The performance of numerical weather prediction (NWP) systems is generally poorer at higher latitudes compared to other regions (e.g. Jung et al., 2016). This is due to various reasons, including scarcity of in situ observations and challenges related to utilization of satellite remote sensing observations, such as the polar night, lack of data from geostationary satellites, and difficulties in interpreting microwave signals originating from snow, ice, and clouds (Scarlat et al., 2018). Other reasons are complexity of surface boundary conditions due to rapidly changing sea ice concentration (Valkonen et al., 2014), rapidly developing mesoscale systems, such as polar lows and explosive cyclones (Pezza et al., 2016), and complexity of subgrid-scale physical processes related, among others, to stable boundary layer (Nigro et al., 2017), mixed-phase clouds