Gsi-rttov wind profile microwave temperature sounder assimilation data quality control method

Abstract

The application discloses a GSI-RTTOV atmospheric microwave temperature meter assimilation data quality control method. The method constructs a three-stage series quality control link of a read-in stage quality control, a radiation calculation stage quality control and an increment stage quality control: the read-in stage performs screening on observation space-time rationality, scanning geometry, brightness temperature range, cloud amount and channel quality; the radiation calculation stage performs double checking on the profile and the radiation result before and after the RTTOV calling; the increment stage performs coarse detection, scattering / rain detection, terrain quality control and surface adaptive threshold quality control based on the increment of the observed brightness temperature and the simulated brightness temperature; the quality control results of each stage are linked through channel-level quality control marks and observation weights, and the qualified channels retain the weights to enter the assimilation, and the rejected channels have the weights reset to zero. The application can effectively reject the rain and strong scattering contaminated observation while retaining the reliable brightness temperature observation information, realizes the channel-level selective assimilation, and improves the brightness temperature assimilation quality and the observation utilization rate.

Description

Technical Field

[0001] This invention belongs to the field of numerical weather prediction satellite data assimilation technology, and in particular relates to a quality control method for assimilated data from the GSI-RTTOV wind and cloud microwave thermometer. Background Technology

[0002] The accuracy of numerical weather prediction largely depends on the utilization level of various observational information by the data assimilation system. Gridded Statistical Interpolation (GSI) is one of the most widely used assimilation systems in operational numerical weather prediction, achieving optimal fusion of observational and model background fields by minimizing a variational cost function. Satellite microwave observations can penetrate non-precipitating cloud regions to provide atmospheric temperature and humidity information, making them an indispensable source of observations for numerical weather prediction.

[0003] The microwave thermometers MWTS-2 and MWTS-3, carried by the Fengyun-3 satellite, have 13 and 17 detection channels respectively. These are autonomously launched vertical atmospheric temperature sounding payloads, primarily used to assimilate atmospheric temperature and thermosphere structure information. In the GSI data assimilation system, the observation operator calls the RTTOV (Fast Radiative Transfer Model) to calculate channel simulated brightness temperature and Jacobian. The variational cost function uses the inverse variance of observation weights to measure the contribution of each observation. The quality control stage determines whether an observation and its channel can be included in the cost function, making it a crucial step in determining the assimilation quality.

[0004] Microwave thermometer observations are significantly affected by surface emissivity, topography, clouds, and precipitation. Precipitation and ice particle scattering can cause the brightness temperature of high-frequency windows to deviate significantly from clear-sky conditions. The model background field representation error is large in complex topographic regions, and the surface emissivity simulation of non-oceanic surfaces has high uncertainty. Therefore, observations must undergo strict quality control before assimilation.

[0005] Existing microwave thermometer assimilation quality control technologies suffer from the following shortcomings: First, inspections are mostly concentrated in a single stage, and readout inspections lack adaptation to instrument scanning geometry and cloud precipitation scenarios. Abnormal observations are prone to entering radiative transfer calculations, wasting computational resources and introducing the risk of computational failure. Second, radiative transfer calculations lack dual protection for input and output, and abnormal profiles or abnormal forward modeling results may enter assimilation and contaminate the cost function and Jacobian. Third, incremental stage quality control relies heavily on a single coarse detection threshold, lacking physical identification of precipitation and strong scattering contamination, as well as adaptive adaptation to terrain and underlying surface types, resulting in both missed and false rejections. Fourth, quality control results are often processed by discarding or retaining entire observations, lacking channel-level selective assimilation capabilities. Reliable temperature channel information is discarded along with contaminated channels, resulting in low observation utilization. Fifth, quality control conclusions lack unified labeling records and diagnostic statistics, making it difficult to support operational monitoring, effect evaluation, and continuous optimization of quality control thresholds.

[0006] Therefore, there is a need in this field for a data assimilation quality control method that adapts to the observation characteristics of microwave thermometers, covers the entire link of radiation calculation and incremental data, and supports channel-level weight linkage. Summary of the Invention

[0007] In view of this, the present invention aims to propose a quality control method for assimilated data of the GSI-RTTOV Fengyun microwave thermometer, in order to solve the following technical problems: First, how to progressively filter out anomalous observations in order of increasing computational cost, so that anomalous observations are eliminated before entering radiative transfer calculations, thereby reducing computational waste and the risk of computational failure. Secondly, how to implement dual protection for the input and output ends of radiative transfer calculation to ensure the reliability of simulated brightness temperature, sensitivity, and Jacobian after accompanying back mapping; Third, how to identify precipitation, strong scattering, and complex terrain pollution based on physical mechanisms and underlying surface characteristics, so as to reduce false rejections while ensuring the accuracy of rejections; Fourth, how to link quality control conclusions with observation weights at the channel level to achieve selective assimilation at the channel level, thereby improving observation utilization while ensuring data quality; Fifth, how to uniformly mark, record, and diagnose quality control conclusions to support business monitoring, assimilation effect evaluation, and continuous optimization of quality control thresholds.

[0008] To achieve the above objectives, the technical solution of the present invention is implemented as follows: This invention provides a method for quality control of assimilated data from the GSI-RTTOV Fengyun microwave thermometer, comprising: Observation read-in and read-in stage quality control steps: Read in the Fengyun satellite MWTS observation data, perform read-in stage quality control on the observation time, spatial location, scan position number, observation zenith angle, brightness temperature of each channel and cloud cover information, and write the observations that pass the quality control into the GSI observation array; Quality control steps during the radiation calculation phase: Convert the model temperature profile, surface variables, and MWTS channel configuration into the RTTOV input structure. Before calling the RTTOV forward model or K matrix, perform a pre-call check to check whether the profile temperature and specific humidity are non-numerical and whether they are within the physical range. After passing the check, perform forward modeling or K matrix calculation to obtain the simulated brightness temperature and sensitivity. After the calculation, perform a post-call check to check for returned error status and whether the simulated brightness temperature and sensitivity are non-numerical. Incremental phase quality control steps: Calculate the increment after deducting the deviation correction based on the difference between the brightness temperature observed by MWTS and the simulated brightness temperature. Perform incremental phase quality control for the temperature detection channel. Incremental phase quality control includes coarse detection, scattering / precipitation detection, topographic quality control and surface adaptive threshold quality control. The quality control criteria and observation weight inverse variance linkage steps are as follows: Update the channel-level quality control criteria according to the quality control results of each stage; retain the observation weight inverse variance calculated according to the observation error variance for qualified channels to enter GSI assimilation; set the observation weight inverse variance to zero for rejected channels so that they do not participate in assimilation. Among them, the quality control of the read-in stage, the quality control of the radiation calculation stage, and the quality control of the incremental stage are executed in series. The quality control of the radiation calculation stage is executed for observations that have passed the quality control of the read-in stage, and the quality control of the incremental stage is executed for observations that have passed the quality control of the radiation calculation stage.

[0009] Furthermore, quality control during the read-in phase includes: Satellite identifier matching: Only observations corresponding to the configured satellites are retained; Scan position range check: For MWTS-2, the scan position number is limited to 1 to 90; for MWTS-3, the scan position number is limited to 1 to 98; and edge scan positions are eliminated according to the upper and lower limits of the edge scan positions configured by the instrument. Spatiotemporal legality check: The observation time must fall within the assimilation time window, and the observation point must be located within the assimilation domain under the regional assimilation mode; Zenith angle consistency verification: The zenith angle is estimated based on the satellite orbital altitude and scanning geometry. The estimated zenith angle is compared with the observed zenith angle. If the deviation exceeds 1 degree, the observation is discarded. The observed zenith angle is limited to not exceeding the maximum zenith angle allowed by the sensor. Zenith angle sign correction is performed on the first half of the scanning field of view. When the satellite is MWTS-2, the scanning position does not exceed 45 degrees of the field of view; when the satellite is MWTS-3, the scanning position does not exceed 49 degrees of the field of view.

[0010] Furthermore, quality control during the read-in phase also includes: Brightness temperature physical range check: Limit the brightness temperature of each channel to between 50K and 550K. If all channels are out of range or fail, discard the observation. If only a few channels are out of range or fail, mark the channel as invalid, set the channel-level quality control flag of the channel to channel configuration rejection, and set the observation weight inverse variance of the channel to zero so that the channel will no longer participate in subsequent radiation calculations and incremental stage quality control. The remaining channels are processed normally. Cloud precipitation quality control: Read the wind and cloud amount field. If the cloud amount is greater than 0.25, discard the observation. If the cloud amount is missing and the missing value is negative, discard the observation. Sparsity score: Read the channel quality factor and adjust the observation sparsity score.

[0011] Furthermore, before the quality control of the radiation calculation stage, the RTTOV interface initialization is also included: generating coefficient channel indexes based on the GSI effective channel list, reading the MWTS instrument radiation transfer coefficient file, obtaining the reference half-layer pressure and vertical layer number, and initializing the calculation data structure including atmospheric profile structure, channel-profile joint structure, emissivity and reflectivity structure, radiation quantity structure and transmittance structure. The pre-call checks also include: verifying that the temperature is within the physical range of 50K to 400K, and verifying that the half-layer air pressure is reasonable; Post-call checks also include: performing a reasonableness check on the Jacobian following the accompanying backmap; If either the pre-call check or the post-call check fails, the channel-level quality control flag for all channels of the observation will be set to radiation calculation failure, and the inverse variance of the observation weights for all channels will be set to zero.

[0012] Furthermore, the increment equals the MWTS observed brightness temperature minus the simulated brightness temperature minus the channel bias prediction term; The coarse inspection includes: judging the increment channel by channel. When the absolute value of the increment of a certain channel is greater than 3K, the channel-level quality control flag of that channel is set to coarse inspection rejection, and the error factor and observation weight inverse variance of that channel are set to zero.

[0013] Furthermore, the scattering / precipitation detection includes: when the brightness temperatures of both the 1st and 10th channels of the MWTS are valid, the scattering index is calculated as the brightness temperature of the 1st channel of the MWTS minus the brightness temperature of the 10th channel. If the scattering index reaches a preset threshold, all channels are rejected. If the scattering index is invalid, the scattering / precipitation detection is skipped. An invalid scattering index means that the brightness temperature of the 1st or 10th channel of the MWTS is missing or invalid. The 1st channel is a low-frequency channel, and the 10th channel is a high-frequency channel.

[0014] Furthermore, the preset threshold is determined using an offline statistical calibration method: the distribution of scattering indicators under the two types of samples is statistically analyzed using historical clear sky samples and confirmed precipitation samples, and the preset threshold is selected between the two types of distributions according to the business-permitted missed rejection rate and false rejection rate; the preset threshold can be updated and configured according to the instrument and season.

[0015] Furthermore, topographic quality control includes: when the background field elevation of the observation point is greater than 2500 meters, setting the channel-level quality control flag for all channels to topographic rejection and setting the inverse variance of the observation weights for each channel to zero. The adaptive threshold quality control for the surface includes: determining the surface type of the observation point as ocean, land, ice, or snow, selecting the corresponding emissivity sensitivity threshold and surface temperature sensitivity threshold, judging the sensitivity of each channel to surface emissivity and surface temperature. The sensitivity is the Jacobian element of the channel's simulated brightness temperature to surface emissivity and surface skin temperature. Channels with sensitivity exceeding the corresponding threshold are rejected. For non-oceanic scenes, the emissivity sensitivity threshold and surface temperature sensitivity threshold are relaxed to three times the threshold for oceanic scenes.

[0016] Furthermore, the linking step between quality control indicators and the inverse variance of observation weights also includes: The channel-level quality control indicators are recorded by category, which includes qualified, radiation calculation failure, coarse detection rejection, cloud precipitation rejection, terrain rejection, and channel configuration rejection. When only some channels in the same observation field are rejected, only the qualified channels participate in assimilation; The number of rejections of various types is counted and written into the diagnostic statistics array, and channel-level quality control flags are recorded for business monitoring and assimilation effect evaluation.

[0017] Furthermore, the quality control method for GSI-RTTOV Fengyun microwave thermometer assimilation data is executed by a quality control system deployed within the GSI data assimilation system. This quality control system includes: The MWTS observation readout quality control module is used to perform quality control during the observation readout and readout phases. The RTTOV interface quality control module is used to perform the conversion to RTTOV input structure, pre-call checks, forward or K matrix calculations, and post-call checks in the quality control of the radiation calculation stage. The MWTS incremental quality control module is used to perform incremental calculations and incremental phase quality control. The quality control flag management module is used to update channel-level quality control flags and retain or zero out the inverse variance of observation weights. The assimilation diagnostic module is used to count the number of rejections and record channel-level quality control indicators.

[0018] Compared with existing technologies, the GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method of the present invention has the following beneficial effects: First, by constructing a three-stage sequential quality control link of reading, radiation calculation, and incremental analysis, and arranging each stage in order of increasing computational cost and information dependence, the reading stage eliminates most abnormal observations with almost zero computational cost, the radiation calculation stage ensures the reliability of forward modeling input and output, and the incremental stage completes fine judgment based on physical mechanisms to achieve progressive filtering, thereby improving the accuracy of quality control while saving computational resources.

[0019] Secondly, by adapting the reading-in stage to the characteristics of microwave thermometers, and by using scanning position range limitations, first half-scan zenith angle sign correction, brightness temperature physical range limits, and cloud cover thresholds, the observations of scanning geometric anomalies, calibration anomalies, and cloud precipitation pollution can be excluded from the assimilation process as early as possible.

[0020] Third, by setting up double checks before and after the RTTOV call, the input end of the radiative transfer model and the output end of the assimilation system are protected respectively, and the Jacobian rationality check after the accompanying back mapping is covered, ensuring the reliability of the variational assimilation gradient information and improving the stability of business operation.

[0021] Fourth, through multiple incremental stage rules such as coarse detection, scattering index, terrain elevation and surface adaptive threshold, the scattering index constructed by the brightness temperature difference between low and high frequency channels can physically identify precipitation and strong scattering pollution. The surface adaptive threshold reduces false rejections by relaxing the threshold for surface classification and non-oceanic scenes. The terrain quality control excludes observations with excessive background error. The judgment granularity matches the spatial scale of pollution, taking into account both the accuracy of rejection and the utilization rate of observations.

[0022] Fifth, through the linkage mechanism between quality control indicators and the inverse variance of observation weights, qualified channels are weighted according to the variance of observation errors, while the weight of rejected channels is reset to zero. When a portion of the same field of view is rejected, only qualified channels participate in assimilation, thereby achieving selective assimilation at the channel level. This effectively improves the utilization rate of observations by making full use of reliable temperature detection information while eliminating abnormal observations.

[0023] Sixth, by statistically analyzing the number of rejections of various types and recording channel-level quality control indicators through the diagnostic output process, the quality control process becomes auditable and traceable, supporting business monitoring, assimilation effect evaluation, and continuous optimization of quality control thresholds.

[0024] In summary, this invention establishes a quality control system adapted to the characteristics of microwave thermometers, which can effectively reject contaminated observations in precipitation scenarios while retaining reliable temperature channel assimilation information and improving the assimilation quality of temperature channels. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the overall process of multi-stage quality control for the assimilation of the Fengyun microwave thermometer provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the quality control process for the microwave thermometer observation and reading stage provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the quality control process for the GSI-RTTOV radiation calculation stage provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the incremental quality control process for microwave thermometers provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the linkage mechanism between quality control indicators and observation weights provided in an embodiment of the present invention. Detailed Implementation

[0026] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0027] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0028] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0030] To facilitate understanding, the terminology and application environment involved in the various embodiments of the present invention will be explained below.

[0031] GSI refers to the Grid Statistical Interpolation Data Assimilation System, one of the operational data assimilation systems in numerical weather prediction. It achieves the assimilation of various meteorological observations by minimizing a variational cost function. RTTOV refers to the Fast Radiative Transfer Model, used to calculate the simulated brightness temperature, spherical transmittance, and related sensitivities of satellite observation channels from atmospheric profiles, surface conditions, and observation geometry. The K-matrix refers to the Jacobian matrix used in radiative transfer calculations, reflecting the relationship between the simulated brightness temperature of the channel and input quantities such as the atmospheric profile. The Jacobian array refers to the data structure in the assimilation system that stores the sensitivity of observations to state variables, provided by the observation operator through adjoint back mapping. MWTS refers to the Microwave Thermometer, with MWTS-2 and MWTS-3 mounted on the Fengyun-3 satellite, possessing 13 and 17 detection channels respectively, primarily used for vertical atmospheric temperature sounding and thermospheric structure sounding. BUFR refers to the Binary Universal Representation Format for meteorological data; 1B radiation data files are stored in this format. The inverse variance of observation weights refers to the weight of an observation in the variational cost function, usually taken as the reciprocal of the observation error variance. When the inverse variance of the weights is set to zero, the corresponding observation no longer contributes to the cost function, which is equivalent to rejecting the observation. The quality control flag refers to the status identifier maintained for each channel, used to record the quality control conclusions of that channel at each stage. The error factor refers to the coefficient characterizing the configuration of observation errors during assimilation. When it is set to zero along with the inverse variance of the weights, the corresponding channel no longer participates in the cost function and Jacobian calculations. Channel-level refers to the processing granularity on a single detection channel, while field-of-view level refers to the processing granularity on all channels of a single observation. Field of view refers to all channel observations acquired by a microwave thermometer in a single scan at a single observation point; full field-of-view rejection means rejecting all channel observations within that field of view.

[0032] Example 1: Overall Flow of Quality Control Method This embodiment provides a method for quality control of assimilated data from the GSI-RTTOV Fengyun microwave thermometer. For example... Figure 1 As shown, this method is designed for microwave thermometer observations on the Fengyun-3 satellite. After completing forward modeling of temperature-related channels in the GSI data assimilation system via the GSI-RTTOV radiation transfer interface, a three-stage cascaded quality control link is constructed: quality control during the read-in stage, quality control during the radiation calculation stage, and quality control during the increment stage. The quality control results of each stage are linked to the observation weights through channel-level quality control flags to achieve channel-level or full-field-of-view rejection. This method includes the following steps: Step S1: Observation Reading and Quality Control During Reading. The GSI observation reading module calls the data reading program according to the observation type, reads MWTS observations from the BUFR format 1B radiation data file, performs quality control on the observation time, spatial location, scan position number, observation zenith angle, brightness temperature of each channel, and cloud cover information during the reading stage, and writes the observations that pass the quality control into the GSI observation array.

[0033] Step S2, RTTOV interface initialization. The radiation observation operator completes RTTOV initialization before the start of the observation cycle, generates coefficient channel indexes, reads the MWTS instrument radiative transfer coefficient file, obtains the reference half-layer pressure and vertical layer number, and initializes each calculation data structure.

[0034] Step S3, quality control before calling the profile conversion and radiation calculation stage. Convert the model background field temperature, water vapor absorber, surface type and surface skin temperature, observation zenith angle and azimuth angle into the atmospheric profile and channel-profile joint description required by RTTOV, and check whether the profile temperature and specific humidity are non-numerical and within the physical range before calling the RTTOV forward modeling or K matrix module.

[0035] Step S4 involves quality control after the RTTOV forward modeling or K-matrix calculation and radiation calculation stages are called. After checking, the RTTOV forward modeling module or K-matrix module is called to complete the calculation of simulated brightness temperature, layer transmittance, and profile sensitivity for each channel. After the call, the returned error status, simulated brightness temperature, profile sensitivity, and Jacobian after backmapping are checked.

[0036] Step S5, Incremental Calculation. For each channel that has passed the radiation calculation quality control, calculate the increment after deducting the antenna deviation correction. The increment is equal to the observed brightness temperature minus the simulated brightness temperature minus the channel deviation prediction term.

[0037] Step S6, Incremental quality control. A dedicated incremental quality control program is invoked for each microwave thermometer observation type, sequentially executing coarse detection, scattering / precipitation detection, topographic quality control, and surface adaptive threshold quality control.

[0038] Step S7: Linking and assimilating quality control flags with observation weights. Each channel maintains an independent quality control flag. The three-stage quality control is executed sequentially, and the corresponding flag is updated after rejection at any stage. For qualified channels, the inverse variance of the observation weights calculated based on the observation error variance is retained and entered into the GSI cost function and Jacobian calculation. For rejected channels, the inverse variance of the observation weights is set to zero, and the number of rejections of each type is counted and written into the diagnostic statistics array.

[0039] In the above process, the three stages are arranged in ascending order of computational cost and information dependence: the input stage uses only observation metadata and the observations themselves, eliminating most obviously anomalous observations with almost zero computational cost; the radiation calculation stage relies on radiative transfer calculations to ensure the reliability of forward modeling input and output; the incremental stage relies on observations and simulated brightness temperature to complete a fine-grained judgment based on physical mechanisms. The quality control results of each stage are not directly used to delete observation data, but are uniformly linked with the observation weights at the end of the process: qualified channels enter the GSI assimilation main process after assimilation weight adjustment; rejected channels enter the quality control rejection branch, their observation weights are reset to zero, and they no longer participate in assimilation analysis; when only some channels in the same observation field are rejected, the remaining qualified channels can still enter the assimilation process. This ensures that MWTS temperature probe information is fully utilized as much as possible while guaranteeing data quality.

[0040] Example 2: Specific Implementation of Quality Control in the Read-in Stage like Figure 2 As shown, quality control during the read-in phase is completed before the observation data enters the assimilation operator. The system first parses the MWTS observation file, extracting fields such as satellite identifier, observation time, scan position, and brightness temperature of each channel, and then performs various checks in groups. Figure 2 The various checks are categorized into four groups: basic legality, observation angle, brightness temperature check, and cloud precipitation quality control. All four groups of checks must be passed before the observation can be written into the GSI observation array. In this embodiment, step S1 specifically includes the following sub-steps.

[0041] Step S1.1, Satellite Identifier Matching. Compare the satellite identifiers in the observation file with the configured satellite identifiers, retain only the observations corresponding to the configured satellites, and discard the observations with inconsistent sources.

[0042] Step S1.2, scan position range check. For MWTS-2, the scan position is limited to between 1 and 90, and for MWTS-3, the scan position is limited to between 1 and 98. Edge scan positions are eliminated according to the upper and lower limits of the edge scan positions configured by the instrument to avoid edge field observations with large scanning geometric errors from entering the assimilation.

[0043] Step S1.3, Spatiotemporal validity check. The observation time must fall within the assimilation time window, and the observation point must be located within the assimilation domain under the regional assimilation mode. Observations that do not meet the time or space conditions are discarded.

[0044] Step S1.4, Zenith Angle Consistency Verification. The zenith angle is estimated based on the satellite orbital altitude and scanning geometry. This estimate is compared with the observed zenith angle; if the deviation exceeds 1 degree, the observation is discarded, and the zenith angle is limited to the maximum allowable zenith angle of the sensor. Zenith angle sign correction is performed on the first half of the scanning field of view. Specifically, sign correction is performed on the field of view for MWTS-2 scanning positions not exceeding 45° and MWTS-3 scanning positions not exceeding 49° to ensure the continuity and consistency of the scanning angle sequence.

[0045] Step S1.5, Brightness Temperature Physical Range Check. Physical range limits are applied to the brightness temperature of each channel, requiring it to be between 50K and 550K. If all channels exceed the range or fail, the observation is discarded. If only a few channels exceed the range or fail, the channel is marked as invalid, its channel-level quality control flag is set to channel configuration rejection, and its observation weight inverse variance is set to zero. This channel will no longer participate in subsequent radiation calculations and incremental stage quality control; the remaining channels are processed normally.

[0046] Step S1.6, Cloud and Precipitation Quality Control. Read the wind and cloud cover field. If the cloud cover is greater than 0.25, discard the observation. If the cloud cover is missing and the missing value is negative, also discard the observation. This is to exclude observations contaminated by clouds and precipitation during the reading stage.

[0047] Step S1.7: Channel quality factor participates in sparsity scoring. The channel quality factor is read and the observation sparsity score is adjusted. The channel quality factor is taken from the quality flag field attached to the observation data and is used to characterize the reliability of the channel data quality. This ensures that observations with better channel quality have a higher retention priority during sparsity, balancing observation density and data quality.

[0048] Observations that pass all the above checks, along with brightness temperature, scanning geometry, and surface attribute information for 13 or 17 channels, are written into the GSI observation array for use by the radiation observation operator. Observations rejected entirely during the read-in phase are flagged with appropriate quality control indicators based on the reason for rejection, such as channel configuration rejection or cloud / precipitation rejection, and are included in the diagnostic statistics; these observations are not written into the GSI observation array.

[0049] As can be seen from this embodiment, the quality control during the read-in stage has the following characteristics: First, the quality control checkpoint is moved forward, and all checks are completed before the observation enters the assimilation operator, avoiding the waste of radiative transfer calculation and assimilation resources by bad data; Second, the rejection granularity is matched with the nature of the contamination, and the brightness temperature check only discards the entire observation when all channels fail, and the failure of individual channels is marked at the channel level and handled by weight linkage, while the cloud precipitation quality control directly discards the entire observation when it is identified as scene-based contamination; Third, the check items are adapted to the characteristics of the MWTS instrument, and the scanning position range limit and the first half-scan zenith angle sign correction are both set for the cross-track scanning geometry of the microwave thermometer.

[0050] Example 3: Specific Implementation of RTTOV Interface Initialization In this embodiment, step S2 specifically includes: the radiation observation operator completes RTTOV initialization before the start of the observation cycle; when the RTTOV radiation transfer engine is enabled, a coefficient channel index is generated based on the GSI effective channel list, the MWTS instrument radiation transfer coefficient file is read, and the reference half-layer pressure and vertical layer number are obtained; simultaneously, the atmospheric profile structure, channel-profile joint structure, emissivity and reflectivity structure, radiation quantity structure, and transmittance structure, and other data structures required for calculation are initialized; and the microwave emissivity spectrum can be used to provide initial surface emissivity values ​​for each channel, providing a reasonable initial surface state for radiation transfer calculation. The above initialization is completed once before the start of the observation cycle, providing unified coefficient support and data structure preparation for subsequent line-by-line observation calculations, avoiding the overhead of repeated initialization.

[0051] Example 4: Specific Implementation of Profile Conversion and Pre-call Quality Control In this embodiment, step S3 specifically includes the following: In the profile conversion section, for each MWTS observation that has passed the read-in quality control, the GSI-RTTOV interface converts the model background field temperature, water vapor absorber, surface type and surface skin temperature, observation zenith angle and azimuth angle into the atmospheric profile and channel-profile joint description required by RTTOV. The conversion process includes: reconstructing the vertical pressure layer based on the reference half-layer pressure; interpolating the model layer temperature along the logarithmic pressure to the RTTOV layer; interpolating the specific humidity after inverse transformation to the water vapor profile; and setting the surface type, elevation and observation geometry.

[0052] In the pre-call check section, such as Figure 3 As shown, the following checks are performed before calling the RTTOV forward modeling or K-matrix module: Check for non-numerical values ​​in the profile temperature and specific humidity; verify that the temperature is within the physical range of 50K to 400K; verify that the half-layer pressure is reasonable. If any check is abnormal, diagnostic information is recorded, the channel-level quality control flag for all channels of that observation is set to radiation calculation failure, the inverse variance of the observation weights for all channels is set to zero, and forward modeling is no longer performed on that observation. These pre-call checks are set at the input of the radiative transfer calculation to prevent abnormal profiles from entering the radiative transfer model and causing calculation failure or error propagation.

[0053] It should be noted that the 50K to 550K range in the brightness temperature physical range check during the read-in stage applies to the channel observed brightness temperature, while the 50K to 400K range in the pre-call check during the radiation calculation stage applies to the profile temperature after model interpolation. The two have different objects of application and are not contradictory.

[0054] Example 5: Specific Implementation of Forward Modeling Calculation and Post-Invocation Quality Control The RTTOV forward modeling module is called when GSI is running in observation operator mode or performing observation simulation. The K matrix module is called when GSI is running in variational assimilation mode and requires Jacobian. The checking logic for pre-call and post-call checks is applicable to both types of calls.

[0055] like Figure 3 As shown, after pre-call checks, the RTTOV forward modeling module or K-matrix module is called to complete the calculation of simulated brightness temperature, layer transmittance, and profile sensitivity for each channel. The calculated simulated brightness temperature is backfilled into the GSI simulated brightness temperature array, the transmittance is backfilled into the layer transmittance array, and the sensitivity is written into the Jacobian array via adjoint backmapping for variational assimilation.

[0056] After the call, post-call checks are performed, including: determining whether the returned error status was successful; checking for non-numerical values ​​in the simulated brightness temperature and profile sensitivity; and performing a reasonableness check on the Jacobian after backmapping, eliminating obviously unreasonable sensitivity values. If any check fails, the observation is marked as a radiometric calculation failure, the channel-level quality control flag for all channels is set to radiometric calculation failure, the inverse variance of the observation weights for all channels is set to zero, diagnostic information is retained but not used for assimilation; if all checks pass, the simulated brightness temperature and transmittance are output, and incremental quality control begins.

[0057] As can be seen from Examples 4 and 5, the quality control in the radiation calculation stage sets up checks at both the input and output ends of the radiation transfer calculation, forming a double protection: checking the protected model input before calling and checking the protected assimilation output after calling, to prevent bad simulated brightness temperature and sensitivity from entering the cost function and Jacobian; and since profile anomalies affect all channels, the rejection granularity in this stage is the entire field of view, that is, the weights of all channels in the entire observation are reset to zero, so that the failure nature matches the rejection granularity.

[0058] Example 6: Specific Implementation of Incremental Calculation and Incremental Stage Quality Control like Figure 4 As shown, incremental phase quality control is performed after obtaining the observed brightness temperature and the RTTOV simulated brightness temperature, including four types of detection: incremental calculation and coarse detection, scattering / precipitation detection, terrain quality control and surface adaptive threshold quality control.

[0059] In step S5, for each channel that has undergone quality control through radiation calculations, the increment after deducting antenna bias correction is calculated. The increment is equal to the observed brightness temperature minus the simulated brightness temperature minus the channel bias prediction term. The increment reflects the degree of deviation between the observation and the model on that channel and serves as a unified metric for quality control in the increment stage. The antenna bias correction is a preset antenna pattern error correction term based on the instrument model; the channel bias prediction term is provided by the bias correction scheme, and in this embodiment, it is obtained by multiplying the correction coefficients output by variational bias correction or static bias correction with the corresponding prediction factor.

[0060] Step S6 specifically includes the following sub-steps.

[0061] Step S6.1, coarse detection. Incremental values ​​are evaluated channel by channel. When the absolute value of an increment in a channel exceeds 3K, the channel is determined to have significantly deviated from the model background. The channel-level quality control flag for that channel is set to coarse detection rejection, and its error factor and the inverse variance of the observation weights are set to zero, thus achieving channel-level rejection.

[0062] Step S6.2, Scattering / Precipitation Detection. When the brightness temperatures of both Channel 1 and Channel 10 are valid, the scattering index is calculated as the brightness temperature of Channel 1 minus the brightness temperature of Channel 10, i.e., the brightness temperature of the low-frequency channel minus the brightness temperature of the high-frequency channel. Precipitation and ice particle scattering significantly reduce the brightness temperature of the high-frequency window channel, while the low-frequency channel is relatively less affected. Therefore, the scattering index can effectively identify observations contaminated by precipitation and strong scattering. If the scattering index reaches the preset threshold, the observation is determined to be contaminated by precipitation or scattering, and all channels are rejected. When the scattering index is invalid, i.e., the brightness temperature of Channel 1 or Channel 10 is missing or invalid, the detection is skipped to avoid false rejections due to missing data. The preset threshold can be determined by offline statistical calibration, i.e., by statistically analyzing the distribution of the scattering index under the two types of samples using historical clear-sky samples and confirmed precipitation samples, and selecting a threshold between the two distributions according to the operationally permissible missed rejection rate and false rejection rate. The preset threshold can be updated and configured according to the instrument and season, where Channel 1 is the low-frequency channel and Channel 10 is the high-frequency channel.

[0063] Step S6.3, Terrain Quality Control. When the background field elevation of the observation point is greater than 2500 meters, it is determined that the observation is located in a complex terrain area, and its background field representation error and surface scattering simulation error are both large. The channel-level quality control flag of all channels is set to terrain rejection, and the inverse variance of the observation weight of each channel is set to zero.

[0064] Step S6.4, Adaptive Threshold Quality Control for Surface. Based on the surface type of the observation point—ocean, land, ice, or snow—corresponding emissivity sensitivity thresholds and surface temperature sensitivity thresholds are selected. The sensitivity of each channel to surface emissivity and surface temperature is assessed, and channels with sensitivities exceeding the corresponding thresholds are rejected. For non-oceanic scenarios, these thresholds are relaxed to three times the thresholds for oceanic scenarios to accommodate the uncertainty of complex underlying surface emissivity models and reduce false rejections. In this embodiment, the sensitivity of each channel to surface emissivity and surface temperature is taken from the RTTOV calculation results, specifically the Jacobian element of the channel's simulated brightness temperature on surface emissivity and surface skin temperature; in alternative embodiments, it can also be obtained from an offline configured sensitivity table. This step, through surface classification and threshold relaxation, adapts the quality control criteria to underlying surface characteristics, balancing rejection accuracy and observation utilization.

[0065] In this embodiment, the four types of detection are executed in the following order: coarse detection, scattering / precipitation detection, terrain quality control, and surface adaptive threshold quality control. Each detection independently updates the channel-level quality control flag, and the rejection of any detection does not affect the execution of subsequent detections or the accumulation of flags. In alternative embodiments, the execution order of each detection can be adaptively adjusted according to operational needs. Among them, coarse detection is a channel-level judgment, scattering / precipitation detection and terrain quality control are field-of-view judgments, and surface adaptive threshold quality control provides adaptive judgment criteria according to the surface type, with the judgment granularity matching the spatial scale of the pollution source.

[0066] Example 7: Specific Implementation of Linking Quality Control Markers with Observation Weights like Figure 5 As shown, this embodiment uses a linkage mechanism between quality control indicators and observation weights to uniformly map the quality control results at each stage into assimilated and usable channel-level weights. Step S7 specifically includes the following:

[0067] In the quality control mark maintenance section, each channel maintains an independent quality control mark, initialized to a qualified state. The three stages of quality control—reading, radiation calculation, and incremental control—are executed sequentially, and the corresponding mark is updated after any stage is rejected. Quality control marks are recorded by category, including qualified, radiation calculation failed, coarse detection rejected, cloud precipitation rejected, terrain rejected, and channel configuration rejected, making quality control conclusions recordable, statistically significant, and traceable.

[0068] In the weighting adjustment rules: for qualified channels, the inverse variance of the observation weight is calculated based on the observation error variance and incorporated into the GSI cost function and Jacobian calculation; for rejected channels, the inverse variance of the observation weight is set to zero, so that it does not contribute to the cost function; when some channels within the same observation field of view are rejected, only qualified channels participate in assimilation. Through the above rules, the quality control criteria are translated into the weighting language of variational assimilation, achieving soft rejection and channel-level selective assimilation.

[0069] In the diagnostic output section, the number of rejections of various types is counted and written into the diagnostic statistics array. At the same time, channel-level quality control flags are recorded for business monitoring and assimilation effect evaluation, and to provide statistical basis for the continuous optimization of subsequent quality control thresholds.

[0070] To illustrate the effectiveness of the linkage mechanism, three typical scenarios are given below. First, a channel passes all stages of quality control, its flag remains qualified, its observation weight is retained, and it enters the Jacobian calculation. Second, a channel is rejected by coarse detection during the incremental stage because the absolute value of the increment is greater than 3K; its observation weight is reset to zero, and it does not participate in assimilation, while other qualified channels within the same field of view are still assimilated normally. Third, an observation fails due to radiometric calculation failure; all channels are flagged as radiometric calculation failed, the entire field of view is rejected, and a diagnostic mark is added. These three scenarios collectively demonstrate that the quality control flag and observation weight are linked, ensuring that assimilation uses only reliable observations.

[0071] Example 8: Quality Control System, Electronic Equipment, and Storage Media This embodiment provides a quality control system for implementing the methods described in any one of Embodiments 1 to 7, including an MWTS observation read-in quality control module, an RTTOV interface quality control module, an MWTS incremental quality control module, a quality control flag management module, and an assimilation diagnosis module. Specifically, the MWTS observation read-in quality control module performs the read-in stage quality control of Embodiment 2, completing observation parsing, various legality checks, and writing of qualified observations; the RTTOV interface quality control module performs initialization, profile conversion, and pre- and post-call checks of Embodiments 3 to 5; the MWTS incremental quality control module performs incremental calculation and various tests of Embodiment 6; the quality control flag management module maintains channel-level quality control flags and performs linkage adjustment between flags and the inverse variance of observation weights; the assimilation diagnosis module counts the number of various rejections and records channel-level quality control flags, outputting diagnostic information. Each module can be implemented by hardware, software, or a combination thereof; when implemented in software, it can be stored as a program module in a computer-readable storage medium; each module can also be integrated or split into multiple units according to implementation needs.

[0072] This embodiment also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the quality control method described in any one of embodiments one to seven.

[0073] This embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the quality control method described in any one of embodiments one to seven. The storage medium can be any type of medium capable of storing program code, such as a read-only memory, random access memory, magnetic disk, or optical disk.

[0074] Example 9: Alternative Example In alternative embodiments of the present invention, the thresholds in the above embodiments are preferred operational configurations and can be adaptively adjusted according to the instrument model, assimilation system version, and operational requirements. Specifically, the coarse detection threshold, cloud cover threshold, terrain elevation threshold, and relaxation factor for non-oceanic scenes can all be configured. The execution order of each check within the read-in phase quality control and the execution order of each detection within the incremental phase quality control can be adaptively adjusted without affecting the update results of the channel-level quality control flags. Read-in phase quality control, radiation calculation phase quality control, and incremental phase quality control can be deployed in the same process of the GSI data assimilation system or deployed in different processes. The quality control methods of each embodiment can also be extended to the quality control of other trans-track scanning microwave thermometer observations. The technical features of the above embodiments can be combined with each other without conflict, and the resulting embodiments also fall within the scope of the present invention and are protected by the present invention. The above alternative and combined implementations do not depart from the concept of the present invention and are within the scope of protection of the present invention.

[0075] Beneficial effects In combination with the above embodiments, the present invention has the following beneficial effects.

[0076] Firstly, through the three-stage sequential quality control link of reading, radiation calculation, and incremental as described in Examples 1 to 7, each stage is arranged in order of increasing computational cost and information dependence. The reading stage eliminates most abnormal observations with almost zero computational cost, the radiation calculation stage ensures the reliability of forward modeling input and output, and the incremental stage completes fine judgment based on physical mechanisms to achieve progressive filtering, thereby improving the accuracy of quality control while saving computational resources.

[0077] Secondly, by adapting the read-in stage of the MWTS instrument to the characteristics of Example 2, and by using the scanning position range limitation, the first half-scan zenith angle sign correction, the brightness temperature physical range limit and cloud cover threshold, the observation of scanning geometric anomalies, calibration anomalies and cloud precipitation pollution can be eliminated as early as possible.

[0078] Third, by implementing the dual checks set before and after the RTTOV call in Examples 4 and 5, the model input end and the assimilation output end are protected respectively, and the Jacobian rationality check after the accompanying back mapping is covered, ensuring the reliability of the variational assimilation gradient information and improving the stability of business operation.

[0079] Fourth, through various rules such as coarse detection, scattering index, terrain elevation and surface adaptive threshold in Example 6, the scattering index constructed by the brightness temperature difference between low and high frequency channels can physically identify precipitation and strong scattering pollution. The surface adaptive threshold reduces false rejections by relaxing the threshold for non-oceanic scenes through surface classification. The terrain quality control excludes observations with excessive background error, and the judgment granularity matches the spatial scale of pollution.

[0080] Fifth, through the linkage mechanism of quality control mark and observation weight inverse variance in Example 7, qualified channels are weighted according to the observation error variance, and the weight of rejected channels is reset to zero. When some channels are rejected, only qualified channels participate in assimilation, thereby achieving selective assimilation at the channel level. While eliminating abnormal observations, reliable temperature detection information is retained to the maximum extent, effectively improving the observation utilization rate.

[0081] Sixth, by using the diagnostic output link in Example 7 to count the number of rejections of various types and record channel-level quality control indicators, the quality control process becomes auditable and traceable, supporting business monitoring, assimilation effect evaluation and continuous optimization of quality control thresholds.

[0082] In summary, this invention forms a quality control system adapted to the characteristics of MWTS, which can effectively reject contaminated observations in precipitation scenarios, while retaining reliable temperature channel assimilation information and improving the quality of temperature channel assimilation.

[0083] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A quality control method for assimilated data from the GSI-RTTOV Fengyun microwave thermometer, characterized in that, include: Observation read-in and read-in stage quality control steps: Read in the Fengyun satellite MWTS observation data, perform read-in stage quality control on the observation time, spatial location, scan position number, observation zenith angle, brightness temperature of each channel and cloud cover information, and write the observations that pass the quality control into the GSI observation array; Quality control steps during the radiation calculation phase: Convert the model temperature profile, surface variables, and MWTS channel configuration into the RTTOV input structure. Before calling the RTTOV forward model or K matrix, perform a pre-call check to check whether the profile temperature and specific humidity are non-numerical and whether they are within the physical range. After passing the check, perform forward modeling or K matrix calculation to obtain the simulated brightness temperature and sensitivity. After the calculation, perform a post-call check to check the returned error status and whether the simulated brightness temperature and sensitivity are non-numerical. Incremental phase quality control steps: Calculate the increment after deducting the deviation correction based on the difference between the brightness temperature observed by MWTS and the simulated brightness temperature. Perform incremental phase quality control for the temperature detection channel. Incremental phase quality control includes coarse detection, scattering / precipitation detection, topographic quality control and surface adaptive threshold quality control. The quality control criteria and observation weight inverse variance linkage steps are as follows: Update the channel-level quality control criteria according to the quality control results of each stage; retain the observation weight inverse variance calculated according to the observation error variance for qualified channels to enter GSI assimilation; set the observation weight inverse variance to zero for rejected channels so that they do not participate in assimilation. Among them, the quality control of the read-in stage, the quality control of the radiation calculation stage, and the quality control of the incremental stage are executed in series. The quality control of the radiation calculation stage is executed for observations that have passed the quality control of the read-in stage, and the quality control of the incremental stage is executed for observations that have passed the quality control of the radiation calculation stage.

2. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, Quality control during the read-in phase includes: Satellite identifier matching: Only observations corresponding to the configured satellites are retained; Scan position range check: For MWTS-2, the scan position number is limited to 1 to 90; for MWTS-3, the scan position number is limited to 1 to 98; and edge scan positions are eliminated according to the upper and lower limits of the edge scan positions configured by the instrument. Spatiotemporal legality check: The observation time must fall within the assimilation time window, and the observation point must be located within the assimilation domain under the regional assimilation mode; Zenith angle consistency verification: The zenith angle is estimated based on the satellite orbital altitude and scanning geometry. The estimated zenith angle is compared with the observed zenith angle. If the deviation exceeds 1 degree, the observation is discarded. The observed zenith angle is limited to not exceeding the maximum zenith angle allowed by the sensor. Zenith angle sign correction is performed on the first half of the scanning field of view. When the satellite is MWTS-2, the scanning position does not exceed 45 degrees of the field of view; when the satellite is MWTS-3, the scanning position does not exceed 49 degrees of the field of view.

3. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, Quality control during the read-in phase also includes: Brightness temperature physical range check: Limit the brightness temperature of each channel to between 50K and 550K. If all channels are out of range or fail, discard the observation. If only a few channels are out of range or fail, mark the channel as invalid, set the channel-level quality control flag of the channel to channel configuration rejection, and set the observation weight inverse variance of the channel to zero so that the channel will no longer participate in subsequent radiation calculations and incremental stage quality control. The remaining channels are processed normally. Cloud precipitation quality control: Read the wind and cloud amount field. If the cloud amount is greater than 0.25, discard the observation. If the cloud amount is missing and the missing value is negative, discard the observation. Sparsity score: Read the channel quality factor and adjust the observation sparsity score.

4. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, Before the quality control of the radiation calculation stage, the RTTOV interface initialization is also included: generating coefficient channel indexes based on the GSI effective channel list, reading the MWTS instrument radiation transfer coefficient file, obtaining the reference half-layer pressure and vertical layer number, and initializing the calculation data structure including atmospheric profile structure, channel-profile joint structure, emissivity and reflectivity structure, radiation quantity structure and transmittance structure. The pre-call checks also include: verifying that the temperature is within the physical range of 50K to 400K, and verifying that the half-layer air pressure is reasonable; Post-call checks also include: performing a reasonableness check on the Jacobian following the backmap; If either the pre-call check or the post-call check fails, the channel-level quality control flag for all channels of the observation will be set to radiation calculation failure, and the inverse variance of the observation weights for all channels will be set to zero.

5. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, The increment equals the MWTS observed brightness temperature minus the simulated brightness temperature, and then minus the channel bias prediction term; The coarse inspection includes: judging the increment channel by channel. When the absolute value of the increment of a certain channel is greater than 3K, the channel-level quality control flag of that channel is set to coarse inspection rejection, and the error factor and observation weight inverse variance of that channel are set to zero.

6. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, Scattering / precipitation detection includes: when the brightness temperatures of both channel 1 and channel 10 of the MWTS are valid, the scattering index is calculated as the brightness temperature of channel 1 minus the brightness temperature of channel 10 of the MWTS. If the scattering index reaches a preset threshold, all channels are rejected. If the scattering index is invalid, the scattering / precipitation detection is skipped. An invalid scattering index means that the brightness temperature of channel 1 or channel 10 of the MWTS is missing or invalid. Among them, channel 1 is a low-frequency channel and channel 10 is a high-frequency channel.

7. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 6, characterized in that, The preset threshold is determined by offline statistical calibration: the distribution of scattering index under the two types of samples is statistically analyzed using historical clear sky samples and confirmed precipitation samples, and the preset threshold is selected between the two types of distributions according to the business-allowed missed rejection rate and false rejection rate. The preset thresholds can be updated according to the instrument and the season.

8. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, Topographic quality control includes: when the background field elevation of the observation point is greater than 2500 meters, setting the channel-level quality control flag of all channels to topographic rejection and setting the inverse variance of the observation weight of each channel to zero; The adaptive threshold quality control for the surface includes: determining the surface type of the observation point as ocean, land, ice, or snow, selecting the corresponding emissivity sensitivity threshold and surface temperature sensitivity threshold, judging the sensitivity of each channel to surface emissivity and surface temperature. The sensitivity is the Jacobian element of the channel's simulated brightness temperature to surface emissivity and surface skin temperature. Channels with sensitivity exceeding the corresponding threshold are rejected. For non-oceanic scenes, the emissivity sensitivity threshold and surface temperature sensitivity threshold are relaxed to three times the threshold for oceanic scenes.

9. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to claim 1, characterized in that, The steps for linking quality control indicators with the inverse variance of observation weights also include: The channel-level quality control indicators are recorded by category, which includes qualified, radiation calculation failure, coarse detection rejection, cloud precipitation rejection, terrain rejection, and channel configuration rejection. When only some channels in the same observation field are rejected, only the qualified channels participate in assimilation; The number of rejections of various types is counted and written into the diagnostic statistics array, and channel-level quality control flags are recorded for business monitoring and assimilation effect evaluation.

10. The GSI-RTTOV Fengyun microwave thermometer assimilation data quality control method according to any one of claims 1 to 9, characterized in that, The method is executed by a quality control system deployed within the GSI data assimilation system. The quality control system includes: The MWTS observation readout quality control module is used to perform quality control during the observation readout and readout phases. The RTTOV interface quality control module is used to perform the conversion to RTTOV input structure, pre-call checks, forward or K matrix calculations, and post-call checks in the quality control of the radiation calculation stage. The MWTS incremental quality control module is used to perform incremental calculations and incremental phase quality control. The quality control flag management module is used to update channel-level quality control flags and retain or zero out the inverse variance of observation weights. The assimilation diagnostic module is used to count the number of rejections and record channel-level quality control indicators.

Images