Comparison between the Statistical Method and Artificial Neural Networks in Estimating Ammonia Emissions from Naturally Ventilated Dairy Cattle Buildings

Abstract

The dairy industry is one of the major contributors to global ammonia emissions, posing increased concerns to producers, professionals and local communities. This study aimed to develop an approach to estimate ammonia emission level from dairy cattle buildings. An investigation on ammonia emission from dairy cattle buildings with different floor types and manure handling systems in Denmark was carried out. Ammonia emission rate was determined by measured ammonia concentrations and calculated values of air exchange rates. The overall measured ammonia emission rates were in the range of 8.9 to 46.8 g/(HPUd) with a standard deviation from 3.9 to 25.6 g/(HPUd) at the air temperatures from 2.3 to 22.4 oC. A statistical model based on the regression analysis as well as the artificial neural networks approach were developed to predict ammonia emission rates from naturally ventilated dairy cattle buildings. The results indicate that the estimated ammonia emission levels by both approaches have good consistencies with measured rates, and the artificial neural network model is shown to provide a better prediction of the emission than the statistical method.