Accurate estimates of ammonia (NH 3) emissions are needed for reliable predictions of fine particulate matter (PM 2.5) by air quality models (AQMs), but the current estimates contain large uncertainties in the temporal and spatial distributions of NH 3 emissions. In this study, the US EPA Community Multiscale Air Quality (CMAQ) modeling system is applied to study the contributions of the agriculture–livestock NH 3 (AL-NH 3) emissions to the concentration of PM 2.5 and the uncertainties in the total amount and the temporal variations of NH 3 emissions and their impact on the formation of PM 2.5 for August and December 2002. The sensitivity simulation results show that AL-NH 3 emissions contribute significantly to the concentration of PM 2.5 , NH 4 + , and NO 3 À ; their contributions to the concentrations of SO 4 2À are relatively small. The impact of NH 3 emissions on PM 2.5 formation shows strong spatial and seasonal variations associated with the meteorological conditions and the ambient chemical conditions. Increases in NH 3 emissions in August 2002 resulted in 410% increases in the concentrations of NH 4 + and NO 3 À ; reductions in NH 3 emissions in December 2002 resulted in 420% decreases in their concentrations. The large changes in species concentrations occur downwind of the high NH 3 emissions where the ambient environment is NH 3 -poor or neutral. The adjustments in NH 3 emissions improve appreciably the model predictions of NH 4 + and NO 3 À both in August and December, but resulted in negligible improvements in PM 2.5 in August and a small improvement in December, indicating that other factors (e.g., inaccuracies in meteorological predictions, emissions of other primary species, aerosol treatments) might be responsible for model biases in PM 2.5 .