In-cloud ice mass accretion on wind turbines is a common challenge that is faced by energy companies operating in cold climates. On-shore wind farms in Scandinavia are often located in regions near patches of forest, the heterogeneity length scales of which are often less than the resolution of many numerical weather prediction (NWP) models. The representation of these forests—including the cloud water response to surface roughness and albedo effects that are related to them—must therefore be parameterized in NWP models used as meteorological input in ice prediction systems, resulting in an uncertainty that is poorly understood and, to the present date, not quantified. The sensitivity of ice accretion forecasts to the subgrid representation of forests is examined in this study. A single column version of the HARMONIE-AROME three-dimensional (3D) NWP model is used to determine the sensitivity of the forecast of ice accretion on wind turbines to the subgrid forest fraction. Single column simulations of a variety of icing cases at a location in northern Sweden were examined in order to investigate the impact of vegetation cover on ice accretion in varying levels of solar insolation and wind magnitudes. In mid-winter cases, the wind speed response to surface roughness was the primary driver of the vegetation effect on ice accretion. In autumn cases, the cloud water response to surface albedo effects plays a secondary role in the impact of in-cloud ice accretion, with the wind response to surface roughness remaining the primary driver for the surface vegetation impact on icing. Two different surface boundary layer (SBL) forest canopy subgrid parameterizations were tested in this study that feature different methods for calculating near-surface profiles of wind, temperature, and moisture, with the ice mass accretion again following the wind response to surface vegetation between both of these schemes.