Abstract:The real-time sea surface elevation data obtained by GNSS buoy observation contains tides, waves and white noise, and their periods are quite different. The tide level information can be extracted through filtering. Sea level is a non-stationary and nonlinear time series, and the commonly used threshold filtering, low-pass filtering, wavelet filtering, etc. are not applicable. The Empirical Mode Decomposition (EMD) is suitable for sea level analysis, but it is prone to modal aliasing and component redundancy. In this paper, based on the improved Multi-index Multi-scale Permutation Entropy algorithm (MMPE), combined with Variational Modal Decomposition (VMD) and Complete Ensemble Empirical Modal Decomposition (CEEMD), two noise reduction smoothing models are proposed, VMD+MMPE and CEEMD+MMPE. By analyzing the tidal sequence results of GNSS buoy data and comparing them with the reconstructed tidal level change results of wavelet analysis, the standard deviations are 1.67 cm and 1.39 cm, and the correlation coefficients are 0.9996 and 0.9998. The MMPE values are 0.8988 and 0.9737, respectively. At the same time, through the time-frequency analysis of the three-dimensional spectrogram, the experimental results show that the algorithm based on CEEMD+MMPE has better time-frequency aggregation, and has higher feasibility and effectiveness in GNSS tide survey data processing.