Research On Atmospheric Parameter Retrieval Method Based On Ground-based Microwave Radiometer

Retrieval of atmospheric parameters based on ground-based microwave radiometer is an important research direction in the field of Meteorology and climate.In this paper,the algorithm and its improvement of ground-based microwave radiometer for retrieving temperature and humidity profiles in clear sky are studied.The main research contents include:(1)Based on the radiative transfer theory,atmospheric millimeter wave transmission model and atmospheric remote sensing equation,the absorption spectrum curve of the atmosphere at sea level is drawn,and the model of retrieving atmospheric parameters by ground-based radiometer under clear sky condition is established,and the data pair of "radiosonde data radiometer brightness temperature value" is obtained through forward simulation.(2)A standard BP(back propagation)neural network is built.Aiming at the shortcomings of BP algorithm,such as slow convergence speed,easy to fall into the local minimum and unable to jump out,combined with the advantages of global search of genetic algorithm,a method of using GABP(genetic algorithm back propagation)algorithm to retrieve the temperature and humidity profile is proposed.The historical radiosonde information is used as the learning sample to train the neural network,and the trained network is used to retrieve the brightness temperature of the radiometer in the test sample.The retrieved atmospheric parameters are compared with the actual data provided by radiosonde.The results show that when the target accuracy is the same,the number of iterations of GABP algorithm is reduced by 4000 times compared with the standard BP algorithm.(3)The retrieval errors of atmospheric temperature profile and water vapor density profile are analyzed.On the basis of GABP algorithm,momentum BP learning rule and adaptive learning rate BP learning rule are integrated.The improvement is made from three aspects:learning rule,seasonal classification of samples and introduction of surface parameters.The results show that the retrieval errors of troposphere below 1000 m are improved.