Applied Energy 111 (2013) 1239–1249

Contents lists available at SciVerse ScienceDirect

Applied Energy

journal homepage: www.elsevier .com/locate /apenergy

A simulation model for the design and analysis of wood pellet supplychains

0306-2619/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.apenergy.2013.06.026

⇑ Corresponding author.E-mail address: taraneh.sowlati@ubc.ca (T. Sowlati).

Mahdi Mobini a, Taraneh Sowlati b,⇑, Shahab Sokhansanj c,d

a Industrial Engineering Research Group, Department of Wood Science, University of British Columbia, 2943-2424 Main Mall, Vancouver, BC V6T-1Z4, Canadab Industrial Engineering Research Group, Department of Wood Science, University of British Columbia, 2931-2424 Main Mall, Vancouver, BC V6T-1Z4, Canadac Department of Chemical and Biological Engineering, University of British Columbia, 2360 East Mall, Vancouver, BC V6T 1Z3, Canadad Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA

h i g h l i g h t s

� Developed a simulation model to evaluate the supply chain of a pellet mill.� Incorporated uncertainties into the model.� Applied the model to a real case study in Canada.� Used the case study to verify and validate the model.� Analyzed different scenarios for the pellet mill.

a r t i c l e i n f o

Article history:Received 3 January 2013Received in revised form 12 June 2013Accepted 14 June 2013Available online 23 July 2013

Keywords:SimulationWood pelletsSupply chainBioenergyRenewable energy

a b s t r a c t

During the past decade, the global trade of wood pellets has been growing. Rapid increases in the produc-tion and consumption of wood pellets, and predictions on its increased demand in the near future haveformed a competitive global market. Several studies have focused on the economic, environmental, andtechnological aspects of wood pellet production and consumption. In this paper, a simulation model isdeveloped to enhance and facilitate the studies concerning the design and analysis of wood pellet supplychains. The scope of the model covers the entire supply chain from sources of raw materials to the endcustomers, providing a framework for assessment of the supply chains. The model includes uncertainties,interdependencies between stages of the supply chain, and resource constraints, which are usually sim-plified or ignored in previous studies. The outputs of the model include the amount of energy consumedin each process and its related CO2 emissions, and the cost components of delivered wood pellets to thecustomers. The model was applied to an existing supply chain located in BC, Canada. The estimated costof wood pellets was 69.27 $ t�1 at the pellet mill’s gate and 101.33 $ t�1 at customers’ locations. Distri-bution of wood pellets to the customers contributed about 30.65% to total costs. Raw material procure-ment and transportation accounted for 29.16% of the total delivered cost, while pellet productioncontributes 40.19% to the total delivered cost. The energy consumption and CO2 emission along the sup-ply chain were estimated at 568.93 kW h t�1 and 136.91 kg t�1, respectively. The results of scenario-based analysis showed that by changing the drying fuel from sawdust to bark, about 1.5% cost reductionwas achievable. Blending 10% bark in the whitewood feedstock reduced the estimated cost to 96.51 $ t�1

(4.75% reduction).� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Wood pellets are densified form of biomass usually made fromby-products of primary wood processing facilities. The moisturecontent of wood pellets is less than 10% and its bulk density isaround 650 kg m�3. Wood pellets have uniform cylindrical shape

and are recognized as a standardized international commodity.The required infrastructure for handling, storage, and transporta-tion of wood pellets are similar to other common commoditiessuch as coal, wood chips, and grain. These properties make it easierto store, handle, and utilize wood pellets compared to other formsof biomass [1].

During the past decade, trade of wood pellets has had a majorshare in the global bioenergy market. Rapid increases in the pro-duction and consumption of wood pellets, and predictions on itsincreased demand in the future have formed a competitive global

1240 M. Mobini et al. / Applied Energy 111 (2013) 1239–1249

market. Total global consumption of wood pellets in 2011 was14.4 million tonnes, from which about 80% were consumed in Eur-ope [2]. High fossil fuel prices, fossil fuel taxes, and incentives forrenewable fuels in the European countries have led to a steadygrowth in the wood pellet market since 1990s. The Europeandemand is expected to grow and could potentially reach to20–50 million tonne per year by 2020 [3]. Also, North Americanand Asian markets are expected to substantially grow as a resultof legislative supports and greenhouse gas (GHG) reduction poli-cies [3,4].

In 2010, the total wood pellet production in Europe was9.2 million tonnes, which was equal to 61% of the global produc-tion. Canada is also a major producer with annual productioncapacity of 3.3 million tonne [3]. About 90% of pellets producedin Canada are exported to Europe where they are used for powerand heat generation [5]. The United States has been experiencingrapid expansions of wood pellet production in the southeast regionto take advantage of the European market [6]. Furthermore, newpotentials for production and consumption of wood pellets aredemonstrated in other biomass-rich regions, such as Russia, Brazil,Australia and New Zealand [3].

Different types of raw material have been tested and suggestedfor pellet production including wood residues (such as sawdust,shavings, wood chips, and bark), forest harvesting residues, andagricultural biomass (such as alfalfa and grain.) [8,12–14]. Produc-tion of wood pellets in different geographical regions, using differ-ent types or mixtures of raw material and employing differenttechnologies in the production, has been the focus of previousstudies [7–10]. Mani et al. [10] estimated 51 $ t�1 as the cost of pel-let production when wet sawdust (40% moisture content) was usedas raw material and dried shavings was used as the drying fuel.Additionally, economics of wood pellet production in BritishColumbia, Canada were studied by Peng et al. [11]. The productioncost was estimated when mill residues, harvesting residues, bark,or mountain pine beetle infested wood was used as the raw mate-rial. It was shown that the final cost of pellets was significantlyinfluenced by the type and cost of raw material and the size ofthe plant.

A multi-criteria decision making approach was used by Sultanaand Kumar [15] to choose the type of raw material for pellet pro-duction based on economic, environmental, and technical factors(eleven criteria were defined). It was shown that pellets made fromwood biomass were preferred based on the defined criteria. Eco-nomics of producing pellets from agricultural biomass in WesternCanada has been studied by Sultana et al. [14]. Three scenariosbased on the average yield of farms were developed and it wasshown that the cost of energy from agricultural biomass was notcompetitive with that of natural gas. Many other studies including[4,16–19] were concerned with utilization of wood pellets as a so-lid biofuel competing with other sources of energy.

Previous studies concerned with estimating the cost of woodpellet production, including those mentioned above, mostly usedstatic deterministic modeling approaches that ignore the dynam-ics, interdependencies, and uncertainties along the supply chain;therefore, the applicability of their results in decision making islimited. For example, neglecting the variations in moisture contentmight lead to incorrect estimations of available feedstock, andtransportation and processing costs. The effects of a machine fail-ure on the subsequent process cannot be included explicitly in astatic model. Moreover, the design of a wood pellet supply chainis largely influenced by the geographical-dependent parameters,such as available types, cost, and quality of raw materials, marketspecifications, and access to the infrastructures. This limits thecredibility of obtained results in each study to the specific case.

In this study, a discrete event simulation model, called PelletSupply Chain (PSC) model, is developed to facilitate the design

and analysis of the wood pellet supply chains, by overcoming theabovementioned limitations. Using the simulation model, it wouldbe possible to incorporate uncertainties in the analysis. Also, theconnections and interactions between different stages of the sup-ply chain are considered as the model provides an integrated per-spective of the chain. The model could be applied to any given area,provided that the input data with proper quality are available. Thesimulation model is intended as a customizable framework fordesigning the wood pellet supply chains. Decisions concerningthe strategic design of the supply chain, such as choosing the bestlocation of the pellet plant amongst a set of potential locations,could be supported by the model. The cost of delivered pellets tothe customers could be evaluated under different configurationsof the supply chain and the sensitivity analysis could be performedto provide a comprehensive viewpoint for the decision makers.Furthermore, the simulation model could be used to evaluate dif-ferent alterations in the existing supply chains, e.g., changing thetypes of raw material or drying fuel, storage policies, number ofloading/unloading stations, and number and size of the buffers inthe processing stage.

The developed simulation model is applied to a case study inBC, Canada. It is used to estimate the cost of delivered pellets todifferent customers and to estimate the energy inputs and CO2

emissions along the supply chain. The model is used to assesstwo alterations in the supply chain: (1) use a different fuel typeto generate the required heat for the drying process within the pel-let mill, and (2) use a different mixture of raw materials for pelletproduction.

2. Wood pellet supply chain

Fig. 1shows a typical wood pellet supply chain. It starts fromsources of raw materials. Availability, quality, and cost of rawmaterials have decisive effects on the business viability and designof the supply chain [20]. Shavings and sawdust are the most de-sired raw material due to their small particle size, low ash content,and low moisture content. Currently sawdust, a by-product of thesawmill industry, is the main raw material in the pellet production[8].

Transportation of raw material to the pellet plant is the nextstage in the supply chain. Due to low bulk density and high mois-ture content of raw materials, transportation can significantly con-tribute to the final cost of wood pellets. In some instances, thepellet plants are located at sawmills gate and pipelines are usedto convey the raw materials to the pellet plant, hence raw materialtransportation is eliminated. Covered and outdoor storage of rawmaterials are both used in the pellet production supply chains.Wet material could be stored outside, but drier materials such asshavings and dried sawdust are stored in covered storage areasor silos. Depending on the type of raw material, a series of pro-cesses are required for densification of biomass. Drying, size reduc-tion, pelletizing, cooling, screening, and packaging processes aretypically seen in pellet mills. Prior to feeding the raw material intopelletizers, its moisture content and particle size should fall withinappropriate ranges. As a general rule, the moisture content of rawmaterials should be in the range of 12–17% wet basis [13]. The ro-tary dryer is the most commonly used technology in the drying ofwood biomass since it can be used to dry high volumes of materialswith wide range of initial moisture content [21–23]. The two mostimportant drying process characteristics are the feed moisture con-tent and the feed flow rate [22]. In the real systems, the feed mois-ture content cannot be controlled since it depends on the suppliersand materials characteristics. It is usually the practice to keep theflow rate of materials as constant as possible and adjust otherparameters, such as temperature of the heating media, in order

Fig. 1. Schematic of a wood pellet supply chain.

M. Mobini et al. / Applied Energy 111 (2013) 1239–1249 1241

to obtain uniform moisture content in the discharged materials.Raw materials are hammer milled using a 4 mm screen for produc-ing pellets with 6 mm diameter [8]. Pellet mills (press) operate at atemperature of 90–100 �C [24]. Pellets are hot and soft upon exitfrom the pellet mill. Therefore cooling is needed for hardening it.Wood pellets are usually stored in silos or warehouses or directlybagged for distribution.

According to Pellet Fuel Institute (PFI) standard, three types ofpellets: premium, standard, and utility pellets are defined [25]. Inthe premium category, the maximum ash content is 1%. The max-imum allowable ash content for the standard and utility categoriesare 2% and 6%, respectively. Since the ash content of raw materialsdoes not change during the production, the production of differenttypes of wood pellets depends on the mixture of raw material.

Wood pellets are distributed in many forms including con-sumer-bags (15–18 kg), big-bags (500 kg), containers, railcars,and ocean vessels [8]. Also, residential market deliveries are doneusing tank trucks, especially in Europe where proper storage facil-ities are available in customers’ locations [8].

3. The simulation model

The scope of the model includes the whole supply chain, whichstarts from sources of raw materials (suppliers) and ends with thecustomers. Procurement, transportation and storage of raw mate-rial, pellet production, and distribution are the major activitiesalong the supply chain that are included in the model. The esti-mates of time, cost, emission, and energy consumption associatedwith each of these activities are provided while taking into accountthe dynamics of the system. The model is developed using theExtendSim simulation software [26], which is an object orientedsimulation environment. The supply chain entities are developedas modules and can contain several sub-modules. Depending onthe module and its role, properties and interactions with othermodules are defined. Discrete-event and discrete-rate modelingapproaches are employed in the development of the model. Theinteractions between different parts of the supply chain are mod-eled as discrete-event. The flow of materials and interdependencybetween the processes inside the pellet mill are modeled using thediscrete-rate approach [27]. The model verification is done throughwalkthroughs, debugging of the codes, and rigorous test runs. Inorder to validate the model, it is applied to a real case and the re-sults are compared with those of the real system.

3.1. Supply chain entities and modules

Raw material suppliers, pellet mills, customers, and transporta-tion vehicles are considered as supply chain entities. A set of prop-erties and interactions is defined for each entity according to theirrole in the supply chain. Fig. 2 depicts the structure of the simula-tion model including the flow of information and interactionsbetween the different modules. Table 1 lists the most important in-put parameters and outputs associated with each module in thesimulation model. The simulation model uses a relational database

to communicate data between the modules. Activities and pro-cesses along the chain include the physical processes and activitiesin addition to scheduling events leading to the occurrence of theactivities.

The road transportation module is composed of three mainactivities (assignment of transportation orders to the vehicles,composition of routes for vehicles for which dispatch requirementsare met, and transportation cost calculations) and a schedulingevent. The scheduling event in the transportation module startswhenever there is a new transportation order generated by pelletmill module either for raw material transportation or for pellet dis-tribution. Another occurrence that could trigger a scheduling eventis when a vehicle has been released from previously assigned tripand has become available. Also, periodic (hourly) schedulingevents are considered. If a certain conditions apply (number ofavailable vehicles is more than zero, there are available vehiclescapable of handling the order, and dispatch requirements aremet) the route is composed and vehicle is dispatched.

3.1.1. SuppliersA pellet mill may receive its raw materials from different types

of suppliers. For each supplier, types of raw material, their quality,and quantity over time are inputs of the model. The location, pro-duction and capacity of suppliers, their operating shifts, transpor-tation mode, storage capacity, and loading rate are other inputfactors of the model. To keep track of the characteristics of thematerial moving through the system, each entity is associated witha material characteristic record. The data on weight, volume,moisture content, ash content, bulk density, heat value, and parti-cle size are data fields maintained in this record. The main outputsof the suppliers’ module are the amount and cost of purchased rawmaterials.

3.1.2. Pellet millThe raw material storage, drying, size reduction, pelletizing,

cooling, and wood pellet storage processes are included in the pel-let mill module. The plant capacity depends on the installed num-ber and capacity of the equipment which are inputs of thesimulation model. Furthermore, the model takes the sequence ofoperations as an input to the model that could be different for dif-ferent types of materials.

The costs related to the pellet plant are composed of several ele-ments: capital cost, operating (e.g., costs of energy, fuel, lubricants,consumables, and labor) and maintenance costs. The capital cost isannualized using the capital recovery factor formula (Eq. (1)),where, ACC is the annual capital cost, TCC is the total capital cost,i is the annual interest rate, and N is the service life of the plant.To estimate the capital cost of a piece of equipment, the powerlaw suggested by Perry et al. [28] is used. Eq. (2) shows the rela-tionship used to calculate the capital cost of a piece of equipmentwith capacity2 relative to a given cost of a piece of equipment withcapacity1.

ACC ¼ TCC � ið1þ iÞN

ð1þ iÞN � 1ð1Þ

Fig. 2. Structure of the simulation model. The model includes transportation, supplier, pellet mill, and customer modules. The flow of information and interactions betweenthe modules are shown in the figure.

Table 1Inputs and outputs of the simulation model.

Module Input parameters Outputs

Suppliers � Biomass production rate� Quality measures of raw materials� Cost of raw materials� Location� Transportation mode

� Total wet and dried weight of purchased raw materials� Cost of purchased raw materials

Pellet plant � Capacity (throughput)� Annual operating days� Location� Sequence of operations inside the plant� Equipment failure rates and repair time� Mixture of raw material� Fuel used for heat generation� Electricity cost� Investment cost

� Raw material consumption� Processing costs� Plant productivity� Total weight of each type of products� Energy consumption (electricity and heat)

Customers � Demand profile� Location� Transportation mode

� Demanded weight of products� Delivered weight and average cost of delivered products� Demand fulfillment rate

Transportationcompany

� Unit cost of transportation per kilometer($ km�1)� Unit cost of transportation per hour ($ h�1)� Number and type of trucks� Capacity of trucks� Fuel consumption rate (lit km�1)

� Total transportation distance/time� Total transportation costs of raw materials (from each supplier) and pellets (to each

customer)� Total transportation costs of pellets� Weight and transportation cost of delivered raw materials from each supplier� Weight and transportation cost of delivered pellets to each customer� Emissions� Energy/ fuel consumption (cost)

1242 M. Mobini et al. / Applied Energy 111 (2013) 1239–1249

Cost2 ¼ Cost1capacity2

capacity1

� �scale factor

ð2Þ

Operating costs are calculated separately for each process.Depending on the case, the labor and overhead costs are estimatedand used in the calculations.

3.1.2.1. Raw material storage. The storage of raw materials is mod-eled by incorporating the storage capacity, number of unloading

stations, unloading rates, and tracking in/out flows. The storagecapacity for each type of raw material is an input parameter ofthe model. Queues might form in front of the unloading stationand the in-line time of the vehicles are estimated in this case.

3.1.2.2. Drying process. The drying process diagram, Fig. 3, de-scribes how the drying process is modeled. The input rate of mate-rials is kept constant as long as the raw materials are available, and

Fig. 3. The drying process flowchart.

M. Mobini et al. / Applied Energy 111 (2013) 1239–1249 1243

the variations in the feed moisture content are dealt with by con-trolling the weight of the fuel materials fed to the burner.

Electricity and heat are used in the drying process. Electricity isconsumed by the belt drive, fans and other electrical componentsof the dryer. In the simulation model, the dryer’s total power(kW) is used to calculate the electricity consumption over timeinterval of Dt as shown in Eq. (3); where E is the price of electricity($ kW h�1), P is the equipment power (kW), SF is the simultaneityfactor of the equipment (%), and W is the processed weight duringthe time interval. The simultaneity factor is an empirical value thatadjusts the power consumption of the equipment according to thenominal power and the operating conditions [8,9].

Electricity cost$

t

� �¼ E� P � SF � Dt

Wð3Þ

Pellet mills in BC usually burn biomass to generate the requiredheat. The required amount of heat depends on the amount of waterthat should be evaporated which in turn depends on the initial andtarget moisture contents of the material. Eq. (4) is used in the sim-ulation model to estimate the amount of water that should beevaporated in order to reach the target moisture content. In thisequation, Weva. is the weight of water that should be evaporated(t), Win is the wet weight of material fed into the dryer (t), dWin

is the dried weight of in feed materials (t), and MCt is the targetmoisture content after the dryer (%). The heat demand of the dryer(HD) is defined as the amount of heat required to evaporate onetonne of water, which is introduced as a specification of the dryersystem and is provided by the manufacturers [8]. Eq. (5) is used tocalculate the required heat (RHeat) to obtain the target moisturecontent.

Weva: ¼Win �dWin

1�MCtð4Þ

RHeat ¼Weva: � HD ð5Þ

The cost of consumed heat during a given time interval is calcu-lated by multiplying the heat requirement by the unit cost of theenergy ($ kW h�1). When biomass is used as the drying fuel, theweight of the required biomass fuel during Dt is calculated basedon the heat demand of the dryer, heat value and moisture contentof the fuel, and the efficiency of the burner as expressed in Eq. (6).Herein, WFuel is the required weight of fuel (t), HVFuel is the heat va-lue of the fuel (kW h t�1), MCFuel is the moisture content of the fuel,and c is the efficiency of the burner (%).

WFuel ¼RHeat

HVFuel � ð1�MCFuelÞ � cð6Þ

3.1.2.3. Size reduction (grinding), pelletizing, and cooling. The sizereduction process is usually performed by hammer mills whensawdust and shavings are used in pellet production. Pelletizationof the ground materials is performed in pelletizers (pellet mills).Counter flow coolers are commonly used in pellet productionindustry. Eq. (3) is used in the simulation model to estimate the en-ergy consumption and cost in the processes.

3.1.3. CustomersThe input data of the model include the demand profile of the

customer, its location, and available transportation mode. It is pos-sible to include demand uncertainties and fluctuations into thesimulation model. Delivered cost of wood pellets is estimated foreach customer in addition to the related energy input and CO2

emissions.

3.1.4. TransportationThe location and available transportation mode for each

supplier and customer are inputs of the model. Travel distancesare calculated using MS Mappoint [29]. Travel times are calculatedbased on travel distances and vehicle’s speed. The fuel consump-tion of vehicles is calculated based on the consumption rate(lit km�1). The energy consumption and generated emissions dueto road transportation are calculated based on the fuel consump-tion. Scheduling of the road transportation is performed consider-ing the vehicles number and availability (idle, busy, off-shift),capacity (volume and weight), hours of operations, and restrictions(e.g. type of materials that can be transported), and also type ofraw material (different types cannot be mixed during transporta-tion). The cost of road transportation is based on an hourly rate.When a trip is made between several suppliers or customers, thetransportation cost is shared between them based on their shareof the load weight. Schedule of the rail transportation is dictatedby the rail transportation company and the same is true for shiptransportation. Therefore, the schedules are inputs of the simula-tion model. Cost calculations for Rail and ship transportation costsare calculated using transportation rates.

Transportation orders generated by the pellet mill module (fordelivering the raw materials to the plant or delivering pellets tothe customers) are assigned with a due data that is used for sched-uling the transportation activities. The scheduling events are trig-gered whenever a truck becomes available, a new transportationorder is received, a specific amount of time is elapsed since the lastscheduling time, or availability of materials (raw materials or woodpellets) are changed. At each scheduling event, the orders aresorted based on the due dates and customers priorities. The ordersare then allocated to the available trucks and trucks are dispatchedwhen the required materials are available at the source.

4. Case study

The case study considered in this work is a pellet producer inBC, Canada. The input data and information are mostly providedby the company. In case the required data were not available, thedata are taken from the literature.

4.1. Supply and demand

The pellet mill in this case study has five suppliers listed inTable 2. Suppliers are wood processing plants that provide twotypes of raw material: sawdust and shavings. They operate fivedays a week and two shifts per day. Two of the suppliers are closeto the pellet mill and the raw material is conveyed to the pellet millusing a pipe; while the materials from the other three suppliers aretransported by trucks.

Table 2List of suppliers.

Supplier Transportation mode Distance to pellet mill (km) Availability (t day�1)a Cost ($ dt�1)a

Sawdust Shavings Sawdust Shavings

1 Truck 25.1 U (280, 300)b U (100, 140) 10.00 15.002 Truck 52.4 – U (50, 70) – 15.003 Truck 55.4 – U (30, 50) – 15.004 Air conveyed – – U (20, 40) – 15.005 Air conveyed – U (240, 260) – 10.00 –

a Data provided by the pellet mill.b Uniform distribution function.

Table 3Input data for customers.

Customer Distance (km) Transportation mode Annual demand (t)

Domestic1 125 Truck 7300Domestic2 205 Truck 7300Export 840 Railcar 146,000

Table 4The pellet plant input data and assumptions.

Item Value/explanations

Nominal throughput 20 t h�1

Working hours/days 24 h 365 days a yearElectricity cost 100 $ MW h�1

Annual interest rate 8%Cost of spare parts,

lubricants, and otherconsumables

8 $ t�1a

Personnel needs and salary There are 14 full-time personnel, with theannual cost of $60 k per person, operatingthe plant in three shifts. Also, two foremanand a supervisor are employed that cost$75 k and $90 k per person, respectively

a Data obtained from the pellet mill.

1244 M. Mobini et al. / Applied Energy 111 (2013) 1239–1249

For each supplier and each type of raw material, a daily avail-ability rate is taken as the input in the model. Uniform distributionfunctions are used in the model to reflect the fluctuations in thedaily available amounts. Based on the supply contract, the pelletmill has to take all the raw materials produced by the suppliersregardless of whether or not the pellet mill has the demand forthe wood pellets. The suppliers are contracted in a way that the de-mand of the pellet mill during the full-load operation is satisfied.The costs are based on dry tonne (dt) of materials as shown in Ta-ble 2. The moisture content of each load is calculated based onsamples taken from it and the suppliers are paid based on the driedweight of delivered raw material.

The initial analysis on the received data (maximum likelihoodestimates) show that moisture content of sawdust delivered tothe plant follows a Weibull distribution with location value of21, scale value of 9.02, and shape value of 3.39, and for shavings,the best fitted distribution was a Weibull distribution with locationvalue of 8, scale value of 3.32, and shape value of 2.02 (shown inFig. 4). The same analyses were conducted on the bulk density ofthe delivered materials and the best fitted distribution functionswere used in the simulation model.

The pellet company considered in this case study sells 90% of itsproducts to European customers through long-term contracts. Thewood pellets for European market are loaded into 90 tonne railcarsat the pellet mill and transported 840 km to the North Vancouverport, where the pellets are stored and then transported by oceanvessels. In addition to the export market, the pellet mill sells about15 k tonne of bagged pellets domestically. Pellet bags weigh about18 kg and are palletized in 0.5 tonne batches. In this study, it is as-sumed that the pellets are delivered to two distribution centers inthe area. Table 3 shows the input data regarding the customers inthe case study.

4.2. Pellet mill

The general input data and assumption regarding the pellet millare given in Table 4. The pellet plant (mill) operates seven days a

Fig. 4. Random distribution functions fitted to the empirical data

week and 24 h a day (365 days per year) with 20 t h�1 nominalcapacity. Sawdust is processed in a drum dryer and is fed intothe hammer mill while shavings bypass the dryer and are directlyfed to the hammer mill. The required heat for the drying process isgenerated in a 16 Mbtu furnace that burns wet sawdust. Groundmaterials are processed with four pelletizers and are fed into thecoolers. After separating the fines in the shaker screens, the woodpellets are stored in the storage bins or packaged and then trans-ported to the customers by railcars or trucks.

The capital cost of the pellet mill is estimated based on the dataprovided by Sultana et al. [14] as shown in Table 5. The cost figuresare adjusted for currency exchange and inflation rate. The ex-change rate of 1.12 (US$/CAD$) and the annual rate of inflation

for moisture content of sawdust (left) and shavings (right).

Table 5Capital cost estimation for the pellet plant.

Item Amortizationperioda

Capital cost for6 t h�1 plantb

Estimated capital costfor 20 t h�1 plant

Numberof pieces

Annualized cost

Dryer 15 $430,000.00 $1,788,924.31 1 $208,999.21Hammer mill 15 $150,000.00 $978,565.74 2 $114,325.39Pellet mill 15 $350,000.00 $2,878,614.31 4 $336,307.20Cooler 15 $170,000.00 $1,097,768.33 2 $128,251.77Screener 15 $18,300.00 $90,476.93 1 $10,570.38Bagging system 15 $450,000.00 $2,306,670.85 1 $269,487.31Conveyers, tanks, etc. 15 $1,113,000.00 $6,591,947.31 1 $770,134.20Land, infrastructure, construction and other 25 – $4,300,000 – $402,818.75

a Ref. [8].b Ref. [14].

Table 6Input data for the equipment.

Equipment Heat demandðkW h t�1

ev:w:ÞFeed rate(t h�1)

Power(kW h)

Fuel Burner efficiency(%)

Simultaneity factor(%)

TBFd TTRe

Specification of the dryerDrum

Dryer1,300a 25a 270a Wet

sawdustU (50-60)b U (85–100)c Exponential

(25 h)Exponential(1 h)

Equipment Number of pieces Capacity (t/h) Power (kW) Simultaneity factor (%) TBF TTR

Specification of size reduction, pelletization, and cooling equipmentHammer mill 2 10 370a U (85–100)a Exponential (15 h) Exponential (0.5 h)Pellet mill 4 5 300a U (85–100)a Exponential (10 h) Exponential (1 h)Cooler 2 10 25a U (85–100)a Exponential (30 h) Exponential (0.5 h)

a Data obtained from the pellet mill.b Ref. [32].c Ref. [8].d Time between failure.e Time to repair.

M. Mobini et al. / Applied Energy 111 (2013) 1239–1249 1245

of 1.21% are used in the calculations [30,31]. The total capital costwas estimated at $20 M for the nominal capacity of 20 t h�1.

Specifications of the equipment are shown in Table 6. The heatrecovery factor in the biomass burner is assumed to follow a uniformdistribution with 50% lower and 60% upper bounds [32]. The repairand maintenance in the pellet plant has a significant impact on plantup-time and the final cost of produced pellets [7,8]. The data on themachine failure and required maintenance are scarce. Nilsson [33]used exponential distribution functions to describe the failure ofthe bailers and loaders. In this research, the same approach is usedto count for the down-time of equipment. It is assumed that the TimeBetween Failure (TBF) and Time To Repair (TTR) for the equipmentfollow probability functions shown in Table 6. The parameters areset after consultation with industry experts.

4.3. Raw material transportation and wood pellet distribution

Transportation of raw materials from suppliers to the pelletplant and delivery of the wood pellets to the domestic customersis outsourced to a trucking company. The transportation companyworks seven days a week and two shifts per day (7 a.m. to 10 p.m.).It is assumed that there are 10 similar trucks available. The fuelconsumption rate and the average speed of the trucks in each tripfollow a uniform distribution function with lower and upperbounds shown in Table 7.

The rail transportation to the harbor costs 28 $ t�1, which in-cludes transportation cost and all related fees such as railcar rent-

Table 7Trucks input data.a

Weight capacity (t) Volume capacity (m3) Fuel cons. rate (Lit km�1) Average s

45 113 U (.30, .35) U (60, 80)

a Data obtained from the pellet company.

als and insurance. In practice, railcars are filled and delivered to theport on a daily basis. The pellet mill contract with the Europeancustomers is to deliver the wood pellets to the ocean vessels, afterthat the customers are responsible for ocean transport and costs.

In BC, the energy intensity factor of electricity is 24 gCO2 kW h�1 [34] that is used to calculate the emissions due to elec-tricity consumption at the plant. When burning biomass to gener-ate the heat required in the drying process, a factor of301.8 g kW h�1 is used to calculate the amount of CO2 emissions[35]. The energy consumed by trucks and generated emissionsfrom it are calculated based on the energy content of diesel(9.96 kW h lit�1) and CO2 emissions of 2.7 kg from one liter of die-sel [36]. The energy consumption of 0.09 kW h(t km)�1 and thegenerated CO2 emission of 16.38 g(t km)�1 are assumed for railtransportation [37].

5. Results and discussions

The simulation time is set to one year and the model was run for50 iterations. The number of iterations was calculated based on the95% confidence level in the estimated value of the total production.

5.1. Supply chain costs

The total production was estimated at 156.87 kt of pellets withstandard deviation of 370 (95% confidence interval: 156.87 k ±

peed (km h�1) Unloading rate (m3 h�1) Loading rate (m3 h�1) Cost ($ h�1)

200 200 114

Table 10Outputs regarding the distribution of wood pellets.

Customer Deliveredweight (t)

Transportationcost ($)

Unittrans.cost($ t�1)

Deliveredcost($ t�1)

Deliveredcost($ GJ�1)

Domestic1 7300.00 205,353.22 28.13 97.40 5.43Domestic2 7300.00 317,781.99 43.53 112.80 6.28Export 141,494.40 3,961,843.20 28.00 97.27 5.42

1246 M. Mobini et al. / Applied Energy 111 (2013) 1239–1249

725.20). This equals to 89.54% of the nominal capacity which com-plies with the values reported by other authors for plant availabil-ity in a year. In the literature, 85–91% plant availability factor iscommonly reported [7–9]. The contribution of different cost itemsin the aggregate cost structure of the supply chain is shown inTable 8.

Standard deviation of the annual cost elements is shown inTable 8. The histogram of the estimated annual cost of the supplychain drawn from 50 runs of the simulation model is shown inFig. 5. The annual cost of the supply chain was estimated withina range of 14.59 M$ and 14.67 M$ at 95% confidence level (usingthe central limit theorem). The small standard deviations (Table 8)for this case study are due to the long term contracts of the pelletmill with suppliers and customers, which may not be the case formany small sized pellet mills.

Raw material procurement and transportation constituted29.16% of the total cost. Total wet weight of delivered raw materi-als to the pellet mill was about 203.02 kt ($155.52 kt dried weight).

Table 8Supply chain costs.

Item Annualcost (k$)

Stn. dev. Unitcost($ dt�1)

Unitcost($ GJ�1)

Raw material 1839.78 2679.80 $11.83 $0.59Raw material transportation 2427.51 4850.32 $15.61 $0.78

Pellet productionDrying 109.53 253.15 $0.78 $0.04Size reduction 491.94 1106.31 $3.50 $0.17Pelletization 797.56 1745.47 $5.67 $0.28Cooling 35.48 54.96 $0.25 $0.01Annualized capital investment 2240.89 0.00 $15.94 $0.80Personnel 1080.00 0.00 $7.68 $0.38Spare parts and other

consumables1124.67 2643.88 $8.00 $0.40

Pellets distribution 4484.98 9902.64 $32.06 $1.60Total 14,632.34 20,148.87 $101.33 $5.06

Fig. 5. Histograms and best fitted distribution functions for annual cost

Table 9Quantities and cost of raw material delivered to the pellet plant.

Supplier Sawdust Shavings

Weight (kt) Material (k$) Trans. (k$) Weight (kt) Cost (k$)

1 73.86 523.50 1007.03 30.02 401.042 0 0.00 0.00 15.63 208.823 0 0.00 0.00 10.43 139.344 0 0.00 0.00 7.83 104.625 65.25 462.46 0.00 0.00 0.00Total 139.11 985.96 1007.03 63.91 853.82

Table 9 shows the aggregated results in terms of weight and cost ofsupplied raw materials from each supplier to the pellet plant. Onaverage, each tonne of raw materials delivered to the plant costs21.02 $ t�1 (equivalent to 27.44 $ dt�1). Delivered sawdust andshavings cost 14.33 $ t�1 and 35.59 $ t�1, respectively.

Drying, size reduction, pelletization, and cooling processes cost10.20 $ dt�1 of wood pellets, which compose 9.80% of the totalcosts. The costs related to production of wood pellets (includingprocessing, capital investment, personnel, and consumables) is41.83 $ t�1, which is 40.19% of the final cost. The estimated costof wood pellets at the gate of the pellet mill is 69.27 $ dt�1. Distri-bution of wood pellets accounts for 30.65% of the total cost, onaverage 32.06 $ t�1 (see Table 8). Table 10 lists the deliveredweight and distribution cost of wood pellets for each customer.Delivered wood pellets to the export port costs 97.27 $ t�1. Thecost of delivered wood pellets to domestic customers depends onthe distance from the pellet mill as shown in Table 10.

5.2. Energy consumption and CO2 emissions

The energy consumption and CO2 emissions along the supplychain are listed in Table 11. The total annual energy consumptionwas 89.19 GW h. On average, the energy input was568.42 kW h t�1. The total annual CO2 emission was 21.46 tonne,an average of 136.76 kg per tonne of wood pellets.

The most energy intensive part of the supply chain was the dry-ing process with an average of 398.38 kW h t�1 including heat and

of the supply chain (left) and delivered cost (right) of wood pellets.

Total

Trans. (k$) Weight (kt) Material (k$) Trans. (k$) Ave. Cost ($ t�1)

651.08 103.88 924.54 1658.11 24.86454.63 15.63 208.82 454.63 42.44314.77 10.43 139.34 314.77 43.55

0.00 7.83 104.62 0.00 13.360.00 65.25 462.46 0.00 7.09

1420.48 203.02 1839.78 2427.51 21.02

Table 11Energy consumption and CO2 emissions.

Item Energy use (kW h) Energy use (kW h t�1pellets) CO2 emission (t) CO2 (kgCO2 t�1

pellets

Drying 62,591,069.80 399.00 18,585.70 118.48Grinding 4,919,408.34 31.36 118.07 0.75Pelletization 7,975,567.67 50.84 191.41 1.22Cooling 354,754.66 2.26 8.51 0.05Trucking 2,314,441.10 14.75 627.41 4.00Rail transportation 11,093,160.56 70.71 1,946.85 12.41Total 89,248,402.14 568.93 21,477.95 136.91

Table 12Outputs of the drying process.

Heat(GW h)

Electricity(GW h)

CO2

emissions(t)

Processedweight (kt)

Evap.water (kt)

Fuelweight(kt)

61.50 1.10 18.59 93.30 26.02 17.45

M. Mobini et al. / Applied Energy 111 (2013) 1239–1249 1247

electricity. Table 12 lists the results for the drying process. It wasestimated that 26.02 kt of water was evaporated that required61.50 GW h of energy. The consumed fuel weight and estimatedCO2 emissions due to drying process are shown in Table 12.

Truck and rail transportation was also an important contributorto the supply chain CO2 emissions. Table 13 lists the outputs of themodel for truck transportation. The total travel distance was774,579 km and total travel time 25,883 h. The total fuel consump-tion was 232, 374 l. The emission generated due to the truck trans-portation was 627.41 t CO2 and the total energy consumption was2.31 GW h. The CO2 emission due to rail transportation was 1950tonne and the energy consumption was about 11.10 GW h.

5.3. Scenario analysis

Two potential modifications to the supply chain are investi-gated in this study: (1) changing the drying fuel from sawdust tobark and hog fuel, and (2) producing lower quality pellets for theinternational market. These two alternations were suggested bythe pellet company president. In order to evaluate the effects ofthese modifications, three scenarios were developed.

Table 13Truck transportation outputs.

Cost(M$)

Fuel (Lit) Distance(km)

Traveltime (h)

CO2

emission (t)Energy(GW h)

2.95 232,373.60 774,578.68 25,882.87 627.41 2.31

Table 14Suppliers input data for Scenario1 and Scenario 2.

Supplier Production (t day�1)

Sawdust Shavings Bark

Scenario 1 1 U (210, 230) U (100, 140) –2 – U (50, 70) –3 – U (30, 50) –4 – U (20, 40) –5 U (240, 260) – U (80, 90)

Scenario 2 1 U (280, 300) U (100, 140) –2 – U(60, 80) –3 – – –4 – – –5 U (240, 260) – U (90, 110)

5.3.1. Changing the drying fuelThe sawdust which is burnt and used as the drying fuel could be

converted to wood pellets to generate more profit for the company,if a sensible and low cost drying fuel was available. One option is toburn bark. The pellet mill considered in this study could receivebark from adjacent sawmills free of charge. However, in order tomake the transition to use bark, the company has to improve itscombustor and emission control system. To do so, a capital invest-ment of 1.5 M$ is required. The simulation model was used to eval-uate this option.

In order to run this scenario, the input parameters of the modelwere modified. The moisture content of bark is assumed to follow aUniform (40%, 55%) distribution function [38]. When consumingbark as the drying fuel, the total amount of required sawdust andshavings would be less than those for the base case. There weretwo possible scenarios: (a) to purchase less sawdust, or (b) to pur-chase less shavings. There is a trade-off between the changes in thematerial procurement cost and transportation cost, which dependon moisture content and bulk density, from one hand, and cost ofraw material on the other hand. Sawdust is cheaper than shavingsper dried tonne delivered, but has higher moisture content that in-creases the transportation cost. The lower bulk density of shavings(on average 130.83 kg m�3) compared with sawdust (on average226.21 kg m�3) is another factor that is taken into account. Toquantify the differences between these two options, two scenarioswere run using the simulation model. In Scenario 1, the amount ofsawdust procurement from Supplier 1 is reduced. In Scenario 2, itis assumed that the amount of purchased shavings is reduced andmore sawdust is used in the mixture of raw materials. Table 14shows the input parameters for Scenarios 1 and 2. The daily avail-abilities of sawdust and shavings at the suppliers’ locations arechanged and bark is supplied from Supplier 5.

The obtained results for the base case and Scenarios 1 and 2 arecompared in Table 15. The amount of bark consumed in Scenario 1and Scenario 2 were about 21.71 kt and 25.10 kt, respectively.More bark was required in Scenario 2 due to the higher moisturein the sawdust. Compared to the base case scenario, the final costof delivered pellets would be slightly reduced by about 1% and 1.5%in Scenario 1 and Scenario 2, respectively. The energy consumptionand CO2 emissions in Scenario 1 were not significantly differentthan the base case scenario. However, the energy consumptionand CO2 emissions in Scenario 2 compared to the base case wereincreased by 10.67% and 13.17%, respectively. The increases weredue to the higher drying process needed as more sawdust wasused.

5.3.2. Producing lower quality pelletsBC wood pellet producers have traditionally been located in

close proximity of sawmills and other wood processing plants.Thus, they have had easy access to abundant sawdust and shavingsand have produced premium quality pellets [39]. Bark is lessexpensive than sawdust and shavings and using bark in the blendof raw materials for pellet production increases the heat value andmechanical strength of the pellets [12]. The PFI standard for wood

Table 15Costs of raw material procurement and transportation.

Scenario Fuel weight (kt) Annualizedcapital ($ t�1)

Raw materialcost ($ t�1)

Raw materialtransportation cost ($ t�1)

Final cost($ t�1)

Energy consumption(kW h t�1)

CO2 emissions(kg t�1)

Base case 17.45 $15.94 $11.83 $15.61 $101.33 568.93 136.91Scenario 1 21.71 $16.94 $11.10 $14.30 $100.31 567.63 136.54Scenario 2 25.10 $16.92 $10.52 $14.30 $99.81 629.61 154.93Scenario 3 30.34 $16.93 $9.37 $12.54 $96.51 724.73 183.46

1248 M. Mobini et al. / Applied Energy 111 (2013) 1239–1249

pellets allows utility grade wood pellets to contain less than 6% ashcontent. In Scenario 3, it is assumed that the pellet plant uses 10%bark in its raw material composition to produce lower qualitywood pellets. The required bark material is supplied from the adja-cent sawmills without any truck transportation.

The simulation results show that adding bark to the raw mate-rial mixture in addition to using bark as the fuel in the drying pro-cess leads to a cost saving of about 4.75% (4.82 $ t�1) in the finalcost of wood pellets. This reduction is due to lower procurementand transportation of raw materials as shown in Table 15. The en-ergy consumption and CO2 emissions in Scenario 3 were, respec-tively, 27.38% and 34.00% higher than those in the base casescenario due to the additional drying requirement for added barkin the raw material mixture.

6. Conclusions and future directions

A discrete event simulation model, called PSC, was developed tofacilitate the design and analysis of the wood pellet supply chains.In comparison to the static modeling, simulation modeling enablesto incorporate uncertainties, interactions between actors, andinterdependencies along the supply chain. However, it requiresmore data and information about the supply chain. Different as-pects of the system, including procurement and transportation ofraw material, processes inside the pellet plant, and distributionof the pellets, are captured in the PSC model, while consideringthe dynamics of the system. The model was applied to a case studyto estimate the delivered cost of the wood pellets to domestic cus-tomers and an international port. The average cost of delivered pel-lets was estimated at 101.33 $ t�1. Raw material procurement andtransportation accounted for 29.16% of the cost. Processing, capitalinvestment, spare parts and consumables, and personnel costs con-stituted 40.19% of the total supply chain costs. The distribution ofwood pellets contributed 30.65% to the total cost.

In addition to the base case scenario, three scenarios weredeveloped and analyzed to evaluate the effects of changing thedrying fuel in the pellet mill and changing the mixture of rawmaterials used for pellet production. It was estimated that about4.75% reduction in the final cost of delivered fuel was achievableby changing the type of fuel for the drying process and the mixof raw material, however, the total energy consumption and CO2

emissions were increased.The obtained results from the simulation model could be more

accurate if improved functional relationships were used for eachoperation. For instance, the effects of using different types andquality of raw material on the performance of the equipment arenot considered in the model since the required data were notavailable. Therefore, by developing functions that represent therelationship between the performance of each piece of equipmentand the type of raw material used in the process, it would bepossible to improve the precision of the results. Additionally, thesources of uncertainties in the real system are not limited to thoseconsidered in this study. Providing more data on the input param-eters could also increase the reliability of the outputs of the model.

The simulation model introduced in this manuscript could beused as a decision support tool to evaluate modifications in

existing supply chains and to design new chains. The outputs ofthe model could be used in conducting feasibility analysis, riskanalysis, and life cycle analysis on the given supply chain.

Acknowledgements

This research is funded partly through the University of BritishColumbia Graduate Fellowship to the senior author, by NaturalSciences and Engineering Research Council of Canada (NSERC),BC Ministry of Forest, and Wood Pellet Association of Canada. Thiswork continues the development of Integrated Biomass SupplyAnalysis & Logistics (IBSAL) [40] by adding the wood pellet supplychain model to it. The Office of Biomass Program of the US DOE isacknowledged for supporting the development of IBSAL model atthe Oak Ridge National laboratory and at the University of BritishColumbia. We also acknowledge the great support of the pelletmill’s President in providing the required data and informationfor our modeling and validating our results.

References

[1] Wolf A, Vidlund A, Andersson E. Energy-efficient pellet production in the forestindustry—a study of obstacles and success factors. Biomass Bioenergy2006;30:38–45.

[2] Murray G. Canadian perspective on EU and other export markets; 2012.[3] Cocchi M. Global wood pellet industry market and trade study; 2011.[4] Stelte W, Sanadi AR, Shang L, Holm JK, Ahrenfeldt J, Henriksen UB. Recent

developments in biomass pelletization – a review. bioresources, vol. 7; 2012.[5] Murray G. Canadian wood pellet industry perspectives; 2012.[6] Pirraglia A, Gonzalez R, Saloni D. Techno-economical analysis of wood pellets

production for US manufacturers. Bioresources 2010;35:5.[7] Uasuf A, Becker G. Wood pellets production costs and energy consumption

under different framework conditions in Northeast Argentina. BiomassBioenergy 2011;35:1357–66.

[8] Obernberger I, Thek G. The pellet handbook; the production and thermalutilization of pellets. 1st ed. London & Washington, DC: Earthscan; 2010.

[9] Thek G, Obernberger I. Wood pellet production costs under Austrian and incomparison to Swedish framework conditions. Biomass Bioenergy2004;27:671–93.

[10] Mani S, Sokhansanj S, Bi X, Turhollow A. Economics of producing fuel pelletsfrom biomass. Appl Eng Agric 2006;22:421–6.

[11] Peng JH, Bi HT, Sokhansanj S, Lim JC, Melin S. An economical and marketanalysis of Canadian wood pellets. Int J Green Energy 2010;7:128–42.

[12] Melin S. Bark as feedstock for production of wood pellets; 2008.[13] Spelter H, Toth D. North American wood pellet sector; 2009.[14] Sultana A, Kumar A, Harfield D. Development of agri-pellet production cost

and optimum size. Bioresour Technol 2010;101:5609–21.[15] Sultana A, Kumar A. Ranking of biomass pellets by integration of economic,

environmental and technical factors. Biomass Bioenergy 2012;39:344–55.[16] Duncan A, Pollard A, Fellouah H. Torrefied, spherical biomass pellets through

the use of experimental design. Appl Energy 2013;101:237–43.[17] Kohl T, Laukkanen T, Järvinen M, Fogelholm C. Energetic and environmental

performance of three biomass upgrading processes integrated with a CHPplant. Appl Energy 2013;107:124–34.

[18] Schmidt J, Leduc S, Dotzauer E, Kindermann G, Schmid E. Cost-effective CO2

emission reduction through heat, power and biofuel production from woodybiomass: a spatially explicit comparison of conversion technologies. ApplEnergy 2010;87:2128–41.

[19] Sikkema R, Steiner M, Junginger M, Hiegl W, Hansen MT, Faaij A. The Europeanwood pellet markets: current status and prospects for 2020. Biofuels BioprodBiorefining 2011;5:250–78.

[20] Lu N, Rice RW. Characteristics of wood fuel pellet manufacturers and marketsin the United States, 2010. Forest Prod J 2011;61:310–5.

[21] Pang S, Mujumdar AS. Drying of woody biomass for bioenergy: dryingtechnologies and optimization for an integrated bioenergy plant. DryingTechnol 2010;28:690–701.

M. Mobini et al. / Applied Energy 111 (2013) 1239–1249 1249

[22] Thibault J, Duchesne C. Evaluation of simple control strategies for rotarydryers. Drying Technol 2004;22:947–62.

[23] Li H, Chen Q, Zhang X, Finney KN, Sharifi VN, Swithenbank J. Evaluation of abiomass drying process using waste heat from process industries: a case study.Appl Therm Eng 2012;35:71–80.

[24] Di Giacomo G, Taglieri L. Renewable energy benefits with conversion of woodyresidues to pellets. Energy 2009;34:724–31.

[25] PFI Standards Committee. Pellet fuels institute standard specification forresidential/commercial densified. Fuel; 2011.

[26] Imagine That. ExtendSim simulation software; 2011.[27] Krahl D. ExtendSim advanced technology: discrete rate, simulation; 2009.[28] Perry RH, Maloney JO, Green D. Perry’s chemical engineers’ handbook. 8th

ed. NY, USA: McGraw-Hill; 1997.[29] Microsoft. Mappoint; 2012.[30] Bank of Canada. Inflation calculator. <http://www.bankofcanada.ca/rates/

related/inflation-calculator/> [accessed December 2012].[31] Bank of Canada. Exchange rates. <http://www.bankofcanada.ca/rates/

exchange/> [accessed December 2012].[32] Office of utility technologies. Renewable energy technology characterizations;

1997. TR-109496.

[33] Nilsson D. SHAM—a simulation model for designing straw fuel deliverysystems. Part 1: model description. Biomass Bioenergy 1999;16:25–38.

[34] Environment Canada. Electricity intensity tables. <http://www.ec.gc.ca/ges-ghg/default.asp?lang=En&n=EAF0E96A-1> [accessed December 2012].

[35] Environmental Protection Agency (EPA). The compilation of air pollutantemission factor: stationary point and area sources; 1995. EPA No.: AP-42.

[36] Natural Resources Canada. The 2011 fuel consumption guide. <http://oee.nrcan.gc.ca/publications/transportation/fuel-guide/2011/index.cfm>[accessed December 2012].

[37] Magelli F, Boucher K, Bi HT, Melin S, Bonoli A. An environmental impactassessment of exported wood pellets from Canada to Europe. BiomassBioenergy 2009;33:434–41.

[38] Kehbila T. Evaluation of primary wood processing residues for bioenergy inBritish Columbia [dissertation]. The University of British Columbia; 2010.

[39] Tumuluru JS, Sokhansanj S, Lim CJ, Bi T, Lau A, Melin S, et al. Quality of woodpellets produced in British Columbia for export. Appl Eng Agric2010;26:1013–20.

[40] Sokhansanj S, Kumar A, Turhollow AF. Development and implementation ofintegrated biomass supply analysis and logistics model (IBSAL). BiomassBioenergy 2006;30:838–47.