Atmospheric temperature and humidity profile retrieval method and device based on satellite hyperspectral data

Abstract

The application provides a kind of atmospheric temperature and humidity profile retrieval method and device based on satellite hyperspectral data, comprising: constructing background temperature and humidity profile using numerical mode historical forecast data, and calculating the background error between it and actual observation;Substitute the background profile into the radiation transfer model to simulate satellite radiation value, and obtain radiation simulation error according to satellite hyperspectral observation;Calculate the nonlinear distance correlation coefficient of each channel radiation simulation error and each pressure layer background temperature and humidity error, and use the first M channels with the highest correlation coefficient as the retrieval preselected channel of the layer;The background error and related data are divided into first and second data sets.Use the first data set, for each pressure layer, use the radial basis function neural network model of 1 to M channel number to establish the relationship between radiation error and background temperature and humidity error;Use the second data set to evaluate the retrieval error under each channel combination, select the optimal channel combination as the standard of minimum error, and thus obtain the final optimized neural network retrieval model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorological prediction, and particularly relates to an atmospheric temperature and humidity profile inversion method and device based on satellite hyperspectral data. BACKGROUND

[0002] The vertical profile of atmospheric temperature and humidity is a key parameter for climate monitoring, weather analysis and numerical weather prediction, and its accurate acquisition is crucial for improving prediction accuracy and understanding the evolution mechanism of weather systems. Traditionally, the atmospheric temperature and humidity profile is mainly obtained by radio sounding observation, but the distribution of sounding sites is sparse, the spatial coverage is uneven, and especially in the ocean, plateau and remote areas, the data is severely lacking, which is difficult to meet the modern meteorological business demand of high precision and high spatio-temporal resolution. With the development of satellite remote sensing technology, especially the emergence of hyperspectral infrared remote sensors, it provides a revolutionary means for obtaining global high-precision and continuous atmospheric temperature and humidity profile.

[0003] The hyperspectral infrared vertical sounder can obtain infrared radiation information of the earth-atmosphere system with extremely high spectral resolution, and the number of channels can reach thousands, far exceeding traditional sounders, which can capture more detailed atmospheric absorption and emission characteristics, and significantly improve the inversion ability of temperature and humidity profile. The spaceborne infrared hyperspectral vertical sounder receives the infrared radiation signal of the earth-atmosphere system, which needs to be converted into temperature and humidity profile by inversion algorithm. The early inversion methods mainly include statistical regression method and physical inversion method. The statistical regression method is based on historical observation data to establish the statistical relationship between radiation and atmospheric parameters, such as eigenvector method and ridge regression solution. This kind of method has fast calculation speed and strong stability, but ignores the physical process of radiation transfer, and the inversion accuracy is limited; the physical inversion method is to solve the radiation transfer equation and construct the merit function to seek the optimal solution, such as one-dimensional variational method. This kind of method has high inversion accuracy, but the calculation is complex and time-consuming, which causes serious pressure to real-time business.

[0004] In recent years, with the rapid development of artificial intelligence technology, machine learning algorithms have been gradually introduced into the field of atmospheric science. The BP (Back Propagation, back propagation) neural network adopts error back propagation mechanism, realizes nonlinear mapping through Sigmoid or Tanh activation function, has the advantages of simple structure and easy implementation, but the training speed is slow, easy to fall into local optimum, and sensitive to network initial weight; the convolutional neural network adopts multi-layer convolution and pooling structure to extract spectral features, uses convolution kernel to process channel sequence, has the advantages of strong automatic feature extraction ability and accurate capture of local spectral features, but the network structure is complex, the parameter tuning is difficult, and the training requires large amount of computing resources; the multi-objective genetic algorithm can optimize long-wave and medium-wave channels by constructing double-objective function, and introduce physical constraint conditions to invert temperature and humidity profile, has the advantages of clear physical meaning and independence from large amount of historical training data, but the calculation complexity is high and the convergence speed is slow. SUMMARY

[0005] In view of the technical problems existing in the prior art, the present application provides an atmospheric temperature and humidity profile inversion method and device based on satellite hyperspectral data, which quickly inverts multi-level atmospheric temperature and humidity information by using an inversion model constructed by a radial basis function (RBF) according to real-time information observed by a satellite.

[0006] The difficulty of the present application is:

[0007] 1) Optimal selection of hyperspectral channels. The number of channels of a hyperspectral infrared vertical sounder can reach thousands, for example, the FY4B-GIIRS in China has 1690 channels, and the sensitivity of each channel to temperature and water vapor is different at different pressure layers. Atmospheric radiation transmission is a nonlinear process, therefore, how to establish an effective channel selection standard according to the inversion requirements of the temperature and humidity profile, and determine the optimal channel combination, is the primary difficulty faced by the present method. Too many channels will introduce redundant information and noise, significantly increasing the computational complexity, which is not conducive to business application; and too few channels may lose key information, resulting in a decrease in inversion accuracy. Therefore, the best balance between inversion accuracy and computational efficiency must be sought, and the channel combination with the largest amount of information and the smallest redundancy must be selected.

[0008] 2) Determination of RBF network base function and model structure. The selection of the base function directly determines the approximation ability, generalization performance and computational efficiency of the network. For example, the Gaussian function has good local response characteristics and is suitable for capturing the local mode of the atmospheric state in the feature space; while the multi-quadratic function has globality, and the change of any data point may affect the entire model output. Therefore, the selection of the base function and the determination of the key hyperparameters such as the number of centers on this basis are the core design difficulties of constructing a high-performance inversion model.

[0009] 3) The physical design of the input variables is an important challenge faced by the present application. The atmospheric radiation transmission process under various weather conditions presents strong nonlinearity, which greatly increases the difficulty of pure data-driven learning algorithms, and easily leads to insufficient generalization ability and large errors of the model in actual application. Therefore, how to break through the limitations of traditional "end-to-end" mapping, integrate the physical prior knowledge of atmospheric radiation transmission (such as introducing the model background field, brightness temperature deviation and other physical quantities) into the design of the input variables, to construct input features with stronger constraints and more explicit physical meaning, so as to reduce the nonlinear complexity of the learning problem, is a key challenge to improve the inversion accuracy, robustness and physical consistency of the neural network.

[0010] The present application provides an atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data, comprising:

[0011] The temperature and humidity profile of the numerical model historical prediction is taken as a background profile, and the error between the background profile and the actual observation profile is taken as a background error;

[0012] The background profile is substituted into a radiation transmission model to obtain simulated radiation values of all high spectral channels of the satellite, and the error between the radiation values observed by the satellite and the simulated radiation values is taken as a radiation simulation error;

[0013] The nonlinear distance correlation coefficient between the radiation simulation error of each high spectral channel of the satellite and the background error on each pressure layer is calculated, and for each pressure layer, M channels are selected as preselected channels for temperature and humidity profile retrieval according to the correlation coefficient from large to small;

[0014] A part of the background error and the corresponding radiation simulation error are selected to construct a first data set, and another part of the background error and the corresponding background profile, actual observation profile and radiation simulation error are selected to construct a second data set;

[0015] A radial basis function neural network model is constructed using the first data set and the preselected M high spectral channels, and the radial basis function neural network model is optimized using the second data set;

[0016] The optimized radial basis function neural network model is used for atmospheric temperature and humidity profile retrieval to obtain the final temperature and humidity profile.

[0017] According to the atmospheric temperature and humidity profile retrieval method based on satellite high spectral data provided by the application, the method comprises:

[0018] The nonlinear distance correlation coefficient between the background error on each pressure layer in the data set and the radiation simulation error of each high spectral channel of the satellite is calculated;

[0019] For the background error on each pressure layer, the satellite channels are sorted according to the nonlinear distance correlation coefficient from large to small, and the first M channels are selected as preselected retrieval channels;

[0020] A radial basis function neural network model is constructed according to the background error in the first data set and the radiation simulation error of the preselected channel.

[0021] According to the atmospheric temperature and humidity profile retrieval method based on satellite high spectral data provided by the application, the nonlinear distance correlation coefficient between the background error on each pressure layer in the first data set and the radiation simulation error of all high spectral channels of the satellite is calculated by the following formula:

[0022]

[0023] Wherein, is a sequence of background errors a nonlinear distance correlation coefficient between the radiation simulation error sequence and the sequence A and B represent bicentral distance matrices.

[0024] The application provides a method for retrieving atmospheric temperature and humidity profiles based on satellite hyperspectral data, and the method comprises the following steps: constructing a radial basis function neural network model according to background errors in a first data set and selected radiation simulation errors by the following formula:

[0025]

[0026]

[0027] wherein, represents a vector formed by radiation simulation errors of m preselected channels, 1≤m≤M; k1 and k2 are vector identifiers, and ||.|| represents the Euclidean distance between two vectors; is an Euler radial basis function value, and D is a radial basis function matrix, is the first data set, is a vector formed by background errors of the j-th pressure layer in the first data set, is a coefficient of the radial basis function neural network model at the j-th pressure layer using m hyperspectral channels.

[0028] The application provides a method for retrieving atmospheric temperature and humidity profiles based on satellite hyperspectral data, and the method comprises the following steps: optimizing the radial basis function neural network model by using a second data set, and the optimization comprises the following steps:

[0029] retrieving temperature and humidity profiles of each pressure layer by using the radial basis function neural network model with different numbers of preselected channels according to the second data set, and calculating the root mean square error of retrieval errors;

[0030] for each pressure layer, selecting a satellite channel combination corresponding to the minimum value in the root mean square error as an optimal satellite channel combination for temperature (humidity) retrieval of the pressure layer.

[0031] The application provides a method for retrieving atmospheric temperature and humidity profiles based on satellite hyperspectral data, and the method comprises the following steps: retrieving temperature and humidity profiles of each pressure layer by using the radial basis function neural network model with different numbers of preselected channels according to the second data set by the following formula:

[0032]

[0033] wherein, is the first data set, ​​​is a second data set, H is a matrix radiation simulation error of the kth profile in the background error and radiation simulation error in the background error radial basis function between a row vector composed of K s is the number of profiles in the background error is a coefficient of a radial basis function neural network model using the radiation simulation error of the first m channels of the preselected channels as the basis to correspond to the background error of the jth pressure layer, is the kth background profile in the jth pressure layer temperature or humidity is the value after inversion.

[0034] The present application also provides an atmospheric temperature and humidity profile inversion device based on satellite hyperspectral data, comprising:

[0035] a background error calculation module, configured to calculate the error between the background temperature and humidity profile and the actual observation profile as the background error, with the temperature and humidity profile predicted by the numerical model as the background profile;

[0036] a radiation simulation error calculation module, configured to calculate the error between the corresponding satellite observed radiation value and the simulated radiation value of all satellite hyperspectral channels by substituting the background profile into the radiation transfer model to obtain the simulated radiation value of all satellite hyperspectral channels, so as to obtain the radiation simulation error;

[0037] a hyperspectral channel preselection module, configured to calculate the nonlinear distance correlation coefficient between the temperature and humidity background error of each pressure layer and the radiation simulation error of each hyperspectral channel, and sort the hyperspectral channels according to the correlation coefficient from large to small, and select the first M channels as the preselected channels.

[0038] a sample data set division module, configured to select a part of the background error and the corresponding radiation simulation error to construct a first data set, and another part of the background error and the corresponding background profile, the actual observation profile and the radiation simulation error to construct a second data set;

[0039] a construction and optimization module, configured to construct a radial basis function neural network model using the first data set, and optimize the radial basis function neural network model using the second data set;

[0040] an inversion module, configured to perform atmospheric temperature and humidity profile inversion using the optimized radial basis function neural network model to obtain the final temperature and humidity profile.

[0041] The application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for retrieving atmospheric temperature and humidity profile based on satellite hyperspectral data according to any one of the above.

[0042] The application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the method for retrieving atmospheric temperature and humidity profile based on satellite hyperspectral data according to any one of the above.

[0043] The application further provides a computer program product comprising a computer program, wherein the computer program is executable by a processor to implement the method for retrieving atmospheric temperature and humidity profile based on satellite hyperspectral data according to any one of the above.

[0044] The application provides a method and device for retrieving atmospheric temperature and humidity profile based on satellite hyperspectral data, which utilizes satellite hyperspectral data to retrieve atmospheric temperature and humidity profile based on radial basis function (RBF), and according to real-time information observed by a satellite, multi-level atmospheric temperature and humidity information is quickly retrieved according to a constructed retrieval model.

[0045] The application has the following advantages:

[0046] 1) The application has a significant advantage in retrieval efficiency. Traditional variational assimilation methods need to solve complex radiation transfer equations multiple times through iterative algorithms, which has a huge amount of calculation and is difficult to meet the high timeliness application demand. The application is based on a pre-trained neural network model, and the retrieval process only needs one forward propagation, which can achieve much faster retrieval speed than the variational method, and provides key technical support for real-time weather analysis and forecast.

[0047] 2) The application is more flexible and targeted in the retrieval strategy. The variational method seeks global optimal estimation of the overall atmospheric profile, which is difficult to optimize for specific needs. In contrast, the application allows optimization of the retrieval accuracy of all meteorological height layers during the training phase. In addition, through systematic stratified statistical verification of the retrieval results, the pressure layers that are significantly improved by the application relative to the numerical model background field can be clearly identified, so that users can more confidently adopt the higher quality retrieval information in these layers, thereby improving the practical value of the product.

[0048] 3) With the continuous improvement of the accuracy of numerical weather prediction, higher requirements are put forward for satellite remote sensing retrieval products: the retrieval error needs to be systematically better than the numerical model background field, so as to produce a positive effect in assimilation. The existing pure data-driven neural network method is constrained by the climatological statistics of the training set, and cannot fully utilize real-time and high-precision model prediction information.

[0049] The innovation of the present application is that a hybrid inversion framework based on RBF neural network is provided. The core of the framework is two unique designs: first, the temperature and humidity profile error provided by the numerical model and its simulated radiation error are taken together as network input; second, a statistical relationship model between the background field error and the simulated radiation error is explicitly established in the network structure. This design makes the network not only learn the mapping from the observed profile, but also focus on learning how to use the brightness temperature deviation to correct the background field deviation. This is equivalent to embedding the idea of variational assimilation into the neural network, so as to make the final inversion profile statistically superior to the initial background field, thereby significantly improving the value and reliability of the inversion product for model assimilation. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Figure 1 is a flowchart of the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data provided by the present application;

[0052] Figure 2 is a schematic diagram of the distribution of the number of channels of GIIRS used for inversion of temperature and water vapor mixing ratio at different pressure layers in the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data provided by the present application;

[0053] Figure 3 is a distribution of hyperspectral channels used for inversion of temperature at each pressure in part (a) and a distribution of hyperspectral channels used for inversion of humidity at each pressure in part (b) of the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data provided by the present application;

[0054] Figure 4 is a comparison of the root mean square error RMSE of the background profile and the temperature profile inverted by the RBF neural network in part (a) and a ratio of the root mean square error RMSE of the RBF neural network inverted temperature profile to the root mean square error RMSE of the background profile in part (b) of the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data provided by the present application;

[0055] Figure 5is the root mean square error RMSE of part (a) of the atmospheric temperature and humidity profile retrieval method based on satellite hyperspectral data Water vapor back mixing ratio background profile and RBF neural network retrieval water vapor profile respectively obtained and (b) Root mean square error RMSE of part RBF neural network retrieval water vapor profile Relative to the root mean square error RMSE of the background profile;

[0056] Figure 6 is the structure schematic diagram of the atmospheric temperature and humidity profile retrieval device based on satellite hyperspectral data provided by the application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme in the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0058] The following will be combined Figure 1 A method for retrieving atmospheric temperature and humidity profile based on satellite hyperspectral data is described, comprising:

[0059] Step 101, taking the temperature and humidity profile of the numerical model historical prediction as the background profile, calculating the error between the background profile and the actual observation profile as the background error;

[0060] Step 102, substituting the background profile into the radiation transfer model to obtain the simulated radiation value of all hyperspectral channels of the satellite, calculating the error between the satellite observed hyperspectral radiation value and the simulated radiation value as the radiation simulation error;

[0061] Step 103, calculating the nonlinear distance correlation coefficient of the temperature (humidity) background error and the radiation simulation error of each hyperspectral channel of the satellite on each pressure layer, selecting the first M channels according to the correlation coefficient from large to small as the inversion preselected channel of the temperature (humidity) of the pressure layer;

[0062] Step 104, constructing a first data set according to part of the background error and the corresponding radiation simulation error, and constructing a second data set by using another part of the background error and the corresponding background profile, actual observation profile and radiation simulation error;

[0063] Step 105, using the first data set, using the corresponding M inversion preselected channel combinations for each pressure layer, constructing a radial basis function neural network model between the radiation simulation error and the background temperature and humidity profile error of each inversion preselected channel combination;

[0064] Step 106, using the second data set to count the temperature and humidity inversion error of M different inversion pre-selected channel combinations on each air pressure layer, and selecting the inversion pre-selected channel combination with the smallest inversion error as the standard to select the best channel combination;

[0065] Step 107, using the selected best channel combination to perform atmospheric temperature and humidity profile inversion on the corresponding radial basis function neural network model to obtain the final temperature and humidity profile.

[0066] The relationship between the satellite observed radiation value and the vertical profile of the physical variable of the earth's atmosphere is defined as follows:

[0067]

[0068] Wherein, represents the radiation value of the atmosphere column outward of a certain latitude and longitude position of the earth observed by the satellite M0 high-spectral channels, which is a function of the atmospheric temperature and humidity and other physical parameters, which are uniformly denoted by Satellite data inversion is an inverse problem, that is, how to solve the atmospheric profile The value on different air pressure layers, and the present application refers to the inversion of temperature and humidity.

[0069] Among numerous machine learning methods, the radial basis function (RBF) neural network has achieved a good balance between precision and efficiency due to its unique advantages, and has become an important technical path in high-spectral temperature and humidity profile inversion. It uses radial basis function as the activation function of the hidden layer, has the single-hidden-layer local approximation characteristic, and exhibits unique value in temperature and humidity profile inversion, showing the balanced advantages of precision and efficiency. This method has a solid mathematical theoretical basis, strong generalization ability, and excellent adaptability to unseen data, and the network structure characteristics can effectively avoid the problem that the BP network is easy to fall into local minimum value. Some studies have shown that compared with the BP neural network, the training time is shortened to 1 / 4, which significantly improves the calculation efficiency.

[0070] With the continuous development of numerical models, the accuracy of weather forecasting has reached a high level, which requires that the error of satellite remote sensing inversion products needs to be systematically better than the background field of numerical models, so as to produce a positive effect in weather analysis and data assimilation. Therefore, the present embodiment will introduce the temperature and humidity profile predicted by the numerical model as the background profile of inversion, use the radial basis function (RBF) to establish a mathematical network model of the radiation error and the background profile error simulated by the atmospheric radiation transfer model R(X), and then realize the inversion of temperature and humidity profile.

[0071] First, according to the historical data of the numerical model prediction The dataset consists of forecast profiles (such as 6-hour temperature forecast profiles) and corresponding observed profiles for the same time period, as well as radiation information from satellite observations (such as brightness temperature). The dataset is defined as follows:

[0072]

[0073] in, This represents the k-th real-world profile observed by the satellite. The corresponding radiation value vector is given by Composed of observed radiation values ​​from each channel; This is the background profile predicted by the corresponding numerical model. Both the actual and background profiles have observed and predicted values ​​at the same L pressure layers. This represents the background error of the k-th background profile at the j-th pressure layer.

[0074] The radiation simulation error of the background profile is calculated using the radiative transfer model. The background profile predicted by the numerical model is then compared. Substituting into the radiative transfer mode (such as the RTTOV mode) yields The simulated radiation values ​​of each satellite channel are used to obtain the simulated radiation error. = -R( Two datasets were constructed using background error and radiation simulation error. Among them, E s Used to construct a basis vector model, E t The following definitions are used to optimize the model:

[0075]

[0076] in: and These are the sample sizes of Es and Et, respectively. .

[0077] Radial basis function (RBF) is constructed using radiation simulation errors, and the coefficients of the inversion model are solved.

[0078] This embodiment introduces the temperature and humidity profiles predicted by numerical models as the background profiles for inversion, and uses radial basis functions to establish a mathematical network model between radiation simulation errors and background profile errors, thereby achieving rapid and accurate inversion of temperature and humidity profiles.

[0079] Based on the above embodiments, this embodiment constructs a radial basis function neural network model, including:

[0080] Calculate the nonlinear distance correlation coefficient between the background error at each pressure layer and the radiation simulation error of each satellite hyperspectral channel in the union of the first and second datasets;

[0081] For each air pressure layer, the satellite channels are sorted in descending order of the nonlinear distance correlation coefficient, and the first M channels are selected as the inversion preselected channels, and the simulation errors thereof are used as the basis vectors of the inversion model;

[0082] A radial basis function neural network model is constructed according to the background error in the first data set and the radiation simulation errors of the first m channels of the preselected channels, 1≤m≤M.

[0083] On the basis of the above embodiment, the nonlinear distance correlation coefficient between the background error in the first data set and the radiation simulation error of the satellite channel is calculated according to the following formula in the embodiment:

[0084] (4)

[0085] wherein, is the nonlinear distance correlation coefficient between the background error sequence and the radiation simulation error sequence ; , and represent the row average, the column average and the overall average, respectively, and b is represented in the same way; A and B represent the bicoherence distance matrix.

[0086] On the basis of the above embodiment, a radial basis function neural network model is constructed according to the background error in the first data set and the selected radiation simulation error according to the following formula in the embodiment:

[0087]

[0088] (6)

[0089] wherein, 1≤m≤M, indicates a vector composed of the radiation simulation errors of the first m channels of the preselected channels; k1 and k2 are vector identifiers, and ||.|| represents the Euclidean distance between two vectors; is the Euler radial basis function value, and D is a symmetric radial basis function matrix, is the first data set, is the first data set in the jth air pressure layer, and the length of the background error sequence composed of the vector, is the coefficient of the radial basis function neural network model, and the subscript j represents the air pressure layer position, and m represents the number of hyperspectral channels used.

[0090] On the basis of the above-mentioned embodiments, the radial basis function neural network model is optimized by using the second data set in this embodiment, comprising:

[0091] The temperature and humidity profiles of each pressure layer are retrieved by using the radial basis function neural network model constructed according to different satellite channel numbers m (1≤m≤M) according to the second data set;

[0092] The root mean square error of the retrieval error of the second data set when using different preselected satellite channel numbers m is calculated;

[0093] For each pressure layer, the satellite channel number corresponding to the minimum value in the root mean square error is selected as the optimal satellite channel number for the temperature (humidity) profile retrieval of the pressure layer.

[0094] On the basis of the above-mentioned embodiments, the radial basis function neural network model is optimized by using the second data set in this embodiment, comprising:

[0095]

[0096] Wherein, is the first data set, is the second data set, and H matrix is the radiation simulation error of the kth profile in and the radial basis function between the radiation simulation error in is a row vector composed of is the number of samples in is the coefficient of the radial basis function neural network model using m channels corresponding to the jth pressure layer, is the temperature or humidity value of the kth background profile of the jth pressure layer in is the corresponding retrieval result. For the retrieval of variables of the jth layer, different satellite channel numbers m are used, and formula (7) is used to obtain the results of all samples in The RMSE (root mean square error) of the retrieval error is calculated, and the satellite channel number corresponding to the minimum RMSE is selected as the optimal channel number for retrieving variables in the jth layer.

[0097]

[0098] ​​​The FY4B-GIIRS radiance data under clear sky conditions are used to retrieve temperature and humidity profiles. GIIRS has M0=1690 high spectral channels, of which 1-723 are long-wave channels and 724-1690 are mid-wave channels. 960 temperature and humidity profiles under clear sky conditions from August to December 2024 are selected as the truth to build the model and test the model. The background profiles are from the 6-hour forecast of a numerical model, from 1000 hPa to 0.01 hPa, with 41 vertical pressure layers, consistent with the truth profiles. The radiative transfer model is RTTOV-4.0. In the experiment, the high spectral channels are selected for optimization, and the experiment is designed to use at most the first M=300 channels with larger distance correlation coefficients for optimization.

[0099] Figure 2 The number of high spectral channels used by the RBF neural network model using FY4B-GIIRS radiance data to retrieve atmospheric temperature and humidity (mixing ratio) profiles at different pressure levels is given. The x-axis represents the number of channels, and the y-axis represents the pressure level. The red column represents temperature, and the green column represents water vapor mixing ratio. For temperature retrieval, only 300 channels were used to retrieve 0.03 hPa temperature, and 299 channels were used to retrieve 600 Pa temperature. The number of satellite channels used at other pressure levels is much lower than 300, with a minimum of 12 channels used for radiation value. For humidity retrieval, the channel number distribution from 1000 hPa to 100 hPa is given, as the water vapor content at higher levels is almost zero and the retrieval is not meaningful. There are 7 pressure levels using 300 channels, and the number of channels used above 300 hPa is significantly reduced. The number of channels used below 850 hPa is also close to 300.

[0100] Figure 3 The distribution of high spectral channels used for temperature and water vapor retrieval at each pressure level is given. For temperature retrieval, channels 1-250 and 1450-1690 are used more, and for channels in the middle interval, more are used above 1 hPa and between 850 hPa and 700 hPa, and fewer are used at other levels. For humidity retrieval, channels used below 400 hPa are mainly distributed between long-wave channels 100-400 and mid-wave channels 1100-1500.

[0101] Figure 4 The root mean square error RMSE of temperature retrieval for 480 test samples (part (a) of Figure 4 and its relative change with the background error (part (b) of Figure 4(See section (b)). The figure shows that the background error provided by the numerical model increases from low to high levels, especially above 10 hPa. The RMSE of the mid-to-low-level temperature is around 0.2 K, indicating that the temperature background profile predicted by the numerical model has reached a high level. The inversion error shows that, except for 30 and 200 hPa, the temperature RMSE is better than the background field, with an average of 0.3946 K. The improvement is most significant at 300 and 1 hPa, which are 0.65 and 0.56 times the background error, respectively. The overall RMSE of the temperature profile is 87.5% of the background error, a reduction of 12.5%.

[0102] Figure 5 The given value is the RMSE (Reference Mean Squared Score) derived from the water vapor mixing ratio inversion. Figure 5 (a) and its relative change with background error Figure 5 (See section (b)). The figure shows that the error in water vapor mixing ratio was improved from 100 hPa to 100 hPa. Water vapor is mainly distributed in the lower to middle layers of the atmosphere, and there is very little water vapor above 400 hPa. The figure shows that the RMSE improvement is particularly significant below 850 hPa, approximately 85% of the model background error, which is very meaningful for weather analysis and data assimilation. At higher pressure levels, the RMSE improvement is not significant, which is related to the very sparse water vapor at higher altitudes.

[0103] The atmospheric temperature and humidity profile inversion device based on satellite hyperspectral data provided by the present invention is described below. The atmospheric temperature and humidity profile inversion device based on satellite hyperspectral data described below can be referred to in correspondence with the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data described above.

[0104] like Figure 6 As shown, the device includes a background error calculation module 601, a radiation simulation error calculation module 602, an inversion channel selection module 603, a dataset partitioning module 604, a construction module 605, an optimization module 606, and an inversion module 607, wherein:

[0105] Background error calculation module 601 is used to use the temperature and humidity profiles of the numerical model’s historical forecasts as background profiles, and to calculate the error between the background profiles and the actual observed profiles as background error.

[0106] The radiation simulation error calculation module 602 is used to substitute the background profile into the radiation transfer mode to obtain the simulated radiation values ​​of all hyperspectral channels of the satellite, and calculate the error between the corresponding satellite observed radiation value and the simulated radiation value as the radiation simulation error;

[0107] The inversion channel screening module 603 is configured to calculate a non-distance correlation coefficient between the background error of each pressure layer and the radiation error of each channel, and screen out preselected satellite channels for inversion on each pressure layer according to the correlation coefficients from large to small.

[0108] The data set division module 604 is configured to construct a first data set according to a part of the background error and corresponding radiation simulation error, and construct a second data set according to another part of the background error and corresponding background profile, live observation profile and radiation simulation error.

[0109] The construction module 605 is configured to construct a radial basis function neural network model by using the first data set.

[0110] The optimization module 606 is configured to optimize the radial basis function neural network model by using the second data set.

[0111] The inversion module 607 is configured to perform atmospheric temperature and humidity profile inversion to obtain final temperature and humidity profiles by using the optimized radial basis function neural network model.

[0112] The embodiment introduces the temperature and humidity profiles predicted by the numerical mode as the background profiles for inversion, and establishes a mathematical network model between the radiation error simulated by the atmospheric radiation transmission mode R(X) and the background profile error by using the radial basis function, so as to realize fast and accurate inversion of the temperature and humidity profiles.

[0113] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for inverting atmospheric temperature and humidity profiles based on satellite hyperspectral data, characterized in that, include: The temperature and humidity profiles from historical numerical model forecasts are used as background profiles, and the error between the background profiles and the observed profiles is calculated as the background error. Substitute the background profile into the radiative transfer model to obtain the simulated radiative value, and calculate the error between the satellite-observed hyperspectral radiative value and the simulated radiative value as the radiative simulation error. Calculate the nonlinear distance correlation coefficient between the radiation simulation error of each hyperspectral channel of the satellite and the background error of temperature and humidity on each barysphere. Select the first m channels from largest to smallest according to the correlation coefficient as the inversion pre-selected channel combination for the temperature and humidity of that barysphere, where m is an integer from 1 to M, and M is a positive integer greater than 1. A portion of the background error data and the corresponding radiation simulation error data are selected to construct the first dataset, while another portion of the background error data, the corresponding background profile, the actual observation profile, and the radiation simulation error are used to construct the second dataset. Using the first dataset, for each pressure layer, M corresponding inversion pre-selected channel combinations are used to construct a radial basis function neural network model between the radiation simulation error and the background temperature and humidity profile error of each inversion pre-selected channel combination. The temperature and humidity inversion errors of M different inversion pre-selected channel combinations on each pressure layer were statistically analyzed using the second dataset. The inversion pre-selected channel combination with the smallest inversion error was used as the standard to select the best channel combination. The radial basis function neural network model corresponding to the selected optimal channel combination is used to invert the ambient temperature and humidity profiles to obtain the final temperature and humidity profiles.

2. The method for inverting atmospheric temperature and humidity profiles based on satellite hyperspectral data according to claim 1, characterized in that, The nonlinear distance correlation coefficient between the radiation simulation error of each hyperspectral channel of the satellite and the background errors of temperature and humidity at each pressure layer is calculated using the following formula: ； in, Construct a sequence for background error With radiation simulation error sequence The nonlinear distance correlation coefficient between them; , and Represent The average of the middle row, the average of the column, and the overall average, for , and Represent The average of the middle rows, columns, and the overall average; A and B denote the bicentric distance matrix.

3. The method for inverting atmospheric temperature and humidity profiles based on satellite hyperspectral data according to claim 1, characterized in that, The radial basis function neural network model between the radiation simulation error and the background temperature and humidity profile error for each inversion preselected channel combination is constructed using the following formula: ； ； in, Let k1 and k2 be vectors representing the radiation simulation errors of m channels in each inversion preselected channel combination, where 1 ≤ m ≤ M; k1 and k2 are vector identifiers, and ||.|| represents the Euclidean distance between the two vectors. These are the Euler radial basis function values, and D is the radial basis function matrix. This is the first dataset. It is a vector composed of the background error of the j-th pressure layer in the first dataset. This represents the coefficients of the radial basis function neural network model corresponding to the inversion preselected channel combination containing m channels at the j-th pressure layer.

4. The method for inverting atmospheric temperature and humidity profiles based on satellite hyperspectral data according to claim 1, characterized in that, The inversion error is the root mean square error.

5. The method for inverting atmospheric temperature and humidity profiles based on satellite hyperspectral data according to claim 1, characterized in that, The temperature and humidity profiles of each pressure layer are inverted using the following formula based on the second dataset and radial basis function neural network models under different combinations of inversion preselected channels: ； in, This is the first dataset. This is the second dataset, and the H matrix is... Radiation simulation error of the k-th profile and Medium radiation simulation error radial basis functions between The row vector formed, K s yes The number of mid-profile samples These are the coefficients of the radial basis function neural network model corresponding to the inversion preselected channel combination containing m channels at the j-th pressure layer. yes The value of the k-th background profile in the j-th pressure layer. It is the result of the inversion.

6. A device for inverting atmospheric temperature and humidity profiles based on satellite hyperspectral data, characterized in that, include: The background error calculation module is used to use the temperature and humidity profiles of the historical forecasts of the numerical model as background profiles, and calculate the error between the background profile and the actual observed profile as the background error. The spectral radiation simulation error calculation module is used to substitute the background profile into the radiative transfer mode to obtain the simulated radiation value, and calculate the error between the satellite-observed hyperspectral radiation value and the simulated radiation value as the radiation simulation error. The hyperspectral channel pre-selection module is used to calculate the nonlinear distance correlation coefficient between the radiation simulation error of each hyperspectral channel of the satellite and the background errors of temperature and humidity at each barosphere. The module selects the top m channels from largest to smallest correlation coefficient as the pre-selected channel combination for the inversion of temperature and humidity at that barosphere, where m is an integer from 1 to M, and M is a positive integer greater than 1. The sample dataset partitioning module is used to select a portion of the background error and the corresponding radiation simulation error to construct the first dataset, and another portion of the background error and the corresponding background profile, actual observation profile and radiation simulation error to construct the second dataset; The module is used to construct a radial basis function neural network model between the radiation simulation error and the background temperature and humidity profile error for each of the M inversion preselected channel combinations for each pressure layer, using the first dataset. The optimization module uses the second dataset to statistically analyze the temperature and humidity inversion errors of M different inversion pre-selected channel combinations on each pressure layer, and selects the optimal channel combination based on the inversion error with the smallest inversion error. The inversion module is used to invert the ambient temperature and humidity profiles using the radial basis function neural network model corresponding to the selected optimal channel combination, and obtain the final temperature and humidity profiles.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the atmospheric temperature and humidity profile inversion method based on satellite hyperspectral data as described in any one of claims 1 to 5.

Images