http://www.iceandclimate.nbi.ku.dk/rese ... diffusion/Diffusion: The limit to δ18O-based ice core dating
Snow is slowly compressed into ice in the upper 80 meters of an ice sheet (read more about the process here). During this process, water vapour can move relative to the ice in the open pores between the snow grains, thereby smoothing the annual δ18O cycles. This diffusion process smoothes the δ18O signal and even erases the annual signal if the annual layers are thinner than 15-20 cm. In ice cores from sites with less than 15 cm of precipitation (measured in equivalents of compacted ice, not snow) per year, the annual cycle in δ18O will be obliterated, and dating based on annual δ18O oscillations is therefore not possible. This is the case for areas in north-eastern Greenland where the annual precipitation rate is significantly lower than 20 cm. For ice cores drilled in areas with about or slightly more than 20 cm of precipitation, diffusion will also blur the annual cycles, but it is possible to retrieve the annual cycle using diffusion correction techniques.
Very slow diffusive processes also take place deeper in the ice sheets. These processes slowly weaken the annual δ18O oscillations as the ice gets older and the layers thin due to the flow of ice.
Due to the diffusion processes,
the limit of safe annual layer detection using δ18O / δD measurements is about 8500 years ago in the DYE-3 ice core. More favourable conditions at the summit of the Greenland ice sheet has permitted successful identification of annual layers from δ18O data in more than 14,000 year old ice from the GRIP ice core, while the NGRIP and NEEM ice cores cannot in general be dated using δ18O data alone.