Intrinsic feature extraction in the COI of wavelet power spectra of climatic signals

Since the wavelet power spectra are distorted at data boundaries (the cone of influence, COI), using traditional methods, one cannot judge whether there is a significant region in COI or not. In this paper, with the help of a first-order autoregressive (AR1) extension and using our simple and rigorous method, we can obtain realistic significant regions and intrinsic feature in the COI of wavelet power spectra. We verify our method using the 300 year record of ice extent in the Baltic Sea.