Content uploaded by Joanna Makulska
Author content
All content in this area was uploaded by Joanna Makulska on Jun 07, 2014
Content may be subject to copyright.
Ann. Anim. Sci., Vol. 10, No. 4 (2010) 333–348
APPLICATION OF MATHEMATICAL MODELLING IN BEEF HERD
MANAGEMENT – A REVIEW* *
Anna Stygar, Joanna Makulska
Department of Cattle Breeding, University of Agriculture in Kraków, Al. Mickiewicza 24/28,
30-059 Kraków, Poland
Corresponding author: rzmakuls@cyf-kr.edu.pl
Abstract
In this paper the possibilities and the advantages of the application of mathematical modelling
technique to the management of beef cattle are discussed. The management of beef herd is inse-
parably associated with making decisions concerning such activities as replacement/culling, in-
semination, feeding and marketing of animals. The present survey includes models that are part of
a whole-farm management strategy; the models of farm enterprise, consisting of both cow-calf and
fattening segments; the models of either cow-calf or fattening segment; and the models of single
animals. In spite of a considerable variety of beef models there are many similarities in their math-
ematical formulation. Methodologically, beef models can be classified into optimization and simu-
lation models. The methods commonly used in constructing and solving the optimization models
are linear programming and dynamic programming expanded with Markov decision processes.
Dynamic programming and Markov decision processes are also very often applied to solve the
stochastic simulation models. Most models adopt the dynamic approach, which is relevant consid-
ering the changes in herd composition and single animal performance over time. The necessity of
taking into account random variation, typically found in animal breeding and production, causes
that a large number of models are stochastic. Mathematical models of beef cattle management are
used mainly as research tools and teaching aids, and still not so many of them are applied directly
in supporting decisions in commercial beef herds.
Key words: mathematical modelling, optimization, simulation, herd management, beef cattle
Beef farming systems across the world show considerable differences in the eco-
nomic efficiency of beef production. Profitability of beef cattle enterprises depends
on many factors including biological performance of animals, management strate-
gies, natural conditions and marketing possibilities. However, it is influenced to the
highest degree by costs and returns and a key element seems to be beef price (Bruce
et al., 1999 a, b; Wolfová et al., 2004). In Poland, the relatively low live weight
prices and high costs of production are, according to many opinions, a significant
limitation to the development of beef cattle breeding. After Poland’s accession to the
* This work was conducted as research project no. DS-3245/KHB/10.
A. Stygar and J. Makulska
334
European Union in 2004, beef prices increased and their periodical fluctuations have
Because farmers have rather little influence on beef prices, attempts to improve
the economic results of beef production through better management of the herd have
been undertaken. Management comprises decision-making for the purpose of accom-
plishing desired goals and objectives. In order to make optimal decisions concerning
a certain unit (animal, group, herd, farm), farm managers need the knowledge about
the present state of the unit, the relation between the factors used and the resulting
effects given the present state, personal preferences and all restraints of legal, eco-
nomic, physical or personal kind. The decision process is further complicated by
the fact that most of the knowledge is associated with uncertainty (Kristensen and
Jørgensen, 1999).
The choice of the optimal decisions can be supported by the use of mathematical
modelling. Mathematical model is a simplified representation of a system aimed at
detecting the quantitative relationships between variables and predicting the effects
of their changes, assuming a compromise between accuracy and tractability. The
main advantages of model based decision support include the ability to take indi-
vidual conditions into account, a concise framework for combination of informa-
tion from different sources, direct representation of uncertainty and efficient search
algorithms for determination of optimal decisions. Models may further contribute
with extensive sensitivity analyses concerning optimal decisions, deviating condi-
tions and parameter values (Jalvingh, 1992; Kristensen and Jørgensen, 1999; Pla,
2007). Advanced computational methods and computers used in the modelling
process allow considering more aspects of a decision compared to the advice given
by experts (e.g. veterinarians or agricultural consultants) or application of general
norms, standards and recommendations. The decisions can be taken at different plan-
ning horizons: long-term (strategic), medium-term (tactical) and short-term (opera-
tional) (Jalvingh, 1992; Kristensen and Jørgensen, 1999).
Numerous investigations indicate the usefulness of mathematical modelling to
support beef cattle management decisions (Table 1). The models described in litera-
ture can be classified as:
– models that are part of a whole-farm management strategy
– the models of farm enterprise, consisting of both cow-calf and fattening (e.g.
feedlot) segments
– the models of either cow-calf or fattening segments
– the models of single animals
Table 1. Reviewed beef cattle models
Authors Year Method used Characteristics of the model
1 2 3 4
Werth et al. 1991 Simulation Model for evaluation of the influence of reproductive
performance and management decisions on net income
in beef production.
Keele et al. 1992 Simulation Model to predict the effects of level of nutrition on
composition of empty body gain in beef cattle.
Mathematical modelling in beef herd management 335
Table 1 – contd.
1 2 3 4
Lamb et al. 1992 Simulation Model for evaluation of mating systems involving five
breeds in integrated cow-calf-feedlot production enter-
prise.
Koots and Gibson 1998 Simulation Bio-economic herd level model to estimate the eco-
nomic values for beef production traits.
Kilpatrick and
Steen
1999 Simulation Model for prediction of beef cattle growth and carcass
composition.
Makulska
and Kristensen
1999 Markov
process
Model to optimize fattening strategy of an individual
young bull and a group of bulls.
Pang et al. 1999 Simulation Model for evaluation of the effects of calving season
and weaning age on bio-economic efficiency of beef
herd.
Tess and Kolstad 2000 Simulation Model to simulate dynamic relationships among beef
cattle genotypes, physiological states, forage quality
and management in range environments of Montana,
USA.
Nielsen and
Kristensen
2002 Multi-level
Markov
process
Model for determination of optimal decisions in or-
ganic steer production – including winter feed level,
grazing strategy and slaughtering policy.
Pihamaa and
Pietola
2002 Optimization
Dynamic
programming
Model for determination of optimal beef cattle man-
agement under agricultural policy reforms in Finland.
Williams and
Jenkins
2003 Simulation A dynamic model of metabolizable energy utilization
in growing and mature cattle.
Hoch and
Agabriel
2004 Simulation A dynamic model to estimate beef cattle growth and
body composition.
Costa and
Rehman
2005 Optimization
Linear
programming
Model for maximization of the asset value of cattle and
the economic returns from beef production systems of
Central Brazil.
Rotz et al. 2005 Simulation Model for simulating feed intake, animal performance
and manure excretion in beef farms.
Veysset et al. 2005 Optimization
Linear
programming
Model for determination of optimal combination of
production activities (animal and grassland) under
many constraints in France.
Wolfová et al. 2005 Simulation Model for economic evaluation of beef bulls’ utiliza-
tion in a variety of production systems in the Czech
Republic.
Crosson et al. 2006 Optimization
Linear
programming
Model for determination of the optimal beef produc-
tion systems in Ireland.
Havlik et al. 2006 Optimization
Linear
programming
Model for agri-environmental policy analysis involv-
ing suckler cow farm production system in the White
Carpathians, Czech Republic.
Makulska 2006 Optimization
Dynamic
programming
Model for supporting the decision processes in bull fat-
tening.
A. Stygar and J. Makulska
336
Table 1 – contd.
1 2 3 4
Oltjen and Ahmadi 2006 Optimization
Linear
programming
Model for ration formulation and projection of profit or
loss in a feedlot system.
Villalba et al. 2006 Simulation Model for stochastic simulation of mountain beef cattle
systems in Spanish Pyrenees.
Gradiz et al. 2007 Simulation Model for integration of beef cow–calf production sys-
tem with sugarcane production in Japan.
Reisenauer
Leesburg et al.
2007 Simulation Model for evaluation of calving seasons and marketing
strategies in beef enterprises located in Northern Great
Plains, USA.
Mathematical methodologies used in livestock herd modelling
Methodologically, livestock models, including beef cattle models, can be divid-
ed into optimization and simulation models (Figure 1). Optimization models allow
determining optimal outcomes given the objective function of expected utility or
function of profit that is maximized subject to production alternatives, prices and
resources availability. Simulation models are developed mainly to improve the un-
derstanding of the systems by studying their behaviour under different conditions.
They calculate the expected utility under a given set of parameters and decision
rules. When simulation models are used to determine “optimal” strategies the aim is
to find the optimal set of decision rules given the precision in the current knowledge
of the parameters (Kristensen and Jørgensen, 1996, 1999; Pla, 2007).
Within optimization and simulation two model categories can be distinguished:
deterministic and stochastic. In deterministic models the assumptions are that the
real system has no random variation while in stochastic models the variation of vari-
ables and parameters is represented through appropriate probability distributions.
The stochastic approach is usually more suited to solve the models of livestock
management since their parameters are associated with animals utilisation and there-
fore have a random component.
Models can also be classified as static and dynamic. In static models time is not
included as variable, so these models are not able to simulate the behaviour of the
system over time and therefore are hardly relevant to livestock management prob-
lems. On the contrary, dynamic models have time as an important driving variable
(Javlingh, 1992).
Linear and dynamic programming (with Markov decision processes) are the
mathematical methods often used in constructing and solving the optimization mod-
els. Dynamic programming and Markov processes are also very often applied to
solve the dynamic stochastic simulation models. Other methods, like decision graphs,
Bayesian networks and Monte Carlo simulations method are not so frequently used
for model based decision support and therefore they are not described in this paper
(Kristensen and Jørgensen, 1999).
Mathematical modelling in beef herd management 337
Figure 1. Methodological classification of livestock models
A. Stygar and J. Makulska
338
Linear programming (LP) is a method of mathematical programming in which
the objective is to maximize or minimize linear function subject to the restrictions
normally referred to as resource constraints. Four basic assumptions are essential to
determine whether LP is applicable to a particular problem and whether it will pro-
vide a meaningful and precise answer (Jalvingh et al., 1997):
– additivity and linearity in input and output coefficients;
– divisibility in resources and products;
– finiteness of alternative processes and resource restrictions;
– single-valued expectations.
The method most frequently used to solve linear programming problems is the
simplex algorithm. It specifies each step that is to be taken during the solution proc-
ess, and is actually a trial-and-error procedure for problem solving. However, it is
constructed in such a way that each trial results in an improved answer. The al-
gorithm guarantees that, if an optimal value exists, it will be found within a finite
number of steps (Heady and Chandler, 1958).
The information on the economic contribution of various resources to the measure
of performance (e.g. profit) is very useful for farm managers and animal producers.
The simplex method provides this information in the form of shadow prices for the
respective resources. The shadow price for a given resource measures the marginal
value of this resource, that is, the rate at which profit would be increased (slightly)
with the increase of the amount of this resource (Jalvingh et al., 1997).
Dynamic programming (DP) has become widely accepted as one of the main tools
for optimization (Bellman, 1957; Kristensen, 1987). DP is applicable to processes
involving a sequence of decisions over a given, finite or infinite period of time (plan-
ning horizon) split into stages. At each stage, the state of the process is observed and
a decision concerning the process has to be made. To solve finite stage decision prob-
lems the most commonly used method is value iteration. It consists in maximizing
(or minimizing) a value function, representing the expected total rewards (outcomes)
from the present stage until the end of the planning horizon. Optimal decisions de-
pending on stage and state are determined backwards step by step as those maximiz-
ing (or minimizing) the value function. This way of determining an optimal policy
is based on the Bellman principle of optimality: “An optimal policy has the property
that whatever the initial state and initial decision are, the remaining decisions must
constitute an optimal policy with regard to the state resulting from the first deci-
sion” (Bellman, 1957). For infinite stage problems the most relevant optimization
technique is policy iteration. This method was introduced by Howard (1960) who
combined the dynamic programming with the mathematically well-established no-
tion of a Markov chain. The combination, characterized by sequential and stochastic
approach, was named Markov decision process (MDP). Usually Markov decision
processes are considered as optimization models. However, since MDP takes into
account the probabilistic nature of herd, it can also be applied in dynamic stochastic
simulations. Use of simulation in Markov decision processes allows comparing the
consequences of different non-optimal and optimal policies (Kristensen, 1994; Pla,
2007).
Mathematical modelling in beef herd management 339
Optimization models
Linear programming
Livestock management applications of linear programming (LP) are numerous
(e.g. Glen, 1980; Olson et al., 1980; Costa and Rehman, 2005; Veysset et al., 2005;
Crosson et al., 2006; Havlik et al. 2006; Oltjen and Ahmadi, 2006). Most often LP is
used for whole-farm planning (Veyseet et al., 2005; Costa and Rehman, 2005; Cros-
son et al., 2006; Havlik et al., 2006) and for formulation of least-cost feeding rations
(Glen, 1980; Oltjen and Ahmadi, 2006).
Some recent examples of the application of linear programming to beef cattle
management are connected with the necessity of re-evaluating optimal systems in
beef farms after the reform of the Common Agricultural Policy (CAP) in 2003.
The models developed in Ireland (Grange Beef Model) and in France (Opt’INRA)
are used to identify the best solutions in these new circumstances. The aim of the
Grange Beef Model is to determine the optimal Irish beef production system based
on management alternatives focusing on grazing temperate grassland, within the
livestock and feeding specifications (Crosson et al., 2006). The French optimization
model Opt’INRA, described in 2005 by Veysset et al. (2005), determines the opti-
mal combination of various production activities (including animals and grassland).
The objective of the Opt’INRA model is to maximize the gross margin of the farm
subjected to numerous constraints (agronomic, agri-environmental, CAP, farm area,
housing, animal production, etc.). By taking into account organic farming limita-
tions and nitrogen balance Opt’INRA model was used to study the adaptation of the
production system and economic consequences of the transition of a cattle suckler
system to organic farming.
The important problems of a sustainable development of beef production and
environment protection were also reflected in the models of Havlik et al. (2006) and
Costa and Rehman (2005). Havlik et al. (2006) developed a linear model called Beef
and Grassland Biodiversity Production Optimisation Model (BEGRAB_PRO.1).
This model was used to analyse the organic suckler cow farms in the Protected Land-
scape Area of the White Carpathians, the Czech Republic. BEGRAB_PRO.1 enables
accounting not only for beef but also for biodiversity production. Biodiversity pro-
duction is depicted by a system of technical constraints, which represent limitations
and tasks to be respected in order to produce particular environmental goods. Costa
and Rehman (2005) devised the bi-criteria model aimed at maximization of the as-
set value of cattle and maximization of the economic returns from Brazilian beef
production systems, in the situation of a rapid spread of pasture degradation. In the
model the different attitudes of the farmers towards overgrazing, pasture costs and
capital availability were analysed. The model was also used to test various hypoth-
eses to explain the overgrazing behaviour.
The examples of the application of linear programming to feeding ration formula-
tion are the model and computer program of beef cattle management (TAURUS) de-
veloped by Oltjen and Ahmadi (2006), and the model of bull fattening devised by Makul-
ska (2006). Generally, TAURUS is intended to formulate the least-cost rations and
to project a profit or loss in feedlot operations. Output of the computer program
consists of six parts: cost and performance, ration composition, price ranges, nutrient
A. Stygar and J. Makulska
340
analysis of the ration, equations, and nutrient analysis of feeds in the ration. The
program predicts also days on feed, live weight, carcass yield, carcass quality and the
digestible energy of five different feed groups used in beef cattle diet formulation. In
the model of Makulska (2006) the least-cost feeding rations for all defined combi-
nations of fattened bull body weight and daily gain were formulated. The rations pro-
vided adequate levels of both energy and protein, within the limits of dry matter
intake.
Dynamic programming and Markov decision processes
Historically, first applications of the dynamic programming method to beef cattle
management were found in the papers by Bonnieux (1969), Nelson (1969) and Meyer
and Newett (1970). In 1972 Kennedy developed the dynamic model of beef bull fat-
tening and marketing. The model consisted of two decision systems: a live weight se-
quencing system and a ration composition system (Kennedy, 1972). A similar dyna-
mic programming problem, i.e. determination of the optimal feeding and marketing
strategies for pasture-fed beef cattle, was described in the model of Clark and Kumar
(1978).
The more recent research on the optimization of the bull fattening process by
means of dynamic programming is the model of Pihamaa and Pietola (2002). The
aim of this model was to increase the returns from Finnish beef production through
the determination of the optimal feeding and time of slaughtering under alternative
policy, price and forage cost scenarios. An important input to the model were subsi-
dies which still considerably influence optimal carcass weight and farmers’ income
in Finland. The dynamic programming, as a tool for gaining a valuable insight into
the factors that determine the efficiency of fattening process of young bulls, was
also described by Makulska (2006). In her model two decision problems were taken
into account: composition of feeding rations and strategy of fattening. Approaches
adopted in the dynamic optimization of fattening strategy included: fattening to the
assumed slaughter weight, fattening with the assumed duration, and cyclic fattening
with the replacement option.
Although Kennedy (1972) has found dynamic programming to be a flexible tool
for dealing with the dynamic problems of animal production he also mentioned the
drawbacks of this method. The main drawback manifests itself in the possibility of
handling only rather small models, with a few hundred states. This is connected with
the problem referred to as the “curse of dimensionality” that can be described in the
following way (Kristensen, 1994): If several variables are considered simultaneously
and each variable is considered at a realistic number of levels, the state space grows
to prohibitive dimensions and model becomes very large. Furthermore, in livestock
management a hierarchical structure of decisions is often faced. The decisions are
made not only at different levels that are mutually dependent but also at different
time horizons. These aspects contribute even more to the dimensionality problem
than the extension of state space mentioned above. In such situation a conside-rably
increased computer memory is required. More complicated multi-state processes,
involving decisions with varying time horizon, can be optimized by means of an
efficient DP algorithm, i.e. the hierarchical Markov process (HMP), developed by
Mathematical modelling in beef herd management 341
Kristensen (1988). HMP is defined as a series of Markov decision processes called
sub-processes built together in one main Markov process (hierarchic structure of
decision processes), so that each stage in the main process represents a sub-proc-
ess. Markov decision programming technique has been mainly applied to solve the
animal replacement problem, defined by Van Arendonk (1984) as follows: If the
asset (animal) is used in production process, it is relevant to examine at regular time
intervals whether the present asset (animal) should be replaced or it should be kept
for an additional period.
In 1999 an attempt to use hierarchic Markov process to optimize fattening strat-
egy of an individual bull and a group of bulls was undertaken by Makulska and
Kristensen (1999). A special emphasis was put on supporting the decision when to
terminate the fattening process. The optimization of fattening strategy considered
various breeds (beef, dairy and crossbred bulls), two scales of production (single-
animal level – small farms, and group level – large farms) and different intensities of
fattening (intensive, semi-intensive, extensive).
In order to circumvent the “curse of dimensionality” in Markov decision pro-
gramming models to more satisfactory extent than with just two levels (main proc-
ess and sub-processes), Kristensen and Jørgensen (2000) introduced the notion of
a multi-level hierarchical Markov process. The basic idea of the multi-level design
is to expand stages of the main (founder) process to a so-called child process, which
again may expand stages further to new child processes leading to multiple levels.
For representation and solution of multi-level hierarchical Markov processes Java
software system (MLHMP) has been developed by Kristensen (2003).
The example of the application of multi-level hierarchic Markov process is the
model of steer management presented in the papers of Nielsen and Kristensen (2002)
and Nielsen et al. (2004). This model is a four-level hierarchical Markov process
with decisions defined only at three levels. The optimized decisions concern:
– grazing strategy (permanent or ryegrass/white clover pasture);
– feeding level in winter (high and low);
– finishing strategy (age 19–27 months);
– time of slaughter (age 19–30 months).
The objective is to optimize economically the organic steer production at single-
animal level. In order to calculate technical and economic key figures (feed intake,
body weight gain and net returns) characterizing the optimal policy a probabilistic
Markov chain simulation was used (Nielsen and Kristensen, 2007).
Simulation models
Simulation models are well suited to dealing with variability and complexity of
animal production. They can be divided into three categories: whole-herd models
with emphasis on management strategies, physiological models of whole herds, and
physiological models of single animals (Kristensen and Jørgensen, 1996). Beef cattle
simulations are often simplified by disregarding variability beyond that created by
the model’s deterministic equations (Shafer et al., 2007). Such approach yields lower
levels of simulated variability than that typically occurring in nature. Nevertheless,
numerous examples of deterministic models are found in literature (Lamb et al.,
A. Stygar and J. Makulska
342
1992 a, b, c, 1993; Koots and Gibson, 1998; Pang et al., 1999 a, b; Rotz et al., 2005;
Wolfová et al., 2005). One of them is the Integrated Farm System Model (IFSM)
developed by Rotz et al. (2005). This whole-farm simulation model is a compre-
hensive presentation of farm production where beef breeding constitutes one of the
branches. It incorporates a beef herd sub-model with other farm components such as
crop growth, harvest, storage, feeding, grazing and manure handling. IFSM enables
predicting nutrient requirements, feed intake, growth rate and manure excretion for
all animal groups making up a beef herd. Since the IFSM integrates many biological
and physical processes in the farm production system it can be a useful tool for eva-
luation and comparison of the long-term performance, economics and environmen-
tal impacts of beef production system. The whole-farm approach is also used in the
model of integrated beef cow-calf and sugarcane production in Tanegashima Island,
Japan (Gradiz et al., 2007). This model simulates the total requirement for energy
and protein, and subsequent losses of nitrogen via faeces and urine, throughout the
reproduction cycle of a mature cow and the growing stages of her calf.
Deterministic simulation models were applied to evaluate biological and
economic efficiency of purebreds, two-breed and three-breed rotational crossbreds,
involving five cattle breeds in the USA. The following approaches were assumed:
cow-calf segment (Lamb et al., 1992 a); feedlot segment (Lamb et al., 1992 b); and
integrated cow-calf-feedlot system (Lamb et al., 1992 c). Moreover, Lamb et al.
(1993) tried to account for genetic trends within breeds included in the simulations
and to examine variability in the average carcass performance.
Other applications of deterministic simulation to model beef cattle production
systems are those presented by Koots and Gibson (1998), Pang et al. (1999 a, b) and
Wolfová et al. (2005). The model of Koots and Gibson (1998) derived economic
values for genetic improvement of multiple traits in the integrated beef enterprise.
Modelling a complete beef production system (as opposed to cow-calf and feedlot
segments separately) was necessary to reflect a situation where market signals flow
down to those making the breeding decisions. That is, although payment is based on
carcass value, animals must flow through both cow-calf and feedlot segments. The
model of Pang et al. (1999 a, b), called Alberta Beef Production Simulation System
(ABPSS), also described the situation of complete beef production. It was composed
of herd inventory, nutrient requirements, forage production and economic sub-mo-
dels. The herd inventory sub-model was used to evaluate population dynamics and
feed requirements in the herd. The nutrient requirements sub-model simulated nu-
trients and feed requirements for calves and cows depending on their physiological
status (maintenance, growth, lactation and gestation) and the climatic conditions.
The use of forage production sub-model allowed predicting forage growth rate, cattle
grazing rate, available forage biomass, and total hectares required for grazing. The
economic sub-model measured bio-economic efficiency, as net return per cow, by
subtracting total cost from total return. The ABPSS model was also applied to simu-
late the influence of calving season and weaning age on the bio-economic efficiency
of beef production systems (Pang et al., 1999 a). Wolfová et al. (2005) developed the
bio-economic deterministic simulation model to evaluate the utilization of bulls in
a variety of production systems in the Czech Republic. The model can simulate life-
Mathematical modelling in beef herd management 343
cycle production of beef cow herds with and without integrated feedlot system. It can
be a valuable tool for the optimization of mating, culling and other management and
marketing strategies in various beef production systems. To simulate herd dynamics
the Markov chain approach was adopted. The herd was described in terms of ani-
mal’s states and possible transitions among particular states at different stages. The
model algorithms served as a basis to write the computer program ECOWEIGHT.
This program was used to estimate marginal economic values for 16 traits in four dif-
ferent management systems of beef bulls in the Czech Republic. ECOWEIGHT was
also applied by Krupa et al. (2005) to calculate the economic weights for produc-
tion and functional traits of Slovakian Simmental cattle under alternative marketing
strategies.
In order to simulate more realistic levels of variability occurring in animal pro-
duction systems and in life of single animals, a variability created by the determinis-
tic equations of the model should be supplemented by that stochastically generated
(Shafer et al., 2007). Both deterministic and stochastic simulation were used in the
model of Werth et al. (1991). The model was devised to evaluate how reproductive
performance interacts with management practices to influence net income in a cow-
calf operation for one year of production. The stochastic dynamic model was applied
to simulate the reproduction performance of the cow-herd. Outputs from the stochas-
tic model were used as inputs into the deterministic cow-herd economic simulation
model that calculated the net income.
A dynamic stochastic model for simulating mountain beef cattle systems in the
Spanish Pyrenees was described by Villalba et al. (2006). The model was used to
assess four feeding strategies during the winter period in the conditions of autumn
cal-ving. Special attention was devoted to the evaluation of the relationship between
nutrition and reproductive performance of cows, considering the information about
production and reproduction variability of the studied groups of animals. Tess and
Kolstad (2000 a) developed a stochastic model aimed at simulation of the dynamic
relationships among beef cattle genotypes, physiological states, forage quality and
management in range environments of Montana (USA). Forage intake, energy and
protein metabolism, growth, reproduction, lactation and differences in chemical
body composition were simulated for individual animals over complete life cycles.
The model was applied to evaluate the response of the production and marketing sys-
tem to the changes in breeding and management strategies (Tess and Kolstad, 2000
b). In 2007 a similar bio-economic computer model of cow-calf enterprise was sed
by Reisenauer Leesburg et al. (2007 a, b) to assess various calving seasons and dif-
ferent calf marketing strategies under conditions of the Northern Great Plains (USA).
The simulated ranch utilized a rotational breeding system based on Hereford and
Angus.
The simulation methods were also applied to devise the models of growth and
metabolism of beef cattle. An example is the model described in the papers of Keele
et al. (1992) and Williams et al. (1992 a, b), developed to predict composition of
empty body gain of several breeds of beef cattle fed at different levels of nutrition.
Also, Kilpatrick and Steen (1999) simulated the influence of the feeding regime
(either silage only or supplemented with concentrates) on beef cattle growth and
A. Stygar and J. Makulska
344
carcass composition. Their model provided information on the most economic level
of concentrate feeding to achieve the animal growth and quality of carcass com-
position required. Williams and Jenkins (2003) used the simulation model to pre-
dict heat production attributable to maintenance and support metabolism in growing
and mature purebred and crossbred cattle. Hoch and Agabriel (2004 a, b) designed
a mechanistic dynamic model in which beef cattle growth and body composition
were simulated for different animal types (sex, breed) under various nutritional con-
ditions. Mechanistic modelling does not require much data for model development
but it assumes basic understanding of the process. The equations in the mechanistic
model are derived from some theory or hypothesis about the fundamental nature
of the process. This is in contrast to other quantitative models, which use the equa-
tions derived from observations in the real world, but not necessarily representing
any understanding of the casual mechanism at work (Morris, 2006). The model of
Hoch and Agabriel (2004 a, b) was constructed on the basis of variations of body
protein and lipid contents. Proteins and lipids in carcass and non-carcass tissues were
distinguished to account for different energy metabolism of these two components
of the body. Evolution of each compartment was determined by the instantaneous
balance between synthesis and degradation, which depends on the physiological age
of the animals and on metabolizable energy supply. Empty and full body weights
were deduced from protein and lipid contents through allometric equations which in
biology are used to describe the morphological evolution of species, and are based
on the relation between an organism’s size and the size of any part of the organism
(Warriss, 2000).
In conclusion, mathematical modelling can be applied to solve complex decision
problems appearing in livestock management. Most often decisions concern feeding,
insemination, marketing, culling/replacement of animals.
Beef cattle models display many similarities in their mathematical formulation
despite a considerable variety in relation to the undertaken problem. Methods that
are commonly employed in modelling are optimization and simulation. Many mod-
els adopt the dynamic approach, which is relevant considering the changes in herd
composition and single animal performance over time. The necessity of taking into
account a random variation, typically occurring in animal breeding and production,
causes that a large number of models are stochastic.
Nowadays the mathematical models of beef cattle management are used mainly
as research tools and teaching aids. Unfortunately, still not so many models are ap-
plied directly in supporting decisions in commercial herds. In order to create an ef-
ficient decision support system for beef cattle, further development of the adequate
methodology maximizing the farmer’s utility is necessary. Since various methods
have different properties a main challenge will probably be to combine some meth-
ods. Taking into consideration more and more variables and parameters results in
the present models becoming very large. Therefore, a high priority should be given
to the circumvention of the dimensionality problems. Another important issue is
data acquisition and their transformation for filtering and organization of databases.
The scarcity of real farm data often significantly hinders the estimation of model
parameters at the herd level and the external validation of the devised model. An
Mathematical modelling in beef herd management 345
extensive research on the methodology observed in last years is not accompanied by
a simultaneous development of software tools and the increase of farmers’ experi-
ence in computer handling. Hence, a more widespread use of model-based decision
support methods at the farm level demands a close cooperation between researchers,
and advisory and training services. Equally important is the accessibility to powerful
personal computers and the development of comprehensible software well fitted to
the needs of herd managers.
References
B e l l m a n R. (1957). Dynamic programming. Princeton University Press, Princeton.
B o n n i e u x F. (1969). Dynamic programming and Markov chains analysis applied to management of
a veal calf fattening operation (in French). Rech. Econ. Soc. Rur., 2: 29–40.
B r u c e L.B., T o r e l l R.C., H u s s e i n H.S. (1999 a). Profit prediction in cow/calf operations.
1. CowCost software program. J. Prod. Agric., 12: 644–647.
B r u c e L.B., T or e l l R.C., H u s s e in H.S. (1999 b). Profit prediction in cow/calf operations.
2. Influence of major management practices. J. Prod. Agric., 12: 647–649.
C l a r k J., K u m a r S. (1978). Planning beef production: An application of dynamic programming.
Rev. Market. Agricult. Econom., 46, 3: 315–326.
C o s t a F.P., R e h m a n T. (2005). Unravelling the rationale of ‘overgrazing’ and stocking rates in the
beef production systems of Central Brazil using a bi-criteria compromise programming model. Agr.
Syst., 83: 277–295.
C r o s s o n P., O’ K ie l y P., O ’ Ma r a F.P., W a l l a ce M. (2006). The development of a math-
ematical model to investigate Irish beef production systems. Agr. Syst., 89: 349–370.
G l e n J.J. (1980). A parametric programming method for beef cattle ration formulation. J. Operat. Res.
Soc., 31: 689–698.
Gradiz L., Sugimoto A., Ujihara K., Fukuhara S., Kahi A.K., Hirooka H. (2007).
Beef cow–calf production system integrated with sugarcane production: Simulation model develop-
ment and application in Japan. Agr. Syst., 94: 750–762.
H a v l i k P., J a c q u e t F., B o i s s o n J.-M., H e j d u k S., V e s e l y P. (2006). Mathematical pro-
gramming models for agri-environmental policy analysis: A case study from the White Carpathians.
Agricult. Econom. - Czech, 52: 51–66.
H e a d y E.O., C h a n dl e r W. (1958). Linear programming methods. The Iowa State University Press,
Ames.
H o c h T., A g a b r i e l J. (2004 a). A mechanistic dynamic model to estimate beef cattle growth and
body composition: 2. Model description. Agr. Syst., 81: 1–15.
H o c h T., A g a b r i e l J. (2004 b). A mechanistic dynamic model to estimate beef cattle growth and
body composition: 1. Model evaluation. Agr. Syst., 81: 17–35.
H o w a r d R. (1960). Dynamic Programming and Markov Processes. The M.I.T. Press, Cambridge,
Massachusetts.
J a l v i n g h A.W. (1992). The possible role of existing models in on-farm decision support in dairy
cattle and swine production. Livest. Prod. Sci., 31: 351–365.
J a l v i n g h A.W., D i j k h u i z en A.A., R e n k e m a J.A. (1997). Linear programming to meet
management targets and restrictions. In: Animal Health Economics. Principles and Applications.
A.A. Dijkhuizen, R.S. Morris. (Eds). Post Graduate Foundation in Veterinary Science, University
of Sydney, pp. 69–84.
K e e l e J.W., W i l l i a m s C.B., B e n n e tt G.L. (1992). A computer model to predict the effects of
level of nutrition on composition of empty body gain in beef cattle. I. Theory and development.
J. Anim. Sci., 70: 841–857.
K e n n e d y J.O.S. (1972). A model for determining optimal marketing and feeding policies for beef
cattle. J. Agric. Econ., 23, 2: 147–159.
A. Stygar and J. Makulska
346
K i l p a t r i c k D.J., S t e e n R.W.J. (1999). A predictive model for beef cattle growth and carcass com-
position. Agr. Syst., 61: 95–107.
K o o t s K.R., G i b s o n J.P. (1998). Economic values for beef production traits from a herd level bio-
economic model. Can. J. Anim. Sci., 78: 29–45.
K r i s t e n s e n A.R. (1987). Optimal replacement and ranking of dairy cows determined by a hierarchic
Markov process. Livest. Prod. Sci., 16: 131–144.
K r i s t e n s e n A.R. (1988). Hierarchic Markov processes and their applications in replacement models.
Eur. J. Oper. Res., 35: 207–215.
K r i s t e n s e n A.R. (1994). A survey of Markov decision programming techniques applied to the ani-
mal replacement problem. Eur. Rev. Agric. Econ., 21: 73–93.
K r i s t e n s e n A.R. (2003). A general software system for Markov decision processes in herd manage-
ment applications. Comput. Electron. Agr., 38: 199–215.
K r i s t e n s e n A.R, J ø r g e n s e n E. (1996). Textbook notes of herd management: Basic concepts.
Dina Notat No. 48.
K r i s t e n s e n A.R., J ø r g e n s e n E. (1999). Decision support models. Proc. of 25th International
Dairy Congress, 21–24 September 1998, Aarhus, Denmark. Vol. III, Future Milk Farming,
pp. 145–163.
K r i s t e n s e n A.R, J ø r g e n s e n E. (2000). Multi-level hierarchic Markov processes as a framework
for herd management support. Ann. Oper. Res., 94: 69–89.
of traits for Slovakian Pied cattle under different marketing strategies. Czech J. Anim. Sci., 50:
483–492.
L a m b M.A., T e ss M.W., R o b i s o n O.W. (1992 a). Evaluation of mating systems involving
five breeds for integrated beef production systems: I. Cow-calf segment. J. Anim. Sci., 70:
689–699.
L a m b M.A., T e s s M.W., R o b i s o n O.W. (1992 b). Evaluation of mating systems involving five
breeds for integrated beef production systems: II. Feedlot segment. J. Anim. Sci., 70: 700–713.
L a m b M.A., T e s s M.W., R o bi s o n O.W. (1992 c). Evaluation of mating systems involving five
breeds for integrated beef production systems: III. Integrated system. J. Anim. Sci., 70: 714–722.
L a m b M.A., T e s s M.W., R o b i s o n O. W. (1993). Evaluation of mating systems involving five
breeds for integrated beef production systems: IV. Accounting for variability and genetic trends.
J. Anim. Sci., 70 : 689–699.
M a k u l s k a J. (2006). Application of the mathematic modelling in the support of the decision pro-
M a k u l s k a J., K r i s t e n s e n A.R. (1999). Economic optimization of bull fattening. (In:) Perspec-
tives of Modern Information and Communication Systems in Agriculture, Food Production and
Environmental Control. G. Schiefer, R. Helbig, U. Rickert (Eds). Bonn, pp. 443–449.
M e y e r C.F., N e w e t t R.J. (1970). Dynamic programming for feedlot optimization. Manage. Sci.,
16, 6 : 683–688.
M o r r i s T.R. (2006). An introduction to modelling in the animal sciences. In: Mechanistic Modelling
in Pig and Poultry Production. C. Fisher and T. Morris (Eds). CABI Publishing, pp. 1–5.
N e l s o n A.G. (1969). The feasibility of an information system: the beef feedlot case, Unpublished PhD
thesis, Purdue University, Lafayette.
N i e l s e n B.K., Kristensen A.R. (2002). A model for simultaneous optimization of feeding level and
slaughtering policy of organic steers. In: Proc. of First European Workshop on Sequential Decisions
under Uncertainty in Agriculture and Natural Resources. INRA, Toulouse, France, pp. 27–32.
N i e l s e n B.K., K r i s t e n s e n A.R., T h am s b o r g S.M. (2004). Optimal decisions in organic steer
production – a model including winter feed level, grazing strategy and slaughtering policy. Livest.
Prod. Sci., 88: 239–250.
N i e l s e n B.K., K r i s t e n s e n A.R. (2007). Optimal decisions in organic beef production from steers
– Effects of criterion of optimality and price changes. Livest. Sci., 110: 25–32.
O l s o n T.A., W i l l h a m R.L., B o e h l j e M.D. (1980). Linear programming as a tool for planning
beef cattle breeding experimentation. J Anim. Sci., 51: 847–859.
market with the particular attention paid to the changes of the purchase prices and their relations
Mathematical modelling in beef herd management 347
between the years 2003–2007 (in Polish). Stowarzyszenie Ekonomistów Rolnictwa i Agrobiznesu,
Rocz. Nauk., X, 4: 288–293.
O l t j e n J.W., A h m a d i A. (2006). TAURUS Beef Cattle Ration Formulation and Evaluation Soft-
ware: Windows Version., Proc. of 4th World Conference on Computers in Agriculture and Natural
Resources. 23–25 July 2006, Orlando, Florida, USA, pp. 09–413.
P a n g H., M a k a r e c h i a n M., B a s a r a b J.A., B e r g R.T. (1999 a). Application of dynamic simu-
lation model on the effect of calving season and weaning age on bioeconomic efficiency. Can.
J. Anim. Sci., 79: 419–424.
P a n g H., M a k a r e c h i a n M., B a s a r a b J.A., B e r g R.T. (1999 b). Structure of a dynamic simu-
lation model for beef cattle production systems. Can. J. Anim. Sci., 79: 409–417.
P i h a m a a P., P i e t o l a K. (2002). Optimal beef cattle management under agricultural policy reforms
in Finland. Agr. Food Sci., Finland, 11: 3–11.
P l a L.M. (2007). Review of mathematical models for sow herd management. Livest. Sci., 106:
107–119.
R e i s e n a u e r L e e s b ur g V.L, T e s s M.W., G r i f f i t h D. (2007 a). Evaluation of calving
seasons and marketing strategies in Northern Great Plains beef enterprises: I. Cow-calf systems.
J. Anim. Sci., 85: 2314–2321.
R e i s e n a u e r L e e s b u r g V.L, T e s s M.W., G r i f f i t h D. (2007 b). Evaluation of calving sea-
sons and marketing strategies in Northern Great Plains beef enterprises: II. Retained ownership
systems. J. Anim. Sci., 85: 2322–2329.
R o t z C.A., B u c k m a s t e r D.R., C o m e r f o r d J.W. (2005). A beef herd model for simulating feed
intake, animal performance, and manure excretion in farm systems. J. Anim. Sci., 83: 231–242.
S h a f e r W.R., Bo u r d o n R.M., E n n s R.M. (2007). Simulation of cow-calf production with and
without realistic level of variability. J. Anim. Sci., 85: 332–340.
T e s s M.W., K o l s t a d B.W. (2000 a). Simulation of cow-calf production systems in a range environ-
ment: I. Model development. J. Anim. Sci., 78: 1159–1169.
T e s s M.W., K o l s t a d B.W. (2000 b). Simulation of cow-calf production systems in a range environ-
ment: II. Model evaluation. J. Anim. Sci., 78: 1170–1180.
V a n A r e n d o n k J.A.M. (1984). Studies on the replacement policies in dairy cattle. I. Evaluation of
techniques to determine the optimum time for replacement and to rank cows on future profitability.
V e y s s e t P., B e b i n D., L h e r m M. (2005). Adaptation of Agenda 2000 (CAP reform) and optimi-
sation of the farming system of French suckler cattle farms in Charolaise area: a model-based study.
Agr. Syst., 83: 179–202.
V i l l a l b a D., C a s a s u s I., Sa n z A., B e r n u e s A., E s t a n y J., R e v i l l a R. (2006). Stochastic
simulation of mountain beef cattle systems. Agr. Syst., 89: 414–434.
W a r r i s s P.D. (2000). Meat Science An Introductory Text. CABI Publishing.
W e r t h L.A., A z z a m S.M., N i e ls e n M.K., K i nd e r J.E. (1991). Use of a simulation model to
evaluate the influence of reproductive performance and management decisions on net income in
beef production. J. Anim. Sci., 69: 4710–4721.
W i l l i a m s C.B., K e e l e J.W., B e n n e t t , G.L. (1992 a). A computer model to predict the effects
of level of nutrition on composition of empty body gain in beef cattle. II. Evaluation of the model.
J. Anim. Sci., 70: 858–866.
W i l l i a m s C.B., K e e l e J.W., W a l d o D.R. (1992 b). A computer model to predict empty body
weight in cattle from diet and animal characteristics. J. Anim. Sci., 70: 3215–3222.
W i l l i a m s C.B., J e n k i n s T.G. (2003). A dynamic model of metabolizable energy utilization in
growing and mature cattle. I. Metabolizable energy utilization for maintenance and support metabo-
lism. J. Anim. Sci., 81: 1371–1381.
for beef cattle used in different production system: 1. Model development. Livest. Prod. Sci., 95:
201–215.
of the economic efficiency of beef cattle production systems. Czech J. Anim. Sci., 49: 357–37.
Accepted for printing 5 X 2010
A. Stygar and J. Makulska
348
ANNA STYGAR, JOANNA MAKULSKA
Zastosowanie modelowania matematycznego w zarządzaniu stadem bydła mięsnego – artykuł
przeglądowy
-
-
-
namiczne z rozszerzeniem o tzw. procesy decyzyjne Markova. Programowanie dynamiczne i procesy
-
-
badawcze i pomoce dydaktyczne. Szersze ich wykorzystanie do wspomagania decyzji podejmowanych
-
gramowania.