Abstract. Correct estimation of the firn lock-in depth is essential for correctly linking gas and ice chronologies in ice core studies. Here, two approaches to constrain the firn depth evolution in Antarctica are presented over the last deglaciation: outputs of a firn densification model, and measurements of δ¹⁵N of N₂ in air trapped in ice core, assuming that δ¹⁵N is only affected by gravitational fractionation in the firn column. Since the firn densification process is largely governed by surface temperature and accumulation rate, we have investigated four ice cores drilled in coastal (Berkner Island, BI, and James Ross Island, JRI) and semi-coastal (TALDICE and EPICA Dronning Maud Land, EDML) Antarctic regions. Combined with available ice core air-δ¹⁵N measurements from the EPICA Dome C (EDC) site, the studied regions encompass a large range of surface accumulation rates and temperature conditions.

Our δ¹⁵N profiles reveal a heterogeneous response of the firn structure to glacial–interglacial climatic changes. While firn densification simulations correctly predict TALDICE δ¹⁵N variations, they systematically fail to capture the large millennial-scale δ¹⁵N variations measured at BI and the δ¹⁵N glacial levels measured at JRI and EDML – a mismatch previously reported for central East Antarctic ice cores.

New constraints of the EDML gas–ice depth offset during the Laschamp event (~41 ka) and the last deglaciation do not favour the hypothesis of a large convective zone within the firn as the explanation of the glacial firn model–δ¹⁵N data mismatch for this site. While we could not conduct an in-depth study of the influence of impurities in snow for firnification from the existing datasets, our detailed comparison between the δ¹⁵N profiles and firn model simulations under different temperature and accumulation rate scenarios suggests that the role of accumulation rate may have been underestimated in the current description of firnification models.