Robust estimation of background noise and signal detection in climatic time series

摘要: We present a new technique for isolating climate signals in time series with characteristic ‘red’ noise background which arises from temporal persistence. This is estimated by ‘robust’ procedure that, unlike conventional techniques, largely unbiased the presence of immersed noise. Making use multiple-taper spectral analysis methods, further provides distinction between purely harmonic (periodic) signals, and broader-band (‘quasiperiodic’) signals. The effectiveness our signal detection demonstrated synthetic examples that simulate variety possible periodic quasiperiodic red apply methodology to historical paleoclimate examples. Analysis ≈ 3 million year sediment core reveals significant components at known astronomical forcing periodicities 100 peak. roughly 1500 tree-ring reconstruction Scandinavian summer temperatures suggests on near-century timescale, an interdecadal 16–18 within interannual El Nino/Southern Oscillation (ENSO) band, quasibiennial timescale. 144 record Great Salt Lake monthly volume change broad band variability, ENSO-timescale peaks, annual cycle its harmonics. Focusing detail global-average surface temperature record, we find highly secular trend relative background, weakly ENSO band. Decadal are marginally this series.