Long-term prediction and comparison of sea-level change based on the SSA and MGF model
基于SSA和MGF的海面变化长期预测及对比

Tide gauges and satellite altimetry are the two measurements techniques of present-day sea level change,and tide gauges provide sea-level variations with respect to the land on which they lie.The predictions of sea-level changes are affected by modeling,methods,data length,data quality and other factors,which cause the uncertainties of prediction.Based on the monthly average tidal records of six tidal gauge stations in East China since the 1950s,Mean Generation Function(MGF) and Singular Spectrum Analysis(SSA) are employed to discuss the stability of long-term prediction.MGF model is built with each station's initial data of over 20 years,and the subsequent data are used to undertake comparative multi-experiments and tests.As a result,these prediction experiments testify that MGF exhibits more favorable and steady long-term prediction.Therefore,based on SSA denoised series,the MGF model is used to predict the sea-level changes of each station on the monthly scale till the year 2050.All stations take on obvious fluctuated rising trend.The calculated annual mean series indicate that the upper limit of the fluctuated sea-level changes can be no more than 20 cm.The velocity of the sea-level changes show periodity and fluctuant with prominent differences in the ascending and degressive segments accompanying with obviously spatial variation.Compared with the previous research findings,whose results are primarily done under linear hypothesis and may show limitation to some extent.Owing to the fluctuations and irregularities of sea-level changes,the prediction conclusion adopted by SSA and MGF methods are relatively comparable,and have favorable long-term prediction potential in terms of methodology and experimental results.Natural forcing,which is a combination with anthropogenic forcing plays an important role in the sea-level change,and there are still enormous uncertainties about sea-level change and its prediction.The integrated prediction system should be constructed with the consideration of multiple factors.Furthermore,comprehensive and comparative research is also needed.