MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.

Sirotenko and V.A. Romanenkov

MATHEMATICAL MODELS OF AGRICULTURAL SUPPLY

O.D. Sirotenko
All Russian Institute of Agricultural Meteorology, Obninsk, Russia

V.A. Romanenkov
All Russian Institute for Fertilizers and Agricultural Soil Science, Moscow, Russia

Keywords: Economic-mathematical models, decision making, optimization, simplex

method, sown areas, fertilizer distribution, livestock rations, precision agriculture.

Contents

1. Introduction
2. Models and decision making in agriculture

S
TE S
3. Mathematical models of optimization and allocation of sown areas

R
AP LS
4. Mathematical models of fertilization optimization
5. Complex optimization of resource allocation in crop growing
6. Economic-mathematical models of optimization of structure of herds and flocks
C EO
7. Economic-mathematical models of optimization of rations of cattle feeding
8. Economic-mathematical models of optimization of combination of several branches
in a farm
9. Economic efficiency of precision agriculture farm application
Glossary
E

Bibliography
H

Biographical Sketches
PL O
M SC

Summary

Environmental protection principles which have emerged in the modern agriculture

specify strict limits for the adopted farming systems. At the same time the solution of
SA NE

food supply problem for the growing world population remains an acute task. Therefore,
various resource-saving programs acquire a great significance for the sustainable
agriculture development. This paper describes an approach for solution of the above
U

mentioned problems, based on mathematical modeling and computer technology

approaches.

The optimization problems are generalized for land, fertilizer, irrigation water, facilities,
labor use in plant-growing, the herd structure and feeding rations in livestock.
Mathematical methods of resource utilization optimization have been used in practice
from the end of the Second World War. First mathematical programming approaches
include the method of linear programming (simplex method). From that time, the
models of agricultural processes and the optimization methods have been developed
quickly. At the moment deterministic statements of problems prevail, but the number of
problems increases with allowance for uncertainty elements connected with weather and
climate solved by methods of stochastic programming. Agroecosystem management
needs rather complex dynamical models. Based on the specific problem or the specific
parameters a wide range of methods from classical analytical studies of the systems of

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

differential equations to numerical simulation was suggested within the expert systems
of decision making support. Managing field variability by using information technology
tools for handling information offer a possibility to solve the optimization task based on
the development of site-specific farming methods in agriculture.

1. Introduction

The utilization of natural resources in the existing technologies of the agriculture

management is close to the saturation condition. The exponential growth of
irreplaceable energy, water and fertilizer consumption is observed per every extra
production unit. Simultaneously, the supply of essential natural resources, necessary for
food production, steadily decreases per capita.

Therefore, various resource-saving programs (lands, fresh water, energy raw materials,
phosphoric and potassium fertilizers, etc.) acquire a great significance for sustainable

S
TE S
agriculture development. The methods of enhancement of resource utilization efficiency
in agriculture based on application of mathematical models of the corresponding

R
AP LS
processes and optimization methods play the key role in these programs.

Modern computer expert systems for supporting decision making in agriculture based
C EO
on developed and complex dynamical models of agroecosystem productivity are the key
systems in this field of studies. There is a natural transition from linear to nonlinear
objective functions and restrictions and from statistical to dynamical models. The role of
different stochastic problems of resource allocation optimization connected with the
E

application of current and prognostic information, in particular meteorological, steadily

H

increases.
PL O

One more important point is practical realization of ideas of the so-called precision
M SC

(exact) farming which results in a sharp rise in the volume of applied information
(increase in the space and time resolution of data on soil conditions). These data are
used to enhance the efficiency of fertilization, to optimize the plant protection means.
SA NE

The decision of optimization problems is facilitated by available software packages

realizing different sets of numerical algorithms of optimization. The availability of these
packages, however, is a necessary but not sufficient condition for an effective solution
U

of new problems. The analysis of physical peculiarities of applied models and the
possibility of utilization of the methods of mathematical analysis, including classical
ones, are also essential. It is very important for more thorough understanding of the
essence of problems to be solved where the use of standard packages promotes only a
little.

2. Models and decision making in agriculture

Problems of resources allocation arise in all cases when it is essential to implement a

variety of activities under the restrictions associated either with resources availability or
with the possibility of their utilization. In these situations the problem of optimal
resources allocation between the forms of activities is important. Many such problems
can be solved by mathematical programming or other related methods. It is assumed

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

here that all parameters in the economic-mathematical model (resources, technical-

economic coefficients and coefficients of the objective function) are deterministic, a
priori known quantities. Problems considering the risk and uncertainty factors, in which
some (and in the general case all) parameters are random values, are known as
stochastic problems of resources allocation. Stochastic is the integral attribute of
agricultural production connected with the uncertain character of the environment and,
primarily, weather conditions.

Consider the statement of a general problem of linear programming in a matrix form:

find

F(x) = C*X max at AX < B, X >0.

Here the matrix A and the vectors B and C are deterministic; it is the most frequently

S
encountered statement of a problem. In stochastic problems A, B and C may be random.

TE S
Problems of stochastic programming differ considerably in the objective function. The

R
AP LS
following objective functions are used:

a) Mathematical expectation of the quantity in a linear form:

C EO
F(x) = M C*X max (min) ,

where M is the symbol of mathematical expectation. It is the primary objective function

E
H

for problems, the optimization criterion of which is the maximum gross production,
maximum profit, etc.
PL O
M SC

b) The probability that the linear form exceeds a certain fixed level:

P[C*X > k ] k] max,

SA NE

where k is a given number. For example, the problem is to find the maximum
probability that the profit will be not less than a given number k.
U

c) Dispersion of a linear form:

X*DX min ,

where D is the quadratic matrix, whose elements are the yield dispersions and
covariances, X* is the transposed vector X. This optimization criterion is reduced to the
option of a plan, in which dispersion of a certain index (gross production, income) has
minimum value. Decrease in the effective index variability of agricultural production,
i.e. its stability rise seems to be extremely important.

d) Linear combination of mathematical expectation and dispersion of linear form:

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

C*X X*DX max,

where is the penalty for dispersion unit. Here mathematical expectation of a linear
form is maximized, however, the penalty for dispersion unit is introduced. When using
this criterion of optimality, the choice of , i.e. the objective assessment of dispersion
effect on the plan quality is difficult.

For solving problems in stochastic programming two approaches are possible: the
desired plan of the problem can be considered as a deterministic or random vector. In
the first case it means that the decision obtained is applied constantly (e.g. during some
years despite observed agrometeorological conditions). The problem may be formulated
in such a way that the decision obtained earlier is corrected or reconsidered depending
on the observed conditions and/or results of forecasting. The specific importance of
such problems keeps growing, with the prospects of application of flexible technologies,

S
TE S
adapted and changed in conformity to forecaster conditions.

R
AP LS
The next example illustrates the case when the decision of the problem is beyond the
limits of mathematical programming. Table 1 presents the income matrix characterizing
cultivation efficiency of four crops k differed in their water requirements.
C EO
F K
K1 K2 K3 K4
F1 5 0 -1 -2
E

F2 2 3 1 -1
H

F3 -4 1 5 -1
PL O

F4 -1 0 1 1
M SC

Table 1. Cultivation efficiency (conditional unit) of different crops depending on water

supply

Positive numbers here correspond to the cases when crop cultivation yields a good
SA NE

return, and negative numbers show losses.

F
U

1 2 3 4
F1 0.05 0.02 0.02 0.01 0.10
F2 0.03 0.10 0.03 0.04 0.20
F3 0.02 0.03 0.03 0.05 0.40
F4 0.05 0.03 0.02 0.20 0.30
0.15 0.18 0.37 0.30 1.0

Table 2. Conjugation matrix for four-phase forecast of summer precipitation

Imagine that a farmer obtains the forecast of summer precipitation done in good time
and, depending on this forecast, decides what crop is preferable. Forecast characteristics
and climatic recurrences of different gradations of summer precipitation are presented in
Table 2.

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

Multiplication of the income matrix

5 2 4 1
0 3 1 0
Uij 4x4 =
1 1 5 1
2 1 1 1

by the conjugation matrix presented in Table 2

0.05 0.02 0.02 0.01

0.03 0.10 0.03 0.04
Pij =
4x4 0.02 0.03 0.03 0.05
0.05 0.03 0.02 0.02

S
TE S
R
AP LS
gives the matrix-product

0.18 0.15 1.06 0.27

C EO
0.11 0.33 0.39 0.17
Uij = .
4x4 0.13 0.26 1.53 0.48
0.10 0.14 0.35 0.09
E
H

The maximum elements in each column are underlined here. Considering the position of
PL O

these elements in lines, the following rule can be formulated for optimal decision
making in the economy:
M SC

with small predicted quantity of precipitation (gradation F1 ) it is desirable to

cultivate the most drought-resistant crop K1 ;
SA NE

if moderate precipitation is expected over the year (F2) the crop K2 will be the most
rewarding;
in the event of forecasts of heavy (F3) and very heavy (F4) precipitation the crop K3
is to be cultivated.
U

So, in three of the four cases it is profitable to trust the forecast; in the fourth it is
desirable to cultivate the crop K3 instead of the crop K4 according to the forecasts.

By summing the maximum elements in a matrix-product, calculation of the mean

income under an optimal strategy application gives 2.52 conditional units for a farm
whereas with full reliance on the forecast this value would be 2.13 conditional units.

In the case when the mathematical model is dynamic and can be presented by
differential equations, the corresponding problems of management can be solved by
classical methods, e.g. using the Pontryagin maximum principle. The dynamics of grass
biomass under multi-cutting management for hay making was described by Sirotenko:

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

dm m m2
=D , m(0) = m0 ,
dt m+F

Ro Ro M
D= ,F= ,m= ,
1 + Rr k b Ro M max

where M is the dry crop biomass and Mmax is its maximum value, D, R0, Rr, ,k and b
are constants, m = m0 when t=0.

The following task of optimal management is set:

m(t)u(t)dt max,
0

S
TE S
R
AP LS
dm
= f ( m ) U m , m 0 = m ( T ) = m (0),0 u < U .
dt
C EO
Here:

m m2 df
f (m) = D , f (0) = f (1) = 0, = 0 at m = = F(1 + F) F .
m+F dm
E
H

Management u(t) is the specific rate of hayingof phytomass, and U denotes the
PL O

maximum rate.
M SC

Shlyachkova showed that decision of the task is possible based on the Pontryagin
maximum principle:
SA NE

0, t (0, max),

U (t ) = f () , t ( max, t T ),
Um, t (t T , T ),

U

where max is the time during which the rate of crop growth reaches maximum,
t T denotes the beginning of the time period appropriate for the maximum number of
cuttings , T is the end of the cutting period.

Another interesting example of management is proposed by Noy Meir I, who formulates

the following equation for pasture biomass dynamics

dV V V Vr
= gV(1 ) Cm H,
dt Vm (V Vr ) + ( Vk Vr )

where V, Vm are, respectively, the current green biomass and its maximum value, kg m-

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

2
; g is the maximum specific growth V, day-1 ; Cm is the maximum rate of biomass
consumption, kg (animal per day)-1 ;Vr is the biomass of unpalatable crop residuals, kg
m-2 ; Vk is the Michaelis-Menthen constant specifying biomass equal to of that
required for an animal intake to be sated ; H is the grazing capacity, animals m-2.

The stability of the system under permanent pasture regime is verified from the balance
dV
condition =0 with determination H and V values at which this balance is possible.
dt
It turned out that

* gV* (Vm V* ) (V* + Vk 2Vr )

H =
Cm Vm (V* Vr )

and the system has the only point of balance

S
TE S
R
AP LS
gVK
H* < ,
Cm
C EO
which is evaluated as safe capacity. The latter term denotes permissible carrying
capacity of the pasture which can guarantee absence of manifestation of system
degradation phenomena.
E

The methods of decision making in agriculture with available simulation models should
H

be considered in particular. These models can be used together with numerous

PL O

mathematical methods of economic analysis, e.g. linear and dynamic programming,

models of irrigation regime optimization, and models of risk and uncertainty analysis.
M SC

Consider here the application of EV and MGSD analyses. So, in two plans A and B, for
which mathematical expectations are E ( A) , E ( B ) and square deviations are V ( A ) ,
V ( B ) , respectively; A exceeds if
SA NE

E ( A ) = E ( B) and V(A)<V(B)
U

or if

V ( A ) = V ( B ) and E(A)>E(B).

The plan A is said to be more EV- effective than B. The attractiveness of a plan
increases with E and reduces with V. To compare different plans they can be presented
in the EV diagram where E is the ordinate and V is the abscissa. The EV-analysis of risk
assumes that the objective function (income, yield, etc.) follows the normal or at least
symmetric law of probability distribution.

Of two plans A and B the plan A is more preferable, i.e. the plan A is more effective
according to the MGSD analysis, if

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

E ( A ) E ( B) and E ( A ) ( ) ( ) ( ) ,

where ( ) , ( ) are the Gini coefficients of distributions A and B.

The Gini coefficient of a property equals to one-half of the mean Gini difference which
denotes the mean absolute difference for a couple of values of the variable (the number
of differences for the matrix row with n elements is equal to 1 n ( n 1) ). This
2
approach based on Gini coefficients has certain advantages as compared to the EV-
analysis.

The dynamical CERES-Maize model was used by Rithveet for solving the problem of
optimizing the dates of sowing and fertilization for maize in Gainesville, USA.
Meteorological information for 10 years was used in 6 versions of model simulation

S
TE S
runs (for two sowing dates 8 March and 8 April and three rates of fertilizers), 6 x 10 =
60 runs in all. Numerical experiments have shown that the maximum yield and the

R
AP LS
maximum income are expected at the early date of sowing and fertilization of 60 kg ha-1.
This strategy, however, is more risky as compared to that of late sowing with similar
fertilization of 60 kg ha-1 (if for the first version ( ) ( ) = 195, for the second one
C EO
this index is 372). Almost zero probability of losses (negative profit) also shows the
advantages of the second version whereas for the first version this possibility is
estimated as 8% (table 3).
E

-
H

-
PL O

-
M SC

TO ACCESS ALL THE 37 PAGES OF THIS CHAPTER,

Visit: http://www.eolss.net/Eolss-sampleAllChapter.aspx
SA NE

Bibliography
U

Barnard C.S. and Nix J.S. (1973) Farm planning and Control Cambridge: Cambridge University Press.
[Discussed are the models applied to planning at a farm level. Considered is the allocation of different
resources such as agricultural lands, labor, facilities, capital as well as organization of individual forms of
production].
Beneke R. Winterboer R. (1973) Linear programming. Application to agriculture Ames: Iowa State
University Press. [Practical examples of linear programming approach application for different farm
optimization activities and detailed description of the theory of the method]
Dalton G.E. (ed) (1975) Study of Agricultural Systems. London: Applied Science Publishers
[Comprehensive review of mathematical models of agricultural production system].
France J., Thornley J.Y.M. (1984) Mathematical Models in Agriculture. A Quantitative Approach to
Problems in Agriculture and Related Sciences, 335 pp., Butterworths Co (Publishers) LTD. [Different
problems of development and application of mathematical models in agriculture from deterministic
dynamic models of crop productivity, pest development and diseases to simulation of meat and egg
production, farm planning and control].

Encyclopedia of Life Support Systems (EOLSS)

MATHEMATICAL MODELS OF LIFE SUPPORT SYSTEMS Vol. II - Mathematical Models of Agricultural Supply - O.D.
Sirotenko and V.A. Romanenkov

Poluektov R.A. (1991) Dynamic models of agroecosystem (in Russian), 312 pp. Leningrad,
Hydrometeoizdat, Russia. [Designing and application of simulation dynamic models of energy and mass
exchange and productivity of anthropogenic eco-agroecosystems].
Tana Y.A. (1971). Operations Research An Introduction. New York: MacMillan. [Well-known manual
on operations research and mathematical programming].
Thornton P.K., Hoogenboom G. (1994). A computer Program to Analyze Single-Season Crop Model
Outputs. Agronomy Journal. Vol. 86, 5, p. 860-868. [Considered are the computer simulation models
of the growth, development and yield of annual crops and the software to perform simple analyses of such
experiments and to identify those treatments that are productive, stable, economically attractive and
suitable for the purposes of the investigator].
Zhukovsky E.E. (1981). Meteorological Data Information and Economic Decisions (in Russian).
Leningrad: Gidrometeoizdat, 303 pp. [Systematic statement of the bases and methods of application of
agrometeorological information in economic decision-making. General principles are presented for
developing meteorological-ecological models, including agriculture].

Biographical Sketches

S
TE S
Oleg D.Sirotenko is Head of Department of the All Russian Institute of Agricultural Meteorology. His

R
AP LS
main scientific interests include mathematical modeling of energy-mass exchange in the soil-plant-
atmosphere system and development of models describing weather and climate influence on crop
productivity, and assessment of global climatic change and greenhoouse impact on agriculture. He is the
lead author on the assessment by the IPCC (Intergovermental Panel on Climate Change) of potential
C EO
impact on agriculture, the author of the IIASA/UNEP research project on climate change and agriculture.
He is Chairman of the Working Group on Relationship between Climate and Sustainable Agricultural
Production within the WMO Commission for Agricultural Meteorology.

Vladimir A.Romanenkov is Head Researcher of Soil Productivity and Fertilization Systems Laboratory,
E

All-Russian Institute for Fertilizers and Agricultural Soil Science of the Russian Academy of Agricultural
H

Science. His research interests include sustainable agriculture principles for different farm systems,
PL O

mathematical dynamic and static models of soil organic matter and cropping systems, construction and
updating of the international electronic database of field experiments, analysis of spatial variation
information from agricultural landscapes and time trends under different management intensity. He has
M SC

experience in the development of postgraduate programs of site-specific farming in field experimentation.

He has experience also in international scientific cooperation - Participation in several international EU
and the Royal Society projects to promote co-ordination for linking, validation and calibration of existing
dynamic simulation models describing the soil / crop system.
SA NE
U

Encyclopedia of Life Support Systems (EOLSS)