| Popis: |
article i nfo Analytic climate models have provided the means to predict potential impacts on future climate by anthro- pogenic changes in atmospheric composition. However, future climate development will not only be influ- enced by anthropogenic changes, but also by natural variations. The knowledge on such natural variations and their detailed character, however, still remains incomplete. Here we present a new technique to identify the character of natural climate variations, and from this, to produce testable forecast of future climate. By means of Fourier and wavelet analyses climate series are decomposed into time-frequency space, to extract information on periodic signals embedded in the data series and their amplitude and variation over time. We chose to exemplify the potential of this technique by analysing two climate series, the Svalbard (78°N) sur- face air temperature series 1912-2010, and the last 4000 years of the reconstructed GISP2 surface tempera- ture series from central Greenland. By this we are able to identify several cyclic climate variations which appear persistent on the time scales investigated. Finally, we demonstrate how such persistent natural vari- ations can be used for hindcasting and forecasting climate. Our main focus is on identifying the character (timing, period, amplitude) of such recurrent natural climate variations, but we also comment on the likely physical explanations for some of the identified cyclic climate variations. The causes of millennial climate changes remain poorly understood, and this issue remains important for understanding causes for natural cli- mate variability over decadal- and decennial time scales. We argue that Fourier and wavelet approaches like ours may contribute towards improved understanding of the role of such recurrent natural climate variations in the future climate development. |
Full Text from ScienceDirect