High-temporal-spatial-resolution three-dimensional water vapor assimilation fusion method based on multi-source data

Abstract

The invention belongs to the field of meteorological science and technology, and relates to a high-temporal-spatial-resolution three-dimensional water vapor assimilation fusion method based on multi-source data, which is realized based on an LAPS mode and comprises the following steps: acquiring multi-source data of a research area and processing the multi-source data; the multi-source data comprises reanalysis data, ground meteorological data, high-altitude meteorological data, foundation GNSS water vapor data and FY-4 satellite AGRI image data; performing assimilation fusion on the processed FY-4 satellite AGRI image data and water vapor products thereof, ground meteorological data, high-altitude meteorological data and foundation GNSS water vapor data by adopting a multi-grid variation assimilation method and taking reanalysis data as a background field through an LAPS mode; according to different data combinations of high-altitude meteorological data and FY-4 satellite AGRI image data, multiple schemes of three-dimensional water vapor assimilation fusion are formed based on assimilation fusion of multi-source data, an optimal scheme is established, and three-dimensional water vapor data with high temporal-spatial resolution are generated.

Description

Technical Field

[0001] The present invention relates to the field of meteorological science and technology, and in particular to a high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data. Background Art

[0002] Water vapor is one of the most active components in the atmosphere. Although it accounts for a very small proportion of the atmosphere (only 0.1%-3%), its spatial distribution is extremely uneven and changes rapidly over time. It is a necessary condition for the formation of clouds and precipitation, and plays an extremely important role in weather evolution. At the same time, water vapor is also a greenhouse gas. Through atmospheric advection and vertical transport, evaporation and condensation, it affects atmospheric radiation and solar radiation, and further affects the water cycle and energy balance between the ground and the atmosphere. Therefore, water vapor is an important object for weather forecasting, climate change and even global change monitoring.

[0003] Currently, there are three main sources of water vapor data: model reanalysis data; ground-based observations, such as water vapor data obtained from sounding instruments, GNSS receivers, sun photometers, microwave radiometers, and lidar. Chinese patents with publication numbers CN107843943A, "A three-dimensional water vapor detection method based on function basis" and CN111881581A, "A method and system for establishing a three-dimensional water vapor grid model", have disclosed methods for detecting three-dimensional water vapor with ground-based GNSS receivers and air-based satellite remote sensing inversion, such as obtaining water vapor data through microwave remote sensing, thermal infrared remote sensing, near-infrared remote sensing, and GNSS occultation. Among these water vapor data, each has its own advantages and disadvantages. Model reanalysis data is globally timed, with acceptable vertical resolution, but low horizontal resolution and accuracy. Ground-based water vapor data are accurate, and except for the sounding instrument, which only has two observations at 0 and 12 o'clock universal time every day, other instrument data have a high observation frequency, but the number of observation sites is relatively small and unevenly distributed, making it impossible to analyze the spatial distribution characteristics of atmospheric water vapor. Polar-orbiting satellite remote sensing inversion obtains the total atmospheric water vapor, with a high horizontal spatial resolution, such as the Chinese invention "A Lan" with publication number CN105300880A. The spatial resolution of water vapor inverted by Landsat8 satellite in the "Inversion Method of Atmospheric Water Vapor Content from dsat8 Thermal Infrared Data" is as high as 100 meters, but the temporal resolution is slightly lower, and most polar-orbiting satellites only have 1-2 water vapor data per day; while the temporal resolution of water vapor inverted by geostationary satellites is high, but usually only low-layer, middle-layer, high-layer and total water vapor data, and the vertical resolution is low, such as the water vapor products mentioned in the Chinese invention patent "A Method for Producing Total Water Vapor and Layered Water Vapor Products from Geostationary Meteorological Satellites" with publication number CN115525854A. Therefore, the differences in temporal, horizontal and vertical resolutions of different water vapor data lead to limited application of atmospheric water vapor data in meteorology. The current satellite remote sensing technology mainly fuses satellite images of different resolutions at the pixel level, and is unable to fuse data from different sources, time dimensions and vertical space dimensions. An effective three-dimensional water vapor fusion method is urgently needed. Summary of the invention

[0004] The prior art has the problem of multi-source heterogeneity and inconsistent performance of water vapor data. The purpose of the present invention is to overcome the above-mentioned defects of the prior art and propose a high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data.

[0005] In order to achieve the above object, the present invention proposes a high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data, which is implemented based on the LAPS mode and includes:

[0006] Step 1) Obtain and process multi-source data of the study area; the multi-source data include: reanalysis data, ground meteorological data, high-altitude meteorological data, ground-based GNSS water vapor data and FY-4 satellite AGRI image data;

[0007] Step 2) Using the multi-grid variational assimilation method, through the LAPS mode, with the reanalysis data as the background field, the processed FY-4 satellite AGRI image data and its water vapor products, surface meteorological data, high-altitude meteorological data and ground-based GNSS water vapor data are assimilated and fused;

[0008] Step 3) For different data combinations of high-altitude meteorological data and FY-4 satellite AGRI image data, based on the assimilation and fusion of multi-source data in step 2), multiple schemes of three-dimensional water vapor assimilation and fusion are formed, and the best scheme is determined from them to generate three-dimensional water vapor data with high temporal and spatial resolution.

[0009] Preferably, the ground meteorological data includes: ground station number, longitude, latitude, altitude, sea level pressure, surface air temperature, dew point temperature, maximum temperature, minimum temperature, wind speed and wind direction;

[0010] The high-altitude meteorological data include sounding data, GNSS occultation atmospheric temperature and humidity profile data and FY-4 satellite GIIRS atmospheric vertical detection products.

[0011] Preferably, the processing of step 1) includes:

[0012] Step 1-1) Perform data quality control on ground meteorological data, remove outliers in the data through standard deviation analysis, fill in missing values ​​through neighbor interpolation, and generate NetCDF format;

[0013] Step 1-2) For high-altitude meteorological data, quality control is performed on different levels of data, outliers are removed and missing values ​​are filled, and they are uniformly processed into NetCDF format so that each type of high-altitude data contains station number, latitude, longitude, altitude, atmospheric pressure, geopotential height, air temperature, dew point temperature, wind direction and wind speed;

[0014] Step 1-3) According to the ground-based GNSS water vapor data, the static delay ZHD and the zenith wet delay ZWD are obtained respectively, and the atmospheric water vapor content PWV is converted by the conversion coefficient Π;

[0015] Step 1-4) For the FY-4 satellite AGRI image data, select 1 visible light reflectance, 2 medium-wave infrared brightness temperature and 2 long-wave infrared brightness temperature data and generate NetCDF format.

[0016] Preferably, the steps 1-3) include:

[0017] The static delay ZHD is calculated using the Hopfield model as follows:

[0018] ZHD=77.6×10 -6 P s [40137+148.72(T s -273.16)-h s ] / (5T s )

[0019] Where P s is the atmospheric pressure at the observation station, in hPa, T s is the absolute temperature at the observation station, in K, h s is the elevation of the observation station, in m;

[0020] According to the zenith tropospheric total delay ZTD and the static delay ZHD, the zenith wet delay ZWD is obtained by the following formula:

[0021] ZWD=ZTD-ZHD

[0022] The conversion coefficient Π is calculated based on the surface temperature and historical meteorological data:

[0023]

[0024] In the formula, R v is the gas constant of water vapor, ρ w is the density of liquid water, k'2 and k3 are empirical constants, T m is the weighted average temperature of the atmosphere, in K, satisfying:

[0025]

[0026] In the formula, p v is the water vapor pressure, in hPa, estimated from the ground temperature T0: T m =70.2+0.72T0.

[0027] Preferably, the step 2) comprises:

[0028] Step 2-1) Regionalization of the study area: setting the map projection and spatial resolution of the study area, and producing basic geographic information data of the study area;

[0029] Step 2-2) Background field data fusion: In the LAPS mode, the Barnes objective analysis method is used for the reanalysis data. Bilinear interpolation is used in the horizontal direction and linear interpolation of the height coordinate is used in the vertical direction. The horizontal and vertical spaces are interpolated to the fine grid size when localized. Point-to-point linear interpolation of grid points is used in time to obtain the background field data in NetCDF format including the three-dimensional space fusion field and the ground two-dimensional fusion field;

[0030] Step 2-3) Fusion analysis of ground meteorological data: Using the processed ground meteorological data, in the LAPS mode, the discrete ground meteorological observation values ​​are interpolated to the localized fine grid points through the distance weight method of the Gaussian weight function, and the ground meteorological data fusion results including the ground pressure, sea level pressure, reference height pressure, ground air temperature, ground dew point temperature, ground wind, vorticity, divergence and potential temperature on the localized fine grid are obtained;

[0031] Step 2-4) 3D temperature simulation analysis: obtain the ground temperature field, pressure field data and background 3D space fusion field data obtained by the LAPS model ground analysis, use the Barnes objective analysis method to set reasonable influence radius and weight for various observation values, interpolate to the fine grid size when localized, and then use the atmospheric vertical profile temperature field in the high-altitude meteorological data to calculate the height field; adjust the boundary layer temperature according to the temperature field obtained by the ground analysis, and adjust the potential temperature at different atmospheric heights to ensure that the high-level potential temperature is greater than or equal to the low-level potential temperature, and obtain the 3D temperature field, 2D boundary layer height field and pressure field;

[0032] Step 2-5) 3D cloud analysis: Based on the fusion analysis of ground meteorological data and the background 3D space fusion field, the preliminary guess field of 3D cloud analysis is obtained, the vertical distribution of clouds, and the approximate horizontal analysis field are obtained; then, the cloud top height field is calculated by performing 3D temperature analysis on the satellite cloud top temperature field, and the cloud top height field is interpolated to the preliminary guess cloud of 3D cloud analysis, thereby determining the cloud top height and 3D cloud analysis field;

[0033] Step 2-6) Three-dimensional water vapor assimilation and fusion analysis: Use the LAPS model to calculate an initial three-dimensional water vapor analysis field using background field data; then use the multi-grid three-dimensional continuous variation technology to correct the initial water vapor field based on high-altitude data, satellite water vapor products, ground-based GNSS total water vapor, LAPS three-dimensional cloud cover analysis results, and FY-4 / AGRI visible-infrared channel data to obtain three elements: three-dimensional specific humidity, three-dimensional relative humidity, and total water vapor.

[0034] Preferably, the step 2-3) comprises:

[0035] In the LAPS model, the distance weight method is used to interpolate the data of discrete meteorological stations to obtain a relatively smooth initial field. The interpolated value of the meteorological observation station position in the initial field is recorded as Satisfy the following formula:

[0036]

[0037] Among them, B (xk,yk) is the observation value of the kth observation station around the interpolation station, x k ,yk is the coordinate of the kth observation station, is the weight function of the kth observation station, c is the filtering constant, r k is the distance between the observation sites, n is the total number of observation sites;

[0038] The interpolation error E of each observation is calculated according to the following formula: (xk,yk) : in, is the interpolated value of the kth meteorological observation station in the initial field; and the Gaussian distance weighted interpolation is performed on the error values ​​of each station, and the interpolated error field is added to the initial interpolated field to obtain the new grid point value

[0039]

[0040] in: is the modified weight function, and G is the filtering constant.

[0041] Preferably, in the multi-grid three-dimensional continuous variation technique of step 2-6), the objective function J on the nth grid is (n) for:

[0042]

[0043] In the formula, Y=Y obs -HX b , X=X a -X b , O is the observation error covariance matrix; X b is the background field vector; X a is the analytical field vector; Y obs is the observation field vector; H is the bilinear interpolation operator from the model grid to the observation point; X represents the correction vector relative to the model field vector, which is calculated by the multi-grid three-dimensional continuous variation technique; Y is the difference between the observation field and the model field; (n) represents the nth grid, and N is the grid multiplicity;

[0044]

[0045] Run the STMAS multigrid fusion module in LAPS mode and start the analysis from the coarsest grid. When n = 1, Y (1) is the deviation between the observed value and the corresponding value of the background field, X (n-1) For J (n-1) An approximate solution to J (n-1) After X (n-1) is interpolated to the nth layer of finer grids and Y is calculated (n) , X (n) By minimizing J (n) Come to seek solutions;

[0046] By changing the grid size, the observation data are analyzed in sequence from long wave to short wave, so as to extract various scale information in the observation. The final analysis result is the superposition of the analysis results of each heavy grid. The result of three-dimensional water vapor assimilation and fusion is three elements: three-dimensional specific humidity, three-dimensional relative humidity and total water vapor.

[0047] Preferably, the step 3) comprises:

[0048] Aiming at different data combinations of high-altitude meteorological data and FY-4 satellite AGRI image data, the 0 and 12 o'clock sounding data of the study area that were not assimilated were used as reference to evaluate and test the three-dimensional water vapor accuracy obtained by each assimilation and fusion scheme. By calculating the correlation coefficient R, root mean square error RMSE, mean absolute percentage error MAPE and mean deviation MB between the two, the accuracy of the specific humidity and relative humidity of each layer and the total water vapor obtained by each assimilation and fusion scheme was evaluated, and the optimal three-dimensional water vapor assimilation and fusion scheme was determined to generate three-dimensional water vapor data with high temporal and spatial resolution.

[0049] Compared with the prior art, the advantages of the present invention are:

[0050] The present invention, with the help of the STMAS fusion module in the LAPS mode of the local analysis and forecasting system, adopts a multi-grid variational assimilation method to reanalyze the data background field, based on the multi-source data of star-air-ground: ground meteorological data, high-altitude data (sounding data, GNSS occultation atmospheric profile products, FY-4 satellite GIIRS temperature and humidity profiles), ground-based GNSS water vapor, FY-4 satellite data and its water vapor products, invents a high temporal and spatial resolution three-dimensional water vapor assimilation and fusion method, which can realize three-dimensional water vapor data with a spatial resolution of 4 km horizontally and 25 hpa vertically on an hourly basis, solves the problems of multi-source heterogeneity and inconsistent performance of current water vapor data, provides high temporal and spatial resolution water vapor data for numerical weather forecasting and meteorological research, and can better serve meteorological applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is the technical process of the present invention based on the high temporal and spatial resolution three-dimensional water vapor assimilation fusion of multi-source data;

[0052] Figure 2 It is the spatial distribution map of three-dimensional specific humidity in the study area;

[0053] Figure 3 It is the three-dimensional spatial distribution map of relative humidity in the study area;

[0054] Figure 4 It is the spatial distribution map of total water vapor in the study area. DETAILED DESCRIPTION

[0055] Aiming at the current problems of multi-source heterogeneity and inconsistent performance of water vapor data, in order to better introduce water vapor data into numerical weather forecasting to improve forecast accuracy, the present invention will use the STMAS fusion module in the LAPS (Local Analysis Prediction System) mode of the regional analysis and forecasting system, adopt a multi-grid variational assimilation method, and use reanalysis data as the background field to assimilate and fuse Fengyun-4 (FY-4) satellite images and their water vapor products, ground meteorological data, high-altitude meteorological data and ground-based GNSS water vapor data, establish the best assimilation and fusion scheme for multi-source data, and generate high-temporal and spatial resolution three-dimensional water vapor data with a horizontal resolution of 4km and a vertical resolution of 25hPa hourly, so as to provide a high-resolution data set for numerical weather forecasting and climate research of the meteorological department.

[0056] Compared with other current mesoscale numerical models, such as MM5, WRF, RAMS and GRAPSE, LAPS has better portability, scalability, and high-efficiency calculation characteristics. It can assimilate as much observation data as possible and achieve kilometer-level resolution data that is closer to the actual meteorological situation. Using the STMAS fusion module in the LAPS mode, the assimilation and fusion of multi-source water vapor data can be achieved.

[0057] The present invention specifically includes the following 3 steps:

[0058] (1) Input of multi-source data

[0059] The multi-source data that need to be input in the present invention mainly include: reanalysis data NCEP / GFS, ground meteorological data including ground station number, longitude, latitude, altitude, sea level pressure, surface air temperature, dew point temperature, maximum temperature, minimum temperature, wind speed and wind direction, high-altitude meteorological data including sounding data, GNSS occultation atmospheric temperature and humidity profile data and FY-4 satellite GIIRS atmospheric vertical detection products, ground-based GNSS water vapor data, and FY-4 satellite AGRI image data.

[0060] (2) Processing of multi-source data

[0061] The present invention processes multi-source data mainly for ground meteorological data, high-altitude meteorological data, ground-based GNSS water vapor data and FY-4 satellite AGRI image data. The processing flow includes data quality control, processing of missing values ​​and outliers, and standardization of data formats.

[0062] ① Processing of ground meteorological data: Data quality control is performed on the ground meteorological data obtained from each meteorological station in the study area. Outliers in the data are eliminated through standard deviation analysis, missing values ​​are filled through neighbor interpolation, and finally the ground meteorological data is generated into a NetCDF format that can be processed by the LAPS model. NetCDF (Network Common Data Format) is a file format used to store multidimensional scientific data (variables) such as temperature, humidity, air pressure, wind speed and wind direction.

[0063] ② Processing of high-altitude meteorological data: The high-altitude meteorological data mainly include sounding data, GNSS occultation atmospheric temperature and humidity profile data, and FY-4 / GIIRS atmospheric vertical detection products. Since the format, data set elements, and number of atmospheric profile layers of each type of data are different, quality control is performed on the different levels of each type of high-altitude data, outliers are removed, and missing values ​​are filled. Finally, they are uniformly processed into the standard format NetCDF that can be read by the LAPS model, so that each type of high-altitude data contains the station number, latitude, longitude, altitude, atmospheric pressure, geopotential height, temperature, dew point temperature, wind direction and wind speed, which become the input parameters for the three-dimensional temperature, three-dimensional cloud and three-dimensional water vapor simulation analysis in the LAPS model.

[0064] ③Processing of ground-based GNSS water vapor data

[0065] When the radio signals transmitted by GNSS satellites pass through the atmosphere, they are affected by the refraction of the ionosphere and troposphere, resulting in a delay relative to the straight-line propagation in a vacuum. The atmospheric delay can be divided into ionospheric delay and tropospheric delay. The ionospheric delay can be eliminated by using dual-frequency receiver observations. After eliminating the ground-based GNSS receiver clock error and satellite clock error by using relative differential processing, the tropospheric delay and positioning parameters can be accurately calculated using precise ephemeris and ground-based GNSS receiver observation data. The zenith tropospheric total delay (ZTD) consists of static delay (ZHD) and wet delay (ZWD). ZHD is relatively regular and can be calculated from ground meteorological data according to a certain model. For example, the Hopfield model calculates ZHD as:

[0066] ZHD=77.6×10 -6 P s [40137+148.72(T s -273.16)-h s ] / (5T s )

[0067] Where: P s is the atmospheric pressure at the observation station (hPa), T s is the absolute temperature at the observation station (K), h s is the elevation of the observation station (m).

[0068] The zenith wet delay ZWD can be calculated by subtracting the zenith static delay ZHD from the zenith tropospheric total delay ZTD, ZWD = ZTD-ZHD, and converted to the atmospheric water vapor content PWV = Π·ZWD through a conversion coefficient Π.

[0069] The conversion factor Π can be calculated using the surface air temperature and historical meteorological data:

[0070]

[0071] Where: R v is the gas constant of water vapor, which is 461.524 J / kg·K; ρ w is the density of liquid water, which is 1000 kg / m 3 , k'2 and k3 are empirical constants, which are 3.8×10 5 K / hpa and 16.52K / hpa. T m is the weighted average temperature of the atmosphere (K), calculated by the following formula:

[0072]

[0073] Where: p v is the water vapor pressure (hPa), integrated along the vertical path of the entire atmosphere. m It is necessary to know the distribution of the integral path T, which is difficult to operate in practice. Therefore, it is usually estimated by the ground temperature T0: T m =70.2+0.72T0.

[0074] ④ Processing of FY-4 satellite image data

[0075] The 14 channels of the Fengyun-4 satellite Advanced Geostationary Radiometric Imager (AGRI) cover visible-near infrared, short-wave infrared, medium-wave infrared and long-wave infrared, and can be used for remote sensing of clouds / water vapor (Table 1). The LAPS mode can process satellite data of 1 visible light, 2 medium-wave infrared and 2 long-wave infrared, and perform assimilation and fusion analysis of 3D clouds and water vapor. After geometric correction, radiation correction and uniform resampling of the spatial resolution of each AGRI channel to 4 km, 1 visible light reflectance, 2 medium-wave infrared brightness temperature and 2 long-wave infrared brightness temperature data are selected to generate the NetCDF format that the LAPS mode can process. By constructing different combinations (Table 2), continuous experiments are conducted in the simulation analysis of 3D clouds and 3D water vapor to determine the best combination.

[0076] Table 1 Performance parameters of the Fengyun-4 satellite Advanced Geostationary Radiometric Imager (AGRI)

[0077]

[0078]

[0079] Table 2 FY-4 / AGRI band combination scheme

[0080]

[0081] (3) Three-dimensional water vapor assimilation and fusion based on multi-source data

[0082] Three-dimensional water vapor assimilation and fusion includes seven processes: regionalization of the study area, background field data fusion, ground meteorological data fusion analysis, three-dimensional temperature analysis, three-dimensional cloud analysis, water vapor analysis and accuracy verification of three-dimensional water vapor data.

[0083] ① Regionalization of the study area: Set the map projection and spatial resolution of the study area, and produce basic geographic information data of the study area.

[0084] ② Background field data fusion: When water vapor assimilation is performed, a first guess field, namely the background field, is required. Using large-scale, low-resolution reanalysis data as background field data, it is necessary to use the Barnes objective analysis method, adopt bilinear interpolation in the horizontal direction, and linear interpolation of height coordinates in the vertical direction, and interpolate the horizontal and vertical space to the fine grid size when localized. Point-to-point linear interpolation of grid points is used in time, that is, the reanalysis data of the previous and next time periods are used to obtain the grid point values ​​at the required time between them through time linear interpolation. The result of background field data fusion will generate a three-dimensional space fusion field and a ground two-dimensional fusion field in NetCDF format.

[0085] ③ Fusion analysis of ground meteorological data: Using the ground meteorological data after quality control, the Barnes objective analysis method is used in the LAPS mode, that is, the discrete ground meteorological observation values ​​are interpolated to the fine grid points during localization through the distance weight method of the Gaussian weight function. It is specifically divided into two steps: First, the data of the discrete meteorological stations are interpolated to a relatively smooth initial field using the distance weight method, and the interpolated value of the meteorological observation station position in the initial field is recorded as The interpolation method is expressed as:

[0086]

[0087] Among them, B (xk,yk) is the kth observation value around the interpolation station, x k ,y k are the observation station coordinates, is the weight function, c is the filter constant, r k The distance between the observation sites. Then the interpolated value Compared with the observed value B (xk,yk) Subtract and calculate the interpolation error of each observation The error values ​​of each station are interpolated with Gaussian distance weights, and the interpolated error field is added to the initial interpolation field to obtain the new grid point value.

[0088]

[0089] in: is the modified weight function, and G is the filtering constant.

[0090] The results of the fusion analysis of ground meteorological data will produce physical quantities such as surface pressure, sea level pressure, reference height pressure, surface air temperature, surface dew point temperature, surface wind, vorticity, divergence, potential temperature, etc. on a localized time-fine grid, providing input parameters for the three-dimensional temperature, three-dimensional cloud and three-dimensional water vapor simulation analysis in the LAPS model.

[0091] ④ Three-dimensional temperature simulation analysis

[0092] The three-dimensional temperature analysis algorithm is mainly carried out in the following three steps:

[0093] The first step is to obtain the ground temperature field, pressure field data and background three-dimensional spatial fusion field data obtained by the LAPS model ground analysis. The Barnes objective analysis method is used to set reasonable influence radius and weight for various observation values, and interpolate them to the fine grid size during localization.

[0094] The second step is to calculate the height field based on the static relationship using the vertical profile temperature field of the atmosphere in the high-altitude data; then adjust the boundary layer temperature based on the temperature field obtained from the ground analysis, as well as the potential temperature at different atmospheric heights to ensure that the high-level potential temperature is greater than or equal to the low-level potential temperature.

[0095] The third step is to calculate the pressure and height of the top of the boundary layer.

[0096] The results of the three-dimensional temperature simulation analysis will obtain a three-dimensional temperature field, a two-dimensional boundary layer height field and an air pressure field, which will provide input parameters for the three-dimensional temperature, three-dimensional cloud and three-dimensional water vapor simulation analysis in the LAPS model.

[0097] ⑤3D cloud analysis

[0098] For the simulation analysis of three-dimensional clouds, firstly, on the basis of the fusion analysis of ground meteorological data and the background three-dimensional space fusion field, the preliminary guess field of three-dimensional cloud analysis is obtained, and the vertical distribution of clouds and the approximate horizontal analysis field are given; secondly, the cloud top height field is calculated by performing three-dimensional temperature analysis on the satellite cloud top temperature field, and then the cloud top height field is interpolated to the preliminary guess cloud analysis field to determine the cloud top height and the three-dimensional cloud analysis field. Due to the lack of ground-based radar data in the study area, the present invention does not perform radar data correction on three-dimensional cloud amount, and is mainly divided into the following steps:

[0099] In the first step, based on the initial cloud amount and cloud base information provided by conventional observations, a cloud thickness is assumed to obtain the cloud top height and cloud profile. It is usually assumed that the cloud thickness is 1000m, with a buffer zone of 500m above and below. Then, using the Barnes objective analysis method, the weight is the distance function from the grid to the observation point, and the original observations with sparse data are interpolated to the fine grid during localization to produce a continuous cloud field. At this time, the rationality of the cloud existence is judged based on the environmental stability and corrections are made. In this way, the three-dimensional cloud structure of the first stage can be obtained.

[0100] The second step is to use the brightness temperature observed by the long-wave infrared channel of the FY-4 satellite to make adjustments. First, based on the three-dimensional temperature analysis and the cloud field obtained by Barnes interpolation, an expected brightness temperature is calculated; then, based on the difference between the expected brightness temperature and the actual brightness temperature observed by the long-wave infrared channel of the FY-4 satellite, the three-dimensional cloud field obtained in the first stage is adjusted. If the brightness temperature observed by the satellite is lower than the expected brightness temperature, the cloud top height and satellite cloud amount are calculated using the satellite channel data, and then the satellite calculation results are compared with the cloud bottom height and cloud top height calculated by ground analysis, and the three-dimensional cloud amount is corrected. The area of ​​high clouds is based on the data observed by satellite, while the area of ​​low clouds mainly depends on the data of high altitude and ground observations. Finally, a brightness temperature value is calculated using the analyzed cloud field, which is compared with the brightness temperature value of the long-wave infrared channel of the FY-4 satellite, and the cloud amount is further adjusted to further balance the two.

[0101] The third step is to use the visible light data from the FY-4 satellite to correct the cloud cover. Since the visible light satellite data has a high spatial resolution, it can enhance the description of low clouds and see thin and sparse clouds. Therefore, the visible light data is interpolated onto a localized grid, and then the cloud brightness value is normalized and corrected according to the sunlight and satellite observation geometry to obtain cloud cover data with uniform brightness.

[0102] ⑥ Three-dimensional water vapor assimilation fusion analysis

[0103] The main process of three-dimensional water vapor assimilation and fusion is to first calculate an initial three-dimensional water vapor analysis field using the model background field data; then use high-altitude data, satellite water vapor products, ground-based GNSS total water vapor, LAPS three-dimensional cloud analysis results, and FY-4 / AGRI visible-infrared channel data to correct the initial water vapor field.

[0104] Since the background error covariance matrix B plays a very important role in the assimilation effect in the assimilation system, it is usually described by a Gaussian function, and the e-fold scale is constructed by the related scales in the longitude and latitude. However, the related scales may vary from region to region and are difficult to estimate. For a given observation system, areas with sparse observations can only provide long-wave information of errors, while areas with dense data can provide both long-wave information of errors and short-wave information of errors. An ideal data assimilation system hopes to obtain an accurate background field error covariance matrix that varies with space and time, so that the long-wave error information of the entire analysis area and the short-wave error information of the data-intensive area can be corrected, but in fact it is unlikely. The present invention introduces a three-dimensional continuous variation technology of multiple grids to perform three-dimensional water vapor assimilation and fusion of multi-source data, controls the related scale of variation by the number of grid points, and adjusts the influence range of observation data by changing the thickness of the grid, thereby playing the role of adjusting the background field error covariance matrix, eliminating the confusion of long and short error information between different scales, improving the results of assimilation analysis, and reducing the demand for computing resources.

[0105] In the multigrid three-dimensional continuous variation technique, the objective function on each grid takes the following form:

[0106]

[0107] In the formula: Y=Y obs -HX b , X=X a -X b , O is the observation error covariance matrix; X b is the background field vector; X a is the analytical field vector; Y obs is the observation field vector; H is the bilinear interpolation operator from the model grid to the observation point; X represents the correction vector relative to the model field vector, calculated by the variational assimilation system; Y is the difference between the observation field and the model field; n is the nth grid, and N represents the grid multiplicity.

[0108]

[0109] Run the STMAS multigrid fusion module in LAPS mode and start the analysis from the coarsest grid. When n = 1, Y (1) is the deviation between the observed value and the corresponding value of the background field, X (n-1) For J (n-1) An approximate solution to J (n-1) After X (n-1) is interpolated to the nth layer of finer grids and Y is calculated (n) , X (n) By minimizing J (n)By changing the grid size, the observation data are analyzed from long wave to short wave in turn, so as to achieve the purpose of extracting information of various scales in the observation. In this multi-scale analysis process, the observation error covariance matrix O of all scales (n) It is the same as the complete observation data, because it reflects the observation error covariance and is independent of the multi-grid layer. The final analysis result is the superposition of the analysis results of each multi-grid:

[0110]

[0111] The results of three-dimensional water vapor assimilation and fusion are three elements: three-dimensional specific humidity, three-dimensional relative humidity and total water vapor.

[0112] ⑦ Accuracy test of three-dimensional water vapor data

[0113] When using the LAPS mode to perform three-dimensional water vapor assimilation and fusion of multi-source data, there are three types of high-altitude data: sounding data, GNSS occultation atmospheric temperature and humidity profile data, and FY-4 / GIIRS atmospheric vertical detection products. There are multiple combinations of FY-4 / AGRI's visible light, medium-wave infrared, and long-wave infrared data. Therefore, multiple schemes for three-dimensional water vapor assimilation and fusion are formed.

[0114] The 0 and 12 o'clock sounding data of the study area that are not involved in assimilation are used as reference to evaluate and test the accuracy of the three-dimensional water vapor obtained by each assimilation and fusion scheme. By calculating the correlation coefficient R, root mean square error RMSE, mean absolute percentage error MAPE, and mean deviation MB between the two, the accuracy of the specific humidity and relative humidity of each layer and the total water vapor obtained by each assimilation and fusion scheme is evaluated to determine the best three-dimensional water vapor assimilation and fusion scheme.

[0115] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0116] Example

[0117] The embodiment of the present invention provides a three-dimensional water vapor assimilation fusion method with high temporal and spatial resolution based on multi-source data, which specifically includes:

[0118] (1) Regionalization of the study area

[0119] Taking the Qinghai-Tibet Plateau as an example, as the study area, the latitude 25.0 ~ 47.0 °, longitude 76.5 ~ 105.5 ° range is regionalized, the map projection mode is set to Lambert, the spatial resolution is 4000m, and the input files such as surface, topography, ground temperature, and albedo are imported. The localization script file is run in the LAPS mode. After the study area is regionalized, the information including the scope of the study area, surface utilization type, surface albedo, topography, soil type, soil temperature, slope, etc. is generated, which lays the foundation for other data fusion and simulation analysis in the LAPS mode.

[0120] (2) Preprocessing of multi-source data in the study area

[0121] The multi-source data of the study area were preprocessed, mainly for ground meteorological data, high-altitude data, ground-based GNSS water vapor data and FY-4 satellite AGRI image data. The processing flow included data quality control, processing of missing values ​​and outliers, standardization of data formats, and finally the NetCDF format that can be processed by the LAPS mode.

[0122] (3) Background field data fusion

[0123] The US NCEP / GFS forecast data with a spatial resolution of 0.25° and a time interval of 6 hours was used as the background field data. The background field fusion program was run in the LAPS mode to complete the fusion of the background field data in the study area, and generate hourly three-dimensional data fusion fields and ground data fusion fields in NetCDF format. The three-dimensional data fusion field contains 43 layers of background field height, temperature, specific humidity, u-wind speed, v-wind speed, and m-wind speed data; the ground data fusion field contains the background field ground u-wind speed, v-wind speed, temperature, dew point temperature, sea level pressure, ground pressure, specific humidity, and precipitation data.

[0124] (4) Analysis of surface meteorological data

[0125] Using ground meteorological data, the ground analysis program is run in the LAPS mode to correct the ground data lgb fused with the background field. The generated surface analysis results include uv wind speed, surface temperature, surface dew point, relative humidity, sea level pressure, temperature advection, potential temperature, surface pressure, vorticity, mixing ratio, convergence, divergence, visibility and other data, which can be used for other analyses in the LAPS mode.

[0126] (5) Simulation analysis of three-dimensional temperature

[0127] When performing three-dimensional temperature analysis in the LAPS mode, the input parameters mainly include: background field fusion data, surface temperature data for surface analysis, and high-altitude temperature, humidity, and pressure atmospheric profile data. Here, the three-dimensional temperature analysis program is run for three types of high-altitude data (sounding data, GNSS occultation atmospheric temperature and humidity profile data, and FY-4 / GIIRS atmospheric vertical detection products), and the results generate a three-dimensional temperature grid file lt1, a two-dimensional boundary layer thickness file pbl, and a temperature observation file tmg for internal drawing. Compare the results of the simulation analysis of the three types of high-altitude data, and select the appropriate results as the input parameters for three-dimensional cloud and water vapor analysis.

[0128] (6) Simulation analysis of three-dimensional clouds

[0129] When performing three-dimensional cloud analysis in the LAPS mode, the input parameters mainly include: the results of background field fusion lga and lgb, the results of ground meteorological observation data preprocessing lso, the results of three-dimensional temperature analysis, the results of FY-4 satellite original image preprocessing lvd, including the reflectivity data of visible-near infrared channels, and the brightness temperature data of medium-wave infrared and long-wave infrared channels. The three-dimensional cloud simulation analysis can obtain the three-dimensional cloud amount field lc3, the two-dimensional maximum cloud / snow cover product lcv, and the cloud base and cloud top height products lcb. Among them, the three-dimensional cloud amount field lc3 is the input parameter of the three-dimensional water vapor assimilation fusion analysis. Different three-dimensional temperature analysis results are obtained from different high-altitude data, and different three-dimensional cloud amount fields can be generated in combination with different FY-4 satellite band combinations.

[0130] (7) Assimilation and fusion analysis of three-dimensional water vapor

[0131] The assimilation and fusion analysis of three-dimensional water vapor is to assimilate the background field fusion data, ground meteorological data, high-altitude meteorological data, FY-4 satellite data and its water vapor products, ground-based GNSS water vapor data, and three-dimensional cloud field in the LAPS model using three-dimensional grid continuous variation to obtain three-dimensional water vapor fields, including three data sets: three-dimensional specific humidity lq3, three-dimensional relative humidity field lh3, and atmospheric total precipitable water field lh4. In the model assimilation, there are three types of high-altitude meteorological data, multiple three-dimensional cloud field data, and different FY-4 satellite data combinations, which will form different water vapor assimilation and fusion schemes. Taking the data at 0 and 12 o'clock universal time as an example, using some sounding data that are not involved in assimilation as reference data, the results of different water vapor assimilation and fusion schemes are tested for accuracy. The correlation coefficient R, root mean square error RMSE, mean absolute percentage error MAPE, and mean deviation MB are selected as evaluation indicators to evaluate the accuracy of each layer's specific humidity and relative humidity, as well as the total water vapor obtained by each assimilation and fusion scheme, and determine the best three-dimensional water vapor assimilation and fusion scheme. Using the best assimilation and fusion scheme, the three-dimensional water vapor spatial distribution map of the Qinghai-Tibet Plateau at 12:00 on January 4, 2016 was generated, as shown in the figure below: Figure 2 Shown is the spatial distribution of total water vapor in the study area; Figure 3 This is the specific humidity distribution diagram of the typical pressure layer in the study area; Figure 4 This is the relative humidity distribution diagram of the typical pressure layer in the study area.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data, implemented based on the LAPS model, including: Step 1) Obtain multi-source data of the study area and process it; The multi-source data include: reanalysis data, surface meteorological data, high-altitude meteorological data, ground-based GNSS water vapor data and FY-4 satellite AGRI image data; Step 2) Using the multi-grid variational assimilation method, through the LAPS mode, with the reanalysis data as the background field, the processed FY-4 satellite AGRI image data and its water vapor products, surface meteorological data, high-altitude meteorological data and ground-based GNSS water vapor data are assimilated and fused; Step 3) For different data combinations of high-altitude meteorological data and FY-4 satellite AGRI image data, based on the assimilation and fusion of multi-source data in step 2), multiple schemes of three-dimensional water vapor assimilation and fusion are formed, and the best scheme is determined from them to generate three-dimensional water vapor data with high temporal and spatial resolution.

2. The high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data according to claim 1 is characterized in that: The ground meteorological data include: ground station number, longitude, latitude, altitude, sea level pressure, surface air temperature, dew point temperature, maximum temperature, minimum temperature, wind speed and wind direction; The high-altitude meteorological data include sounding data, GNSS occultation atmospheric temperature and humidity profile data and FY-4 satellite GIIRS atmospheric vertical detection products.

3. The high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data according to claim 2 is characterized in that: The processing of step 1) includes: Step 1-1) Perform data quality control on ground meteorological data, remove outliers in the data through standard deviation analysis, fill in missing values ​​through neighbor interpolation, and generate NetCDF format; Step 1-2) For high-altitude meteorological data, quality control is performed on different levels of data, outliers are removed and missing values ​​are filled, and they are uniformly processed into NetCDF format so that each type of high-altitude data contains station number, latitude, longitude, altitude, atmospheric pressure, geopotential height, air temperature, dew point temperature, wind direction and wind speed; Step 1-3) According to the ground-based GNSS water vapor data, the static delay ZHD and the zenith wet delay ZWD are obtained respectively, and the atmospheric water vapor content PWV is converted by the conversion coefficient Π; Step 1-4) For the FY-4 satellite AGRI image data, select 1 visible light reflectance, 2 medium-wave infrared brightness temperature and 2 long-wave infrared brightness temperature data and generate NetCDF format.

4. The high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data according to claim 1 is characterized in that: The steps 1-3) include: The static delay ZHD is calculated using the Hopfield model as follows: ZHD=77.6×10 -6 P s [40137+148.72(T s -273.16)-h s ] / (5T s ) Where P s is the atmospheric pressure at the observation station, in hPa, T s is the absolute temperature at the observation station, in K, h s is the elevation of the observation station, in m; According to the zenith tropospheric total delay ZTD and the static delay ZHD, the zenith wet delay ZWD is obtained by the following formula: ZWD=ZTD-ZHD The conversion coefficient Π is calculated based on the surface temperature and historical meteorological data: In the formula, R v is the gas constant of water vapor, ρ w is the density of liquid water, k'2 and k3 are empirical constants, T m is the weighted average temperature of the atmosphere, in K, satisfying: In the formula, p v is the water vapor pressure, in hPa, estimated from the ground temperature T0: T m =70.2+0.72T0.

5. The high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data according to claim 1 is characterized in that: The step 2) comprises: Step 2-1) Regionalization of the study area: setting the map projection and spatial resolution of the study area, and producing basic geographic information data of the study area; Step 2-2) Background field data fusion: In the LAPS mode, the Barnes objective analysis method is used for the reanalysis data. Bilinear interpolation is used in the horizontal direction and linear interpolation of the height coordinate is used in the vertical direction. The horizontal and vertical spaces are interpolated to the fine grid size when localized. Point-to-point linear interpolation of grid points is used in time to obtain the background field data in NetCDF format including the three-dimensional space fusion field and the ground two-dimensional fusion field; Step 2-3) Fusion analysis of ground meteorological data: Using the processed ground meteorological data, in the LAPS mode, the discrete ground meteorological observation values ​​are interpolated to the localized fine grid points through the distance weight method of the Gaussian weight function, and the ground meteorological data fusion results including the ground pressure, sea level pressure, reference height pressure, ground air temperature, ground dew point temperature, ground wind, vorticity, divergence and potential temperature on the localized fine grid are obtained; Step 2-4) 3D temperature simulation analysis: obtain the ground temperature field, pressure field data and background 3D space fusion field data obtained by the LAPS model ground analysis, use the Barnes objective analysis method to set reasonable influence radius and weight for various observation values, interpolate to the fine grid size when localized, and then use the atmospheric vertical profile temperature field in the high-altitude meteorological data to calculate the height field; adjust the boundary layer temperature according to the temperature field obtained by the ground analysis, and adjust the potential temperature at different atmospheric heights to ensure that the high-level potential temperature is greater than or equal to the low-level potential temperature, and obtain the 3D temperature field, 2D boundary layer height field and pressure field; Step 2-5) 3D cloud analysis: Based on the fusion analysis of ground meteorological data and the background 3D space fusion field, the preliminary guess field of 3D cloud analysis is obtained, the vertical distribution of clouds, and the approximate horizontal analysis field are obtained; then, the cloud top height field is calculated by performing 3D temperature analysis on the satellite cloud top temperature field, and the cloud top height field is interpolated to the preliminary guess cloud of 3D cloud analysis, thereby determining the cloud top height and 3D cloud analysis field; Step 2-6) Three-dimensional water vapor assimilation and fusion analysis: Use the LAPS model to calculate an initial three-dimensional water vapor analysis field using background field data; then use the multi-grid three-dimensional continuous variation technology to correct the initial water vapor field based on high-altitude data, satellite water vapor products, ground-based GNSS total water vapor, LAPS three-dimensional cloud cover analysis results, and FY-4 / AGRI visible-infrared channel data to obtain three elements: three-dimensional specific humidity, three-dimensional relative humidity, and total water vapor.

6. The high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data according to claim 5 is characterized in that: The steps 2-3) include: In the LAPS mode, the distance weight method is used to interpolate the data of discrete meteorological stations to obtain a relatively smooth initial field. The interpolated value of the meteorological observation station position in the initial field is recorded as Satisfy the following formula: Among them, B (xk,yk) is the observation value of the kth observation station around the interpolation station, x k ,y k is the coordinate of the kth observation station, is the weight function of the kth observation station, c is the filtering constant, r k is the distance between the observation sites, n is the total number of observation sites; The interpolation error E of each observation is calculated according to the following formula: (xk,yk) : in, is the interpolated value of the kth meteorological observation station in the initial field; and the Gaussian distance weighted interpolation is performed on the error values ​​of each station, and the interpolated error field is added to the initial interpolated field to obtain the new grid point value in: is the modified weight function, and G is the filtering constant.

7. The high temporal and spatial resolution three-dimensional water vapor assimilation fusion method based on multi-source data according to claim 5 is characterized in that: The multi-grid three-dimensional continuous variation technique in step 2-6) is used to calculate the objective function J on the nth grid. (n) for: In the formula, Y=Y obs -HX b , X=X a -X b , O is the observation error covariance matrix; X b is the background field vector; X a is the analytical field vector; Y obs is the observation field vector; H is the bilinear interpolation operator from the model grid to the observation point; X represents the correction vector relative to the model field vector, which is calculated by the multi-grid three-dimensional continuous variation technique; Y is the difference between the observation field and the model field; (n) represents the nth grid, and N is the grid multiplicity; Run the STMAS multigrid fusion module in LAPS mode and start the analysis from the coarsest grid. When n = 1, Y (1) is the deviation between the observed value and the corresponding value of the background field, X (n-1) For J (n-1) An approximate solution to J (n-1) After X (n-1) is interpolated to the nth layer of finer grids and Y is calculated (n) , X (n) By minimizing J (n) Come to seek solutions; By changing the grid size, the observation data are analyzed in sequence from long wave to short wave, so as to extract various scale information in the observation. The final analysis result is the superposition of the analysis results of each heavy grid. The result of three-dimensional water vapor assimilation and fusion is three elements: three-dimensional specific humidity, three-dimensional relative humidity and total water vapor.

8. The high temporal and spatial resolution three-dimensional water vapor assimilation and fusion method based on multi-source data according to claim 1 is characterized in that: The step 3) comprises: Aiming at different data combinations of high-altitude meteorological data and FY-4 satellite AGRI image data, the 0 and 12 o'clock sounding data of the study area that were not assimilated were used as reference to evaluate and test the three-dimensional water vapor accuracy obtained by each assimilation and fusion scheme. By calculating the correlation coefficient R, root mean square error RMSE, mean absolute percentage error MAPE and mean deviation MB between the two, the accuracy of the specific humidity and relative humidity of each layer and the total water vapor obtained by each assimilation and fusion scheme was evaluated, and the optimal three-dimensional water vapor assimilation and fusion scheme was determined to generate three-dimensional water vapor data with high temporal and spatial resolution.

Images