Building a Bayesian network to identify key intervention points for improving nitrogen efficiency in New Zealand dairy farm systems

Gina Lucci1, Cecile DeKlein2, Vicki Burggraaf1, Diana Selbie1 and David Pacheco3

1 AgResearch, Private Bag 3123, Hamilton 3240, New Zealand
2 AgResearch, Puddle Alley, Private Bag 50034, Mosgiel 9053, New Zealand
3 AgResearch, Private Bag 11008, Palmerston North, 4442, New Zealand


Nitrogen (N) losses from New Zealand dairy farms are, in part, due to inefficiencies in N use within the system. Nitrogen cycling in pastoral dairy farming systems is complex, and understanding the interactions and interdependencies of N sources, N use and processes that control N losses will enable a more targeted approach to improving the overall N efficiency of the system. Bayesian Network (BN) modelling is an alternative to conventional modelling as it can evaluate complex multifactor problems using both forward and backward reasoning (cause-to-effect, and effect-to-cause), as well as assign probabilities to different outcomes. We developed a BN to identify the relative contribution of different components within a NZ dairy system to N leaching losses. An initial analysis revealed that the BN model can be a valuable tool for understanding how elements of the dairy N system fit together and their relative importance to overall N loss. Preliminary results also show that N leaching was most affected by feed N content and DM intake as opposed to the breed and weight of the cow. After further validation of the model it will be used to assess how current systems can be changed to meet N leaching targets, and to identify future strategies for improving N efficiency that target the key intervention points.