Emissions of nitrogen oxide (NOx = NO + NO2) from the photolysis of nitrate (NO−3 ) in snow affect the ox- idising capacity of the lower troposphere especially in re- mote regions of high latitudes with little pollution. Current air–snow exchange models are limited by poor understand- ing of processes and often require unphysical tuning param- eters. Here, two multiphase models were developed from physically based parameterisations to describe the interac- tion of nitrate between the surface layer of the snowpack and the overlying atmosphere. The first model is similar to previous approaches and assumes that below a thresh- old temperature, To, the air–snow grain interface is pure ice and above To a disordered interface (DI) emerges cov- ering the entire grain surface. The second model assumes that air–ice interactions dominate over all temperatures be- low melting of ice and that any liquid present above the eutectic temperature is concentrated in micropockets. The models are used to predict the nitrate in surface snow con- strained by year-round observations of mixing ratios of ni- tric acid in air at a cold site on the Antarctic Plateau (Dome C; 75◦06′ S, 123◦33′ E; 3233 m a.s.l.) and at a rel- atively warm site on the Antarctic coast (Halley; 75◦35′ S, 26◦39′ E; 35 m a.s.l). The first model agrees reasonably well with observations at Dome C (Cv(RMSE) = 1.34) but per- forms poorly at Halley (Cv(RMSE) = 89.28) while the sec- ond model reproduces with good agreement observations at both sites (Cv (RMSE) = 0.84 at both sites). It is therefore suggested that in winter air–snow interactions of nitrate are determined by non-equilibrium surface adsorption and co- condensation on ice coupled with solid-state diffusion inside the grain, similar to Bock et al. (2016). In summer, however, the air–snow exchange of nitrate is mainly driven by solvation into liquid micropockets following Henry’s law with contributions to total surface snow NO−3 concentrations of 75 and 80 % at Dome C and Halley, respectively. It is also found that the liquid volume of the snow grain and air– micropocket partitioning of HNO3 are sensitive to both the total solute concentration of mineral ions within the snow and pH of the snow. The second model provides an alter- native method to predict nitrate concentration in the surface snow layer which is applicable over the entire range of envi- ronmental conditions typical for Antarctica and forms a basis for a future full 1-D snowpack model as well as parameteri- sations in regional or global atmospheric chemistry models.