【摘要】 海平面变化具有非平稳性、非线性以及多时间尺度等特性,对未来海平面变化进行准确预测较为困难,为此提出一种集成SSA与LSTM的组合预测模型,利用AVISO提供的1993—2020年的格网化海平面高度异常SLA数据,对全球海平面变化进行了短期预测分析。首先利用SSA分解提取原始SLA序列的长期趋势、周期项和残差等子序列,降低原始序列的复杂度,然后对各子序列分别构建LSTM模型进行预测,最后将子序列预测值重构得到最终预测结果。经与LSTM直接预测、SSA-ARIMA组合模型等方法对比,SSA-LSTM组合模型预测效果更为理想。基于SSA-LSTM组合模型对2021—2025年全球海平面变化趋势的预测结果表明:该时间段全球海平面上升速率约为3.96 mm/a。更多还原

【Abstract】 Sea level change is characterized by non-stationarity, nonlinearity and multi-time scale, it is difficult to predict the future sea level change accurately. The paper proposes a combined prediction model integrating SSA and LSTM, a short-term forecast analysis of global sea level change is carried out using the sea level anomalies(SLA) data from 1993 to 2020 provided by AVISO. Firstly, SSA decomposition was used to extract sub-sequences such as long-term trends, periodic terms and the residual of the SLA data to reduce the complexity of the original sequence. Then, LSTM models were constructed for each sub-sequence to predict the future changes. Finally the predicted values of all the sub-sequences were reconstructed to obtain the final prediction result of SLA. Compared with the direct prediction of LSTM and SSA-ARIMA combined model, the SSA-LSTM combined model has better prediction effect. Based on the SSA-LSTM combined model, the global sea level rise rate in 2021-2025 is about 3.96 mm/a.更多还原