Method and system for joint inversion of precipitation micro-physical parameter profile based on active radar and passive microwave

Abstract

The invention provides a rainfall micro-physical parameter profile inversion method and system based on active radar and passive microwave joint, and the method comprises the steps: carrying out the constraint of particle micro-physical parameters through employing the active radar and passive microwave detection multi-source observation data, and building a dual optimization inversion model; an observation operator of the inversion model comprises micro-physical characteristic changes of different types of particles and an electromagnetic wave attenuation effect, and inversion uncertainty caused by particle micro-physical assumption and the attenuation effect can be reduced, so that optimal state variables of different types of precipitation particles on a profile can be obtained through inversion. And the inversion precision of the rainfall micro-physical parameter profile is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of precipitation radar data processing and application, and particularly relates to a method and system for jointly inverting precipitation microphysical parameter profiles based on active radar and passive microwave. Background Art

[0002] High-precision precipitation parameter profiles are key to studying cloud and precipitation processes and, consequently, improving weather and climate models. Satellite-based remote sensing and inversion can provide regional and even global estimates of precipitation's three-dimensional structure and microphysical properties, providing crucial data for weather, climate, and extreme event forecasting. Optimizing the utilization of precipitation information obtained from active and passive microwave satellite-based precipitation detection and improving precipitation parameter profile inversion algorithms are crucial for obtaining high-precision precipitation profile products.

[0003] Currently, spaceborne sensors used for cloud and precipitation profile detection and inversion primarily include active and passive microwave remote sensing. Spaceborne precipitation radars are representative active microwave payloads capable of acquiring the vertical structure of precipitation parameters. my country's FY-3G satellite, launched in April 2023, carries a dual-frequency precipitation radar operating in the Ku and Ka bands. This dual-frequency precipitation radar can jointly invert precipitation droplet profile information using two bands, thereby improving precipitation estimation accuracy. Spaceborne microwave imagers are representative passive microwave payloads. They offer a wide range of wavelengths and high sensitivity. Using multi-band radiometric measurements, they enable three-dimensional detection of layered cloud precipitation at multiple altitudes. For example, the microwave imagers (MWRIs) carried by the Fengyun-3 satellite and the GPM satellite microwave imager (GMI) utilize multiple microwave channels located in atmospheric windows and near water vapor absorption lines to capture information such as the water content of hydrometeors. Spaceborne radars and passive microwaves play an important role in detecting the three-dimensional structure of global clouds and precipitation.

[0004] Current spaceborne precipitation radars use dual-frequency inversion methods to obtain precipitation parameter profiles such as raindrop size and precipitation intensity. However, these profiles are based on assumptions about the microphysical properties of rainfall particles. The particle microphysical model and scattering characteristics are overly simplistic, failing to consider the complex variations in the microphysical model of non-liquid particles. This results in limited inversion accuracy for non-liquid particles. Furthermore, radar electromagnetic wave attenuation significantly affects the reflectivity factor. Spaceborne radars typically use a ground-based reference method to correct for attenuation of the radar reflectivity factor. This attenuation correction error can propagate further into the microphysical parameter inversion.

[0005] Passive microwave inversion primarily relies on statistical empirical methods and physical modeling to retrieve information such as precipitation parameters. The accuracy of the inversion depends on the construction of a priori databases and the calculation of radiative transfer processes. The official GPM satellite product (2021, Global Precipitation Measurement (GPM) mission algorithm theoretical basis document GPROF2021) combines DPR and GMI data based on Bayesian theory. It further adjusts the DPR attenuation using GMI-derived total water content and cloud liquid water, resulting in more accurate precipitation inversions. Grecu et al. (2016, The GPMcombined algorithm) combine DPR and GMI data, using a statistical optimization process to minimize a cost function to obtain high-precision rainfall intensity profiles and raindrop spectrum profiles. The primary objective of the GPM satellite's active and passive microwave joint inversion is precipitation. Key physical assumptions and output parameters focus on rainfall particles. Failure to consider precipitation microphysical parameter profiles can lead to significant errors due to radar electromagnetic wave attenuation and the assumptions underlying the hydrometeor particle microphysical model used in the inversion. Summary of the Invention

[0006] In order to solve the problems existing in the prior art, the present invention provides a method and system for jointly inverting precipitation microphysical parameter profiles using active and passive data.

[0007] The specific technical solutions of the present invention are as follows:

[0008] A method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave joint inversion includes the following steps:

[0009] 1) Based on active radar observations, a discriminant method is used to classify hydrometeors at different heights along the precipitation profile to obtain different types of particles.

[0010] 2) Based on the microphysical property modeling of the different types of particles, the state variables of the different types of particles are converted into simulated observation data through forward simulation, and the simulated observation data are used to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, where x represents the state variable of the particle;

[0011] 3) establishing a regularized inversion model for precipitation microphysical parameter profiles, iteratively solving the regularized inversion model using a gradient descent method based on the observation operator H(x) and active observation data obtained through observations from a spaceborne dual-frequency radar, to obtain a set of state variables for different types of particles;

[0012] 4) Input the state variable sets of different types of particles obtained by regularized inversion into the atmospheric radiation transfer equation, and obtain the simulated microwave brightness temperature of different frequency channels through simulation calculation. At the same time, obtain the observed microwave detection data of different frequency channels with temporal and spatial matching to obtain the corresponding observed microwave brightness temperature of different frequency channels;

[0013] 5) Establishing a least squares estimation model, inputting the simulated microwave brightness temperature and the corresponding observed microwave brightness temperature for inversion, calculating the least squares solution of the state variables of different types of particles, that is, obtaining the optimal state variables of each type of particle, and combining the optimal states of each type of particle to obtain the optimal precipitation microphysical parameter profiles at different altitudes.

[0014] The present invention is further designed to:

[0015] The precipitation product based on active radar observations in step 1) uses a discriminant method to classify hydrometeors at different heights on the precipitation profile to obtain different types of particles, specifically:

[0016] 1.1) Obtain observational data of the dual-frequency radar precipitation profile and background atmospheric temperature information for the precipitation area to be inverted. The background atmospheric temperature information is matched with the observational data of the dual-frequency radar precipitation profile through interpolation calculation. The matched observational data includes a ground snowfall identification product, a melt layer top and melt layer bottom detection product, a precipitation type classification product, and temperature information. Select data from the observational data where the precipitation detection indicator is greater than 0;

[0017] 1.2) Check whether the ground snowfall indicator product in the selected data is greater than 0; if it is greater than 0, proceed to step 1.3; if not, proceed to step 1.4;

[0018] 1.3) Determine whether the height corresponding to the pixel in the selected data is less than the height corresponding to a temperature of 0°; if so, the pixel corresponds to wet snow; if not, the pixel corresponds to dry snow / ice crystals;

[0019] 1.4) Based on the melt layer top and melt layer bottom detection products, determine whether the melt layer top MLT and melt layer bottom MLB are detected for each profile line; if so, proceed to step 1.5; if not, proceed to step 1.6;

[0020] 1.5) Further determine the precipitation type corresponding to this profile;

[0021] If it is stratiform precipitation type S, pixels below the bottom of the melt layer are rain, pixels above the bottom of the melt layer but below the top of the melt layer are wet snow, and pixels above the top of the melt layer are dry snow / ice crystals.

[0022] If the precipitation type of this profile is convective precipitation C, the precipitation below the bottom of the melt layer is rain, and the precipitation above the bottom of the melt layer is graupel.

[0023] 1.6) Further determine the precipitation type corresponding to this profile. If it is stratiform precipitation type S, the hydrometeors on the profile are primarily dry snow / ice crystals or wet snow. Furthermore, based on the temperature information corresponding to each pixel on the profile, determine the specific hydrometeors corresponding to each pixel. A temperature greater than 0 indicates wet snow, and a temperature less than 0 indicates dry snow / ice crystals. If it is convective precipitation type C, the hydrometeors on the profile are primarily graupel.

[0024] The present invention is further designed to:

[0025] In step 2), the microphysical properties of the different types of particles are modeled, and the state variables of the different types of particles are converted into simulated observation data through forward simulation. The simulated observation data is used to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, specifically:

[0026] 2.1) Using ground-based remote sensing observations and in situ microphysical observations, calculate the microphysical properties of different types of hydrometeors, including dry snow / ice crystals, wet snow, graupel, and rain.

[0027] 2.2) Based on the microphysical properties of different types of particles counted in 2.1), microphysical property models of different types of particles are established, wherein the microphysical property models include a shape model, a density model, and a particle spectrum model; the particle spectra of different types of particles are all modeled as normalized gamma spectra;

[0028] 2.3) Calculate the dielectric constant ε of different types of particles. For hydrometeors composed of ice crystals and rain, use the Debye equation; for hydrometeors composed of dry snow, wet snow, and graupel, use the Maxwell-Garnet equation.

[0029] 2.4) Based on the shape model and density model in the microphysical property model established in 2.2) and the particle dielectric parameters calculated in 2.3), the scattering characteristics of different types of particles in the Ku band and Ka band are calculated using the T-matrix and DDA tools. The scattering characteristics include the backscattering cross section σ b and extinction cross section Q ext ;

[0030] 2.5) Based on the particle spectrum model in the microphysical characteristics model established in 2.2) and the particle scattering characteristics calculated in 2.4), simulated observation data including radar reflectivity factors and integrated attenuation paths in the Ku and Ka bands are calculated:

[0031]

[0032] Where, K = ε-1 / ε+2, λ represents the wavelength of Ku band or Ka band, Zλ Indicates the reflectivity factor corresponding to wavelength λ, K ext,λ represents the attenuation coefficient corresponding to wavelength λ, N(D) represents the particle spectrum, σ b,λ (D) represents the backscattering cross section of a particle with a diameter of D corresponding to wavelength λ, Q ext,λ (D) represents the attenuation cross section of a particle with a diameter of D corresponding to a wavelength of λ, PIA λ represents the integrated attenuation path corresponding to wavelength λ, D max 、D min It represents the maximum and minimum values of the particle diameter, r represents the distance from the detection target to the radar, and R represents the total distance from the radar to the ground;

[0033] 2.6) Statistical analysis of radar observation parameters and particle state variables W and D calculated in 2.5) m Based on the relationship between the state variation of particles and the simulated observation data established by the fitting method, the observation operator H(x) is simulated to take into account the changes in the microphysical properties of different types of particles. The observation operator is modeled in the form of a polynomial:

[0034]

[0035] Where n represents the number of polynomial coefficients to be fitted; c i , e i Respectively represent the coefficients of the i-th term in each formula; in the observation operator H(x), x represents the state variable of the particle, x=[D m ,W], W represents the water content of the particles, D m represents the mass-weighted diameter.

[0036] The present invention is further designed to:

[0037] In step 3), the regularized inversion model of precipitation microphysical parameter profile is established as follows:

[0038]

[0039] Where argmin(f(x)) is the state variable x of the particle that makes the function f(x) reach its minimum value, σ R The covariance matrix of the observation error of δy=yH(x) is represented, β is the regularization constant (β takes an empirical value, which can be 0.01, 0.1, or 0.5), x b represents the prior background estimate of the particle's state variables, σ B represents the prior error covariance matrix.

[0040] The present invention is further designed to:

[0041] In step 3), a regularized inversion model of precipitation microphysical parameter profile is established based on the observation operator H(x), active observation data and prior information x b , the regularized inversion model is iteratively solved using the gradient descent method to obtain the state variable sets of different types of particles, specifically:

[0042] 3.1) Setting the prior background microphysical parameters x in the regularized inversion model of precipitation microphysical parameter profiles b , and the background error covariance matrix σ B and the observation error covariance matrix σ R , where the prior background microphysical parameters x b Based on the statistical average of microphysical parameters observed by ground drop spectrum, σ B and σ R Obtained by statistically analyzing the mean square error of state variables and radar observations respectively;

[0043] 3.2) Initialize the state variable x0, regularization constant β, convergence condition error e (set to be less than 0.1), and number of iterations k in the regularized inversion model;

[0044] 3.3) Substitute the state variables into the observation operator to obtain the observed value, and calculate the difference between the active observation data in step k and the observed value obtained by the observation operator

[0045] 3.4) Take the partial differential of the cost function and determine the iteration step size:

[0046]

[0047] 3.5) Update based on iterative step size

[0048] 3.6) Determine whether the iteration has converged. If and If the difference between y is less than e, or the number of iterations is reached, the iteration stops, otherwise, return to step 3.3); k represents the actual radar observation at step k, represents the estimated value of the state variable at time k; represents the estimated value of the state variable at time k+1;

[0049] 3.7) When the iteration converges or reaches the number of iterations, the calculation results are output

[0050] The present invention is further designed to:

[0051] In step 3), the active observation data N represents the number of pixels on the profile, Z Kurepresents the reflectivity factor detected by Ku-band radar, Z Ka Represents the reflectivity factor detected by Ka-band radar, PIA Ka represents the attenuation integral path of Ka-band radar detection. DFR represents the difference between the radar echoes observed in Ku-band and Ka-band (dual frequency ratio). DFR = 10log 10 Z Ku -10log 10 Z Ka .

[0052] The present invention is further designed to:

[0053] In step 5), a least squares estimation model is established, specifically as follows:

[0054]

[0055] Where x * represents the optimal state variable of the particle, represents the state variables of the particle obtained based on regularized inversion.

[0056] The system for inverting precipitation microphysical parameter profiles based on active radar and passive microwave joint inversion includes the following modules:

[0057] The particle classification module uses a discriminant method to classify hydrometeors at different heights on the precipitation profile based on precipitation products observed by active radar to obtain different types of particles.

[0058] An observation operator construction module is configured to model the microphysical properties of the different types of particles, convert the state variables of the different types of particles into simulated observation data through forward simulation, and use the simulated observation data to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, where x represents the state variable of the particle;

[0059] A state variable solving module establishes a regularized inversion model for the precipitation microphysical parameter profile. Based on the observation operator H(x) and active observation data, the regularized inversion model is iteratively solved using a gradient descent method to obtain a set of state variables for different types of particles. The active observation data is obtained through observation data from a spaceborne dual-frequency radar.

[0060] The brightness temperature acquisition module inputs the state variable set of different types of particles obtained by regularization inversion into the atmospheric radiation transfer equation, and obtains the simulated microwave brightness temperature Tb of different frequency channels through simulation calculation. sim At the same time, the observation microwave detection data of different frequency channels matching the radar detection time and space are obtained to obtain the corresponding observation microwave brightness temperature Tb of different frequency channels. obs ;

[0061] Profile inversion module, establishes the least squares estimation model, inputs the simulated microwave brightness temperature Tb sim and observed microwave brightness temperature Tb obs The inversion is performed to calculate the least squares solution of the state variables of different types of particles, that is, the optimal state variables of each type of particle are obtained. The optimal state variables of each type of particle are combined to obtain the optimal precipitation microphysical parameter profiles at different heights.

[0062] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the above method.

[0063] A computer-readable storage medium stores a computer program, which implements the steps of the above method when executed by a processor.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] The present invention provides a method for jointly inverting precipitation microphysical parameter profiles based on dual-frequency precipitation radar and multi-channel microwave imager observation data. The method utilizes multi-source observation data from active radar and passive microwave detection to constrain particle microphysical parameters and establish a dual optimization inversion model. The observation operator of the inversion model incorporates the changes in microphysical properties of different types of particles and the electromagnetic wave attenuation effect, which can reduce the inversion uncertainty caused by particle microphysical assumptions and attenuation effects. The optimal state variables of different types of precipitation particles on the profile can be inverted, thereby improving the inversion accuracy of the precipitation microphysical parameter profile.

[0066] Because the profile contains different types of hydrometeor particles, the method of this invention classifies these particles into five types: dry snow / ice crystals, wet snow, graupel, and rain. The observation operator calculations consider variations in each particle's microphysical properties, such as shape, density, and spectral distribution. A set of state variables is inverted for each particle, and the optimal state variables for each particle are selected using the least-squares method. This overcomes the problem of reduced microphysical parameter inversion accuracy in existing methods, which rely on the use of a single particle type and simple particle microphysics assumptions to invert the entire profile.

[0067] Radar electromagnetic waves significantly attenuate when passing through precipitation, and this attenuation significantly affects the reflectivity factor. Existing methods typically use a ground-based reference method to correct for radar reflectivity attenuation. This attenuation correction error propagates further into the microphysical parameter inversion, significantly impacting the inversion results. The observation operator of the present invention is derived from simulation calculations, which already account for the attenuation effect. The observation data itself includes attenuation, and since both simulation and observation include attenuation, no additional correction is required during the regularized inversion, improving inversion accuracy.

[0068] The observation operator includes a microphysical model of attenuation and particle type changes. A dual-optimization inversion model of dual-frequency radar regularization and microwave imager least squares estimation is established to reduce inversion uncertainty, obtain the optimal solution of the microphysical parameters of different types of particles, and then obtain the optimal precipitation microphysical parameter profile. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 The process of jointly retrieving precipitation microphysical parameter profiles based on active radar and passive microwave;

[0070] Figure 2 Flowchart for classification of hydrometeor particles based on active radar;

[0071] Figure 3 Flowchart for solving the regularized inversion model using gradient descent;

[0072] Figure 4 The vertical profile of the reflectivity factor of the GPM DPR observation case on June 10, 2019, and the matching GMI observation brightness temperature; (a) is the radar reflectivity factor of the Ku-band radar, (b) is the reflectivity factor of the Ka-band radar, and (c) is the brightness temperature observed by different GMI channels;

[0073] Figure 5 Comparison between the inversion algorithm results of the present invention and the official GPM inversion algorithm results;

[0074] Figure 5 In the figure, (a) is the water content W obtained by inversion based on the inversion algorithm proposed in the present invention, and (b) is the mass-weighted diameter D obtained by inversion based on the inversion algorithm proposed in the present invention. m , (c) is the Ku-band radar reflectivity factor calculated based on the microphysical parameters inverted by the inversion algorithm proposed in this invention, (d) is the water content W inverted based on the GPM official inversion algorithm, and (e) is the mass-weighted diameter D inverted based on the GPM official inversion algorithm. m , (f) is the Ku-band radar reflectivity factor calculated based on the microphysical parameters inverted by the GPM official inversion algorithm. DETAILED DESCRIPTION

[0075] In order to make the purpose and technical solutions of the present invention clearer and more complete, the present invention is further described in detail below in combination with specific implementation cases and with reference to the accompanying drawings.

[0076] To address the current problems in inverting spaceborne precipitation microphysical parameter profiles, this paper proposes a dual-optimization estimation method based on the combined detection of a dual-frequency precipitation radar and a multi-channel microwave imager. The method takes into account the differences in the scattering characteristics and sensitivity of particles to microwave detection of different frequencies, as well as the complementarity of active and passive microwave detection. Based on the physical nature of precipitation parameter profile inversion, the inversion algorithm establishes a regularized inversion model based on the dual-frequency radar and solves the inverse problem of physics and mathematics to obtain the set of microphysical state variables of different types of hydrometeor particles. Furthermore, combined with multi-channel passive microwave detection, the least squares optimization estimation is used to further constrain the microphysical properties of different types of particles, thereby obtaining the optimal precipitation microphysical parameter profile.

[0077] The regularized inversion model directly solves the inverse problem, reducing the various assumptions of previous algorithms and taking into account the prior microphysical properties and observation operators of different particle types. By iteratively solving the regularized equation, it can obtain optimized inversion results for the microphysical parameters of different particle types along the profile. Compared with previous algorithms, the present invention improves inversion accuracy.

[0078] The present invention first uses active radar for optimal inversion to obtain sets of state variables for different particle types (particles of different shapes and densities can each be inverted to obtain a set of W and Dm. For example, different snow densities correspond to different observation operators, each yielding a set of W and Dm). Then, using passive microwave detection constraints, the optimal state variables corresponding to snow with a density of 0.1 (for example) are determined. Through these two optimization inversions, the optimal W and Dm for this particle are obtained.

[0079] The existing inversion methods do not consider the changes in the microphysical properties of so many types of particles. They all assume that the particles are spherical and have the same density, and the corresponding W and Dm can be obtained using the Bayesian method.

[0080] Example 1:

[0081] The method of the present invention for inverting precipitation microphysical parameter profiles based on active radar and passive microwave is as follows: Figure 1 As shown, the following steps are included:

[0082] 1) Precipitation products based on active radar observations. These precipitation products can be obtained from official data products of spaceborne dual-frequency radars. A discriminant method is used to classify hydrometeors at different heights along the precipitation profile to obtain different types of particles.

[0083] 2) Based on the microphysical properties of different types of particles, the microphysical state variables of different types of particles are converted into simulated observation data through forward simulation. The simulated observation data are used to construct an observation operator H(x) that takes into account the changes in the microphysical properties of different types of particles, where x represents the state variable of the particle;

[0084] 3) establishing a regularized inversion model for precipitation microphysical parameter profiles, iteratively solving the regularized inversion model using a gradient descent method based on the observation operator and active observation data obtained through observation data from a spaceborne dual-frequency radar, to obtain a set of state variables for different types of particles;

[0085] 4) Input the state variable set of different types of particles obtained by regularization inversion into the atmospheric radiation transfer equation, and obtain the simulated microwave brightness temperature Tb of different frequency channels through simulation calculation sim At the same time, the observation microwave detection data of different frequency channels that match the radar detection (active observation data) in time and space are obtained. The observation microwave detection data is obtained based on the passive microwave detection brightness temperature of the satellite-borne microwave imager, and the corresponding observation microwave brightness temperature Tb of different frequency channels is obtained. obs ;

[0086] 5) Establish a least squares estimation model and input the simulated microwave brightness temperature Tb sim and observed microwave brightness temperature Tb obs The inversion is performed to calculate the least squares solution of the state variables of different types of particles, that is, the optimal state variables of each type of particle are obtained. The optimal state variables of each type of particle are combined to obtain the optimal precipitation microphysical parameter profiles at different heights.

[0087] Example 2:

[0088] This embodiment is further designed based on the first embodiment in that:

[0089] The meteorological parameter identification of the precipitation product based on active radar observation in step 1) is combined with the background temperature information, and the hydrometeor particle type classification is performed on the particles at different heights of the precipitation parameter profile of the precipitation scene observed by the GPM satellite using the discriminant method. Figure 2 As shown, the meteorological parameter identifiers such as the precipitation detection identifier flagPrecip, the ground snowfall identifier flagSurfaceSnow, the melt layer top MLT and the melt layer bottom MLB are specifically:

[0090] 1.1) Obtain observational data of the dual-frequency radar precipitation profile and background atmospheric temperature information for the precipitation area to be inverted. The background atmospheric temperature information is matched with the observational data of the dual-frequency radar precipitation profile through interpolation calculation. The matched observational data includes a ground snowfall identification product, a melt layer top and melt layer bottom detection product, a precipitation type classification product, and temperature information. Select data from the observational data where the precipitation detection flag (flagPrecip) is greater than 0;

[0091] 1.2) Check whether the surface snowfall flag product (flagSurfaceSnow) in the selected data is greater than 0; if it is greater than 0, go to step 1.3; if not, go to step 1.4;

[0092] 1.3) Determine whether the height corresponding to the pixel in the selected data is less than the height corresponding to a temperature of 0°; if so, the pixel corresponds to wet snow; if not, the pixel corresponds to dry snow / ice crystals;

[0093] 1.4) Based on the melt layer top and melt layer bottom detection products, determine whether the melt layer top MLT and melt layer bottom MLB are detected for each profile line; if so, proceed to step 1.5; if not, proceed to step 1.6;

[0094] 1.5) Further determine the precipitation type corresponding to this profile;

[0095] If the precipitation type is stratiform, the hydrometeors on the profile are mainly dry snow / ice crystals, wet snow, and rain. The corresponding hydrometeors at different heights are further determined based on the heights of the pixels on the profile. Pixels below the bottom of the melt layer are rain, pixels above the bottom of the melt layer but below the top of the melt layer are wet snow, and pixels above the top of the melt layer are dry snow / ice crystals.

[0096] If the precipitation type of this profile is convective precipitation C, then the hydrometeors on the profile are mainly graupel and rain. Furthermore, based on the heights of different pixels on the profile, the corresponding hydrometeors at different heights are determined. Hydrometeors below the bottom of the melt layer are rain, and hydrometeors above the bottom of the melt layer are graupel.

[0097] 1.6) Further determine the precipitation type corresponding to this profile. If it is stratiform precipitation type S, the hydrometeors on the profile are primarily dry snow / ice crystals or wet snow. Furthermore, based on the temperature information corresponding to different pixels along the profile, determine the specific hydrometeors corresponding to each pixel. A temperature greater than 0 indicates wet snow, and a temperature less than 0 indicates dry snow / ice crystals. If it is convective precipitation type C, the hydrometeors on the profile are primarily graupel.

[0098] Example 3:

[0099] This embodiment is further designed on the basis of the second embodiment in that: in step 2), based on the microphysical properties modeling of the different types of particles, the state variables of the different types of particles are converted into simulated observation data through forward simulation, and the observation operator H(x) considering the changes in the microphysical properties of the different types of particles is constructed with the simulated observation data, specifically:

[0100] 2.1) Using ground-based remote sensing observations and in situ microphysical observations, calculate the microphysical properties of different types of hydrometeors, including dry snow / ice crystals, wet snow, graupel, and rain.

[0101] 2.2) establishing microphysical property models for different types of particles based on the microphysical properties of the different types of particles collected in 2.1), wherein the microphysical property models include a shape model, a density model, and a particle spectrum model;

[0102] In the shape model, the shape model of ice crystals / dry snow is set to disk, column, snowflake, bullet flower, and ellipsoid; wet snow is set to ellipsoid and snowflake; graupel and raindrops are both set to ellipsoid;

[0103] In the density model, the density of dry snow is 0.1g / cm 3 , the ice crystal density is 0.92g / cm 3 , the raindrop density is 1g / cm 3 The density of hail is 0.4g / cm 3 , wet snow density is modeled as ρ = aD b , where D represents the particle diameter, a and b are constants, b is -1, a changes with the degree of melting, a=a u / (1-FR), where a u is a constant of 0.143, FR refers to melting ratio;

[0104] In the particle spectrum model, the particle spectra of different types of particles are modeled as normalized Gamma spectra;

[0105] The normalized Gamma spectrum is expressed as follows:

[0106]

[0107] Where N(D) represents the particle spectrum, that is, the number of particles corresponding to particles of different diameters D in a unit volume, ρ represents the particle density, W represents the water content of the particles, and D m represents the mass-weighted diameter, μ is the particle spectrum shape parameter, and μ is set to a constant. Since both radar and passive microwave observations depend on the particle spectrum, the two unknowns in the particle spectrum: W and D m As the state variable of the particle.

[0108] 2.3) Calculate the dielectric constant ε of different types of particles. For hydrometeors composed of ice crystals and rain, use the Debye equation; for hydrometeors composed of dry snow, wet snow, and graupel, use the Maxwell-Garnet equation.

[0109] 2.4) Based on the shape model and density model in the microphysical property model established in 2.2) and the particle dielectric parameters calculated in 2.3), the scattering characteristics of different types of particles in the Ku band and Ka band are calculated using the T-matrix and DDA tools. The scattering characteristics include the backscattering cross section σ b and extinction cross section Qext ;

[0110] 2.5) Based on the particle spectrum model in the microphysical characteristics model established in 2.2) and the particle scattering characteristics calculated in 2.4), the simulated observation data including the radar reflectivity factor (unit: mm) in the Ku band and Ka band are calculated. 6 .m -3 ), integral attenuation path:

[0111]

[0112] Where, K = ε-1 / ε+2, λ represents the wavelength of Ku band or Ka band, Z λ Indicates the reflectivity factor corresponding to wavelength λ, K ext,λ represents the attenuation coefficient corresponding to wavelength λ, N(D) represents the particle spectrum, σ b,λ (D) represents the backscattering cross section of a particle with a diameter of D corresponding to wavelength λ, Q ext,λ (D) represents the attenuation cross section of a particle with a diameter of D corresponding to a wavelength of λ, PIA λ represents the integrated attenuation path corresponding to wavelength λ, D max 、D min It represents the maximum and minimum values of the particle diameter, r represents the distance from the detection target to the radar, and R represents the total distance from the radar to the ground.

[0113] 2.6) Statistical analysis of radar observation parameters and particle state variables W and D calculated in 2.5) m The relationship between the state of the particles (W, D m ) and simulated observation data (Z ku , Z ka , PIA), the observation operator H(x) is obtained by simulation to consider the changes in the microphysical properties of different types of particles, where the observation operator is modeled in the form of a polynomial:

[0114]

[0115] Where n represents the number of polynomial coefficients to be fitted; c i , e i Respectively represent the coefficients of the i-th term in each formula; in the observation operator H(x), x represents the state variable of the particle, x=[D m ,W], W represents the water content of the particles, D m represents the mass-weighted diameter.

[0116] Example 4:

[0117] This embodiment is further designed on the basis of the third embodiment in that:

[0118] In step 3), a regularized inversion model of precipitation microphysical parameter profile is established based on the observation operator H(x), active observation data, and prior information x b , the regularized inversion model is iteratively solved using the gradient descent method to obtain the state variable sets of different types of particles, specifically:

[0119] The regularized inversion model of precipitation microphysical parameter profiles is established as follows:

[0120]

[0121] Where argmin(f(x)) is the state variable x of the particle that makes the function f(x) reach its minimum value, σ R The covariance matrix of the observation error of δy=yH(x) is represented, β is the regularization constant (β takes an empirical value, which can be 0.01, 0.1, or 0.5), x b represents the prior background estimate of the particle's state variables, σ B represents the prior error covariance matrix.

[0122] The gradient descent method is used to regularize the inversion model and iteratively solve the state variables of different types of hydrometeor particles. Since non-liquid particles such as ice crystals / dry snow, wet snow, and graupel have different shapes and densities, each shape and density corresponds to an observation operator, thereby solving the state variables of each type of particle. Therefore, each shape and density of non-liquid particles will be inverted to obtain a set of state variables (W, D m ), thus forming a set of state variables. The iterative solution process of the gradient descent method is as follows Figure 3 The specific steps are as follows:

[0123] 3.1) Setting the prior background microphysical parameters x in the regularized inversion model of precipitation microphysical parameter profiles b , and the background error covariance matrix σ B and the observation error covariance matrix σ R , where the prior background microphysical parameters x b Based on the statistical average of microphysical parameters observed by ground drop spectrum, σ B and σ R Obtained by statistically analyzing the mean square error of state variables and radar observations respectively;

[0124] 3.2) Initialize the state variable x0, regularization constant β, convergence condition error e (set to be less than 0.1), and number of iterations k in the regularized inversion model;

[0125] 3.3) Substitute the state variables into the observation operator to obtain the observed value, and calculate the difference between the active observation data in step k and the observed value obtained by the observation operator

[0126] 3.4) Take the partial differential of the cost function and determine the iteration step size:

[0127]

[0128] 3.5) Update based on iterative step size

[0129] 3.6) Determine whether the iteration has converged. If and If the difference between y is less than e, or the number of iterations is reached, the iteration stops, otherwise, return to step 3.3); k represents the actual radar observation at step k, represents the estimated value of the state variable at time k; represents the estimated value of the state variable at time k+1;

[0130] 3.7) When the iteration converges or reaches the number of iterations, the calculation results are output

[0131] Embodiment 5:

[0132] This embodiment is further designed on the basis of the fourth embodiment: in step 3), the active observation data N represents the number of pixels on the profile, Z Ku represents the reflectivity factor detected by Ku-band radar, Z Ka Represents the reflectivity factor detected by Ka-band radar, PIA Ka represents the attenuation integral path of Ka-band radar detection. DFR represents the difference between the radar echoes observed in Ku-band and Ka-band (dual frequency ratio). DFR = 10log 10 Z Ku -10log 10 Z Ka .

[0133] Example 6:

[0134] This embodiment is further designed on the basis of the fourth embodiment in that:

[0135] Step 4) Input the state variable set of different types of particles obtained by regularized inversion into the atmospheric radiation transfer equation, and use the fast atmospheric radiation transfer model RTTOV software to calculate and obtain the microwave brightness temperature Tb of particles with different shapes and densities in different frequency channels of GMI. sim, and based on the temporal and spatial matching of passive microwave detection data, the microwave brightness temperature Tb observed in different channels was obtained. obs .

[0136] Step 5) Establish a least squares estimation model, input the simulated microwave brightness temperature and the observed microwave brightness temperature for inversion, calculate the least squares solution of the state variables of different types of particles, that is, obtain the optimal state variables of each type of particle, combine the optimal states of each type of particle to obtain the optimal precipitation microphysical parameter profiles at different heights, and establish the least squares estimation model as follows:

[0137]

[0138] Where x * represents the optimal state variable of the particle, represents the state variables of the particle obtained based on regularized inversion.

[0139] Embodiment seven:

[0140] The system for inverting precipitation microphysical parameter profiles based on active radar and passive microwave is characterized by including the following modules:

[0141] The particle classification module uses a discriminant method to classify hydrometeors at different heights on the precipitation profile based on precipitation products observed by active radar to obtain different types of particles.

[0142] An observation operator construction module is configured to model the microphysical properties of the different types of particles, convert the state variables of the different types of particles into simulated observation data through forward simulation, and use the simulated observation data to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, where x represents the state variable of the particle;

[0143] A state variable solving module establishes a regularized inversion model for the precipitation microphysical parameter profile. Based on the observation operator H(x), active observation data, and prior information, the regularized inversion model is iteratively solved using a gradient descent method to obtain a set of state variables for different types of particles. The active observation data is obtained through observation data from a spaceborne dual-frequency radar.

[0144] The brightness temperature acquisition module inputs the state variable set of different types of particles obtained by regularization inversion into the atmospheric radiation transfer equation, and obtains the simulated microwave brightness temperature Tb of different frequency channels through simulation calculation. sim At the same time, the observation microwave detection data of different frequency channels (passive microwave detection brightness temperature of satellite-borne microwave imager) matching the radar detection time and space are obtained to obtain the corresponding observation microwave brightness temperature Tb of different frequency channels. obs ;

[0145] Profile inversion module, establishes the least squares estimation model, inputs the simulated microwave brightness temperature Tb sim and observed microwave brightness temperature Tb obs The inversion is performed to calculate the least squares solution of the state variables of different types of particles, that is, the optimal state variables of each type of particle are obtained. The optimal state variables of each type of particle are combined to obtain the optimal precipitation microphysical parameter profiles at different heights.

[0146] Embodiment 8:

[0147] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the method described in any of the above embodiments.

[0148] Embodiment 9:

[0149] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described in any of the above embodiments.

[0150] Test example:

[0151] like Figure 4 and Figure 5 This example uses Python to invert the precipitation microphysical parameters W and Dm using GPM DPR and GMI observational data. This example uses a stratiform precipitation event that occurred over the ocean on June 10, 2019.

[0152] Figure 4 (a) and (b) are the vertical profiles of radar reflectivity factors observed in the DPR Ku and Ka bands in the example; Figure 5 (c) shows the brightness temperatures of the nine channels corresponding to this profile observed by GMI in this example. This example is a typical example of stratiform precipitation. Particle type identification reveals that the area below the melt layer (around 3.5 km) is composed of raindrops, the middle of the melt layer is primarily wet snow, and the area above the melt layer is dry snow.

[0153] Figure 5 (a) and (b) are the state parameters W and D of the particle obtained by inversion using the method of the present invention. m Vertical section; Figure 6 (d) and (e) are W and D obtained by inversion of the DPR official algorithm. m vertical section of Figure 5 (c) and (f) are the Ku-band radar reflectivity factors calculated using the microphysical parameters inverted using the method of the present invention and the official algorithm, respectively.

[0154] contrast Figure 5 (a), (b) and Figure 5 In (c) and (d), it can be found that the D inverted by the method of the present invention is m The vertical distribution of W is more consistent with the echo distribution structure observed by radar. The melting layer shows a larger D due to the melting effect. m and W, which is consistent with the theory. m Substitute the Ku-band reflectivity factor obtained in the reflectivity factor calculation equation ( Figure 5 (c)) and the reflectivity factor after attenuation correction of radar detection ( Figure 4 (a) shows significantly better consistency. The reflectivity factor and echo structure are significantly closer to the reflectivity factor calculated from the microphysical parameters inverted by the official algorithm and the radar observation, demonstrating the higher accuracy of the proposed method. The proposed algorithm considers the microphysical models of different types of hydrometeor particles in the forward operator of the dual-optimization inversion model, reducing inversion uncertainty caused by variations in the particle microphysical models. Furthermore, the inversion model incorporates the electromagnetic wave attenuation effect, avoiding the errors introduced by fitting the attenuation coefficient in the linear attenuation correction method used in traditional inversion algorithms. As a result, the proposed method inverts the microphysical vertical profiles with higher accuracy.

Claims

1. A method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave, characterized in that: The following steps are involved: 1) Based on active radar observations, a discriminant method is used to classify hydrometeors at different heights along the precipitation profile to obtain different types of particles. 2) Based on the microphysical property modeling of the different types of particles, the state variables of the different types of particles are converted into simulated observation data through forward simulation, and the simulated observation data are used to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, where x represents the state variable of the particle; 3) establishing a regularized inversion model for precipitation microphysical parameter profiles, iteratively solving the regularized inversion model using a gradient descent method based on the observation operator H(x) and active observation data obtained through observations from a spaceborne dual-frequency radar, to obtain a set of state variables for different types of particles; 4) Input the state variable sets of different types of particles obtained by regularized inversion into the atmospheric radiation transfer equation, and obtain the simulated microwave brightness temperature of different frequency channels through simulation calculation. At the same time, obtain the observed microwave detection data of different frequency channels with temporal and spatial matching to obtain the corresponding observed microwave brightness temperature of different frequency channels; 5) Establishing a least squares estimation model, inputting the simulated microwave brightness temperature and the corresponding observed microwave brightness temperature for inversion, calculating the least squares solution of the state variables of different types of particles, that is, obtaining the optimal state variables of each type of particle, and combining the optimal state variables of each type of particle to obtain the optimal precipitation microphysical parameter profiles at different altitudes.

2. The method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave according to claim 1, characterized in that: The precipitation product based on active radar observations in step 1) uses a discriminant method to classify hydrometeors at different heights on the precipitation profile to obtain different types of particles, specifically: 1.1) Obtaining dual-frequency radar precipitation profile observation data and background atmospheric temperature information for the precipitation area to be inverted. The background atmospheric temperature information is matched with the dual-frequency radar precipitation profile observation data through interpolation calculation. The matched observation data includes ground snowfall identification products, melt layer top and melt layer bottom detection products, precipitation type classification products, and temperature information; Select the data with precipitation detection flag greater than 0 in the observation data; 1.2) Check whether the ground snowfall indicator product in the selected data is greater than 0; if it is greater than 0, proceed to step 1.3; if not, proceed to step 1.4; 1.3) Determine whether the height corresponding to the pixel in the selected data is less than the height corresponding to a temperature of 0°; if so, the pixel corresponds to wet snow; if not, the pixel corresponds to dry snow / ice crystals; 1.4) Based on the melt layer top and melt layer bottom detection products, determine whether the melt layer top MLT and melt layer bottom MLB are detected for each profile line; if so, proceed to step 1.5; if not, proceed to step 1.6; 1.5) Further determine the precipitation type corresponding to this profile; If it is stratiform precipitation type S, pixels below the bottom of the melt layer are rain, pixels above the bottom of the melt layer but below the top of the melt layer are wet snow, and pixels above the top of the melt layer are dry snow / ice crystals. If the precipitation type of this profile is convective precipitation C, the precipitation below the bottom of the melt layer is rain, and the precipitation above the bottom of the melt layer is graupel. 1.6) Further determine the precipitation type corresponding to this profile. If it is stratiform precipitation type S, the hydrometeors on the profile are primarily dry snow / ice crystals or wet snow. Furthermore, based on the temperature information corresponding to each pixel on the profile, determine the specific hydrometeors corresponding to each pixel. A temperature greater than 0 indicates wet snow, and a temperature less than 0 indicates dry snow / ice crystals. If it is convective precipitation type C, the hydrometeors on the profile are primarily graupel.

3. The method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave according to claim 1, characterized in that: In step 2), the microphysical properties of the different types of particles are modeled, and the microphysical state variables of the different types of particles are converted into simulated observation data through forward simulation. The simulated observation data is used to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, specifically: 2.1) Using ground-based remote sensing observations and in situ microphysical observations, calculate the microphysical properties of different types of hydrometeors, including dry snow / ice crystals, wet snow, graupel, and rain. 2.2) Based on the microphysical properties of different types of particles counted in 2.1), microphysical property models of different types of particles are established, wherein the microphysical property models include a shape model, a density model, and a particle spectrum model; the particle spectra of different types of particles are all modeled as normalized gamma spectra; 2.3) Calculate the dielectric constant ε of different types of particles. For hydrometeors composed of ice crystals and rain, use the Debye equation; for hydrometeors composed of dry snow, wet snow, and graupel, use the Maxwell-Garnet equation. 2.4) Based on the shape model and density model in the microphysical property model established in 2.2) and the particle dielectric parameters calculated in 2.3), the scattering characteristics of different types of particles in the Ku band and Ka band are calculated using the T-matrix and DDA tools. The scattering characteristics include the backscattering cross section σ b and extinction cross section Q ext ; 2.5) Based on the particle spectrum model in the microphysical characteristics model established in 2.2) and the particle scattering characteristics calculated in 2.4), simulated observation data including radar reflectivity factors and integrated attenuation paths in the Ku and Ka bands are calculated: Where, K = ε-1 / ε+2, λ represents the wavelength of Ku band or Ka band, Z λ Indicates the reflectivity factor corresponding to wavelength λ, K ext,λ represents the attenuation coefficient corresponding to wavelength λ, N(D) represents the particle spectrum, σ b,λ (D) represents the backscattering cross section of a particle with a diameter of D corresponding to wavelength λ, Q ext,λ (D) represents the attenuation cross section of a particle with a diameter of D corresponding to a wavelength of λ, PIA λ represents the integrated attenuation path corresponding to wavelength λ, D max 、D min It represents the maximum and minimum values of the particle diameter, r represents the distance from the detection target to the radar, and R represents the total distance from the radar to the ground; 2.6) Statistically analyze the relationship between the radar observation parameters calculated in 2.5) and the particle state variables. Establish a quantitative relationship between the particle state variables and the simulated observation data based on a fitting method. Simulate and obtain an observation operator H(x) that takes into account the variations in the microphysical properties of different particle types. The observation operator is modeled as a polynomial: Where n represents the number of polynomial coefficients to be fitted; c i , e i Respectively represent the coefficients of the i-th term in each formula; in the observation operator H(x), x represents the state variable of the particle, x=[D m ,W], W represents the water content of the particles, D m represents the mass-weighted diameter.

4. The method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave joint according to claim 1, characterized in that: In step 3), the regularized inversion model of precipitation microphysical parameter profile is established as follows: Where argmin(f(x)) is the state variable x of the particle that makes the function f(x) reach its minimum value, σ R represents the covariance matrix of the observation error of δy=yH(x), β is the regularization constant, x b represents the prior background estimate of the particle's state variables, σ B represents the prior error covariance matrix.

5. The method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave according to claim 4, characterized in that: In step 3), a regularized inversion model of precipitation microphysical parameter profile is established based on the observation operator H(x), active observation data and prior information x b , the regularized inversion model is iteratively solved using the gradient descent method to obtain the state variable sets of different types of particles, specifically: 3.1) Setting the prior background microphysical parameters x in the regularized inversion model of precipitation microphysical parameter profiles b , and the background error covariance matrix σ B and the observation error covariance matrix σ R , where the prior background microphysical parameters x b Based on the statistical average of microphysical parameters observed by ground drop spectrum, σ B and σ R Obtained by statistically analyzing the mean square error of state variables and radar observations respectively; 3.2) Initialize the state variable x0, regularization constant β, convergence condition error e (set to be less than 0.1), and number of iterations k in the regularized inversion model; 3.3) Substitute the state variables into the observation operator to obtain the observed value, and calculate the difference between the active observation data in step k and the observed value obtained by the observation operator 3.4) Take the partial derivative of the cost function and determine the iteration step size: 3.5) Update based on iterative step size 3.6) Determine whether the iteration has converged. If and If the difference between y is less than e, or the number of iterations is reached, the iteration stops, otherwise, return to step 3.3); k represents the actual radar observation at step k, represents the estimated value of the state variable at time k; represents the estimated value of the state variable at time k+1; 3.7) When the iteration converges or reaches the number of iterations, the calculation results are output 6. The method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave joint according to claim 1, characterized in that: In step 3), the active observation data N represents the number of pixels on the profile, Z Ku represents the reflectivity factor detected by Ku-band radar, Z Ka Represents the reflectivity factor detected by Ka-band radar, PIA Ka represents the attenuation integral path of Ka-band radar detection. DFR represents the difference between the radar echoes observed in Ku-band and Ka-band (dual frequency ratio). DFR = 10log 10 Z Ku -10log 10 Z Ka .

7. The method for inverting precipitation microphysical parameter profiles based on active radar and passive microwave according to claim 1, characterized in that: In step 5), a least squares estimation model is established, specifically as follows: Where x * represents the optimal state variable of the particle, represents the state variables of the particle obtained based on regularized inversion.

8. A system for inverting precipitation microphysical parameter profiles based on active radar and passive microwave, characterized by: Includes the following modules: The particle classification module uses a discriminant method to classify hydrometeor particles at different heights on the precipitation profile based on precipitation products observed by active radar, obtaining different types of particles. An observation operator construction module is configured to model the microphysical properties of the different types of particles, convert the state variables of the different types of particles into simulated observation data through forward simulation, and use the simulated observation data to construct an observation operator H(x) that takes into account the changes in the microphysical properties of the different types of particles, where x represents the state variable of the particle; A state variable solving module establishes a regularized inversion model for the precipitation microphysical parameter profile. Based on the observation operator H(x) and active observation data, the regularized inversion model is iteratively solved using a gradient descent method to obtain a set of state variables for different types of particles. The active observation data is obtained through observation data from a spaceborne dual-frequency radar. The brightness temperature acquisition module inputs the state variable set of different types of particles obtained by regularization inversion into the atmospheric radiation transfer equation, and obtains the simulated microwave brightness temperature Tb of different frequency channels through simulation calculation. sim , and simultaneously obtain the observational microwave detection data of different frequency channels that match time and space, and obtain the corresponding observational microwave brightness temperature Tb of different frequency channels obs ; Profile inversion module, establishes the least squares estimation model, inputs the simulated microwave brightness temperature Tb sim and observed microwave brightness temperature Tb obs The inversion is performed to calculate the least squares solution of the state variables of different types of particles, that is, the optimal state variables of each type of particle are obtained. The optimal state variables of each type of particle are combined to obtain the optimal precipitation microphysical parameter profiles at different heights.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images