Dust cycle is an important component of the earth system and has been explicitly represented in the current earth system models(ESMs).However,there are still large uncertainties in the models representation of dust emission processes.Therefore it is desirable to identify the model bias and thus improve it.Dust emission parameterizations in ESMs are generally evaluated by long-term(one-year or multi-year)simulations of the model in which biases in dust aerosol cannot be easily attributed to the meteorological states or the parameterizations themselves.This study presents a complementary process-oriented evaluation of dust emission parameterizations in the Community Earth System Model(CESM)by applying the model to simulate an observed severe dust storm during 19–22 March 2010 over East Asia.The different parameterizations accounting for the effect of land-surface characteristics are assessed through sensitive experiments with different roughness correction factor(fλ)and source erodibility index(S).The simulation results are compared with each other and evaluated with surface synoptic observations(SYNOP)and station observation of dust concentrations.We show that generally this model can capture the main dust emission regions and the evolution of main dust plumes in Mongolia and northern China,even without considering the constraint of roughness elements(i.e.,fλ set to 1).With the constraint of fλ taken into account,dust emission scheme effectively alleviates or eliminates the overestimation of dust emission over the vegetated area and consequently the model mostly captures better the magnitude of surface dust concentrations at its nearby and downwind stations.Besides,time variation of dust concentration is also closer to the observation at most stations.The incorporation of a geomorphic S(Sg)tends to concentrate the dust emission over the low-lying basins including terrain depressions and river-valleys,thereby producing hot spots of dust distribution which are much stronger than observation.It also shifts some dust emission regions and thus could not capture the evolution of surface dust concentration at stations of Dalanzadgad and Wulatezhongqi,and these biases could be ascribed to the deficiencies of Sg in representing the potential ability of land surface for wind erosion.Overall the inclusion of fλ on the threshold friction velocity could satisfactorily provide realistic representation of the effect of land surface characteristics for the severe dust storm.