Currently, very simple parameterizations for the seasonal variation of the ammonia (NH3) emissions are used in most air pollution models. This limits the capability to reproduce observed seasonal variations in the NH3 concentrations in rural regions with agricultural activity. A dynamical parameterization of the temporal variation of NH3 emission for application into air pollution models has therefore been proposed. The obtained improvements are demonstrated by implementing a dynamical NH3 emission parameterization into the large-scale air pollution model ACDEP. Meteorology, information about agricultural practice, and a simple crop growth model are the basis for the parameterization. Fifteen additive functions are used to describe the different parts of the NH3 emissions from agriculture, and a 16th function is used to describe NH3 emission from catalysts used in road traffic. The new parameterization is tested by comparison of model results with NH3 measurements from Danish monitoring stations situated in agricultural areas during the years 1999--2001. Improvements in correlation coefficient of calculated diurnal mean concentrations from 0.42 to 0.70 have been achieved, and a monthly correlation coefficient up to 0.98 has been obtained for the Tange station in 2000. This validation shows that the new parameterization gives a substantial improvement in the model calculations of the NH3 concentrations in areas with high agricultural activities. Finally, the validation showed that optimum results are obtained by including a temperature correction on the gauss functions representing the emission sources.