Method and system for inverting total rainfall amount of clear sky atmosphere by using FY-4B / AGRI data

Through the processing of FY-4B/AGRI data and the application of random forest algorithm, the problem of inversion of Fengyun-4B satellite data in the total amount of atmospheric precipitation is solved, and high-precision inversion of the total amount of atmospheric precipitation in clear sky is achieved, improving the The accuracy of weather forecast and climate simulation.

CN119001922APending Publication Date: 2024-11-22NANJING UNIV OF INFORMATION SCI & TECH +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411105130.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing technology lacks effective methods to invert the total amount of atmospheric precipitable (TPW) using Fengyun 4 B (FY-4B) satellite data, especially under clear sky conditions, and the traditional methods have relatively good spatial and temporal resolutions. Low, difficult to meet high-precision weather forecasting and climate simulation requirements.

Method used

FY-4B/AGRI data is used to obtain and process bright temperature data and water vapor reanalysis products, combine cloud detection products, use random forest algorithms to perform time and space matching and clear sky identification, filter clear sky data, and use infrared splitting windows and water vapor absorption bands. The bright temperature difference in inversion TPW improves spatial and temporal resolution.

Benefits of technology

It realizes high-precision total precipitation inversion of clear sky and atmospheric precipitation, provides low-cost, high-temporal and spatial resolution water vapor data, makes up for the shortcomings of traditional methods, and improves the accuracy of short-term weather forecasts and climate simulations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119001922A_ABST
    Figure CN119001922A_ABST
Patent Text Reader

Abstract

The invention provides a method and a system for inverting the total rainfall amount of clear sky atmosphere by using FY-4B / AGRI data, and the method and the system for inverting the total rainfall amount of the atmosphere introduce brightness temperatures of three water vapor absorption wavebands and four atmosphere window wavebands as auxiliary data while using an infrared split window to invert TPW. Wherein a channel infrared split window sensitive to atmospheric rainfall is used, the water vapor content is inverted by using the difference of brightness temperature values of the channel infrared split window, three water vapor absorption wavebands respectively respond to water vapor of a high layer, a middle layer and a low layer of a troposphere, four atmospheric window wavebands provide more earth surface radiation information, and the total rainfall amount of the clear sky atmosphere can be inverted with high precision.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of satellite meteorological applications, and relates to a method and a system for inverting the total amount of clear-sky atmospheric precipitable water by using FY-4B / AGRI data. Background Art

[0002] Water vapor is one of the most important greenhouse gases in the Earth's atmosphere and plays an important role in regulating the Earth's climate change. Studying the distribution, changes and influencing factors of water vapor can help us better understand the operating mechanism of the climate system, and then predict and assess the trends and impacts of climate change. Total Precipitable Water (TPW) refers to the total amount of water vapor contained in a vertical column of a unit cross-section extending from the surface to the top of the atmosphere. It is an important indicator factor for characterizing the water vapor content in the atmosphere.

[0003] At present, the measurement methods of TPW mainly include radiosonde method, ground-based detection method and remote sensing inversion method. Radiosonde method is a method of observing the atmosphere by using a radiosonde carried by a sounding balloon. Although the observation accuracy is extremely high, water vapor profiles can only be obtained twice a day; ground-based detection methods mainly include ground-based GNSS and ground-based microwave radiometer detection, but the stations are extremely rare in some remote areas and on the ocean; compared with the first two methods, remote sensing inversion method has the advantages of large observation range and high frequency, and can invert the spatiotemporal distribution of water vapor at regional or global scales.

[0004] Fengyun-4 is a widely used geostationary orbit quantitative remote sensing meteorological satellite. At present, there are few studies on TPW inversion developed for Fengyun-4B (FY-4B), and there is no official TPW product officially announced. Therefore, the research on TPW inversion of FY-4B has important practical significance for improving the accuracy of weather forecasts and climate simulations and monitoring climate change. In addition, the algorithm for TPW inversion by geostationary meteorological satellite remote sensing is mainly based on thermal infrared band data. However, thermal infrared electromagnetic waves cannot penetrate clouds, so only clear sky TPW can be inverted. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention proposes a method and system for inverting the total amount of precipitable water in the clear sky atmosphere using FY-4B / AGRI L1 data.

[0006] The specific technical solutions of the present invention are as follows:

[0007] A method for inverting the total amount of precipitable water in the clear sky atmosphere using FY-4B / AGRI data comprises the following steps:

[0008] S1, obtain the full disk data of FY-4B / AGRI in the area to be studied in a preset period, read the pixel brightness data of the specified band and convert it into brightness temperature data to obtain a brightness temperature data set, which contains the latitude and longitude, time and brightness temperature information of each pixel;

[0009] S2, obtain the TCWV data of the water vapor reanalysis product corresponding to the study area and time period, obtain the TPW data of each grid point, and calculate the secant value of the satellite observation zenith angle of each grid point for angle correction to obtain the TPW data set, which contains the latitude and longitude, time, TPW value and secant value of the satellite observation zenith angle of each grid point;

[0010] S3: performing spatiotemporal matching on the brightness temperature dataset and the TPW dataset. During the spatiotemporal matching, based on longitude and latitude, the data in the brightness temperature dataset is read, each grid point of the TPW dataset is traversed, and the pixel closest to each grid point in the brightness temperature dataset at the same time is searched, and the TPW dataset corresponding to each grid point is merged with the brightness temperature dataset of the corresponding pixel to obtain a matching dataset of brightness temperature and TPW;

[0011] S4, using the cloud detection product CLM corresponding to the study area and time period to distinguish clear sky, filter out the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtain the matching data set screened by the clear sky threshold;

[0012] S5, using the matching dataset screened by the clear sky threshold, modeling and inversion of TPW were performed based on the random forest algorithm.

[0013] Furthermore, in step S1, reading the pixel brightness data of the specified band includes: the brightness data of the pixels in the 9th band to the 15th band.

[0014] Furthermore, in step S5, the importance of each parameter in inverting TPW is measured by the importance score of each parameter:

[0015] For the matching data set selected by the clear sky threshold, the brightness temperature data of the 9th to 11th band, the brightness temperature data of the 12th to 15th band, the brightness temperature difference between the 13th and 14th bands, the secant value of the satellite observation zenith angle, the longitude, latitude and time parameters are calculated by X. J To represent, J = 1, 2, 3, ..., 12, assuming that the model has a total of I decision trees, C categories, the parameter importance score is represented by S, and the Gini index is represented by GI, then the steps for calculating the importance of each parameter are as follows:

[0016] Calculate the GI index of the i-th tree node q:

[0017]

[0018] In the formula, C means there are C categories, P qc Indicates the proportion of category c in node q;

[0019] Calculation parameter X J The importance of node q in the i-th tree, that is, the change in GI before and after node q branches:

[0020]

[0021] In the formula, and Respectively represent the GI index of the two new nodes before and after the branch;

[0022] Assume that the parameter X J The nodes that appear in decision tree i are set Q, then X J The importance of the i-th tree is:

[0023]

[0024] Assuming there are I trees in the model, the parameter importance score is:

[0025]

[0026] Finally, the importance scores obtained in the previous step are normalized to obtain the importance scores S of each parameter. J (J=1,2,3,...,12):

[0027]

[0028] Furthermore, in step S2, the water vapor reanalysis product TCWV data adopts ERA5 water vapor reanalysis product TCWV data; in step S4, the cloud detection product CLM adopts FY-4B cloud detection product CLM.

[0029] Furthermore, the specific process of step S1 is as follows:

[0030] First, the FY-4B data row and column number and longitude and latitude lookup table is downloaded from the Fengyun satellite remote sensing data service website, and the row and column number of each pixel in the full disk data is converted into longitude and latitude. Then, according to formula (1) and formula (2), the brightness values ​​of the pixels in the 9th to 15th bands are converted into brightness temperature data respectively. Finally, the data of each pixel is read and input to generate text to obtain the brightness temperature data set.

[0031] The brightness value DN is converted to the radiant brightness L by formula (1) λ , calculate the brightness temperature data T by formula (2):

[0032] L λ=SCALE*DN+OFFSET (1)

[0033]

[0034] In the formula, SCALE is the slope at the corresponding wavelength, OFFSET is the intercept at the corresponding wavelength, λ is the wavelength, and L λ is the radiation brightness, C1=3.7418×108W·m -2 μm 4 ,C2=14388μm·K.

[0035] Furthermore, the specific process of step S2 is as follows:

[0036] First, the ERA5 water vapor reanalysis product TCWV was downloaded from the European Centre for Medium-Range Weather Forecasts. The secant value VZA of the satellite observation zenith angle corresponding to each grid point was calculated using formula (3) to obtain the TPW dataset.

[0037]

[0038] In the formula, r is the distance from the satellite to the center of the earth, d is the distance from the satellite to the grid point, and the intermediate variable γ is calculated by formula (4):

[0039]

[0040] In the formula, is the latitude of the grid point, λ is the longitude of the grid point, s is the longitude of the satellite's subsatellite point.

[0041] Furthermore, the specific process of step S4 is as follows:

[0042] Download the FY-4B cloud detection product (CLM) from the Fengyun satellite remote sensing data service website. In this product, 0 represents clouds, 1 represents possible clouds, 2 represents possible clear sky, 3 represents clear sky, 126 represents null value, and 127 represents fill value. Use the FY-4B data row and column number and longitude and latitude lookup table to assign longitude and latitude to the cloud detection CLM product. According to the longitude and latitude, read and write the cloud detection product into a new column of the matching data set in step S3, and only retain the row data with numbers 1, 2 or 3 in the column, so as to obtain the matching data set with cloud data excluded, that is, the matching data set filtered by the clear sky threshold.

[0043] Furthermore, the specific process of step S5 is as follows:

[0044] First, according to the longitude and latitude, 70% of the matching data set that has been screened by the clear sky threshold is randomly selected as the training set and 30% as the test set. The training set is used as the original data set for the establishment of the random forest model.

[0045] Then, use Bootstrap to randomly select m samples from the original training set with replacement sampling, perform n sampling times in total, generate n training sets, and train n decision tree models as random forest models;

[0046] Finally, the data of the test set were input as independent variables into the established random forest model, and the TPW data calculated by the model were output;

[0047] The output TPW data were matched with the TPW data in the ERA5 water vapor reanalysis product TCWV data, and the accuracy indicators of the two were calculated to evaluate the accuracy of FY-4B / AGRI's TPW inversion based on the random forest algorithm. The data input as independent variables into the established random forest model included the brightness temperature data of the 9th to 11th bands, the brightness temperature data of the 12th to 15th bands, the brightness temperature difference between the 13th and 14th bands, the secant value of the satellite observation zenith angle, TPW data, longitude, latitude and time.

[0048] A system for retrieving the total amount of clear-air atmospheric precipitable water using FY-4B / AGRI data, including the following parts:

[0049] The brightness temperature data processing module obtains the full disk data of FY-4B / AGRI in the area to be studied in the preset period, reads the pixel brightness value of the specified band and converts it into brightness temperature data to obtain a brightness temperature data set, which contains the latitude and longitude, time and brightness temperature information of each pixel;

[0050] The TPW data processing module obtains the TCWV data of the water vapor reanalysis product corresponding to the study area and time period, obtains the TPW data of each grid point, and calculates the secant value of the satellite observation zenith angle of each grid point for angle correction to obtain the TPW data set. The TPW data set contains the latitude and longitude, time, TPW value and secant value of the satellite observation zenith angle of each grid point;

[0051] A brightness temperature and TPW data matching module performs spatiotemporal matching on the brightness temperature dataset and the TPW dataset. During the spatiotemporal matching, the data in the brightness temperature dataset is read based on longitude and latitude, and each grid point of the TPW dataset is traversed to search for the pixel closest to each grid point in the brightness temperature dataset at the same time, and the TPW dataset corresponding to each grid point is merged with the brightness temperature dataset of the corresponding pixel to obtain a matching dataset of brightness temperature and TPW;

[0052] The clear sky discrimination module uses the cloud detection product CLM corresponding to the study area and time period to discriminate the clear sky, and screens out the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtains the matching data set screened by the clear sky threshold;

[0053] The TPW inversion module uses the matching dataset screened by the clear sky threshold to model and invert TPW based on the random forest algorithm.

[0054] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the method as described above.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The method and system for inverting the total amount of atmospheric precipitable water in clear sky using FY-4B / AGRI data introduces brightness temperatures of three water vapor absorption bands and four atmospheric window bands as auxiliary data while using an infrared split window to invert the total amount of atmospheric precipitable water TPW. The infrared split window of a channel sensitive to atmospheric precipitable water is used to invert the water vapor content using the difference in brightness temperature values. The three water vapor absorption bands respond to water vapor in the upper, middle and lower layers of the troposphere, respectively. The four atmospheric window bands provide more surface radiation information, and the total amount of atmospheric precipitable water in clear sky can be inverted with high precision.

[0057] Based on the high temporal and spatial resolution of FY-4B / AGRI data, the present invention creates brightness temperature data with a temporal resolution of up to 1h and a spatial resolution of up to 4km, which are accurately registered and fused with the corrected ERA5 water vapor product; the mid-infrared and thermal infrared brightness temperature data in the 9th to 15th bands are used to invert the total amount of atmospheric precipitable water during the day and at night; at the same time, compared with traditional thermal infrared method, radiosonde method and remote sensing inversion method, the spatial resolution and temporal resolution are higher.

[0058] The method and system for inverting the total amount of atmospheric precipitable water in clear sky of the present invention selects infrared split window data of brightness temperature difference between bands 13 and 14 in consideration of the influence on the sensitivity of atmospheric precipitable water; considers the comprehensiveness of the data and contains more surface information, selects water vapor information of water vapor absorption bands of multiple tropospheric layers in bands 9 to 11 at the same time, and provides four atmospheric window band data of bands 12 to 15 with more surface radiation information, thereby inverting the total amount of atmospheric precipitable water in clear sky with high precision.

[0059] The present invention uses FY-4B / AGRI data to invert the total amount of atmospheric precipitable water, solving the problem that the radiosonde method and ground-based detection method in the prior art are restricted by economic, geographical environment and technical factors, as well as the low spatial and temporal resolution of observation. It provides users with low-cost, high-precision and high spatial and temporal resolution water vapor data, making up for the shortcomings of traditional detection instruments and meteorological stations in monitoring water vapor, and is beneficial to improving the accuracy of short-term weather forecasts and climate simulations. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1It is the flow chart of FY-4B / AGRI inversion TPW in the embodiment;

[0061] Figure 2 It is a scatter plot of the correlation between FY-4B / AGRI-TPW and ERA5-TPW in the ocean area in the embodiment;

[0062] Figure 3 It is a scatter plot of the correlation between FY-4B / AGRI-TPW and ERA5-TPW in the land area of ​​the embodiment;

[0063] Figure 4 is the parameter importance score output by the random forest algorithm in the ocean area;

[0064] Figure 5 is the parameter importance score output by the random forest algorithm in the land area.

[0065] Importance score, range parameters, latitude, longitude; Total Precipitable Water (TPW); TCWV (total volumetric water vapor content). DETAILED DESCRIPTION

[0066] The present invention is further described and illustrated below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0067] Embodiment 1:

[0068] The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data of the present invention comprises the following steps:

[0069] S1, obtain the full disk data of FY-4B / AGRI in the area to be studied in a preset period, read the pixel brightness value of the specified band and convert it into brightness temperature data to obtain a brightness temperature data set, which contains the latitude and longitude, time and brightness temperature information of each pixel;

[0070] S2, obtain the TCWV data of the water vapor reanalysis product corresponding to the study area and time period, obtain the TPW data of each grid point, and calculate the secant value of the satellite observation zenith angle of each grid point for angle correction to obtain the TPW data set, which contains the latitude and longitude, time, TPW value and secant value of the satellite observation zenith angle of each grid point;

[0071] S3: performing spatiotemporal matching on the brightness temperature dataset and the TPW dataset. During the spatiotemporal matching, based on longitude and latitude, the data in the brightness temperature dataset is read, each grid point of the TPW dataset is traversed, and the pixel closest to each grid point in the brightness temperature dataset at the same time is searched, and the TPW dataset corresponding to each grid point is merged with the brightness temperature dataset of the corresponding pixel to obtain a matching dataset of brightness temperature and TPW;

[0072] S4, using the cloud detection product CLM corresponding to the study area and time period to distinguish clear sky, filter out the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtain the matching data set screened by the clear sky threshold;

[0073] S5, using the matching dataset screened by the clear sky threshold, modeling and inversion of TPW were performed based on the random forest algorithm.

[0074] The data are in the following forms: Brightness temperature dataset:

[0075]

[0076] TPW dataset:

[0077] longitude latitude time TPW VxDV 94.75 32.25 1073068 6.6104243 1.758693185 95 32.25 1073068 7.190581885 1.750265572 95.25 32.25 1073068 8.180766456 1.741957855 95.5 32.25 1073068 9.337409743 1.733768034 ... ... ... ... ...

[0078] Matching datasets for brightness temperature and TPW:

[0079]

[0080] Cloud detection product data:

[0081]

[0082]

[0083] After assigning longitude and latitude, the cloud detection product data set is as shown in the table above, where 0 represents clouds, 1 represents possible clouds, 2 represents possible clear sky, 3 represents clear sky, 126 represents null value, and 127 represents fill value.

[0084] Embodiment 2:

[0085] like Figure 1 As shown in FIG. 1 , the method for inverting the total atmospheric precipitable water (TPW) by FY-4B / AGRI of the present invention is as follows:

[0086] S1: Obtain the full disk data of FY-4B / AGRI L1 in the area to be studied in the preset period, read the data of the specified band, perform geometric correction on it, and then convert the pixel brightness value into brightness temperature to obtain the brightness temperature dataset.

[0087] In addition to the brightness temperature data of the infrared split window bands (10.80μm and 12.00μm) commonly used to invert TPW, the brightness temperature data in this dataset also includes brightness temperature data of three water vapor absorption bands (6.25μm, 6.95μm and 7.42μm) and four atmospheric window bands (8.55μm, 10.80μm, 12.00μm and 13.30μm). The data of the seven bands used are shown in Table 1:

[0088] Table 1 FY-4B / AGRI water vapor related bands

[0089] Band Central wavelength (μm) Bandwidth(μm) Spatial resolution (km) 9 6.25 5.80~6.70 4 10 6.95 6.75~7.15 4 11 7.42 7.24~7.60 4 12 8.55 8.3~8.8 4 13 10.80 10.30~11.30 4 14 12.00 11.50~12.50 4 15 13.3 13.00~13.60 4

[0090] The specific steps of step S1 are as follows:

[0091] The FY-4B / AGRI L1 (4KM) full disk data was downloaded from the Fengyun satellite remote sensing data service website. The data format is HDF, including 15 bands, and the number of pixel rows and columns in each band is 2748×2748.

[0092] Read the data of the specified band in the downloaded data, including the brightness data of the 9th to 15th bands. For further processing of the data of the specified band, first download the "FY-4B data row and column number and longitude and latitude lookup table 4km (133°E)" file from the Fengyun satellite remote sensing data service website for geometric correction, where 133°E is the longitude of the subsatellite point of the FY-4B satellite. That is, according to the lookup table, the row and column number of each pixel is converted into longitude and latitude, that is, each pixel is assigned longitude and latitude. Then perform radiometric calibration on the data, and convert the DN values ​​of the pixels in the 9th to 15th bands (mid-infrared and thermal infrared bands) into brightness temperature according to formulas (1) and (2) respectively. Finally, the data of each pixel is read and input into text. The first column is longitude, the second column is latitude, the third column is time, the fourth to sixth columns are brightness temperatures of bands 9 to 11, the seventh to tenth columns are brightness temperatures of bands 12 to 15, and the eleventh column is the brightness temperature difference between bands 13 and 14. The brightness temperature dataset is obtained.

[0093] The brightness value DN is converted to the radiant brightness L by formula (1) λ , and then calculate the brightness temperature data T by formula (2):

[0094] L λ =SCALE*DN+OFFSET (1)

[0095]

[0096] Where SCALE and OFFSET are the slope and intercept at the corresponding wavelength, λ is the wavelength, and L λis the radiation brightness, C1=3.7418×108W·m -2 μm 4 ,C2=14388μm·K.

[0097] S2: Obtain the TCWV data of the ERA5 water vapor reanalysis product for the same study area and the same period, obtain the TPW data of each grid point, and calculate the secant value of the satellite observation zenith angle of each grid point for angle correction, and obtain the TPW data set containing the TPW data and the secant value of the satellite observation zenith angle at the grid point with longitude and latitude as coordinates. The specific process of this step is as follows:

[0098] Download the ERA5 water vapor reanalysis product TCWV from the European Center for Medium-Range Weather Forecasts. The product file format is NC. Read and write it into text. The first column is longitude, the second column is latitude, the third column is time, and the fourth column is TPW. At the same time, calculate the satellite observation zenith angle VZA corresponding to each grid point. The calculation formula is:

[0099]

[0100] Where r is the distance from the satellite to the center of the earth, d is the distance from the satellite to the ERA5 grid point, and the calculation formula of the intermediate variable γ is as follows:

[0101]

[0102] In the formula, is the latitude of the ERA5 grid point, λ is the longitude of the ERA5 grid point, s is the longitude of the satellite subsatellite point. Calculate the secant value of the satellite observation zenith angle of each grid point and write it into the fifth column of the text to obtain the ERA5 dataset.

[0103] In step S2, the TCWV product provides TPW in the form of grid points, with a spatial resolution of 0.25°×0.25°, a temporal resolution of 1 h, and a horizontal and vertical grid number of 1440×721.

[0104] S3: Perform spatiotemporal matching of the brightness temperature dataset and the TPW dataset to obtain the TPW of each grid point in the study area and the brightness temperature of the corresponding pixel in the AGRI image, i.e., the brightness temperature and TPW matching dataset. The specific process of this step is as follows:

[0105] Based on longitude and latitude, the grid points of the TCWV product are traversed to search for the pixel closest to the grid point in the AGRI image at the same time, and the TPW dataset corresponding to the grid point is merged with the brightness temperature dataset of the corresponding pixel, and then read and written into the same text.

[0106] S4: Use the FY-4B cloud detection product CLM in the same study area and the same period to identify clear skies, remove the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtain the matching data set screened by the clear sky threshold.

[0107] The specific process of this step is as follows:

[0108] Download the FY-4B cloud detection product CLM from the Fengyun satellite remote sensing data service website. In the description of this product, 0 represents clouds, 1 represents possible clouds, 2 represents possible clear sky, 3 represents clear sky, 126 represents null value, and 127 represents fill value. Similarly, use the "FY-4B data row and column number and longitude and latitude lookup table 4km (133°E)" file to assign longitude and latitude to the CLM product. According to the longitude and latitude, read and write the product into a new column in the merged dataset in S3, and only keep the row data with numbers 1, 2 or 3 in the column, and get the matching dataset with cloud data removed, that is, the matching dataset after clear sky discrimination.

[0109] S5: Using the corresponding latitude, longitude, time and other data in the matching data set after clear sky discrimination, model and invert TPW in land and ocean areas based on the random forest algorithm. The specific operations of this step are as follows:

[0110] First, according to the longitude and latitude, 70% of the matching data set that has been screened by the clear sky threshold is randomly selected as the training set and 30% as the test set. The training set is used as the original data set for establishing the random forest model.

[0111] Then use Bootstrap to randomly select m samples from the original training set with replacement sampling, perform n sampling times in total, generate n training sets, and train n decision tree models as random forest models.

[0112] Finally, the brightness temperature, longitude and latitude, and time of the test set were input as independent variables into the established random forest model. The TPW calculated by the output model was matched with the TPW provided by the ERA5 water vapor reanalysis product TCWV, and the accuracy indicators of the two were calculated to evaluate the accuracy of the TPW inversion based on the random forest algorithm of FY-4B / AGRI.

[0113] In step S1, the brightness temperature data set includes: the first column is longitude, the second column is latitude, the third column is time, the fourth to sixth columns are the brightness temperatures of the 9th to 11th bands, the seventh to tenth columns are the brightness temperatures of the 12th to 15th bands, and the 11th column is the brightness temperature difference between the 13th and 14th bands.

[0114] In step S2, the TPW data set: the first column is longitude, the second column is latitude, the third column is time, the fourth column is TPW, and the fifth column is the secant value of the satellite observation zenith angle.

[0115] Embodiment three:

[0116] This example is further designed as follows: The specific process of modeling and inverting TPW based on the random forest algorithm in step S5 is as follows:

[0117] First, according to the longitude and latitude, 70% of the matching data set that has been screened by the clear sky threshold is randomly selected as the training set and 30% as the test set. The training set is used as the original data set for the establishment of the random forest model.

[0118] Then, Bootstrap is used to randomly select m samples from the original training set with replacement sampling, a total of n samplings are performed to generate n training sets, and n decision tree models are trained as random forest models;

[0119] Finally, the data of the test set are input as independent variables into the established random forest model, and the TPW data calculated by the model are output; and the TPW data in the ERA5 water vapor reanalysis product TCWV data are matched, and the accuracy indicators of the two are calculated to evaluate the accuracy of FY-4B / AGRI's TPW inversion based on the random forest algorithm. The data input as independent variables into the established random forest model include the brightness temperature data of the 9th to 11th bands, the brightness temperature data of the 12th to 15th bands, the brightness temperature difference between the 13th and 14th bands, the secant value of the satellite observation zenith angle, longitude, latitude and time, and the output is TPW data.

[0120] For the inverted TPW, the root mean square error RMSE, mean absolute error MAE and correlation coefficient R are used as accuracy indicators to evaluate and analyze the error of FY-4B / AGRI TPW. The specific formula is:

[0121]

[0122] In the formula, y o For FY-4B / AGRI-TPW, is the average value of FY-4B / AGRI-TPW, y r For ERA5-TPW, is the average value of ERA5-TPW in mm, and n is the total number of observations.

[0123] Embodiment 4:

[0124] This example is further designed to measure the importance of each parameter in the inversion of TPW by the importance score of each parameter in the random forest algorithm in step S5:

[0125] The Gini index, also known as the Gini impurity coefficient, is a common tool for classifying attribute purity. The importance score is calculated based on the Gini index. For the brightness temperature data of the 9th to 11th bands, the brightness temperature data of the 12th to 15th bands, the brightness temperature difference between the 13th and 14th bands, the secant value of the satellite observation zenith angle, longitude, latitude and time, a total of 12 parameters are used in X J To represent, where J = 1, 2, 3, ..., 12, assuming that the model has a total of I decision trees, C categories, the parameter importance score is represented by S, and the Gini index is represented by GI, then the steps for calculating the importance of each parameter are as follows:

[0126] Calculate the GI index of the i-th tree node q:

[0127]

[0128] In the formula, C means there are C categories, P qc Indicates the proportion of category c in node q.

[0129] Calculation parameter X J The importance of node q in the i-th tree, that is, the change in GI before and after node q branches:

[0130]

[0131] In the formula, and Respectively represent the GI index of the two new nodes before and after the branch.

[0132] Assume that the parameter X J The nodes that appear in decision tree i are set Q, then X J The importance of the i-th tree is:

[0133]

[0134] Assuming there are I trees in the model, the parameter importance score is:

[0135]

[0136] Finally, the importance scores obtained in step 4 are normalized to obtain the importance scores S of each parameter. J (J=1,2,3,...,12):

[0137]

[0138] Embodiment five:

[0139] This example provides a system for inverting the total amount of clear-air atmospheric precipitable water using FY-4B / AGRI data, which is used to implement the method of the present invention and includes the following parts:

[0140] The brightness temperature data processing module obtains the full disk data of FY-4B / AGRI in the area to be studied in the preset period, reads the pixel brightness value of the specified band and converts it into brightness temperature data to obtain a brightness temperature data set, which contains the latitude and longitude, time and brightness temperature information of each pixel;

[0141] The TPW data processing module obtains the TCWV data of the water vapor reanalysis product corresponding to the study area and time period, obtains the TPW data of each grid point, and calculates the secant value of the satellite observation zenith angle of each grid point for angle correction to obtain the TPW data set. The TPW data set contains the latitude and longitude, time, TPW value and secant value of the satellite observation zenith angle of each grid point;

[0142] A brightness temperature and TPW data matching module performs spatiotemporal matching on the brightness temperature dataset and the TPW dataset. During the spatiotemporal matching, the data in the brightness temperature dataset is read based on longitude and latitude, and each grid point of the TPW dataset is traversed to search for the pixel closest to each grid point in the brightness temperature dataset at the same time, and the TPW dataset corresponding to each grid point is merged with the brightness temperature dataset of the corresponding pixel to obtain a matching dataset of brightness temperature and TPW;

[0143] The clear sky discrimination module uses the cloud detection product CLM corresponding to the study area and time period to discriminate the clear sky, and screens out the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtains the matching data set screened by the clear sky threshold;

[0144] The TPW inversion module uses the matching dataset screened by the clear sky threshold to model and invert TPW based on the random forest algorithm.

[0145] Embodiment five:

[0146] This example provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute the method steps of the present invention.

[0147] Test Example:

[0148] The land research area selected in this example is 26°-60°N, 60°-117°E; the ocean research area is 12°-21°N, 113°-118°E.

[0149] Based on the method of the present invention, the clear sky TPW inversion is performed on the FY-4B / AGRI remote sensing image data from June 2022 to May 2023. Five remote sensing images are evenly selected in each month, all at UTC 4h, and there are 15 remote sensing images in each quarter, for a total of 60 images. The selected data sets also include the ERA5 water vapor reanalysis product TCWV and the FY-4B / AGRI cloud detection product CLM in the same area and at the same time. The data processing flow is as follows Figure 1 shown.

[0150] Plot the correlation scatter plots of TPW provided by FY-4B / AGRI-TPW and ERA5 water vapor reanalysis product TCWV in the marine study area. Figure 2 As shown. Figure 2 It can be seen that most of the matching points are concentrated near 1:1, indicating that the FY-4B / AGRI-TPW product and ERA5-TPW have good consistency; from the accuracy index, the correlation coefficient between FY-4B / AGRI-TPW and ERA5-TPW is as high as 0.989, the root mean square error is 2.279mm, and the mean absolute error is 1.280mm. In summary, FY-4B / AGRI can invert clear sky TPW with high accuracy in the ocean area.

[0151] The correlation scatter plots of the TPW provided by FY-4B / AGRI-TPW and ERA5 water vapor reanalysis product TCWV in the land study area are shown in the figure. Figure 3 As shown. Figure 3 It can be seen that most of the matching points are concentrated near 1:1, indicating that the FY-4B / AGRI-TPW product and ERA5-TPW have good consistency; from the accuracy index, the correlation coefficient between FY-4B / AGRI-TPW and ERA5-TPW is as high as 0.997, the root mean square error is 0.972mm, and the mean absolute error is 0.483mm. In summary, FY-4B / AGRI can invert clear sky TPW with high accuracy in land areas.

[0152] At present, the atmospheric precipitable water content (same area and same time) inverted by the traditional thermal infrared split window algorithm has a mean absolute error MAE of about 6 mm and a correlation coefficient R of about 0.8; compared with this, the method for inverting atmospheric precipitable water provided by the present invention has a MAE of 1.280 mm and a correlation coefficient R of 0.989 in the marine research area, which has extremely high accuracy; the present invention has a higher accuracy in the land area, with a MAE of 0.483 mm and a correlation coefficient R of 0.997. It can be seen that the method provided by the present invention realizes the high-precision inversion of clear sky atmospheric precipitable water content.

[0153] Figure 4 and Figure 5are the parameter importance scores output by the random forest algorithm in the ocean and land areas, respectively. Figure 4 and Figure 5 It can be seen that although the split window has the greatest impact on the model regardless of the ocean or land area, the influence of the water vapor absorption band and the atmospheric window band on the model cannot be ignored, especially in the ocean area, which has important practical significance for the high-precision inversion of TPW.

[0154] The above-mentioned embodiment is only a preferred solution of the present invention, but it is not intended to limit the present invention. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all technical solutions obtained by equivalent replacement or equivalent transformation fall within the protection scope of the present invention.

Claims

1. A method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data, characterized in that: The following steps are involved: S1, obtain the full disk data of FY-4B / AGRI in the area to be studied in a preset period, read the pixel brightness data of the specified band and convert it into brightness temperature data to obtain a brightness temperature data set, which contains the latitude and longitude, time and brightness temperature information of each pixel; S2, obtain the TCWV data of the water vapor reanalysis product corresponding to the study area and time period, obtain the TPW data of each grid point, and calculate the secant value of the satellite observation zenith angle of each grid point for angle correction to obtain the TPW data set, which contains the latitude and longitude, time, TPW value and secant value of the satellite observation zenith angle of each grid point; S3: performing spatiotemporal matching on the brightness temperature dataset and the TPW dataset. During the spatiotemporal matching, based on longitude and latitude, the data in the brightness temperature dataset is read, each grid point of the TPW dataset is traversed, and the pixel closest to each grid point in the brightness temperature dataset at the same time is searched, and the TPW dataset corresponding to each grid point is merged with the brightness temperature dataset of the corresponding pixel to obtain a matching dataset of brightness temperature and TPW; S4, using the cloud detection product CLM corresponding to the study area and time period to distinguish clear sky, filter out the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtain the matching data set screened by the clear sky threshold; S5, using the matching dataset screened by the clear sky threshold, modeling and inversion of TPW were performed based on the random forest algorithm.

2. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 1, characterized in that: In step S1, reading the pixel brightness data of the specified band includes: the brightness data of the pixels in the 9th band to the 15th band.

3. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 2 is characterized in that: In step S5, the importance of each parameter in inverting TPW is measured by the importance score of each parameter: For the matching data set selected by the clear sky threshold, the brightness temperature data of the 9th to 11th band, the brightness temperature data of the 12th to 15th band, the brightness temperature difference between the 13th and 14th bands, the secant value of the satellite observation zenith angle, the longitude, latitude and time parameters are calculated by X. J To represent, J = 1, 2, 3, ..., 12, assuming that the model has a total of I decision trees, C categories, the parameter importance score is represented by S, and the Gini index is represented by GI, then the steps for calculating the importance of each parameter are as follows: Calculate the GI index of the i-th tree node q: In the formula, C means there are C categories, P qc Indicates the proportion of category c in node q; Calculation parameter X J The importance of node q in the i-th tree, that is, the change in GI before and after node q branches: In the formula, and Respectively represent the GI index of the two new nodes before and after the branch; Assume that the parameter X J The nodes that appear in decision tree i are set Q, then X J The importance of the i-th tree is: Assuming there are I trees in the model, the parameter importance score is: Finally, the importance scores obtained in the previous step are normalized to obtain the importance scores S of each parameter. J (J=1,2,3,...,12):

4. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 1, characterized in that: In step S2, the water vapor reanalysis product TCWV data adopts the ERA5 water vapor reanalysis product TCWV data; in step S4, the cloud detection product CLM adopts the FY-4B cloud detection product CLM.

5. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 1, characterized in that: The specific process of step S1 is as follows: First, the FY-4B data row and column number and longitude and latitude lookup table is downloaded from the Fengyun satellite remote sensing data service website, and the row and column number of each pixel in the full disk data is converted into longitude and latitude. Then, according to formula (1) and formula (2), the brightness values ​​of the pixels in the 9th to 15th bands are converted into brightness temperature data respectively. Finally, the data of each pixel is read and input to generate text to obtain the brightness temperature data set. The brightness value DN is converted to the radiant brightness L by formula (1) λ , calculate the brightness temperature data T by formula (2): L λ =SCALE*DN+OFFSET (1) In the formula, SCALE is the slope at the corresponding wavelength, OFFSET is the intercept at the corresponding wavelength, λ is the wavelength, and L λ is the radiation brightness, C1=3.7418×108W·m -2 μm 4 ,C2=14388μm·K.

6. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 1, characterized in that: The specific process of step S2 is as follows: First, the ERA5 water vapor reanalysis product TCWV was downloaded from the European Centre for Medium-Range Weather Forecasts. The secant value VZA of the satellite observation zenith angle corresponding to each grid point was calculated using formula (3) to obtain the TPW dataset. In the formula, r is the distance from the satellite to the center of the earth, d is the distance from the satellite to the grid point, and the intermediate variable γ is calculated by formula (4): In the formula, is the latitude of the grid point, λ is the longitude of the grid point, s is the longitude of the satellite's subsatellite point.

7. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 1, characterized in that: The specific process of step S4 is as follows: Download the FY-4B cloud detection product (CLM) from the Fengyun satellite remote sensing data service website. In this product, 0 represents clouds, 1 represents possible clouds, 2 represents possible clear sky, 3 represents clear sky, 126 represents null value, and 127 represents fill value. Use the FY-4B data row and column number and longitude and latitude lookup table to assign longitude and latitude to the cloud detection CLM product. According to the longitude and latitude, read and write the cloud detection product into a new column of the matching data set in step S3, and only retain the row data with numbers 1, 2 or 3 in the column, so as to obtain the matching data set with cloud data excluded, that is, the matching data set filtered by the clear sky threshold.

8. The method for inverting the total amount of clear-sky atmospheric precipitable water using FY-4B / AGRI data according to claim 1, characterized in that: The specific process of step S5 is as follows: First, according to the longitude and latitude, 70% of the matching data set that has been screened by the clear sky threshold is randomly selected as the training set and 30% as the test set. The training set is used as the original data set for the establishment of the random forest model. Then, use Bootstrap to randomly select m samples from the original training set with replacement sampling, perform n sampling times in total, generate n training sets, and train n decision tree models as random forest models; Finally, the data of the test set were input as independent variables into the established random forest model, and the TPW data calculated by the model were output; The output TPW data were matched with the TPW data in the ERA5 water vapor reanalysis product TCWV data, and the accuracy indicators of the two were calculated to evaluate the accuracy of FY-4B / AGRI's TPW inversion based on the random forest algorithm. The data input as independent variables into the established random forest model included the brightness temperature data of the 9th to 11th bands, the brightness temperature data of the 12th to 15th bands, the brightness temperature difference between the 13th and 14th bands, the secant value of the satellite observation zenith angle, TPW data, longitude, latitude and time.

9. A system for inverting the total amount of precipitable water in the clear sky atmosphere using FY-4B / AGRI data, characterized in that: It includes the following parts: The brightness temperature data processing module obtains the full disk data of FY-4B / AGRI in the area to be studied in the preset period, reads the pixel brightness value of the specified band and converts it into brightness temperature data to obtain a brightness temperature data set, which contains the latitude and longitude, time and brightness temperature information of each pixel; The TPW data processing module obtains the TCWV data of the water vapor reanalysis product corresponding to the study area and time period, obtains the TPW data of each grid point, and calculates the secant value of the satellite observation zenith angle of each grid point for angle correction to obtain the TPW data set. The TPW data set contains the latitude and longitude, time, TPW value and secant value of the satellite observation zenith angle of each grid point; A brightness temperature and TPW data matching module performs spatiotemporal matching on the brightness temperature dataset and the TPW dataset. During the spatiotemporal matching, the data in the brightness temperature dataset is read based on longitude and latitude, and each grid point of the TPW dataset is traversed to search for the pixel closest to each grid point in the brightness temperature dataset at the same time, and the TPW dataset corresponding to each grid point is merged with the brightness temperature dataset of the corresponding pixel to obtain a matching dataset of brightness temperature and TPW; The clear sky discrimination module uses the cloud detection product CLM corresponding to the study area and time period to discriminate the clear sky, and screens out the data in the matching data set of brightness temperature and TPW in step S3 that are in a cloudy state, and obtains the matching data set screened by the clear sky threshold; The TPW inversion module uses the matching dataset screened by the clear sky threshold to model and invert TPW based on the random forest algorithm.

10. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 8.

Citation Information

Cited By

  • Atmospheric precipitable water quantity inversion method and system based on multi-source satellite remote sensing data

    CN121564572A