海平面变化的小波和自回归模型集成预测试验

Prediction experiment of sea-level changes based on the wavelet analysis and auto-regressive model

针对海面变化预测时间序列模型中趋势组份和周期（准周期）组份的提取和预测问题，基于吴淞站1955-2001年月平均潮位序列，采用小波分析（wA）与自回归（AR）模型相结合的方案，对小波分解的不同尺度分量序列，借助于时间序列模型进行分量预测，再对它们进行叠加建立预测模型，进行了月平均潮位预测试验。以1955～1996年数据为基础建立模型，1997～2001年数据作为验证，结果表明两种方法的结合使用显示了较好的效果，具有较高的精度。

Based on the monthly average tidal records of Wusong tidal gauge station from 1955 to 2001, a prediction experiment is made using wavelet analysis （WA） and auto-regressive（AR） models in this paper. WA provides a way of analyzing the local behavior of functions, it may be used to isolate the sea-level change records to reveal the hierarchy of features and the scaling behavior. Since components are the filtered versions of the raw data,their behavior is more regular than that of the original signal and more predictable accordingly. Using AR model to predict each component in practice we can take advantage of each significant component, the summation of all the predicted components is the predict results. Analyzing the data of 1955- 1996 to predict the sea-level change of 1997 2001, comparing the prediction results with the original data, we find that these two series are fairly comparable. It was shown that the integration use of these two methods provides an efficient way to predict sea-level changes.