Retrieval Method Study Of Atmospheric Parameters Of Microwave Radiometer Based On 1D-Var Algorithm

In recent years,passive microwave remote sensing technology has become a research hotspot in the field of atmospheric sounding due to its low power consumption,low volume,high reliability,high-resolution advantages and all weather condition.For passive microwave remote sensing technology,how to achieve high-precision detection of atmospheric parameters at a long distance and without contact is one of the main challenges faced by passive microwave remote sensing technology in the field of atmospheric sounding.As the main representative of passive remote sensing instruments,ground-based microwave radiometer has the ability to detect atmospheric temperature and humidity profiles simultaneously.In this doctoral thesis,the observations of ground-based microwave radiometer are used as the main object to establish an atmospheric temperature and humidity profile retrieval system,with the aim of improving the observation accuracy of atmospheric temperature and humidity profiles.Thus,this doctoral thesis focuses on relevant techniques involves in achieving high-precision atmospheric temperature and humidity profile.The main works of this doctoral thesis are as follows:(1)This doctoral thesis firstly analyzes the basic components of the atmosphere and the vertical distribution of atmospheric temperature and humidity profiles.And according to the theory of atmospheric microwave transmission,the physical relationship between the radiated energy of oxygen(water)molecules in the atmosphere and the temperature and humidity is analyzed in the microwave band.Further,through the absorption characteristics of water vapor molecules and oxygen molecules,a mathematical model about the relationship between atmospheric radiation brightness temperature and temperature and humidity is established,and the downstream radiation brightness temperature values corresponding to the oxygen and water vapor profiles in the atmosphere at different frequency bands are calculated.Next,the MonoRTM(Monochromatic radiative transfer model),an internationally widely used atmospheric radiative transfer model,is introduced.The radiative brightness temperatures in the microwave band calculated using actual data are used to show the utility of the MonoRTM model.The retrieval methods for ground-based microwave radiometer inversion of atmospheric parameters are analyzed,and the advantages and shortcomings of different methods as well as the conditions of applicability are compared.(2)In the second part of this doctoral thesis,the study focuses on the retrieval method of atmospheric parameters based on BP neural network.A revision method for the sample data set is proposed for the problem that there is a geographical difference between the observation location of ground-based microwave radiometer and the launch location of sounding balloon,which leads to the deviation of the observation information from the training sample.The BP neural network retrieval algorithm training set data is pre-processed to eliminate abnormal samples,and the correlation between the brightness temperature and atmospheric parameters in the sample data is improved by revising the simulated brightness temperature in the sample data,and the revised training samples are used to train the BP neural network model.The inversion results are evaluated and analyzed,and conclusions are given.(3)In the third part of this doctoral thesis,to solve the problem of insufficient generalization ability of BP neural network inversion method in complex background environment,the retrieval study of the physical retrieval algorithm of atmospheric temperature and humidity profile based on one-dimensional variational algorithm is carried out.The one-dimensional variational retrieval method solves the ill-posed problem in the atmospheric microwave radiative transfer function by add a prior information using as constraints.For the problem that the background error covariance matrix in a prior information is not of full rank,using the minimum variance estimation method.Further,the influence of the iteration step size on the inversion results in the optimization scheme of the objective function is analyzed,and a Gaussian Newton iteration method with a step size factor is proposed to realize the inversion of atmospheric temperature and humidity profiles by a one-dimensional variational algorithm under clear sky conditions.(4)At last,the problem of large bias and scattering of iterative results in the inversion of atmospheric parameters over the ocean by the one-dimensional variational method is investigated.It’s found that the one-dimensional variational method in the iterative process of retrieving the atmospheric temperature profile,the calculation of simulated brightness temperature in the water vapor band shows large deviations,making the iterative results not converge.To solve this problem,a combined one-dimensional variational retrieval method has been proposed to overcome the non-convergent problem caused by the simulated brightness temperature error in the water vapor band.The new background error covariance matrix consists of temperature and water vapor background error covariance matrix.Meanwhile,due to the lack of the radiosonde data,there is a calculating problem for the background error covariance matrix,the reanalysis data are used instead of the radiosonde data,the errors in the calculation process can be reduced by constraining the iterative equations using the Gauss-Newton iteration method with correction factor mentioned in the previous chapter.Finally,the performance of the one-dimensional variational retrieval method under oceanic conditions is evaluated and analyzed based on the results.