||Ammonia (NH3) emission from animal manure contributes to air pollution and ecosystem degradation, and the loss of reactive nitrogen (N) from agricultural systems. Estimates of NH3 emission are necessary for national inventories and nutrient management, and NH3 emission from field-applied manure has been measured in many studies over the past few decades. In this work, we facilitate the use of these data by collecting and organizing them in the ALFAM2 database. In this paper we describe the development of the database and summarise its contents, quantify effects of application methods and other variables on emission using a data subset, and discuss challenges for data analysis and model development. The database contains measurements of emission, manure and soil properties, weather, application technique, and other variables for 1895 plots from 22 research institutes in 12 countries. Data on five manure types (cattle, pig, mink, poultry, mixed, as well as sludge and "other") applied to three types of crops (grass, small grains, maize, as well as stubble and bare soil) are included. Application methods represented in the database include broadcast, trailing hose, trailing shoe (narrow band application), and open slot injection. Cattle manure application to grassland was the most common combination, and analysis of this subset (with dry matter (DM) limited to <15%) was carried out using mixed- and fixed-effects models in order to quantify effects of management and environment on ammonia emission, and to highlight challenges for use of the database. Measured emission in this subset ranged from <1% to 130% of applied ammonia after 48 h. Results showed clear, albeit variable, reductions in NH3 emission due to trailing hose, trailing shoe, and open slot injection of slurry compared to broadcast application. There was evidence of positive effects of air temperature and wind speed on NH3 emission, and limited evidence of effects of slurry DM. However, random-effects coefficients for differences among research institutes were among the largest model coefficients, and showed a deviation from the mean response by more than 100% in some cases. The source of these institute differences could not be determined with certainty, but there is some evidence that they are related to differences in soils, or differences in application or measurement methods. The ALFAM2 database should be useful for development and evaluation of both emission factors and emission models, but users need to recognize the limitations caused by confounding variables, imbalance in the dataset, and dependence among observations from the same institute. Variation among measurements and in reported variables highlights the importance of international agreement on how NH3 emission should be measured, along with necessary types of supporting data and standard protocols for their measurement. Both are needed in order to produce more accurate and useful ammonia emission measurements. Expansion of the ALFAM2 database will continue, and readers are invited to contact the corresponding author for information on data submission. The latest version of the database is available at http://www.alfam.dk.