A novel technique for micro-siting and forecasting appropriate to wind farms has been developed. The modelling approach was tested with a selection of different physical options and initialization fields at a complex geographical location characterised by a contrasting diurnal meteorological cycles. An improvement of 10% in wind speed forecasting was achieved by coupling a mesoscale model to a microscale model; the latter was based on Large Eddy Simulation (LES). The platforms for meso and microscale modelling were the Weather Research and Forecasting (WRF) and OpenFOAM models respectively.