Proceedings of the

4th International Conference on Information Systems, Logistics and Supply Chain

CREATIVE LOGISTICS FOR AN UNCERTAIN WORLD
ILS 2012 – Quebec (Canada), August 26-29

Integrating Optimization and Simulation for

Supply Chain Tactical Planning in the Forest Products Industry

M. Arabi1, J. Gaudreault2, M. Nourelfath3, J. Favreau4, M. Morneau-Pereira5

FORAC Research Consortium, Université Laval
1065, Av. De la Médecine, Québec (QC), G1V 0A6, CANADA
1
[email protected]
2
[email protected]
3
[email protected]
4
[email protected]
5
[email protected]

Abstract: This paper addresses a tactical planning problem in a supply chain network in the forest products industry.
We present a mathematical model that aims at integrating data coming from simulation tools (FPInterface and Optitek)
with an optimization model (LogiOpt). This mathematical model aims at planning at the same time harvesting,
transportation and production operations.

Keywords: Supply chain, Tactical planning, Forest products industry.

1 Introduction according to numerous factors (Rönnqvist, 2003). These

factors include: the forest area where raw material
comes from, harvesting techniques and the choice of the
The job of the supply planner is to produce a tactical mill used to transform wood.
plan that will allow overall efficient operations. The
planner’s goal is to run the logistics network in an In the forest industry, tactical planning is usually done
optimized way, in order to set up the maximum value manually and the manager’s decisions are based mainly
from the wood resources (D'Amours et al., 2008). The on his experience and intuition. Forestry problems are
plan must state: (1) the quantities of timber to harvest characterized by a large number of possible scenarios
and in what location, (2) the selected harvesting and but the planner usually reduces its analysis to a limited
transportation systems, (3) where to transform the wood number of possibilities.
fiber, (4) types and quantities of logs to be processed in
each mill and methods by which logs will be Recently, software simulation tools have been
transformed, and finally (5) the expected quantities of developed in Canada to help the forest industry make
finished products. Since many options are available to better decisions based on more detailed information. For
the planner, this task requires many calculations. example, FPInterface and Optitek simulation softwares,
developed by FPInnovations, allow forest companies to
This problem is particularly complex since the quantity evaluate different scenarios regarding harvesting and
of finished products from the mill’s output varies sawing processes.

* Corresponding Author: [email protected] These simulators allow for the evaluation of costs and
productivity levels. Moreover, they predict the finished
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

products which can be obtained depending on the wood operations synchronization in the network helps
resource, its physical characteristics and technologies generating a greater value from the network and
used in the mill for wood transformation. resources.

Until now, these tools were only used in order to assess

the profitability of a man-made plan. What we propose
in this article is to have the simulator feed a Harvest area A Harvest area B Harvest area C
mathematical model, thus producing an optimized plan.
Log2 Log1 Log3 Log5
This plan would take into account the network capacity, Log1 Log2

product demand and variations in product resources.

A lot of research has focused on the use of optimization

for planning forest operations (e.g. Epstein et al., 1999;
Martell et al., 1998; Karlsson et al., 2003).
Mill A Mill B
The use of mathematical models to plan forest
operations began in the 1960s (Beaudoin et al., 2008).
Since then, various types of models have been
Lumber A Lumber B Lumber B Lumber C Lumber D
developed to address different aspects of the wood
supply chain. Some are designed for specific activities
such as skidding (Carlsson et al., 1998) or transport Figure 1. Simplified example of a forest product supply chain
(Wightman et al., 1990; Weintraub et al., 1996), while
other research has integrated several activities in a
single model to capture the possible synergies between 2.2 Harvest areas
them. For example, (Burger et al., 1995) integrated
harvesting, storage and transport in their approach while
A harvest area is a forest area containing a certain
(Karlsson et al., 2004) also included allocation of
volume of wood to harvest. It is characterized by
harvest teams in theirs.
different distributions in terms of species, stem
diameters, etc. The harvest areas are generally modeled
In comparison with existing research, this article
using a sampling process.
provides two original elements. First, the system models
the whole supply chain in order to integrate forest and
During harvesting, trees are cut and branches are
production related decisions. Additionally, the system
removed. Trees are bucked (or cross-cut) into logs of
connects: (1) a harvesting operations simulator
different dimensions and qualities. Many types of logs
(FPInterface), (2) a sawing process simulator (Optitek),
(or log classes) are obtained simultaneously (Ordre des
and (3) an optimization module establishing the tactical
ingénieurs forestiers du Québec, 2009).
plan.
Harvesting can be done using different harvesting
In recent years, there has been a great amount of interest
systems (or harvesting mode). For a given area, the
in the coupling of simulation/optimization systems in
proportion of each log class obtained depends on forest
research and industrial communities. The paper presents
characteristics as well on the selected harvesting system.
the mathematical model and architecture of the system,
integrating it with the simulation subsystems. Section 2
Figure 2 shows a harvesting area that has been
describes the problem in detail. Section 3 briefly
harvested using a certain harvesting system, the
presents the current planning approach used in the
different resulting logs as well as the tree species ratio
industry as well as available simulation tools. Our
harvested in this area. Using simulation, it is possible to
integrated simulation and optimization solution is then
anticipate the quantity of each log class which would be
presented in Section 4. Sections 5 and 6 present the
obtained using each different harvesting system.
mathematical model as well as the solution validation.

2.3 Transportation
2 Problem
Logs are first stacked along the forest roads and are then
transported to compatible mills (volumes of wood from
2.1 Forest products supply chain a given block can be allocated to many different mills).
In this transportation stage, it is necessary to determine
The lumber production supply chain is a complex the volume to transport, what type of equipment
system made up of harvest areas in forests, harvesting (transportation mode) is to be used, what the best
and transportation systems, and mills. Figure 1 shows a routing decisions, etc. are. There exist different types of
simplified example of a supply chain (a real network trucks to move logs with various lengths. The
may contains hundreds of harvest areas). Better transportation mode will also vary depending on the
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

geographical location of the harvesting areas. over the finished products obtained (Gaudreault et
Transportation can be done with a single transportation al., 2010). Moreover, the quantities associated with each
mode or a combination of different ones. product will be different according to the harvest area
and the harvesting system that were used.

In general, a sawmill is composed of three separate units

Harvest area: 3
Harvesting system: 5 Log1: 20 m3
which are called the sawing unit, the drying unit and the
Initial volume: 100 m 3 planing unit, as shown in Figure 4. Each of these units
Harvesting variable Cost: 10$/ m 3 has its own theoretical maximum capacity. The total
Harvesting Fixed Cost:1000$
capacity of the mill depends on the unit with the lowest
20 % capacity, i.e. the bottleneck (Gaudreault et al., 2009;
Log 2: 20 m 3
Lemieux et al., 2009).
20 %
The capacity can be expressed in terms of
15%
Log 3: 35 m 3 transformation capacity (logs total volume), production
capacity (lumber total volume), or machine availability
10 % (time unit). Machine availability allows taking into
Log 4: 20 m 3 account the fact that we do not obtain the same level of
productivity depending on the type of log that is
processed. This information can be obtained by
simulation (see Section 3.2).
log yard

Figure 2. Harvesting of an area using a given harvesting

2.5 Tactical planning
system (harvesting mode)
Tactical planning consists in defining the proposed use
of the material, resources and network over a planning
2.4 Sawmilling horizon ranging from one to two years. The planning
horizon is usually divided into periods of one month
The sawmilling unit transforms logs into finished (Carlsson et al., 2008).
products called lumbers. From a given log, several types
of lumbers are produced at the same time (it is said to be The plan specifies what harvest areas will be exploited,
a divergent process with coproduction). Log production how they will be harvested (harvesting system) and the
is an automated process in which each log is quantity of each log that will be obtained.
scanned and then cut to generate the highest possible
value. Figure 3 shows an example of a mill production The plan also specifies which type of logs will be
matrix (the ratio of each expected finished product from transported to each mill, and how mills will be
a log class is given). configured (transformation mode). Moreover, an
aggregated production plan for the mills will be
Log 2
generated (expected consumption and production, per
Log volumes :10 m 3 period).
Harvest area : A
Harvesting system: 3

Mill1 Sawing Drying Planing

Sawmilling Modes: 2
Fixed Cost :1000$ (1) (2) (3)
Transportation variable Cost: 5$/unit

Expected Quantities (Logs)

(production matrix) 20 %
25 % (Finished product)
15%
Figure 4. Sawmilling complex

Taking into account the features of the problem

10 Lumber
5 Lumber features presented in the previous section (e.g., the
Sale : 22$/unit
Sale: 25$/unit 8 Lumber quantities of each product produced by the sawmill
Sale : 20$/unit
depending on the harvested area, harvesting system,
Figure 3. An example of mill finished products sawmilling mode and so on), producing a plan is not an
easy task. It is however necessary in order to: (1)
maximize the value generated by the supply chain, and
In most mills, production lines can be configured using if necessary (2) make a diagnosis in order to reorganize
different modes. For each mode, there is a different the network (for example, adding capacity to a certain
associated production matrix. This gives some control mill).
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

Optitek and FPInterface use 3D representations of the

stems and logs (Zhang et al., 2005; Goulet, 2006).
When Optitek is connected to FPInterface, the entire
3 Current situation supply chain can be simulated. FPInterface can compute
replenishment costs and Optitek can compute revenues.
In industrial practice, planning is still largely done
based on intuition and without mathematical
programming support. In other words, the elaboration of
the tactical plan is generally made manually with a trial
and error approach. Different tools and software
technologies have been developed to help wood supply
chain planning at the tactical and operational levels.

To help managers make better decisions, FPInnovations

have developed two softwares that assist the planning
and control of activities related to the exploitation of the
wood resource: FPInterface and Optitek.

3.1 FPInterface

FPInterface is a platform performing simulations of

forest operations (FPInnovations, 2008). It models the
forest supply chain, harvest areas and wood inventories, Figure 6. Optitek simulator (FPInnovations)
as well as harvesting and transportation systems.

FPInterface computes costs associated to supplying a 3.3 Limitations of the current system
mill from a given harvest area and identifies which type
of log would be obtained using different harvesting
system. The combination of both applications allows
anticipating the economic value generated by a
Users can visualize data through a map, which makes harvesting area allocated to a mill. It therefore helps the
interpreting results easier (Figure 5). manager to establish a reasonable plan.

However, this process is done by a trial and error

approach. The user must specify a plan manually and it
can only evaluate a single plan per simulation.
Reconfiguration of the system can be quite long and
trying all possible solutions is practically impossible.
Consequently, since the selection of a plan is based on
the experience and intuition of the user, he can neglect
certain solutions, because they do not seem to offer
enough potential, while in reality it would have been a
good choice.

There is also no indication about the gap in revenue that

separates the plan from what the optimal solution would
be. In addition it is very difficult to assess the impact of
Figure 5. FPInterface software a change (real or potential) which could occur in the
network (e.g., shut down of a mill).

3.2 Optitek
4 Proposed optimization-simulation
Optitek (Figure 6) is a simulator for sawmill operations. integration
Each machine in the production line can be modeled
through different modules. When a log description is We propose that an optimization module, called
provided, Optitek can forecast the lumber production LogiOpt, be added to the integrated FPInterface-Optitek
which would be obtained from the mill. system. Figure 7 shows a schematic view of the
interconnections between the simulation tools and
optimization module as well as the data flow between
them.
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

gb,r Harvesting variable costs that is generated by a

All the possible “elementary operations” that can be unit of harvested volume (m3) of log in a
included into a plan are simulated individually: For each harvest area b using harvesting system r.
valid combination of harvest area and harvesting
system, the system generates 3D profiles of logs that ρb,r,f,u Transportation cost of a unit volume (m3) of a
would be obtained. Optitek then performs a sawing product between harvest area f and mill u.
simulation for each mill of the network for each of those
alternative forest products. b,r, f,p,n,u The products quantity p produced by mill u
which is in sawmilling mode n and consumes
These results are then fed into the optimization module, product f harvested from block b by harvesting
which generates an optimal tactical plan (the model system r.
takes into account mills capacity and the fact that the
source of the raw material has an influence on the  u,t The maximum volume consumed by the mill u
lumber types and quantities). Finally, the plan is (representing a production capacity) at period t.
returned to FPInterface in order to be displayed to the
user.  u ,t The maximum capacity [in Foot Board Meas-
ure (FBM)] consumed by the mill u in period t.
This capacity gives the maximum number of
5 Mathematical model units of (FBM) volume that can be generated
by the mill.
In this section, the LogiOpt mathematical model is cu Sawmilling fixed cost of the mill u.
presented. Data sets are introduced first, followed by
parameters and variables used to express the model. wu Sawmilling variable costs of operation of the
Finally the model formulation is described. It should be
mill u (per m3).
noted that to develop this model, the specific
requirements of FPInnovations tools were taken into Φp Incomes for a unit of finished product p.
account. For example, for each finished product from a
specific harvest area, harvesting methods and the ib,r, f,0 The quantity of logs f, harvested using system r
transformation mill should be known.
which is available along the road of harvest
area b at the beginning of the planning horizon.
Sets
B Harvest areas (b ϵ B) jb,r, f,u,0 The quantity of harvested logs f generated by
harvesting system r which is in stock at the
R Harvesting systems (r ϵ R)
mill u at the beginning of the planning horizon.
F Log types (f ϵ F)
P Product (lumber) (p ϵ P) Decision variables
U Mills (u ϵ U) Vb,r,t The total volume (m3) harvested from area b
N Sawmilling modes n:{1,…,N} using harvesting system r in period t.
T Periods t=1,…,T H b,r,f,t The volume of the logs f (m3) obtained from
harvest area b using system r in period t.
Parameters
X b,r,f,n,u,t The volume of logs f (m3) consumed by the
sb The total volume (m3) of wood (initial volume) mill u (configured in mode n) during period t.
which can be extracted from harvest area b. This log has been harvested in a harvest area b
using harvesting system r.
kb,r, f The generated volume (m3) of log f by using
harvesting systems r in harvest area b. Qb,r,f,u,t The volume of log f harvested using system r
which is transported between harvest area b
ξ b,r The value of indirect sales (in $) of harvesting and mill u in period t.
area b using harvesting system r. Indirect sales
correspond to products that are sold to third Zp,n,u,t The quantity of lumber p manufactured by mill
parties which do not need to be transported to a u configured in mode n during period t.
sawmill in our supply chain. I b,r,f,t The quantity of logs f harvested using system r
b Harvesting Fixed Cost associated with harvest in stock in area b at the end of period t.
area b in the planning horizon. J b,r,f,u,t The quantity of logs f harvested using system r
in stock at mill u at the end of period t.
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

Ym,t Constraints
The binary variable takes the value 1 when the
mill u is configured as sawmilling modes Maximum allowable harvest:
m :{ 1… M} at the period t, 0 otherwise.
   V b,r ,t  s b b  B (1)
bB rR t{1,..,T }

Objective function
Harvested volumes:
The objective function is to maximize profit by taking
into account sales, transportation costs, harvest costs H b,r,f ,t  k b,r,f Vb,r,t b  B, r  R (2)
and mills costs. The harvesting fixed costs and
f  F , t  1,..., T
sawmilling fixed costs have been taken into account
only in order to calculate the net value of the supply
chain.
Sawmill capacity:

MAX [  s b  
    X b,r , f ,n,u ,t   u,t u U , t  1,..., T (3)
 V b,r ,t ]    b bB rR f F nN
bB rR t{1,...,T } bB

[     Z p,n,u ,t   p]   Z p,n,u,t  u,t u U , t  1,..., T (4)

pP nN uU t{1,...,T } pP nN

 
[    V b,r ,t   g b,r     b,r ]
bB rR t{1,...,T }  bB rR 
Mill operations:
[      Q b,r , f ,u ,t   b,r , f ,u ]
Z p , m , u , t     ( X b , r , f , m ,u ,t   b , r , f , p , m , u )
bB rR f F uU t{1,...,T }
p  P (5)
[    X b,r , f ,n,u ,t  wu ]   cu bB rR f F
bB rR f F nN uU t{1,...,T } uU m  1,..., M , u U , t  1,...,T

FPInterface U,N, фp,u

Optitek
B, R, U, N, T,Ψb , ρb,r, gb,r Harvesting
User
Simulation F Sawmilling
αu,t, βu,t, cu , w, фp,u Interface
sb , ib,r,f,0 , jb,r,f,u,0u Simulation

N
Vb,r,t T μb,r,f,p,n,u
Sb kb,r,f P
Hb,r,f,t F Cu
Ψb ξr,b U
Xb,r,f,n,u,t wu фp,u
Qb,r,f,u,t ρb,r
Zp,n,u,t gb,r
Y n,t αu,t
I b,r,f,t βu,t
Jb,r,f,u,t Cu
Wu
ib,r,f,0
jb,r,f,u,0

LogiOpt
Optimization

Figure 7. Integration of simulators and optimization module

 ILS’2012 - August 26-29, 2012 - Quebec - Canada

Constraints (3) and (4) set the capacity of the mill in

 Y n,t  1 u U , t  1,..., T (6) terms of consumption and production.
nN

Constraint (5) sets the relationship between the con-

X b,r , f ,n,u,t  Y n,t  M b  B, r  R (7)
sumed volume at the mill and the produced volume.

f  F , n=1,,N, u U , t  1,...,T Constraints (6) and (7) ensure that a mill can be config-
ured according to only one sawmilling mode.
Where M is a large positive number.
Constraints (8) and (9) calculate the inventory of prod-
Equilibrium flow (harvest area): ucts for each harvest area and constraints (10) and (11)
calculate inventory at the mill.

I b,r , f ,1  i b,r , f ,0  H b,r , f ,1   Qb,r , f ,u,1 (8)

6 Validation
uU
b  B, r  R, f  F
In this section, the verification and validation of
I b,r , f ,t  I b,r , f ,t 1  H b,r , f ,t   Qb,r , f ,u ,t (9) mathematical model are presented. The model was
uU
implemented in ILOG OPL Studio version 6.3 and was
b  B, r  R, f  F , t  2,..,T solved in CPLEX 12.1 using a 3 GHz processor with 2
GB RAM.
Equilibrium flow (mill):

J b,r , f ,u,1  j b,r , f ,u,0  Qb,r , f ,u,1   X b,r , f ,n,u,1 (10) 6.1 Validation of the optimization model
nN with a simplified case.

b  B, r  R, f  F , u U To validate the mathematical model, we first developed

a small case study (see Figure 8). This case study
J b,r , f ,u,t  J b,r , f ,u,t 1  Qb,r , f ,u,t   X b,r , f ,n,u,t (11) allowed us to assess the validity of the model with our
nN research partners.
b  B, r  R, f  F , u U , t  2,..,T
The case study’s network consists of two harvest areas
(b1, b2) and two mills (u1, u2). The initial volumes of
Non-negativity constraint harvest areas are s1 and s2 respectively. The harvest
areas have a single harvesting mode (r1 and r2,
Vb,r,t  0 b  B, r  R, t  1,..,T (12) respectively). No log of type f2 comes from b2. The mill
u2 does not produce lumber of type p2 and does not
Hb,r,f,t  0 b  B, r  R, f  F , t  1,..,T (13) accept log (f2).

The optimization’s results were compared with expected

results. In addition, the model was also expanded step
X b,r,f,n,u,t  0 b  B, r  R, f  F , n  N , (14) by step for different total numbers of elements of the
u U , t  1,.., T model (harvest areas and logs and finished product) to
ensure the accuracy of the results.

Qb,r,f,u,t  0 b  B, r  R, f  F , u U (15) Afterwards, the model was evaluated by FPInterface’s

t  1,.., T experts to assess how our model met their requirements
(this test case consisted of 264 harvest areas, 9 logs, 3
mills and 18 finished products).
Z p,n,u,t  0 p  P, n  N , u U , t  1,..,T (16)
It is noteworthy that to better evaluate and compare the
optimization’s results, the decisions variables are
Constraint (1) sets the harvested volume to observe the presented graphically with reports. A results example is
maximum allowable volume. presented in Figure 9. It shows the volumes of wood
that are transported to a given mill, for each log type.
Constraint (2) establishes a link between the harvested
volumes and the quantity of each log type which will be
available for transportation to the mill.
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

s1 : 100 m 3 s2 : 50 m 3
b1 ψ 1 : 1000$ ψ2 : 1000$ b2
g1 : 10$ g2 : 10$
ξ1,1 :5$ ξ 2,1 :5$

K1,1,1 :50 %
K2,1,1 :100 %
K1,1,2 : 45 %

f1 f1
f2

ρ 2,1,1,2 :50$/m 3
ρ 1,1,2,1 :50$/m 3

N: 1 N: 1
u1 c:1500$ c:1500$ u2
w:10$/m 3 w:10$/m 3
α:100 m 3 α:50 m 3

μ1,1,1,2,1,1,1:10 unit/m3
μ1,1,1,1,1,1:12 unit/m3 μ1,1,1,2,1,1,1:15 unit/m3
μ1,1,2,2,1,1,1:5 unit/m3
μ1,1,2,1,1,1,1:5 unit/m3 μ2,1,1,1,1,1,1:10 unit/m3
μ2,1,1,2,1,1:10 unit/m3
μ2,1,1,1,1,1:10 unit/m3

p1
p2 p2
Ф1,2 : 10$/unit
Ф2,1 : 15$/unit
Ф1,1 : 10$/unit
Figure 8. Simplified example of the forest product network

6.2 Validation of the optimization-

simulation system with an industrial
case
30371 30399
30000
We initiated another validation phase (still ongoing) 25000
Transported Volume (m3)

consisting in producing a tactical plan based on data

provided by a large Canadian forest products company. 20000

This case involves hundreds of harvest areas, three mills 15000

(two mills have the same distance from several harvest 10000 8133 7745 7815
areas).
5000 3616
1389 1654
583
Dozens of log types and lumber products are involved. 0
f1 f2 f3 f4 f5 f6 f7 f8 f9
We are still working on the data validation and we will Type of Logs
soon be able to compare the plan generated by the Figure 9. The transported volume of the logs for one mil.
model with scenarios generated manually by a human.
 ILS’2012 - August 26-29, 2012 - Quebec - Canada

7 Conclusion Epstein, R., Morales, R., Seron, J., Weintraub, A.,

(1999). Use of OR systems in the Chilean forest
This project aims to enable managers in defining inte- industries, Interface, 29(1), 7-29.
grated tactical plans in the forest products industry. This FPInnovations, (2008). Guide de l’utilisateur de
work has the following original features: the proposed FPInterface.
system models the supply chain as a whole and there- Gaudreault, J., Forget, P., Frayret, J.M., Rousseau, A.,
fore, allows an integrated optimization of forest and Lemieux, S., D'Amours, S., (2010). Distributed
industrial operations. operations planning in the lumber supply chain:
Models and coordination. International Journal of
Moreover, the combinatorial nature of the problem is Industrial Engineering: Theory, Applications and
taken into account by our approach (i.e., the final prod- Practice 17(3).
ucts, the harvesting system, the selected mill and its Gaudreault, J., Frayret, J.M., Pesant, G. (2009).
configuration). All of this is made possible because of Distributed Search for Supply Chain Coordination.
our integration with a harvest simulator (FPInterface) Computers in Industry, 60(6), 441-451.
and a sawmilling simulator (Optitek). Goulet, P., (2006) Optitek: User’s Manual. Document
E-4130. Forintek Canada Corp. Quebec City, QC.
This coupling of simulation and optimization is of great Karlsson, J., Rönnqvist, M., Bergström, J., (2003).
interest from a scientific and an industrial perspective. Short-term harvest planning including scheduling of
In the near future, decision makers will be able to see harvest crews. International Transactions in
the optimal solution directly on a forest map in FPInter- Operational Research 10: 413-431.
face and they will be able to modify the proposed solu- Karlsson, J., Rönnqvist, M., Bergstrôm, J., (2004). An
tion and evaluate the economic impacts associated with optimization model for annual harvest planning.
different options. Canadian Journal of Forest Research 34, 1747-1754
Lemieux, S., D’Amors, S., Gaudreault, J., and Frayret,
For example, it will be possible to evaluate the impact J.M., (2009), Agent-base simulation to anticipate
of modifying a mill capacity more easily. This way, the impacts of tactical supply chain decision-making in
tactical plan can be quickly readjusted and the manager the lumber industry, CIRRELT, 51, working
can see how the rest of the supply chain should adapt in document, Canada.
order to take advantage of these changes, thus providing Martell, D.L., Gunn, E.A., Weintraub, A., (1998).
him with a more versatile approach. Forest management challenges for operational
researchers. European Journal of Operational
Research, 104, 1-17.
Ordre des ing nieurs forestiers du u bec, (2009).
8 References Manuel de foresterie 2e Éditions Multimonde., 1249-
1290.
Beaudoin, D., LeBel, L., Jean-Marc Frayret, J.M., Rönnqvist, M., (2003). Optimization in forestry.
(2006) Tactical supply chain planning in the forest Mathematical Programming. B(97), 267-284.
products industry through optimization and scenario- Weintraub, A., Epstein, R., Morales, R., Serón, J., and
based analysis. Canadian Journal of Forest Research, Traverso, P., (1996). A Truck Scheduling System
2006, 37(1) 128-140. Improves Efficiency in the Forestry Industry.
Burger, D. H., and Jamnick, M. S., 1995. Using linear Interfaces 26(4) , 1-12.
programming to make wood procurement and Wightman, R., and Jordan, G. 1990. Harvest distributon
distribution decisions. The Forestry Chronicle. planning in New Brunswick. CPPA Woodlands Pap.
71(1), 89-96. August: 19-22.
Carlsson, D., D’Amours, S., Martel, A., Rönnqvist, M. Zhang, S.Y., Tong, Q.J., (2005). Modeling lumber
(2008). Decisions and methodology for planning the recovery in relation to selected tree characteristics in
wood fibre flow in the forest supply chain. Recent jack pine using sawing simulator Optitek. Annals of
developments in supply chain management. Koster, Forest Science. 62(3), 219-228.
R., Delfmann, W., (Eds.) Helsinki School of
Economics: 11-39.
Carlsson, D., Rönnqvist, M., Westerlund, M., (1998).
Extraction of logs in forestry using operations
research techniques, Report LiTH-MAT-R-98-17,
Linköping University, Sweden.
D'Amours, S., Rönnqvist, M., Weintraub, A., (2008)
Using Operational Research for Supply Chain
Planning in the Forest Products Industry,
Information Systems and Operational Research, 46
(4), 265-281.