JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101,NO. C7, PAGES 16,401-16,419, JULY 15, 1996 Observation of autumn freeze-up in the Beaufort and Chukchi Seas using the ERS I synthetic aperture radar D. P. Winebrenner Applied Physics Laboratory, College of Ocean andFishery Sciences, University of Washington Seattle B. Holt Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California E.D. Nelson• Applied Physics Laboratory, College of Ocean andFishery Sciences, University of Washington Seattle Abstract. We present observations of the transition in sea ice backscattering signatures from late summer into early winter in the Beaufort and Chukchi Seas during1991 and 1992, using data from theERSI synthetic aperture radar (SAR). Weemploy both analyzed surface temperature fields and direct observations of near-surface temperature from drifting buoys. Consistent with previous surface-based observations, backscattering from sea ice surviving at theend of summer increases strongly and rapidly when temperatures fall below freezing forthe final timein theautumn, apparently because of increased volume scattering from bubbles in the upper part of the ice. Areally averaged backscattering sometimes overshoots typical winter multiyear ice values by 1-2 dB, before settling back over a period of approximately a week. The backscattering-temperature link forms a sufficient basis foran algorithm to estimate retrospectively, from timeseries of SARimages, the date of freeze-up (defined here to bethedate onwhich all liquid water in thebubbly upper layer of surviving sea ice freezes and remains frozen forthe duration of theensuing autumn and winter). We present a prototype algorithm and use it to estimate freeze-up dates in nine Lagrangian cells in' the Beaufort Sea during the autumn of 1992.We observe a 12-day spread in freeze-up dates between latitudes of approximately 73øN and 82øN, withdates in the northernmost cells of August 29-30 andthose in the southernmost cells of September 7. Two cells at latitudes of 75øN-77øN appear to freeze-up earlierthan cells further north but present uncertainties limit the significance of thisobservation. Introduction The two most significant seasonal transitions under- gone by Arctic seaice are the onset of melting in the spring and the end of the melt season in autumn. The onset of snow melt marks the beginningof a sharp re- duction of surface albedo from winter values and thus increased absorption and heating of the iceby shortwave radiation [Maykut, 1986]. The decrease in air temper- aturesbelowfreezing at the end of summer marksthe autumn freeze-up and thus the end of freshwater input to the upperocean, an increase in surface albedo, and, in most areas, the minimum extent of the ice cover. Taken together, thesetwo seasonal transitions define Now at U.S. West NewVector Group, Englewood, Colorado. Copyright 1996 by the AmericanGeophysical Union. Paper number 96JC01292. 0148-0227/96/96JC-01292$09.00 the length of the melt season (or, alternatively, the length of the ice growth season), which is an impor- tant environmental parameter in determining the ice massbalanceand polar energybudgets[Maykut and Untersteiner, 1971;Ebert and Curry, 1993].Variability in the length of the melt season accounts for much of the variability in modeledaverage annual ice thickness [Hiikkinen andMellor,1990]. Changes in the Arctic ice cover associated with sum- mer melting produce significantchanges in microwave and opticalsignatures from their (typically stable) win- ter values [Carsey, 1985; Holt and Digby, 1985; Drinkwa- ter and Carsey, 1991; Gogineniet al., 1992; Robinson et al., 1992; Schwartz et al., 1994; Winebrenner et al., 1994]. The 10-year study by Robinson et al. showed significant regional and interannual differences in the timing of melt. Variations in ponding, drainage, ice de- cay, and wind roughening of openwater are apparently responsible for variable optical and microwavesigna- tures throughout the summer [Gogineni et al., 1992]. 16,401