Examination of the ammonia emission coefficient of liquid manure associated with its field application based on machine learning
To quantify the effects of environmental factors,fertilization techniques,physical and chemical characteristics of manure on ammonia(NH3)emission coefficient after liquid manure return to field.This study used the method of meta-analysis to explore the effects of 16 factors,including soil properties,liquid manure properties,and farmland management measures,on soil NH3 emission coefficient.We analyzed 52 publications and a total of 454 observation studies.The machine learning toolbox in the MatLab package was used to train and construct a prediction model for the NH3 emission coefficient.According to the meta-analysis,the type,dry matter content,pH of liquid manure,and application method were the most important factors affecting the NH3 emission coefficient.Among 26 trained models,the most effective model was the Gaussian process model(exponential GPR)with a determination coefficient(R2)of 0.64 and root-mean-square error(RSME)of 0.067.The correlation coefficient between the predicted and actual NH3 emission coefficients reached 0.91.This model could not only predict the impact of manure application technology on the NH3 emission coefficient successfully but also systematically compared the pre-treatment methods of liquid manure and the effects of its dry matter content,pH,and other physicochemical properties on the NH3 emission coefficient after field application.However,the variation in the NH3emission coefficient for different soil textures was relatively low.In conclusion,the prediction model for the NH3 emission coefficient can not only indicate the effects of fertilization technology,environmental factors,and other factors on NH3 emissions,but also systematically elucidate the physicochemical characteristics and management methods for liquid manure with respect to the NH3 emission coefficient after field application.
万方数据
维普
