ABSTRACT:A new downscaling method has been developed to improve forecasts of near-surface temperature. This involves applying a correction to forecast temperatures to account for the difference in height between the terrain in the forecast model and the real terrain, using an estimate of the lapse rate of temperature. The strongest variations in lapse rate are found to occur overnight, being a function of cloud cover and geostrophic wind speed. These variables similarly influence the night time lapse rate in the UK Met Office Unified Model forecast at 4 km resolution. Therefore, a simple algorithm has been employed to estimate the lapse rate at a given location, based on forecast model temperature and elevation fields. The algorithm has a positive impact on performance in stable conditions, compared to assuming an adiabatic lapse rate at all times. Crown