Abstract: The interpolation results of 2 m temperature forecast products produced by the meso-scale operational model in winter of 2014 in Liaoning province were corrected using methods of 7-day bias correction (7DBC) and running mean bias correction (RMBC).The correction results were compared to products of MOS (model output statistics) forecast.The forecast accuracy of the three methods,MOS forecast,7DBC and RMBC,were analyzed.The results show that forecast accuracy of the two bias correction methods is higher than that of NWP (numerical weather prediction) interpolation method.The RMBC method is better than the 7DBC method.As to 24 h maximum temperature forecast,the forecast accuracy of the RMBC method is the highest.For minimum temperature forecast,the forecast accuracy of RMBC is higher than that of NMC (National Meteorological Center) MOS forecast at 08:00,while less than MOS forecast at 20:00.The method of RMBC needs accumulated data of 1 to 15 days.Compared to MOS method,the method of RMBC needs less data and is easy to operate,which is more suitable for areas without long-term records to conduct numerical model revisions for temperature.