Agriculture Reference
In-Depth Information
the need for research in these areas to be in-
tegrated as far as possible.
Discussion
If modelling techniques are to make a last-
ing contribution to poultry science and to
the poultry industry, then communication
amongst the participants in this field of en-
deavour is essential. This will of course take
many forms but publication in refereed
journals and the sharing of models may cre-
ate special requirements for this area of re-
search. In the age of the internet it should be
possible to develop these.
Most journals now accept modelling
papers and editors are responsible for set-
ting the standards required. However, if
the aim is to enable modellers to benefit
from each other's work, then the quality of
the paper's content becomes very import-
ant. Refereeing of modelling papers re-
quires a high level of technical knowledge
and the commitment of a great deal of
time. If the general criterion for a paper
describing experimental work is that the
reader should be able to repeat the experi-
ment, an analogous criterion for a paper
describing a model is that it should be pos-
sible to recreate the model from the infor-
mation given. Because of technical issues
of computer programming this may not be
literally possible but the information con-
tent should be at this level. To date this
has rarely been achieved.
Ideally a paper describing a model
should explicitly present the theories and
assumptions on which the model rests, the
detailed flow of information through the
model, individual mathematical forms and
equations, the logic and process of param-
eterization and the parameter values. Add-
itional information about sensitivity and
model validation may also be available. The
emphasis should be on the process and not
on the end result.
In many cases modelling will require
that previously published experimental data
are available for re-calculation or re-use.
Rosen (2006) has emphasized the need for
additional reporting of conditions and re-
sults in trials which might be useful in a
future holo- or meta-analysis. In developing
his effective energy scale, which has been
widely used in modelling, Emmans (1994)
Real-time control models
Recursive control algorithms are well devel-
oped for control of physical processes and
environments, and interest in applying
these ideas to broiler production seems to
have been stimulated by two developments:
the use of two feeds, usually involving a
whole cereal; and the need to control feed-
ing in some circumstances. In the first case
the problem is to control the mixture of the
two feeds presented to the birds taking ac-
count of their performance, and in the se-
cond case to control the feed supply to
achieve a given growth trajectory. Real-time
measurements of feed intake, water intake,
body weight and environmental variables
can be used to provide the control.
Filmer (2001) and Stacey
et al
. (2004)
describe the application of a commercial
automatic growth and nutrition control sys-
tem for broiler production. This was con-
ceived as part of an integrated management
system (IMS) for broiler production (Frost
et al
., 2003). Published research in this area
has come from the Catholic University of
Leuven, Belgium. Aerts
et al
. (2003a,b)
demonstrated the control of growth trajec-
tory through both quantitative and qualita-
tive dietary control. Similar approaches
have been used to model heat production
under various temperature and light inten-
sity regimes (Aerts
et al
., 2000).
This seems to be an area where the
modelling is ahead of the implementation.
The IMS described by Frost
et al
. (2003)
has not developed further because of the
difficulty of maintaining a high level of
day-by-day management input and also
maintaining sensitive equipment in a
broiler farm environment. A scale-up of the
control system devised by Aerts
et al
.
(2003b) from the laboratory to pens of 1500
birds led to a deterioration in accuracy of
control (Cangar
et al
., 2007). If such engin-
eering and management problems can be
overcome then a rapid implementation of
the modelling seems to be possible.
