(Table 1) Seasonal sea ice concentration and the observed changes from 1972 to 2007 for the Southern Beaufort Sea

A reliable data set of Arctic sea ice concentration based on satellite observations exists since 1972. Over this time period of 36 years western arctic temperatures have increased; the temperature rise varies significantly from one season to another and over multi-year time scales. In contrast to most of Alaska, however, on the North Slope the warming continued after 1976, when a circulation change occurred, as expressed in the PDO index. The mean temperature increase for Barrow over the 36-year period was 2.9°C, a very substantial change. Wind speeds increased by 18% over this time period, however, the increase were non-linear and showed a peak in the early 1990s. The sea ice extent of the Arctic Ocean has decreased strongly in recent years, and in September 2007 a new record in the amount of open water was recorded in the Western Arctic. We observed for the Southern Beaufort Sea a fairly steady increase in the mean annual amount of open water from 14% in 1972 to 39% in 2007, as deduced from the best linear fit. In late summer the decrease is much larger, and September has, on average, the least ice concentration (22%), followed by August (35%) and October (54%). The correlation coefficient between mean annual values of temperature and sea ice concentration was 0.84. On a monthly basis, the best correlation coefficient was found in October with 0.88. However, the relationship between winter temperatures and the sea ice break-up in summer was weak. While the temperature correlated well with the CO2 concentration (r=0.86), the correlation coefficient between CO2 and sea ice was lower (r=-0.68). After comparing the ice concentration with 17 circulation indices, the best relation was found with the Pacific Circulation Index (r=-0.59).

Alkenones and SST of surface sediments (Appendix)

We have analysed alkenones in 149 surface sediments from the eastern South Atlantic in order to establish a sediment-based calibration of the U37K' paleotemperature index. Our study covers the major tropical to subpolar production systems and sea-surface temperatures (SST's) between 0° and 27°C. In order to define the most suitable calibration for this region, the U37K' values were correlated to seasonal, annual, and production-weighted annual mean atlas temperatures and compared to previously published culture and core-top calibrations. The best linear correlation between U37K' and SST was obtained using annual mean SST from 0 to 10 m water depth (U37K' = 0.033 T + 0.069, r**2 = 0.981). Data scattering increased significantly using temperatures of waters deeper than 20 m, suggesting that U37K' reflects mixed-layer SST and that alkenone production at thermocline depths was not high enough to significantly bias the mixed-layer signal. Regressions based on both production-weighted and on actual annual mean atlas SST were virtually identical, indicating that regional variations in the seasonality of primary production have no discernible effect on the U37K' vs. SST relationship. Comparison with published core-top calibrations from other oceanic regions revealed a high degree of accordance. We, therefore, established a global core-top calibration using U37K' data from 370 sites between 60°S and 60°N in the Atlantic, Indian, and Pacific Oceans and annual mean atlas SST (0-29°C) from 0 m water depth. The resulting relationship (U37K' = 0.033 T + 0.044, r**2 = 958) is identical within error limits to the widely used E. huxleyi calibrations of and attesting their general applicability. The observation that core-top calibrations extending over various biogeographical coccolithophorid zones are strongly linear and in better accordance than culture calibrations suggests that U37K' is less species-dependent than is indicated by culture experiments. The results also suggest that variations in growth rate of algae and nutrient availability do not significantly affect the sedimentary record of U37K' in open ocean environments.