利用SSA(奇异谱分析)对测站坐标时间序列进行数据处理,包括数据补缺、趋势项和周期项的识别和提取.奇异谱分析是一种从时间序列的动力重构出发并与经验正交函数相联系的统计技术,可以较好地从含噪声的有限尺度时间序列中提取趋势和周期等信息,目前已应用于多种时间序列的分析中.将奇异谱分析的优点应用到GPS时间序列分析中:利用奇异谱分析对中国地壳运动观测网络GPS数据服务提供的测站坐标(NEU)时间序列(以BJFS(北京房山)站为例)进行补缺,各向插补的均方误差均为mm级;根据降噪重构序列提取坐标时间序列中的趋势成分(N方向~11.688mm·年-1,E方向29.585mm·年,U方向2.557mm·年-1)和周期成分(N,U方向上存在年周期和半年周期,U方向上还存在着1.5年和0.25年左右的周期,E方向上只存在年周期);对完整序列进行重构降噪,即从原始序列中提取有用信息而丢弃一些干扰信息,起着平滑作用.试验结果表明,BJFS站的各方向上均存在显著的变动周期和明显的趋势,也有较多的噪声信息.
Abstract: Singular spectrum analysis(SSA) was introduced to data processing of station coordinates, including treatment of missing values, identification and extraction of trends and cycles. Singular spectrum analysis, a statistical technique with reconstruction and empirical orthogonal function, could effectively extract trends and cycles from the limited scale of the time series with noise. Hence, SSA was applied to various time series analysis such as the process of GPS time series. Coordinate time series from crustal movement observation network of China GPS data services were used by SSA for treatment of missing values, and the mean square errors (MSE) of the interpolation in every direction were at millimeter degree, reconstruction and extraction of trend (N: -11.688mm·a 1 E:29.585 mm·a-1; U:2.557 mm·a 1) and cycle components (N:0.5a, la; E.. la; U:O. 25a, 0.5a, la, 1.5a). The results show that obvious trend, significant cycles and noise do exist in each direction of Beijing Fangshan Station (BJFS).