Modelling firn density at Dye-2 and KAN_U, two sites in the percolation zone of the Greenland ice sheet

Abstract. Modelling firn density is crucial for understanding the current mass balance of the Greenland ice sheet (GrIS) and predicting its future. As snowmelt increases in a warming climate on the GrIS, accurate information about firn density and its variability over time and space becomes increasingly important in the percolation zone. Previous research indicates that none of the existing models accurately simulate firn properties at sites with varying snowmelt rates due to the limited knowledge of liquid water percolation and the complexity of firn densification. Here, we enhance the representation of the firn densification model based on the Community Firn Model (CFM) by (i) using the stressed-based dry-firn densification scheme and recently published parametrizations for firn characteristics, (ii) modifying the expression of irreducible water in the Darcy-flow scheme, and (iii) allowing the time step to be adaptive. The improved model is employed at two climatologically distinct sites, Dye-2 and KAN_U, two sites in the southwest percolation zone of the GrIS, to evaluate its performance. The modelled firn depth-density profiles at the two study sites generally agree well with the in situ measurements obtained from 16 firn cores drilled between 2012 and 2019. At Dye-2, with comparatively high accumulation and low snowmelt rates, the model simulates thin ice lenses and/or ice layers in the top 10 m of the firn column. The modelled firn density aligns with the average level of observations, with the relative bias in density ranging from 0.36 % to 6 %. At the KAN_U site, characterized by relatively low accumulation and high snowmelt rates, the model captures high-density layers (~917 kg · m^-3) caused by the refreezing of liquid water. The observed ice slabs are partly reproduced, and the relative bias in density between simulations and observations at all 8 cores is within ±5 %.