Ammonia emissions pose significant challenges for both environmental protection and human health. A substantial portion of these emissions occurs after field fertilization. Accurate prediction of these emissions is essential for national inventories and for identifying effective mitigation strategies. Although several static machine learning models have been developed to estimate final cumulative emissions, the potential benefits of dynamic machine learning to improve these predictions remain unknown. To address this gap, we compared 13 static models (1 random forest, 12 neural networks) and 33 dynamic models (7 random forests and 26 recurrent neural networks). The best performing model was a recurrent neural network, achieving an average mean absolute error (MAE) of 4.56 kgN/ha (95 % CI = [4.17, 4.95]), corresponding to a decrease in MAE of 13.6 % and 17.7 % compared to the best static neural network and the static random forest, respectively.