Title: | A Segmented Method to Predict Lifetime Productivity and Optimise Culling Across Age Classes in Commercial Sheep Flocks |
Contributor(s): | Richards, Jessica Susan (author) ; Kinghorn, Brian (supervisor) ; Van Der Werf, Julius (supervisor) ; Atkins, Kevin David (supervisor) |
Conferred Date: | 2020-03-12 |
Copyright Date: | 2019-08 |
Handle Link: | https://hdl.handle.net/1959.11/57146 |
Abstract: | | Opportunities for increasing productivity and profitability within the sheep
industry are constantly increasing with advances in genetic, measurement and
management technologies. The challenge is for producers to find the key
opportunities for their own production system and identify which are the most
beneficial in their environment and to meet their target markets. Available
decision support tools for replacement and culling decisions in Merino sheep
flocks are limited to single selection events and managing whole age classes until
culling at a fixed age.
A method was developed to predict lifetime production changes in cohorts of
animals in response to various strategies for selection and culling across multiple
age classes. Simple deterministic predictions assume selection from normally
distributed data, which would not work in scenarios where selection has already
occurred and the distribution of animals within selected cohorts is not normal.
Therefore, a new deterministic approach was developed by predicting from
segments of phenotypes within a distribution and then combining the outcomes
across overlapping segments for whole flock or cohort outcomes. This enables
selection from skewed distributions of animals, as a result of previous culling or
selection events. Deterministic prediction was preferred to stochastic simulation
because it provided a single, repeatable outcome for each specific strategy. This
provides an ideal platform for applying an optimisation algorithm for efficiently
finding the best strategy among numerous possibilities to meet an objective. For
this purpose we used a Differential Evolution algorithm. The importance of
applying such an algorithm increases as the complexity of possible combinations contributing to the objective increases, such as multiple factors influencing the
value of an economic objective.
This so-called “segmented method” was initially developed for prediction of
continuous traits, but was unsuitable for predicting discrete categorical data, such
as reproduction. Therefore, a method for examining discrete traits on the same
platform was also developed, where the traits were modelled using an underlying
continuous liability scale with thresholds between lambing subclasses to
determine response in reproduction in each age class. By integrating these
adjustments for discrete traits into the segmented method, responses in discrete
traits could also be predicted from skewed distributions as a result of selection
across multiple age classes.
This thesis describes the development of this new prediction method and the
associated optimising algorithm using examples for increasing productivity
within Merino flocks, specifically optimising wool profitability or increasing
reproduction performance. This method provides opportunity to better understand
consequences of selection and management decisions on cohorts of the flock at
different age classes and the population as a whole. The method developed also
provides a base upon which to add more components (such as additional traits,
targeted management applications or treatment options) that will act as a stimulus
for researchers and extension staff to ask further questions on the potential value
of future information sources, as well as to identify better ways of using currently
available information. Although this study has focussed on the benefits of
application within a commercial Merino flock the method is applicable to all
livestock industries by replacing relevant parameters and altering objectives.
Publication Type: | Thesis Doctoral |
Fields of Research (FoR) 2020: | 300305 Animal reproduction and breeding 300302 Animal management |
Socio-Economic Objective (SEO) 2020: | 100412 Sheep for meat 100413 Sheep for wool |
HERDC Category Description: | T2 Thesis - Doctorate by Research |
Description: | | Please contact rune@une.edu.au if you require access to this thesis for the purpose of research or study.
Appears in Collections: | School of Environmental and Rural Science Thesis Doctoral
|