基于小波分析和LS-SVM的积雪厚度多星融合反演

Multi-star fusion inversion of snow thickness based on wavelet analysis and LS-SVM

针对目前利用单星反演雪深存在的不足,该文研究GPS信噪比(SNR)观测量中卫星反射信号分离模型的改进,以及积雪厚度反演模型,首先利用小波分析有效分离卫星反射信号,进而通过最小二乘支持向量机(LS-SVM)实现积雪厚度的多星融合反演。以美国板块边界观测计划提供的监测数据为例,经实验表明:小波分析能够提高卫星反射信号的分离精度,有效抑制了异常跳变现象,单星反演结果与雪深参考值之间相关系数r均大于0.547,相对于传统方法至少提高了8.9%;多星融合反演结果与雪深参考值吻合性较好,当融合卫星达到6颗时,其反演结果与雪深参考值之间相关系数r=0.948,相对于单颗卫星至少提高了6.4%。

In view of the shortcomings of using single-star inversion snow depth,this paper studied the improvement of satellite reflection signal separation model in GPS signal to noise ratio(SNR)observation,and the snow thickness inversion model,firstly using wavelet analysis.The satellite reflection signal was effectively separated,and then the multi-star fusion inversion of snow thickness was realized by a least squares support vector machine(LS-SVM).Taking the monitoring data provided by the US plate boundary observation program as an example,experiments showed that wavelet analysis could improve the separation accuracy of satellite reflection signals,effectively suppress the abnormal jump phenomenon,and the correlation coefficient between single star inversion results and snow depth reference values.Both were greater than 0.547,which was at least 8.9%higher than the traditional method;the multi-satellite fusion inversion results were in good agreement with the snow depth reference value.When the fusion satellite reached 6,the inversion result was related to the snow depth reference value.The correlation coefficient was 0.948,which at least increased 6.4%over a single satellite.