探索性GPS全球电离层TEC数据分析

In this paper we apply spatial statistics for mining and analyzing global ionospheric TEC data published by IGS center. The spatial non-stationary property of the global TEC is discovered by analyzing three-dimensional map, histogram, semivariogram and counting the statistics such as mean, median, skewness, kurtosis, etc. The global trend surface of TEC can be modeled by the Spherical Harmonics and 9-order is the most suitable by comparing residuals' statistical characteristics after least-squares fitting. After removing the trend, the semivarigram of the residuals doesn't show stationary character. After doing second-order difference in longitudinal direction and then doing first-order difference in latitudinal direction, the residual result shows the white noise property, and which indicates the local and/or small scale of variation is a three order of partial random walking spatial process. And after three order's difference, error's standard deviation decreased from 9.221TECU to1.175TECU .At the same time, spatial outliers (speckle noises) are detected on TEC data by using the nearest neighbor method based on the Delaunay triangulation and its spatial statistical distribution is found obeying monotonic decreasing. For the TEC data obtained at different epoch, its spatial distribution can be describe with the fore-mentioned mixed model including global trend, local variation and abnomaly, that is, the TEC global trend surface can be fitted by the Spherical Harmonics, and the local (small scale)change can be described with the three-order of partial random walk process, also disturbed by speckle noises obeying monotonic decreasing.