关键词: EOF分解, EEMD分解, 滑动傅里叶分析, 波内频率, 时空重构

Abstract:

A statistical forecast model suitable for nonlinear climatic series is established adopting the Ensemble Empirical Mode Decomposition (EEMD) and Sliding Fourier Analysis (SDFA). Empirical Orthogonal Decomposition(EOF) is firstly employed on climatic pentad anomaly to obtain their time series, and then EEMD is employed on these time series to get their Intrinsic Mode Functions (IMF), the primary Fourier spectrum signals which composed the major component of each IMF are extracted through the model building with SDFA, which equivalently indicates the achieving of primary frequencies of climatic signals. With the reconstruction of IMFs and the residual, the predicted time series can be obtained and be used for further rebuilding of spatial and temporal fields to accomplish the final forecasting. Such ideas is then applied in the wind velocity forecasting in Xinjiang. With the assessment of anomaly correlation coefficients (ACC), the skill score (SS) and anomaly sign score (R), the results show that the proposed algorithm model in ideas mentioned above can capture the primary frequencies of climate variation and give a good prediction on time coefficient of velocity series, and ultimately, a fairly good spatial and temporal wind distributions in Xinjiang are achieved successfully. During the whole prediction, the method shows a good skill, the average SS for the former 40 pentads exceeds 0.5 and the average ACC within 36 pentads surpass 0.4.

Key words:  EOF analysis, EEMD method, sliding Fourier analysis, intrawave frequency, temporal reconstruction