A sand-dust numerical model coupled with MM5(PSU/NCAR mesoscale model) -CUCSDust (CMA Unified Chemistry System-Dust) is introduced. CUCSDust includes such detail physical processes as dust production, transport, growth, coagulation, dry and wet deposition, which can simulate and forecast the initial and sand-dust concentration of sand and dust storm. A parallel scheme is also designed to reduce the run-time which is most important for real-time forecasting. Based on CUCSDust, a numerical model forecasting system is established in northeast Asia, and is applied to China Meteorological Administration (CMA) for operational prediction of sand and dust storm from March, 2006. There were 31 storms occurred in northeast Asian in 2006 and the system nicely predicted 29 storms. A prediction result of severe sand and dust storm occurred in 26-28, March, 2006 is chosen, which shows no difference for sand-dust spatial distribution between model outputs and remote sensing monitoring imagers and indicates that the model system has the capability of real-time forecasting sand and dust storm in northeast Asia.