Estimating ammonia (NH₃) volatilization from cropland can identify potential environmental risks and is also important for mitigating NH₃ loss, since cropland is the major anthropogenic source of NH₃. Reducing uncertainties around NH₃ volatilization estimates is a key need. In this study, both a single national model and agro-region-specific models were developed to embody the variation of NH₃ volatilization associated with environmental attributes and management practices for cropland in various agro-regions of China. A database with 951 field measurements across China was established to develop and evaluate the model, in which 75% of the measurements were used for model fitting and 25% were used for evaluation. The agro-region specific model had a better performance than the single national scale model, and the agro-region-specific model could successfully simulate NH₃ emissions with R² values of 0.41–0.93 as a function of crop type, N fertilizer type and rate, timing of fertilization, soil pH, clay content, bulk density, precipitation, air temperature, and soil total nitrogen content. Inputting high-resolution spatial data into the model, the NH₃ emissions from China's cropland were estimated to be 4.31 Tg NH₃N in the year 2015 with a 95% confidence interval of 3.64–5.64 Tg NH₃N, of which paddy rice cultivation accounted for 44%, and maize and wheat cultivation contributed 20% and 16%, respectively. Notable spatial variability was also found in NH₃ emissions from China's cropland due to regional differences in climate, soil and management patterns. This study showed that the accuracy of estimation of farmland NH₃ emissions can be improved by taking meteorological, soil and management variables into account and inputting high-spatial-resolution data into the modeling.