Agriculture Reference
In-Depth Information
and records of the main parameters are
readily available both from meteorological
data and from poultry houses. However, a
description of the 'effective' environment,
as experienced by the bird, is much more
complex and is a problem that has not been
completely solved despite the best efforts of
the environmental physiologists. Inter-
actions with feathering in particular, and to
a lesser extent with stocking density, are im-
portant. Modelling the physiological and
behavioural adaptations to high environ-
mental temperatures and other stresses is
likely to be very difficult. Our experience
(EFG Software) is that the balance between
heat production and heat loss to the envir-
onment in broilers is the major limiting fac-
tor governing response to diet under prac-
tical conditions, so this is a very critical
area for modelling. Other aspects of the
physical environment that need to be re-
solved include diurnal variation, light:dark
patterns and stocking density.
Most of the parameters describing the
availability of energy and amino acids from
the feed are likely to be model inputs. For
practical use these need to draw on the ex-
tensive data and experimentation that are
used in the industry. For energy, metaboliz-
able energy is a convenient input, although
energy transactions in the model will have
to be expressed in terms of a net energy scale
if heat production is to be calculated. For the
same reason dietary fat will have to be con-
sidered as a resource. For amino acids the
scale that is used needs to be compatible
with the assumptions that are made about
their rates of utilization. This topic has been
critically discussed by Moughan (2003).
For modelling growing birds consider-
ation of nutrients other than energy and
amino acids seems to be unnecessary. In the
laying hen where calcium and phosphorus
absorption is one of the control mechanisms
governing shell quality, then these add-
itional nutrients and absorption processes
may be important (Kebreab
et al
., 2009).
Modelling mineral utilization may also be
useful in designing anti-pollution strategies.
Although gut function and digestion play a
major role in practical broiler nutrition, to
date, ways of handling this in a mechanistic
model have not been suggested. If such work
is undertaken, some early work on digestion
in the pig may be useful (Bastianelli
et al
.,
1996; Rivest
et al
., 2000). In addition, a huge
amount of effort has gone into the simula-
tion of digestion in ruminants.
The prediction of feed intake is a cen-
tral necessity in any mechanistic model of
animal production systems. Models in
which feed intake is a direct or indirect in-
put to the model are unlikely to be flexible
enough for practical use. A possible excep-
tion to this is the idea of an adjusted refer-
ence feed intake as used by Quentin (2004).
Feed intake will be influenced by factors in
the bird description, in the environment
and in the feed. This remains one of the
most challenging areas requiring further de-
velopment in mechanistic modelling. Once
the basic problem of prediction is solved,
characteristics of the feed such as 'palatability'
and bulk (limiting gut fill) will need to be
considered.
Existing mechanistic models of broiler
growth simulate the performance of a single
animal, which is assumed to be at the mean
of a population. If the parameters used to
describe genetic potential growth are from
known correlated distributions, then popu-
lations of birds can be generated and simu-
lations of population response carried out
(Gous and Berhe, 2006). Whether other
model elements, for example, the efficien-
cies with which nutrients are utilized, should
be seen as stochastic, remains an open ques-
tion for lack of any hard information. If this
is considered, then the covariance amongst
different model components needs to be
included. Simply introducing stochasticity
as a statistical device to all elements in a
model probably has little value. If one of the
model outputs is to reflect some sort of grad-
ing scheme such as percentage rejects, then
the question of non-normal distributions
may need to be considered.
The other issues that arise in creating a
model of broiler production include mortal-
ity (usually considered as an input), calibra-
tion or adjustment for known stressors or for
disease and economic modules. The assem-
bling of all these elements into a single pro-
gram is a significant enterprise emphasizing
