Non-Linear Analysis Methods Applied To Observational And Simulated Climatic Time Series

The results of a simulated CO2 (C) and a global ice volume (V) time series, derived from a simple relaxation model of the glacial-interglacial cycles[1], have been analyzed using non-linear techniques. On a first approximation, we have compared simulated time series with the corresponding observational time series, obtaining correlations of 0.88 between the proxy-record delta O-18([2]) and simulated V, and 0.79 between the reconstructed atmospheric CO2 concentration([2-7]) and simulated C, used to quantify the maximum observational variance that this simple model is able to explain.Fourier transform, wavelet transform, cross-wavelet transform, wavelet coherence and cross-recurrence analysis are useful tools to quantify the performance of a model at reproducing the dynamics embedded in observational time series. The analysis shows that the model reproduces closely the dynamics embedded in the ice volume time series, but for the CO2 case the coherence between simulated and observational CO2 is only sporadic, indicating that both time series do not follow the same dynamical behavior, although, in the deglacial periods the two carbon series become dynamically close.The analysis reinforces the hypothesis that some specific mechanisms included in the model are able to closely reproduce the glacial-interglacial oscillations and thus suggests which specific mechanisms should be more seriously investigated in the climate system. These techniques may be applied to other climatic time series to quantify the performance of a model simulating the dynamics of the climate system.