This paper presents and evaluates an inverse model for estimating ammonia emission from agricultural land. The method is based on an analytical model derived from the advection-diffusion equation, assuming power law profiles for wind speed and diffusivity. A three-dimensional model and a two-dimensional model are evaluated. The hypotheses of flux-driven or concentration-driven emissions are also tested. The model is evaluated against three datasets covering a range of ammonia fluxes, field geometry/size and measurement techniques. The sensitivity and the uncertainty of the method is also evaluated with a MonteCarlo approach, as well as based on existing datasets. Finally, the capability of the method to work with time-integrated concentrations (e.g. using diffusive concentration samplers) is also evaluated. The inverse model gives estimations of the ammonia emissions within a few per cent of the measurements. Moreover, the method is mainly sensitive to the concentration, the friction velocity and the thermal stratification of the atmosphere. The two-dimensional approaches give similar results to the three-dimensional one, provided the field is large enough. The concentration-driven hypothesis is similar to the flux-driven hypothesis for a fetch greater than approximately 20 m. The results are discussed in comparison with the previous approaches: the Theoretical Profile Shape (TPS or Zinst approach) and the backward Lagrangian Stochastic model (BLS).