Abstract: Sand and dust storms are significant natural disasters which affect East Asia and China in spring, occurring from March to May. Performances of CMA-CUACE/Dust, an operational Asian sand and dust storm numerical forecasting system of CMA since 2006, are analyzed. It’s found that the model overestimates in Central Asia, underestimates in northern Mongolia, and diffuses too quickly in downwind areas far away from the source area, especially in Northern China, Korean Peninsula and Japan, resulting in low peak values or less lingering time there for very extreme sand and dust storm events. A scale-adaptive dust emission scheme is applied by resolving the mean wind speed of the model grid into sectional one which can account for values larger than the mean value by Weibull integration function. This significant aspect is crucial because the dust emission is the third power of the wind in the dust emission scheme. New wind erosion database is also adopted which consists of the updated desertification by using twenty-year surface data and new parameters deduced from site observations in the heart of Gobi Desert, determining the size distribution of the emitted dust together with the soil texture data sampling from main deserts in China.After evaluation for the strongest sand and dust episode of 13-17 in March 2021 in East Asia in the past decade, and the consistent run in the same operational environment from 1 March to 31 May in 2023, it is found that the updated CMA-CUACE/Dust effectively improves disadvantages such as overestimation in Central Asia, underestimation in northern Mongolia, and rapid dissipation in China. The predicted peak dust concentration of the extreme episode closely matches observations in China both in the source area and in Shanghai after a 4-day transportation. Threat score (TS) of three-month forecast run also indicates that the improved model shows good consistency and continuity in forecast results across various forecast lengths. TS for 1-5 days with different forecast lengths is significantly higher than that of the previous operational system and surpasses those of Korean dust model—ADAM (the Asian Dust Aerosol Model). Furthermore, the missing rate is significantly reduced, while the false alarm rate remains almost unchanged. TS for episodes beyond the level of sand and storms are all above 0.3, with some exceeding 0.5. All these findings show that the improved model performs much better than the previous one.