Volcanic eruptions can cause damage to land and people living nearby, generate high concentrations of toxic gases, and also create large plumes that limit observations and the performance of forecasting models that rely on these observations. This study investigates the use of micro- to meso-scale simulation to represent and predict the convection, transport, and deposit of volcanic pollutants. The case under study is the 2007 eruption of the Piton de la Fournaise, simulated using a high-resolution, coupled lava/atmospheric approach (derived from wildfire/atmosphere coupled code) to account for the strong, localized heat and gaseous fluxes occurring near the vent, over the lava flow, and at the lava–sea interface. Higher resolution requires fluxes over the lava flow to be explicitly simulated to account for the induced convection over the flow, local mixing, and dilution. Comparisons with air quality values at local stations show that the simulation is in good agreement with observations in terms of sulfur concentration and dynamics, and performs better than lower resolution simulation with parameterized surface fluxes. In particular, the explicit representation of the thermal flows associated with lava allows the associated thermal breezes to be represented. This local modification of the wind flow strongly impacts the organization of the volcanic convection (injection height) and the regional transport of the sulfur dioxide emitted at the vent. These results show that explicitly solving volcanic activity/atmosphere complex interactions provides realistic forecasts of induced pollution.