Performance of meso-scale numerical weather prediction model HARMONIE-AROME was assessed for a summer-time heatwave in a high-latitude coastal city of Turku in SW Finland. Representativeness of the model's ECOCLIMAP-SG based physiographic land cover specification with a resolution of 750 m was assessed against the 20 m resolution CORINE Land Cover (CLC) dataset. Moreover, the modelled 2 m air temperatures were compared with temperature observations recorded at 74 sites of the local climate observation network. Correlation analysis between the model's physiographic data (PGD) and CLC shows a statistically significant (p ≤ 0.01) match between the datasets in 76% of the land cover types, indicating mainly a good representativeness of the HARMONIE-AROME's land cover parameterization in the study area. The cases of mismatch are mostly related to the differences in ECOCLIMAP-SG and CLC classification principles and/or resolution, and thus can not be considered to indicate poor representativeness of the ECOCLIMAP-SG classification in the area. The HARMONIE-AROME modelled temperatures were on average slightly higher than the observed temperatures, and the difference was largest for daily minimum temperatures. Heterogeneity of land cover and topography inside a 750 m grid cell was related to larger difference between the modelled and observed temperature. •The T modelled by HARMONIE-AROME was accurate; only 0.8 °C warmer than the observed T.•The largest site-specific differences were 2.1 °C / 0.4 °C (same/opposite direction).•Diurnally, the difference between the modelled and observed T was largest at night.•ECOCLIMAP-SG reflects well the land cover (CLC) in the study area.•The best correspondence between ECOCLIMAP-SG and CLC was on natural uninhabited areas.