Abstract: In the analysis of crustal movement and deformation, it is generally believed that mass movement is featured by rigid motion, and the status of the block is considered as remarkably stable. However, due to local tectonic movement of the block and the gross error, abnormal movement will be observed from a few survey station. Only by removing these abnormal stations can the reasonable analysis results of the long-term movement trend and motion law of the crust in the study area be obtained. For a long time, the most common way to screen station is to see if the difference value between the posterior estimate of the residual standard deviation and residual error is beyond some limiting value. This method fails to eliminate abnormal stations efficiently. This paper proposes to use the mean shift model to eliminate the anomaly stations, and compare it with the quasi calibration method and the traditional residual-standard-deviation-based simple LS estimation method. The fitting analysis result of the recent velocity field data of GPS observation of China’s tectonic environment monitoring network in the Circum-Bohai-sea area shows that the iterative processusing the mean drift amount can not only screen the stations robustly, but also avoid the impact raised by selection inaccuracy of Quasi-Accurate-Observationand is able to obtain a higher velocity field fitting accuracy compared with the quasi-calibration method.

Key words: crustal movement, least squares method, anomaly station, mean shift model