Agriculture Reference
In-Depth Information
As model developers, scientists play a
key role in creating the link between science,
decision makers and users. While the deci-
sion makers will easily formulate their needs
on a long-term basis, the scientists need to
understand how this will be reflected in
terms of model equations and, later, in terms
of model application. It is the role of scientists
to understand the users' needs and expect-
ations on a short-, medium- and long-term
basis, and to develop an adaptive model that
can be continuously improved.
Gathering the stakeholders together be-
fore starting the model framework descrip-
tion may represent the first important step
of model development projects to ensure
all needs and expectations are known and
understood.
level (Rivera-Torres
et al
., 2011c). However,
protein and lipid turnover rates are difficult
and expensive to determine, and the data
are scarce and relatively old when consider-
ing the rapid advances in genetic selection
(Kang
et al
., 1985). Alternatively, Rivera-
Torres
et al
. (2011c) proposed indirect esti-
mates of the protein and lipid turnover
parameters by manual calibration using
protein and lipid retention rates of turkeys
with different breeds and genders.
The challenge to scientists is to determine
the most appropriate method for model
simulation and application. While the model
of Rivera-Torres
et al
. (2011c) used protein
and lipid turnover rates to gain flexibility in
the description of growth rate, it also allowed
the simulation of macronutrient composition
effects on energy and nutrient utilization
(i.e. amino acids, glucose, fatty acids, acetyl
coenzyme-A) for protein and lipid gain.
This model is particularly interesting for re-
search application, while more complicated
for commercial application. Indeed, in-
creasing the level of complexity of models
is usually related to an increasing number
of input data required to calibrate the model
and ensure an appropriate level of precision
of the outputs
(Fig. 8.1
). As an example, be-
cause the model of Rivera-Torres
et al
. (2011c)
simulated macronutrient utilization and
ATP production, data on energy utilization
were generated to calibrate the model in terms
of nutrient oxidation rates (i.e. fatty acids,
amino acids, glucose, acetyl coenzyme-A) to
ensure an adequate level of precision of the
outputs (e.g. protein and lipid retention,
maintenance heat production, carbon diox-
ide production).
Framework
The model framework defines the inputs,
outputs and flows and compartments of the
model. It also defines how genetic potential
and nutrient requirements are simulated.
Independently of the approach used to simu-
late genetic potential, most of the current
poultry models are based on the assumption
that genetic potential enables the determin-
ation of feed intake as a result of dietary
nutrient and environmental constraints
(Emmans and Fisher, 1986). To evaluate gen-
etic potential accurately there is a need to
evaluate the nutrient partitioning of the ani-
mal regularly through body composition
analyses (Rivera-Torres
et al
., 2011b; Rivera-
Torres and Ferket, 2012; Murawska, 2013)
or calorimetry measurement (Rivera-Torres
et al
., 2010, 2011a).
Recent body composition analyses of
turkeys has facilitated the estimation of the
parameters of the Gompertz equations and
of allometric relations (Rivera-Torres
et al
.,
2011b; Rivera-Torres and Ferket, 2012) for
further use in turkey models. The main ad-
vantage of these methods is that the param-
eters of the allometric and Gompertz equations
can be estimated by regression analysis. Also,
the simulation of protein and lipid turnover
rates (Danfaer, 1991) is an interesting method
to simulate turkey growth potential at a metabolic
Model Development
Development strategy
While stakeholders may expect models to
deliver a high level of accuracy, they may also
expect the model to consider a large number
of inputs in order to account for a large
spectrum of effects (e.g. individual amino
acid level, ambient temperature, stocking
