The backscattering signal of laser ceilometer as a typical nonlinear and nonstationary signal is susceptible to be polluted by noise. Aiming at this problem, the ensemble empirical mode decomposition (EEMD) denoising method is applied. Firstly, we use EEMD to decompose the noise signal and analyze the decomposition of the IMF component, then find out the larger component of IMF. Finally, we reconstruct the IMF component and the rest of the components signal after using SavitzkyGolay (SG) filter. The simulation and experiment results show that compared with the traditional empirical mode decomposition(EMD) method, the signaltonoise ratio based on the EEMD method after processing increases 1.695 dB, the mean square error decreases by an average of more than 30%. It is shown that the method is suitable for nonlinear and nonstationary characteristics for the scattering echo signal processing, and able to provide the high signaltonoise ratio of the initial data by laser ceilometer for the next level cloud base height inversion.