The dependence of dust production on model grid-space resolution is investigated using the Navy's operational mesoscale meteorological model with an imbedded dust emission model. The study covers a 2-week period of strong dust storms in April 1998 in the major dust source area of East Asia. The modeled surface winds at grid resolutions of 20, 40, 60, and 80 km are verified against observational data. At all resolutions the model has a positive bias in wind speed that decreases as resolution increases. Dust fluxes that are proportional to the fourth power of the friction velocity (u*) and the third power of the wind speed are calculated at all four resolutions and compared. Compared with the 20-km resolution u*-driven flux, which is deduced to be the most accurate, the u*-driven flux on the coarser grids overestimate the flux, with the 80 km being 60% higher for individual events and nearly 20% higher in the total dust production for the entire study period. The wind-driven flux misses the smaller events due to the lack of a dependence on stability and wind shear, when compared with the timing of surface dust observations, and has differences of up to 70%, when compared with the 20-km u*-driven flux. Averaging over space and time tends to reduce the differences among grids and might support modeling at coarse resolution. ¿ 2001 American Geophysical Union