A novel Bayesian methodology to estimate unobserved traits of cattle for AI assisted prec

"The traditional method of feeding livestock is to feed them according to the requirements of the 'average' individual within a group and feed them as a group. This means that some individuals of the group receive less nutrients than they require, with the obvious consequences on their health and welfare, whereas some others are fed above their requirements, meaning that the excess nutrients are excreted in the environment. Recent advances in automation technology now allows individuals to be identified and to be fed differently whilst they are still in their group.
In order to feed individuals according to their needs, one needs to have information about some traits which are difficult to measure. For a growing animal these include their maximum capacity for growth and at what stage they are in their growth trajectory (degree of maturity). We propose that these individual traits can be inferred from traits that are measured routinely such as bodyweight and food intake, through a novel statistical methodology. The hypothesis is that the approach will lead to better estimates of individual animal nutrient requirements which will enhance the resilience and sustainability of their system of production.

1. The student will interrogate a large database of several thousand beef cattle owned our Industry sponsor , which record serial measurements of several traits of individuals from birth to slaughter. As also the frequency of these measurements varies between individuals, the aim will be to select a subset of individuals suitable for the purposes of subsequent steps.
2. The student will estimate the empirical average and distribution of individual (unobserved) traits characterising growth potential and degree of maturity in the individuals of the sub-population, through the application of a Bayesian approach, based on the principles of inverse modelling using the traits measured routinely.
3. The database will contain additional traits for some individuals within the population of beef cattle, such as body composition at slaughter. (i) Formal statistical diagnostics will be used to assess the biological plausibility of the estimates of the values of the unobserved traits for each individual within the population, including comparisons when the numbers of measured traits increase. (ii) Formal statistical comparisons of the distributions of the biological values of the unobserved traits will be made against the actual measurements of these traits on a subset of individuals (validation), ie the individuals where such measurements were recorded.
4. The nutrient requirements of the individuals within the population will be estimated during their growth trajectory. Energy and protein requirements will be estimated in the first instance and their mean values for the population will be compared against the values of the average individual as estimated by the recently updated feed tables for UK beef cattle. These tables are currently used as the guidelines by the beef cattle Industry.
5. Finally, the consequences of feeding strategies where targeting the nutrient requirements for the average individual against feeding the individuals within a population will be simulated in terms of nutrient excretion and environmental impact. This will lead to strategies for precision management of beef cattle, which enhance animal health, welfare and performance, whilst minimising their environmental impact."