Spatiotemporal Decomposition Pattern Of Spatiotemporal Datasets Based On A Combination Of Empirical Mode Decomposition With Empirical Orthogonal Function

Decomposition of spatiotemporal pattern is an important method in the analysis on geophysical spatiotemporal datasets.Usually,empirical orthogonal function(EOF)is the most effective decomposition method with the characteristic of biorthogonality.For either stationary or non-stationary spatiotemporal series,in the past several decades,investigators have decomposed the two kinds of series into different spatiotemporal patterns using EOF.However,a careful analysis showed that the decomposition patterns of geophysical spatiotemporal series in the presence of non-stationary are unstable since the variance and covariance of the series change with time.And these patterns cannot be used directly to predict the observed variables.Additionally,the portion of temporal field decomposition through classical EOF consists of a variety of intrinsic signals with different oscillation frequency,which fails to build the physical relationship between oscillation frequency signal and external influencing factors.In this paper,based on a combination of empirical orthogonal function method with empirical mode decomposition,a new decomposition method,i.e.,emperical orthogonal decomposition based on EMD,abbreviated as EODE,is first proposed.And then,this method is utilized to decompose spatiotemporal patterns of monthly mean precipitation in Taihu Lake Basin,China.As a result,the reliability of the method is verified by several experiments of artificial dataset and precipitation records.The main topics of this paper are discussed as follows:1)Analysis on the characteristics of the existing decomposed methods on geophysical spatial-temporal datasets.Firstly,this paper recalled the concept and fundamental theory of empirical mode decomposition and empirical orthogonal function.And then,compared the disadvantages and the advantages of EOF and EMD.Next,summarized the key points of these two methods,laid the theoretical framework of the proposed decomposition method in the end.2)Proposition of the geophysical spatial-temporal decomposition method.The patterns through empirical orthogonal function showed that the average or weighted average of spatial-temporal series is usually one of the main temporal patterns.Therefore,it is assumed that the average of spatial temporal series in the entire study area is the dominant temporal pattern.First,we decomposed the average series into a variety of intrinsic mode functions(IMFs)and a residue by means of EMD.And then,these IMFs are viewed to be independent variables.Next,the association between time series and intrinsic mode functions in every geographic observation station is obtained,describing spatial differentiation of the geographic observation variables.3)Numerical simulation and instance validation of the proposed spatiotemporal pattern decomposition method.Firstly,relevant and irrelevant spatial-temporal datasets are generated through ARFIMA.The proposed method is then applied to decompose the artificically geophysical spatial-temporal series.The result showed that spatiotemporal patterns of artificial datasets are completely consistent with theoretical analysis.Next,we employed the proposed method to decompose the monthly mean precipitation in Taihu Lake Basin from January 1957 to May 2015.The acquired patterns in Taihu Lake Basin are verified by the other known decomposition approaches.The investigation results show that the proposed decomposition method is reliable and can be applicable for geophysical spatiotemporal datasets in the presence of non-stationary and long-range correlation.The method has the capacity to extract more hidden time-frequency information on spatiotemporal patterns.