. Groups, each group contains Each sampling point was obtained, and the horizontal and vertical polarization channels were respectively... Group of horizontal polarization time series and A set of vertical polarization time series. This embodiment takes... For 256, If it is 8, then Approximately 1.024 seconds.

[0020] Next, a Doppler spectral transform is performed on each time series, with a window function applied to each sample before the transform. When directly transforming a time series of finite length, strong echo components will diffuse into other channels through sidelobes. Spectral polarization demixing requires resolving several velocity components with significantly different amplitudes within the same distance range; without suppression, weak components will be overwhelmed by the sidelobes of strong components. Let the first... Group of horizontal polarization time series in pulse sequence number The echo complex signal at the location is After adding a window, perform the transformation: In the formula For the first Group of horizontal polarization channels in channel The complex spectrum on, For window functions, The imaginary unit, For a single group of points, The Doppler velocity channel number is used; the vertical polarization channel is obtained similarly. This embodiment uses a Blackman window. Its highest sidelobe is about 58 dB lower than the main lobe, and it has a strong ability to isolate adjacent velocity components with large amplitude differences. The trade-off is a slightly wider main lobe and a slight loss of resolution. Alternatively, a Hanning window or Hamming window, which has a narrower main lobe and higher resolution but weaker sidelobe suppression, can be used. The above transformation can be implemented in engineering by Fast Fourier Transform.

[0021] Each Doppler velocity channel Corresponding to a specific Doppler velocity ;in For channel The corresponding Doppler velocity, For Doppler velocity resolution, substitute the values ​​from this embodiment. m / s, total number of Doppler velocity channels 256. The aforementioned difference between rain and dry snow is approximately 3.5 m / s, equivalent to about 28 channels, which is sufficient to represent different channels under this set of parameters. Maximum measurable velocity m / s, measurable velocity range is m / s to m / s. To increase resolution, you can increase... However, a fixed residence time reduces the number of groups available for averaging. A compromise between the two requires combining spectral estimation to determine stability.

[0022] After obtaining the complex spectra of each group, the power spectrum is calculated separately and averaged between groups. For the horizontal polarization channel, the... Group power spectrum The average power spectrum is Similarly, the vertical polarization channel is obtained. Power spectrum With average power spectrum All values ​​are plotted with the Doppler velocity channel on the x-axis and echo power on the y-axis. Averaging across multiple groups is used because precipitation echoes are random signals, and the power spectrum of a single group fluctuates greatly with its standard deviation being on the same order of magnitude as the mean; directly calculating the spectral differential reflectivity and spectral correlation coefficient based on this would be highly unstable. Averaging the power spectra of the independent power spectra can reduce the estimated normalized variance by approximately The ratio of low signal-to-noise ratio is particularly important for micro-radars with low transmission power and already low signal-to-noise ratio.

[0023] The cross-spectrum required for the spectral correlation coefficient must be averaged in a complex form while preserving phase, which differs from power spectral averaging. cross spectrum of groups ; superscript Represents complex conjugation; the average cross spectrum is obtained by averaging the cross spectra of each group among the groups. The reason for retaining the complex number first and then averaging is that the cross-spectral phase of the uncorrelated echoes between the horizontal and vertical channels is randomly distributed among the groups. Averaging the complex number makes them cancel each other out and only retains the phase-stable correlated components. If the amplitude is taken first and then averaged, the random phase terms cannot be canceled out, and the spectral correlation coefficient will be systematically overestimated and lose the ability to distinguish phase states.

[0024] Based on this, three spectral polarization quantities distributed along the Doppler velocity channels are given. The spectral reflectivity is obtained from the average power spectrum of the horizontally polarized channel after distance correction and calibration. ;in For channel Spectral reflectance (linear value) at [location]. Radar constants determined for system calibration (combined transmit power, antenna gain, wavelength, pulse width, and receive gain, etc.). This is the distance from the radar to the distance storage. The noise level of the horizontal polarization channel; multiplied by Compensate for the two-way spread attenuation of echo power with distance, subtract In engineering, noise floor elevation is often addressed using logarithmic methods. (Unit: dBZ) The spectral differential reflectance is the logarithm of the ratio of the horizontal and vertical average power spectra in the same channel: ;in For channel Spectral differential reflectance at a given location (in dB). This represents the noise level of the vertical polarization channel. The system differential reflectivity bias (in dB, obtained by calibration and subtracted) is due to the inconsistency in the gains of the two channels. The spectral correlation coefficient is the normalization of the amplitude of the average cross spectrum over the geometric mean of the two denoised average power spectra. ;in For channel The spectral correlation coefficient at the point is between 0 and 1. The noise level is deducted from the denominator, but not from the numerator, because the noise in the two receiving channels is independent, and their cross spectrum tends to 0 after averaging multiple sets. The average cross spectrum itself does not contain noise bias. However, the noise in the average power spectrum is added with positive values. If it is not deducted from the denominator, the spectral correlation coefficient will be suppressed in the channel with a low signal-to-noise ratio, and the spectral differential reflectance will be pulled to 0, thus masking the true polarization characteristics.

[0025] All of the above formulas require noise level. and This embodiment estimates the noise level from the Doppler velocity channels containing only background noise using an objective segmentation method: all Doppler velocity channels of a given channel are arranged in ascending order of power, and candidates are gradually included in the candidate noise set starting from the lowest power. For each candidate included, it is checked whether it satisfies the statistical relationships expected of white noise. ;in and , respectively, represent the mean and variance of the power within the candidate set. The number of groups participating in the average, for the time... The normalized variance of the white noise after group averaging should be equal to The inclusion process stops when the candidate set exactly satisfies this relation, using the mean value at that point. As the noise level, it is obtained for both the horizontal and vertical channels respectively. , Those channels whose power is significantly higher than the noise level and are continuous with each other constitute the Doppler velocity range occupied by the precipitation particle swarm in that distance reservoir. Alternatively, the average of the lowest power channels (e.g., the lowest 10%) can be taken as the noise level, or measurements can be taken individually on distant, precipitation-free distance reservoirs and then used for near-distance reservoirs.

[0026] In actual observations, if the Doppler velocity of precipitation particles exceeds... m / s to Within the measurable velocity range of m / s, the average power spectrum folds at the velocity boundaries. This is because sampling at the pulse repetition frequency causes the Doppler spectrum to... For periodic repetition, the velocity distribution crossing the boundary is folded back from one side of the boundary to the other. This manifests as a spectrum closely adhering to one side of the velocity boundary, separated from the main spectrum segment adhering to the other side by a region containing only noise levels. After identifying this folded spectrum segment, it is shifted along the Doppler velocity axis by the span of a measurable velocity range (i.e., (m / s) to make it continuous with the main spectrum segment, restoring the true falling velocity distribution. Whether there is a noise-only interval between the two spectrum segments is the basis for determining that it is indeed a fold rather than two closely spaced true falling velocity components; when the two spectrum segments are adjacent and there is no noise interval between them, this criterion is no longer applicable, and such corner cases can be handled separately in implementation.

[0027] Thus, for each range library within the detection range, three curves—spectral reflectance, spectral differential reflectance, and spectral correlation coefficient—are obtained along the Doppler velocity channel, providing input for subsequent separation of mixed phases and component-by-component discrimination. As an alternative implementation, before calculating the aforementioned spectral polarization quantities, a notch filter can be placed near the zero Doppler velocity to suppress ground clutter; alternatively, the average power spectrum and average cross spectrum can be moderately smoothed between several adjacent range libraries or several adjacent Doppler velocity channels to stabilize the spectral correlation coefficient estimation, at the cost of correspondingly reducing the range or Doppler velocity resolution; when it is necessary to expand the measurable velocity range without folding, staggered pulse repetition frequency transmission can be used as an alternative to folding and shifting.

[0028] Following step 1, the spectral reflectance of each distance has been obtained along the Doppler velocity channel. Spectral differential reflectance Spectral correlation coefficient These three curves have been used to estimate the noise levels of the horizontal and vertical polarization channels. and ,in This refers to the Doppler velocity channel number. Step 2 handles the overlapping of precipitation particles of various phases with different falling velocities within the same range library. The reason for separating them in the Doppler velocity domain is that the single resolution of a small radar with a wide beam is very large, often simultaneously covering several types of particles with significantly different falling velocities; the conventional method of obtaining a single spectral correlation coefficient for the entire range library only yields a suppressed mixed value, indicating mixing but not indicating where to separate them. However, at a set elevation angle, particles with different falling velocities are projected onto different positions on the Doppler velocity axis (guaranteed by the projection relationship in Step 1). By examining the spectral correlation coefficient point by point along the Doppler velocity channel, the boundary between the two phases can be located and separated accordingly.

[0029] Before proceeding with separation, first define the Doppler velocity range actually occupied by precipitation echoes within the reservoir at that distance. The criterion for whether a channel contains precipitation echoes is based on how much it exceeds the noise level: via The standard deviation of the average power fluctuation of the channels containing only noise after incoherent averaging is approximately Therefore, when the average power of the horizontal polarization channel satisfies It was initially believed that the channel contained precipitation echoes, among which For the horizontal polarization channel in the channel Average power on This represents the number of groups participating in the average (8 in this example). Substitute the values... Approximately 1.06, which is the requirement higher Approximately 3.1 decibels. Connect adjacent channels that meet this condition to form a continuous region, and denot the left and right channel numbers as follows: , , interval This refers to the Doppler velocity range occupied by the precipitation particle swarm within the reservoir at that distance; isolated peaks above the threshold and separated on both sides by channels containing only noise are removed as noise spikes.

[0030] Subsequently, the spectral correlation coefficient curve was examined along the Doppler velocity channel within the Doppler velocity range to find the local minimum formed by its initial decrease followed by an increase. The physical basis for using the valley value to mark the phase boundary is that the spectral correlation coefficient reflects the consistency of backscattering in both polarization directions of all particles falling into the same channel. When particles in a channel belong to the same phase state, their shape, orientation, and dielectric properties are statistically consistent, the two polarization echoes are in stable proportion and phase coherent, and the spectral correlation coefficient is close to 1. When particles of two phase states fall into the same channel at similar falling velocities (such as low-density large snowflakes and small raindrops having similar falling velocities), their polarization scattering characteristics are quite different, the coherence of the combined echoes from the two channels decreases, and the spectral correlation coefficient decreases accordingly. The closer the echo power contributed by the two phase states is to being equal, the more significant the decrease in coherence. Therefore, the deepest valley value corresponds precisely to the transition position where the two phase states are evenly matched, that is, their boundary.

[0031] The presence of a single local minimum is insufficient to determine a phase boundary, because the spectral correlation coefficient is estimated from a finite sample and exhibits statistical fluctuations. Even within a single phase state, subtle local dips can occur due to these fluctuations. To distinguish true phase boundary dips from random fluctuations, an adaptive fluctuation threshold is set for each local minimum. The standard deviation of the fluctuation in the spectral correlation coefficient estimate can be estimated using the following formula: In the formula For channel The standard deviation of the spectral correlation coefficient estimate Let be the signal-to-noise ratio (SNR) of the horizontally polarized echo in this channel. This formula shows that the closer the spectral correlation coefficient is to 1, the more stable the estimate and the smaller the fluctuations; the lower the SNR, the larger the fluctuations. The latter point is particularly crucial for micro-radars with low transmit power and inherently low SNR, as it allows the threshold to automatically adjust channel-by-channel according to echo quality rather than taking a fixed value, widening it at weak echoes and tightening it at strong echoes, neither missing the true phase boundary nor being misled by noise dips. For a channel located... For a local minimum, find the nearest local maximum on both sides of it, and denote them as channels. and ,like Figure 5 As shown, the valley depth is defined as ;when It was then believed that the depression exceeded the estimated fluctuations, among which As a confidence factor, this embodiment uses 3, which corresponds to approximately 3 times the standard deviation in confidence.

[0032] Using the fluctuation threshold only eliminates random fluctuations, but it cannot rule out another type of true depression that is not a phase boundary: turbulence or wind shear within the beam can broaden the spectrum and lower the spectral correlation coefficient, creating depressions exceeding the fluctuation threshold even within a single phase state. Wide beams and long-range columbariums are precisely where turbulence and shear are most severe. Therefore, a further screening is performed on the local minima that pass the fluctuation threshold to examine whether the polarization characteristics on both sides of the depression are truly different. For example... Figure 6 As shown, the basis for this is: if the two sides are indeed two different phases, then the spectral differential reflectance will exhibit different orders of magnitude on both sides, or the spectral correlation coefficient will be different on both sides; if the two sides are the same phase and the depression is only caused by turbulence or shear, then the spectral differential reflectance and spectral correlation coefficient should be similar on both sides. Specifically, in the channel and Take a neighborhood segment from each channel (e.g., 3 to 5 adjacent channels) and calculate the average spectral differential reflectance, denoted as . and And estimate the fluctuations on both sides according to the following formula, and then synthesize them into the difference test threshold: In the formula For channel The standard deviation (in dB) of the differential reflectance estimate is 4.343. This formula is used to convert relative fluctuations expressed as the natural logarithm into decibels, and it holds true when the spectral correlation coefficient is high. When the difference in reflectance between the two spectral sides satisfies... or the correlation coefficient level of both sides of the spectrum and When the difference exceeds its corresponding fluctuation, it is determined that the two sides are two types of particles with different phase states, and the channel is closed. The trough of the spectral correlation coefficient is identified, and its corresponding Doppler velocity channel is determined as the separation boundary. Conversely, when the spectral differential reflectance on both sides is close to the spectral correlation coefficient level, the depression is determined to originate from turbulence or wind shear and is retained within the same spectral component. Adding a comparison of the spectral correlation coefficient level in addition to the spectral differential reflectance is to address adjacent phases such as rain and hail, where both sides have low spectral differential reflectance and are difficult to distinguish based on spectral differential reflectance alone. In this case, the difference in the spectral correlation coefficient level can serve as a redundant criterion to avoid missed detections.

[0033] like Figure 3 As shown, suppose the spectral reflectance of a certain distance exhibits a bimodal structure, with the lower Doppler velocity side representing slower-falling particles and the higher velocity side representing faster-falling particles. According to the correspondence in step 1... (Unit: meters per second) The lower speed side is approximately 0.5 to 1.5 meters per second, corresponding to channels 132 to 140; the higher speed side is approximately 2.5 to 4.5 meters per second, corresponding to channels 148 to 164. A trough appears between the two peaks at 2.0 meters per second, i.e., channel 144. Figure 4 As shown, the spectral correlation coefficient approaches 1 at both peaks and dips to a trough at channel 144, which is the separation boundary. Based on this, the distance library is divided into spectral component 1 and spectral component 2. The local maximum on the left is at channel 138 with a spectral correlation coefficient of 0.975 and a spectral differential reflectance of 0.3 dB. The local maximum on the right is at channel 152 with a spectral correlation coefficient of 0.98 and a spectral differential reflectance of 2.0 dB. The spectral correlation coefficient drops to 0.86 at the trough at channel 144. Fluctuation threshold: when the spectral correlation coefficient is 0.975 and the signal-to-noise ratio is high. Approximately 0.0125, Approximately 0.037, valley depth Exceeding the threshold, a true concave shape is observed. Bilateral differentiation: Difference in spectral differential reflectance. The corresponding synthesis threshold is approximately 1.0 dB. Exceeding this threshold, the two sides are determined to be of different phases, and channel 144 is designated as the separation boundary. As a contrast, another distance contains only rapidly falling particles of the same type. However, due to turbulence at channel 144, the spectral correlation coefficient is reduced from 0.98 to 0.92, and the differential reflectance on both sides is 2.0 and 2.1 dB respectively. Although the valley depth of 0.06 exceeds the fluctuation threshold, the difference in differential reflectance on both sides is only 0.1 dB, far below the synthesis threshold of approximately 1.0 dB. Therefore, it is judged as a turbulence-induced depression, and no separation boundary is set. This pair of positive and negative examples illustrates that both the fluctuation threshold and the differentiation between the two sides are indispensable.

[0034] After determining all separation boundaries, set the two endpoints of the Doppler velocity range. , Together with all separation boundaries, this serves as the dividing line, along the Doppler velocity. The channel is sequentially divided into multiple adjacent velocity intervals, ensuring each channel belongs to a unique interval (the channel containing the separation boundary is assigned to its right interval to guarantee no overlap or omission). Each velocity interval is thus a spectral component. In the example above, only channel 144 has a separation boundary, resulting in two spectral components: For those who fall more slowly For those falling faster, their respective phase states will be determined later component by component. If no local minimum is confirmed within the Doppler velocity range, the entire range is considered a single velocity interval, corresponding to a single spectral component, indicating only one phase state.

[0035] The separation results are robust to horizontal wind, which is an advantage of placing the separation in the Doppler velocity domain and treating it as a relative position. When the horizontal wind within the resolution volume is approximately uniform, the component of the horizontal wind along the radar azimuth is... After projection, the same Doppler velocity shift is superimposed on all precipitation particles, with a translation amount of... ,in The component of horizontal wind along the radar azimuth, To set the pitch angle, For Doppler velocity resolution. When 2 meters per second The translation amount at 45 degrees is approximately Each channel, that is, the entire spectrum, along with the valleys and each separation boundary, is shifted by about 11 channels. Since all spectral components are moved as a whole by the same shift, their relative order and relative intervals remain unchanged, and the position of the separation boundary relative to each spectral component also remains unchanged. Therefore, separation does not require prior knowledge or deduction of horizontal wind. What turbulence and shear destroy is precisely the premise of the same shift within the resolution body, which has been blocked outside the separation boundary by the two sides.

[0036] As an alternative implementation, a moving average of three Doppler velocity channels can be applied to the spectral correlation coefficient curve before detecting local minima to suppress fragmented false minima caused by estimation fluctuations, at the cost of slightly dulling the true valley values; in addition to taking the neighborhood average, the neighborhood median can also be used to weaken the influence of individual abnormal channels; confidence factor In addition to using 3, the value can be appropriately increased in volume scans with a generally low signal-to-noise ratio to more strictly suppress phase missegmentation. To avoid mistakenly cutting a relatively wide transition region into two adjacent separation boundaries, the minimum channel spacing between two adjacent valleys can be limited to retain the two closest local minimum depths. The width of the lateral neighborhoods can also be adjusted according to the Doppler velocity resolution and the degree of spectral broadening. All of the above methods fall within the process of determining the separation boundary based on the valley value of the spectral correlation coefficient, and then dividing the range library into multiple spectral components based on the separation boundary. The resulting multiple spectral components are then used as the objects for subsequent component-by-component phase discrimination.

[0037] Following step 2, for each distance, the Doppler velocity range occupied by the precipitation particle swarm has been determined along the Doppler velocity. It is divided into several adjacent velocity intervals, each velocity interval being a spectral component. The sequence number remains the same as the Doppler velocity channel number. Step 3 determines the phase state corresponding to each spectral component. This differs from conventional methods in that conventional hydrogel classification calculates the single differential reflectance and correlation coefficient for the entire range library and then uses fuzzy logic criteria. Ranges falling into a mixed phase category can only be classified as a single phase or a general mixed category. However, after the separation in Step 2, the spectral components here are already separated, each containing only a single phase. Polarization features can be extracted for each spectral component before further discrimination, ultimately providing multiple coexisting phase states within the same range library.

[0038] First, extract three types of quantities for discrimination from each spectral component. Then, for the occupies the velocity range... spectral components, of which , Given the left and right channel numbers of this interval, sum the average power of the horizontally polarized channels after denoising along the Doppler velocity to obtain the component reflectivity: In the formula For the first The component reflectance (linear value) of each spectral component. For the horizontal polarization channel in the channel Average power on This represents the noise level of the horizontal polarization channel. For radar constants, This represents the distance from the radar to the distance source. The summation over the interval, rather than taking the value of a single channel, is because particles of the same phase are distributed across multiple channels according to their falling velocity; accumulating their power restores the overall echo strength of this type of particle. Component reflectivity is mainly used to define velocity intervals and assist in discrimination, not as the primary criterion. Similarly, the average power after denoising of the horizontal and vertical polarization channels within the interval is summed separately, and the logarithm of their ratio yields the component differential reflectivity: In the formula For the first Component differential reflectance (in decibels) of each spectral component. For vertical polarization channels in channel Average power on This represents the noise level of the vertical polarization channel. This is the system differential reflectance bias (in decibels, consistent with the one deducted in step 1). Component differential reflectance reflects the statistical flattening of particles within that spectral component: flattened raindrops have a larger positive value, while nearly spherical graupel, hail, or low-density dry snow and ice crystals have a value close to 0.

[0039] The component correlation coefficient needs to be obtained by using the average cross spectrum that preserves the phase in step 1. The calculation method does not simply average the correlation coefficients of each channel spectrum, but rather first sums the average cross spectrum of the complex spectrum and the two denoised average powers within the interval, and then normalizes them. In the formula For the first The component correlation coefficient of each spectral component takes a value between 0 and 1. The numerator is the amplitude of the sum of the average cross spectra within the interval, and the denominator is the geometric mean of the sum of the denoising power of the two channels. The method of summing the complex cross spectra first and then taking the amplitude, rather than taking the amplitude of each channel first and then averaging, is similar to the complex averaging of the cross spectra in step 1: the cross spectra of each channel in the same phase are stable and consistent, and the in-phase superposition is preserved during complex summation. If the amplitude is taken first, phase information is discarded, and the overall coherence cannot be reflected. A high component correlation coefficient indicates that the particle morphology within that spectral component is consistent and the phase is simple. Hail and wet snow, due to surface water inclusions or irregular shapes, have reduced coherence between the two polarized echoes, resulting in relatively low component correlation coefficients.

[0040] Next, the relative order of each spectral component along the fall velocity is determined, which is a key step that distinguishes this step from conventional classification. Absolute fall velocity is not used directly because Doppler velocity contains both fall velocity and horizontal wind components. The absolute position of a single spectral component can be shifted by the unknown horizontal wind, and inferring absolute fall velocity from absolute Doppler velocity is unreliable. Step 2 has shown that the horizontal wind exerts the same offset on all spectral components within the same distance range, and the relative order and interval of each spectral component remain unchanged. Therefore, the relative speeds between spectral components are robust and independent of the horizontal wind. Based on this, it is agreed that: when a water particle falls at a set pitch angle, its radial velocity increases along a certain defined direction; this direction is taken as the defined direction. Within the same distance range, the spectral component corresponding to a faster-falling particle is closer to the defined direction, and the slower-falling particle is closer to the opposite end. In practice, simply sorting along the defined direction using the center channel of each spectral component's velocity interval yields the relative order of each spectral component along the fall velocity within that distance range, without any prior knowledge of horizontal wind. Using a numerical example from Step 2: this distance range has two spectral components, each occupying... (Central passageway 137) and (Center approximately 156); Taking the direction of increasing Doppler velocity as the determining direction, the latter is the one falling faster and the former is the one falling slower, and the relative order of their falling speeds is thus determined.

[0041] After obtaining the component differential reflectance, component correlation coefficient, and their relative order along the falling velocity for each spectral component, the phase state is determined using fuzzy discrimination. Fuzzy discrimination, rather than hard thresholding, is used because the polarization characteristics of each phase state overlap numerically and have no clear boundaries. Using membership degrees with continuous values ​​between 0 and 1 to characterize the degree of support of the feature values ​​for each phase state is more in line with physical reality and more tolerant of estimation fluctuations than hard thresholding. For the six candidate phase states—rain, wet snow, dry snow, ice crystals, graupel, and hail—membership functions are established on the three features: component differential reflectance, component correlation coefficient, and relative order along the falling velocity. The focus of the discrimination is on component differential reflectivity, component correlation coefficient, and the relative order along the falling velocity, with component reflectivity used as an auxiliary factor: X-band experiences double-path attenuation and differential attenuation when passing through rain areas, which systematically reduces the absolute magnitude of component reflectivity, especially after heavy precipitation at a distance. Using its absolute level as the main criterion is prone to overall misclassification. Component differential reflectivity is the ratio of the two channels, and most of it cancels out when the attenuation is divided by the same path. The component correlation coefficient is a normalized quantity and is unrelated to the absolute strength of the echo. The relative order along the falling velocity depends only on the order between spectral components and is unrelated to the power level. All three are insensitive to attenuation. Using them as the main criteria is precisely to avoid the destruction of absolute reflectivity by X-band attenuation.

[0042] Each membership function is given in trapezoidal form as piecewise straight lines, which is easy to implement and has clear boundaries, such as Figure 8 As shown. Figure 8 In the diagram, (a) represents the membership function on the component differential reflectance, and (b) represents the membership function on the component correlation coefficient. Taking the component differential reflectance as an example, its membership function is denoted as... For rain and ice crystals with a relatively large positive preference value, a trapezoid with an open lower bound is adopted: For values ​​less than 0.5 dB, a value of 0 is used; between 0.5 and 1.5 dB, the value linearly increases from 0 to 1; and for values ​​greater than 1.5 dB, a value of 1 is used. For wet snow, dry snow, sleet, and hail, where the preference is close to 0, a central trapezoidal approach is adopted: a value of 1 is used between -0.3 and +0.3 dB, and the value linearly decreases to 0 by 1.0 dB on each side. The membership function of the component correlation coefficient is denoted as... For rain, dry snow, ice crystals, and graupel with simple phases and high component correlation coefficients, a value above 0.98 is assigned as 1, a linear increase is taken between 0.95 and 0.98, and a value below 0.95 is assigned as 0. For wet snow and hail with relatively low component correlation coefficients, a triangle with peak values ​​around 0.90 is selected, with a bulge between 0.85 and 0.95, a peak value of 1, and a linear decrease if deviating from this range. The peak values ​​for wet snow and hail are set around 0.90 because their coherence to polarized echoes decreases due to water inclusion or irregular shapes, but remains significantly higher than non-meteorological clutter. A value around 0.90 distinguishes them from pure phases without mistakenly incorporating them into clutter. The relative order term along the falling velocity does not give the membership degree as an absolute value, but rather assigns a value based on the spectral component's ranking among spectral components in the same distance library, and its membership function is denoted as... For fast-falling rain, hail, and sleet, the closer to the definite direction, the closer the membership degree is to 1; the opposite is true for slow-falling dry snow, wet snow, and ice crystals. When a distance library has only one spectral component and the relative order cannot be compared, If each candidate phase state is set to 1, this term will not participate in the distinction, and the discrimination will degenerate into being determined solely by the component differential reflectivity and the component correlation coefficient.

[0043] The inclusion of the relative order along the falling velocity is to resolve some phase pairs that are difficult to distinguish based on polarization alone. For example... Figure 7 As shown, both rain and ice crystals can exhibit large component differential reflectance and high component correlation coefficients, and their polarization values ​​are similar when viewed individually. However, rain falls faster than ice crystals. Dry snow and graupel both exhibit component differential reflectance close to 0 and high component correlation coefficients, but graupel is denser and falls faster than loose dry snow. Wet snow and hail both exhibit component differential reflectance close to 0 and relatively low component correlation coefficients, but hail falls significantly faster than wet snow. For these three pairs of phase states, the relative order of falling velocity provides a decisive distinguishing dimension beyond polarization. Therefore, the relative speed between spectral components, a byproduct of spectral separation, becomes the basis for discrimination.

[0044] For each spectral component, multiply the membership degrees obtained from its three features for a particular candidate phase to obtain the membership degree: In the formula For the first Each spectral component belongs to the first The degree of belonging to a candidate phase, indicated by the superscript. Indicate the membership function of each candidate phase. Choose one of the following phase types: rain, wet snow, dry snow, ice crystals, sleet, or hail. Use multiplication instead of addition: multiplication requires all three features to be satisfied simultaneously to obtain a high membership degree; if any feature strongly disagrees (membership degree approaches 0), the candidate phase is rejected. Addition can cause a high score for one feature to mask a disagreement for another, leading to misclassification. The phase with the highest membership degree among the six candidate phase types is taken as the phase of this spectral component. In the formula For the first The phase state of each spectral component is determined. This indicates the selection of the candidate phase with the highest membership degree. Continuing with the previous numerical example, the slower-falling spectral component has a component differential reflectance of 0.3 dB and a component correlation coefficient of 0.975. Among candidate phases with intermediate component differential reflectance, it has a high membership degree and is located at the slower-falling end, so the highest membership degree is for dry snow. The faster-falling spectral component has a component differential reflectance of 2.0 dB and a component correlation coefficient of 0.98. Among candidate phases with relatively large component differential reflectance and a high component correlation coefficient, it has a high membership degree and is located at the faster-falling end, so the highest membership degree is for rain. Therefore, this distance library is judged to be a coexistence of dry snow above and rain below, which is consistent with the physical picture of the beam crossing the melting layer. It should also be added that for large particles falling faster (such as large raindrops or melting hail) that enter the resonance region in the X-band and change their polarization characteristics, their abnormal component differential reflectance is confined to an independent spectral component near its own definite direction end, without contaminating the discrimination of other spectral components. This is another advantage of component-by-component discrimination.

[0045] The phase states determined by each spectral component within the same range library are summarized to obtain the coexisting phase state combinations for that range library; when there is only one spectral component, the coexisting phase state combinations contain only one phase state. Finally, the discrimination is extended to the entire detection space: the radar scans angle by angle according to azimuth at a set elevation angle, and advances along each radial direction according to range library. Steps 1 to 3 are performed on each range library to obtain its coexisting phase state combinations. Then, the coexisting phase state combinations of all range libraries are spliced ​​together according to their spatial positions determined by their azimuth, range, and set elevation angle to obtain the spatial distribution of precipitation particle phase states within the radar's detection range. Since the set elevation angle is always between horizontal and vertical, each range library simultaneously satisfies the two conditions required in step 1: the fall velocity can be projected and the spectral differential reflectivity can be retained. The discrimination conditions are consistent throughout space, and the resulting phase state distribution is physically self-consistent.

[0046] As an alternative implementation, the inflection point values ​​of the trapezoidal and triangular membership functions can be shifted as a whole based on local raindrop spectrum observations or scattering calculations to adapt to different climate zones and radar calibration states. When the radar is equipped with reliable temperature profiles or zero-degree layer height information, it can be added as an additional membership function to the product of membership degrees, but this is not necessary for implementing this method. The six candidate phase states can be added or removed as needed. In addition to outputting the category label, the phase state obtained can also be output along with the component reflectance, component differential reflectance, component correlation coefficient, and the velocity range it occupies for subsequent quantitative precipitation estimation. All of this does not exceed the processing procedure of extracting the component differential reflectance and component correlation coefficient for each spectral component, combining the relative order of each spectral component along the falling velocity, calculating the membership degree from the membership function and taking the largest one as the phase state, and then summarizing them into a combination of coexisting phase states and splicing them into a spatial distribution.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. 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. Such 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 method for identifying the phase state of precipitation particles in an X-band dual-polarization micro-weather radar, characterized in that, This method is based on an X-band dual-polarization micro-weather radar that detects objects by simultaneously transmitting and receiving horizontal and vertical polarization signals, and includes the following steps: Step 1, Spectral polarization demixing: Scan the detection range at a set pitch angle between the horizontal and vertical directions. For each range library within the detection range, obtain the spectral reflectivity, spectral differential reflectivity, and spectral correlation coefficient distributed along the Doppler velocity channel from the echoes of that range library in the horizontal and vertical polarization channels. Step 2, Separation of mixed phases in the spectral domain: Based on the valley value of the spectral correlation coefficient formed by the first decrease and then increase of the spectral correlation coefficient along the direction of Doppler velocity, the separation boundary is determined. The distance library is divided into multiple spectral components corresponding to a single phase along the direction of Doppler velocity using the separation boundary as the boundary. Step 3, fuzzy discrimination of unmixed components: For each spectral component, obtain the component differential reflectance and component correlation coefficient, and obtain the relative order of each spectral component along the falling velocity. Find the membership degree of each spectral component to each candidate phase state on the membership function and obtain the membership degree. Take the candidate phase state with the largest membership degree as the phase state of the corresponding spectral component. Summarize the phase states of all spectral components in the same distance library into the coexisting phase state combination of the distance library. Piece together the coexisting phase state combinations of all distance libraries according to azimuth, distance and set elevation angle to obtain the spatial distribution of precipitation particle phase states.

2. The method according to claim 1, characterized in that, The beamwidth corresponding to the antenna aperture of the X-band dual-polarization micro-weather radar used allows a single resolution element to accommodate precipitation particles of various phases. The transmit power of the X-band dual-polarization micro-weather radar used keeps the signal-to-noise ratio of the single-channel echo from the long-range reservoir at a level that causes fluctuations in the estimation of the spectral correlation coefficient.

3. The method according to claim 1, characterized in that, Step 1, which involves obtaining spectral reflectance, spectral differential reflectance, and spectral correlation coefficient, includes: continuously acquiring multiple sets of horizontally polarized time series and multiple sets of vertically polarized time series from the same distance library within a dwell time; applying window functions to each set of horizontally polarized time series and each set of vertically polarized time series to suppress spectral leakage; and then performing Doppler spectral transformation on each set; averaging the Doppler power spectrum obtained from each set of horizontally polarized time series along the Doppler velocity channel to obtain the average power spectrum of the horizontal polarization channel; and averaging the Doppler power spectrum obtained from each set of vertically polarized time series along the Doppler velocity channel. The average power spectrum of the vertical polarization channel is obtained by averaging the power spectrum of the horizontal polarization time series and the echo complex signal of the same vertical polarization time series. The cross spectrum of each group of cross spectrums is then averaged along the Doppler velocity channel to obtain the average cross spectrum. The spectral reflectivity is obtained from the average power spectrum of the horizontal polarization channel, and the spectral differential reflectivity is obtained from the average power spectra of the horizontal and vertical polarization channels. The spectral correlation coefficient is obtained from the average cross spectrum together with the average power spectra of the horizontal and vertical polarization channels. Finally, the noise level of the same distance library is estimated from the Doppler velocity channel containing only background noise.

4. The method according to claim 3, characterized in that, The Doppler power spectrum is plotted with the Doppler velocity channel on the x-axis and the echo power on the y-axis. Each Doppler velocity channel corresponds to a Doppler velocity value determined by the radar wavelength and pulse repetition mode. When the Doppler velocity of precipitation particles exceeds the measurable velocity range of the radar, the Doppler power spectrum folds at the velocity boundary. A folded spectrum segment, which is attached to one velocity boundary and separated from the main spectrum segment near the other velocity boundary by a region containing only noise level, is shifted along the Doppler velocity axis by the span of a measurable velocity range, so that the folded spectrum segment is continuously connected to the main spectrum segment.

5. The method according to claim 3, characterized in that, Step 2 includes: within each range cell, connecting Doppler velocity channels with spectral reflectance higher than the noise level into a continuous region, which is defined as the Doppler velocity range occupied by the precipitation particle swarm within that range cell; within the Doppler velocity range, identifying a local minimum formed by the spectral correlation coefficient first decreasing and then increasing along the Doppler velocity direction; when the decrease in this local minimum relative to the adjacent local maximum exceeds the estimated fluctuation of the spectral correlation coefficient itself, and the spectral differential reflectance takes different values ​​on both sides of this local minimum, or the spectral correlation coefficient takes different values ​​on both sides of this local minimum, then this local minimum is identified as the spectral... The valley value of the correlation coefficient is used to determine the Doppler velocity channel corresponding to the valley value of the spectral correlation coefficient as the separation boundary. The valley value of the spectral correlation coefficient corresponds to the boundary between two types of precipitation particles with similar falling velocities but different phases. When the values ​​of the spectral differential reflectance on both sides of the local minimum are similar, and the values ​​of the spectral correlation coefficient on both sides of the local minimum are similar, the local minimum is retained within the same spectral component. The endpoint of the Doppler velocity range and all the obtained separation boundaries are used as the boundary. The distance library is divided into multiple adjacent velocity intervals along the Doppler velocity, and each velocity interval is a spectral component.

6. The method according to claim 5, characterized in that, Under the condition that the horizontal wind is approximately uniform within the resolution corresponding to the distance library, the horizontal wind applies the same Doppler velocity shift to all precipitation particles within the resolution, causing the spectrum of Doppler velocity variation within the distance library to be translated as a whole along the Doppler velocity axis, while the relative order and relative spacing of all separation boundaries and multiple spectral components along the Doppler velocity axis remain unchanged.

7. The method according to claim 5, characterized in that, Step 3 includes: for each spectral component, summing the echo power of the average power spectrum of the horizontal polarization channel in each Doppler velocity channel of the velocity interval occupied by the spectral component along the Doppler velocity to obtain the component reflectivity of the spectral component. The component reflectivity characterizes the echo strength of the precipitation particle swarm corresponding to the spectral component; summing the echo power of the average power spectrum of the horizontal polarization channel and the vertical polarization channel in each Doppler velocity channel of the velocity interval along the Doppler velocity respectively, and obtaining the component differential reflectivity of the spectral component from the summation result; and combining the values ​​of the average cross spectrum in each Doppler velocity channel of the velocity interval along the Doppler velocity, and obtaining the component correlation coefficient of the spectral component from the combination result together with the summation result of the two.

8. The method according to claim 7, characterized in that, At a set pitch angle, the falling precipitation particles generate Doppler velocity components along the Doppler velocity axis. Precipitation particles falling faster tend to be biased towards the same end of the Doppler velocity axis, and the direction of that biased end is taken as the definite direction. Within the same distance range, the spectral component corresponding to a faster falling precipitation particle is closer to the end of the definite direction, while the spectral component corresponding to a slower falling precipitation particle is closer to the end opposite to the definite direction. Under the condition that the horizontal wind within the resolution corresponding to the distance range is approximately uniform, the horizontal wind applies the same Doppler velocity offset to each spectral component within the same distance range, and the relative order of each spectral component along the definite direction remains unchanged. Thus, the relative order of each spectral component along the falling velocity within the same distance range is obtained.

9. The method according to claim 8, characterized in that, Membership functions are established for six candidate phases: rain, wet snow, dry snow, ice crystals, graupel, and hail. They are based on the relative order of component differential reflectivity, component correlation coefficient, and falling velocity, respectively, to adapt to the influence of X-band two-way propagation through rain zones on the reflectivity order. In terms of component differential reflectance, the membership degree of rain and ice crystals increases as the component differential reflectance is greater than 0, while the membership degree of wet snow, dry snow, graupel and hail increases as the component differential reflectance approaches 0. In terms of component correlation coefficient, the membership degree of rain, dry snow, ice crystals and graupel increases as the component correlation coefficient approaches 1, while the membership degree of wet snow and hail increases as the component correlation coefficient decreases relative to spectral components close to 1 within the same distance range. Along the relative order of falling velocity, the membership degree of rain, hail and graupel increases as the corresponding spectral component approaches the end of the defined direction, while the membership degree of dry snow, wet snow and ice crystals increases as the corresponding spectral component approaches the end opposite to the defined direction. The candidate phases of rain and ice crystals, dry snow and graupel, and wet snow and hail are similar to each other in terms of component differential reflectance and component correlation coefficient. The two candidate phases in each pair are distinguished from each other by means of the relative order of falling velocity.

10. The method according to claim 9, characterized in that, For each spectral component, the membership degree of the spectral component to each candidate phase is obtained from the membership function. For the same candidate phase, the membership degrees obtained from the component differential reflectivity, component correlation coefficient and relative order along the falling velocity are multiplied to obtain the degree of belonging. The candidate phase with the highest degree of belonging is taken as the phase of the precipitation particle swarm corresponding to the spectral component, and the velocity range occupied by the spectral component is taken as the falling velocity range of the precipitation particles corresponding to the spectral component. Large particles that fall faster and whose polarization characteristics change when entering the resonance region in the X-band are separated into independent spectral components near one end of a certain direction, so that the polarization characteristics of the large particles can be distinguished within the velocity range occupied by the independent spectral components. When one spectral component is obtained in the range library, the coexisting phase combination of the range library contains one type of phase.

Images