Agriculture Reference
In-Depth Information
too often ignored is the distinction between
potential and observed growth. Observed
growth is the outcome of interactions be-
tween the animal's genotype and both the
internal and external environment in which
it is kept. Potential growth, expressed in a
non-limiting environment, is a concept that
can reasonably be assumed to reflect the
genotype of the animal and nothing else.
There can be no general rule about the na-
ture of observed growth and hence no rea-
son why observed growth data should con-
form to any particular mathematical form.
Those many studies in which different growth
equations are compared to observed data
sets seem to the present writer to not en-
hance our understanding of growth. Parks
(1982) raised this issue in a different way
by considering the mathematical properties
that a growth function may have. He lists sev-
eral cases, one of which that growth is con-
tinuous and possesses continuous rates of
change of all orders leading to justification
of a growth function widely accepted by
most people writing in this field. Parks ar-
gued that this case should be accepted be-
cause (i) there appears to be no evidence to
the contrary, and (ii) it is intuitively the most
promising of the cases considered. However,
Parks' assumption that these properties of
growth curves could be used to study the ef-
fects of environmental and non-environmental
factors on growth seems to the present writer
to not be justified. Alternative approaches to
modelling irregular growth patterns are sug-
gested by the work of Roush
et al
. (1994),
who explored an analytical approach to peri-
odicity or chaos in broiler growth data, and
Talpaz
et al
. (1991) who modelled growth fol-
lowing a period of feed restriction.
The weakness of arguing for the study of
potential growth is that non-limiting environ-
mental conditions are difficult to define and
to confirm in any particular experimental cir-
cumstances. It is easy to say that 'non-limiting'
environments must be used, but difficult to
ensure that they apply to all birds at all times
in an experiment. These problems cannot be
solved completely, although several experi-
ments of this kind have been reported and
they seem to be based on reasonable assump-
tions for practical modelling. Once a growth
function has been selected, then deviations
from the smooth function can be used to iden-
tify periods in which growth might not have
been unrestricted (Ferguson and Gous, 1993).
Studies of growth parameters under as-
sumed non-limiting conditions have been re-
ported for broilers by Stilborn
et al
. (1994),
Hancock
et al
. (1995), Gous
et al
. (1996, 1999),
Hruby
et al
. (1996), Wang and Zuidhof (2004)
and Sakomura
et al
. (2005, 2006). Similar ex-
periments have not been reported for turkeys,
although Emmans (1989) considers the prob-
lem. Additional information on turkeys is
available from Hurwitz
et al
. (1991) and Porter
et al
. (2010).
Empirical models of poultry
production systems
A wide range of approaches to the empirical
modelling of biological systems is described
by Roush (2006) and their application to the
poultry industry is discussed by Roush
(2001). Amongst the tools listed by Roush (2006)
are: (i) stochastic and fuzzy logic models;
(ii) non-linear dynamics (chaos); (iii) regres-
sion analysis and response surface method-
ology; (iv) artificial neural networks; (v) genetic
algorithms; (vi) Kalman filter; and (vii) lin-
ear, chance constrained, goal, and quadratic
programming. Elegant examples of each of
these procedures have been demonstrated
by Roush and his colleagues but their appli-
cation to the solution of poultry science
problems or for commercial decision making
remains elusive. Ad hoc application of sin-
gle analytical techniques to single problems
(e.g. Faridi
et al
., 2013) seems to the present
author to emphasize the limitations rather
than the usefulness of these methods.
Empirical models are typically based
on commercial data or on experimental data
from pen trials, but can be distinguished
from the methods discussed above by the
fact that they consider part or all of a poultry
production system and are aimed at im-
proved commercial decision making. The
distinction between these models and, for
example, meta-analysis is not absolute and,
as usual, the distinction between empirical
