| In order to determine the accurate state of high-altitude meteorological elements,so as to predict the future weather and early warning of high-risk weather,it is urgent to establish a high-precision meteorological background field.At present,in order to obtain high-quality and high-resolution atmospheric background field,downscaling of large-scale and low-resolution global reanalysis data is one of the feasible and effective method.At present,in order to obtain fine and high-resolution atmospheric background field,downscaling of large-scale and lowresolution global reanalysis data is one of the feasible and effective means.The objective of this paper is to establish a high-precision temperature background field in China’s coastal areas,with a height of 0～10km,a vertical resolution of 100m and a time resolution of 12h.And test the results.Firstly,this paper introduces the current conventional global atmospheric reanalysis data,and then based on the global reanalysis data launched by the National Environmental Prediction Center of the United States,the cubic spline interpolation method,inverse distance weight interpolation method,Cressman interpolation method and artificial neural network method are used to establish the high-resolution temperature background field in China’s coastal areas,Finally,the results are applied to Newton’s nonlinear iterative physical inversion method to verify the effectiveness of the background field.The first part of this paper introduces the sources of atmospheric meteorological data used in this paper and the theoretical methods used in this paper,and analyzes the principles of various spatial interpolation methods,which provides a theoretical basis for the later experiments.Then,a special meteorological state in the atmosphere-temperature inversion phenomenon is introduced,and several causes of temperature inversion are summarized.In addition,two error analysis methods,root mean square error and frechet distance,are proposed,which provides a quantitative analysis method for the later results.The second part of this paper attempts to use the traditional spatial interpolation method to establish the temperature background field of high-resolution NCEP data on the basis of the above theory.Through the design experiments of the parameters of different methods,the optimal parameters of each interpolation algorithm are finally determined.On this basis,the temperature background field with a vertical resolution of 100m at the height of 0～10km is established.By comparing the experimental results of various methods,the accuracy and similarity of the results are quantitatively analyzed.Then,based on the background field,the experimental data are classified according to seasons,and the laws and reasons of coastal atmosphere changing with seasons are analyzed;The experimental data are divided into three layers according to height:0～2km,2～6km and 6～10km.The variation law and reason of coastal atmosphere with height are analyzed.In the third part of this paper,BP neural network is used to establish the background field.By selecting appropriate network parameters,BP neural network is established and trained.The differences and similarities between the results of neural network method and traditional spatial interpolation method are analyzed.It is found that there is a large error in the near ground data of the global reanalysis data of the U.S.Environmental Prediction Center.By introducing the historical weather forecast data as an additional input to the network,the ground temperature error is corrected,and the feasibility and effectiveness of the corrected results are verified.Finally,based on the above work,the principle of Newton nonlinear iterative method is briefly described.Based on the background field obtained by inverse distance weight method and neural network method,the atmospheric inversion experiment of Newton nonlinear iterative method is carried out,and the quality of the background field is analyzed from the perspectives of convergence speed and final result error. |
