Oscillatory behavior in Arctic sea ice concentrations

We have done a frequency analysis of the 9-year record of sea ice concentrations obtained with the NASA Nimbus 7 scanning multichannel microwave radiometer (SMMR). Because of the periodic revisit time of the SMMR, we are able to examine recurrence of shorter-period oscillations, as aliased by the revisit frequency, and longer-period oscillations per se. A number of findings result from a sequence of two analysis schemes, as follows. We have first determined, then removed, the mean, linear trend, and seasonal cycle (the latter composed of five harmonics of the annual cycle) from the time series for each pixel of gridded SMMR ice concentration data in the Arctic, using multiple linear regression. We find that the 9-year trends are most positive in the Beaufort and East Greenland Seas, whereas the most negative areas are in the Barents Sea and the sea ice surrounding Svalbard; this is similar to trend patterns in the lengths of the sea ice season reported earlier by Parkinson [1992]. In the residual data the standard deviation ranges from 12% to 16% in the central Arctic basin, considerably above the estimated wintertime ice concentration accuracies of 5% in that region; this is in part due to fluctuations in the ice concentration rather than noise. Secondly, we have used Fourier analysis and multiple-taper filtering techniques to identify tentatively some of these fluctuations as tidal components or multiyear components similar to those found in the El Nino-Southern Oscillation Index, and we show their spatial distribution. However, there are some high-amplitude, periodic components that we have not yet been able to identify, we show the spatial distribution of one of these components.