Method for extracting subtropical high parameters based on meteorological satellite multispectral imaging observation

Abstract

The application discloses a subtropical high parameter extraction method based on meteorological satellite multispectral imaging observation, which comprises the following steps: S1, projecting target data into a stationary meteorological satellite observation field to calculate a subtropical high influence area, and performing space-time matching on the subtropical high influence area, historical observation data and sea-land template data; S2, randomly sampling a space-time matching data set, selecting time or space discrete points, and dividing a plurality of sampling areas according to different sky states; S3, dividing sample sets of the plurality of sampling areas into a training set and a test set, respectively establishing an identification inversion model by using an XGBoost classification decision tree, and identifying an optimal model by using a Bayesian iteration optimization; S4, inputting real-time stationary observation data, and obtaining an influence probability of each point by using the model to identify the subtropical high influence area; S5, dividing according to a probability threshold value; S6, converting the division result by projection, applying Canny edge detection to a specified range, and extracting a subtropical high outer boundary; and S7, calculating an area index or a strength index of the subtropical high.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of meteorological remote sensing technology, and specifically to a method for extracting subtropical high pressure parameters based on multispectral imaging observations of meteorological satellites. Background Art

[0002] The subtropical high pressure (hereinafter referred to as the "subtropical high") is a key large-scale weather system that controls the tropical and subtropical regions. It has a close interactive relationship with the tropical and mid-high latitude circulation systems. A large number of studies have shown that the intensity and position changes of the subtropical high not only regulate the distribution of droughts and floods and the activities of the main rain belt in the subtropical region, but also have a significant impact on the movement paths of low-latitude weather systems such as typhoons and climate anomalies. my country is located on the west coast of the Pacific Ocean and the southeastern end of Eurasia. Weather and climate changes are closely related to the dynamics of the subtropical high pressure in the western Pacific. This system not only dominates the spatial distribution of my country's rain belts, but also plays a vital role in the total amount of precipitation and the start and end time of the rainy season. Therefore, accurately identifying the location, structure and intensity evolution characteristics of the subtropical high pressure and in-depth understanding of its activity patterns play a key role in improving the refinement of short-term climate prediction and weather forecasting.

[0003] In weather and climate analysis, the influence of a subtropical high is generally defined as the area enclosed by the 5880 gpm contour line in the 500 hPa geopotential height field. This method is simple and straightforward and is a common expression used in weather analysis and forecasting. In the past, identification of subtropical highs was typically based on manual analysis of radiosonde data collected twice daily, morning and evening. Due to the sparse distribution of radiosonde stations and the lack of offshore observations, radiosonde observations provide only a rough estimate of the subtropical high's influence, making it difficult to accurately characterize its internal structure, particularly its structural characteristics over the offshore region. With the advancement of numerical forecasting technology, numerical forecasts based on 500 hPa geopotential height analysis are now increasingly used. Although numerical forecast models can produce meteorological fields with higher temporal and spatial resolution than conventional radiosonde observations, significant discrepancies between forecasts and observations persist due to issues with the model's initial values. Reanalysis data, which integrates observations with numerical forecasts, is currently considered the "optimal" reflection of atmospheric conditions. However, reanalysis data are typically only available with a lag of at least one day, failing to meet the timeliness requirements of weather monitoring and forecasting.

[0004] FY-4 series geostationary satellite is the second generation of FY geostationary meteorological satellite in China. The application goal of the series is to face the meteorological observation demand of China Meteorological Administration around 2020. The instrument design index is equivalent to the latest generation of GOES-R and Himawari series of geostationary meteorological satellites in the United States and Japan. FY-4A and FY-4B satellites have been launched. The advanced geosynchronous radiation imager (AGRI) (hereinafter referred to as imager) carried on the FY-4B satellite has a spectral range of 0.45-13.6 μm, and a total of 15 channels, including 3 visible and near-infrared channels (spectral range of 0.45~0.9 μm), 3 short-wave infrared channels (spectral range of 1.36~2.35 μm), and 9 mid-long wave infrared channels (spectral range of 3.5~13.8 μm). In addition to the visible and near-infrared channels, the mid-long wave infrared channels have a nadir horizontal resolution of 4 km. In addition, based on the infrared channels of the FY-4 satellite imager, the FY-4 satellite can quantitatively obtain the outgoing longwave radiation (OLR) of the earth-atmosphere system observed from space. The outgoing longwave radiation can well reflect the large-scale upward (low OLR) and downward motion (high OLR) of the atmospheric system, and thus is very helpful for revealing the subtropical high activity.

[0005] In summary, the current traditional method of identifying the subtropical high based on sounding or numerical prediction and reanalysis data cannot accurately and efficiently reflect the subtropical high structure and position change, so it is necessary to propose a subtropical high automatic identification method based on high spatiotemporal resolution geostationary meteorological satellite observation data to meet the urgent needs of business. SUMMARY

[0006] Therefore, the present application provides a subtropical high parameter extraction method based on meteorological satellite multispectral imager observation. At present, the identification of subtropical high mainly relies on numerical prediction data, and there is no similar published result on automatic identification and parameter extraction of subtropical high based on meteorological satellite multispectral imager data, and the method of the present application fills this gap.

[0007] In order to achieve the above purpose, the embodiments of the present application provide the following technical solutions:

[0008] A subtropical high parameter extraction method based on meteorological satellite multispectral imaging observation, comprising the following steps:

[0009] S1. Project the target data into the field of view of the stationary meteorological satellite observation by using the bilinear interpolation method, calculate the observed subtropical high influence area according to the set potential height as the division standard, and perform spatio-temporal matching on the subtropical high influence area data observed at the data time, the historical long sequence stationary meteorological satellite observation data and the sea-land template data to obtain a spatio-temporal matching data set.

[0010] S2. Random sampling method is used on the spatio-temporal matching data set to select time or space discrete sample points, wherein in the spatial sampling process, the outgoing longwave radiation (OLR) of the stationary meteorological satellite is used as the division index, and the different celestial body states in the meteorological classification are divided into multiple sampling areas.

[0011] S3. The sample set of the multiple sampling areas is divided into a training set and a test set according to a certain proportion, and an XGBoost classification decision tree method is used to establish an identification inversion model, and in the training process, a Bayesian iteration optimization method is used, and the AUC mean of cross-validation is used as the objective function to build an optimal subtropical high identification model.

[0012] S4. Input real-time stationary meteorological satellite imager observation data, use the optimal subtropical high identification model to identify the subtropical high influence area in the satellite observation field according to the following formula to obtain the subtropical high influence probability of each point;

[0013]

[0014] In the formula, x i represents the prediction characteristic factor of the i-th point; represents the prediction probability of the k-th tree; is the sum of the prediction probabilities of the K trees in the i-th point.

[0015] S5. According to the subtropical high influence probability of each point, each point is divided into a subtropical high influence area and a non-subtropical high influence area according to a set probability threshold, and after traversing all points, a subtropical high identification result is obtained.

[0016] S6. The subtropical high identification result is converted by projection to convert from satellite nominal projection to equi-latitude-longitude grid data, and the subtropical high influence area in the equi-latitude-longitude grid data is regionally segmented and numbered, and the Canny edge detection method is applied to the segmented area in the specified range to extract the outer boundary of the subtropical high.

[0017] S7. The subtropical high index is calculated according to the grid spacing of the subtropical high identification result and the segmented area, including the west ridge point index, the north boundary index, the area index or the intensity index.

[0018] The embodiments of the present application have the following advantages:

[0019] The method for extracting a subtropical high pressure parameter based on meteorological satellite multispectral imager observation has important application prospects for studying the multi-scale structure of the subtropical high pressure, improving the water level of short-term climate prediction and weather forecasting by using the spatiotemporally continuous subtropical high pressure product obtained from the stationary meteorological satellite data. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the following embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only exemplary, and for those skilled in the art, other drawings can be obtained from the provided drawings without creative labor.

[0021] The structures, proportions, sizes, etc. shown in the specification are only used to cooperate with the content disclosed in the specification for understanding and reading by those skilled in the art, and are not used to limit the limiting conditions of the embodiments of the present application, so they do not have technical significance. Any modification of structure, change of proportion relationship or adjustment of size, which does not affect the effects and purposes that can be achieved by the present application, should still fall within the scope of the technical content disclosed by the present application.

[0022] Figure 1 The method flowchart of the subtropical high pressure parameter extraction method based on meteorological satellite multispectral imager observation of the present application;

[0023] Figure 2 The flowchart of the subtropical high pressure parameter extraction method based on meteorological satellite multispectral imager observation of the present application;

[0024] Figure 3 The ROC curve diagram of the subtropical high pressure area recognition result generated by the method of the present application based on FY4B / AGRI satellite data, taking the subtropical high pressure recognition result obtained from ERA5 reanalysis data as the benchmark true value. DETAILED DESCRIPTION

[0025] The embodiments of the present application will be described below by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0026] As shown in Figures 1-2 A subtropical high pressure parameter extraction method based on meteorological satellite multispectral imager observation, the steps are as follows:

[0027] S1. Project the target data into the field of view of the geostationary meteorological satellite observation using the bilinear interpolation method, calculate the observed subtropical high influence area according to the set potential height as the division standard, and perform spatiotemporal matching of the observed subtropical high influence area data with the historical long sequence geostationary meteorological satellite observation data and the sea-land template data based on the data time of the observed subtropical high influence area, to obtain a spatiotemporal matching data set.

[0028] Specifically, the target data in this embodiment is the data of the set potential height in the reanalysis data or the sounding data, such as the data of 500 hPa (Z H , unit: potential ten meters gpm). The target data is projected into the field of view of the geostationary meteorological satellite observation using the bilinear interpolation method, and data with the same resolution as the geostationary meteorological satellite infrared channel observation is formed. The formula for calculating the observed subtropical high (STH) is:

[0029] ;

[0030] In the formula, G is the set potential height, unit: potential meters gpm, and G is preferably 5880 gpm.

[0031] Then, the spatiotemporal matching of the observed subtropical high influence area data with the historical long sequence geostationary meteorological satellite observation data (including the brightness temperature (Tb) of the medium wave and long wave infrared channels of the geostationary meteorological satellite imager, the satellite zenith angle (θ), and the outgoing long wave radiation (OLR) of the geostationary meteorological satellite) and the sea-land template information is performed based on the time of the observed subtropical high influence area data, to obtain a spatiotemporal matching data set.

[0032] S2. Random sampling method is used on the spatiotemporal matching data set to select time or space discrete sample points. In the spatial sampling process, the outgoing long wave radiation OLR of the geostationary meteorological satellite is used as the division index, and the sky state is divided into three sampling areas according to the different weather classifications, i.e., clear sky sampling area, little cloud sampling area, and much cloud sampling area. In order to ensure sample balance, the sample collection is performed according to the principle that the sampling number in the subtropical high and non-subtropical high areas under three different sky conditions is equal, and finally three sample data sets are generated, i.e., clear sky sample set, little cloud sample set, and much cloud sample set. The OLR is used as the threshold value for the clear sky, little cloud, and much cloud area division standard in this technology as follows:

[0033]

[0034] Based on the static meteorological satellite observation data feature parameter extraction, according to the time and location of the sampling point, the information collected includes: satellite observation time, longitude, latitude, land identification, satellite zenith angle cosine (cosθ), satellite zenith angle sine (sinθ), high layer water vapor channel (WV1) brightness temperature (wavelength 5.8-6.7μm), middle layer water vapor channel (WV2) brightness temperature (wavelength 6.75-7.15μm), low layer water vapor channel (WV3) brightness temperature (wavelength 7.24-7.60μm), long wave infrared channel 1 (IR1) brightness temperature (wavelength 8.3-8.8μm), wave infrared channel 2 (IR2) brightness temperature (wavelength 10.3-11.3μm), long wave infrared channel 3 (IR3) brightness temperature (wavelength 11.5-12.5μm), double channel brightness temperature difference 1 (BTD1) (WV1-WV3), double channel brightness temperature difference 2 (BTD2) (WV3-IR1), double channel brightness temperature difference 3 (BTD3) (IR1-IR2), double channel brightness temperature difference 4 (BTD4) (WV1-IR2), double channel brightness temperature difference 5 (BTD5) (IR2-IR3), three channel brightness temperature difference 1 (TBTD1) ((WV1-WV3)+(WV2-WV3)), and two types of OLR index, one is 24 hours, 12 hours, 6 hours, 3 hours and 1 hour OLR anomaly index; the second type is 10 days, 5 days, 24 hours, 12 hours, 6 hours, 3 hours and 1 hour OLR standard deviation index, a total of 30 parameters. Among them, the calculation formulas of OLR anomaly index (OAI: OLR Abnormal Index) and standard deviation index (SOI: Standard OLR deviation Index) are as follows:

[0035]

[0036] In the formula,

[0037] OLR represents the satellite observation OLR of a certain place at the current time;

[0038] OLR represents the satellite observation OLR of a certain place in a time period before the current time;

[0039] Then, OLR represents the standard deviation of the satellite observation OLR of a certain place in a time period before the current time.

[0040] t is respectively 10 days, 5 days, 24 hours, 12 hours, 6 hours, 3 hours and 1 hour.

[0041] S3. The sample set of the three sampling areas of clear sky, few clouds and cloudy is divided into training set and test set according to a certain proportion, and an XGBoost classification decision tree method is used to establish an identification inversion model. In the model training process, a Bayesian iterative optimization method is used, and the mean AUC (an evaluation index of the performance of a classification model) of cross-validation is used as the objective function to build the optimal model for identifying the subtropical high. If the AUC value increases, the identification inversion model is subjected to a new round of iteration, and new feature parameters are added. When the AUC value stops increasing, the iteration is ended, and the current model is the optimal model for identifying the subtropical high. The calculation formula of the mean AUC is as follows:

[0042]

[0043] wherein, represents the ranking number of the positive sample i;

[0044] represents the number of positive samples;

[0045] represents the number of negative samples.

[0046] S4. Real-time static meteorological satellite imager observation data is input, and the optimal model for identifying the subtropical high is used to identify the subtropical high influence area in the field of view of satellite observation, so as to obtain the subtropical high influence probability of each point.

[0047] Specifically, S3 obtains the subtropical high recognition optimal model under clear sky, few clouds and cloudy conditions, wherein the prediction factors in the subtropical high recognition optimal model under clear sky and few clouds conditions include: satellite observation time, longitude, latitude, land-sea identification, satellite zenith angle cosine (cos θ), satellite zenith angle sine (sin θ), high-level water vapor channel (WV1) brightness temperature (wavelength 5.8-6.7 μm), middle-level water vapor channel (WV2) brightness temperature (wavelength 6.75-7.15 μm), low-level water vapor channel (WV3) brightness temperature (wavelength 7.24-7.60 μm), long-wave infrared channel 1 (IR1) brightness temperature (wavelength 8.3-8.8 μm), long-wave infrared channel 2 (IR2) brightness temperature (wavelength 10.3-11.3 μm), long-wave infrared channel 3 (IR3) brightness temperature (wavelength 11.5-12.5 μm), double-channel brightness temperature difference 1 (BTD1) (WV1-WV3), double-channel brightness temperature difference 2 (BTD2) (WV3-IR1), double-channel brightness temperature difference 3 (BTD3) (IR1-IR2), double-channel brightness temperature difference 4 (BTD4) (WV1-IR2), double-channel brightness temperature difference 5 (BTD5) (IR2-IR3), three-channel brightness temperature difference 1 (TBTD1) ((WV1-WV3)+(WV2-WV3)), and 24-hour, 12-hour, 6-hour, 3-hour and 1-hour OLR anomaly indexes, totaling 23 parameters. The optimal model under cloudy conditions further includes 10-day, 5-day, 24-hour, 12-hour, 6-hour, 3-hour and 1-hour OLR standardized deviation indexes in addition to the above prediction factors, totaling 30 parameters.

[0048] The subtropical high recognition optimal model under clear sky, few clouds and cloudy conditions is respectively inputted with real-time static meteorological satellite imager observation time, longitude and latitude, satellite zenith angle, channel brightness temperature, OLR and land-sea template information, and the subtropical high recognition optimal model under clear sky, few clouds and cloudy conditions is used to recognize the subtropical high influence area in the satellite observation field of view according to the following formula to obtain the subtropical high influence probability of each point:

[0049]

[0050] In the formula, x i represents a prediction characteristic factor;

[0051] represents the prediction probability of the kth tree;

[0052] The final prediction result of a certain point is the sum of the prediction probabilities of the K trees.

[0053] S5. For each point, the subtropical high influence probability is divided into subtropical high influence area and non-subtropical high influence area according to the set probability threshold. After traversing all points, the subtropical high recognition result is obtained. Specifically, based on the subtropical high influence probability of each point obtained in S4, the probability of 50% is taken as the threshold. The points with probability < 50% are classified as non-subtropical high influence area and marked as 0; otherwise, as subtropical high influence area and marked as 1. After traversing all points, the subtropical high recognition result is obtained.

[0054] In this step, in order to ensure the spatial continuity of the subtropical high and remove small isolated high pressure bodies in space, morphological closing operation is used to fill the hollow area, and then Gaussian convolution filtering is used to low-pass filter the subtropical high area to remove noise and isolated points.

[0055] The formula of the Gaussian convolution filtering is as follows:

[0056]

[0057] In the formula, x and y represent the row and column coordinates relative to the center point of the filtering window, respectively.

[0058] σ represents the standard deviation of the Gaussian distribution.

[0059] S6. The subtropical high recognition result is converted by projection from satellite nominal projection to equi-latitude and longitude grid data. The subtropical high influence area in the equi-latitude and longitude grid data is regionally segmented, and the segmented regions are numbered. The Canny edge detection method is applied to the segmented regions within the specified range to extract the outer boundary of the subtropical high. The specified range is the user-specified region, such as the range north of 10°N and west of 150°E.

[0060] S7. The area index or intensity index of the subtropical high is calculated according to the grid spacing of the subtropical high recognition result and the segmented regions. The area index is calculated according to the latitude and longitude range of the specified region, and the calculation formula is as follows:

[0061]

[0062] In the formula, is the subtropical high and non-subtropical high influence area determined by the subtropical high probability obtained by inversion in step eight, represented by 0 and 1;

[0063] is the latitude;

[0064] dx and dy are the grid spacing.

[0065] The intensity index is calculated according to the latitude and longitude range of the specified area, the relative volume of the subtropical high main area in the area, and the relative size of the probability value is used to represent, the calculation formula is:

[0066]

[0067] In the formula, The subtropical high and non-subtropical high influence area determined by the subtropical high probability size obtained by inversion in step eight is represented by 0 and 1;

[0068] The latitude is;

[0069] dx and dy are the grid distances;

[0070] The probability size of each grid point in the subtropical high identification area.

[0071] The west ridge point index of the subtropical high in the specified area is obtained according to the extracted outer boundary and influence area of the subtropical high in S7; the west ridge point index is the longitude value of the westernmost point of the outer boundary of the subtropical high main body in the area, if the westernmost point longitude is less than the western boundary longitude of the area, the west ridge point index is set to the western boundary longitude of the area, if it does not exist, the west ridge point index is the eastern boundary longitude of the area.

[0072] The north boundary index of the subtropical high in the specified area is obtained according to the extracted outer boundary and influence area of the subtropical high in S7; the north boundary index is the average value of the latitude of the outer boundary of the subtropical high main body to the polar side in the area according to the latitude and longitude range of the specified area.

[0073] The experiment of the application of the technology:

[0074] And taking the western Pacific subtropical high as an example, the flowchart of extracting the subtropical high west ridge point index, north boundary index, area index and intensity index is shown, and the processing process is as follows:

[0075] 1. Read FY4B / AGRI L1B data and positioning information; the data is in HDF5 format, the time resolution is 15 minutes, the subpoint is located at (0°N, 105°E), the full disc is observed, and the coverage range is the eastern hemisphere area. Among them, channel 9 (wavelength 5.8-6.7 μm), channel 10 (wavelength 6.75-7.15 μm), channel 11 (wavelength 7.24-7.60 μm), channel 12 (wavelength 8.3-8.8 μm), channel 13 (wavelength 10.3-11.3 μm), channel 14 (wavelength 11.5-12.5 μm) channel count value, scaling factor, satellite zenith angle, latitude and longitude data, etc.

[0076] 2. Read the FY4B / AGRI outgoing longwave radiation product; the data is in netcdf format, the time resolution is 15 minutes, the sub-satellite point is located at (0°N, 105°E), the full disc is observed, the coverage range is the eastern hemisphere region, and the spatial resolution is 4km. It contains the row and column numbers of each pixel point in the satellite observation field and the outgoing longwave radiation (OLR) parameter value based on the FY4B / AGRI inversion, the unit is W m -2 According to the method described in the application, it is necessary to read the OLR product data with a time interval of 15 minutes within 10 days before the current time.

[0077] 3. Read the sea-land template data; the data is static data, the spatial resolution is 4km, and the coverage range is global. The data contains latitude and longitude information, and sea-land identification information.

[0078] 4. Calculate the brightness temperature (Tb ch9 , Tb ch10 , Tb ch11 , Tb ch12 , Tb ch13 , Tb ch14 ) of channel 9, 10, 11, 12 and channel 13 and 14 using the channel count value and the scaling coefficient of channel 9 (wavelength 5.8-6.7 μm), channel 10 (wavelength 6.75-7.15 μm), channel 11 (wavelength 7.24-7.60 μm), channel 12 (wavelength 8.3-8.8 μm), channel 13 (wavelength 10.3-11.3 μm), channel 14 (wavelength 11.5-12.5 μm) in FY4B / AGRI L1B data. And on this basis, the double-channel brightness temperature difference (BTD1) of channel 9 and channel 11, the double-channel brightness temperature difference (BTD2) of channel 11 and channel 12, the double-channel brightness temperature difference (BTD3) of channel 12 and channel 13, the double-channel brightness temperature difference (BTD4) of channel 9 and channel 13, the double-channel brightness temperature difference (BTD5) of channel 13 and channel 14, and the three-channel brightness temperature difference (TBTD1) of channel 9, 10 and 11 are calculated.

[0079] 5. Calculate the 24-hour (OAI24h), 12-hour (OAI12h), 6-hour (OAI6h), 3-hour (OAI3h) and 1-hour (OAI1h) OLR anomaly index using the OLR data read in the FY4B / AGRI data; and the 10-day (SOI10d), 5-day (SOI5d), 24-hour (SOI24h), 12-hour (SOI12h), 6-hour (SOI6h), 3-hour (SOI3h) and 1-hour (SOI1h) OLR standardized deviation index, the calculation formula is as follows:

[0080]

[0081] 6. Constructing subtropical high retrieval models under clear sky, light cloud and heavy cloud conditions: Based on the subtropical high area extracted from the FY4B / AGRI long time series satellite observation data and OLR product and ERA5 reanalysis data, the subtropical high sample data set under clear sky, light cloud and heavy cloud sky conditions is established by the method of spatio-temporal matching and random sampling. Among them, the clear sky, light cloud and heavy cloud sky conditions are divided according to the threshold value of OLR≧250W m -2 , 235W m -2 <OLR<250W m -2 , OLR≦235W m -2 . The matching data set covers the time range from March 1, 2024 to January 31, 2025, and the space coverage is taken from the region with FY4B / AGRI satellite zenith angle≦70°. On the basis of the constructed sample set, the training set and the test set are divided according to the ratio of 7:3, and the XGBoost classification decision tree method is used to establish the identification retrieval model. In the model training process, the Bayesian iterative optimization method is used, and the AUC mean (classification model performance evaluation index) of 5 times cross-validation is taken as the objective function to obtain the optimal model: STH_clearsky_bestmodel.m (clear sky), STH_cloudy_bestmodel.m (light cloud), STH_heavycld_bestmodel.m (heavy cloud). In this example, the main parameters of the optimal subtropical high retrieval model under clear sky, light cloud and heavy cloud conditions obtained based on the XGBoost method are shown in the following table:

[0082] Table 1. Main parameter list of optimal subtropical high retrieval model under clear sky, light cloud and heavy cloud conditions

[0083]

[0084] Using the subtropical high retrieval model under clear sky, light cloud and heavy cloud conditions of FY4B satellite obtained in step 6, and the current time prediction factors obtained in steps 3-5, including satellite observation time, longitude, latitude, land-sea identification, satellite zenith angle cosine (cosθ), satellite zenith angle sine (sinθ), channel 9, 10, 11, 12 and channel 13 and 14 brightness temperature (Tb ch9 , Tb ch10 , Tb ch11 , Tb ch12 , Tb ch13 , Tb ch14), the brightness temperature difference of two channels (BTD1, BTD2, BTD3, BTD4, BTD5) and the brightness temperature difference of three channels (TBTD1), the OLR anomaly index (OAI24h, OAI12h, OAI6h, OAI3h and OAI1h) and the OLR standard deviation index (SOI10d, SOI5d, SOI24h, SOI12h, SOI6h, SOI3h and SOI1h) information, the subtropical high influence probability of the current observation time of FY4B / AGRI is obtained by inversion.

[0085] The inversion result is output by pixel and the subtropical high influence probability is represented by different colors, the value range is between 0%-100%, and the smaller the probability is, the lower the possibility of the area being the subtropical high influence area is. The subtropical high and non-subtropical high areas are divided by taking the probability of 50% as the threshold, the morphological closing operation is applied to the inversion result to fill the hollow area, and then the Gaussian convolution filter is used for low-pass filtering of the subtropical high area to remove noise and isolated points. According to the spatial resolution of FY4B / AGRI, the kernel in the closing operation is selected as a 3x3 rectangular template, as shown below. The Gaussian kernel size in the Gaussian filter is selected as 15x15, ;

[0086]

[0087] The subtropical high recognition result after noise removal is converted from satellite nominal projection to 0.25°x0.25° equal latitude and longitude grid data through projection conversion. On this basis, the subtropical high is segmented and the segmented areas are numbered. The Canny edge detection method is applied to the subtropical high area in the specified area range to extract the outer boundary of the subtropical high in the specified area. In this example, the kernel window used for Canny edge detection is 3x3.

[0088] According to the extracted outer boundary and influence area of the subtropical high, the west ridge point index (GD), north boundary index (GB), area index (GM) and intensity index (GQ) of the western Pacific subtropical high are calculated in this example. According to the definition, the western Pacific region is the region west of 150°E and north of 10°N. The calculation method is as follows:

[0089] West ridge point index (GD): taking 10°N and 150°E as the south boundary and east boundary, the longitude value of the westernmost point of the outer boundary of the main body of the subtropical high is found. If the westernmost longitude is less than 70°E, the west ridge point index is set to 70°E, and if it does not exist, the west ridge point index is 150°E.

[0090] North boundary index (GB): the latitude average value of the outer boundary of the north side of the main body of the subtropical high is calculated with 10 °N and 150 °E as the south boundary and the east boundary.

[0091] Area index (GM): the area of the main body of the subtropical high is calculated with 10 °N and 150 °E as the south boundary and the east boundary, and the calculation formula is as follows:

[0092]

[0093] wherein, is the subtropical high and non-subtropical high influence area determined by the probability size of the subtropical high obtained by inversion, and 0 and 1 are used to represent. is the latitude, and dx and dy are the grid distances.

[0094] Intensity index (GQ): the relative volume of the main body of the subtropical high is calculated with 10 °N and 150 °E as the south boundary and the east boundary. According to the relationship between probability and intensity, in the present application, the relative size of the probability value is used to represent the relative volume of the subtropical high. The calculation formula is as follows:

[0095]

[0096] wherein, is the subtropical high and non-subtropical high influence area determined by the probability size of the subtropical high obtained by inversion, and 0 and 1 are used to represent; is the latitude, and dx and dy are the grid distances; is the probability size of each grid point in the subtropical high identification area.

[0097] Experimental results

[0098] In the present experiment, FY4B observation data before and after April 8, 2025, which did not participate in the model training, were used for testing. The data include:

[0099] ① FY4B / AGRI L1B product at 0000 UTC on April 8, 2025, the data file name is:

[0100] FY4B-_AGRI--_N_DISK_1050E_L1-_FDI-_MULT_NOM_20250408000000_20250408001459_4000M_V0001.HDF. (The data contains 6 infrared channel brightness temperature data required in the present application);

[0101] ② FY4B / AGRI GEO data (containing satellite zenith angle information) at 00000 UTC on April 8, 2025, the data file name is:

[0102] FY4B-AGRI-N_DISK_1050E_L1-GEO-MULT_NOM_20250408000000_20250408001459_4000M_V0001.HDF;

[0103] ③ 15-minute interval current time and its previous 10 days within the continuous FY4B OLR retrieval products, a total of 960 data files, data file name: FY4B-AGRI-N_DISK_1050E_L2-OLR-MULT_NOM_20250408000000_20250408001459_4000M_V0001.NC, FY4B-AGRI-N_DISK_1050E_L2-OLR-MULT_NOM_20250407234500_20250407235959_4000M_V0001.NC, FY4B-AGRI-N_DISK_1050E_L2-OLR-MULT_NOM_20250407233000_20250407234459_4000M_V0001.NC ......

[0104] ④ Sea-land template data, data file name:

[0105] landseamask_FY4BNOM.nc

[0106] The file name of the model is:

[0107] STH_clearsky_bestmodel.m

[0108] STH_cloudy_bestmodel.m

[0109] STH_heavycld_bestmodel.m

[0110] In addition, in order to verify the reliability of the experimental results, the subtropical high area obtained by using the formula in S1 based on ERA5 reanalysis data is also used in the experiment to evaluate the accuracy of the experimental results. The data file name is:

[0111] ERA5_STH_20250408000000.nc

[0112] Figure 3The ROC graph of the identification of the subtropical high pressure of FY4B on April 8, 2025 0000 UTC and the identification result of the same time of ERA5 is verified, and the AUC value is 0.916 (the AUC value is usually between 0-1, and the closer the value is to 1, the better the consistency between the two data). The test result shows that the identification result of the subtropical high pressure obtained by the method has good consistency with the identification result of ERA5.

[0113] The method is based on the West Pacific subtropical high influence area and contour line of FY4B satellite on April 8, 2025 0000 UTC. Based on the extraction result, the West Pacific subtropical high west ridge point index, north boundary index, area index and intensity index calculation method in the application are applied to obtain the West Pacific subtropical high west ridge point index of 74°E, the north boundary index of 20.48°N, the area index of 110×10 5 km 2 , and the intensity index of 382.96×10 6 km 2 at this time.

[0114] In order to verify the reliability of the result, we calculated the West Pacific subtropical high parameters based on the ERA5 reanalysis data, using the West Pacific subtropical high west ridge point index, north boundary index, area index and intensity index calculation method revised by China Meteorological Administration. The calculation method is as follows:

[0115] West ridge point index: on the 500 hPa weather map, the most west longitude value of the 5880 gpm contour line on the west side of the main body of the subtropical high in the range of north of 10°N and west of 150°E. If the most west longitude is less than 70°E, the west ridge point index is set to 70°E, and if it does not exist, the west ridge point index is 150°E;

[0116] North boundary index: on the 500 hPa weather map, the average latitude value of the 5880 gpm contour line on the north side of the subtropical high in the range of north of 10°N and west of 150°E;

[0117] Area index: on the 500 hPa weather map, the area of the region surrounded by the 5880 gpm contour line in the range of north of 10°N and west of 150°E, and the calculation formula is as follows:

[0118]

[0119] Wherein, is the latitude, and dx and dy are the grid distance;

[0120] ;

[0121] Hij represents the geopotential height value of a certain grid point on the 500 hPa geopotential height field.

[0122] Intensity index: the relative volume of the subtropical high body with a geopotential height greater than the 5880 gpm isobaric surface in the range of north of 10 °N and west of 150 °E on the 500 hPa weather map, the calculation formula is as follows:

[0123]

[0124] wherein, dx, dy, n ij and H ij The parameter meanings are the same as in the area index.

[0125] The results show that the west ridge point index of the western Pacific subtropical high calculated based on the ERA5 reanalysis data is 70 °E, the north boundary index is 21.68 °N, the area index is 86 × 10 5 km 2 , and the intensity index is 326.39 × 10 6 km 2 . The numerical values are basically equivalent to the parameters of the western Pacific subtropical high estimated based on the subtropical high identified by the FY4B satellite.

[0126] The above examples show that, since the spatial and temporal resolution of the geostationary meteorological satellite (the spatial resolution of the FY4B infrared channel is 4 km, and the temporal resolution is 15 minutes, and the spatial resolution of the infrared channel of Himawari-8 or GOES-R is 2 km, and the temporal resolution is 5-30 minutes) is much higher than that of the current numerical model and sounding observation, the time and space continuous distribution of the subtropical high product obtained by using the geostationary meteorological satellite data can provide higher spatial and temporal resolution of the multi-scale structure information of the subtropical high, and has important application prospects for analyzing and studying the short-term structural change characteristics of the subtropical high and improving short-term climate prediction and weather forecast.

[0127] Although the present application has been described in detail in the foregoing description with general principles and specific embodiments, some modifications or improvements can be made on the basis of the present application, which is obvious to those skilled in the art. Therefore, these modifications or improvements made on the basis of not deviating from the spirit of the present application, all belong to the scope of protection claimed by the present application.

Claims

1. A method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observations, characterized in that: include: S1. Project the target data onto the observation field of view of a geostationary meteorological satellite. Calculate the affected area of ​​the subtropical high pressure using the set geopotential height as the division standard. Based on the data time of the affected area, perform spatiotemporal matching with the historical long-sequence geostationary meteorological satellite observation data and sea and land template data to obtain a spatiotemporal matching dataset. S2. Randomly sample the spatiotemporal matching dataset, select discrete sample points in time or space, and divide the spatial sampling process into multiple sampling areas according to different sky states in meteorological classification; S3. Divide the sample sets of multiple sampling areas into training sets and test sets, and use XGBoost classification decision trees to establish recognition inversion models, and use Bayesian iterative optimization to construct the optimal recognition model; S4. Input real-time geostationary meteorological satellite imager observation data and use the optimal identification model to identify the subtropical high pressure influence area within the satellite observation field of view according to the following formula, and obtain the influence probability of each point; ； Where, xi is the prediction characteristic factor of point i; is the predicted probability of the kth tree; is the sum of the predicted probabilities of the K trees in the i-th point; S5. Based on the comparison of the influence probability of each point with the probability threshold, each point is divided into the influence area and the non-influence area of ​​the subtropical high pressure, thereby obtaining the identification result of the subtropical high pressure; S6. Convert the identification results from the satellite nominal projection into equal latitude and longitude grid data, segment and number the subtropical high pressure affected area in the equal latitude and longitude grid data, apply Canny edge detection to the segmented areas within the specified range, and extract the outer boundary of the subtropical high pressure; S7. Calculate a subtropical high pressure index based on the subtropical high pressure identification result and the grid spacing of the segmented area, including at least one of a west ridge point index, a northern boundary index, an area index, or an intensity index.

2. The method for extracting subtropical high pressure parameters based on multispectral imaging observations of meteorological satellites according to claim 1, wherein: The formula for calculating the observed subtropical high pressure influence area in S1 is: ; Where G is the set geopotential height, and the unit is geopotential meter (gpm).

3. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, wherein: In S2, different sky conditions in meteorological classification are divided into clear sky, partly cloudy and cloudy, i.e., three sampling areas. The regional division standard of the three sampling areas is performed with OLR as the threshold. 。 4. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, wherein: The information collected by the sampling area in S2 includes: satellite observation time, sampling area longitude, sampling area latitude, sea and land identification, satellite zenith angle cosθ, satellite zenith angle sine sinθ, upper water vapor channel brightness temperature, middle water vapor channel brightness temperature, lower water vapor channel brightness temperature, long-wave infrared channel brightness temperature, dual-channel brightness temperature difference, three-channel brightness temperature difference, and OLR anomaly index and OLR standardized deviation index for multiple time periods.

5. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, wherein: If the AUC value increases in S3, the model undergoes a new round of iteration and adds new characteristic parameters. When the AUC value stops increasing, the iteration ends and the optimal model for identifying subtropical high pressure is obtained. The AUC calculation formula is: ； Where, Represents the ranking number of positive sample i; represents the number of positive samples; Indicates the number of negative samples.

6. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, characterized in that: In order to ensure the spatial continuity of the subtropical high pressure, the spatially isolated small-scale high-pressure bodies are removed in S5, and the hole areas are filled by morphological closing operation. Then, the subtropical high pressure area is low-pass filtered using Gaussian convolution filtering to remove noise and isolated points. The formula of the Gaussian convolution filtering is as follows: ； Where x and y represent the row and column coordinates relative to the center point of the filter window respectively; Represents the standard deviation of the Gaussian distribution.

7. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, characterized in that: The area index of the subtropical high pressure in S7 is calculated based on the longitude and latitude range of the specified area, and the calculation formula is: ； Where, The subtropical high pressure and non-subtropical high pressure influence areas determined by the subtropical high pressure probability obtained by inversion in step 8 are represented by 0 and 1; is latitude; dx and dy are the distances between grids.

8. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, characterized in that: The intensity index of the subtropical high pressure in S7 is calculated based on the longitude and latitude range of the specified area, and is expressed by the relative size of the probability value. The calculation formula is: ； Where, The subtropical high pressure and non-subtropical high pressure influence areas determined by the subtropical high pressure probability obtained by inversion in step 8 are represented by 0 and 1; is latitude; dx and dy are the distances between grids; is the probability size of each grid point in the subtropical high pressure identification area.

9. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, characterized in that: In said S7, according to the extracted outer boundary and affected area of ​​the subtropical high pressure, a western ridge point index of the subtropical high pressure in the specified area is obtained; The west ridge point index is to find the longitude value of the westernmost point of the outer boundary of the subtropical high pressure body in the specified area according to the longitude and latitude range of the area. If the longitude of the westernmost point is less than the longitude of the western boundary of the area, the west ridge point index is set as the longitude of the western boundary of the area. If it does not exist, the west ridge point index is the longitude of the eastern boundary of the area.

10. The method for extracting subtropical high pressure parameters based on meteorological satellite multispectral imaging observation according to claim 1, characterized in that: In said S7, based on the extracted outer boundary and affected area of ​​the subtropical high pressure, a northern boundary index of the subtropical high pressure in the specified area is obtained; The northern boundary index is calculated based on the longitude and latitude range of a specified area, and is the average latitude of the outer boundary of the main body of the subtropical high pressure toward the polar side in the area.

Images