Abstract: We analyzed the characteristics of the correlated noise in position time series for fiducial GPS stations of the Crustal Motion Observation Network of China. The spectral index for powerlaw noise was derived using the maximum likelihood estimator. We took into account more types of noise models, and found that, besides flicker noise, about 50% east components can be described by a lsquo;random walk noise plus first-order Gauss-Markov noisersquo; model, which is inconsistent with the results from global solutions produced by other agencies. This reveals that the difference in processing strategy can induce varied correlated noise. The magnitudes of white noise and flicker noise are spatially correlated. The sites in southeastern coastal region have larger noise than inland northwestern ones. When considering the correlated noise, the estimation errors of velocity will increase by over an order. The velocity uncertainties are usually less than1 mm/a for horizontal components and larger for vertical ones.