The Analysis Of External Forcing Scale Characteristics Of Time Series And Its Influence On Forecasting Ability

The non-stationary of the climate system has been generally recognized.The root cause of this non-stationary is due to the driving force changes over time.Therefore,the extraction and analysis methods of external forced characteristics have also gained more and more attention.Slow feature analysis(SFA)is an effective method for extracting slow-changing feature from fast-changing signal.Its proposal enriches the means of reconstruction of non-stationary system's driving force signal.The SFA method has also achieved initial success in meteorological applications.There need more researches for the information contained in the driving force signal found by this method.In this paper,we constructed two-dimensional non-stationary system model based on Henon chaotic mapping,and tried to test the ability of reconstructing driving force signal from two-dimensional and complex non-stationary system by SFA method Firstly.Then,we use SFA method to extract the driving force signal from the real non-stationary time series.Analysis of the scale features and physical background of driving force signal by wavelet transform technology.Finally,combining the SFA method,reconstruct the nonlinear time series prediction model.Explore the effect of external compulsion on predicting ability.Through a series of experimental analysis,the following results are obtained:(1)The SFA can successfully extract the driving force signal from the non-stationary time series with one time-varying parameter.We also extracted the driving force signal from the non-stationary time series with two time-varying parameters by SFA and wavelet transform technology.(2)By using SFA method,we reconstructed the driving force signal of Beijing air temperature.Wavelet transformation technique was then used to analyze the scale structure of the derived driving force.We found the driving force consists of six scales.The influence of solar activity and ocean activity on temperature was also been found.(3)Combining the SFA method,reconstruct the nonlinear time series prediction model.We use this model to predict the monthly mean temperature anomaly in Mohe Station.We choose two scales signals from driving force signals and added them into prediction model.The results show that this method can improve the accuracy of the prediction.(4)Adding Nino3 SST and Solar flux index into the reconstruction of nonlinear time series prediction.The results show that adding the real driving force signals can improve the forecasting ability effectively.