Research On Hyperspectral Microwave Detection Technology Of Atmospheric Temperature Profile

Atmospheric temperature profile is the vertical distribution of atmospheric temperature with height,which is a very important part of atmospheric state parameters.Temperature profile is the basis of studying other atmospheric parameters,and plays an important role in the research fields of numerical weather forecast,extreme weather warning and meteorological analysis.With the rapid development of numerical weather prediction,there is a higher demand for spaceborne microwave radiometer.Nowadays,the vertical detection accuracy of microwave radiometer in orbit satellite can't meet the demand of scientific research and practical application.At present,the microwave temperature radiometer generally has more than ten channels,so it is impossible to carry out fine atmospheric vertical detection.In this context,the development of hyperspectral microwave radiometer has become an inevitable trend.In this thesis,I have carried out the research on the hyperspectral resolution microwave simulation brightness temperature data,including the hyperspectral data processing and atmospheric temperature profile inversion.The main research work includes:Firstly,the theory of atmospheric radiative transfer is analyzed in detail.The forward model is built by solving the absorption coefficient by mpm93 model.The simulation bright temperature data needed for subsequent experiments are generated by using the forward model.At the same time,the principle of atmospheric temperature profile inversion method is discussed,which provides a theoretical basis for the subsequent inversion experiments.Second,this thesis introduces the definition and significance of the weight function in detail.The difference of weight function between hyperspectral data and MWTS-II observation data is compared,which shows the advantage of hyperspectral data in atmospheric vertical detection.In this thesis,two methods are used to preprocess hyperspectral data,which are PCA and channel selection.The principle of the two methods is deduced in detail,and the pretreatment effect is analyzed.Third,the implementation and effect of statistical inversion,neural network inversion and physical inversion are analyzed.The inversion method is verified by using the observation data of MWTS-II microwave temperature radiometer.And the results show that the inversion method is effective.The comparison experiments of different inversion methods are carried out by using hyperspectral simulation bright temperature data of single observation point and mwts-ii simulation bright temperature data.Fourth,the experimental results show that increasing the number of channels can reduce the inversion error.When the number of channels is more than 200,the inversion error is not sensitive to the increase of the number of channels.The brightness temperature noise will affect the inversion accuracy.The noise experiment shows that adding a small brightness temperature noise will make the inversion error larger.The experiment of non atmospheric state factors show that increasing the non atmospheric state factors such as latitude,terrain height and surface temperature can effectively improve the inversion accuracy.The research work in this thesis lays a foundation for the design of microwave radiometer and the use of high spectral resolution detection data in microwave band in the future.