论文部分内容阅读
海洋表面温度(sea surface temperature,SST)对气候有着很大影响,但其所具有的非线性、无明显周期、强随机性等特点,给SST预测分析带来了很大的困难。本文将互补集合经验模态分解(complementary ensemble empirical mode decomposition,CEEMD)与支持向量机(support vector machine,SVM)相结合来研究对海洋表面温度异常(SSTA)的预报,并从预报准确性、可预报时长、不同起报时间对预报精度影响等方面设计了多组数值实验。实验结果显示CEEMDSVM方法预测12个月SSTA的效果较好,平均绝对误差在0.3°C左右,相关系数达到了0.85,而且试验中未出现春季预报障碍问题。
The sea surface temperature (SST) has a great influence on the climate, but its characteristics such as non-linearity, no obvious periodicity and strong randomness bring great difficulties to SST prediction analysis. In this paper, the prediction of ocean surface temperature anomaly (SSTA) is studied by combining CEE with complementary ensemble empirical mode decomposition (CEEMD) and support vector machine (SVM) Forecast time, the impact of different reporting time on the accuracy of forecasting, etc. designed a number of numerical experiments. The experimental results show that the CEEMDSVM method is effective in predicting 12-month SSTA with an average absolute error of about 0.3 ° C and a correlation coefficient of 0.85, and no spring forecast obstacle occurred in the experiment.