CN110378074A - Infrared satellite radiance data cloud detection method of quality control based on particle filter - Google Patents

Abstract

The invention relates to a particle filter-based infrared satellite radiance data cloud detection quality control method, the specific steps are as follows: (1) construct a satellite field of view, define the model layer and the cloud coverage ratio of each model layer; (2) fit the whole sky (3) Update and normalize all particles; (4) Compare the simulated brightness temperature and clear sky brightness temperature to determine whether the channel is not affected by clouds, so as to complete the particle filter-based Quality control of infrared satellite radiance data cloud detection. Due to its non-parametric characteristics, the present invention effectively avoids the condition that the random quantity must satisfy the Gaussian distribution when dealing with nonlinear filtering problems, and then can express a wider distribution than the Gaussian model, and also has a certain effect on the nonlinear characteristics of variable parameters. Stronger modeling capabilities. The algorithm of the invention is fast and effective, and the obtained cloud detection identification can provide effective reference information for business identification of weather systems and numerical data assimilation.

Description

Quality Control Method for Cloud Detection of Infrared Satellite Radiance Data Based on Particle Filter

technical field

The invention relates to the technical field of application of satellite meteorological data in earth sciences, in particular to a particle filter-based infrared satellite radiance data cloud detection quality control method.

Background technique

At present, there are three basic cloud detection methods in the world:

1) CO2 slice method: Smith and Frey (1990) calculated cloud top pressure and effective emissivity by CO2 slice method and carried out cloud detection on ATOVS atmospheric infrared detector. The CO2 slice method needs to assume that there are only thin opaque clouds in the atmosphere, and the calculation method is also relatively complicated.

2) The method of using synchronized cloud products: Guan Li (2007) used the MODIS (Moderate-Resolution Imaging Spectro-radiometer) secondary cloud cover product synchronized with AIRS (the Advanced InfraRedSounder) to determine the field of view affected by clouds. Based on the AIRS cloud detection scheme of Goldberg et al. (2003), Chen Jing et al. (2011) combined the characteristics of the GRAPES-3DVAR system and the AIRS instrument to detect clouds in the ocean and land fields of view respectively. The realization of this type of method requires two types of data that require high spatio-temporal matching (Li et al.2005).

3) Minimization method: Huang et al. (2004) proposed the minimum local emissivity variance method, by calculating the local variance of the cloud spectral emissivity of a specific pressure layer in the background field to obtain the optimal estimate of the monolayer cloud emission . Also based on the variational method, Auligné et al. (2013a, 2013b) and Xu et al. (2013, 2015) proposed Multivariate and Minimum Residual (MMR). The MMR cloud detection method builds a cost function, uses a minimization algorithm to fit the observations, and simulates the cloud amount parameters of each model layer. Whether the channel is polluted is determined based on the difference between the emissivity value under clear sky and the simulated all-sky condition. This method requires a convergent and iterative process, and the amount of calculation is relatively large.

In recent years, ideas based on probability density functions such as ensemble Kalman Filter (KF for short) and Particle Filter (PF for short) have been widely used in many meteorological fields such as assimilation and inversion of atmospheric ducts. Using samples to represent the probability density distribution can effectively avoid the conditional constraints that random quantities must meet the Gaussian distribution when dealing with nonlinear filtering problems, and then can express a wider distribution than the Gaussian model, and also have stronger suggestions for the nonlinear characteristics of variable parameters. modeling capabilities. The invention firstly applies the idea of particle filter algorithm in probability theory to cloud detection field of satellite data in meteorology, and provides necessary input information for quality control of infrared satellite data assimilation.

Contents of the invention

The technical problem to be solved by the present invention is to provide a cloud detection quality control method for infrared satellite radiance data based on particle filtering to solve the problem of large amount of calculation and complex calculation of the cloud detection algorithm in the prior art.

In order to solve the above technical problems, the technical solution of the present invention is to provide a particle filter-based infrared satellite radiance data cloud detection quality control method, and its innovative point is that it specifically includes the following steps:

(1) Construct a satellite field of view, construct initial cloud cover profile particles in the field of view, and define each particle as a complete set of model stratus cloud coverage percentage combinations, and use c=c ¹ ,c ² ,. .., c ⁿ represents the effective cloud coverage ratio of each model layer in the satellite field of view, wherein c ⁰ is the clear sky ratio, n is the number of model layers, and in the satellite field of view, the cloud coverage ratio of the kth layer of the model layer is c ^k , and the cloud coverage ratio in the clear sky area is 1-c ^k ; carry out the inversion for the first time, and set the equal probability of clouds to be evenly distributed in each model layer, the formula is:

Among them, the background field of other time times of the test is obtained from the inversion results of the previous time times through model prediction.

(2) Through the effective cloud coverage ratio of each model layer defined in step (1), the whole-sky radiance value is obtained by fitting Expressed as:

Among them, the total number of model layers is n+1, that is, the total number of particles is n+1, including the effective cloud coverage ratio c ¹ ,...,c ⁿ and clear sky percentage c ⁰ of n model layers;

(3) The posterior probability of the particle c ^k on the pattern layer of the kth layer is Then the update equation of the particle c ^k is:

Among them, σ is the observation error, is the observed radiance at wavenumber v, is the radiance rate value simulated by the radiative transfer model CRTM at wavenumber v when a blackbody cloud is placed at the kth layer of the model layer;

The posterior probability of particle c ⁰ is Then the update equation of c ⁰ is:

in, is the radiance value calculated by the radiative transfer model under clear sky conditions; after updating each particle, normalize all particles to ensure that the sum of the cloud coverage ratio and clear sky ratio of each model layer is 1, and for all particles The formula for normalization is:

(4) For any channel, if the difference between the brightness temperature simulated under cloudy conditions and the simulated brightness temperature under clear sky conditions based on the cloud parameter inversion method is less than 1% of the brightness temperature under clear sky conditions, then the channel can be judged as unsafe. The channel affected by the cloud, so as to complete the control of cloud detection quality of infrared satellite radiance data based on particle filter.

Further, each particle defined in the step (1) is a set of complete model layer cloud coverage percentage combinations expressed as: each particle represents that all clouds are in the model layer represented by the particle, and the last particle represents the complete model layer clear sky.

Further, the inversion of the first time is set in the step (1) because there is no cloud inversion result of the previous time in the assimilated background field, that is, cold start.

Further, in the step (2), for any height plane of the k-th layer mode layer in the field of view, the area covered by clouds blocks the upward radiation from the area below the plane with a ratio c ^{k k} , and the remaining 1 -c ^k 's upward radiation works.

Further, in the step (4), it is judged that the channel is not affected by the cloud. The specific criteria for determining the channel are:

Compared with the prior art, the present invention has the beneficial effects of:

Due to its non-parametric characteristics, the particle filter-based infrared satellite radiance data cloud detection quality control method of the present invention effectively avoids the condition that the random quantity must satisfy the Gaussian distribution when dealing with nonlinear filtering problems, and can then express the ratio Gaussian The wider distribution of the model also has a stronger modeling ability for the nonlinear characteristics of variable parameters. The algorithm of the invention is fast and effective, and the obtained cloud detection identification can provide effective reference information for business identification of weather systems and numerical data assimilation.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings that need to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can also be obtained based on these drawings without creative effort.

Fig. 1 is a schematic flow chart of the particle filter-based infrared satellite radiance data cloud detection quality control method of the present invention.

Fig. 2 is a schematic diagram of cloud coverage percentages corresponding to n model layers in the field of view of the present invention.

Detailed ways

The technical solutions of the present invention will be clearly and completely described below through specific embodiments.

The particle filter-based infrared satellite radiance data cloud detection quality control method provided by the present invention is shown in Figures 1 and 2, and its specific steps are as follows:

(1) Construct a satellite field of view, construct initial cloud amount profile particles in the field of view, and define each particle as a complete set of model stratus cloud coverage percentage combinations, that is, each particle represents that all clouds are in the particle The model layer represented by , and the last particle represents a completely clear sky, and c=c ¹ ,c ² ,...,c ⁿ represents the effective cloud coverage ratio of each model layer in the satellite field of view, where c ⁰ is The proportion of clear sky, n is the number of model layers, in the field of view of the satellite, the cloud coverage ratio of the kth layer of the model layer is c ^k , and the cloud coverage ratio of the clear sky area is 1-c ^k ; The cloud inversion result of the second time, that is, the cold start, inverts the first time, and sets the cloud to be evenly distributed in each model layer with equal probability. The formula is:

Among them, the background field of other time times of the test is obtained from the inversion results of the previous time times through model prediction.

(2) Through the effective cloud coverage ratio of each model layer defined in step (1), the whole-sky radiance value is obtained by fitting Expressed as:

Among them, the total number of model layers is n+1, that is, the total number of particles is n+1, including the effective cloud coverage ratio c ¹ ,...,c ⁿ and clear sky percentage c ⁰ of n model layers; for this field of view Any height surface of , the area covered by clouds blocks the upward radiation from the area below the plane with a ratio of c ^k , and the remaining 1-c ^k upward radiation is effective;

(3) The posterior probability of the particle c ^k on the pattern layer of the kth layer is Then the update equation of the particle c ^k is:

Among them, σ is the observation error, is the observed radiance at wavenumber v, is the radiance rate value simulated by the radiative transfer model CRTM at wavenumber v when a blackbody cloud is placed at the kth layer of the model layer;

The posterior probability of particle c ⁰ is Then the update equation of c ⁰ is:

in, is the radiance value calculated by the radiative transfer model under clear sky conditions; after updating each particle, normalize all particles to ensure that the sum of the cloud coverage ratio and clear sky ratio of each model layer is 1, and for all particles The formula for normalization is:

(4) For any channel, if the difference between the brightness temperature simulated under cloudy conditions and the simulated brightness temperature under clear sky conditions based on the cloud parameter inversion method is less than 1% of the brightness temperature under clear sky conditions, then the channel can be judged as unsafe. The channel affected by the cloud, wherein the specific criterion for judging that the channel is not affected by the cloud is: In this way, the quality control of infrared satellite radiance data cloud detection based on particle filter is completed.

The embodiment described above is only a description of the preferred implementation of the present invention, and is not intended to limit the concept and scope of the present invention. Various modifications and improvements made should fall within the scope of protection of the present invention, and the technical content claimed for protection in the present invention has been fully recorded in the technical requirements.

Claims (5)

1. The infrared satellite radiance data cloud detection quality control method based on particle filter, it is characterized in that: specifically comprise the following steps: (1) Construct a satellite field of view, construct initial cloud cover profile particles in the field of view, and define each particle as a complete set of model stratus cloud coverage percentage combinations, and use c=c ¹ ,c ² ,. .., c ⁿ represents the effective cloud coverage ratio of each model layer in the satellite field of view, wherein c ⁰ is the clear sky ratio, n is the number of model layers, and in the satellite field of view, the cloud coverage ratio of the kth layer of the model layer is c ^k , and the cloud coverage ratio in the clear sky area is 1-c ^k ; carry out the inversion for the first time, and set the equal probability of clouds to be evenly distributed in each model layer, the formula is: Among them, the background field of other time times of the test is obtained from the inversion results of the previous time times through model prediction. (2) Through the effective cloud coverage ratio of each model layer defined in step (1), the whole-sky radiance value is obtained by fitting Expressed as: Among them, the total number of model layers is n+1, that is, the total number of particles is n+1, including the effective cloud coverage ratio c ¹ ,...,c ⁿ and clear sky percentage c ⁰ of n model layers; (3) The posterior probability of the particle c ^k on the pattern layer of the kth layer is Then the update equation of the particle c ^k is: Among them, σ is the observation error, is the observed radiance at wavenumber v, is the radiance rate value simulated by the radiative transfer model CRTM at wavenumber v when a blackbody cloud is placed at the kth layer of the model layer; The posterior probability of particle c ⁰ is Then the update equation of c ⁰ is: in, is the radiance value calculated by the radiative transfer model under clear sky conditions; after updating each particle, normalize all particles to ensure that the sum of the cloud coverage ratio and clear sky ratio of each model layer is 1, and for all particles The formula for normalization is: (4) For any channel, if the difference between the brightness temperature simulated under cloudy conditions and the simulated brightness temperature under clear sky conditions based on the cloud parameter inversion method is less than 1% of the brightness temperature under clear sky conditions, then the channel can be judged as unsafe. The channel affected by the cloud, so as to complete the control of cloud detection quality of infrared satellite radiance data based on particle filter. 2. the infrared satellite radiance data cloud detection quality control method based on particle filtering according to claim 1, is characterized in that: each particle defined in the described step (1) is a group of complete model stratus cloud coverage percentages The combination is expressed as: each particle represents that all clouds are in the model layer represented by this particle, and the last particle represents a completely clear sky. 3. the infrared satellite emissivity data cloud detection quality control method based on particle filtering according to claim 1, is characterized in that: the inversion of setting first time time in the described step (1) is because in the background field of assimilation There is no cloud inversion result from the previous time, that is, cold start. 4. the infrared satellite emissivity data cloud detection quality control method based on particle filtering according to claim 1, is characterized in that: in described step (2) for any kth layer model layer in the field of view In the height plane, the area covered by clouds blocks the upward radiation from the area below the plane with a proportion of c ^k , and the remaining 1-c ^k upward radiation is effective. 5. the infrared satellite emissivity data cloud detection quality control method based on particle filtering according to claim 1, is characterized in that: in described step (4), it is judged that described channel is the channel not affected by cloud and the concrete discriminant criterion is :