科研之友
微信
新浪微博
Facebook
分享链接
摘要
The current understanding of hydroclimatic processes is largely based on time series analysis of observations such as river discharge. Although records of these variables are often nonlinear and nonstationary, they have been commonly analyzed by classical methods designed for linear and/or stationary data. This study investigates the possibility of analyzing hydroclimatic time series using a novel data-driven method named Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), which is suitable for nonlinear and nonstationary signals. CEEMDAN is here applied to a monthly mean discharge record (1904-2010) of the Parana River (South America). The results obtained in this way are interpreted by comparing them with CEEMDAN decompositions of other records such as climate index time series. It is found that Parana flow modes consist of (i) annual and intraannual oscillations reflecting the rainfall seasonality of different Parana Basin sectors, and (ii) interannual to interdecadal changes linked to climate cycles like El Nino/Southern Oscillation, the North Atlantic Oscillation, and the Interdecadal Pacific Oscillation. A nonlinear trend of Parana discharge is found and reveals a monotonic increase that could be attributed to global warming and anthropogenic land-cover changes. The spectral separation of modes obtained using CEEMDAN is cleaner than that achieved by the Ensemble Empirical Mode Decomposition technique. This makes it easier to interpret CEEMDAN results. Hence, CEEMDAN is proposed as a powerful method for extracting physically meaningful information from hydroclimatic data. Key Points A new empirical decomposition extracts meaningful modes from hydroclimatic data The spectral separation of modes is cleaner than that achieved by EEMD A nonlinear trend of river flow could be linked to global warming
- 出版日期2014-2-16
