Development of new computational models for predicting nitrogen excretion in dairy cattle

  • Xianjiang Chen

Student thesis: Doctoral Thesis


The nitrogen (N) excretion from dairy production systems can cause environmental pollution and affect human health, which has promoted the growing interest over the past decades to develop strategies for reducing N excretion in dairy cows. Therefore, it is critical to have capacity to accurately predict/mitigate N excretion from dairy cows, in order to enhance economic stability and reduce environmental impacts of dairy farming. The immediate objective of this thesis is to develop updated and more effective regression and computational models for evaluating N excretion from dairy cattle.

In this thesis, analysis of variance and restricted maximum likelihood components analysis methods were performed to evaluate and compare the N utilisation efficiency between modern Holstein-origin dairy cows and earlier populations, and also among the dairy cattle at different physiological stages. Besides, machine learning based algorithms and linear mixed model were used to explore dynamic causal influence among factors affecting N utilisation of dairy cattle and develop new regression and computational models for evaluating N excretion from dairy cattle.

The findings of the study indicate that modern dairy cows could utilise consumed N more efficiently and using previous models developed from old cow data to predict N excretion for modern dairy herds might over-estimate their N excretion in urine and total manure. The physiological stages of dairy cattle had great influence on its N utilisation efficiency. Dairy calves could utilise feed N more efficiently than growing and mature cattle, and lactating cows had a lower proportion of excretion than growing cattle and dry cows.

Finally, a number of new models for predicting N excretions from dairy cattle were developed. These were: 1) linear and multiple linear regression models for predicting N excretions from modern dairy cows and from different physiological stages of Holstein-Friesian dairy cattle; 2) artificial neural network-based models for prediction of manure N excretion of dairy cows using either N intake or live weight and milk yield as primary explanatory variables; and 3) Bayesian network model for capturing relationships among factors that influence N utilisation efficiency of dairy cows and performing causal influence among predictors.

These models may provide effective tools for optimizing the management of feed N resource for current dairy production and developing strategies to reduce N excretion in dairy production systems for the achievement of sustainable and environment-friendly dairy production under grassland-based farming conditions.
Date of AwardJun 2023
Original languageEnglish
SponsorsDepartment of Agriculture, Environment and Rural Affairs Northern Ireland
SupervisorHuiru (Jane) Zheng (Supervisor), Haiying Wang (Supervisor) & Tianhai Yan (Supervisor)


  • Manure
  • Artificial neural networks
  • Machine learning
  • Livestock

Cite this