Recent studies have shown that the global mean sea level (GMSL) is accelerating. For improved process understanding and sea level projections, it is crucial to precisely estimate the GMSL acceleration due to externally-forced global climate change. For that purpose, the internal climate variability-related signal of the GMSL needs to be removed from the GMSL record. In the present study, we estimate how the observed GMSL rate has evolved with time over the altimetry era (1993-present), with the objective of determining how it is influenced by the interannual variability. We find that the GMSL rate computed over 5-year moving windows, displays significant interannual variability around 6–7 years and 12–13 years, preventing from robust acceleration estimation. To remove from the observed GMSL time series, the interannual variability, possibly related to internal climate modes, like ENSO, PDO, IOD, NAO or AMO, we use two methods previously widely applied in the literature: (1) multiple linear regression of the GMSL against some climate indices, and (2) Empirical Orthogonal Function (EOF) decomposition of the gridded sea level data to isolate the interannual signal. Although the interannual signal of the corrected GMSL time series is reduced, a cycle around 6–7 years still remains in the GMSL rate. We discuss possible sources of the remaining 6-7-year cycle, including the limitation of the methods used to remove the interannual variability.

最近的研究表明，全球平均海平面（GMSL）正在加速。为了提高对过程的了解和海平面预测，精确估算由于外部强迫的全球气候变化导致的GMSL加速至关重要。为此，需要从GMSL记录中删除GMSL与内部气候变化相关的信号。在本研究中，我们估计在高海拔时代（1993年至今）中观测到的GMSL率如何随时间变化，目的是确定它如何受到年际变化的影响。我们发现，按5年移动窗口计算的GMSL率在6-7年和12-13年之间显示出显着的年际变化，从而无法进行可靠的加速估计。为了从观测到的GMSL时间序列中去除年际变化，可能与内部气候模式有关，例如ENSO，PDO，IOD，NAO或AMO，我们使用以前在文献中广泛应用的两种方法：（1）GMSL对某些气候指数的多元线性回归；（2）经验正交函数（EOF）分解网格化的海平面数据以隔离年际信号。尽管校正后的GMSL时间序列的年际信号减少了，但GMSL率仍保持6-7年左右的周期。我们讨论了剩余的6-7年周期的可能来源，包括用于消除年际变化的方法的局限性。（2）对网格化海平面数据进行经验正交函数（EOF）分解，以分离出年际信号。尽管校正后的GMSL时间序列的年际信号减少了，但GMSL率仍保持6-7年左右的周期。我们讨论了剩余的6-7年周期的可能来源，包括用于消除年际变化的方法的局限性。（2）对网格化海平面数据进行经验正交函数（EOF）分解，以分离出年际信号。尽管校正后的GMSL时间序列的年际信号减少了，但GMSL率仍保持6-7年左右的周期。我们讨论了剩余的6-7年周期的可能来源，包括用于消除年际变化的方法的局限性。