Abstract: A new method is presented based on the characteristics of the satellite clock errors,i.e.the bias,drift,acceleration and periodicity of the clock.First the new method constructs a reasonable model.The model includes a quadratic polynome and the periodic terms.Then IGS satellite clock error corrections are used and lost data are repaired with the quadratic interpolation and the noise of the residuals after the quadratic fitting are reduced by the wavelet transform at last the dominant periodic terms of the residual variations are found by using the spectrum analysis.On the basis of them,precise fitting and predicting of GPS satellite clock errors are achieved.Experimental results of a few days demonstrate the feasibility and effectiveness of the new method.