Agriculture Reference
In-Depth Information
500
D21
D63
D105
D35
D77
D49
D91
50
50
500
Simulated average daily feed intake (g/day), log transformation
Fig. 8.2.
Observed and simulated average daily feed intake (g/bird/day) of male turkeys as a function of age
(day
21
to
105,
D21 to D105). Values below the dotted line refer to an overestimation of the model outputs,
while values on the dotted line are in agreement with the observations.
deviate from realistic values when challen-
ging the outputs with extreme inputs (e.g. de-
ficient levels of amino acids, unbalanced
amino acid to energy ratio). By understanding
the robustness of a model the user is capable
of defining the limitations for its application.
For example, low density diets are expected
to result in greater feed intake, but a robust
model should also account for feed intake
limitation due to feed bulk and intake cap-
acity. Although robust concepts have been
developed in broiler chicken models (EFG
Software, 1995) to account for perturbations
such as heat dissipation capacity and feed
intake capacity, no estimate of such equation
parameters has yet been proposed in turkeys.
Robustness can be evaluated by per-
forming sensitivity analyses, which consist
of identifying the parameters that have a
major impact on model outputs. Sensitivity
analysis enables a better understanding of
the behaviour of the model depending on the
parameter values. The value of the sensitive
parameters should be as accurate as possible,
while adjusted parameters should not be
sensitive to changes. Also, scientists and
users should identify the sensitive param-
eters of the model to ensure an appropriate
interpretation of the outputs. Some of the
most sensitive parameters in turkey models
are the parameters associated with energy
and nutrient requirements. Rivera-Torres
et al
. (2011a) showed that energy require-
ments for maintenance (i.e. per kilogramme
of metabolic weight) were 16% lower in fe-
males than in males. Because energy require-
ment regulates feed intake when no nutrient
is limiting in the diet it is important that tur-
key models account for gender-based differ-
ences in maintenance requirements as they
strongly impact feed intake and, therefore,
weight gain. Also, an accurate estimate of the
relationship between water and protein mass
is key to estimating body weight and feed in-
take. Indeed, as water mass represents close
to two-thirds of the whole body mass, an
overestimate of the water-to-protein ratio
may result in an overestimate of body weight
as a result of the excessive estimation of
water retention (Rivera-Torres
et al
., 2011c).
Application
Conditions for success
Before training users in model applica-
tion, documentation should be prepared
to: (i) present model theory for a basic
