X-band meteorological radar accurate rainfall inversion method based on backscattering differential phase

Abstract

The invention discloses an X-band meteorological radar accurate rainfall inversion method based on a backscattering differential phase. The method comprises the following steps: firstly, determining a distribution structure of each rainfall estimation equation of a place according to a raindrop spectrum of the place observed by a raindrop spectrometer; secondly, performing T matrix simulation by utilizing the observed raindrop spectrum, and establishing a rainfall estimation equation of the place; then, a spatial sampling method of radar variables is determined, sampling is carried out, and sampled data is preprocessed; then, the rainfall rate is inverted, and the precision of the rainfall rate is verified; and finally, inverting the one-hour rainfall cumulant and verifying the precision of the one-hour rainfall cumulant. According to the invention, the precision of rainfall inversion of the X-band meteorological radar is improved, and a guarantee is provided for accurate rainfall real-time forecast.

Description

Technical Field

[0001] The invention relates to the technical field of meteorological radars, and in particular to an X-band meteorological radar precise rainfall inversion method based on backscatter differential phase. Background Art

[0002] In recent years, with the urgent need for local severe convective rainfall monitoring and the continuous expansion of corresponding rainfall products, more and more X-band meteorological radars have played an important role in rainfall observation. Compared with S-band and C-band radars, X-band radars have significant advantages in monitoring fast-developing storms at close range, in addition to their small size and low cost, due to their higher resolution and greater mobility.

[0003] Since the shape of raindrops gradually flattens as the size of raindrops increases, the relationship between radar variables and rain rate inevitably depends on the assumed model of the relationship between the raindrop axis ratio and the diameter of raindrops of equal volume. The influence of many factors such as rainfall type and air flow changes makes the various raindrop shape models proposed and the simulations based on observations more or less different from the actual situation. Therefore, making a priori assumptions about raindrop shapes will undoubtedly bring artificial errors to rainfall inversion. Differential reflectivity Z DR It is insensitive to the number concentration of raindrops (the total number of raindrops per unit volume) and is highly correlated with the shape of raindrops. It is a good parameter for obtaining the equal volume diameter and mass-weighted average diameter of raindrops. DR It helps to alleviate the uncertainties associated with the variation of raindrop spectrum and unknown raindrop shape, thereby reducing the overestimation of rainfall when large raindrops or melted hail occur, and is widely used in various rainfall inversion methods. However, as a power measurement, Z DR It is affected by radar calibration errors, attenuation, partial beam blocking and wet radome deviation, and these effects may not be eliminated in a timely and effective manner. DR The main problems in performing rainfall inversion.

[0004] Backscatter differential phase δ hv It is an important polarization parameter of X-band weather radar, because in the X-band, the scattering of electromagnetic waves by moderate rain is already in a non-Rayleigh state. hv is significant. DR Same, δ hv It is also insensitive to the number concentration of the raindrop spectrum and is highly correlated with the raindrop shape, helping to mitigate uncertainties associated with the variation of the raindrop spectrum and the unknown raindrop shape. This can reduce the overestimation of rainfall when large raindrops or melting hail occur. Importantly, as a non-power measurement, δ hv Therefore, in δ hvWhen the estimated value is accurate, the hv The X-band weather radar rainfall inversion is appreciable and feasible. Summary of the invention

[0005] The present invention provides an X-band meteorological radar rainfall inversion method based on backscatter differential phase, which mainly solves the technical problem of solving the problem that the existing method cannot avoid the rainfall inversion deviation caused by radar system deviation, and the problem that factors such as raindrop shape in heavy rainfall scenarios reduce the rainfall inversion accuracy, thereby improving the accuracy of radar rainfall (especially heavy rainfall) inversion.

[0006] In order to solve the above technical problems, the present invention provides an X-band meteorological radar accurate rainfall inversion method based on backscatter differential phase, comprising:

[0007] Step 1: According to the raindrop spectrum of a certain place observed by the raindrop spectrometer, determine the distribution structure of each rainfall estimation equation in the place;

[0008] Step 2: Use the observed raindrop spectrum to perform T matrix simulation and establish the rainfall estimation equation for the area;

[0009] Step 3: Determine the spatial sampling method of radar variables and implement sampling, and pre-process the sampled data;

[0010] Step 4: Invert the rainfall rate and verify its accuracy;

[0011] Step 5: Invert the 1-hour rainfall accumulation and verify its accuracy.

[0012] Furthermore, in step 1, the method for determining the distribution structure of the rainfall estimation equation is:

[0013] According to the raindrop spectrum of a certain place observed by the raindrop spectrometer, the horizontal polarization attenuation rate A of the radar variable in that place is determined. H (dB / km), differential phase shift rate K DP (° / km), backscatter differential phase δ hv The relationship between (°) and rainfall rate R (mm / h), that is, the rainfall estimation equation R (A H ), R L (K DP ) [liquid], R L (K DP ,δ hv ) [liquid], R M (K DP )

Ice water mixed

[0014] Furthermore, the step 2 includes the following sub-steps:

[0015] Step 2.1, radar variable theory simulation:

[0016] Assume that the shape and size model of raindrops is

[0017] γ=0.9951+0.0251D-0.03644D 2 +0.005303D 3 -0.0002492D 4 (1)

[0018] Where γ is the ratio of the major axis to the minor axis of the raindrop, and D (mm) is the diameter of the raindrop of equal volume. The direction of the raindrop is a two-dimensional axisymmetric Gaussian distribution, with an average tilt angle of 0 and a width of the tilt angle distribution of 10°. The T matrix method is used to calculate the theoretical value of the radar variable for non-spherical particles with a fixed direction.

[0019] Step 2.2: Establish the rainfall estimation equation:

[0020] Appropriate weight regression was performed in the logarithmic domain to improve the fitting results. H or K DP , the rainfall estimation equation is obtained using standard weighted least squares fitting. For two variables K DP and δ hv Each data is weighted by the rainfall rate calculated by the “Iscov” function in Matlab. H Dependence on temperature, R(A H ) is temperature-dependent. The optimal power-law relationship between horizontal and vertical polarization at different temperatures in the X-band is much more stable than that in the S-band and C-band, and it hardly changes when the temperature exceeds 20°C. At the same time, if a single R(A) is used in the X-band, H ) relationship, a temperature change of 5°C will result in indistinguishable deviations. Therefore, a single R(A) can be used in the X-band H ) relationship, ignoring the influence of radar wavelength and temperature. The form of the rainfall estimation equation is

[0021]

[0022] Among them, a1, a2, b1, b2, c1, c2, c3, d1, d2, e1, e2, and e3 are all constant coefficients.

[0023] Furthermore, the step 3 includes the following sub-steps:

[0024] Step 3.1: Determine the spatial sampling method:

[0025] For the spatial sampling method of radar data, 9 distance bins were considered, including the location and front and rear positions of the rain gauge, as well as the radials on both sides of the radial. The rainfall rates of the 9 distance bins were inverted using radar variables, and the rainfall rate with the smallest absolute difference from the rain gauge measurement was selected for comparison. and Inverting the rainfall rate in the corresponding area or directly averaging the rainfall rate in space can reduce K DP and δ hv However, in order to overcome the problem of different variables measured by radar and rain gauges and to make sure that radar can detect rainfall measured by rain gauges, the average radar rainfall rate across multiple range bins is not considered.

[0026] Step 3.2, data preprocessing:

[0027] In order to avoid ground clutter, bio-echoes and abnormal propagation, the hv A was performed within the range of >0.85 H , K DP and δ hv The estimation of the rainfall and the subsequent rainfall inversion are performed. The echoes from the areas affected by the non-uniform beam filling are also removed. In addition, speckle filtering is performed to reduce the influence of isolated points and segments.

[0028] Furthermore, in step 4, the method for verifying the rainfall rate inversion accuracy is:

[0029] The commonly used rainfall estimation equation R(Z H ) and R(Z H ,Z DR ) as a comparison with the present invention, is expressed as:

[0030]

[0031] where ζ H and DR are the reflectance Z in linear units H (dBZ) and differential reflectivity Z DR (dB), f1, f2, g1, g2, g3 are all constant coefficients.

[0032] Furthermore, the step 5 includes the following sub-steps:

[0033] Step 5.1, Calculation of rainfall accumulation in 1 hour:

[0034] Although the temporal and spatial sampling methods have been optimized and the good performance of the rain gauge has been ensured, local rainfall peaks with high temporal and spatial variation characteristics may still lead to underestimation or overestimation of rainfall rates, as the heaviest rainfall may occur without radar measurements. Since the spatial variability of the time integral is weak, the influence of different sampling scales is relatively less important, so further comparison of the radar-inverted 1-hour rainfall accumulation with the 1-hour rainfall accumulation measured by the rain gauge is still considered. The radar-inverted 1-hour rainfall accumulation is calculated as follows

[0035]

[0036] Where R hour is the time series of rainfall rate {R i ,i=1,…,n} accumulated rainfall in one hour, R i It only expresses the rainfall rate at the lowest elevation angle of the radar; R i and R i+1 The average value represents the value from t i and t i+1 The rainfall rate in the starting time window; t st and t ed are the start time and end time of the time series respectively; t i and t i+1 For R i and R i+1 The start or end time of the

[0037] Step 5.2: Verification of the accuracy of the inversion of 1-hour rainfall accumulation:

[0038] The statistics of root mean square error (RMSE), normalized bias (NB) and Pearson correlation coefficient (CORR) are used as indicators of algorithm performance, expressed as:

[0039]

[0040] Where RA is the 1-hour rainfall accumulation observed by the rain gauge or retrieved by the radar. The unit of RMSE is mm, the unit of NB is %, and CORR is dimensionless. The horizontal line and the angle brackets represent the average value.

[0041] Beneficial effects:

[0042] 1. Although the radar variable differential reflectivity Z DR It helps to alleviate the uncertainties associated with the variation of raindrop spectrum and unknown raindrop shape. It is widely used in rainfall inversion, but as a power measurement, Z DRThe backscatter differential phase δ is one of the important polarization parameters of X-band meteorological radar. hv It is also insensitive to the number concentration of the raindrop spectrum and is highly correlated with the raindrop shape, which helps to alleviate the uncertainties associated with the variation of the raindrop spectrum and the unknown raindrop shape, and can reduce the overestimation of rainfall when large raindrops or melt hail occur. Importantly, as a non-power measurement, δ hv It can also maintain good performance.

[0043] 2. In the X-band, the scattering of electromagnetic waves by moderate rain is already in a non-Rayleigh state. hv is significant. Therefore, in δ hv When the estimated value is accurate, the hv The X-band weather radar rainfall inversion is appreciable and feasible. The present invention proposes an X-band weather radar accurate rainfall inversion method based on backscatter differential phase, and performs rainfall inversion using backscatter differential phase.

[0044] 3. The present invention is applied to perform radar rainfall inversion based on data of rainfall events observed by X-band weather radar in Oklahoma, USA. The comparison results with rain gauge data show that the accuracy of radar rainfall inversion in the region has been improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a logic diagram of an X-band weather radar accurate rainfall inversion method based on backscatter differential phase according to Embodiment 1 of the present invention;

[0046] Figure 2 The attenuation correction of the first embodiment of the present invention is R(Z H ) The inverted rainfall rate domain (storm observed by X-SAPR SE on October 4, 2017);

[0047] Figure 3 The attenuation correction of the first embodiment of the present invention is R(Z H ,Z DR ) The inverted rainfall rate domain (storm observed by X-SAPR SE on October 4, 2017);

[0048] Figure 4 The attenuation correction and calibration error compensation of the first embodiment of the present invention are shown in Figure 1 at an elevation angle of 2.5°. H ) The inverted rainfall rate domain (storm observed by X-SAPR SE on October 4, 2017);

[0049] Figure 5 The attenuation correction and calibration error compensation of the first embodiment of the present invention are shown in Figure 1 at an elevation angle of 2.5°. H ,Z DR ) The inverted rainfall rate domain (storm observed by X-SAPR SE on October 4, 2017);

[0050] Figure 6 The R(A) of the embodiment 1 of the present invention at an elevation angle of 2.5° H ) and R L (K DP ) The inverted rainfall rate domain (storm observed by X-SAPR SE on October 4, 2017);

[0051] Figure 7 The rainfall rate domain inverted by the FAB method of the present invention at an elevation angle of 2.5° in Example 1 of the present invention (the storm observed by X-SAPR SE on October 4, 2017);

[0052] Figure 8 The R(A) of the embodiment 1 of the present invention at an elevation angle of 2.5° H ) and R L (K DP ) inverted rainfall rate domain (storm observed by X-SAPR SW on October 6, 2016);

[0053] Fig. 9 The R(A) of the embodiment 1 of the present invention at an elevation angle of 2.5° H ), R L (K DP ) and R M (K DP ) inverted rainfall rate domain (storm observed by X-SAPR SW on October 6, 2016);

[0054] Fig.10 The rainfall rate domain inverted by the FAB method of the present invention at an elevation angle of 2.5° in Example 1 of the present invention (the storm observed by X-SAPR SW on October 6, 2016);

[0055] Fig.11 The first embodiment of the present invention is composed of R(Z H ,Z DR ) Composite rainfall rate domain obtained by merging the rainfall rate domain of the inverted X-SAPR SE with the rainfall rate domain of the nearby S-band radar KINX (storm observed by X-SAPR SW on October 6, 2016);

[0056] Fig.12A composite rainfall rate domain map obtained by combining the rainfall rate domain of X-SAPR SE inverted by the method FAB of the present invention and the rainfall rate domain of the nearby S-band radar KINX (the storm observed by X-SAPR SW on October 6, 2016) in Example 1 of the present invention;

[0057] Fig.13 is R(Z) of the first embodiment of the present invention H ,Z DR ) The inverted 1-hour rainfall accumulation domain (stratiform precipitation observed by XSAPR SE on February 20, 2017);

[0058] Fig.14 This is the 1-hour rainfall accumulation domain (stratigraphic precipitation observed by XSAPR SE on February 20, 2017) inverted by the method FAB of the present invention in Example 1 of the present invention. DETAILED DESCRIPTION

[0059] In order to make the purpose and technical solution of the present invention more clearly understood, the present invention is further described in detail below through specific implementation methods in combination with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0060] Embodiment 1:

[0061] In this embodiment, a non-uniformly distributed scanning Doppler radar network of the U.S. Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) project located at the Southern Great Plains (SGP) site in Oklahoma, USA, is used, which contains three X-band scanning ARM precipitation radars (X-SAPRs) placed in a triangular array. X-SAPR is a polarimetric Doppler weather radar that can simultaneously transmit and receive horizontal and vertical polarizations to provide measurement data for identifying precipitation types and inverting rainfall rates. Three laser drop spectrometers (LDIS) and a video drop spectrometer (VDIS) are deployed in the overlapping parts of the three X-SAPRs, and their triangular array structures are opposite to capture detailed information on the size and velocity of individual hydrometeors. At the same time, within the coverage of the X-SAPR, a Pluvio2 weighing bucket rain gauge, three Novalynx tipping bucket rain gauges, and three TR-525M metric heated rain gauges are deployed to support the verification of radar rainfall inversion. Due to beam blocking, data at elevation angles of 0.5°, 1.5°, and 2.5° are selected to estimate rainfall. In addition, the observation radius of X-SAPR is 50 km, so even in the cold season formation, the rainfall inversion performance of the algorithm is mainly affected by the change of raindrop spectrum and possible frozen hydrometeors such as hail, rather than by bright band contamination. During the observation period, 4 raindrop spectrometers collected a large number of observed raindrop spectra.

[0062] This embodiment provides an X-band meteorological radar accurate rainfall inversion method based on backscatter differential phase, which specifically includes the following steps:

[0063] Step 1: According to the raindrop spectrum of a certain place observed by the raindrop spectrometer, determine the distribution structure of each rainfall estimation equation in that place:

[0064] The distribution structure of rainfall estimation equation is determined as follows:

[0065] According to the raindrop spectrum observed by the raindrop spectrometer, the horizontal polarization attenuation rate A of the radar variable in the area is determined. H (dB / km), differential phase shift rate K DP (° / km), backscatter differential phase δ hv The relationship between (°) and rainfall rate R (mm / h), that is, the rainfall estimation equation R (A H ), R L (K DP ) [Liquid], R L (K DP ,δ hv ) [Liquid], R M (K DP )

Ice water mixed

[0066] According to the raindrop spectrum of the area, when R>3mm / h in the X-band, R(K DP ) is less sensitive to the change of raindrop spectrum than R(A H ), so when R<3mm / h, R(A H ) relationship, when R>3mm / h, R L (K DP ) relationship. L (K DP ) compared to R L (K DP ,δ hv ) is less affected by the uncertainty of the raindrop spectrum and the unknown shape of the raindrops. However, in light rain, the approximate spherical shape of the raindrops makes δ hv is too small, or even submerged in the noise, so R L (K DP ,δ hv ) to invert δ hv >T δhv1It is necessary to identify areas where ice and water are mixed and use R M (K DP ) relationship instead of R L (K DP ) relationship. Since the initially rolling dry hail gradually turns into melted hail or large raindrops containing ice cores during the melting process, the uncertainty of the shape at this time will greatly weaken the accuracy of precipitation inversion. Therefore, using R M (K DP ,δ hv ) to invert the rainfall rate. Hail can be obtained by high Z H (there is still potential bias) with low Z DR and ρ hv Combined with recognition, the present invention sets the threshold T ZH =53dBZ, T ZDR =1.5dB, T ρhv =0.92.

[0067] Step 2: Use the observed raindrop spectrum to perform T matrix simulation and establish the rainfall estimation equation for the area, including:

[0068] Step 2.1, radar variable theory simulation:

[0069] Assume that the shape and size model of raindrops is

[0070] γ=0.9951+0.0251D-0.03644D 2 +0.005303D 3 -0.0002492D 4 (13)

[0071] Where γ is the ratio of the major axis to the minor axis of the raindrop, and D (mm) is the diameter of the raindrop of equal volume. The direction of the raindrop is a two-dimensional axisymmetric Gaussian distribution, with an average tilt angle of 0 and a width of the tilt angle distribution of 10°. The T matrix method is used to calculate the theoretical value of the radar variable for non-spherical particles with a fixed direction.

[0072] Step 2.2: Establish the rainfall estimation equation:

[0073] Appropriate weight regression was performed in the logarithmic domain to improve the fitting results. H or K DP , the rainfall estimation equation is obtained using standard weighted least squares fitting. For two variables K DP and δ hv Each data is weighted by the rainfall rate calculated by the “Iscov” function in Matlab. H Dependence on temperature, R(A H) is temperature-dependent. The optimal power-law relationship between horizontal and vertical polarization at different temperatures in the X-band is much more stable than that in the S-band and C-band, and it hardly changes when the temperature exceeds 20°C. At the same time, if a single R(A) is used in the X-band, H ) relationship, a temperature change of 5°C will result in indistinguishable deviations. Therefore, a single R(A) can be used in the X-band H ) relationship, ignoring the influence of radar wavelength and temperature. The form of the rainfall estimation equation is

[0074]

[0075]

[0076] For the threshold T δhv1 and T δhv2 First, according to the raindrop shape model, raindrops with a diameter greater than about 1 mm are non-spherical. hv The simulation of the estimation method and results shows that when the deviation is less than about 0.5°, δ hv The deviation is large, with the maximum deviation being about 0.35°. Third, since most light rain and moderate rain samples (e.g., based on the observed raindrop spectra, the δ hv It accounts for more than half of the total) and plays a decisive role in the fitting, so it is necessary to ensure that too small δ is excluded in the fitting. hv values ​​so that the fitted relationship can better reflect heavy rain or storms. Fourth, in areas where ice and water are mixed, δ hv There are few samples with a value greater than about 2°. Therefore, T δhv1 Set to 0.5°, 1° and 1.5°, and set T δhv2 0.5°, and by analyzing the performance of the algorithm under all three possible combinations, it is found that at least in Oklahoma, T δhv1 It does not seem to have a significant impact on rainfall inversion, and perhaps it can be fixed to a lower value to better play its role. δhv1 and T δhv2 Set to 0.5°.

[0077] Step 3: Determine the spatial sampling method of radar variables and implement sampling, and preprocess the sampled data, including:

[0078] Step 3.1: Determine the spatial sampling method:

[0079] For the spatial sampling method of radar data, 9 distance bins (i.e., radar distance resolution units) were considered, including the distance bin where the rain gauge is located, and 8 distance bins in front, behind, left, right, and diagonal directions. The rainfall rates of the 9 distance bins were inverted using radar variables, and the rainfall rate with the smallest absolute difference from the rain gauge measurement was selected for comparison. Although spatial averaging is used, the rainfall rate of the rain gauge is not very accurate. and Inverting the rainfall rate in the corresponding area or directly averaging the rainfall rate in space can reduce K DP and δ hv However, in order to overcome the problem of different variables measured by radar and rain gauge and to make sure that radar can detect rain measured by rain gauge, the average radar rainfall rate over multiple range bins is not considered.

[0080] Step 3.2, data preprocessing:

[0081] In order to avoid ground clutter, bio-echoes and abnormal propagation, the hv A was performed within the range of >0.85 H , K DP and δ hv The estimation of the rainfall and the subsequent rainfall inversion are performed. The echoes from the areas affected by the non-uniform beam filling are also removed. In addition, speckle filtering is performed to reduce the influence of isolated points and segments.

[0082] Step 4: Invert the rainfall rate and verify its accuracy:

[0083] Verification of rainfall rate inversion accuracy:

[0084] The commonly used rainfall estimation equation R(Z H ) and R(Z H ,Z DR ) is compared with the inversion result of the method FAB of the present invention, and is expressed as:

[0085]

[0086] where ζ H and DR are the reflectance Z in linear units H (dBZ) and differential reflectivity Z DR (dB).

[0087] The rainfall rate domain inverted from the storm observed by X-SAPR SE on October 4, 2017 at the antenna elevation angle EL = 2.5° is as follows: Figure 2 to Figure 7 As shown in Figure 2, due to the calibration error of the radar, only the attenuation is corrected without correcting the calibration deviation. H ) and R(Z H ,Z DR)Inverted rainfall rate domain ( Figure 2 and Figure 3 ) has obvious deviations, and under the influence of partial beam blocking, the rainfall rate is significantly underestimated. After correction, the problem of partial beam blocking still exists, such as Figure 4 and Figure 5 As shown. H and K DP The estimate of is not affected by partial beam blocking, so in [R(A H ),R L (K DP )] domain completely eliminates the deviation caused by blocking ( Figure 6 ). At the same time, [R(A H ),R L (K DP )] domain and R(Z H ) domains have good structural consistency and the same resolution. hv It is also not affected by partial beam blocking, so there is no area where the rainfall rate is underestimated in the rainfall rate domain of the FAB inversion method of the present invention ( Figure 7 ). In addition, under the influence of heavy raindrops, [R(A H ),R L (K DP The rainfall rate in the convective core region of the domain is overestimated. hv As the raindrops get bigger, it increases, including δ hv The estimated value of takes better account of the change in raindrop shape. Therefore, the method FAB of the present invention can reduce [R(A H ),R L (K DP )] domain, although the changes do not appear to be particularly dramatic.

[0088] In areas where ice and water are mixed, apply R M (K DP ) instead of R L (K DP ). The rainfall rates of X-SAPR SW inverted by different methods for the hailstorm observed on October 6, 2016 are shown in the following table. Figure 8 to Figure 10 The black circle is a severe convective cell containing hail, where the rain gauge shows a rainfall rate of 36.3 mm / h. H ),R L (K DP )]domain( Figure 8 ) compared to [R(A H ),R L (K DP ),R M (K DP )]domain( Fig. 9 ) The rainfall rate in the hail mixed area is lower, but still higher than the rain gauge value. DP cannot contain all rainfall information, δ hv contains more raindrop spectrum information, so Fig.10 The rainfall rate domain of the FAB inversion of the method of the present invention is shown to be more consistent with the rain gauge value. Not only the rainfall rate in the convective core is reduced, but also the rainfall rate around the convective core is reduced.

[0089] Since the method FAB of the present invention is not affected by radar calibration errors and partial beam blocking, it can be seamlessly synthesized with rainfall products obtained by nearby radars even if the radar wavelengths are different. Fig.11 and Fig.12 The composite rainfall rate domain map of the event on October 6, 2016 is composed of R(Z H ,Z DR ) and the rainfall rate domain of the X-SAPR SE inverted by the method of the present invention FAB is combined with the rainfall rate domain of the nearby S-band radar KINX. Due to the calibration error of the X-SAPR SE, R(Z H ,Z DR ) domain shows obvious deviation. The rainfall rate domain of X-SAPR SE inverted by the FAB method of the present invention is more consistent with the rainfall rate domain of KINX, and there is no discontinuity on the equidistant line, but the resolution of the rainfall rate domain of KINX is significantly lower than that of X-SAPR SE, resulting in a certain degree of roughness on the equidistant line.

[0090] Step 5: Invert the 1-hour rainfall accumulation and verify its accuracy, including:

[0091] Step 5.1, calculation of rainfall accumulation in 1 hour:

[0092] Although the temporal and spatial sampling methods have been optimized and the good performance of the rain gauge has been ensured, local rainfall peaks with high temporal and spatial variation characteristics may still lead to underestimation or overestimation of rainfall rates, as the heaviest rainfall may occur without radar measurements. Since the spatial variability of the time integral is weak, the influence of different sampling scales is relatively less important, so further comparison of the radar-inverted 1-hour rainfall accumulation with the 1-hour rainfall accumulation measured by the rain gauge is still considered. The radar-inverted 1-hour rainfall accumulation is calculated as follows

[0093]

[0094] Where R hour is the time series of rainfall rate {R i ,i=1,…,n} accumulated rainfall in one hour, R i It only expresses the rainfall rate at the lowest elevation angle of the radar; Ri and R i+1 The average value represents the value from t i and t i+1 The rainfall rate in the starting time window; t st and t ed are the start time and end time of the time series respectively; t i and t i+1 For R i and R i+1 The start or end time of the

[0095] Step 5.2: Verification of the accuracy of the inversion of 1-hour rainfall accumulation:

[0096] The statistics of root mean square error (RMSE), normalized bias (NB) and Pearson correlation coefficient (CORR) are used as indicators of algorithm performance, expressed as:

[0097]

[0098] Where RA is the 1-hour rainfall accumulation observed by the rain gauge or retrieved by the radar. The unit of RMSE is mm, the unit of NB is %, and CORR is dimensionless. The horizontal line and the angle brackets represent the average value.

[0099] For the stratiform precipitation observed by XSAPR SE on February 20, 2017, R(Z H ,Z DR ) and the 1-hour rainfall accumulation domain inverted by the method FAB of the present invention are as follows Fig.13 and Fig.14 As shown. Similarly, in the 1-hour rainfall accumulation domain inverted by the FAB method of the present invention, the influence of the southern sector blocking is completely eliminated. In addition, R(Z H ,Z DR ) domain, the reason is unknown. However, no blocking occurs in the same area of ​​the 1-hour rainfall accumulation domain inverted by the FAB method of the present invention.

[0100] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. An X-band weather radar accurate rainfall inversion method based on backscatter differential phase, characterized in that: The following steps are involved: Step 1: According to the raindrop spectrum of a certain place observed by the raindrop spectrometer, determine the distribution structure of each rainfall estimation equation in the place; Step 2: Use the observed raindrop spectrum to perform T matrix simulation and establish the rainfall estimation equation for the area; Step 3: Determine the spatial sampling method of radar variables and implement sampling, and pre-process the sampled data; Step 4: Invert the rainfall rate and verify its accuracy; Step 5: Invert the 1-hour rainfall accumulation and verify its accuracy.

2. The method according to claim 1, characterized in that In step 1, the method for determining the distribution structure of the rainfall estimation equation is: According to the raindrop spectrum of a certain place observed by the raindrop spectrometer, the horizontal polarization attenuation rate A of the radar variable in that place is determined. H (dB / km), differential phase shift rate K DP (° / km), backscatter differential phase δ hv The relationship between (°) and rainfall rate R (mm / h), that is, the rainfall estimation equation R (A H ), R L (K DP ) [liquid], R L (K DP ,δ hv ) [liquid], R M (K DP )【Ice water mixed】、R M (K DP ,δ hv )The distribution structure of the ice-water mixed state; since both forward scattering (forescatter) and backscattering (backscatter) are used, this method is called FAB.

3. The method according to claim 1, characterized in that The step 2 includes the following sub-steps: Step 2.1, radar variable theory simulation; Step 2.2: Establish the rainfall estimation equation.

4. The method according to claim 1, characterized in that In step 2, the rainfall estimation equation is in the form of Among them, a1, a2, b1, b2, c1, c2, c3, d1, d2, e1, e2, and e3 are all constant coefficients.

5. The method according to claim 1, characterized in that The step 3 includes the following sub-steps: Step 3.1, determine the spatial sampling method; Step 3.2: Data preprocessing.

6. The method according to claim 1, characterized in that In step 3, for the spatial sampling method of radar data, 9 distance bins (i.e., radar distance resolution units) are considered, including the distance bin where the rain gauge is located, and 8 distance bins in front, behind, left, right, and diagonal directions of the distance bin, totaling 8 distance bins; the rainfall rates of the 9 distance bins are inverted using radar variables, and the rainfall rate with the smallest absolute difference from the rain gauge measurement value is selected for comparison.

7. The method according to claim 1, characterized in that In step 3, in order to avoid ground clutter, biological echoes and abnormal propagation, hv A was performed within the range of >0.85 H , K DP and δ hv The estimation and subsequent rainfall inversion are performed; echoes from areas affected by non-uniform beam filling are also eliminated; speckle filtering is also performed to reduce the influence of isolated points and segments.

8. The method according to claim 1, characterized in that In step 4, the method for verifying the rainfall rate inversion accuracy is: The commonly used rainfall estimation equation R(Z H ) and R(Z H ,Z DR ) as a comparison with the present invention, is expressed as: where ζ H and DR are the reflectance Z in linear units H (dBZ) and differential reflectivity Z DR (dB), f1, f2, g1, g2, g3 are all constant coefficients.

9. The method according to claim 1, characterized in that The step 5 includes the following sub-steps: Step 5.1, calculation of rainfall accumulation in 1 hour: Step 5.2: Verify the accuracy of the inversion of 1-hour rainfall accumulation.

10. The method according to claim 1, characterized in that In step 5, the 1-hour rainfall accumulation inverted by radar is calculated according to the following formula: Where R hour is the time series of rainfall rate {R i ,i=1,…,n} accumulated rainfall in one hour, R i It only expresses the rainfall rate at the lowest elevation angle of the radar; R i and R i+1 The average value represents the value from t i and t i+1 The rainfall rate in the starting time window; t st and t ed are the start time and end time of the time series respectively; t i and t i+1 For R i and R i+1 The start or end time of the

Images