Modelling firn thickness evolution during the last deglaciation: constraints on sensitivity to temperature and impurities

The transformation of snow into ice is a complex phenomenon that is difficult to model. Depending on surface temperature and accumulation rate, it may take several decades to millennia for air to be entrapped in ice. The air is thus always younger than the surrounding ice. The resulting gas-ice age difference is essential to documenting the phasing between CO2 and temperature changes, especially during deglaciations. The air trapping depth can be inferred in the past using a firn densification model, or using delta N-15 of air measured in ice cores. All firn densification models applied to deglaciations show a large disagreement with delta N-15 measurements at several sites in East Antarctica, predicting larger firn thickness during the Last Glacial Maximum, whereas delta N-15 suggests a reduced firn thickness compared to the Holocene. Here we present modifications of the LGGE firn densification model, which significantly reduce the model-data mismatch for the gas trapping depth evolution over the last deglaciation at the coldest sites in East Antarctica (Vostok, Dome C), while preserving the good agreement between measured and modelled modern firn density profiles. In particular, we introduce a dependency of the creep factor on temperature and impurities in the firn densification rate calculation. The temperature influence intends to reflect the dominance of different mechanisms for firn compaction at different temperatures. We show that both the new temperature parameterization and the influence of impurities contribute to the increased agreement between modelled and measured delta N-15 evolution during the last deglaciation at sites with low temperature and low accumulation rate, such as Dome C or Vostok. We find that a very low sensitivity of the densification rate to temperature has to be used in the coldest conditions. The inclusion of im-purity effects improves the agreement between modelled and measured delta N-15 at cold East Antarctic sites during the last deglaciation, but deteriorates the agreement between modelled and measured delta N-15 evolution at Greenland and Antarctic sites with high accumulation unless threshold effects are taken into account. We thus do not provide a definite solution to the firnification at very cold Antarctic sites but propose potential pathways for future studies.