Methods to measure gaseous emissions from spatially confined sources often rely on atmospheric dispersion models to relate downwind concentration or flux measurements to the respective emission rate. A backward Lagrangian stochastic (bLS) model was validated, which has been enhanced to incorporate surface dry deposition for trace gases for which this removal process is significant, such as ammonia (NH3). Controlled release experiments were carried out at a grassland site with a parallel release of NH3 and methane (CH4) at a known release rate through a source grid with 36 critical orifices. NH3 and CH4 concentrations and vertical fluxes were measured downwind of the source grid and the bLS model was used to infer the release rates from the measured quantities. The median recovered fractions (ratios between the model-inferred emission rate and the actual release rate) ranged between 0.96 and 1.11 for CH4 and between 0.32 and 0.72 for NH3. The smaller recovered fraction of NH3 compared to CH4 was attributed to NH3 dry deposition loss downwind of the source. Surface deposition velocities were inferred to account for smaller recovered fractions from experiments in which both NH3 and CH4 were measured in parallel. Median surface deposition velocities varied between 1.7 and 5.8 cm s-1 for the individual release days, which correspond to deposition velocities at a commonly used height of 2 m between 0.6 and 1.7 cm s􀀀 1. Compared to literature values this is in the expected range of deposition velocities for grassland sites.