The Study Of Retrieving Lop Nur’s Subsurface Sodium Content Based On GA-PLS,Using PolSAR Technology

Lop Nur is located in the west of Xinjiang, China, acted as an important geographic position along the ancient ‘Silk Road’. It once was a large and famous salt lake in China and its drying-up was always a focus. Lop Nur shows a clear special ‘Ear’ feature on the remote sensing image, which recorded plenty of climate environment information. The objective of this paper is to interpret Lop Nur’s environmental evolution during its drying-up process based on an analysis of its sodium sedimentary feature using remote sensing data. Polarimetric synthetic aperture radar(PolSAR) technology has an apparent advantage in extremely arid regions. Based on ALOS-PALSAR PolSAR image, the polarimetric scattering characters of the lake basin was analyzed in detail. The result shows that, in the ‘Ear’ section of Lop Nur, the bright strips and grey strips have the same kinds of scattering mechanisms, but the scattering process of the bright strips is more complicate; in the lake basin, surface scattering is the main scattering mechanism, but other scattering mechanisms also exist and can’t be ignored. In the ‘Ear’ section, based on the analysis of Cloude decomposition’s 3 component images, it is found that the surface scattering contribution image has a small grayscale dynamic range to express SAR signal’s response of the Ear’s bright-grey strips feature. On the contrary, the multi-scattering contribution image, though stands for a small contribution to the whole scattering mechanism, shows a clear ‘Ear’ feature. Therefore, the author suggests that the multiple scattering is the dominant reason of the ‘Ear’ feature’s formation. Moreover, as multiple scattering mostly happens in the subsurface, so the properties of the subsurface determines PolSAR image’s backscattering intensity.The retrieval technology of polarimetric radar is always an important and difficult item in the application of radar remote sensing. The main innovative point of this paper is the first application of genetic algorithm-partial least squares(GA-PLS) as a modeling method to the PolSAR retrieval field. Based on the polarimetric parameters calculated by Cloude decomposition of ALOS-PALSAR data, this paper firstly build a mathematical model between the polarimetric parameters and Lop Nur’s subsurface sodium content through using GA-PLS algorithm. As a result, the R2 and root-mean-square error of GA-PLS model can reach 0.7 and 9.1 g/kg, which implies the polarimetric parameters selected by GA-PLS can accurately retrieve the subsurface sodium content in Lop Nur. Based on the GA-PLS model, the subsurface sodium content distribution map of Lop Nur can be obtained. Through the analysis of Lop Nur’s subsurface sodium content distribution regularity, it is found that the sodium content distribution shows a clear ‘Ear’ feature. This means the subsurface salt content is the major reason for the bright-grey strips feature on SAR images in Lop Nur. In addition, the subsurface sodium content in Lop Nur along the same strip exhibits an obvious regularity. From the northeast to the southwest section on the same strip, the subsurface sodium content changes from high to low,then to high. In the thesis, the phenomenon is called ‘same strip, different sediments’. Through the contrastive analysis of the strips’ elevation and subsurface sodium content based on previous research results, the author believes this special phenomenon is a result of the lake body’s movement caused by the wind stress in Lop Nur, which leaves a special record both in the strips’ elevation and the subsurface salt sediments. Based on this viewpoint, it is suggested that the wind factor plays a significant contribution to the environment evolution process of Lop Nur.