Agriculture Reference
In-Depth Information
The optimization process is started by
passing initial specifications (Nutrients) to
a feed formulator to determine the ingre-
dient composition at the least cost (For-
mulation). These diets are made available
to the animal in the biology component to
produce a specific animal response, in-
cluding feed intake, growth, feed costs
and carcass characteristics. From these
data the financial outcomes (Economics)
can be generated, which are then for-
warded to the optimizer to complete the
cycle. This process is repeated before
identifying the 'best' solution to meet the
optimization objective. Currently in Wat-
son
®
2.0 the processes to be optimized
include energy content, nutrient density,
amino acid responses, carcass weights
and feeding phases, while the optimiza-
tion objectives include maximizing growth
rates, margin over feed costs, net profit
per pig and minimizing feed:gain, cost/
kilogramme gain, and nutrient excretion.
There will be different optimum solutions
depending on the objective. For example,
the incremental cost of increasing the
dietary amino acid level may not be off-
set by the increase in revenue generated
from improved feed efficiency and/or
higher carcass lean, resulting in differ-
ences in the optimum amino acid re-
quirements between minimizing feed:gain
and maximizing margin over feed cost
objectives. The results of the optimiza-
tion process always provide three pos-
sible solutions including: (i) the single
best solution (e.g. highest MOFC or low-
est feed:gain) called the Optimum solu-
tion; (ii) a solution that can meet the
objective (e.g. improve MOFC by CAN$0.50/
pig) but does not deviate too much from
the current nutrition programme, called
the Minimum solution; and (iii) a pro-
gramme that represents the average of all
solutions that meet the objective (e.g. im-
prove MOFC by $0.50/pig), called the
Average solution.
As individual pigs will have different
optimum performance and economic re-
sponses to amino acid intakes, it is neces-
sary to incorporate this between-animal
variation into the optimization process.
Gous and Berhe (2006) summarized the im-
portance of this issue as follows:
Models of individuals may be adequate for
an understanding of the theory of growth
and feed intake, as well as for 'what-if'
scenario planning. However, for purposes
of optimization, it is imperative to account
for the variation inherent in the system if
a realistic assessment of the population
response is to be simulated.
Within a batch of finishing pigs there is suf-
ficient between-animal variation in protein
and fat deposition, feed intake and subse-
quent efficiency of nutrient utilization to
ensure differences in the optimum nutrient
response between the single average indi-
vidual and the batch mean (Pomar
et al
.,
2003; Brossard
et al
., 2009; de Lange
et al
.,
2012). The challenge for commercial models
used across functional areas is how to intro-
duce stochasticity into the optimization pro-
cess without generating excessively large
amounts of redundant data that take up un-
necessary time and computing resources.
Animal Variation
Most integrated pig growth models are de-
terministic by nature and assume that the
response of the 'average' individual is a good
representation of the population response.
In most practical cases this assumption may
hold true, but there are cases when the
mean response can differ significantly from
the average individual response due to the
variation in growth potential between indi-
vidual animals (Pomar
et al
., 2003; Brossard
et al
., 2009). The extent of these differences
will depend on: (i) the extent of the dif-
ferences between individuals within the
population; (ii) the correlation between the
genetic parameters defining the genotype;
and (iii) the individual animal's ability to
cope with social stressors (Wellock
et al
.,
2003c, 2004). The more individuals vary
within a population (e.g. the larger the vari-
ation in initial starting weight), the more in-
appropriate it is to use the average individual
response as a means of predicting the popu-
lation response. For example, predicting
