Methane and nitrous oxide are key pollutants emitted from agriculture. Primarily the livestock production has a significant share in CH4 emissions. The N2O emissions largely correspond to direct emissions associated with the cultivation of soils. The priority task of agriculture is to develop adaptive solutions enabling the reduction of pollutions in the next years. These capabilities apply to both technological solutions on the farms, as well as improved methods of management and policy tools. Therefore complementary information to the knowledge in the field of the possibilities for reducing CH4 and N2O are extremely valuable. The study of predictions of N2O and CH4 emissions on the basis of different arable crops areas with the use of Flexible Bayesian Models of neural networks was carried out. The decision trees have been designed in order to provide the knowledge and methods that allow the rapid identification of the most important arable crops that affect the quantity of these emissions. On the basis of the conducted analysis, wheat, maize and potatoes in the case of N2O emission and wheat and maize in the case of CH4 emission are the most important differentiating variables.