Study On Atmospheric Temperature Profile Retrieval Methods Based On Microwave Radiometer

Atmospheric sounding is an observation method of various meteorological information,and the atmospheric temperature profiles are important informations of weather forecast and climate research.The microwave radiometer is more and more widely used in atmospheric remote sensing in recent years,because of its high sensitivity,wide detection range and other advantages.Therefore,in this paper,the microwave radiometer microwave transmission process and the method of retrieving the atmospheric temperature and humidity profiles are studied.First of all,this article briefly discuss the structure and basic observation principle of the microwave radiometer by introduce the MP3000 A microwave radiometer.Through researching the basic principle of microwave radiation way,deriving the relationship between brightness temperature and brightness.Through the analysis of radiation in the medium dissemination way,obtaining the radiate transfer equation.According to the characteristics of atmospheric medium,presenting the atmospheric microwave radiation brightness temperature.Secondly,according to the absorption characteristic of oxygen molecular and water vapor in the atmosphere,we elicit the atmospheric microwave radiation model via radiation transfer theory.By comparing the simulation of the brightness temperature with the sounding data of Harbin with the observation data of MP3000 A microwave radiometer,verify the effectiveness and correctness of the algorithm.Thirdly,this article briefly introduce several atmospheric temperature profile retrieval methods,then mainly discuss the neural network retrieval method.We design a temperature profile retrieval model under 10 km height of Peking atmosphere named Net-profile based on neural network.After training and simulating by test samples,we compare the simulation result with the profile data in MP3000 A microwave radiometer and prove that Net-profile simulation result make a good match with the sounding data and observation data of MP3000 A microwave radiometer.Based on the research of Net-profile,we improve the neural network model structure by divide the atmosphere height into 10 layers in certain rule,and design ten separate nets named Net-1~Net-10.The separate neural network model is qualified to retrieval the atmospheric temperature profile,and have almost same performance with Net-profile.However,the training time of the new net is much shorter than that of Net-profile and MP3000 A microwave radiometer.