A Method for Simultaneously Retrieving Wind Speed and Precipitation by a Ku-Band Microwave Scatterometer

Abstract

The present invention relates to the fields of satellite remote sensing, ocean remote sensing, and meteorological technology, and particularly relates to a method for simultaneously retrieving wind speed and precipitation using a Ku-band microwave scatterometer. The method includes: selecting relevant parameters of the wind observation unit of the Ku-band scatterometer that meet the quality control requirements, and inputting them into a specific model established and trained in advance for simultaneously retrieving wind speed and precipitation using the Ku-band scatterometer, so as to achieve the simultaneous estimation of wind speed and precipitation under the set conditions, where the set conditions are that the wind speed is less than 14 m / s and the precipitation is less than 10 mm / h. Based on a general method of support vector machines, the present invention establishes a specific model for identifying the precipitation influence area of the Ku-band scatterometer, removing the precipitation influence on wind speed, and quantitatively estimating precipitation information, so as to achieve the simultaneous estimation of wind speed and precipitation below 14 m / s (meters per second) and 10 mm / h (millimeters per hour) of precipitation.

Description

Technical Field

[0001] The present invention relates to the fields of satellite remote sensing, ocean remote sensing and meteorological technology, and in particular to a method for simultaneously inverting wind speed and precipitation using a Ku-band microwave scatterometer. Background Art

[0002] As a remote sensing sensor capable of efficiently acquiring ocean surface winds, the technology for retrieving ocean surface winds from observations provided by spaceborne microwave scatterometers has undergone significant development over the past 40 years. Most microwave scatterometers used to measure ocean surface winds operate in the C or Ku bands. The presence of precipitation or precipitation clouds can affect the wind-generated echoes from the ocean surface, thereby affecting the wind retrieval results. The impact in the shorter-wavelength Ku band is approximately 10 times greater than in the C band. In current conventional processing, quality control methods are used to eliminate data whose retrieval results deviate significantly from the modeled ocean scene, including data affected by precipitation. Previous studies have shown that while precipitation has complex characteristics and can affect wind retrieval results, its influence on wind measurements also includes precipitation information. Quantifying this information can facilitate wind correction and precipitation information extraction. Currently, a common approach is to estimate precipitation and wind information using physical models based on electromagnetic scattering or Bayesian theory, leveraging precipitation and wind information provided by numerical prediction models. However, due to differences in the temporal and spatial scales of numerical models and scatterometer observations, the obtained results also exhibit the same characteristics. Currently, there is no method to identify precipitation areas, correct wind speeds for precipitation effects, or estimate precipitation based on observations themselves. Summary of the Invention

[0003] The purpose of the present invention is to overcome the defects of the prior art and propose a method for simultaneous inversion of wind speed and precipitation using a Ku-band microwave scatterometer.

[0004] To achieve the above object, the present invention proposes a method for simultaneous inversion of wind speed and precipitation using a Ku-band microwave scatterometer, the method comprising:

[0005] The relevant parameters of the Ku-band scatterometer wind observation unit that meet quality control are selected, and a pre-established and trained specific model for simultaneous inversion of wind speed and precipitation for the Ku-band scatterometer is input to achieve simultaneous estimation of wind speed and precipitation under set conditions. The set conditions are wind speed less than 14 meters per second and precipitation less than 10 millimeters per hour.

[0006] As an improvement to the above method, the Ku-band scatterometer observation wind unit meets the quality control mark for the wind unit with precipitation, and the relevant parameters used include: the quality control factor maximum likelihood estimation residual parameter MLE, the observed wind speed f, the wind direction two-dimensional variational defuzzification reanalysis wind speed less than 14 meters per second, and the precipitation influence factor α; wherein,

[0007] The maximum likelihood estimation residual parameter MLE of the quality control factor is the sum of the normalized Euclidean distances from the observed backscattering coefficient in the wind cell to the geophysical model function.

[0008] The observed wind speed f is the wind speed inverted using the conventional process.

[0009] The precipitation influence degree factor α is defined as the Joss factor in the wind speed objective function in quality control normalized by the difference between the observed wind speed and 18, and satisfies the following formula:

[0010] α = Joss / (f - 18).

[0011] As an improvement of the above method, it is characterized in that the specific model for simultaneous inversion of wind speed and precipitation for the Ku - band scatterometer includes a wind speed correction support vector machine model, a precipitation identification support vector machine model, and a precipitation inversion support vector machine model; among them,

[0012] The wind speed correction support vector machine model is used to output the wind speed after correcting for the influence of precipitation.

[0013] The precipitation identification support vector machine model is used to output a precipitation presence marker, where 0 indicates no precipitation and 1 indicates precipitation.

[0014] The precipitation inversion support vector machine model is used to output the precipitation amount.

[0015] As an improvement of the above method, the method further includes the training step of the wind speed correction support vector machine model; specifically including:

[0016] Using the spatio - temporal matching observation data of the C - band and Ku - band scatterometers as input, and using the spatio - temporal matching observation quality control result of the C - band with the accepted wind speed as the output true value, the wind speed correction support vector machine model is trained.

[0017] As an improvement of the above method, the method further includes the training step of the precipitation identification support vector machine model; specifically including:

[0018] Using the spatio - temporal matching observation data of the C - band and Ku - band scatterometers, selecting the wind cells that simultaneously satisfy the Ku - band observation quality control result of rejection and the C - band observation quality control result of acceptance, and using the relevant parameters corresponding to these wind cells as input; for a precipitation value of 0 mm / h, it is marked as 0, indicating no precipitation, otherwise it is marked as 1, indicating precipitation, and using the mark as the output true value, the precipitation identification support vector machine model is trained.

[0019] As an improvement of the above method, the method further includes the training step of the precipitation inversion support vector machine model; specifically including:

[0020] Using the precipitation - impact - related parameters of the spatio - temporal matching observation data of C - band and Ku - band scatterometers as inputs, and using the area - weighted average value of the remote - sensing precipitation product within the wind cell as the output true value, train the precipitation inversion support vector machine model. The precipitation - impact - related parameters include the maximum likelihood estimation residual parameter MLE of the quality control factor, the observed wind speed f, the two - dimensional variational de - blurred re - analysis wind speed of the wind direction less than 14 m / s, and the precipitation impact degree factor α.

[0021] Compared with the prior art, the advantages of the present invention are as follows:

[0022] Based on the general method of support vector machines, a specific model for identifying the precipitation - affected area of Ku - band scatterometers, removing the precipitation impact on wind speed, and quantitatively estimating precipitation information is established, realizing the simultaneous estimation of wind speed and precipitation below 14 m / s (meters per second) and 10 mm / h (millimeters per hour) of precipitation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of a method for simultaneously retrieving wind speed and precipitation by a Ku - band microwave scatterometer. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] Based on the quality control factor used in the research on the quality control of wind field inversion by microwave scatterometers, using the remote - sensing precipitation product with spatio - temporal matching of the wind field inversion unit of the scatterometer as a reference, using the spatio - temporal matching observation data of C - and Ku - band scatterometers, and based on the general method of support vector machines, a specific model for identifying the precipitation - affected area of Ku - band scatterometers, removing the precipitation impact on wind speed, and quantitatively estimating precipitation information is established, realizing the simultaneous estimation of wind speed and precipitation below 14 m / s (meters per second) and 10 mm / h (millimeters per hour) of precipitation.

[0025] Based on the general method of support vector machines, a specific model for identifying the precipitation - affected area of Ku - band scatterometers, removing the precipitation impact on wind speed, and quantitatively estimating precipitation information is established, realizing the simultaneous estimation of wind speed and precipitation below 14 m / s (meters per second) and 10 mm / h (millimeters per hour) of precipitation.

[0026] The quality control factors used in the study of microwave scatterometer wind field inversion quality control are numerically correlated with precipitation. Due to the longer wavelength, the influence of precipitation or precipitation clouds on the C-band is 10 times weaker than that on the Ku-band. Using the spatio-temporal matching observation data of C and Ku scatterometers, in the wind field where the C-band observation quality is acceptable, there will be inversion units (wind units) that are excluded due to the influence of precipitation. Since the distribution of the observed values of the backscattering coefficient in the scatterometer wind units is not unique, there are various situations that need to be modeled in direct information extraction, making the estimation and elimination of the precipitation influence with complex properties even more complex. However, in the process of wind field inversion, the quality control factor used as the benchmark for quality elimination comprehensively considers the differences between the empirical model used for inversion: the geophysical model function (GMF) that maps the observed sea surface wind field and the observations, or the changes in the spatial uniformity of the wind units caused by precipitation. For the wind units with precipitation, these factors are more statistically representative. At this time, with the help of spatio-temporal matching remote sensing precipitation products as a reference, the changes in the numerical values of the quality control factors caused by the precipitation influence in the Ku-band are analyzed in principle, and the factors correlated with precipitation are sorted out; at the same time, considering the quantities that can provide precipitation information in other scatterometer observations, the mapping from the quality control factor to precipitation and the mapping from these factors to the true wind field can be established to correct the precipitation influence on the wind speed. The analysis and research show that these two mappings have complex non-linear characteristics. Therefore, the present invention selects a general method based on support vector machines, selects suitable inputs, and establishes specific models for identifying the precipitation influence area of the Ku-band scatterometer, removing the precipitation influence on the wind speed, and quantitatively estimating the precipitation information, so as to simultaneously estimate the wind speed and precipitation below 14 m / s (meters per second) and 10 mm / h (millimeters per hour) of precipitation. For the true value of the wind speed correction, the wind speed data that are spatio-temporally matched and C-band quality-accepted (not affected by precipitation) are used; the precipitation amount is provided by the geometrically weighted average remote sensing precipitation product.

[0027] In the research of the present invention, the parameters related to the precipitation influence are established as the maximum likelihood parameter (MLE) of the quality control factor, the observed wind speed, the two-dimensional variational de-fuzzed reanalysis wind speed of the wind direction, and the precipitation influence degree factor. Among them, MLE is defined as the sum of the normalized Euclidean distances from the observed backscattering coefficient in the wind unit to the GMF; the observed wind speed is the wind speed (f) inverted using the conventional process; the precipitation influence degree factor is named α and is defined as the wind speed objective function factor (Joss) in the quality control normalized by the difference between the observed wind speed and 18:

[0028] α = Joss / (f - 18)

[0029] For the support vector machine (SVM) model for wind speed correction, during the training process, the wind speed accepted by the quality of the C-band in the C and Ku-band spatio-temporal matching data is used as the output true value; after establishing the model using two years of C and Ku-band matching data for training, for the conventional observation wind units with two-dimensional variational deblurred reanalysis wind speed less than 14 m / s and without C-band matching data, the parameters of the above-mentioned wind units after quality rejection are used as the input, and the output is the corrected wind speed.

[0030] Since the influence of precipitation may come from adjacent wind units; for the SVM model for precipitation identification, during the training process, the precipitation identification (0 mm / h means no precipitation) of the wind units based on whether the precipitation value of the precipitation product matched in space and time is 0 mm / h is used. For the wind units with precipitation, the output label is 1, otherwise it is 0. Similarly, two years of matching data are used for training. After the model is established, for the conventional observation wind units with two-dimensional variational deblurred reanalysis wind speed less than 14 m / s and without C-band matching data, the parameters of the above-mentioned wind units after quality rejection are used as the input, and the output is the identification of whether there is precipitation in this wind unit.

[0031] For the SVM model for precipitation retrieval, during the training process, the weighted average value of the precipitation amount in this wind unit covered by the precipitation product matched in space and time is used. Similarly, two years of matching data are used for training. After the model is established, for the conventional observation wind units with two-dimensional variational deblurred reanalysis wind speed less than 14 m / s and without C-band matching data, the parameters of the above-mentioned wind units after quality rejection are used as the input, and the output is the precipitation amount in this wind unit.

[0032] The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

[0033] Embodiment 1

[0034] As Figure 1 shown, the embodiment of the present invention proposes a method for simultaneously retrieving wind speed and precipitation of a Ku-band microwave scatterometer. According to the technical solution, the training process for establishing the SVM model uses the wind field and quality control identification of the microwave scatterometer matched in space and time in the C and Ku bands, as well as the two-dimensional variational deblurred reanalysis wind speed and the wind speed values observed in the corresponding two bands. The precipitation product matched in the same space and time is matched with the wind units, and the precipitation amount of the wind units is calculated by the area-weighted method. For the Ku-band wind units with two-dimensional variational deblurred reanalysis wind speed less than 14 m / s and the wind units rejected by MLE quality control, the wind units accepted by MLE quality control of the C-band matched in space and time are selected to establish the above SVM model. The application is carried out for the observation data of the Ku-band scatterometer and the wind units with two-dimensional variational deblurred reanalysis wind speed less than 14 m / s and rejected by MLE quality control.

[0035] The innovation points of the present invention:

[0036] 1) Through analysis, a factor indicating the degree of precipitation influence is proposed, denoted as α, which is the objective function factor (Joss) of two-dimensional variational defuzzification and reanalysis wind speed normalized by the difference between the observed wind speed and 18.

[0037] 2) Determine that the support vector machine model is used for the wind speed correction of the Ku-band microwave scatterometer affected by precipitation. The inputs are the maximum likelihood parameter (MLE) of the quality control factor, the observed wind speed, the two-dimensional variational defuzzification and reanalysis wind speed, and the precipitation influence degree factor α; the output during training is the C-band quality acceptance wind speed that matches in the same time and space, and during application, the inputs are the above input parameters of the conventional Ku-band scatterometer observation wind unit without C-band matching data where the two-dimensional variational defuzzification and reanalysis wind speed is less than 14 m / s, and the output is the precipitation influence corrected wind speed.

[0038] 3) Determine the support vector machine models for identifying the presence of precipitation and estimating the precipitation amount in the wind units of the Ku-band microwave scatterometer affected by precipitation. The inputs are the same as in 2), and the outputs during training are respectively the area-weighted average values of the precipitation presence identification and the remote sensing precipitation product within the wind unit; during application, the inputs are the same as in 2), and the outputs are respectively the precipitation presence identification and the precipitation amount.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the embodiments, those of ordinary skill in the art should understand that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for simultaneously retrieving wind speed and precipitation by a Ku - band microwave scatterometer, the method comprising: Selecting relevant parameters of the wind observation unit of the Ku - band scatterometer that meet the quality control, and inputting them into a specific model for simultaneously retrieving wind speed and precipitation for the Ku - band scatterometer established and trained in advance, to achieve the simultaneous estimation of wind speed and precipitation under the set conditions, where the set conditions are that the wind speed is less than 14 m / s and the precipitation is less than 10 mm / h; Relevant parameters used for the wind units marked as precipitation present in the Ku - band scatterometer observation wind unit that meet the quality control, the relevant parameters including: the maximum likelihood estimation residual parameter MLE of the quality control factor, the observed wind speed f, the two - dimensional variational de - blurred re - analysis wind speed of the wind direction less than 14 m / s, and the precipitation influence degree factor α; where, The maximum likelihood estimation residual parameter MLE of the quality control factor is the sum of the normalized Euclidean distances from the observed backscattering coefficient in the wind unit to the geophysical model function; The observed wind speed f is the wind speed retrieved using the conventional process; The precipitation influence degree factor α is defined as the wind speed objective function factor Joss in the quality control normalized by the difference between the observed wind speed and 18, satisfying the following formula: α = Joss / (f - 18); The specific model for simultaneously retrieving wind speed and precipitation for the Ku - band scatterometer includes a wind speed correction support vector machine model, a precipitation identification support vector machine model, and a precipitation inversion support vector machine model; where, The wind speed correction support vector machine model is used to output the wind speed after correcting for the influence of precipitation; The precipitation identification support vector machine model is used to output a precipitation presence mark, where 0 indicates no precipitation and 1 indicates precipitation; The precipitation inversion support vector machine model is used to output the precipitation amount.

2. The Ku-band microwave scatterometer wind speed and precipitation simultaneous inversion method according to claim 1, wherein The method further includes the training step of the wind speed correction support vector machine model; specifically including: Using the spatio - temporal matching observation data of the C - band and Ku - band scatterometers as input, and using the wind speed accepted by the spatio - temporal matching observation quality control result of the C - band as the output true value, to train the wind speed correction support vector machine model.

3. The Ku-band microwave scatterometer wind speed and precipitation simultaneous inversion method according to claim 1, wherein The method further includes the training step of the precipitation identification support vector machine model; specifically including: Using the spatio - temporal matching observation data of the C - band and Ku - band scatterometers, selecting the wind units that simultaneously meet the quality control result of the Ku - band observation being rejected and the quality control result of the C - band observation being accepted, and using the relevant parameters corresponding to the wind units as input; for the precipitation value of 0 mm / h, marking it as 0, indicating no precipitation, otherwise marking it as 1, indicating precipitation, and using the mark as the output true value, to train the precipitation identification support vector machine model.

4. The Ku-band microwave scatterometer wind speed and precipitation simultaneous inversion method according to claim 1, characterized in that, The method further includes the training step of the precipitation inversion support vector machine model; specifically including: Using the precipitation impact related parameters of the spatio-temporal matching observation data of C-band and Ku-band scatterometers as inputs, and using the area weighted average value of the remote sensing precipitation product within the wind cell as the output true value, train the precipitation inversion support vector machine model. The precipitation impact related parameters include the maximum likelihood estimation residual parameter MLE of the quality control factor, the observed wind speed f, the two-dimensional variational de-fuzzified reanalysis wind speed of the wind direction less than 14 m / s, and the precipitation impact degree factor α.

Images