| The land surface temperature is an important indicator of the Earth's surface energy balance and the greenhouse effect, and is also a key factor in the physical processes of the surface and global scale, However, the retrieval accuracy of surface emissivity directly affects the estimation of surface temperature, So accurate access to land surface emissivity is the key to land surface temperature retrieval. In the paper, The CRTM (Community Radiative Transfer Model) radiative transfer model is used to retrieve the relationship between microwave land surface emissivity and its influencing factors in the desert region by combining the microwave imager data of FY-3 meteorological satellite C-star (FY-3C).Through the improvement of land surface emissivity retrieval accuracy, to and build the algorithm model. The paper is divided into six chapters. The first two chapters introduce the current research progress?research results on the land surface emission algorithm and the CRTM model. The first chapter mainly introduces the physical model, semi-empirical model and data analysis method of handling parameter problem, and describes the application way and limitation of each model in detail. The second chapter introduces the CRTM model and used data in this paper,the focus is on using the CRTM model to restores the true state of the radiative transfer process, as well as the process of the algorithmic program.By summing up the previous research results and the existing algorithm model,we try to develop an algorithmic model for specific problems. In the third chapter, a joint algorithm based on genetic algorithm and Newton method is introduced to retrieve the microwave land surface emissivity model in desert, which is combined with the advantages of genetic algorithm for global efficient search and Newton method for local search, and make full use of the advantages of an algorithm to make up for disadvantages of another algorithm, And the joint algorithm is accessed into CRTM model. In contrast, the modified land surface emissivity simulates the brightness temperature closer to the observed value than the original model's calculated brightness temperature (the land surface emissivity is not changed), and the error of the brightness temperature simulation value has reached 2.13K. At the same time, in order to test the universality of the model, it is used in the other two desert areas, the results also show that the simulated brightness temperature is better than the original model. In order to improve the surface emissivity structure, we add two effects in the original linear model in the four chapter,(The soil moisture at 0.07 m and 0.28 m below ground), the functional relationship model (multi-factor model)with four factors for the microwave land surface emissivity in the desert region is obtained and the surface emissivity calculated by the model is provided to the CRTM model for simulation analysis. It can be seen that the land surface emissivity model of the multi-factor is slightly better than that of the two factors model. After analyzing the linear algorithm model, and then introduce the nonlinear algorithm model In the five chapter, BP neural network algorithm, The original 40 scanning points information is input as forty training samples, and the weights and thresholds are trained by the back adjustment mechanism to obtain the nonlinear functional relationship of the microwave land surface emissivity model. From the retrieval results,the error of the brightness temperature simulation value is about 3.34K. The results are slightly worse than the linear model, which may be due to the limitations of the BP neural network algorithm in dealing with nonlinear problems. |
PDF Full Text Request