Volatilization of ammonia constitutes a globally significant pathway of nitrogen losses in agricultural ecosystems with important impacts on air pollution and climate. The future ammonia emissions will be impacted by changes in global food demand and climate, and estimating these impacts requires a process based model. This paper describes the development and application of the FAN process model, which is incorporated into the land surface component of an earth system model. FAN evaluates ammonia volatilization mechanistically in response to soil temperature, moisture, pH and agricultural practices. A comparison with experimental data shows that FAN can reproduce the different volatilization rates of different types of manures and synthetic fertilizers in a physically meaningful way. The resulting global ammonia emissions and their atmospheric impacts are evaluated in a five-year present-day simulation coupling the emissions to an atmospheric model. Compared to current emission inventories, the simulated emissions are within the range of existing estimates for Europe, North America and China, but about 50-70% higher for India and Latin America, and a factor of 2-3 higher than the current estimates for Africa. The simulated airborne ammonia and ammonium concentrations and wet depositions are compared with a global set of in-situ observations, and with a reference simulation based on the HTAPv2 emission inventory. In most parts of the world, the simulation using FAN captures the observed patterns with a similar skill as the reference simulation. While only limited data are available in areas where the FAN and reference simulations differ the most, the higher ammonia emissions simulated by FAN are consistent with long-term observations in West Africa. In conclusion, the FAN model was found to provide a viable alternative for static emission inventories with the main advantages being consistency with the simulated nitrogen cycle and the ability to simulate the response to environmental conditions on hourly to century timescales and provide climate-consistent emission estimates. The results also stress the need for global observations to better constrain ammonia emissions outside the regions covered by current air pollution monitoring networks.