If it were possible to properly extract seasonal information from ice-core isotopic records, paleoclimate researchers could retrieve a wealth of new information concerning the nature of climate changes and the meaning of trends observed in ice-core proxy records. It is widely recognized, however, that the diffusional smoothing of the seasonal record makes a "proper extraction" very difficult. In this paper, we examine the extent to which seasonal information (specifically the amplitude and shape of the seasonal cycle) is irrecoverably destroyed by diffusion in the firm. First, we show that isotopic diffusion in firn is reasonably well understood. We do this by showing that a slightly modified version of the Whillans and Grootes (1985) theory makes a tenable a priori prediction of the decay of seasonal isotopic amplitudes with depth at the GISP2 site, though a small adjustment to one parameter significantly improves the prediction. Further, we calculate the amplitude decay at various other ice-core sites and show that these predictions compare favorably with published data from South Pole and locations in southern and central Greenland and the Antarctic Peninsula. We then present numerical experiments wherein synthetic ice-core records are created, diffused, sampled, reconstructed and compared to the original. These show that, after, diffusive mixing in the entire firn column, seasonal amplitudes can be reconstructed to within about 20% error in central Greenland but that all information about sub-annual signals is permanently lost there. Furthermore, most of the error in the amplitude reconstructions is due to the unknowable variations in the subannual signal. Finally, we explore how these results can be applied to other locations and suggest that Dye 3 has a high potential fur meaningful seasonal reconstructions, while Siple Dome has no potential at all. An optimal ice-core site for seasonal reconstructions has a high accumulation rate and a low temperature.
