Several studies have shown the importance of aerosols in the Earth's radiative balance. The radiative effect of dust has earlier been quantified in global and regional climate models. Fewer studies have included prediction of aerosols online in numerical weather prediction (NWP) models. Predicting climate effect of aerosols is different from including aerosols actively in weather prediction because climate models give average responses over long timescales, whereas the purpose of weather prediction models is to calculate the right weather on short timescales. In this paper, we run a mesoscale NWP with four different assumptions on dust aerosols: (1) no dust, (2) climatological data set for dust aerosols, (3) forecasted dust using the Dust Entrainment and Deposition model (DEAD) and (4) same as assumption 3 but with decreased single scattering albedo. We show that the assumption on dust is important for predicting ground temperature and convective precipitation in the model. Dust changes the model physics through changing the radiative fluxes. We interpret the changes through analyzing the energy budgets for four zones close to the major dust sources. The zones include both convective and nonconvective areas. In convective areas over ocean, dust can decrease convective activity. The vertical gradient of the aerosols and their single scattering albedo determine how efficient they are. Over land, dust also influences the surface energy budget so that decreased latent heat fluxes result if dust plumes pass over areas where water evaporates from the surface.