基于改进奇异谱分析方法提取GNSS坐标时间序列趋势项及季节项信息

发布时间：2017-12-31 02:27

本文关键词：基于改进奇异谱分析方法提取GNSS坐标时间序列趋势项及季节项信息　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：第一个GNSS连续观测站在1991年1月20日建立以后,世界各国陆续建立更多的GNSS连续观测站。并且随着GNSS测量技术的进步及数据处理中模型精度不断提高,在全球范围内已经累积了 20多年高精度的GNSS连续观测数据。为研究不同时空尺度的地球物理现象如地球自转、区域形变、地震形变监测、冰后回弹、断层滑动及全球板块构造运动提供重要的数据支持。数据预处理是GNSS站点坐标时间序列分析的第一步工作,其主要内容包括三部分:粗差探测与剔除、缺损数据插值及阶跃项改正。在采用小波分析和功率谱分析对GNSS站点坐标时间序列进行分析时,需要GNSS站点坐标时间序列去除线性趋势并为零均值。首先,在时域上采用小波分析从GNSS站点坐标时间序列中提取的季节项进行分析。其次,采用功率谱分析对GNSS观测数据进行分析时,结果发现在低频处的谱能量较大且频谱呈倾斜(斜率趋近于-1),这说明噪声项中包含着闪烁噪声;但随着频率的增加,在高频处谱能量逐渐降低且频谱趋于平缓(斜率趋近于0),这说明噪声项中包含着白噪声。结果得出的闪烁噪声和白噪声模型组合也与GNSS站点坐标时间序列噪声的最佳模型相符合。同时功率谱分析结果表明GNSS坐标时间序列中含有频率(cpy)接近1.0和2.0的季节项。在与GNSS观测技术相关的系统误差及各种地球物理效应共同的影响下,GNSS站点时间序列中可能包含非长期趋势项、阶跃项、噪声项以及振幅随时间变化的季节项。如何将上述信息甚至一些原因不明确的其他信息从GNSS站点时间序列中分离出来是现在时间序列研究的热点。使用传统参数模型去解决这些复杂的问题时具有局限性。作为一种从数据自身出发的无参自适应的奇异谱分析方法可以在没有使用原始数据中任何地球物理现象先验信息的情况下将有用信息从从受到噪声干扰的GNSS站点时间序列中提取出来。为改正传统的奇异谱分析方法(SSA)具有相移现象缺点,本文提出一个改进奇异谱分析方法(SSA-PD)用于拟合GNSS站点时间序列。通过模拟数据计算表明模拟信号和改进奇异谱分析方法重构信号残差的均方根小于1.8 mm,并且改进奇异谱分析方法拟合精度显著优于传统奇异谱分析方法。采用IGS站点坐标时间序列将小波分析方法与改进奇异谱分析方法进行比较,结果表明改进奇异谱分析方法在提取年以及半年季节项要优于小波分析方法。最后对奇异谱分析方法提取GNSS坐标时间序列中的趋势项和季节项进行分析,结果表明GNSS站点时间序列时频特性呈现出了显著的区域性。并对形成的因素进行定性分析。
[Abstract]:The first GNSS continuous observation stations in January 20, 1991 after the establishment of the world began to build more continuous GNSS stations. The accuracy of the model and with the continual improvement and data processing of GNSS measurement technology in the worldwide has accumulated 20 years of continuous observation data of high precision GNSS. To study the effect of different temporal and spatial scales of geophysical phenomena as the earth's rotation, regional deformation, earthquake deformation monitoring, post glacial rebound, fault slip and global plate tectonic motion to provide important data support. Data preprocessing is the first step in the job analysis coordinate time series of GNSS site, its main contents include three parts: detection and elimination of outliers, missing data interpolation and step correction in the wavelet analysis and power spectrum analysis of coordinate time series of GNSS site, GNSS site to coordinate time series and linear trend removal Zero mean. First of all, in the time domain by using wavelet analysis to extract the seasonal item from the GNSS site coordinates in time series analysis. Secondly, by analyzing the power spectrum analysis of GNSS data, the results found in the spectrum of more energy and spectrum at low frequency is inclined (slope closer to -1), indicating that noise contains a flicker noise; but with the increase of frequency, the high frequency spectrum energy decreases gradually and tends to smooth spectrum (slope near 0), indicating that the noise contained in white noise. The best result of the model is a combination of flicker noise and white noise model and the coordinate time series of GNSS site noise is consistent. At the same time, the power spectrum analysis results show that the frequency with GNSS coordinate time series in (CPY) close to 1 and 2 of the season. In correlation with GNSS observation technique and system error of various geophysical effects together under the influence of GN May contain non SS site long-term trends in time series, step, noise and the amplitude changes with time. How will the season information or other information for some reason is not clear from the GNSS site in the time series is now separated from the time series of research hot spots. With the limitations of the traditional model parameters to use to solve these complex problems. The extracted GNSS site time series as a singular spectrum analysis method non parametric adaptive starting from the data itself can not use the original data of any geophysical phenomena under the condition of prior information will be useful information from the subject to noise. In order to correct the traditional singular spectrum analysis method (SSA) with phase shift defects, this paper presents an improved method of singular spectrum analysis (SSA-PD) is used for fitting the GNSS site time series. Through the simulation data calculation table The analog signal and improve the singular spectrum analysis method to reconstruct the signal RMS residuals of less than 1.8 mm, and improved the singular spectrum analysis method fitting accuracy is significantly better than the traditional singular spectrum analysis method. Using the IGS site coordinate time series wavelet analysis method and improved singular spectrum analysis method, the results show that the improved singular spectrum analysis method in the extraction of years the first half of season and items to be superior to the wavelet analysis method. Finally the singular spectrum extraction method of GNSS coordinate time series in the trend and seasonal item analysis, the results show that the GNSS site time series time-frequency characteristics showing a significant region. And the factors for the formation of qualitative analysis.

【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P228.4

【参考文献】