Desert dust is an important aerosol component that produces large uncertainties in assessments of Earth’s radiative budget and global climate change. However, current global climate model (GCM) simulations show that modeled dust poorly captures the spatiotemporal variability of observed dust, inhibiting accurate assessments of aerosol radiative effects. Furthermore, dust emission is a local-scale process varying on scales < 1–10 km, and thus GCMs with typical grid-scales of > 100 km inherently have difficulties capturing dust spatial distribution and its sensitivity to local-scale meteorological variability. To tackle these problems, we develop a new dust emission scheme that includes more aeolian processes, and use the Community Earth System Model (CESM2) as a case study. First, we account for the dissipation of surface wind momentum by roughness elements including plants and rocks, which reduce the wind momentum exerted on the bare soil surface over deserts. Second, we account for the effects of soil particle size distribution (PSD) on dust emission threshold by implementing a realistic soil median diameter inferred from multiple soil PSD observations. Third, we account for intermittent dust emissions induced by boundary-layer turbulence, which further couples dust with boundary-layer dynamics. With more aeolian processes, CESM2 dust emission matches better against satellite retrievals in spatial variability, seasonality, and activation frequency. Modeled dust aerosol optical depth also shows better agreement in spatiotemporal correlations with satellite-derived and ground-based AOD observations. In addition to improving the aeolian processes, we conduct simulations across multiple grid resolutions and show that the high-resolution simulations generally produce a better dust spatial distribution. We then generate a correction map to dust emissions for the coarse-gridded simulations to reduce the scale-dependency of dust emission parameterizations, and results indicate further improvements for coarse-gridded CESM2. Our results suggest including more physical processes in GCMs can lessen bias, improve simulation results, and eliminate the use of empirical source functions. Thus, our scheme could improve assessments of dust impacts on the Earth system and future climate changes.