Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An agent-based supply chain model for strategic analysis in forestry Vahid, Saba 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2012_spring_vahid_saba.pdf [ 1.34MB ]
Metadata
JSON: 24-1.0072459.json
JSON-LD: 24-1.0072459-ld.json
RDF/XML (Pretty): 24-1.0072459-rdf.xml
RDF/JSON: 24-1.0072459-rdf.json
Turtle: 24-1.0072459-turtle.txt
N-Triples: 24-1.0072459-rdf-ntriples.txt
Original Record: 24-1.0072459-source.json
Full Text
24-1.0072459-fulltext.txt
Citation
24-1.0072459.ris

Full Text

An Agent-Based Supply Chain Model for Strategic Analysis in Forestry  by  Saba Vahid  M.A.Sc., The University of British Columbia, 2006 B.Sc., Sharif University of Technology, 2002  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE STUDIES (Forestry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)     December 2011  © Saba Vahid, 2011   ii Abstract An agent-based forest sector model, CAMBIUM 2.0, is developed and applied to case studies of the forest industry in the coastal British Columbia (BC). By combining optimization and simulation, this model allows policy makers and managers to examine the impact of different supply chain (SC) configurations (e.g. establishing new facilities), and changing forest management policies (e.g. harvest restrictions). The forest sector structure and the state of the forest resources that develop over time are a result of autonomous agents interacting with each other while competing for available forest resources needed to manufacture forest products. The thesis is presented in four chapters. Chapter 1 introduces SC modelling concepts and techniques and identifies research objectives and methods. Chapter 2 presents and discusses the structure of the agent-based simulation model and the formulation of the facility location problem, presenting a novel algorithm for integrating the optimization problem with the simulation model. The model is applied to the case of a forest industry SC to establish a new agent. The predictions of the new agent about its profits are not strongly affected by higher levels of information about the cost structure of its competitors, while improving the accuracy of market predictions has a noticeable impact on such predictions. Chapter 3 evaluates the impact of establishing a log sort yard on the profitability of the forest products SC. Considering different market price scenarios, establishing a sort yard does not seem to benefit the forest products SC, mainly because of intense competition for timber. In Chapter 4, CAMBIUM 2.0 is used to investigate the impact of harvest policy changes on the SC performance and the timber supply sustainability. Alternative harvest priorities (e.g. harvesting stands with highest value first) and modifying the harvesting preference of the mills (i.e. harvesting a mix of high and low value stands) improves the timber supply sustainability with less negative economic impacts compared to lowering the harvest limit. The modelling framework developed in this research can be extended to address other research questions such as changing log export policies, setting stumpage prices, or encouraging replanting of economically desirable species.  iii Preface This thesis is based on a series of manuscripts that are published or will be submitted for publication in peer-reviewed journals. I developed the model, gathered the required data from various sources, designed and conducted the simulation experiments, and wrote all of the manuscripts. Dr. Thomas Maness, my PhD supervisor, advised me in the process of model development and validation, as well as designing the scenario analyses. He also edited the manuscripts and is the co-author on all articles. The simulation model developed in my dissertation was based on a model created by Dr. Olaf Schwab and Dr. Thomas Maness in the University of British Columbia (Schwab et al., 2008) which was extended and modified significantly to fit the purpose of my research. My dissertation research was conducted in collaboration with FPInnovations (Vancouver, BC). Mr. Joel Mortyn and Mr. Jack MacDonald from FPInnovations helped in estimating and verifying the collected forest resource and industrial data during model development. Versions of the following chapters have been or will be submitted for publication: Chapter 1: Vahid, S. and Maness, T. (2010). Modelling customer demand in forest products industry supply chains: a review of the literature. International Journal of Simulation and Process Modelling. 6(2):103 – 114. Chapter 2. Vahid, S. and Maness, T. (to be submitted). New Facility Location in a Forest Products Supply Chain Model. Chapter 3. Vahid, S. and Maness, T. (to be submitted). Impact of Establishing a Centralized Sort Yard in Coastal British Columbia. Chapter 4. Vahid, S. and Maness, T. (to be submitted). Impact of Harvest Policy Changes on Sustainability.   iv Table of Contents  Abstract .................................................................................................................................... ii  Preface ..................................................................................................................................... iii  Table of Contents ................................................................................................................... iv  List of Tables ......................................................................................................................... vii  List of Figures ....................................................................................................................... viii  List of Abbreviations ............................................................................................................ xii  Acknowledgments ................................................................................................................ xiii  Dedication ............................................................................................................................. xiv  Chapter 1: Introduction ......................................................................................................... 1  1.1  Forest Products Supply Chain ............................................................................................... 1  1.2  Research Objectives .............................................................................................................. 4  1.3  Supply Chain Models ............................................................................................................ 4  1.4  Forest Products Supply Chain Models .................................................................................. 6  1.4.1  Optimization Models ........................................................................................................ 7  1.4.2  Simulation Models .......................................................................................................... 11  1.5  Facility Location in Supply Chain Models ......................................................................... 16  1.6  Policy Analysis Using Supply Chain Models ..................................................................... 18  1.7  Summary of Relevant Literature ......................................................................................... 19  1.8  Structure of This Thesis ...................................................................................................... 20  Chapter 2: New Facility Location in a Forest Products Supply Chain Model................ 22  2.1  Introduction ......................................................................................................................... 22  2.2  Facility Location in CAMBIUM ......................................................................................... 25  2.2.1  CAMBIUM 2.0 ............................................................................................................... 26  2.2.2  Single Facility Location Optimization Problem ............................................................. 28  Model Assumptions ................................................................................................................. 30  Parameters ............................................................................................................................... 31  Variables .................................................................................................................................. 32  Objective Function .................................................................................................................. 33   v Constraints ............................................................................................................................... 33  2.2.3  CAMBIUM 2.0 Simulation Flow ................................................................................... 37  2.3  Case Study: British Columbia’s Coastal Primary Forest Products Industry ....................... 43  2.3.1  Forest Inventory Data ..................................................................................................... 44  2.3.2  Operational Data for Manufacturing Facilities ............................................................... 45  2.3.3  Scenarios ......................................................................................................................... 48  2.4  Results & Discussion .......................................................................................................... 49  2.4.1  Predicted Profits ............................................................................................................. 49  2.4.2  Observed Profits ............................................................................................................. 51  2.4.3  Predicted and Observed Log Exports ............................................................................. 52  2.5  Conclusions ......................................................................................................................... 54  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia ................................................................................................................................................. 56  3.1  Introduction ......................................................................................................................... 56  3.2  Modelling Sort Yard Operations in CAMBIUM 2.0 .......................................................... 59  3.3  Data and Scenarios .............................................................................................................. 60  3.3.1  Case Study Data .............................................................................................................. 60  3.3.2  Market Price Scenarios ................................................................................................... 60  3.4  Results and Discussion ........................................................................................................ 62  3.4.1  Scenario I ........................................................................................................................ 62  3.4.2  Scenario II ....................................................................................................................... 69  3.5  Conclusions ......................................................................................................................... 73  Chapter 4: Impact of Harvest Policy Changes on Sustainability ..................................... 75  4.1  Introduction ......................................................................................................................... 75  4.1.1  History of Forest Management in BC ............................................................................. 75  4.1.2  Coastal BC Timber Resource ......................................................................................... 77  4.2  Methods ............................................................................................................................... 79  4.3  Data and Scenarios .............................................................................................................. 81  4.3.1  Data ................................................................................................................................. 81  4.3.2  Scenarios ......................................................................................................................... 82  4.4  Results and Discussion ........................................................................................................ 84  4.4.1  Impact of AAC Reduction .............................................................................................. 84  4.4.2  Impact of Harvest Priority Change ................................................................................. 89   vi 4.4.3  Impact of Removing Quality Requirements ................................................................... 93  4.5  Conclusion .......................................................................................................................... 96  Chapter 5: Model Validation and Verification .................................................................. 98  5.1  Introduction ......................................................................................................................... 98  5.2  Verification ......................................................................................................................... 98  5.3  Validation .......................................................................................................................... 102  Chapter 6: Conclusions ...................................................................................................... 110  6.1  Conclusions ....................................................................................................................... 110  6.2  Limitations ........................................................................................................................ 113  6.3  Future Work ...................................................................................................................... 114  References ............................................................................................................................ 117  Appendices ........................................................................................................................... 129  Appendix A . Additional Graphs for Chapter 2 ............................................................................. 129  A.1  Average Agents Profit in the Optimal Solution of Facility Location Problem ............. 129  Appendix B . Additional Graphs for Chapter 3 ............................................................................. 131  B.1  Average Total Harvest .................................................................................................. 131  B.2  Average Total Saw Log Volume Purchased by Sawmills from the Sort Yard ............. 132  Appendix C . Additional Graphs for Chapter 4 ............................................................................. 133  C.1  Base Case ...................................................................................................................... 133  C.2  Scenarios I and II .......................................................................................................... 134  C.3  Scenarios III and IV ...................................................................................................... 134  C.4  Age Distribution of Timber Harvesting Land Base (THLB) for All Scenarios ............ 136   vii List of Tables Table  2.1  Log quality ratio by site index ........................................................................... 41  Table  2.2  Cost, recovery, and other operating assumptions for BC Coast sawmills ......... 46  Table  2.3  Harvesting and transportation cost assumptions for the BC Coast .................... 46  Table  2.4  Cost, recovery, and other operating assumptions for BC Coast sort yards ........ 47  Table  2.5  Log and lumber prices for BC Coast at the beginning of the simulation .......... 47  Table  3.1  Average sawmill product recovery factors for scenario I .................................. 69  Table  3.2  Average sawmill product recovery factors for scenario II ................................. 73  Table  4.1  Log quality ratio by site index and age group .................................................... 80  Table  4.2  Sawmill costs for all scenarios ........................................................................... 81  Table  4.3  Market price for all scenarios ............................................................................ 82  Table  4.4  Scenario descriptions ......................................................................................... 83  Table  5.1  Initial sawmill specifications ............................................................................. 99    viii List of Figures Figure  1.1  A typical forest products supply chain............................................................. 1  Figure  2.1  Value of manufacturing shipments for sawmills and wood preservation sector in BC,Source: Statistics Canada (2011b) ........................................................ 22  Figure  2.2  (a) BC Coast total sawn lumber production and its share of Canadian sawn lumber production, (b) Wood products manufacturing employment level and its share of total employment in BC, Source: Statistics Canada (2011a, 2011c) ......................................................................................................................... 23  Figure  2.3  CAMBIUM 2.0 flow of simulation in every time step ................................... 28  Figure  2.4  Flow of Simulation in CAMBIUM 2.0 ........................................................... 38  Figure  2.5  Strategy selection schematic ........................................................................... 39  Figure  2.6  Forest region and sawmills included in the case study of the coastal BC industry ........................................................................................................... 44  Figure  2.7  Percentage of change in market prices for log and lumber products .............. 48  Figure  2.8  (a) Average profits of existing agents in the optimal solution for the facility location problem, (b) Average profit of new agent in the optimal solution for the facility location problem ........................................................................... 50  Figure  2.9  (a) Average profit (േ 1SD) of existing agents based on simulation results in scenario 3,  (b) Average profit (േ 1SD) of the new agent based on simulation results, scenario 3 ............................................................................................ 52  Figure  2.10  Average log export ratio (േ 1SD) of the new sort yard based on simulation results .............................................................................................................. 54  Figure  3.1  Ratio of saw and pulp logs delivered from the forest and the sort yard ......... 60  Figure  3.2  (a) Framing lumber composite price, (b) Weighted average log price (over all species) for the BC Coast, Source: Random Lengths (2011), BC Ministry of Forests, Land, and Natural Resource Operations (2011c) .............................. 61  Figure  3.3  Market price changes for scenario II .............................................................. 62  Figure  3.4  (a) Average total profits of supply chain members (േ1SD) scenario I – without a sort yard, (b) Average total profits of supply chain members (േ1SD) scenario I – with a sort yard ............................................................................ 63  Figure  3.5  (a) Average total profits of supply chain members with and without the sort yard: scenario I, (b) Average total harvest volume with and without the sort yard: scenario I ................................................................................................ 64  Figure  3.6  (a) Average total imported logs (േ1SD), scenario I – without a sort yard, (b) Average total imported and purchased sort yard logs (േ1SD), scenario I – with a sort yard ................................................................................................ 65  Figure  3.7  Log output and export volume for the sort yard, scenario I ............................ 66   ix Figure  3.8  (a) Average lumber production and capacity levels (േ1SD), scenario I – without a sort yard, (b) Average lumber production and capacity levels (േ1SD), scenario I – with a sort yard ............................................................. 66  Figure  3.9  Log production of the sort yard (േ1SD) for scenario I .................................. 68  Figure  3.10  (a) Average total profits of supply chain members (േ1SD): scenario II – without a sort yard, (b) Average total profits of supply chain members (േ1SD): scenario II – with a sort yard ............................................................ 69  Figure  3.11  Average total harvest volumes with the presence of a sort yard for scenario I and II ............................................................................................................... 70  Figure  3.12   (a) Average total imported logs (േ1SD), scenario II – without a sort yard, (b) Average total imported and purchased sort yard logs (േ1SD), scenario II – with a sort yard ................................................................................................ 71  Figure  3.13  Log output and export volume for the sort yard for scenario II ...................... 71  Figure  3.14   (a) Average lumber production and capacity levels (േ1SD), scenario II – without a sort yard, (b) Average lumber production and capacity levels (േ1SD), scenario II – with a sort yard ............................................................ 72  Figure  3.15  Log production of the sort yard (േ1SD) for scenario II ................................. 73  Figure  4.1  (a) Average total harvest of sawmills (േ1SD) – Base case and scenarios I and II,  (b) Average total volume of imported logs – Base case and scenarios I and II ...................................................................................................................... 85  Figure  4.2  (a) Average total profits of sawmills (േ1SD) – Base case and scenarios I and II, (b) Average total production capacity of sawmills (േ1SD) – Base case and scenarios I and II ............................................................................................. 86  Figure  4.3  Average number of active agents in scenario II. Error bars show the observed maximum and minimum number of active agents in each time interval. ....... 86  Figure  4.4  (a) Average remaining volume of standing timber - Base case and scenarios I and II, (b) Average remaining value of standing timber - Base case and scenarios I and II ............................................................................................. 87  Figure  4.5  Average unit value of remaining timber - Base case and scenarios I and II ... 88  Figure  4.6  (a) Average total harvest of sawmills– Base case and Scenarios III and IV,  (b) Average total volume of imported logs – Base case and Scenarios III and IV ......................................................................................................................... 90  Figure  4.7  (a) Average total production capacity of sawmills – Base case and scenarios III and IV, (b) Average total profits of sawmills – Base case and scenarios III and IV.............................................................................................................. 90  Figure  4.8  Average unit value of remaining timber - Base case and scenarios III and IV ......................................................................................................................... 92   x Figure  4.9  (a) Average total production capacity (േ1SD) of sawmills – Base case and scenario V, (b) Average total profits (േ1SD) of sawmills - Base case and scenario V ....................................................................................................... 93  Figure  4.10  (a) Average total harvest (േ1SD) of sawmills - Base case and scenario V,  (b) Average total volume (േ1SD) of imported logs - Base case and scenario V . 94  Figure  4.11   (a) Average total production (േ1SD) of sawmills – Base case,  (b) Average total production (േ1SD) of sawmills – Scenario V ........................................ 94  Figure  4.12  Average unit value of remaining timber – Base case and scenario V ............ 95  Figure  5.1  (a) Average number of active agents, base case, (b) Average number of active agents, increased production costs. Error bars show minimum and maximum observed number of agents. .......................................................................... 103  Figure  5.2   (a) Average profits of supply chain members (േ1SD), base case, (b) Average profits of supply chain members (േ1SD), increased production costs ......... 104  Figure  5.3  (a) Average lumber production of sawmills (േ1SD), base case, (b) Average lumber production of sawmills (േ1SD), increased production costs ........... 104  Figure  5.4  (a) Average number of active agents, increased prices. Error bars show minimum and maximum observed number of agents, (b) Average lumber production of sawmills (േ1SD), increased prices ........................................ 105  Figure  5.5  Average profits of supply chain members (േ1SD), increased prices ........... 106  Figure  5.6  (a) Average log export ratio (േ1SD) of the sort yard for the base case and the case with lower log prices, (b) Average production capacity (േ1SD) of the sort yard for the base case and the case with lower log prices ..................... 107  Figure  5.7  Average production volume (േ1SD) of high value lumber for the base case and the case with reduced import limit ......................................................... 108  Figure  5.8  (a) Average sort yard profits (േ1SD) for the base case and the case with limited exports, (b) Average log output (േ1SD) of the sort yard for the base case and the case with limited exports .......................................................... 109  Figure A.1  (a) Average profits (േ1SD)of existing agents - scenario 1, (b) Average profit (േ1SD)of new agent, scenario 1 ................................................................... 129  Figure A.2  (a) Average profits (േ1SD) of existing agents - scenario 2, (b) Average profit (േ1SD) of new agent, scenario 2 .................................................................. 129  Figure A.33  (a) Average profits (േ1SD) of existing agents - scenario 3, (b) Average profit (േ1SD) of new agent, scenario 3 .................................................................. 130  Figure B.14  (a) Average Total harvest volume (േ1SD): scenario I - without a sort yard, (b) Average Total harvest volume (േ1SD): scenario I - with a sort yard .......... 131  Figure B.25  (a) Average Total harvest volume (േ1SD): scenario II - without a sort yard, (b) Average Total harvest volume (േ1SD): scenario II - with a sort yard ... 131  Figure B.36  (a) Average Total purchased logs from sort yard (േ1SD): scenario I, (b) Average Total purchased logs from sort yard (േ1SD): scenario II .............. 132   xi Figure C.17  (a) Average total harvest volume of sawmills (±1SD) – Base case, (b) Average total log import volume of sawmills (±1SD) – Base case .............. 133  Figure C.28  (a) Average total profit of sawmills (±1SD) – Base case, (b) Average total production capacity of sawmills (±1SD) – Base case ................................... 133  Figure C.39  (a) Average total harvest volume of sawmills (±1SD) - Scenario I, (b) Average total harvest volume of sawmills (±1SD) – Scenario II ................. 134  Figure C.410  (a) Average total harvest of sawmills (±1SD) - Scenario III, (b) Average total harvest of sawmills (±1SD), scenario IV ...................................................... 134  Figure C. 511 (a) Average total log import volume of sawmills (±1SD), scenario III, (b) Average total log import volume of sawmills (±1SD), scenario IV ............. 135  Figure C.612 (a) Average total production capacity of sawmills (±1SD), scenario three, (b) Average total production capacity of sawmills (±1SD), scenario four ......... 135  Figure C.713 (a) Average total profits of sawmills (±1SD) - scenario III,  (b) Average total profits of sawmills (±1SD) - scenario IV ...................................................... 136  Figure C.814  Average THLB area (±1SD) by age group for the base case ....................... 136  Figure C.915  (a) Average THLB area (±1SD) by age group for scenario I, (b) Average THLB area (±1SD) by age group for scenario II .......................................... 137  Figure C.1016 (a) Average THLB area (±1SD) by age group for scenario III, (b) Average THLB area (±1SD) by age group for scenario IV ........................................ 137  Figure C.1117 Average THLB area (±1SD) by age group for scenario V ........................... 138   xii List of Abbreviations  AAC Annual Allowable Cut ABM Agent-based Modelling BC British Columbia bft Board Feet C & I Criteria and Indicators DES Discrete Event Simulation FL Forest License LP Linear Programming m3 Cubic Meter mbf Thousand Board Feet MIP Mixed Integer Program OR Operations Research SCM Supply Chain Management SD System Dynamics SFM Sustainable Forest Management TFL Tree Farm License TSA Timber Supply Area    xiii Acknowledgments I would like to thank my research advisor, Dr. Thomas Maness, for guiding me throughout my program and for his intellectual and moral support when I faced challenges. This work would not have been possible without him. I would also like to express my gratitude to my supervisory committee, Dr. John Nelson and Dr. Farrokh Sassani, for their extremely helpful ideas and suggestions for improving the work presented in this dissertation. Additionally, I would like to thank the members of my examining committee - Dr. Robert Kozak, Dr. Gary Schajer, and Dr. Woody Chung - for critically reviewing and commenting on my thesis. I am thankful to other graduate students and post doctoral fellows in our research group for their support and encouragements throughout my program. I would like to especially thank Dr. Olaf Schwab who helped me immensely in understanding the fundamentals of the CAMBIUM model and Dr. Cristian Palma who motivated and inspired me when I most needed it. I am also extremely grateful to Mr. Catalin Ristea who always took the time to listen to my problems and offered his assistance when possible. I would also like to express my highest appreciation to the researchers at FPInnovations in Vancouver - Dr. Darrell Wong, Mr. Joel Mortyn, and Mr. Jack MacDonald – for kindly sharing their knowledge and views with me and for answering my endless questions. I could not have endured the pressure of the past few years, if it was not for the unconditional love and support of my family. My lovely parents, Sima and Abolghassem, believed in me when I lost faith and cheered me on when I needed motivation. I am forever indebted to them for all they gave me. My dearest siblings, Hamid and Sepideh, have always inspired me with their love and courage and I am eternally grateful to them for all they have done for me. I went through a long and difficult journey to complete my work and I cannot imagine having done it without the support of my loving partner and companion, Ario. He has given me love, peace, and comfort and I thank him for that. I am enormously thankful to my dearest friends, Nazly and Shora, who have always rooted for me. Thank you Nazly, for being my family away from home and thank you Shora, for being a source of hope and optimism for me.  xiv Dedication  ،مردپ و ردام هب ميدقت  دنا هتشاد رواب ارم هشيمھ هک.  To my parents, who have always believed in me.   1 Chapter 1: Introduction  Chapter 1. Introduction1  1.1  Forest Products Supply Chain The supply chain in the forest products sector is exceedingly complex. Although it includes private companies, it is impacted by a collection of entities that are outside of the supply chain of a typical business. These entities can include government agencies, non- governmental organizations, environmental groups, and community organizations. Outside influences are constantly changing the dynamics of the system. Additionally, the business cycle impacts the forest products sector in a profound way and creates a ripple effect through regional economies. Furthermore, forest policy and international trade agreements are constantly changing and creating uncertainty for investors.  Figure  1.1 A typical forest products supply chain  1 A modified version of this chapter has been previously published: Vahid, S. and Maness, T. (2010). Modelling customer demand in forest products industry supply chains: a review of the literature. International Journal of Simulation and Process Modelling. 6(2):103 – 114. Forest Pulp Mill Sawmill or  Panel Mill Log Sort Yard Export MarketLocal Market Pulp Based  Manufacturing Local Market Export Market Flow of Material/Products Secondary Wood  Products Manufacturing  Chapter 1: Introduction  2 A typical forest industry supply chain is shown in Figure  1.1. The flow starts in the forest where trees are harvested and the branches are removed. Usually, the next step is bucking: cutting trees into transportable length logs based on diameter, length, and quality. The logs are then transported to sawmills, pulp mills, or sort yards, presumably based on the highest valued application. Sort yards are intermediate storage places where long logs from the forest are further sorted based on their grade and sent to appropriate manufacturing processes. Transportation of logs can be by truck, ship, or by log boom (water), depending on the terrain condition and the proximity of roads and water. When logs are received by sawmills, they are sawn into final products (in a multitude of manufacturing steps). If necessary they are also kiln-dried to produce specific products. The market for sawmill products includes construction industries, secondary wood processing plants, and remanufacturing plants among others. Chip residues are produced as a by-product of sawmilling activities and are sold to pulp mills, composite panel plants, or bio-energy plants (Ronnqvist, 2003) . Each of these stages requires decisions that affect the outcome of future steps. For example, the bucking decisions made in the forest will affect the type of products that can be produced from the log. Therefore, it is important that all activities be planned and coordinated jointly (Haartveit & Fjeld, 2002). However, because of the nature of this industry, decisions of the upstream and downstream supply chain members are rarely integrated. Supply Chain Management (SCM) is a concept that can assist companies in achieving integrated planning and operations. SCM is the integrated planning of all business activities including purchasing, manufacturing, warehousing and transportation of raw materials and finished products. Effectively linking these activities can greatly decrease the overall cost of the supply chain by eliminating redundant inventories, increasing throughput and reducing waste within the supply chain (Moyaux et al., 2004; Singer & Donoso, 2007). The forest industry is a big part of the BC economy and a major source of employment in rural communities. The industry accounted for approximately 30% of the provincial exports in 2009 and provided direct and indirect jobs equal to 7% of total employment in 2008 (BC Ministry of Forests, 2010).  However, the industry is facing challenges; wood products manufacturing shipments of BC has decreased sharply during the past two decades, dropping by more than 60% since 1995 (Statistics Canada, 2011b). Employment has also been decreasing, with approximately 10,000 jobs lost since 1995 (Statistics Canada, 2011a). BC  Chapter 1: Introduction  3 forest companies have low average Return on Capital Employed (ROCE), with an average ROCE of 3.8% in 2010, when the cost of capital is approximately 10% to 12% (Hamilton, 2011). As a result, the industry has been having trouble attracting new capital.  This is problematic, since capital investments are one of the main drivers of technological progress. Lack of capital investments, particularly in the coastal BC forest sector, has caused the industry to lag behind other regions in Canada and elsewhere. Outdated manufacturing equipment results in high production costs, and since BC Coast already has higher log and labor costs compared to its competitors, the profitability of the forest industry in this region is further decreased. To remain competitive the BC forest sector needs to find new operating strategies, employing new management techniques in areas such as harvest scheduling, transportation, manufacturing and production planning that will improve the sector’s performance. Establishing new facilities such as log sort yards is one potential strategy that may be worth investigating. Log sort yards have been recommended for improving the utilization of smaller logs (Venn et al., 2009) and benefiting small wood manufacturing businesses (Sunderman, 2003). As previously mentioned, SCM concepts and techniques can be of benefit in this context by providing analytical tools and frameworks to investigate and compare different strategies and investment decisions. Decision support tools can be used to investigate how SCM could be applied to the BC Coast to make the process more efficient. Another source of problems for the forest industry in BC in recent years has been the decreasing availability of old-growth timber which has historically been abundant, especially on the coast (Pearse, 2001). Harvesting practices have favored old-growth timber in the past, resulting in aggressive harvesting of the most accessible high quality old-growth stands (Prudham, 2007; Wilson, 1998). Although this trend originally contributed to the growth of the forest industry, it has also caused a decline in availability of the most desired timber resources, which has resulted in higher timber harvesting costs. Although the harvesting policies on BC have been changed over the years to manage the forest resources in a more sustainable manner (BC Ministry of Forests, 2010), some argue more policy reforms are required to benefit the society as well as the industry in the long run (Burda et al., 1998; Wilson, 1998). Similar to production strategies of the industry, the impact of forest policy changes such as harvest restrictions or export regulations can also be investigated using SCM decision support tools that have been customized for use on the BC Coast.  Chapter 1: Introduction  4 1.2 Research Objectives The objective of this thesis is to provide a decision support tool that can be used to aid policy research regarding the supply chain of the forest products industry. Specifically, this decision support tool will be used to answer the following questions: 1. What would be the economic impact of establishing a new log sorting facility within the existing supply chain of the coastal BC forest industry? a. How does the sort yard impact the profits and operations of the existing facilities? b. How does the sort yard perform through time as the surrounding environment changes? c. If a sort yard is economically efficient, where should the new facility be best located from a set of potential locations? 2. How would a change in harvest policies impact the performance of the supply chain members? a. How would a change in allowable harvest levels impact the profits and operations of the supply chain members? b. How would a change in harvest priority (the order of harvesting forest stands) impact the profits and operations of the supply chain members? 3. How would a change in harvest policies impact the state of forest resources in the region? a. How would a change in allowable harvest levels impact the sustainability of the forest enterprise? b. How would a change in harvest priority (the order of harvesting forest stands) impact the sustainability of the forest enterprise? The remainder of this chapter provides a review of relevant supply chain literature with a focus on forest products industry supply chains. A review of literature on forest products supply chains is presented first, followed by research on facility location within supply chains, and finally previous policy analysis performed using supply chain models. Finally, the thesis contributions and structure are presented. 1.3 Supply Chain Models Supply chains are networks that connect the raw material sources to finished products consumers through manufacturing activities and distribution channels (Santoso et al., 2005;  Chapter 1: Introduction  5 Vila et al., 2006). There are many decisions to be made in a supply chain and modelling techniques can, directly or indirectly, help the decision makers by revealing the consequences of the proposed actions and strategies. The research literature on SCM is rapidly growing, offering different classifications of supply chain models. Depending on the operational level of the questions to be answered, the supply chain models are broken down into strategic, tactical or operational hierarchies. Strategic planning is at the highest level and is concerned with broad-scale decisions over long periods of time. Strategic planning is a process designed to give a firm a competitive advantage over competitors (Gunn, 2007). The strategic plan identifies the types of actions that need to be taken, but does not plan the implementation steps for those actions (Church, 2007). An example of a strategic analysis is deciding on which manufacturing facilities to establish in a production-distribution network. On the other end of the spectrum is operational planning, concerned with the day to day operations of the firm or supply chain, with time spans ranging from a day to a few weeks. For example, scheduling truck routes for transporting logs from specific harvest sites to specific destinations is an example of operational planning. Tactical models can provide a link between the two ends of the decision level spectrum. Tactical models translate the strategies into appropriate operational level targets (Church, 2007). These models ensure that the strategic goals are feasible at operational level. For example, harvest scheduling at the strategic level may identify a certain number of hectares from a certain age class that needs to be harvested on a land base. A tactical model is then used to provide more spatial detail about the specific stands that need to be harvested in a specific order. SCM models are also classified into centralized and decentralized models based on how decisions are made. In centralized supply chain models, all procurement, production, and distribution decisions are made by a central unit, considering the state of the entire system. This ensures a higher level of control and collaboration among all supply chain members and a globally optimum decision. Traditionally, many of the models in SCM literature have utilized centralized decision making. However, sometimes it is not realistic to assume that all decisions can be controlled centrally, especially if the supply chain members do not belong to the same organization. Each firm may aim to maximize its benefits without considering the impact on the whole system. Additionally, different firms may not be willing to share their  Chapter 1: Introduction  6 cost and price information with others. In such cases, decentralized models are more appropriate (Stadtler, 2005). Decentralized supply chain models allow individual supply chain members to make decisions based on their own goals, while still operating in the same environment that inevitably affects all members. This reflects the decision making process in many real world systems and simultaneously decreases the model complexity, particularly in the case of larger supply chains that may be very difficult to model with centralized modelling techniques. Finally, another approach to classifying supply chain models is based on the modelling approach and solution method. Under this classification scheme, supply chain models can be broadly categorized into optimization and simulation models. Optimization models use mathematical programming approaches to find a feasible and “optimal” solution to a supply chain problem such as designing a transportation network, or locating a new plant. Alternatively, simulation models allow the decision makers to see the performance of the supply chain over time under various scenarios and help them understand the inter- relationships between different model components (Shapiro, 2001). Optimization models are mostly centralized, while simulation models can more easily represent decentralized decision making. Simulation and optimization have also been combined for supply chain management in manufacturing industries. In fact, simulation based optimization has become a popular approach, mainly because of its ability to incorporate uncertainty into optimization problems (Fu, 2002; Jung et al., 2004; Mele et al., 2006). In this chapter, the latter classification scheme is used to review the relevant supply chain models. Considering the vast volume of available literature on supply chain models, only the studies focusing on forest industries are presented here. For information on supply chain modelling research in other contexts, the readers may refer to available reviews on the subject (Beamon, 1998; Chan & Chan, 2005; Min & Zhou, 2002; Stadtler, 2005; Thomas & Griffin, 1996). 1.4 Forest Products Supply Chain Models There have been studies that look at each of the operational areas in the forest industry separately, examining the effect of different management scenarios on the performance of individual companies as well as the entire sector in different regions and countries. In recent  Chapter 1: Introduction  7 years great emphasis has been placed on supply chain management as a result of consolidation of upstream and downstream companies in the forest industry. It is a common opinion in today’s forest industry that the supply chain can be improved as a whole if an analysis integrated all the different steps of the wood flow from the forest to the customer (Bredstrom et al., 2004; Ronnqvist, 2003). Although such an analysis would be extremely complicated, even a small improvement in efficiency could result in large financial gains, considering the large volume of wood in a supply chain. In a study on Quebec mills, for example, Moyaux et al. (2004) showed that by effectively managing all nodes in a supply chain, the overall cost can greatly decrease. Another study on the Chilean sawmill industry found that internal supply chain management would increase the profitability of the sawmills by approximately 15% (Singer & Donoso, 2007). 1.4.1 Optimization Models Optimization models identify potential improvements that can be made in a supply chain with regards to a certain performance measure (objective function), such as order fulfillment rate or total profits. Supply chain optimization models prescribe a plan for production and distribution activities of supply chain members that is “optimal”, meaning that no alternative plan can further improve the value of the objective function. In this category of supply chain models, one optimization problem (either deterministic or stochastic) is constructed based on all the constraints and variables of the problem. Interactions among different supply chain members must be translated into constraints in the model. As the size of this optimization problem grows, finding an exact optimal solution becomes a more difficult task and in many cases, approximation techniques and heuristics are needed to find a near-optimal solution. Optimization studies in forestry have mainly focused on individual areas such as harvest scheduling and forest planning (Borges et al., 1999; McDill et al., 2002; Weintraub et al., 1994), sawmill operations (Maness & Adams, 1991; Todoroki & Ronnqvist, 2002), and transportation (Ronnqvist & Ryan, 1995; Weintraub et al., 1995). In recent years however, modelling the entire supply chain has been receiving more attention. Some recent literature reviews in the field have been published (D'Amours et al., 2008; Ronnqvist, 2003; Weintraub & Romero, 2006). The different types of decisions that need to be made in a wood products supply chain have been discussed in length by Ronnqvist (2003).  Chapter 1: Introduction  8 Presenting an extended review of literature, he focused on how Operations Research (OR) and especially optimization can be used in decision support tools in forestry. The author claimed that developing a robust optimization tool is an important part of improving control over the wood-flow. In addition, Ronnqvist emphasized the importance of incorporating uncertainty and environmental issues in forest supply chain optimization tools. A recent review on the use of OR models in forestry and agriculture by Weintraub and Romero (2006) compared models based on problem areas, data, environmental issues, and the impact of OR models. They also recommended accounting for environmental concerns in forestry OR models. More recently, D’Amours et al. (2008) provided a review of strategic, tactical and operational problems in forest industry supply chains and how OR has been used to address such problems. They argued that there was a need for more research on integrating the forest management activities with the forest products supply chains. Combining Linear Programming (LP) and economic equilibrium theory, Gautier et al. (2000) built a model to foresee and explain economic trends facing lumber and paper products in Quebec. Their model consisted of individual LP sub-problems solved for each actor (seller or buyer), and a global master LP problem to find the equilibrium prices and quantities for wood chips. The model was later used by the Quebec ministry of natural resources as a guide in negotiations with the industry. Bredstrom et al. (2004) utilized Mixed-Integer Programming (MIP) to model the supply chain of a pulp manufacturing company in order to facilitate short term decision making. An efficient heuristic technique was proposed for solving the MIP problem and the results showed a lower total cost for the supply chain compared to using manually generated plans. Singer and Donoso (2007) modeled the internal process of sawmills in Chile using Linear Programming (LP). They also modeled the collaborations among sawmills and showed increased profits as a result of internal supply chain management. One limitation of their work however was that they assumed raw material (timber) supply was unlimited which is not realistic. Chauhan et al. (2009) modelled a wood products supply chain with multiple harvest sites (material sources) and multiple mills (demand points). Logs of different length and species were modelled as multiple products, and the resulting MIP problem aimed to minimize the  Chapter 1: Introduction  9 total cost of harvesting and transporting logs while meeting the demand of all mills. The authors compared the performance of branch-and-bound technique with their developed heuristic approach for solving the MIP problem and reported that their heuristic performed well for small scale problems. Although this model is useful in linking the resource characteristics to the production activities of the supply chain members (which is relevant to the purpose of this research), the resulting model is very large and difficult to solve even for a single period. Therefore, it would not be computationally feasible to use such a model formulation for optimizing the supply chain activities over multiple periods. Considering the relatively short planning horizon and the level of model details in the studies discussed above, they are suited for tactical and operational planning, as opposed to strategic analysis. Alternatively, the models presented in the following articles have a strategic focus. Troncoso and Garrido (2005) developed an integrated supply chain model using MIP to analyze the strategic issues of forest industry in Chile. Their objective function minimized the net present value of the total cost, including transportation costs, operation costs, and investment costs. The authors used the special structure of the MIP problem to decompose the model into three individual problems and reduced the solution time significantly. Gunnarsson, Ronnqvist and Carlsson (2006) developed an MIP model that combined facility location and ship routing problem and applied it to the case of a pulp mill in Sweden. The purpose of the model was to meet annual demand for pulp products while minimizing the distribution costs. A heuristic was developed to solve the resulting MIP problem and was shown to perform well within practical time limits. Vila et al. (2006), developed a mixed-integer programming model to study the logistics of divergent process2 supply chains in the lumber industry. They developed a generic model that took into account facility location, technology and capacity selection, upgrading or changing the technologies, temporary shutdowns of facilities, and international markets. They demonstrated the performance of the model by applying it to a lumber company in Quebec. Daugherty and Fried (2007) used mixed-integer programming to model a network of bio- mass energy production facilities. Their model jointly optimized the prescription of fuel  2 Divergent processes are those that convert one type of raw material into several products.  Chapter 1: Introduction  10 treatment for acres of land and determined the location and capacity of energy production facilities. These strategic models are useful for the case of vertically integrated companies that manage all supply chain members in a centralized manner. However, if the objective is to model independent firms that belong to the same supply chain, as is the case for this thesis, these centralized model structures are not sufficient. Additionally, sometimes it is desirable to include some uncertainty in the model to represent the supply chain more realistically. When uncertainty is included in supply chain models, deterministic optimization is no longer helpful. Stochastic programming is one way to address this challenge. Probability distributions are identified for unknown parameters and expected values are used in formulations instead of known values. Stochastic network design becomes more complicated and computationally cumbersome as the number of uncertain parameters grows. Modifications of Benders’ decomposition3 (Benders, 1962) have been previously used in a variety of contexts to solve large MIP problems that include stochastic parameters (Gutierrez et al., 1996; Mirhassani et al., 2000; Santoso et al., 2005). Hultqvist and Olsson (2004) included uncertainty of weather conditions in their model of roundwood procurement in a Swedish pulp and paper supply chain. They showed that the inclusion of uncertainty was beneficial for the supply chain, however the resulting stochastic program was too large to solve to optimality and the conclusions were based on near-optimal solutions. Vila et al. (2007) developed a framework for designing production-distribution networks using a two-stage stochastic programming model, with a case study on lumber industry in Eastern Canada. Their objective was to include the market forces into the design of the network and improve the competitive position of the company involved. Their results showed that the stochastic program could be solved efficiently for moderate size problems, but was much more difficult to solve to optimality when applied to large problems. Vila et al. (2009) combined the production-distribution network design approach of Vila et al. (2006) with the market framework of Vila et al. (2007) and applied to the case of the Eastern Canadian lumber industry. Instead of considering expected future demands with a low  3 Benders’ decomposition is a method of solving large optimization problems by separating variables into a master problem and a sub-problem.  Chapter 1: Introduction  11 probability of occurrence, the proposed stochastic programming model considered several future market environments and deployed the production-distribution network to capture profitable opportunities. The discussed stochastic optimization models may be more realistic compared to their deterministic counterparts, but they are significantly more difficult to solve and still have the assumption of centralized decision making which makes them unsuitable for the purpose of the research in this thesis. 1.4.2 Simulation Models Simulation models do not prescribe an “optimal” design for the supply chain; their utility is in understanding the dynamics of the supply chain and in determining the outcome of different scenarios. In an optimization model, all interdependencies of the supply chain members should be translated into a mathematical program and the resulting model may be very large and complex, especially when there are uncertainties present. Simulation models on the other hand can accommodate the variability in input data more readily (e.g. different log diameters in a sawmill) and are usually easier to comprehend by end-users compared to large optimization problems. Discrete Event Simulation In Discrete-Event Simulation (DES) models, the activities within the supply chain are represented through individual events that are carried out at separate points in time according to a schedule (Kleijnen, 2005; Lee et al., 2002). This type of simulation has traditionally been used for modelling supply chains (Terzi & Cavalieri, 2004). With regards to forest products industry, discrete-event simulation has most often been used to model sawmill operations. A review of early sawmill simulation models is presented in Randhawa et al. (1994). While many of these studies focused on individual stages of production and distribution (Mendoza et al., 1991; Randhawa et al., 1994), some included the entire supply chain (Beaudoin et al., 2007; Lonnstedt, 1986). Most of the recent work on supply chain simulation, however, has been done in the area of multi-agent simulation – as discussed later in this chapter - because of their inherent capacity to easily reflect the complexities of supply chain interactions.  Chapter 1: Introduction  12 Lonnstedt (1986), simulated the forest sector in Sweden to study the dynamics of cost competitiveness in the long term. Based on the results, he suggested policy changes to increase the investment in the industry, for instance lowering taxes or interest rate.  Mendoza et al. (1991) combined optimization and simulation to model a hardwood mill. Their main goal was to develop optimal yet feasible production schedules. First a Linear Programming (LP) problem was solved with the objective of maximizing profits. The resulting optimal log input mix was then fed into a process simulation model to calculate production times and to measure the performance of different machines. Randhawa et al. (1994) developed a discrete-event object-oriented simulation environment that could be used to model sawmills with various configurations.  Lin et al. (1995) studied the benefits of producing green dimension parts directly from hardwood logs by comparing four mill designs using simulation. Baesler et al. (2004) used simulation along with experimental design to identify bottlenecks and factors that affect productivity (number of logs per day) in a Chilean sawmill. Their results pointed to the potential for a 25% increase in production. Beaudoin et al. (2007) combined a deterministic MIP and Monte Carlo sampling methods to support tactical wood procurement decisions in a multi facility company. Their test case results showed that their proposed planning process achieved an average profitability increase of 8.8% compared with an approach based on a deterministic model using average parameter values. Similar to some of the studies presented here, this thesis combines simulation and optimization for the purpose of modelling the supply chain of the forest industry in BC. This approach allows for incorporating the uncertainties, without causing modelling complexities and computational problems. However, although the developed simulation model uses a discrete event simulation engine, it is different from the traditional DES models because it uses individual agents to represent the members of the supply chain. A review of agent-based models follows shortly. System Dynamics System Dynamics (SD) modelling is mainly used for simulating continuous systems (as opposed to discrete event simulation). An SD model is characterized by feedback mechanisms and information delays to help explain the behavior of complex systems. In SD, real-world systems are represented in terms of stock variables (e.g. profit, knowledge,  Chapter 1: Introduction  13 number of people), flow between these stock variables, and the information that determines the flow values. Interacting feedback loops link the stock and flow variables. The resulting model is a system of differential equations and its dynamic behavior is the result of the structure of these feedback loops and delays (Borshchev & Filippov, 2004). It can also be combined with OR techniques to model supply chains. SD approach has been refined and expanded to study supply chain dynamics (Sterman, 2000; Towill et al., 1992). Angerhofer and Angelides (2000) have reviewed the research on SD modelling in supply chain management. Based on their review, System Dynamics can be used in combination with different techniques to be applied in areas such as inventory management, demand amplification, and international supply chain design. SD has rarely been used in modelling forest industry supply chains. Schwarzbauer and Rametsteiner (2001) used SD to analyze the potential impact of Sustainable Forest Management (SFM) certification on forest products in the Western European forest sector. Their results showed only modest changes from SFM-certification in forest products markets. Haartveit and Fjeld (2002) developed the “wood supply game” based on the Sterman Beer Game4 (Sterman, 1984, 1989) as educational material for students in forest logistics courses. The game included four stages in the supply chain from the forest to the lumber or paper retailer. Demand of the end customer was decided based on a random draw and the game demonstrated the distortion of demand as it moved upstream through the supply chain (the bullwhip effect). Jones, Seville, and Meadows (2002) modelled the supply chain of the Northeastern US forest industry. Their main goal was to answer policy question on the economic and environmental sustainability of the lumber industry in that region. Their results showed the capacity of the lumber mills could potentially exceed the available timber resources and feedback mechanisms are required to ensure the sustainability of lumber mill operations. What is important to remember is that SD models are better suited for aggregate views of the system and policy questions at a strategic level. The modelled system is evolved as a result of equations that link stock and flow variables together and it is not always possible to identify  4 In a beer distribution game players are trying to minimize their total costs by managing inventories in the face of uncertain demand (a draw from a card deck). The results of the Sterman beer game showed oscillation and amplification of demands for upstream members (Sterman, 2000).  Chapter 1: Introduction  14 individual behaviors of people or firms. Consequently, SD was not considered suitable for the purpose of this research, since observing the operation of supply chain members under different scenarios is one objective of the research. Alternatively, Agent-Based Modelling (ABM) makes use of individual behavior and characteristics to create a system from the bottom-up. For a comparison between SD models and ABM, readers are referred to Parunak et al. (1998). Agent-based Models Agent-based models aim to investigate how the players within a supply chain interact under changeable policies and rules to create a stable state for all supply chain members, generally using game theory. ABM has attracted a great deal of attention during recent years for the purpose of decentralized planning. Each member of the supply chain can be considered as an agent who is autonomous or semi-autonomous. Agents have predefined characteristics, decisions rules, and objectives, based on which they interact with each other. Each agent tries to maximize its own utility, but has to do so in an environment where all other agents are present. The main advantages of multi-agent systems are their ability to model decentralized complex systems easily, offering increased flexibility without losing efficiency, and providing learning systems that improve over time with better decisions (Forget et al., 2008; Mele et al., 2006). In recent years, ABMs have been increasingly used in the area of supply chain management in a variety of manufacturing contexts (Gjerdrum et al., 2001; Kaihara, 2001, 2003; Lin et al., 1998; Swaminathan et al., 1998). Some studies have combined ABMs with optimization techniques to ensure the local optimality of results for each agent (Gjerdrum et al., 2001; Lee & Lau, 1999) or for the overall supply chain (Dudek & Stadtler, 2005; Mele et al., 2006). For recent reviews of ABM approaches in manufacturing and production planning, the reader is referred to Shen et al. (2006) and Monostori et al. (2006). In the context of forest industries, many recent studies have focused on multi-agent modelling of the supply chain, identifying many improvement opportunities in different geographical regions. Moyaux et al. (2004) used agent-based simulation to model the forest supply chain in Quebec. They used the “wood supply game” developed by Fjeld (2001) to study how local decisions made by companies affected the whole supply chain.  Each company (mills,  Chapter 1: Introduction  15 wholesalers, and retailers) was represented by an intelligent agent that had a specific behavior (ordering scheme) and also the option of collaborating with other agents in decision making. The results showed that the lowest cost of the supply chain was associated with highest collaboration of agents. Based on the results, collaboration and sharing the information was not only best for the supply chain as a whole, but it was also best for each individual company. Frayret et al. (2007) presented a generic architecture to implement distributed advanced planning and scheduling (APS) systems with simulation capabilities. APS systems provide companies with algorithms and models for planning their activities from raw material procurement to distribution (Stadtler, 2005). The APS system in this research utilized agent- based technology, operation research, and constraint programming. The objective of the developed platform was to provide an industrial advanced planning tool, and to study its dynamics and performance. The authors validated their model by applying it to the case of a lumber production company in Canada. The performance of this APS tool under different scenarios was further studied and validated by Lemieux et al. (2009). Forget et al. (2008) used multi-behavior agents for creating a supply chain model and applied it to the lumber industry supply chain. Instead of using agents that are either reactive (have a predefined action for every possible state of the environment) or deliberative (use their knowledge about the environment to make decisions, but have slow reaction times in dynamic environments), they proposed the use of a multi-behavior agents. In their later work, Forget et al (2009) extended their research by comparing the performance of single-behavior and adaptive (multi-behavior) agents under different business environments. They found that performance gains were possible if agents adjusted their behavior  in every situation (e.g. changed their scheduling strategy from fulfilling orders at the latest possible date to a strategy that fulfilled orders as early as possible) instead of using a single one over the entire horizon. Schwab et al. (2009) developed and implemented a strategic agent-based forest sector model (CAMBIUM)  for assessing the impact of demand and resource inventory changes on the structure and economic viability of the forest sector. In CAMBIUM, aggregate structure of the industry emerges as a result of individual company production decisions and stand-level  Chapter 1: Introduction  16 ecological processes. The utility of this model was tested by assessing the impacts of a market downturn in the US forest products market on forest industry structure and mountain pine beetle salvage harvesting in British Columbia, Canada. Simulation results indicated a significant medium-term timber supply shortage; reduced stumpage revenues; intensive cost competition among primary wood-products manufacturers; and a large number of insolvencies in the panel, lumber, and pulp sectors. Agent-based models are helpful tools for both strategic and operational level planning under uncertainty, because of their flexibility and the fact that they are less complicated compared to large centralized stochastic programming models. Instead of creating a complex system of equations or a large number of constraints, the interactions among agents can be developed as a small module that gets executed as many times as necessary. Therefore, according to the advantages of ABM and the strategic nature of the problem addressed in this thesis, ABM seems to be a suitable technique for the purpose of modelling the BC Coast’s forest products supply chain. While all of the discussed agent-based models included the interactions of supply chain members and their behavioral rules, only one model, CAMBIUM by Schwab et al. (2009), incorporated the “ecological” activities of the timber harvesting land base as well. The forest polygons grow throughout the simulation and regenerate after being harvested and the state of the forest changes accordingly. Modelling the growth of the forest allows the end-user to not only study the economic development of the forest sector through time, but also observe the sustainability impacts of such developments. CAMBIUM was also the only model structured to support strategic analysis, while the remaining models were more suitable for operational and tactical planning. 1.5 Facility Location in Supply Chain Models In one way or another, firms have always been faced with the problem of how certain facilities should be located relative to other facilities and their target customers. The term “facility” here includes everything ranging from schools and hospitals to factories and retail stores. Given that the initial setup of any facility incurs significant costs and also causes long term impacts, facility location decisions are mainly considered strategic. While facility  Chapter 1: Introduction  17 location is a well established research area within OR, facility location models have been only gradually proposed within the supply chain context (Melo et al., 2009). A general discrete5 facility location problem involves a set of spatially distributed customers and a set of facilities to serve customer demands. Based on distances, times or costs between customers and facilities, the main question usually is: which facilities should be opened and which customers should they service to minimize total cost? If prices of products are also considered, profit maximization can be used instead of cost minimization. There are many variations of the facility location problem. For example, they can include single or multiple facilities, be deterministic or stochastic, have single-period or multi-period decision horizons, or be competitive or standard. For further information, readers are referred to previous reviews on facility location problems (Daskin et al., 2005; Eiselt et al., 1993; Klose & Drexl, 2005; Louveaux, 1993; Melo et al., 2009; Owen & Daskin, 1998; Plastria, 2001; Snyder, 2006; Vidal & Goetschalckx, 1997). Vidal and Geotschalckx (1997) and Owen and Daskin (1998) reviewed the facility location models and their applications at the strategic level. Klose and Drexl (2005) provided a review of facility location model formulation and solution methods, including continuous, network, and mixed integer programming models. Louveaux (1993) and Snyder (2006) reviewed the literature with a focus on incorporating uncertainty in facility location models. Daskin et al. (2005) and Melo et al. (2009) specifically focused on the theory and applications of facility location models in supply chain management. Finally, Eiselt et al. (1993) and Plastria (2001) provided a review on “competitive facility location” models6. Like every other industry, forest industries are faced with important location choices such as where to place a new lumber mill, a log sort yard, or a pulp production facility. Church et al. (1998) presented a review on various locational issues in the area of forest management. A comprehensive review of all facility location research in forest management and forest products industry is beyond the scope of this work. The focus of this chapter is on models  5 Continuous facility location models usually have a macroeconomic nature, and are not common in supply chain design and modelling (Melo et al., 2009). In this chapter, only discrete facility location models are discussed. 6 In competitive facility location theory, it is assumed that the firm locating the facility has existing or future competition in the area. Such models allow the customers to select the facility they patronize; i.e. new facilities should compete with other facilities for market share (Drezner & Hamacher, 2002; Plastria, 2001).  Chapter 1: Introduction  18 that cover all the stages of the supply chain. Some of the supply chain optimization models that were mentioned in Section  1.4.1 addressed the facility location problem inherent in their structure  (Daugherty & Fried, 2007; Gunnarsson et al., 2006; Troncoso & Garrido, 2005; Vila et al., 2006, 2007). These studies integrated the location problem with the problem of designing and planning the rest of the supply chain, some focusing on minimizing total costs (Gunnarsson et al., 2006; Troncoso & Garrido, 2005), others aiming to maximize the profits or revenues (Daugherty & Fried, 2007; Vila et al., 2006, 2007). The advantage of these integrated models is that the location decision is optimized considering all stages of the supply chain versus only a single echelon. Since one objective of this research is to study the impact of a new facility on the supply chain of the BC Coast industry, a facility location problem must be formulated to find the optimal location of the new facility. However, as mentioned before, in order to reduce the complexity of the model and incorporate random factors, a combination of simulation and optimization is selected as the appropriate modelling approach. While the interactions of the supply chain members follow the structure of an ABM simulation, the facility location is integrated in the form of an optimization problem. 1.6 Policy Analysis Using Supply Chain Models As can be seen from the surveyed research literature, supply chain models have mainly been used to study production and distribution systems and to investigate the impact of various operational policies, such as inventory management and ordering policies (e.g. Moyaux et al. (2004)). However, it is possible to utilize integrated supply chain models for answering relevant public policy and regulation questions, e.g. the impact of trade regulations on the performance of a global supply chain (Cohen & Lee, 1989). Global supply chains are impacted strongly by national and international government regulations (interested readers are referred to Meixell and Gargeya (2005) for a review on global supply chain issues). Global forest sector models have been developed to investigate the impacts of factors such as economic growth, energy prices, trade regulations, transport costs, exchange rates, and environmental policies. Examples of such models are the Timber Assessment Market Model (Adams & Haynes, 1980), the Global Trade Model (GTM) (Kallio et al., 1987), the Global Forest Products Model (Buongiorno et al., 2003), and the  Chapter 1: Introduction  19 EFI-GTM model (Kallio et al., 2004). These models are constructed at the most aggregate level by representing competing economies (mostly based on individual or regional grouping of countries), without explicitly modelling the production processes of forest products manufacturers. Although they provide valuable information on how the forest sector in a region reacts to various external policy changes, these models cannot provide details on the performance of individual economic players which may be of interest. Supply chains that are operating locally may also be impacted by government policies. This is especially true in the case of natural resource industries such as forestry or mining that are usually dealing with publicly owned resources. For example, in recent years Sustainable Forest Management (SFM) practices have been promoted in different regions of the world through government regulations (Wijewardana, 2008). However, integrated supply chain models in a local context have seldom been used to investigate impacts of policy changes. Examples of such models can be seen in the literature related to green supply chains and reverse logistics, (Masui, 2005; Mitra & Webster, 2008). The number of such studies is limited, but growing. A review of green supply chain management research has been provided by Srivastava (2007). Considering the importance of regulatory forces in the forest products industry, it may be beneficial to pay more attention to integrated supply chain models that analyse the impact of such forces in the context of both global and local supply chain operations. This is one of the objectives of the supply chain model developed in this research. 1.7 Summary of Relevant Literature As discussed in previous sections, traditional mathematical programming models have been used to study forest industry supply chains in the past. However, the supply chain of forest products companies is not always entirely and centrally controlled by corporate headquarters. Instead, it is the result of a collection of “agents” interacting in a dynamic system making decisions according to rules, policies, and beliefs. Therefore, using an agent-based supply chain model to represent the decentralized nature of forest product industries provides benefits over traditional centralized modelling methods. A survey of literature also showed that agent-based models have been increasingly used in the context of forest industry supply  Chapter 1: Introduction  20 chain modelling in recent years. Therefore, ABM, combined with an optimization element, was selected as the modelling approach in this thesis. The main objectives of this thesis are i) to investigate the economic activities of supply chain members when a new facility becomes active, and ii) to monitor the state of available forest resources and investigate how they change according to different harvest trends and policies. Considering the reviewed ABM studies, the scope and structure of the model CAMBIUM (Schwab et al., 2009) was found to be the most suitable blueprint for achieving the objectives of this research. However, in order to address all the specific issues related to this research, CAMBIUM was improved and extended as described below. 1.8 Structure of This Thesis The contributions of this thesis to the current body of literature are based on four manuscripts that are published or will be submitted for publication: a review on relevant supply chain modelling literature that was presented in this chapter and three analysis papers, as described below. 1. New Facility Location in a Forest Products Supply Chain Model This manuscript presents the structure of CAMBIUM 2.0 agent-based model. Through a novel algorithm, an agent-based simulation is combined with a facility location optimization problem to establish a new facility within the existing supply chain. Other extensions and improvements in CAMBIUM 2.0 include modelling product and mill differentiations based on log quality, as well as incorporating log import and export activities and a new agent type (sort yard). The optimization problem formulation and the flow of simulation are presented in the manuscript. The results are discussed according to different levels of availability of information for the new (entering) agent and it is investigated if the final location decision changes when more information is available to the agent. 2. Impact of Establishing a Centralized Sort Yard in Coastal British Columbia The second manuscript studies the impact of establishing a new log sort yard on the performance of the existing facilities within the supply chain. Using CAMBIUM 2.0, the profits, harvests, imports, and export levels of the supply chain members are observed and compared before and after setting up a log sort yard. Different scenarios regarding  Chapter 1: Introduction  21 market prices are explored to observe the variations of the supply chain activities in a changing business environment. The feasibility of investing in a log sort yard is discussed considering the current state of resources and the cost structure of the industry. 3. Impact of Harvest Policy Changes on Sustainability The final manuscript uses the developed agent-based model to address the harvest policy change on the performance of the supply chain members. Various policy scenarios regarding the allowable harvest volumes and harvest priorities are analyzed and compared to the status quo. In addition to economic activity of supply chain members, the volume and value of the remaining trees in the forest is monitored under each policy scenario to determine their impact of the forest resources. The results are discussed considering the economic and environmental trade-offs among different scenarios. The rest of this thesis is structured as follows: chapter two explains the structure of CAMBIUM 2.0 and integrating it with facility location module, chapter three investigates the impact of a new log sort yard on a BC Coast’s forest products supply chain, chapter four provide a harvest policy analysis using the developed model, and chapter five presents the concluding remarks and directions for future research. Appendices are also provided for model validation and verification, and for providing additional results from each chapter.  22 Chapter 2: New Facility Location in a Forest Products Supply Chain Model Chapter 2. New Facility Location in a Forest Products Supply Chain 2.1 Introduction The forest industry has historically been an important part of the economy in British Columbia (BC) with access to one of the world’s richest forest regions with a natural stock of old growth timber known for its impressive size, quality and value (Pearse, 2001). However, the industry - especially on the BC Coast- has experienced major downsizing during the past 20 years. Figure  2.1 shows the value of manufactured goods in the BC sawmills and wood preservation sector, clearly demonstrating the financial loss that the industry has suffered during the past decade. 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 1995 1998 2001 2004 2007 2010 V al ue  (b ill io n $) Time (years) Figure  2.1 Value of manufacturing shipments for sawmills and wood preservation sector in BC, Source: Statistics Canada (2011b) With old growth timber becoming scarce, many sawmills on the BC Coast that had focused on large-diameter logs have been permanently closed. Figure  2.2(a) shows the decline in both  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  23 the total volume and the national share of BC Coast’s lumber production from 1995 to 2010 while Figure  2.2(b) shows the same declining trend in the sector’s employment. 0% 2% 4% 6% 8% 10% 12% 14% 0 1 2 3 4 5 6 7 8 9 1995 1998 2001 2004 2007 2010 S ha re  o f to ta l p ro du ct io n in  C an ad a (% ) V ol um e (m ill io n m 3) Time (years) Production Share in Canada 0% 1% 2% 3% 0 10 20 30 40 50 60 1995 1998 2001 2004 2007 2010 S ha re  o f to ta l e m pl oy m en t i n B C  (% ) pe rs on s (th ou sa nd s) Time (years) Persons Employed Share in BC (a) (b)  Figure  2.2 (a) BC Coast total sawn lumber production and its share of Canadian sawn lumber production, (b) Wood products manufacturing employment level and its share of total employment in BC, Source: Statistics Canada (2011a, 2011c) Problems of the industry have been the result of a combination of factors including high delivered log costs, changing resource characteristics, changes in product demand, economic downturns, rising value of the Canadian dollar, outdated manufacturing equipment, and limited access to some export markets as a result of tariff and non-tariff trade barriers. Additionally, BC is experiencing a “fall down” effect as a result of historical logging policies. The fall down effect is the decline in quality and availability of higher value timber as a result of “best first” harvesting activities (Pedersen, 2003). In BC, “Annual Allowable Cut” (AAC)7 is determined based on volume rather than value, different values of various classes of timber (old growth, second growth, and rotten or uneconomic timber) are not reflected in it (Burda et al., 1998; Luckert & Haley, 1995). Therefore, forest areas with higher quality timber get harvested first and lower yield harvest blocks are left for future periods; this effectively means that sustainable financial results cannot be achieved without either significant technological improvements or significant reductions in labor and raw materials cost.  7 AAC is the maximum amount of timber that may be harvested to ensure a sustainable yield from the forest.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  24 It is inevitable that the industry will shift towards using managed, second-growth forests (Pearse, 2001). However, this transition requires new forest management and harvest policies as well as innovative production and management strategies that do not solely rely on large diameter and higher quality of old growth timber. Investment is needed all along the supply chain to bring the sector back to profitability. One potential investment opportunity for the BC Coast is establishing a dedicated sorting facility in the region that can benefit the industry by directing the right log to the right facility and reduce misallocations and potential value loss. Other potential opportunities may be establishing bio-energy plants or investments in flexible manufacturing technologies for sawmills. In order to study the effects of these various configurations and investment decisions on a supply chain, descriptive and normative models are often used (Shapiro, 2001) and can be helpful in the case of the coastal BC industry. Descriptive models, such as demand forecasts and simulation, help decision makers better understand how different members and components of the supply chain interact with each other and with the external environment. Alternatively, normative models, such as mathematical programming, identify the norms to which the supply chain members should conform. Both categories of models can be constructed at the strategic level, meaning they may focus on high level decisions of the supply chain over extended time horizons, such as opening new facilities or establishing transportation networks. These strategic models help expand the knowledge of supply chain decision makers and allow them to better investigate and predict the outcome of their decisions. Accurate descriptive models, including simulation, are necessary but not sufficient to identify actions that improve the performance of the supply chain (Shapiro, 2001). However, they create a foundation for building normative models to prescribe an optimal set of decisions. On the other hand, normative (optimization) models may not always find a feasible solution to supply chain problems, especially when a large number of decision factors and operational constraints exist that can result in complex and computationally exhaustive optimization problems.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  25 An alternative to using either of these approaches alone is to combine them. The combination of these techniques can help modellers accommodate the details of supply chain operations without creating optimization problems that are too difficult – or impossible – to solve. In this chapter, the process of integrating optimization with an agent-based forest sector simulation model is demonstrated and the results are presented for a case study of the BC Coast industry. The integrated simulation model has a facility location optimization module that allows new potential agents to enter the simulation model at different points in time. Contrary to traditional centralized optimization models, the cost and price information available to these new agents for making the location decision is limited. In this study, the impact of various levels of information availability on the final location decision and predictions of the new agents is investigated. The rest of this chapter discusses the structure and mathematics of the model, followed by data descriptions, results, and discussions. Model validation process is described in Chapter 5 for interested readers. 2.2 Facility Location in CAMBIUM CAMBIUM is an agent-based simulation model developed by Schwab et al. (2009) to model the structural developments and evaluate the impact of natural disturbances in British Columbia’s forest products industry. In this model, each economic player in the supply chain (sawmills, pulp mills, etc) is represented as an agent. Under predefined rules, agents interact with each other and use the available forest resources in order to manufacture products and compete in the market. CAMBIUM is a discrete event simulation model; i.e. it consists of distinct actions executed at specified time steps. These actions are representations of the decisions and activities of agents in the model. At each time step of the simulation, corresponding to one year, agents perform planning, harvesting, manufacturing and distribution activities. They also choose an investment “strategy”. Current strategy choices are: investment in process innovation, investment in capacity expansion, or no investment (status quo). These three strategies have been shown to be sufficient for agent differentiation in the model (Schwab, 2008). Agents learn from their past experiences and adjust their strategy choices accordingly. Their learning behavior is modelled using the EWA-Lite learning algorithm (Ho et al., 2001). Each agent’s choices are  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  26 influenced by its responsiveness to the external changes, its risk aversion, its past experience of each strategy, choices made by its competitors, and the expected payoff of each strategy. The model has been programmed in Java, using the Recursive Porous Agent Simulation Toolkit for Java (Repast J) that has been especially developed for modelling agent-based systems (North et al., 2006).  CAMBIUM is a grid-based spatial model where grids (tiles) contain information on either forest resources, or active economic agents. For example, a forest tile includes information on geographical location (latitude, longitude and elevation), species and age distribution, and site quality. An agent tile includes information on location, agent type (sawmill, panel mill, etc.), learning parameters, and operational data such as costs, recovery factors and capacity. 2.2.1 CAMBIUM 2.0 While CAMBIUM successfully represents the developments in a typical forest sector, there are some aspects of the supply chain that it does not currently address. Specifically, five modification steps are made to improve the suitability of CAMBIUM for modelling supply chains, as discussed below. The extended model will henceforth be referred to as CAMBIUM 2.0. Product differentiation: In CAMBIUM 2.0, Instead of modelling lumber as one generic product, two major categories of lumber products are considered: specialty (high value) lumber products and commodity (low value) lumber products. The commodity lumber category itself includes two classes of products to represent the lowest value dimensional lumber and some higher grade commodity lumber that has slightly higher market value. The quality of lumber products is determined based on the specifications of the harvested logs. Consequently, harvested logs are also categorized into two major groups: high value logs required for specialty lumber production, and commodity grade logs required for manufacturing commodity lumber. The quality of the logs is estimated by the site index8 of the harvest block of their origin. It is assumed that all sites have a mix of high and low quality timber and the ratio of high quality timber increases as the site quality becomes  8 Site index is the most common measure of forest site productivity and forest growth. It is defined at the height of the largest diameter dominant or co-dominant tree at a specified age. The age or diameter specifications may differ in various regions but the goal is to provide a numeric description of site productivity that enables forest managers to predict forest stand growth and the yield of timber at harvest.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  27 better. Differentiating lumber products based on log quality generates a more realistic representation of the value of timber resources and final products. Sawmill differentiation: Modelling different lumber products allows for inclusion of different types of sawmills. In CAMBIUM 2.0 sawmills are divided into commodity and specialty mills that process different log grades to produce different lumber products. Specialty mills require high value logs to produce high value lumber while commodity mills can process both high and low quality logs to produce commodity lumber (log quality determines the grade of the commodity product). Specialty mills can also produce commodity lumber if they have an inventory of harvested low quality logs. Modelling different types of sawmills helps to identify the ones that could strive and grow under difficult economic conditions and resource limitations. New facility location: In order to support integrated supply chain design and management, it would be beneficial to extend CAMBIUM so that it includes new agent entry. CAMBIUM 2.0 has an integrated facility location module to locate new players in the model space among a set of potential sites. The inclusion of the facility location optimization problem with the rest of the simulation is shown in Figure  2.3. Section  2.2.2 describes the optimization problem formulation for locating a new facility in the supply chain. This problem is solved in each time step (i.e. every year) until the new facility is located, after which the facility location module is not executed anymore. Sort yard operation: A new type of agent, a sort yard, is defined and modelled in CAMBIUM 2.0. Dedicated sorting facilities allow the distribution of logs to appropriate processing facilities and have the potential to improve the profitability of the forest products supply chain. For example, high quality logs can be directed to mills capable of producing high value lumber products in order to maximize the extracted value from logs. Sorting can also be done at the landing, but sort yards offer some significant advantages with respect to the proximity to manufacturing and export facilities, scanning and optimization technology, and more space for storing sorted logs.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  28 Planning and strategy implementation Harvesting and log procurement Timber processing Market trading New Facility Location Simulation Optimization  Figure  2.3 CAMBIUM 2.0 flow of simulation in every time step Log import and export9: While in some regions importing and exporting logs may not account for a large portion of the forest industry’s business interactions, in other regions, including BC Coast, they play an important role in the industry (for example see: TimberWest Forest Corp. (2010) and Western Forest Products Inc. (2011)). Especially in cases where only a fraction of a sector is modelled, it is logical to assume that the local resources may not be sufficient to meet all the timber demand of the supply chain members. In CAMBIUM 2.0 the mills are allowed to purchase logs from external sources (up to a user defined limit) when the harvest blocks cannot provide all their required timber. Also, if any sort yards are present in the model, they are allowed to export logs to an external destination at the market price of logs. 2.2.2 Single Facility Location Optimization Problem The formal theory on facility location can be traced back to Alfred Weber (1929) who placed a single warehouse on a network to minimize the weighted sum of distances between the warehouse and its customers (Drezner & Hamacher, 2002; Owen & Daskin, 1998; Weber, 1929). Since then, the research literature on facility location has grown enormously in a variety of contexts such as locating hospitals, emergency service units (police stations, ambulances, etc.), manufacturing plants, and transportation hubs.  9 Importing and exporting is defined relative to the boundaries of the model, not national or political boundaries. Therefore, importing logs means purchasing logs from a source that is not part of the modelled system, and exporting log means selling logs to a demand point other than wood processing facilities in the model.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  29 Mathematical programming models have been used extensively for solving facility location problems. These models have especially become popular after the improvements in the computation power and speed of computers in the 1960’s. This is because even basic facility location models can become computationally complex for large problem instances (Drezner & Hamacher, 2002; Owen & Daskin, 1998). A basic location problem, with the objective of minimizing the cost of establishing a new facility and the total distance between the facility and the customers can be described as follows (Drezner & Hamacher, 2002). Model (I): JjIidhxycz n j m i ijijij n j jj     ,Min 1 11  ( 2.1) Iixts n j ij   1.. 1  ( 2.2) JjIiyx jij  ,0  ( 2.3)   JjIiy j  ,1,0  ( 2.4) JjIixij  ,0  ( 2.5) In equation ( 2.1) z is the objective function, representing total setup and transportation costs to be minimized. I and J are the sets of customers and potential facility locations, respectively. Fixed cost of establishing a new facility at location Jj  is shown with cj and hi is the demand of customer i.  The distance between customer i and facility j is shown with dij. yj is a binary variable indicating if a facility is established at location j or not. xij is the proportion of the demand of customer i served by facility j. Constraint ( 2.2) ensures that all of the demand of customer i is met. Constraint ( 2.3) guarantees that only facilities that are established can serve the customers. The last two constraints ensure non-negativity of xij and binary nature of yj. Model (I) is an uncapacitated location model, i.e. facilities are assumed to have unlimited capacity to meet all of the  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  30 demand. Modifying ( 2.3) to represent the capacity limits of each facility would transform model (I) into a capacitated location problem. Jjybhx jj Ii iij   0  ( 2.6) bj is the capacity of facility j, therefore equation ( 2.6) ensures that the facility cannot meet customer demand beyond its capacity. Other modifications possible to model (I) are adding a limit on total number of new facilities. A special case of this is when only one new facility can be located, in which case the problem is known as a “single facility location problem”. Although model (I) aims to minimize total costs, the same problem can be addressed while maximizing the total profits, by including revenues in the objective function. Although traditionally most facility location problems have been formulated to minimize costs (Current et al., 1990), in recent years there has been increasing research on incorporating profits in the objective function (Meng et al., 2009; Shen, 2006). Considering the success of most private and public companies is measured with their profitability, profit maximizing is an appropriate formulating approach for supply chain strategic decisions, such as facility location. A profit oriented objective aims to locate the new facility in a way that maximizes the combined revenues or net profits of all supply chain members. A profit maximizing objective is defined in this research for locating a new agent in a forest products supply chain. Model Assumptions Considering the available case study data, the facility location optimization problem used in this research is constructed based on a supply chain with two types of players: sawmills and sort yards. Sawmills can further be categorized into high-value lumber mills and commodity lumber mills. However, pulp and panel mills can be added to the current formulation with minor modifications. Available timber is also either of high value or commodity value, based on the conditions of the block it’s harvested from. High value mills only process high value logs and commodity mills only process commodity grade logs.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  31 Shortages are allowed in the formulation, i.e. mills and sort yards may not be able to meet all the demand. However, currently no penalty is applied to the unfulfilled demand. Parameters ACt allowable cut volume in period t Capj production capacity of facility j CCjt  amortization cost per unit of output for facility j in period t (outstanding loans to be  repaid)  Cj setup cost for facility j DCt total demand for commodity lumber in period t DHt total demand for high-value lumber in period t DPt  total demand for pulpwood in period t FCjt fixed cost for facility j in period t i interest rate I set of all harvest blocks (indexed by i) Invj inventory limit for facility j J set of all existing and potential facilities (indexed by j) J’ set of all potential facilities (indexed by j) K a subset of J where all members are sort yards (indexed by k) L  number of existing facilities LGt logging cost per unit in period t  LHCt  unit price of commodity lumber in period t LHPt  unit price of high-value lumber in period t LIt  lumber inventory holding cost per unit in period t  PPt  unit price of pulpwood in period t PRF average mill product recovery factor PS pulp share of harvested roundwood SCPt unit price of commodity grade logs in period t SHPt unit price of high value logs in period t SLPS pulp share of sorted logs from the sort yard SLXt export limit for logs in period t St stumpage cost per unit in period t  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  32 SYRF average sort yard recovery factor T decision time horizon (indexed by t) TCt transportation cost per unit of distance per unit of raw material in period t TIt  timber inventory holding cost per unit in period t  VCjt variable cost per unit of output for facility j in period t Vit timber volume in block i in period t Variables dcjt commodity value lumber demand assigned to facility j in period t dhjt high value lumber demand assigned to facility j in period t dpjt pulp log demand assigned to facility j in period t lcIjt  commodity lumber inventory of facility j at the end of period t ݈ܿ݌௝௧  commodity value lumber production of facility j in period t lhIjt  high value lumber inventory10 of facility j at the end of period t ݈݄݌௝௧  high value lumber production of facility j in period t  prjt operating profit11 of facility j in period t pwIjt  pulp log inventory of facility j at the end of period t  ݌ݓ݌௝௧  pulpwood production of facility j in period t rwcijt commodity grade log volume transported from harvest block i to facility j in period t rwhijt high grade log volume transported from harvest block i to facility j in period t ݏ݈ܿ௞௝௧ commodity grade logs transported from sort yard k to facility j in period t slcIkt  commodity grade saw log inventory of facility k at the end of period t  ݏ݈ܿܺ௞௧ commodity grade logs exported from sort yard k in period t ݏ݈݄௞௝௧ high grade logs transported from sort yard k to facility j in period t slhIkt  high value saw log inventory of facility k at the end of period t ݏ݈݄ܺ௞௧ high grade logs exported from sort yard k in period t     10 Since shortages (negative inventories) are allowed in the model, lumber and pulp log inventory variables are in fact unrestricted in sign or urs variables. No shortages for saw logs are considered since there is no explicit external demand for it. A common way to deal with an unrestricted in sign or urs variable is to substitute it with two non-negative variables (e.g. lhIjt = lhIrjt+  - lhIjt­ ). It can be shown that in the optimal solution, only one of the two substitute variables can have a value higher than zero (Winston & Venkataramanan, 2003), resulting in either a positive or a negative value for the original urs variable. 11 Operating profit may be positive or negative, and is considered a urs variable as well which can be substituted with two variables, as mentioned above.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  33 yj ܾ݅݊ܽݎݕ, ൜ 1  if facility ݆ is active    0  if facility ݆ is inactive    Objective Function The supply chain in this research is assumed to be decentralized, meaning that no central decision making entity exists that can enforce a specific flow of raw material and products between the forest, the agents, and the market. Therefore, the decision maker in charge of the new facility does not have the authority to force an “optimal” flow of timber and lumber products among agents that would maximize its own profits only. Consequently, maximizing the profits of only the new facility is not a suitable objective for this optimization problem. The objective of the facility location problem therefore is defined as maximization of the combined profits of all supply chain members. This would help the decision maker in expecting the flow of raw material and products when the profits of other agents are also included in the optimization. Max  z = NPV (operating profits – setup costs) Max z = ∑ ∑ ௣௥ೕ೟ሺଵା௜ሻ೟௧்ୀଵ௝א௃ െ ∑ ܥ௝ݕ௝௝א௃  ( 2.7) Since prjt is a urs variable, equation ( 2.7) can be re-written to include the non-negative substitute variables: Max z = ∑ ∑ ௣௥ೕ೟ శି௣௥ೕ೟ష ሺଵା௜ሻ೟௧்ୀଵ௝א௃ െ ∑ ܥ௝ݕ௝௝א௃  ( 2.8) Constraints Maximum total number of new facilities is one. ෍ݕ௝ ௝א௃ᇱ ൑ 1 ( 2.9) Binary variable for all existing facilities is equal to one, indicating they are active. ݕ௝ ൌ 1                                                ׊݆ א ܬ െ ܬԢ ( 2.10)  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  34 Total profit for a sawmill is equal to the sum of sales revenue minus loan repayments, fixed and variable production costs, harvesting and harvest transportation costs, log purchasing and transportation costs, and inventory holding costs ݌ݎ௝௧ା െ ݌ݎ௝௧ି ൌ ൫݀ ௝݄௧ െ ݈݄ܫ௝௧ି൯ܮܪ ௧ܲ ൅ ൫݀ ௝ܿ௧ െ ݈ܿܫ௝௧ି൯ܮܥ ௧ܲ ൅ሺ݀݌௝௧ െ ݌ݓܫ௝௧ିሻܲ ௧ܲ – ܨܥ௝௧. ݕ௝ – ܸܥ௝௧ሺ݈݄݌௝௧ ൅ ݈݄ ௝ܿ௧ሻ െ ܥܥ௝௧ · ݕ௝ െ ෍ ܶܥ௧൫ݎݓ݄௜௝௧ ൅ ݎݓܿ௜௝௧൯݀௜௝௜אூ  െ෍ ሺܮܩ௧ ൅  ܵ௧ሻ. ൫ݎݓ݄௜௝௧ ൅ ݎݓܿ௜௝௧൯௜אூ   െ෍ ܶܥ௧൫ݏ݈݄௞௝௧ ൅ ݎݓܿ௞௝௧൯݀௞௝௞א௄  െ෍ ܵܪ ௧ܲ. ݏ݈݄௞௝௧௞א௄ െ෍ ܵܥ ௧ܲ . ݏ݈ܿ௞௝௧௞א௄ െ ܮܫ௧ ሺ݈݄ܫ௝௧ ା ൅ ݈ܿܫ௝௧ାሻ െ ܶܫ௧. ݌ݓܫ௝௧ା ׊݆ א ܬ െ ܭ  ( 2.11) Total profit for a sort yard is calculated similar to sawmill profits, where no log purchasing and log transportation costs exist (i.e. sort yards do not buy logs from other sort yards, or from external sources). ݌ݎ௞௧ା െ ݌ݎ௞௧ି ൌ ෍ ܵܪ ௧ܲ. ݏ݈݄௞௝௧௝א௃ି௄ ൅෍ ܵܥ ௧ܲ. ݏ݈ܿ௞௝௧௝א௃ି௄ െ ܥܥ௞௧ · ݕ௞ ൅ሺ݀݌௞௧ െ ݌ݓܫ௞௧ିሻܲ ௧ܲ – ܨܥ௞௧. ݕ௞ – ܸܥ௞௧෍ ൫ ݏ݈݄௞௝௧ ൅ ݏ݈ܿ௞௝௧൯௝א௃ି௄ െ෍ ܶܥ௧ሺݎݓ݄௜௞௧ ൅ ݎݓܿ௜௞௧ሻ݀௜௞௜אூ െ෍ ሺܮܩ௧ ൅ ܵ௧ሻሺݎݓ݄௜௞௧ ൅ ݎݓܿ௜௞௧ሻ௜אூ   െܶܫ௧ ሺݏ݈݄ܫ௞௧ ൅ ݏ݈ܿܫ௞௧ሻ െ ܶܫ௧. ݌ݓܫ௞௧ା ׊݇ א ܭ  ( 2.12) Product demand assigned to individual facilities must be less than or equal to total product demand in each period. ෍ ݀ ௝݄௧௝א௃ି௄ ൑ ܦܪ௧                      ׊ݐ ൌ 1, . . , ܶ ( 2.13) ෍ ݀ ௝ܿ௧௝א௃ି௄ ൑ ܦܥ௧                       ׊ݐ ൌ 1, . . , ܶ ( 2.14) ෍ ݀݌௝௧௝א௃ ൑ ܦ ௧ܲ                           ׊ݐ ൌ 1, . . , ܶ ( 2.15)  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  35 Lumber production of sawmills must be less than or equal to their production capacity. ݈݄݌௝௧ ൅ ݈ܿ݌௝௧ ൑ ݕ௝ · ܥܽ݌௝              ׊݆ א ܬ െ ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.16) Volume of processed log in sort yards must be less than or equal to their production capacity. ∑ ሺݎݓ݄௜௞௧ ൅ ݎݓܿ௜௞௧ሻ௜אூ ݏݕܴܨ ൑ ݕ௞ · ܥܽ݌௞ ׊݇ א ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.17) Inventory levels must be less than or equal to inventory limits for sort yards and sawmills. There is no need to set a limit for pulp logs since pulp log volume is dependent on the volume of saw log and lumber products. ݈݄ܫ௝௧ା ൅ ݈ܿܫ௝௧ା ൑ ݕ௝. ܫ݊ݒ௝                   ׊݆ א ܬ െ ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.18) ݏ݈݄ܫ௞௧ ൅ ݏ݈ܿܫ௞௧ ൑ ݕ௞. ܫ݊ݒ௞            ׊݇ א ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.19) Lumber production is related to the incoming harvested and purchased logs and product recovery factors. ݈݄݌௝௧ ൌ ∑ ݎݓ݄௜௝௧ · ሺ1 െ ܲܵሻ௜אூ  ൅ ∑ ݏ݈݄௞௝௧ · ሺ1 െ ܵܮܲܵሻ௞א௄ ܴܲܨ ׊݆ א ܬ, ׊ݐ ൌ 1, . . , ܶ ( 2.20) ݈ܿ݌௝௧ ൌ ∑ ݎݓܿ௜௝௧ · ሺ1 െ ܲܵሻ௜אூ  ൅ ∑ ݏ݈ܿ௞௝௧ · ሺ1 െ ܵܮܲܵሻ௞א௄ ܴܲܨ ׊݆ א ܬ, ׊ݐ ൌ 1, . . , ܶ ( 2.21)  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  36 Lumber inventory, demand and production in each period are inter-related, as shown in the equation below. ݈݄݌௝௧ ൌ ݀ ௝݄௧ െ ݈݄ܫ௝ሺ௧ିଵሻା ൅ ݈݄ܫ௝௧ା െ ݈݄ܫ௝௧ି ׊݆ א ܬ, ׊ݐ ൌ 1, . . , ܶ ( 2.22) ݈ܿ݌௝௧ ൌ ݀ ௝ܿ௧ െ ݈ܿܫ௝ሺ௧ିଵሻା ൅ ݈ܿܫ௝௧ା െ ݈ܿܫ௝௧ି ׊݆ א ܬ, ׊ݐ ൌ 1, . . , ܶ ( 2.23) ෍ ݏ݈݄௞௝௧௝א௃ି௄ ൅ ݏ݈݄ܺ௞௧ ൌ ෍ ݎݓ݄௜௞௧ ܻܴܵܨ௜אூ ൅ ݏ݈݄ܫ௞ሺ௧ିଵሻ െ ݏ݈݄ܫ௞௧ ׊݇ א ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.24) ෍ ݏ݈ܿ௞௝௧௝א௃ି௄ ൅ ݏ݈ܿܺ௞௧ ൌ ෍ ݎݓܿ௜௞௧ ܻܴܵܨ௜אூ ൅ ݏ݈ܿܫ௞ሺ௧ିଵሻ െ ݏ݈ܿܫ௞௧ ׊݇ א ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.25) Pulp log volume is related to processing activities in both sawmills and sort yards. It is also interrelated with inventory levels and pulp log demand. ݌ݓ݌௝௧ ൌ ෍ ሺݎݓ݄௜௝௧ ൅ ݎݓܿ௜௝௧ሻ௜אூ . ܲܵ ൅෍ ሺݏ݈݄௞௝௧ ൅ ݏ݈ܿ௞௝௧ሻ. ܵܮܲܵ௞א௄ ׊݆ א ܬ െ ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.26) ݌ݓ݌௞௧ ൌ ෍ ሺݎݓ݄௜௞௧ ൅ ݎݓܿ௜௞௧ሻ௜אூ . ܲܵ ׊݇ א ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.27) ݌ݓ݌௝௧ ൌ ݀݌௝௧ െ ݌ݓܫ௝ሺ௧ିଵሻା ൅ ݌ݓܫ௝௧ା െ ݌ݓܫ௝௧ି ׊݆ א ܬ, ׊ݐ ൌ 1, . . , ܶ ( 2.28) If a facility is not active, no harvested or purchased logs should be directed to it. The Big M formulation method is used to represent this logical condition. ෍ ሺݎݓ݄௜௝௧ ൅ ݎݓܿ௜௝௧ሻ ൅෍ ሺݏ݈݄௞௝௧ ൅ ݏ݈ܿ௞௝௧ሻ௞א௄ ൑ ݕ௝ · ܯ௜אூ   ׊݆ א ܬ െ ܭ, ׊ݐ ൌ 1, . . , ܶ ( 2.29) ෍ ሺݎݓ݄௜௞௧ ൅ ݎݓܿ௜௞௧ሻ ൑ ݕ௞ · ܯ ׊݆ א ܭ, ׊ݐ ൌ 1, . . , ܶ௜אூ   ( 2.30)  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  37 Available timber volume in harvest blocks is updated in every period according to harvest levels. Also, harvest levels cannot exceed the available timber volume in a block. Periodic allowable cut restrictions must also be considered. ෍ ሺݎݓ݄௜௝௧ ൅ ݎݓܿ௜௝௧ሻ ൑ ௜ܸ௧௝א௃     ׊݅ א ܫ, ׊ݐ ൌ 1, . . , ܶ ( 2.31) ௜ܸ௧ ൌ ௜ܸሺ௧ିଵሻ െ෍ ሺݎݓ݄௜௞௧ ൅ ݎݓܿ௜௞௧ሻ ׊݅ א ܫ, ׊ݐ ൌ 1, . . , ܶ௝א௃   ( 2.32) ෍ ሺݎݓ݄௜௝௧ ൅ ݎݓܿ௜௝௧ሻ ൑ ܣܥ௧௝א௃    ׊݅ א ܫ, ׊ݐ ൌ 1, . . , ܶ ( 2.33) All variables are continuous and non-negative, except for yj which is a binary variable. 2.2.3 CAMBIUM 2.0 Simulation Flow Figure  2.4 shows a detailed flow of simulation in CAMBIUM 2.0. The space of the model is constructed using 225 hectare tiles (1.5 by 1.5 km squares) which represent either the forest area or the agents. At the start of the simulation, using forest cover datasets and growth and yield curves, forest tiles are generated. Each forest tile belongs to an administrative unit which enforces a maximum allowable harvest limit. Additionally, each forest tile has an associated growth and yield curve which is used in future time steps to grow its available timber volume. These curves should be selected based on the geographical area to be modelled, and can be obtained from government units or private organizations that own and manage the forests. For BC, growth and yield curves are obtained from the BC Ministry of Forests, Lands and Natural Resource Operations (MoF), as explained in the next section. Agent tiles are generated at the same time as forest tiles using location and operational information of existing wood processing facilities. Once the space of the model is generated, a set of actions are performed sequentially at each time step of the simulation. These actions are performed for all active agents.   Chapter 2: New Facility Location in a Forest Products Supply Chain Model  38   Figure  2.4 Flow of Simulation in CAMBIUM 2.0 Generating tiles of the model space and forest inventory Generating active and potential agents and placing active agents on tiles Production planning for current period Strategy selection Bidding for stumpage and saw log price Purchasing saw logs Is it profitable to establish new facility? Solve facility location problem using data for potential facilities Yes Move to the next time period No Place new facility in model space Are there any financially feasible plans? Idling/Removing current agent Moving to next agent Yes No Data flow Program Flow Growth and yield data External market prices Information for existing and potential facilities Timber supply review data Forest inventory and GIS dataset Are there any existing sort yards? Is sort yard selected as roundwood source? Yes Yes Harvesting No No Product Pricing, trading and capital recovery Yes Is roundwood demand fulfilled? No Processing roundwood and manufacturing final product  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  39 1) Each agent calculates the production target for current period based on forecasted demand and prices. The other deciding factors in setting a production target are fixed and variable production costs, available capital, and production capacity. A production target should be financially feasible over the equipment lifetime of the agent. Otherwise, agents adjust the target by decreasing it in increments, until a financially feasible production target is achieved. If no financially feasible production target can be found, the agent becomes idle for the current period. If an agent is idle for three consecutive periods, it becomes inactive and is removed from the simulation. 2) Each agent selects one of the three investment strategies (capacity expansion, process innovation, or sustainment) based on its available capital, past experience, and expected payoff of the strategy (calculated as the net present value of profits for each strategy over an infinite time horizon). At this point, the EWA-Lite algorithm is used for strategy selection (Ho et al., 2001; Schwab, 2008; Schwab et al., 2009). A response function is used determine a selection probability for each strategy, based on the agent’s sensitivity to the change in the business environment and the calculated expected payoff. Once these probabilities are calculated, a random draw determines which strategy is selected12. Figure  2.5 shows a simple schematic of how strategies are selected. Expected strategy payoffs Past experiences Change in business environment Response Function Selection probability Random draw Strategy selection  Figure  2.5 Strategy selection schematic  12 The strategies are sorted in ascending order of their selection probabilities. A cumulative probability is calculated for each strategy as the sum of its own selection probability and those of the strategies with lower selection probabilities (the last strategy would have a cumulative probability equal to one).  A random number from the uniform distribution U(0,1) is drawn. Starting from the first strategy, this random number is compared with the cumulative selection probabilities. The first strategy with a cumulative probability value higher than the random number is selected.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  40 All agents have a randomly selected “success rate”: a number drawn from a uniform distribution between zero and one which is selected at the beginning of the simulation and remains unchanged. The success rate determines the probability of successful implementation of the “process innovation” strategy for each agent. Process innovation allows agents to improve their lumber recovery factors or decrease their variable production costs by investing in research and new technology. The success or failure of agents in improving their production process is determined by a random draw (from a uniform distribution between zero and one) and comparing the random number with the success rate, similar to the process of strategy selection. If the random number is smaller than the success rate value, the strategy is implemented successfully. 3) In order to set the stumpage price for each period, a bidding system is developed in which agents calculate the maximum stumpage price they can afford to pay based on their harvest target and market prices for the final product. After all agents submit their bids, a weighted average of the highest bid value over the latest five periods (years) is set as the stumpage price, with the more recent periods having higher weights. Five years is selected as a time range that provides a reasonable measure to reflect the changes in the stumpage price without the extreme rise and fall patterns. In special cases, such as the case for the BC Coast, if this stumpage price is higher than a maximum limit, that limit is set as the stumpage price. 4) If any sort yard agents are active, similar to the stumpage bidding system, sawmill agents calculate the maximum they can afford to pay for logs from the sort yard. A weighted average of the highest bids over the latest five periods is set as the price for logs. There is an allowed variation limit of this set price from the externally defined market price for logs. If the set price after the bidding is over or under the variation limit, then that limit is set as the price for logs. 5) If any sort yards are active, sawmill agents compare the total cost of harvesting their required timber with the total cost of purchasing it from the sort yard and select the lower cost option. 6) Agents proceed to harvesting timber from harvest units or purchasing logs from active sort yards. The order in which the mills access the harvest blocks is random in each period so that no one mill has an advantage by accessing the closest forest tiles. Specialty mills do not  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  41 harvest the blocks with “poor” site index since these blocks have a low ratio of high to low quality logs. However, when a mill harvests a block, all the available timber on that block must be harvested, regardless of the quality of the timber. Therefore, even harvesting blocks with “medium” or “good” site index results in an inventory of low quality logs for the specialty mills.  In each period, the specialty mill that has low quality logs may produce a limited volume of commodity lumber if it has unused production capacity or if its inventory of low value logs reaches a user defined limit. Table  2.1 Log quality ratio by site index Site Index High Grade Log Share Low Grade Log Share Poor 0.2 0.8 Medium 0.6 0.4 Good 0.9 0.1 The recovery factors are used to convert incoming timber into outgoing product, whether it is sorted logs or lumber. At this point, stumpage, production and transportation costs are deducted from each agent’s working capital. If an agent cannot receive all the required timber to meet its production demand, it then proceeds to “import” logs from an external source (e.g. other local forest management companies or log markets). Total log import of agents in each period cannot exceed a specified limit. 7) Agents calculate a reserve price for their products based on production costs and expected market prices to cover their costs and gain an agent-specific minimum profit. The products are bundled into “trading units”, defined by a product group, volume offered, an owner, a reserve price, and an expiry date. When all agents submit their trading units, the units are sold at the realized price of the market. Currently a linear relationship between quantity and price is used to decide on the market price. However, in cases where the modelled system is not an entire sector in a geographical area, the assumption is that agents’ production levels cannot significantly influence the market price of log or lumber products. Therefore, a variation limit is set within which the prices can fluctuate around the exogenously determined market price. If the reserve price of the trading unit is higher than the realized market price, only a fraction of its volume is  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  42 traded, resulting in a loss for the unit owner. The sales revenue for successful trades is added to the working capital of each agent. Unsold inventories will be returned to the agent and are offered at a user-defined discount rate the next period. Trading units are discarded if they pass their expiry date.  8) In each time step of the model, if there are “potential facilities” to be located, a facility location problem is solved using the formulation in Section  2.2.2. This is because the optimal solution would be different at each time step because of changing resource availability, changing market prices, and possibility of other agents becoming insolvent as the simulation progresses. Therefore, locating a new facility may not generate optimal overall profits in the first few time steps but as the model environment changes, it may become optimal to establish the new facility in one of the candidate sites. When the optimal solution indicates that a new facility is selected, a final test is performed to determine if the net present value of the profits for the new facility is positive in the optimal solution. If this is the case, the facility is entered into the simulation, otherwise the simulation continues and the facility location problem keeps being solved in the following periods. This final test is necessary because the objective of the optimization problem is to maximize the combined profits of all agents which could potentially result in a loss for the new facility at the expense of other agents profits. After the agent is added to the simulation, it will participate in all the above mentioned actions along with other agents. If only one new facility is to enter the model, the optimization problem will not be solved in time steps after the facility is located. However, it is also possible to locate several facilities, in a predetermined order. The optimization problem is solved in each period for the first candidate facility, based on a set of potential sites. Once the first facility is located, it will be removed from the list of candidate facilities along with all the related potential sites. The model then proceeds to the next candidate facility and its set of potential sites and the optimization problem will be solved using the new information until the second facility is located as well. This process continues until all candidate facilities are located or the simulation reaches its final time period. The information for existing facilities and current costs and prices are updated in each period and fed into the problem.  The optimization is a Mixed Integer Program (MIP) that is solved  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  43 using Gurobi Optimizer Java library, version 3.0.3 (2011). This library can easily be integrated combined with CAMBIUM 2.0 model structure. Gurobi Optimizer uses branch and bound solution approach to solve MIPs. 2.3 Case Study: British Columbia’s Coastal Primary Forest Products Industry A case study is carried out in order to investigate the difference in outcome based on perfect versus imperfect information of the decision maker who is locating a new facility in the supply chain. CAMBIUM 2.0 is applied to the case of a forest products supply chain on Vancouver Island, consisting of eight sawmills focusing on different lumber products (two commodity sawmills and six high value sawmills). The new facility to be located is a sort yard. Six Tree Farm Licenses13 (TFLs) are assigned to these sawmills and the sawmills can also purchase logs from additional sources to meet their unfulfilled timber demand. The forest area and sawmill locations included in the study are shown in Figure  2.6, as well as the potential locations for the new facility.  13 Tree farm licenses (TFLs) are forest tenures that provide companies with long-term rights to utilize Crown timber. These licenses convey the right to harvest an allowable annual cut (AAC) from the area, along with strategic and operational management responsibilities including protection, maintaining resource inventories, road building, and reforestation (BC Ministry of Forests, 1997).  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  44 1 Existing sawmills Potential sort yard locations Forest tiles (TFLs) 2 3  Figure  2.6 Forest region and sawmills included in the case study of the coastal BC industry 2.3.1 Forest Inventory Data Total number of tiles for representing the Vancouver Island area is just under 15,000, of which approximately 4,500 are forested tiles. Detailed forest inventory data required for constructing the forest tiles in the model are not publicly available for the six TFLs under study. Therefore, an inventory dispersion algorithm developed by Schwab and Maness (2010) is used to rebuild the forest inventory from publicly available aggregate data for each TFL. Unassigned tiles are fed into the algorithm along with aggregate data. The algorithm uses the geographical information of the tiles (latitude, longitude, and elevation) to populate them with data (timber species and other attributes such as site index, and age class). This ensures that the total assigned areas match the known attributes from the latest forest inventory projections. The aggregate data used as input to the inventory dispersion algorithm are: total timber harvesting land base and non-timber harvesting land base area of TFL and area breakdown by species, age class, site index, and Biogeoclimatic Ecosystem Classification (BEC) zones. In cases where all or some of this information is not available for a TFL, the information  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  45 from an adjacent Timber Supply Area14 (TSA) is used, assuming that the resource characteristics are similar for these adjacent areas. The reports containing aggregate data for TFLs and TSAs are available from the BC Ministry of Forests, Lands and Natural Resource Operations (2011b). Yield curves used for calculating and growing the available timber of each tile in each period are generated using the Variable Density Yield Prediction (VDYP 7) model from the BC Ministry of Forests, Lands and Natural Resource Operations (The Blue Denim Consulting Group, 2007). VDYP 7 is a stand level, empirical growth and yield model that is suitable for predicting the yield for stands with single or multiple species. This growth and yield prediction model was selected based on the decision flowchart provided by the BC MoF (BC Ministry of Forests, 2002). In each period, the yield curves determine the timber volume per hectare (m3/ha) of each forest tile based on the species composition, the site index and the age of the stand. The most recent AAC levels for the TFLs (as of March 2011) are obtained from the BC MoF (BC Ministry of Forests, 2011a). 2.3.2 Operational Data for Manufacturing Facilities Operating costs and recovery factors used for different types of mills are listed in Table  2.2, while cost assumptions related to harvesting activities are shown in Table  2.3. These costs are selected based on expert opinions (Mortyn, J., personal communication, 25 May 2011), annual reports of the represented forest company15 (Western Forest Products Inc., 2010), and previous research on sawmills in Canada and the US (Gibson et al., 2009; International Wood Markets Group, 2007). Lumber recovery factor values for each mill in each simulation run is randomly drawn from the uniform distributions shown in Table  2.2. The values for the distributions are based on expert opinion and the work published by the US Department of Agriculture (Spelter & Alderman, 2005). The sawmill equipment has a lifespan of 20 years after which the sawmill  14 A Timber Supply Area (TSA) is an area of the province created by the BC MoF for the purpose of analysis, planning, and management of timber resources (BC Ministry of Forests, 1997). 15 None of the cost and recovery values used in this work has been directly provided by Western Forest products Inc. or other forest companies in BC. Therefore - while the authors believe the costs and recovery assumptions closely represent the operating conditions of the industry - these values should only be considered as our estimates based on data available from public sources and previously published research.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  46 agent must replace it and pay for the replacement in installments during the following periods. The same rule applies in the case of capacity expansions where agents repay the investment amount in installments. Table  2.2 Cost, recovery, and other operating assumptions for BC Coast sawmills  High Value Mill Commodity Mill Manufacturing cost ($/mbf16 output) 260 200 Fixed cost ($/mbf capacity) 40 40 Recovery factor (bft/m3) Uniform(205,267) Uniform (227,304) Equipment lifespan (year) 20 20 Loan  repayment period (year) 5 5 When the mills harvest timber, 20% of the harvested logs will be pulp logs (pulp share). However, if they purchase logs from the sort yard the pulp share is reduced to 1%, since the sort yard has already sorted out the pulp quality logs. Sawmills are also allowed to “import” logs – buy them from sources not included in the model – up to a maximum of 20% of the period’s total harvest volume. The pulp share of the “imported” logs is 1%, the same as the pulp share of logs from the sort yard. The mills pay the market price of logs to purchase logs from sort yard or external sources, as well as the transportation cost from the purchase point to the mill. In the case of external log sources with no specific location in the model, the transportation distance is assumed to be the same as the most recent forest tile the mill harvested. Table  2.3 Harvesting and transportation cost assumptions for the BC Coast  Cost Logging (tree to truck) ($/m3) 40 Maximum stumpage ($/m3)  5 Transportation (landing to mill) ($/m3.km)  0.15 Any active sort yards in the system also pays stumpage, logging, and transportation costs (from the forest to the sort yard) indicated in Table  2.3. Other costs for the candidate sort yard that is to be located are shown in Table  2.4. The values were selected based on the  16 mbf (thousand board feet or 1000 bft) is a specialized measurement unit for the volume of lumber. One thousand board feet is approximately 2.36 cubic meters (m3).  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  47 USDA reports on sort yard operations (Dramm et al., 2004; Dramm et al., 2002) and the analysis published by the BC MoF (Sterling Wood Group Inc., 2002). These values are used for all candidate sort yards (from which only one can be established), regardless of their geographical location. The recovery factor value is based on the assumption that since sort yard can identify and group logs for various sawing purposes, the ratio of saw logs increases to 85%17 of the harvested timber (vs. 80% for sawmills). Table  2.4 Cost, recovery, and other operating assumptions for BC Coast sort yards  Sort Yard Variable sorting and log processing cost ($/m3output) 15 Fixed cost ($/m3 capacity) 3 Recovery factor (m3sorted log/ m3 harvested timber) 0.85 Equipment lifespan (year) 7 Loan  repayment period (year) 5 Initial prices for final lumber products are shown in Table  2.5. The values are selected based on published reports and surveys for lumber and log prices in BC (BC Ministry of Forests, 2011c, 2011d; International Wood Markets Group, 2007; Random Lengths, 2011). Table  2.5 Log and lumber prices for BC Coast at the beginning of the simulation  Price ($/unit) Specialty lumber products ($/mbf) 700 Commodity lumber –high grade ($/mbf) 325 Commodity lumber – low grade ($/mbf) 250 Specialty grade log ($/m3) 130 Commodity grade log ($/m3) 60 Pulp log ($/m3) 35 In order to study the effect of market price changes on the location decision of the new agent, market prices change for all products during the simulation according to graph in Figure  2.7.  17 This is different than the pulp share of the outgoing logs from the sort yard (1%). It shows how much of the harvested logs by the sort yard are classified as pulp logs and are not sold to the mills.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  48 The price changes are selected to include both increase and decrease in prices over the decision time horizon of the new agent (20 years). 40% 50% 60% 70% 80% 90% 100% 110% 120% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 C ha ng e in  m ar ke t p ric es  o f pr od uc ts  (% ) Time (years) Figure  2.7 Percentage of change in market prices for log and lumber products There are two capacity options and three candidate locations for the new sort yard (six options in total). The capacity options are the output volume of 150,000 and 250,000 m3 per year. Locations are selected based on their proximity to water, harvest blocks, and sawmills, as shown in Figure  2.6. 2.3.3 Scenarios Three scenarios are defined to represent different levels of information available to the new agent for making the location decision. Although the costs and prices are identical in all three scenarios; the perception of the new agent about those costs and prices is different. In the first scenario the new agent is assumed to have limited knowledge of the production costs and recovery factors of the existing agents and uses estimates for making the facility location decision. This is the case where only publicly available information is used along with estimates of the private information of other agents. In this scenario, the new agent underestimates the operational costs of others. Additionally, no change in market prices is expected by the agent. The second scenario represents the case where the new agent has access to the private cost and recovery factors of all other agents, but does not expect any change in market prices of the products. Finally, the third scenario represents a case where the new agent has perfect information about other agents and can correctly predict market  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  49 price changes. The scenarios are designed so the value of additional information in each scenario can be investigated by observing the results. There are several random factors in the simulation including initial values of agent recovery factors, order in which the agents access the forest tiles in each period, investment strategy selection, and the success rate of agents in implementing the “process innovation” strategy. Therefore, in order to capture the effect of these random factors, the simulation is performed 50 times for each scenario and the averages of variables are reported. 2.4 Results & Discussion The model runs at an average speed of approximately 3 simulation years/second, with a single run of 150 years completed in just under a minute.  However, time steps in which the optimization problem is solved take longer to complete. The optimization problem for the case study results in a final problem matrix with approximately 65,000 columns and 2,000 rows (varies among different time steps according to the model environment and active agents). The optimization problem takes less than 20 seconds to solve each time. 2.4.1 Predicted Profits Based on the optimal solution to the facility location problem, a new sort yard is established in all of the simulation runs within the first decade. The optimal location of the sort yard in all three scenarios is “location 3” (see Figure  2.6) - close to harvest blocks - and 250,000 m3 is selected as the optimal annual production capacity. All specialty mills operate at maximum capacity levels while commodity mills - if present in the formulation - are set to be idle (in all simulation runs the commodity mills exit the simulation within the first three years due to cash flow problems. Therefore, if the facility location problem is being solved after they exit, commodity mills won’t be included in the formulation). It should be noted here that the link between the harvesting of high quality and low quality logs is not represented in the optimization problem, as can be seen from the formulation described previously. Therefore, the optimization problem assumes that high quality logs may be harvested without any accompanying low quality logs. However, when the sort yard is established and becomes active in the simulation, the link between high and low quality logs is in effect and results in a mixed harvest for agents. These low quality logs may be processed into commodity lumber by the mills or exported by the sort yard.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  50 The values for the profit of the new agent and the existing agents in the optimal solution are shown in Figure  2.8 for all three scenarios. These values are the results of the facility location problem only and can be viewed as the “predictions” of the new agent about its own profits and the profits of others during the decision time horizon. For better readability, standard deviations are not shown in Figure  2.8. Instead, separate graphs for each scenario showing the standard deviation are presented in Appendix A. A slight decreasing trend in the “predicted” profit is observed for scenario one and two in Figure  2.8. This is due the increased distances of harvestable blocks from the mills over the decision making period which result in increased transportation costs. Profit fluctuations resulted by the market price changes make it more difficult to observe the same declining trend in profits for scenario three. 0 20 40 60 80 100 120 140 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed  p ro fit  o f e xi st in g ag en ts   ( m ill io n$ ) Time (years) 0 2 4 6 8 10 12 14 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed  p ro fit  o f n ew   a ge nt  (m ill io n $) Time (years) (a) (b) Scenario 1 Scenario 2 Scenario 3 Figure  2.8 (a) Average profits of existing agents in the optimal solution for the facility location problem, (b) Average profit of new agent in the optimal solution for the facility location problem It is observed from Figure  2.8 (a) that in scenario one, the new agent underestimates the production costs of other agents and therefore the predicted combined profits is higher compared to scenario two and three. Additionally, the sudden increase in profits in scenario one is due to the new agent’s assumption that all agents have outstanding loans in the first five periods (default period for repaying investment loans) after which their profit level increases. Alternatively, in the two scenarios where there is access to more accurate information, the predicted profit has a more gradual increase, representing the fact that some  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  51 agents have fewer remaining periods of loan repayment. As expected, when price forecasts are available to the new agent, it can correctly predict the changes in the combined profits of other agents. Looking at the predicted profit of only the new agent (Figure  2.8 (b)) it can be seen that the profit values in scenario one and two are exactly the same. This is due to the fact that in both scenarios the new agent has accurate information about its own costs and predicts no change in the market prices. However, this predicted profit correctly reflects market conditions when the agent has access to accurate price forecasts as shown for scenario three. 2.4.2 Observed Profits Figure  2.9 shows the profit of existing agents and the new agent during the first 20 years, as observed through 50 simulation runs. The different assumptions of each scenario are designed to impact the result of the optimization problem and do not affect other aspects of the agents’ operations during the simulation. Therefore, since the location of the new agent (the final result of the optimization problem) is the same for all scenarios, the observed simulation results for scenario one and two have the same values and therefore are not shown in Figure  2.9. Comparing the results of the optimization problem from Figure  2.8 (a) and the simulation runs in Figure  2.9 (a) shows that the detailed annual predictions do not match precisely. However, the overall trend of the profit in the simulation runs is approximately the same as predicted in scenario three, with an increase in profits until year 10 and a decreasing trend afterwards. This is to be expected since the market prices strongly affect the profit of agents, and access to correct price forecasts improves the predictions of the agent about the supply chain profits. The difference in the scale of profits is due to the fact that growth and expansion of agents is not represented in the formulation of the optimization problem and therefore the new agent cannot predict an increase in the size (and consequently the profits) of other agents.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  52 -100 -50 0 50 100 150 200 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P ro fit  o f e xi st in g ag en ts   ( $ m illi on ) Time (years) 0 5 10 15 20 25 30 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P ro fit  o f t he  n ew  a ge nt   ( $ m illi on ) Time (years) (a) (b)  Figure  2.9 (a) Average profit (േ 1SD) of existing agents based on simulation results in scenario 3, (b) Average profit (േ 1SD) of the new agent based on simulation results, scenario 3 Focusing on the profit of the new agent only, shown in Figure  2.9 (b), it is clear that it is different from the predictions in all scenarios in Figure  2.8 (b). This may be because the optimization problem presented in  2.2.2 is a simplistic representation of the agent interactions that are taking place in the simulation model.  The difference in the scale of the profits between the two figures is because of the new agent’s expansion during the simulation. The expansion starts within the first decade of establishment in all simulation runs but the agent cannot predict this when solving the optimization problem. Also the effect of random elements in the simulation cannot be predicted by the agent. For example, the order in which the agents access the harvest blocks is random in the simulation (However, once when they do get access to harvest blocks, they harvest the closest blocks first). If the new agent accesses the forest tiles before others in a period, it could harvest the closest tiles and pay less transportation costs, which would result in higher profits in that period. Therefore, although price drops (in year 11 and 16) result in a decrease in the agent’s profits, the extended capacity of the agent and the random factors in the simulation create large fluctuations of profit between periods. 2.4.3 Predicted and Observed Log Exports In the optimal solution in every scenario all of the sorted logs are directed to export locations, indicating that it is sub optimal for the logs to be sold to sawmills within the supply chain. This is the result of the objective function taking into account the combined profit of all  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  53 agents, rather than only sawmill profits or the new sort yard profit. Since the sort yard revenue from any log sales to the sawmill is equal to the sawmills’ cost of purchasing those logs, the potential benefit of purchasing logs from the sort yard is only in lower harvesting and transportation costs for the sawmills. However, exporting logs means higher revenue for the supply chain with no sawmill costs associated with it. This increased revenue is higher than the decreased costs of harvesting if the mills were to buy the logs from the sort yard and therefore exporting is selected in the optimal solution. Although it might seem that access to on demand high quality logs with low pulp share might encourage sawmills to buy logs from the sort yard, their high production costs prevent them from exploiting this opportunity. Harvesting logs is inexpensive enough that even considering the lower value of the logs, it is still more economic for the mills to harvest mixed stands rather than buy sorted logs. Figure  2.10 shows the ratio of exported logs to total log output of the new sort yard during the first 20 years of the simulation. Comparing the results of the simulation with the predicted results, it can be seen that before the market downturn, most logs are in fact exported to outside destinations. However, in the later periods the log prices decrease and they become more affordable for sawmills while simultaneously harvesting costs are increasing as a result of increasing distance between mills and forest tiles, therefore an increasing portion of the logs is sold to the sawmills within the supply chain. Prescribing exporting logs may also explain the selection of the same location in all three scenarios. It is observed that although different levels of information availability change the predictions of the agent about profit levels, the location decision is not affected. The same location and production capacity is selected in all scenarios. Since all the logs are set to be exported, the selected location of the new sort yard is the closest to the harvest blocks, minimizing the transportation costs from the forest to the sort yard.  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  54 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Lo g Ex po rt ra tio  (% ) Time (years) Figure  2.10 Average log export ratio (േ 1SD) of the new sort yard based on simulation results However, it must be noted that the criteria for locating the new agent in this case study was having a positive Net Present Value (NPV) over the decision time horizon. In many cases the decision maker may be interested in consistent or increasing cash flows and the improved predictions could make a difference in such cases. From what was observed here, it may be beneficial for the decision makers to spend more resources on constructing a strong market forecast, rather than acquiring exact information on the cost structure of their competitors. An exact knowledge of the operational costs of other firms is of little value when paired with inaccurate market predictions. 2.5 Conclusions Supply chain models have been extensively used in many manufacturing and service industries around the world, aiding managers and decision makers in analyzing and understanding the interactions of supply chain members and predicting the outcome of their decisions. In this work an existing strategic agent-based forest sector model, CAMBIUM, was modified and extended in order to address its limitations and enhance its applicability. Specifically, product and mill differentiation based on log quality was incorporated in the model; importing and exporting option for logs was added; and a facility location module  Chapter 2: New Facility Location in a Forest Products Supply Chain Model  55 was integrated in the existing model which would select the optimal location of a new facility in the supply chain from a set of potential sites. The model was applied to the case of a forest supply chain on BC Coast, in order to establish a new sort yard in one of the potential locations on Vancouver Island. In comparing the optimal value of the profit and log allocations with the outcome of the simulation, interesting observations were made. It was seen that access to correct market forecasts will significantly improve the predictive abilities of the facility location module, while access to competitor cost structure did not improve the results as much. It was also observed that the optimization problem represented a simplified version of the supply chain operations and could not predict the capacity expansions or occasional fluctuations of the profits that would happen during the simulation. This issue could be addressed by extending the formulation with more relevant variables and constraints, but in turn could result in a mixed integer program too complex to be solved.  56 Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia Chapter 3. Impact of Establishing a Centralized Sort Yard  in Coastal British Columbia 3.1 Introduction The forest products industry has shown a growing interest in establishing log sort yards in response to declining inventory and quality of timber resources and in an effort to recover more value from available resources (Dramm et al., 2002). Log sort yards are central sites in which harvested logs are unloaded, inspected, merchandized (cut to length), graded, marked and sorted into bins based on species, dimension, quality, and potential usage. The sorted logs are then sold to interested buyers based on a competitive bidding process (Schadendorf & Eddy, 1995). For integrated forest management companies, log sort yards are an alternative to sorting at the harvest sites and can potentially improve log utilization and result in better value recovery of small diameter material and underutilized species. Better merchandizing decisions can be made in a sort yard with professional log graders and equipped with automated bucking systems. Sort yards could also provide improved marketing opportunities for wood and biomass from thinning operations. Furthermore, sort yards can benefit forest-dependent communities by providing a reliable log source that can support local forest product businesses (Dramm et al., 2004; Sunderman, 2003). In British Columbia (BC), log sort yards have been suggested as the means for diversifying and strengthening small wood-manufacturing businesses, and bridging the gap between wood suppliers and wood users (Sunderman, 2003). There have been instances of successful log sort yards in the BC interior region run by the government (e.g. Lumby sort yard) or independently (e.g. Revelstoke community forest) (Schadendorf & Chiocca, 1997). However, there have been only a few quantitative studies on optimizing the location choice (Broad, 1989; Sessions et al., 2005; Sessions & Paredes, 1987) or analyzing sort yard operations  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  57 (Dramm et al., 2004; Dramm et al., 2002; Han et al., 2009; Sunderman, 2003; Venn et al., 2009). Sessions and Paredes (1987) provided a heuristic solution approach for the Mixed Integer Program (MIP) problem of locating multiple sort yards in a network comprised of forest nodes, mill nodes, and sort yard nodes. They presented a two phase solution approach that is less computationally exhaustive compared to traditional solution methods for MIP problems, such as branch and bound algorithm. Alternatively, Broad (1989) provided an MIP formulation of the sort yard location problem and recommended branch and bound as the solution method. Sessions et al. (2005) studied the benefits of central sorting facilities with a case study on Vancouver Island in BC and compared centralized sorting of logs to sorting of logs at the harvest sites. Although their model incorporated the probability of misallocating logs at the landing, their findings show that it is still preferable to sort logs at the landing instead of a centralized sorting facility. The authors claim that this decision may be the result of the cost data used for their case study and may be different for other forest companies. Additionally, the misallocation cost in their work was only related to recognizing the mistake and returning the log to the correct sort. Sessions et al. (2005) suggest that including the cost of mis-manufacturing a product from a mis-allocated log may change the results. Dramm et al. (2002) provided a general overview of log sort yard operations in the United States, including some examples from Canada. They discussed objectives, layout and equipment selection criteria for different types of sort yards. They identified the key elements of a successful sort yard to be a well-conceived business plan, dependable timber supply, product diversity, and a reliable transportation infrastructure. Additionally, they suggested that a certain percentage of high value logs in the log supply mix was required to make the sort yard financially feasible. Sunderman (2003) presented a case study on the formation and operation of the Creston community log sort yard in BC, identifying its success factors and challenges. He recommended close relationships among local community, government and sort yard managers in order to ensure the success of any sort yard. Dramm et al. (2004) discussed marketing and economic concepts related to establishing sort yards and offered assessment tools to study their feasibility. They emphasized the need for detailed planning and analysis before establishing a sort yard. Importance of a mix of high and low quality logs in the timber supply of the sort yard was discussed as well as the role of log market  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  58 fluctuations on the sort yard profitability. Following the work of Dramm et al. (2004), a spreadsheet analysis tool, Log-sort Yard Cash Flow Analysis (LSY), was developed by Bilek (2004) to aid decision makers in feasibility studies of sort yards. Although sort yards focusing on small logs have been associated with low profits, Venn et al. (2009) presented a feasibility study for a potential sort yard in Montana to increase the residual value of timber harvested from forest health restoration treatments. Harvested logs from such treatments are small, low value, and with mixed species. This study suggested that establishing a sort yard substantially increases the residual value of such logs. It also showed that profits are sensitive to log prices; log prices below a certain level will eliminate the added benefits of the sort yard. Han et al. (2009) reported the results of a similar feasibility study on a sort yard in Western Montana dedicated to sorting and selling small diameter logs. Their results showed that under specific market conditions such a sort yard may be profitable, however the profit margin is very sensitive to log prices and the delivered log costs. There have been two previous studies on sort yard operations in the BC interior region. Schadendorf and Eddy (1995) reported the results of a sort yard feasibility study for the Boundary area. Their findings showed that although the undertaking was feasible based on their assumptions, the profits were extremely sensitive to slight changes in log prices and sort yard operating costs. Schadendorf and Chiocca (1997) studied the effect of the Lumby sort yard operations on wood consumption of the local value-added wood products industry during a three year period from 1993 to 1996. Their results showed that although the sort yard was profitable during the study period, the value-added wood products sector did not benefit significantly from the logs sold by the sort yard. The main reason was reported to be the high bid prices because of competition from major licensees, and the lack of consistency in wood supply quality and quantity. Except for the work of Sessions et al. (Sessions et al., 2005) and a BC government report (Sterling Wood Group Inc., 2002) that provided estimates of sort yard investment and operational costs, there are no studies on the feasibility and operations of a log sort yard in the coastal BC. Sessions et al. (2005) focused only on the first stage of the supply chain, which is harvesting the wood and sorting it either at the landing or at the sort yard. Therefore,  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  59 the potential sawmill costs or benefits in relation to the sort yard were not included in their study. Further research on potential benefits and performance of log sort yards on the operations of the entire supply chain is necessary to provide additional evidence in support or against establishing such facilities. In this chapter, using an agent-based simulation model described in Chapter 2, I provide a strategic analysis of the impact of a central sorting facility on the profitability of a forest products supply chain on Vancouver Island, BC. This chapter further describes the methods and data used for the case study followed by results and discussions of two scenarios and concluding remarks. 3.2 Modelling Sort Yard Operations in CAMBIUM 2.0 CAMBIUM 2.0, the forest sector model discussed in Chapter 2, is used to locate and simulate the operations of a sort yard within a forest industry supply chain. CAMBIUM 2.0 is an extended spatial multi-agent model based on CAMBIUM (Schwab, 2008). It combines optimization and simulation of the forest sector supply chain to establish a new sawmill or sort yard within an existing forestry supply chain and simulate the operations of the new and existing firms through time. The structure of the model, the flow of simulation, and the formulation of the facility location problem were presented in Chapter 2. Since CAMBIUM 2.0 is a strategic model, the operations of sort yard and sawmills are modelled at the most aggregate level, only considering the input raw material and the final production output using conversion factors (product recovery factors). The sort yard in this model is represented as a facility that harvests wood from the forest, sorts the logs, and sells them as pulp logs and high grade or commodity grade saw logs to the sawmills and export destinations. The log quality used for sorting logs into high value or commodity value categories is based on site index of their harvest block of origin; more productive sites have a higher proportion of high grade logs. Better bucking and merchandizing decisions in the sort yard are reflected through the lower pulp log output of  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  60 the sort yard compared to direct delivery from the forest to the mills, as shown in Figure  3.1.18 Forest Sawmill Pulp log High and low quality logs used for production Forest Sort Yard Sawmill 20% 80% 15% 85% 1% 99%  Figure  3.1 Ratio of saw and pulp logs delivered from the forest and the sort yard When sawmills harvest wood from the forest, 20% of the harvested timber is categorized as pulp wood. This share is reduced to 15% when a sort yard harvests wood from the forest, indicating that more marginal logs are identified as potential saw logs in the sort yard. Also, when sawmills purchase logs from the sort yard, the pulp share is only 1%, reflecting the advantage of buying sorted logs versus receiving a mix of logs directly from the forest. 3.3 Data and Scenarios 3.3.1 Case Study Data Analyses presented in this chapter are based on the case study of coastal BC forest industry which was presented in Chapter 2. Therefore, all of the forest inventory information and sawmill and sort yard operational data are the same as what was described in Chapter 2. Three locations are considered as potential sort yard locations: Port Alberni, Nanaimo, and Campbell River. The potential locations and existing mills are presented in Figure 2.6. The maximum log import limit for sawmills in this analysis is 400,000 cubic meters (m3). 3.3.2 Market Price Scenarios Two scenarios are explored in this analysis regarding the market prices. In order to investigate the impact of the sort yard on profitability of the supply chain, each price scenario  18 Figure  3.1 shows the ratio of pulp logs and saw logs delivered from different sources to the sawmills. For example, if a sort yard harvests timber from the forest, 15% of its harvest volume is categorized as pulp logs and the rest is sold to the sawmills. However, when the sawmill purchases the logs from the sort yard, 1% of the logs are still categorized as pulp logs, accounting for some mis-sorting at the sort yard.  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  61 is analyzed with and without the presence of the sort yard. Results are reported as averages of values over 50 simulation runs to reduce the impact of random factors in the model. In scenario I (the base case scenario), market prices are constant throughout the simulation. The prices are the same as the prices presented in Table 2.5 in Chapter 2. The purpose of this scenario is to observe how a sort yard impacts the performance of the supply chain. In scenario II, market prices are assumed to be cyclic, in order to investigate the performance of the supply chain in a changing business environment. Figure  3.2 shows the average price of log and commodity lumber in recent decades. It can be seen that the frequency of market shocks (sudden rising or falling of prices) has been increasing, with shock periods having a duration of one or two years, followed by a longer period of a more gradual price change. 0 20 40 60 80 100 120 140 19 90 19 92 19 94 19 96 19 98 20 00 20 02 20 04 20 06 20 08 20 10 A ve ra ge  lo g pr ic e ($ /m 3) Time (year) (a) (b) 0 50 100 150 200 250 300 350 400 450 19 90 19 92 19 94 19 96 19 98 20 00 20 02 20 04 20 06 20 08 20 10 Lu m be r c om po si te  p ric e ($ /m bf ) Time (Year) Figure  3.2 (a) Framing lumber composite price, (b) Weighted average log price (over all species) for the BC Coast, Source: Random Lengths (2011), BC Ministry of Forests, Land, and Natural Resource Operations (2011c) Considering the price cycles observed in log and lumber markets, the change in market prices for scenario II is presented in Figure  3.3, showing a cyclic pattern with short periods of market shocks followed by longer periods of stable prices that are close to the base case scenario. Since the purpose of this scenario is to analyze the supply chain development under market shocks, the frequency of the shocks is higher in the beginning of the simulation. This is because sawmills mostly grow and expand during the early periods of simulation (mainly because the timber resources are limited and do not allow mills to expand constantly), after which price changes may not affect them as much as they would during the earlier periods.  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  62  Figure  3.3 Market price changes for scenario II 3.4 Results and Discussion Results for both scenarios are summarized as average values over five year intervals, error bars show the standard deviations over the five year periods for all 50 simulation runs. As previously mentioned, each scenario is studied twice: with and without locating a sort yard. In the case where a sort yard enters the simulation, Port Alberni is selected in all simulation runs as the optimal location, because it is closest to the harvest blocks and also the mills. 3.4.1 Scenario I Profits and Harvest Levels Figure  3.4 shows the profits of supply chain members with and without establishing a sort yard for scenario I (constant prices). Figure  3.4(a) shows that the profits of sawmills grow within the first 30 years of simulation after which they fluctuate occasionally but do not show a major growth or decline. The same trend can be seen in profits of sawmills and the sort yard when a sort yard enters the simulation, as shown in Figure  3.4(b). The initial combined loss for sawmills is because of the operating losses of the commodity mills which have high operating costs and low value products. These mills cannot economically survive and become insolvent within the first three or four years in all simulation runs, therefore a significant increase in profits is observed between the first and the second data point on the profit curve. 50% 60% 70% 80% 90% 100% 110% 120% 130% 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 10 0 10 5 11 0 11 5 12 0 12 5 13 0 13 5 14 0 14 5 15 0 C ha ng e in  p ric es  (% ) Time (Year)  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  63 (a) (b) -100 -50 0 50 100 150 200 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) Combined Sawmill Profits -100 -50 0 50 100 150 200 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) Combined Sawmill Profits Sort Yard Profits Total Profits  Figure  3.4 (a) Average total profits of supply chain members (േ1SD) scenario I – without a sort yard, (b) Average total profits of supply chain members (േ1SD) scenario I – with a sort yard Comparing the combined sawmill profits in Figure  3.4(a) with the total profits in Figure  3.4(b) does not show a significant difference in profit levels. Therefore, it seems that the addition of the sort yard does not greatly increase the overall profits of the supply chain. This can be better observed in Figure  3.5(a), which shows the total profits of the supply chain members with and without the sort yard (Error bars in Figure  3.5 have been omitted for better readability of the graphs. Standard deviations for profit levels can be observed in Figure  3.4, while variations of harvest levels are presented in Appendix B-1). In order to understand why profits do not increase when a sort yard enters the supply chain, resource limitations should be discussed first. Figure  3.5(b) shows the harvest levels during the simulation time horizon, with and without the presence of the sort yard. It is seen that the harvest volume grows during the first 25 years and fluctuates afterwards in both cases. This is not the result of enforced harvest limits, since the Annual Allowable Cut (AAC) for the modelled forest area is just under six million cubic meters (BC Ministry of Forests, 2011a) and the harvest level never reaches this limit. Alternatively, the reason for the fluctuations of harvest can be explained with the shortage of harvestable timber volume. In the beginning of the simulation, mills rapidly harvest the available timber that matches their required quality while expanding their production capacity, therefore depleting the available good quality and most accessible timber. However, after year 25 when the timber shortage starts, the harvest level is mainly limited by economically available timber volume, which declines for a while  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  64 and then fluctuates as various forest stands (tiles) mature and become available for harvest.It can be seen that the profit shown in Figure  3.5(a) closely mimics the harvest levels, indicating that the profit is strongly affected by available timber for harvest. (a) (b) -100 -50 0 50 100 150 200 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) Without Sort Yard With Sort Yard 2.0 2.5 3.0 3.5 4.0 4.5 5.0 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Without Sort Yard With Sort Yard  Figure  3.5 (a) Average total profits of supply chain members with and without the sort yard: scenario I, (b) Average total harvest volume with and without the sort yard: scenario I Therefore, although it may be expected that adding a sort yard improves the profit levels of the supply chain, the sort yard is competing with sawmills to access the limited timber resource. Figure  3.5(a) shows that addition of a sort yard increases the overall profits while no shortage of timber exists. However, after year 25 when timber becomes scarce, if the sort yard harvests a block, it means sawmills cannot harvest it and will have to buy logs from either the sort yard or external sources which is more costly; hence, little or no increase in overall profits is observed. This is partly because of the cost structure of the forest sector on the coast where stumpage prices are extremely low. This acts as an incentive for the mills to choose harvesting over purchasing logs, except in cases where the expected transportation cost from a distant forest block is high enough to justify the cost of buying the logs from the sort yard. Log imports and exports Apart from harvesting, sawmills have the option of importing logs from external log sources (log markets or local forest companies managing other forest licenses in the area). Figure  3.6 shows the volume of logs imported or purchased from the sort yard in scenario I. Further illustrating the shortage in resources, it can be observed from Figure  3.6(a) that the import  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  65 starts after 25 years, when the first shortage occurs. The import trend is similar and only slightly lower when a sort yard exists in the supply chain, as shown in Figure  3.6(b). Standard deviations for “purchased saw log” volume are presented in Appendix B-2.  Figure  3.6 (a) Average total imported logs (േ1SD), scenario I – without a sort yard, (b) Average total imported and purchased sort yard logs (േ1SD), scenario I – with a sort yard Although sawmills need sources other than harvest blocks to meet their demand for timber, it is observed that the volume of logs purchased from the sort yard fluctuates and is below the log output capacity of the sort yard (see Figure  3.7). As explained in Chapter 2, the decision of sawmills to harvest or purchase logs is determined based on their respective estimated costs. Since harvesting with current stumpage prices (5 $/m3 for stumpage and 45 $/m3 for logging) is much cheaper than purchasing the logs (130 $/m3 for high quality logs), the majority of the mills choose to harvest rather than buy logs from the sort yard. However, once the harvest step occurs and some of the mills do not receive their required timber from the harvest blocks, they cannot revise their decision and choose to purchase from the sort yard, and therefore have to acquire logs from external sources. Consequently, most of the logs from the sort yard are exported, as shown in Figure  3.7 (Error bars are omitted for better readability. Standard deviation is presented for later discussion in Figure  3.9). (a) (b) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Purchased from sort yard Imported 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Imported  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  66 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e  (m illi on  m 3 ) Time (years) Log Production Exported Logs  Figure  3.7 Log output and export volume for the sort yard, scenario I Production levels The production levels and capacity growth of the sawmills and the sort yard shows how the supply chain expands if no market shocks occur. Figure  3.8 shows the lumber production and capacity levels for scenario I, with and without the sort yard. High Value Lumber Production Commodity Lumber Production Total Lumber Production Capacity (a) (b) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m bf ) Time (years) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m bf ) Time (years)  Figure  3.8 (a) Average lumber production and capacity levels (േ1SD), scenario I – without a sort yard, (b) Average lumber production and capacity levels (േ1SD), scenario I – with a sort yard In both cases (with and without the sort yard), commodity lumber production levels are extremely low. This reflects the fact that commodity mills become insolvent and drop out early in the simulation. The observed commodity production is in fact carried out by high value sawmills when their inventory of low quality logs reaches a user defined limit. The low  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  67 levels of commodity production are also the result of the log harvesting and importing assumptions of the model. Since it is assumed that the high quality mills can “import” the logs that match their requirements, and since they do not harvest forest tiles with “poor” site index and high volume of low quality logs, their inventory of low quality logs does not grow rapidly and therefore production levels of commodity lumber remains low. Production and capacity levels of high value mills show their growth and are the result of their investments in capacity expansion. Of course, the production levels and sawmill capacities remain relatively stable after the initial growth period, because even when the capital is available to make further expansions, the timber shortage means that any additional capacity will not be utilized19. Comparing Figure  3.8(a) and Figure  3.8(b) shows a difference in production levels that is to be expected, considering the  competition for resources. Since capacity expansion for agents requires a minimum user defined capacity utilization rate, when sawmills do not receive enough timber to reach that utilization rate, they cannot increase their capacity. Therefore when a sort yard is present and the competition for resources is high, capacity expansion of sawmills is negatively affected. The slow negative trend in Figure  3.8(b) reflects the downsizing of sawmill agents because they cannot access enough timber to sustain a high enough cash flow and pay the fixed costs associated with their unutilized production capacity.  19 In this study a minimum utilization rate of 95% is considered necessary before an agent can expand its production capacity. Utilization rate is calculated as the ratio of production volume to production capacity.  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  68 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e  (m ill io n m 3 ) Time (years) Figure  3.9 Log production of the sort yard (േ1SD) for scenario I Figure  3.9 presents the log production of the sort yard (processing of unsorted timber into sorted logs), showing its growth even after the first 25 years. However, high variability of the data after year 25 shows that these production levels change significantly from one simulation run to the other. Since the agents access the harvest blocks in a random order, sometimes the sort yard may be among the first agents to access the forest and therefore receives the timber it requires for that period. However, if that is not the case, the sort yard may not harvest any timber and has to use only its existing timber inventories for log production. This causes large variations from one simulation run to the other. Product Recovery Factors Sawmill investments expand not only their production capacity, but also improve their product recovery factors through acquiring new production technologies and equipment. Table  3.1 shows the change in the product recovery factors of sawmills during the simulation. Regardless of the presence of the sort yard, the sawmills improve their product recoveries, with most of the improvements occurring during the first 30 years which is the main growth period for agents in the simulation.  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  69 Table  3.1 Average sawmill product recovery factors for scenario I  Product Recovery Factors (bft/m3) in Each Simulation Year  1 10 20 30 50 100 150 Without Sort Yard 236.54 249.70 264.02 265.95 266.54 267.10 267.37 With a Sort Yard 234.05 248.83 263.19 264.82 265.87 266.88 266.97 3.4.2  Scenario II Profits and Harvest Levels Figure  3.10 shows the profit of the supply chain members under scenario II (cyclic market prices). The differences between the two graphs can be explained similar to the results presented for scenario I. Profit in scenario II follows market price cycles as expected, with sudden decreases and increases that match the market shocks. High variability of the data at points that correspond to market shocks is due to averaging the results over five year periods which include different market prices. Comparing Figure  3.10 with Figure  3.4, overall higher profits are observed in the later periods in scenario II compared to scenario I. This is due to the slightly upward trend of the market prices in scenario II. (a) (b) -100 -50 0 50 100 150 200 250 300 350 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) Combined Sawmill Profits -100 -50 0 50 100 150 200 250 300 350 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) Combined Sawmill Profits Sort Yard Profits Total Profits  Figure  3.10 (a) Average total profits of supply chain members (േ1SD): scenario II – without a sort yard, (b) Average total profits of supply chain members (േ1SD): scenario II – with a sort yard Figure  3.11 shows the harvest levels for scenarios one and two, when a sort yard is present (Error bars for harvest levels in scenario II are presented in separate graphs in Appendix B-1). Harvest levels in scenario II do not show any major change from scenario I, except that the timber shortage starts slightly later (year 30 instead of year 25). This may be  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  70 due to the negative impact of cyclic prices on sawmill capacity expansion, which consequently slows their harvesting rate compared to scenario I. The same trend is observed for harvest levels when the sort yard is not present. 2.0 2.5 3.0 3.5 4.0 4.5 5.0 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Scenario One Scenario Two  Figure  3.11 Average total harvest volumes with the presence of a sort yard for scenario I and II Log imports and exports Figure  3.12 shows the log imports of sawmills in scenario II and the volume of logs they purchased from the sort yard. Import levels are similar for scenario II compared to scenario I while purchased logs show more fluctuations. The fluctuations match the price cycle of scenario II (e.g. year 20 and 90). Whenever there is a price fall, which makes sorted logs more affordable to sawmills, the volume of purchased saw logs from the sort yard increases.  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  71  Figure  3.12  (a) Average total imported logs (േ1SD), scenario II – without a sort yard, (b) Average total imported and purchased sort yard logs (േ1SD), scenario II – with a sort yard Similar to scenario I, the major proportion of log output from the sort yard is exported, as shown in Figure  3.13. However, sawmills seem to buy more logs from the sort yard in the earlier periods of the simulation compared to same periods in scenario I. This can be because of price drops in the first 30 years in scenario II which would make the sorted logs more affordable to sawmills. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e  (t ho us an d m 3 ) Time (years) Log Production Exported Logs  Figure  3.13 Log output and export volume for the sort yard for scenario II (a) (b) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Imported 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Purchased from sort yard Imported  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  72 Production levels Figure  3.14 shows the lumber production levels and capacities of sawmills for scenario II. High Value Lumber Production Commodity Lumber Production Total Lumber Production Capacity (a) (b) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m illi on  m bf ) Time (years) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m illi on  m bf ) Time (years)  Figure  3.14  (a) Average lumber production and capacity levels (േ1SD), scenario II – without a sort yard, (b) Average lumber production and capacity levels (േ1SD), scenario II – with a sort yard The same trend and behavior is observed as in scenario I. Although the prices in scenario II have an upward trend, the production levels do not follow this trend, most probably due to harvestable timber shortage and the fact that the sawmills reach their maximum log import limit. Therefore, availability of logs seems to be the main factor limiting further expansion of the mills. An interesting observation can be made by looking at Figure  3.15 which shows the log output of the sort yard in scenario II. Market shocks seem to have a negative effect on the growth of the sort yard as seen by comparing log output levels in scenario I (Figure  3.9) and II. The initial drop in log prices may prevent the sort yard from making capacity expansions, and once the prices increase, the timber shortage results in low capacity utilization rates and therefore, capacity expansion is slower for the sort yard and does not reach the same levels as scenario I. Similar to scenario I, The reason for high fluctuations of log output in Figure  3.15 is the shortage of timber and the random order in which the agents access the forest.  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  73 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e  (m ill io n m 3 ) Time (years) Figure  3.15 Log production of the sort yard (േ1SD) for scenario II Product Recovery Factors Table  3.2 shows the change in the product recovery factors of sawmills during the simulation for scenario II. A trend similar to scenario I is observed, with most of the improvements made during the first 30 years. Table  3.2 Average sawmill product recovery factors for scenario II  Product Recovery Factors (m3/mbf) in Each Simulation Year  1 10 20 30 50 100 150 Without Sort Yard 237.30 247.76 264.36 265.94 266.58 267.19 267.21 With a Sort Yard 236.40 245.77 262.81 265.34 266.48 266.98 267.14 3.5 Conclusions Considering the potential benefits of the sort yard to forest product supply chains, a case study was performed on a coastal BC forestry supply chain using an agent-based forest sector simulation and optimization model, CAMBIUM 2.0. Two price scenarios were considered to investigate the performance of the supply chain under different market conditions. The addition of a sort yard did not seem to improve the profitability of the supply chain as a whole in either of the market price scenarios. The main reason for this is the limited forest resources that all agents need to manufacture their  Chapter 3: Impact of Establishing a Centralized Sort Yard in Coastal British Columbia  74 products. These results match previous research findings on establishing a sort yard on the BC Coast (Sessions et al., 2005). Low stumpage prices (in comparison to higher cost of purchasing the logs) mean that it is much cheaper for sawmills to harvest their timber rather than buy logs from the sort yard. Consequently, the sort yard exports most of its log output to external destinations. Therefore, based on the model structure and cost and price assumptions in this study, although a sort yard may be able to generate profits and expand its production capacity within the modelled supply chain on the BC Coast, it seems that with the current cost structure of the industry, its addition to the supply chain would not impact the existing sawmills positively and would result in high log export levels which may not be desirable for local communities and businesses.  75 Chapter 4: Impact of Harvest Policy Changes on Sustainability  Chapter 4. Impact of Harvest Policy Changes on Sustainability 4.1 Introduction 4.1.1 History of Forest Management in BC British Columbia’s forests cover nearly two thirds of the provincial land base, with an approximate area of 55 million hectares. They are also among  the world’s largest public forests (BC Ministry of Forests, 2010), with 95% of the forest land owned by either provincial or federal governments (BC Ministry of Forests, 2006). Therefore, it is the responsibility of the government to manage and regulate the use of public forests while ensuring their long-term sustainability. Historically, forest companies in BC  accessed the province’s forest resources with minimal intervention from the government (Prudham, 2007). However, as concerns grew over the unregulated harvesting activities, the government established a royal commission in 1945 (Sloan, 1945) which introduced the “sustained yield” of timber and the annual allowable cut (AAC) into the Forest Act of BC (BC Ministry of Forests, 2010).  Area-based tenures, later named Tree Farm Licenses (TFLs), were developed that would transfer specific rights to use public forest land and resources to forest companies in exchange for their commitments to invest in manufacturing facilities, pay stumpage and provide long-term forest management (BC Ministry of Forests, 2006). The central idea of sustained yield was to convert natural forests into normal, “ideal” forests with trees of various ages distributed and growing in a way that could produce equal annual volumes of timber continually without negative impacts on future production (Prudham, 2007). Implementing sustained yield forestry led to the accelerated liquidation of older natural forests with the eventual goal of making the forests more productive with various young stands. This led to increasing public concern over the loss  Chapter 4: Impact of Harvest Policy Changes on Sustainability  76 of non-timber forest values such as recreation or wildlife, and resulted in the establishment of another royal commission in 1976 (Pearse, 1976). The findings and recommendations from this commission led to reforms in timber tenures and the Forest Act of 1979 and the Ministry of Forests Act (BC Ministry of Forests, 2010). New forms of tenure were introduced, including those that would later be known as replaceable and non-replaceable Forest Licenses (FL). These licenses were developed specifically for small sawmill owners and independent manufacturing facilities (BC Ministry of Forests, 2006). The government was now required to ensure “integrated resource management” by taking into consideration the non-timber values of the forest. Further developments in forest management policies of the province included the start of consensus-based land use planning in 1992, the Forest Practices Code of British Columbia Act in 1995, and finally the Forest and Range Practices Act (FRPA) in 2003 which was developed in response to the growing international attention to sustainable forest management (SFM20)  (BC Ministry of Forests, 2010; Hickey & Innes, 2008). Under FRPA, criteria and indicators (C&I) were developed based on global and national C&I (CCFM, 1997, 2003; Montreal Process, 1995) to define, promote, and assess the efforts of government and forests companies towards sustainable management of timber and non-timber forest values (BC Ministry of Forests, 2010). In BC, forest companies gain the right to harvest timber in public forests through tenure agreements with the provincial government, mainly in the form of (Tree farm Licenses) TFLs and Forest Licenses (FLs) (BC Ministry of Forests, 2006). Under the FRPA legislation, tenure holders are held accountable for their operations on public forests, and must demonstrate that they take measures to achieve specified results based on a number of public value indicators (Hickey & Innes, 2008). This is ensured through a management plan - the Forest Stewardship Plan (FSP) - that is submitted by forest tenure holders and must be approved by the provincial government. The FSP provides measurable results  20 Sustainable Forest Management (SFM), according to Natural Resources Canada (Natural Resources Canada, 2007) is defined as “Management that maintains and enhances the long-term health of forest ecosystems for the benefit of all living things while providing environmental, economic, social and cultural opportunities for present and future generations”.   Chapter 4: Impact of Harvest Policy Changes on Sustainability  77 and strategies consistent with government objectives for various forest values (BC Ministry of Forests, 2010). Although efforts are being made by the government to meet the public’s growing demand for sustainable management of timber and non-timber forest values, some argue that the 1940’s “sustained yield” attitude still prevails in the forest industry, especially in setting AAC levels, leading to rapid exhaustion of the province’s natural timber reserve (Burda et al., 1998; Cashore et al., 2000; Luckert & Haley, 1995; Prudham, 2007). The AAC determination is based on sustaining the timber volume available for harvest in the future, without consideration of its value or access cost beyond what is determined to be outside the timber harvesting land base.  The timber harvesting land base excludes non-commercial species, stands of low productivity and areas that are considered to be economically inaccessible. The result of this policy is that most of the accessible, high value timber has been harvested, leaving the hard-to-reach or low value stands for future periods (Wilson, 1987). Therefore, although a certain “volume” of timber is left for future harvest, it may not be economically viable for forest companies to harvest it. This policy has had a negative impact on the coastal BC industry because of the unique characteristics of the forests in the coastal area, including the declining old growth timber inventory and the difficult terrain which makes harvesting uneconomic in some areas. 4.1.2 Coastal BC Timber Resource The coastal BC forests are comprised of mainly of softwood species such as Douglas-Fir (Pseudotsuga menziesii), Western Hemlock (Tsuga heterophylla), Balsam (Abies amabilis) and Western Red Cedar (Thuja plicata), but also have a wide range hardwood species, including Red Alder (Alnus rubra) and Big-leaf Maple (Acer macrophyllum) (BC Ministry of Forests, 2003). Traditionally the coast region has been associated with high value species; however, recently negative trends have been identified in the region including the declining volume of old growth timber, difficult reforestation conditions in many old growth areas, and limited road access which has substantially increased harvesting costs (BC Competition Council, 2006). During the “liquidation-conversion” era of the forest industry in BC, as labelled by Wilson (1998), the old-growth coastal stands were rapidly harvested and converted to  Chapter 4: Impact of Harvest Policy Changes on Sustainability  78 second-growth forests (Prudham, 2007; Thielmann & Tollefson, 2009; Wilson, 1987). This conversion to second-growth forests, paired with a sharp increase in number of protected areas removed from the timber harvesting land base, translates into reduced timber supply over the next couple of decades (Pearse, 2001), which could lead to AAC reduction in the coastal region. Additionally, the practice of high-grading or high retention harvesting (selective harvesting of a stand and taking higher value trees) has resulted in creating uneconomic stands with tress that are no longer profitable to harvest (Forest Practices Board, 2009). A special investigation report by the Forest Practice Board (FPB) in coastal BC region showed that sustainability of timber supply was impacted negatively by high retention harvesting in over half of the forest stands under study (Forest Practices Board, 2009). This was claimed to be the result of harvesting commercially valuable cedar and spruce trees and leaving behind hemlock and balsam trees that have limited prospects for an economically viable future harvest. Additionally, the report suggested limited efforts for reforestation of valuable species was occurring, with management relying instead on natural regeneration. This would encourage hemlock and balsam species which are not profitable to harvest, especially on hard-to-reach,  high cost forest stands (Forest Practices Board, 2009). The results of this study further shows that while current harvest policies may meet some environmental or social values by retaining enough volume of timber, they may not be truly “sustainable” since the future yield of the forest in terms of timber quality is not directly included in harvest planning and policy development. In order to maintain current harvest levels, uneconomic stands need to be harvested and regenerated with desirable species. Otherwise, future timber supply will decline and the sustainable AAC will need to be reduced. Private forest companies may have no financial incentive to harvest uneconomic stands, however it is the responsibility of the government to design and implement policies to move the industry in a direction that guarantees a more sustainable future timber supply that meets economic, environmental, and social objectives. Since the decline of the timber supply, at least over the short term, seems inevitable (Pearse, 2001), both the forest industry and the government must be prepared to face this  Chapter 4: Impact of Harvest Policy Changes on Sustainability  79 timber shortage, while simultaneously trying to design and implement policies that would efficiently make use of available timber resources. Additionally, special efforts must be made to ensure that current harvested areas grow back to a state that can support long term timber needs of the province, considering both the volume, and the value of the resources. This may be done through different policy instruments such as adjusting AAC levels (which only control the volume), or by implementing different harvest priorities that include the value of the resource, such as harvesting a mix of economic and uneconomic stands. Previous studies on the impact of AAC reductions in BC have focused only on the economic impacts of harvest reductions and expectedly have reported mostly negative results (Binkley et al., 1994; Horne et al., 1991; Lax & Parker, 1992). In this Chapter, using an example from a forest products supply chain on the BC Coast, the impact of changes to resource availability and harvest priorities is tested on the supply chain performance as well as the volume and value of the timber that remains in the forest. The rest of this chapter discusses the methods and data, followed by results, discussion, and concluding remarks. 4.2 Methods CAMBIUM 2.0 forest sector model presented in Chapter 2 is used to investigate the impact of changes in resource availability and harvest policy on the wood products manufacturing supply chain, as well as the forest land base. The forested area is represented with approximately 15,000 tiles (each one with a 225 ha area), 4,500 of which are forested. These forested tiles contain either one or two species. Each forest tile has a number of attributes, including site index (a measure of site productivity) and age. The age of the forest tile is used along with a yield curve to calculate the volume of the timber that could be gained by harvesting the tile. As discussed in Chapter 2, the forest inventory data was projected onto the model landscape using an algorithm developed by Schwab and Maness (2010) and from the aggregate information available in the timber supply review reports (BC Ministry of Forests, 2011b). The represented species include balsam, western hemlock, cedar, alder, Douglas-fir, spruce, and pine.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  80 Since the objective of this manuscript is to investigate the impact of different harvest policy and restrictions of the sustainability of the timber resource, it is important to represent the forest resource accurately. However, no public data is available on the distribution of different log qualities for the modelled forest area. Therefore, similar to previous chapters, the proportion of high and low quality logs are estimated using the attributes of the forest tile. However, instead of only considering the site productivity, a combination of site index and age of the forest tile is used in this chapter, to provide a more realistic representation of the available timber quality. The distribution of log quality based on age and site index is shown in Table  4.1. Table  4.1 Log quality ratio by site index and age group Age Group Site Index  Good Medium Poor  High Low High Low High Low                age <= 40   yrs 0.15 0.85 0.10 0.90 0.10 0.90 40   yrs < age <= 60   yrs 0.30 0.70 0.20 0.80 0.15 0.85 60   yrs < age <= 80   yrs   0.45 0.55 0.35 0.65 0.25 0.75 80   yrs < age <= 100 yrs   0.55 0.45 0.40 0.60 0.35 0.65 100 yrs < age 0.70 0.30 0.55 0.45 0.40 0.60 In CAMBIUM 2.0, harvesting activities impact the forested land base by changing both the volume and the value of the remaining standing timber in the forest. In the beginning of each simulation time step, the sum of the volume of all harvestable forest tiles is calculated (harvestable forest tiles are the ones within the “timber harvesting land base” containing trees older than the minimum harvest age). A similar calculation is performed for the value of the available timber in these tiles. The value of the timber is approximated with log prices and according to the proportion of high and low quality logs in the forest tiles, as presented in Equation ( 4.1).  Chapter 4: Impact of Harvest Policy Changes on Sustainability  81 Timber Value=Volume×(1-Pulp Ratio)×High Quality Log Ratio×High Quality Log price Volume×(1-Pulp Ratio)×Low Quality Log Ratio× Low Quality Log price  Volume × Pulp Ratio× Pulp Log Price ( 4.1) 4.3 Data and Scenarios 4.3.1 Data In this analysis, the same BC Coast case study from Chapter 2 is used. Key data are repeated here for the reader’s convenience. Sawmill costs are shown in Table  2.2 and market prices are shown in Table  4.3. Costs and prices are assumed to remain unchanged throughout the simulation. Market prices are used to calculate the value of the standing trees according to Equation ( 4.1). The combined AAC for the six TFLs in the case study is approximately 5.921 million m3. Considering the results of Chapter 2 and 3, establishing a sort yard would result in the majority of its harvested timber being exported. Therefore, in order to eliminate the impact of log exports and to focus on the behavior of the local industry only, no sort yards are present in this analysis. Table  4.2 Sawmill costs for all scenarios  High Value Mill Commodity Mill Manufacturing cost ($/mbf output) 260 200 Fixed cost ($/mbf capacity) 40 40 Logging (tree to truck) ($/m3) 40 40 Maximum stumpage ($/m3) 5 5 Transportation (landing to mill) ($/m3) 0.15 0.15  21 It must be noted that this AAC is based on the most recent information for the modelled TFLs while the forest inventory data was constructed using public information that were not as recent. Therefore, the AAC and the actual available timber may not exactly match.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  82 Table  4.3 Market price for all scenarios  Price ($/unit) Specialty lumber products ($/mbf) 700 Commodity lumber –high grade ($/mbf) 325 Commodity lumber – low grade ($/mbf) 250 Specialty grade log ($/m3) 130 Commodity grade log ($/m3) 60 Pulp log ($/m3) 35  4.3.2 Scenarios Five scenarios are evaluated that modify the resource availability, harvest priorities, and log quality requirements.  The supply chain activities are simulated for 150 years. In order to capture the impact of random elements, 50 simulation runs are performed for each scenario. The presented results are the averages of the variable over 50 runs for consecutive five year intervals. All results are compared with a base case scenario that has AAC of 5.9 million m3 and a log import limit of 250,000 m3. The harvest priority in the base case scenario is “closest first”. In the previous chapters, the high value mills were set to only harvest forest tiles with medium or good site indices. However, since the log quality distribution is now impacted by both age and site index, instead of inspecting the site index of a forest tile, the agents monitor the ratio of high quality logs. If the ratio of the high quality logs in a forest tile is less than 50%, the high value mills will not harvest it. Resource Availability Two scenarios are investigated regarding the AAC levels (scenarios I and II in Table  4.4). Moderate and dramatic reductions in AAC are represented through reducing the AAC by 20% and 40%, respectively. AAC reductions are modelled because AAC and harvest levels in coastal BC have been decreasing during the past two decades (Kiss, 2010; Woodbridge Associates, 2009) and are expected to be reduced even further (Coast Forest Products Association, 2003; Pearse, 2001).  Chapter 4: Impact of Harvest Policy Changes on Sustainability  83 Harvest priority in the first two scenarios is the same as the base case, meaning each agent harvest tiles that are closest to it, as long as the harvested timber meets the quality requirement of the agent. AAC reductions are implemented by reducing the AAC in each individual TFL, rather than enforcing a limit on the combined harvest volume from all TFLs. Table  4.4 Scenario descriptions Scenario Variable Element of Scenario  Name Value I AAC 20%  reduction compared to base case II AAC 40%  reduction compared to base case III Harvest Priority Oldest First IV Harvest Priority Highest Value First V Quality Requirement high value mills harvest high or low value stands Harvest Priority In addition to limiting the harvest volumes, harvesting activities can be controlled through implementing different harvest priorities (i.e. the order in which the forest tiles are harvested). As an alternative to “closest first” harvest priority of the base case scenario, two additional harvest priorities are considered in this analysis, scenarios III and IV, shown in Table  4.4). Scenario III or “Oldest first” priority means that sawmills access the oldest harvestable tiles first, regardless of their distance to the mill.  Scenario IV or “Highest value first” means that sawmills access the forest tiles based on the value of their available timber, calculated based on Equation ( 4.1). All other assumptions in scenarios III and IV, including the AAC level, are the same as the base case. The harvesting step of the model was modified slightly to implement the different harvest priorities of scenarios III and IV. At the beginning of the harvest step in each period, all harvestable forest tiles are arranged in a collection that can be accessed by all agents. This collection is then sorted according to the specific harvest priority, e.g. in descending  Chapter 4: Impact of Harvest Policy Changes on Sustainability  84 order of age. However, the agents do not access these forest tiles in a random order22 (as is done for all other scenarios). Alternatively, the first tile from the collection is selected and the closest mill to this forest tile is selected to harvest it. If the closest mill has no remaining timber demand, the next closest mill is selected and this process goes on until all mills harvest their required timber, or no more forest tiles are left to harvest. While this method of allocating forest tiles aim to prevent the harvesting of forest tiles by agents that are farthest from them, there may still be conditions where a mill has to harvest a forest tile that is closer to another mill (i.e. when the closer mill has no more demand for timber). Quality Requirement Finally, a different approach to harvesting compared to the base case scenario was tested with regards to the log quality requirement. As previously mentioned, in the base case scenario, high value mills will not harvest a forest tile with a ratio of high quality logs less than 50%. In order to see the impact of removing this requirement, scenario V was developed which is similar to the base case, with the only difference being that the quality requirement is removed. On average, for all scenarios, each simulation (150 years) was completed in approximately 20 seconds on a PC with a dual core, 2.1 GHz processor. 4.4 Results and Discussion 4.4.1 Impact of AAC Reduction Impact on Supply Chain Performance Figure  4.1 shows the average volume of harvest and log imports of sawmills under different AAC levels for the base case and scenarios I and II. Standard deviations of the base case values are presented in Appendix C-1 and standard deviations of log imports for scenario I and II are presented in Appendix C-2.  22 As discussed in Chapter 2, the random order of agents in accessing the forest was designed to prevent favoring some agents by giving them earliest access to the forest in every period.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  85 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Scenario IIScenario IBase Case (a) (b)  Figure  4.1 (a) Average total harvest of sawmills (േ1SD) – Base case and scenarios I and II, (b) Average total volume of imported logs – Base case and scenarios I and II A 20% reduction in AAC in scenario I results in a harvest volume relatively close to the base case. The harvest volume falls short of the levels in the base case at approximately year 25, and the gap between the two scenarios widens as the simulation progresses. Year 25 is also the point where the timber shortage starts impacting the industry, causing sawmills to start importing logs in the base case, as well as scenario I. The lower harvest levels in scenario I also result in lower profits and prevent the expansion of sawmills, as seen in Figure  4.2 (Standard deviations for the base case values are presented in Appendix C-1). It is seen that the profit levels in all scenarios remain at a relatively stable level without significantly increasing throughout the simulation. This could be because of low availability of high quality timber, as shown in Table  4.1. High quality timber is required for generating large profits by the mills. Higher levels of low quality timber results in the production capacity being utilized for manufacturing commodity lumber. The lower profits from commodity lumber production prevent the growth of the mills, and impact their ability to import their required high quality logs. Therefore, as shown in Figure  4.1(a), the mills do not reach their log import limit in any of the scenarios.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  86 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m illi on  m bf ) Time (years) Scenario IIScenario IBase Case (a) (b) -120 -100 -80 -60 -40 -20 0 20 40 60 80 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years)  Figure  4.2 (a) Average total profits of sawmills (േ1SD) – Base case and scenarios I and II, (b) Average total production capacity of sawmills (േ1SD) – Base case and scenarios I and II The dramatic AAC reduction of 40% in scenario II strongly impacts the profits of the industry and results in much lower profits and production capacity levels (Figure  4.2).  It also causes an immediate start of the log imports in response to limited harvests (Figure  4.1). However, log imports decrease gradually, mainly because the sawmills downsize or become insolvent as a result of low timber availability and low profits. Figure  4.3 shows that on average, six agents become insolvent in scenario II. The average number of insolvent agents for the base case and scenario I is three and four, respectively. 0 1 2 3 4 5 6 7 8 9 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 N um be r o f A ge nt s Time (years) Figure  4.3 Average number of active agents in scenario II. Error bars show the observed maximum and minimum number of active agents in each time interval.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  87 Impact on the Resource Sustainability In order to see how different harvest restrictions impact the sustainability of the timber resource, we look at the volume and value of the remaining timber in the forest, as shown in Figure  4.4 (standard deviations of all data points are less than 5% of the average value and are not shown since the error bars would be too small to see). Graphs of age distribution of the timber harvesting land base for all scenarios are presented in Appendix C.4. When the AAC is reduced by 20% in scenario I, the levels of remaining volume and value are close to the base case, while a dramatic reduction in AAC in scenario II (40% reduction) causes a stronger upward trend compared to the base case. It is seen that by restricting the annual allowable cut levels, both the value and the volume of the remaining forest increase compared to the base case. This observed increase is expected since both scenarios I and II result in downsizing and lower economic activity of the supply chain members compared to the base case and therefore less harvesting occurs in the forest. Scenario IIScenario IBase Case (a) (b) 150 200 250 300 350 400 450 500 550 600 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) 0 10 20 30 40 50 60 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V al ue  (b ill io n $) Time (years)  Figure  4.4 (a) Average remaining volume of standing timber - Base case and scenarios I and II, (b) Average remaining value of standing timber - Base case and scenarios I and II Another measure that can demonstrate the change in timber resources more clearly is the unit value of the remaining timber ($/m3), calculated by dividing the total value of the remaining timber by the total volume. Figure  4.5 shows the change in unit value of  Chapter 4: Impact of Harvest Policy Changes on Sustainability  88 timber for the base case and scenarios I and II (standard deviations are less than 5% of the average values and are not shown since the error bars would be too small to see). If the mills harvest the forest using a balanced approach, it is expected that while the total volume and value of the remaining timber changes, its unit value should stay relatively stable. However, it is seen from Figure  4.5 that in the base case scenario, where AAC is the highest, the unit value of timber declines more severely compared to scenarios I and II. This is because the mills do not harvest the low quality stands and extract the forest’s high quality timber instead. Therefore, as the mills harvest more forest tiles, there will be less high value timber available in the forest, resulting in a declining unit value of the remaining timber. As a final note on the scenarios, the number of harvested forest stands (tiles) can be compared to see how many of the stands are left untouched because they are uneconomic. In the base case, on average 1511 stands are untouched. This number increases to 1672 and 2428 for scenarios I and II, respectively. This increase is a result of lower harvesting activity in both of these scenarios compared to the base case. 80 81 82 83 84 85 86 87 88 89 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 U ni t V al ue  ($ /m 3 ) Time (years) Scenario I Scenario II Base Case  Figure  4.5 Average unit value of remaining timber - Base case and scenarios I and II  Chapter 4: Impact of Harvest Policy Changes on Sustainability  89 It can be seen that while scenario II benefits the timber resource conditions more than the other two scenarios, it negatively impacts the industry. This negative impact of harvest reductions has been pointed out in previous research as well (Binkley et al., 1994). Therefore, if the harvesting preference of mills remains the same, a balanced approach to use the forest resources for economic purposes may not be achieved only by limiting the harvest volume. Instead of focusing on retaining a certain level of timber volume in the forest it may be more beneficial to try and sustain a uniform flow of income from the forest, which would reflect the quality of the harvested timber. I attempt to do this by modelling other harvest priorities. 4.4.2 Impact of Harvest Priority Change Impact on Supply Chain Performance Figure  4.6 shows the harvest levels and log import volumes and Figure  4.7 shows the production capacity and profit levels for the base case and for scenarios III and IV (all standard deviations are presented in separate graphs in Appendix C-3 for better readability). In scenario III the forest tiles are harvested in descending order of their age, by the closest mill that still has demand for timber. Figure  4.6(a) shows that the harvest volumes in this scenario are lower than the base case levels. This is because of the presence of fewer sawmills since higher transportation costs cause more sawmills to become insolvent during the early years of the simulation (on average four agents become insolvent, three of which are bankrupt within the first 40 years). The log import volume in scenario III is also lower compared to the base case. This again is a result of fewer agents being active, which means an overall lower demand and higher availability of timber. Additionally, since younger forest tiles are not harvested at the beginning of the simulation, they are allowed to grow and therefore more timber is available for harvest in future periods.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  90 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Scenario IVScenario IIIBase Case (a) (b)  Figure  4.6 (a) Average total harvest of sawmills– Base case and Scenarios III and IV, (b) Average total volume of imported logs – Base case and Scenarios III and IV Figure  4.7(a) shows a lag in the growth of the total capacity of sawmills in scenario III, mainly due to initial insolvencies. However, the total capacity reaches the levels of the base case after year 90. The mills in scenario III have access to higher quality timber, since age is one factor in determining the ratio of high to low quality timber. While some mills cannot survive the high transportation costs in scenario III, the ones that remain in business have the opportunity to generate high profits and therefore can afford to expand their capacity, as shown by the growth trend in Figure  4.7(a). Scenario IVScenario IIIBase Case (a) (b) 400 450 500 550 600 650 700 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (th ou sa nd  m bf ) Time (years) -140 -120 -100 -80 -60 -40 -20 0 20 40 60 80 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years)  Figure  4.7 (a) Average total production capacity of sawmills – Base case and scenarios III and IV, (b) Average total profits of sawmills – Base case and scenarios III and IV  Chapter 4: Impact of Harvest Policy Changes on Sustainability  91 Scenario IV represents the case where forest tiles are harvested in descending order of their total value, calculated from Equation ( 4.1). Focusing on scenario IV in Figure  4.6(a), it is seen that the harvest levels are lower compared to the base case. Similar to scenario III, this is partly because of fewer active agents in the model (on average four agents become insolvent in scenario IV). However, Figure  4.7(a) shows that the industry in scenario IV does not downsize as sharply as in scenario III. This shows that the remaining agents gain enough profits - by harvesting highest value stands - to sustain their production capacity levels. The profit level in scenario IV is higher than scenario III and close to the base case, reflecting the economic advantage of harvesting the highest value stands compared to harvesting the oldest stands. In scenario IV, the log imports start with a lag compared to the base case, as shown in Figure  4.7(b). In the base case scenario, mills harvest the closest stands that meet their log quality requirements. There are cases where the total inventory of harvested volume reaches a user-defined limit (total production capacity in this example), without enough high quality logs available to meet the production target. In such cases the mills purchase logs from other sources (i.e. log imports) if they can afford it. In scenario IV, since the mills harvest the highest value stands first, they are able to receive all their required high quality logs through harvesting in the early years. However, once the availability of the high quality logs decline, the log imports start. Access to the highest value stands also prevents the import volume to reach the base case level. Scenarios III and IV have a higher number of tiles that are never harvested compared to the base case (1788 and 1680 forest tiles, respectively, compared to 1511 forest tiles in the base case). This could be because of the presence of a lower number of agents and because older tiles and higher value tiles generate higher harvest volumes which would require fewer tiles to be harvested in order to meet the demand. Impact on the Remaining Forest Figure  4.8 shows the average unit value of timber for the base case and scenario III and IV. Scenario IV shows the lowest decline. The reason is since oldest forest tiles are harvested first, younger forest tiles are allowed to grow and have higher yields in future periods. Also, as the simulation goes forward, some of the forest tiles that are harvested  Chapter 4: Impact of Harvest Policy Changes on Sustainability  92 in early periods grow back and since age impacts the ratio of high quality logs in a forest tile, the value of the tiles grow as well. 80 81 82 83 84 85 86 87 88 89 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 U ni t V al ue  ($ /m 3 ) Time (years) Scenario IVScenario IIIBase Case Figure  4.8 Average unit value of remaining timber - Base case and scenarios III and IV The changing trend of unit value in scenario IV can be explained similar to the case for scenario III. Since some of the “low value” forest tiles that are not harvested in the early years are young stands, their volume growth over time contributes to the slightly higher volume and value of the remaining timber compared to the base case. Additionally, since the high value forest tiles that are harvested in the beginning of the simulation have high site indices, their volume growth over the years is faster compared to the base case where a mix of forest tiles are harvested in each period only based on their distance from the sawmills. However, both scenarios show an improvement in retaining the unit value of timber compared to the base case. This shows that the base case scenario degrades the value of the forest at a faster rate and the decline trend is sharper. In the base case scenario, it is possible for a mill to harvest young stands as long as they have a high ratio of high quality logs, therefore preventing them from growing in volume and value. By looking at the graphs for unit value of timber and profit levels, harvesting the highest value stands (scenario IV) seems to improve the sustainability of the timber resource without a strong negative impact on the industry.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  93 4.4.3 Impact of Removing Quality Requirements Impact on Supply Chain Performance Removing the requirements of not harvesting the stands with low ratio of high quality logs in the base case means that the mills would have to harvest the closest stands, regardless of the quality of the available timber. Therefore, they would potentially harvest more low quality logs and would have to use their production capacity to produce more commodity lumber and less high value products, resulting in lower profits. This is seen in Figure  4.9, comparing the profits and production capacities of the base case and scenario V (standard deviations for profits of the base case is presented in Appendix C-1). 0 100 200 300 400 500 600 700 800 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (th ou sa nd  m bf ) Time (years) Scenario VBase Case (a) (b) -80 -60 -40 -20 0 20 40 60 80 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years)  Figure  4.9 (a) Average total production capacity (േ1SD) of sawmills – Base case and scenario V, (b) Average total profits (േ1SD) of sawmills - Base case and scenario V Lower profits of sawmills in scenario V prevent them from expanding their production capacities as much as they would in the base case. The lower level of harvests and log imports in Figure  4.10 match the lower production capacity of the supply chain members. Lower profits in scenario V result in one more agent becoming insolvent compared to the base case (four agents in comparison to three in the base case). The additional agent goes bankrupt during the second half of the simulation. This can be a contributor to lower capacity and harvest levels towards the end of the simulation. Although lower profits of mills in scenario V prevents them from importing large volumes of log in the early years, the lower harvesting of high quality logs does force them to import high quality logs to  Chapter 4: Impact of Harvest Policy Changes on Sustainability  94 meet their production demand. More harvest of the low value logs also increase the production volume of the commodity lumber in scenario V, as shown in Figure  4.11. 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Scenario VBase Case (a) (b) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (m illi on  m 3 ) Time (years)  Figure  4.10 (a) Average total harvest (േ1SD) of sawmills - Base case and scenario V, (b) Average total volume (േ1SD) of imported logs - Base case and scenario V Removing the log quality requirement results in more un-harvested forest tiles (1700 compared with 1511 in the base case). The reason is fewer agents are present and the lower profits of the remaining agents prevent them from harvesting forest tiles that are distant and generate high transportation costs. 0 100 200 300 400 500 600 700 800 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (th ou sa nd  m bf ) Time (years) 0 100 200 300 400 500 600 700 800 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (th ou sa nd  m bf ) Time (years) High Value Lumber ProductionCommodity Lumber ProductionTotal Production Capacity Figure  4.11  (a) Average total production (േ1SD) of sawmills – Base case, (b) Average total production (േ1SD) of sawmills – Scenario V  Chapter 4: Impact of Harvest Policy Changes on Sustainability  95 Impact on Remaining Forest  Figure  4.12 shows the unit value of remaining timber in the forest when the quality requirement is removed. Scenario V shows a much slower declining trend in the unit value of timber compared to the base case. Additionally, the unit value grows back to its initial level towards the end of the simulation, showing the restoration of timber resources by changing the log quality requirement. This is a noteworthy result, since the harvest volume in the two scenarios is very close, meaning that although a relatively similar volume of timber remains standing in the forest, the base case scenario depletes the higher value stands and continually degrades the remaining forest.  Figure  4.12 Average unit value of remaining timber – Base case and scenario V Comparing all scenarios together, it seems that scenarios V and IV offer some benefits with regards to timber resource conservation, with less negative impact on the industry compared to other scenarios. Removing the harvest quality requirement (as in scenario V), which may be interpreted as banning “cherry-picking” the best stands, shows that if the mills had to harvest a mix of economic and uneconomic stands, the expansion of the industry would be more proportional with the quality of the available resources and the negative impacts on resource sustainability could be avoided. According to the results, the production capacity in scenario V does not decrease sharply, although the profits are notably lower than the base case. However, if financial motivations and subsidies are 80 81 82 83 84 85 86 87 88 89 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 U ni t V al ue  ($ /m 3 ) Time (years) Scenario V Base Case  Chapter 4: Impact of Harvest Policy Changes on Sustainability  96 present to promote a more balanced harvesting approach, the long time result may be a more stable forest industry with access to a higher value timber resource. A combination of AAC restrictions and alternative harvest priorities may also be beneficial. However, it is clear from the presented results that in order to maintain a profitable and growing forest industry, the timber harvesting land base will eventually have to be converted to a younger, second-growth forest. 4.5 Conclusion The shortage of high value timber is negatively impacting the industry in coastal BC since historical harvesting practices have degraded much of the natural timber resource of the region. Therefore, it is beneficial to investigate some alternative harvest policies and restrictions that could potentially improve the existing forest resources of this area for better serving the interests of both the public and the industry. An agent-based forest sector model, CAMBIUM 2.0, was used in this chapter to explore the impacts of four different harvest policies regarding the AAC levels and harvest priorities and requirements on the performance of a forest products supply chain in BC Coast, while monitoring the growth of volume and value of the remaining timber in the forest. The results showed that simple restrictions on harvest levels in the form of AAC reductions impacts the industry negatively while causing an increase in the volume and value of the remaining timber in the forest. This agrees with previous research. It may be inevitable for the industry to downsize during the timber shortage since the increasing cost of accessing the desired timber will force some firms out of business. Prioritizing older forest stands over younger ones increases the access costs substantially and will also impact the industry negatively in the near future, but higher timber availability in the long term will be beneficial to the industry, and allows for a stable volume and value of the remaining timber which is higher than the status quo. Prioritizing higher value stands over lower values ones results in immediate benefits to the industry, but causes a period of moderate downsizing in the future. However, this scenario still benefits timber resource sustainability more than the base case scenario.  Chapter 4: Impact of Harvest Policy Changes on Sustainability  97 Not allowing the mills to select the stands for harvest based on the quality forces them to harvest a mix of economic and uneconomic stands.  This has some negative impacts on the industry, but results in a more sustainable future compared to status quo. Considering the predicted initial financial loss to the industry, it may be worthwhile to subsidize the industry for early losses in order to build up the forest value for the future.  98 Chapter 5: Model Validation and Verification  Chapter 5. Model Validation and Verification  5.1 Introduction In this chapter steps taken for verification and validation of CAMBIUM 2.0 are presented. Verification process ensures that the model is correctly programmed and implemented and is error free, while the validation step ensures the model correctly represents the real world system and that it’s behaving as expected under various parameter changes. 5.2 Verification In order to ensure that no programming errors exist in the model, a simple test is performed to observe the relationship between inputs and output of the manufacturing agents in the model. The input volume of timber (harvested timber) is compared to the sum of output volume (products and waste material) of the agents to make sure all material is accounted for. Simultaneously, the costs and revenues are compared to make sure the profits are calculated correctly and according to the harvested and processed material. Since the verification process for all agents of the model is similar, the verification test of one sawmill is presented here. Similar tests have been performed for other agents. The selected sawmill is a high value mill, with annual production capacity of 132,000 mbf. Sawmill information at the beginning of the test period is presented in Table  5.1. All costs and prices are similar to what was presented in scenario one of Chapter 3.  Chapter 5:Model Validation and Verification  99 Table  5.1 Initial sawmill specifications Recovery Factor Capacity Initial Capital Variable Cost Fixed Cost 227.1123 (bft/m3) 32,000 (mbf/year) $29,472,000 260 ($/mbf) 1,280,000 ($/year) First, the required timber demand for meeting the production target of the mill is calculated. Timber Demand= Production Target×1000 Recovery Factor ൈሺ1-Pulp Log  Ratioሻ ൌ 32000 ×1000 227.11ൈ1-0.2؆ 176,126 m 3  ( 5.1)  During the harvesting step, the mill harvests one forest tile with a total high quality log harvest equal to the timber demand (176,12324 m3). However, since the forest tile has a mix of high and low quality logs (90% high quality and 10% low quality), some low quality logs are also harvested along with the high quality timber (19,570 m3). The transportation cost is calculated based on Equation ( 5.2) (forest tiles are indexed with i). The transportation cost per m3 per Kilometer is $0.15. Total Transportation Cost= Unit Transportation Cost ൈ (Harvest Volume from Tile ݅  ൈ Distance of Tile ݅ from Millሻ ൌ 0.15×(195,692×3.06ሻ ؆ $ 105,821 ( 5.2)  Total logging and stumpage costs are also calculated based on the harvested volume, as shown in Equation ( 5.3). Stumpage is set at $5 per m3 for the period and the logging cost is $40 per m3. Total Logging and Stumpage Cost= ሺUnit Logging Cost൅ Unit Stumpage Costሻ ൈ Total Harvest Volume ൌ 45× 195,692ൌ$ 8,806,151 ( 5.3)  During the manufacturing process, the harvested timber is converted into high value lumber. The sawmill has some low value logs in its inventory, but since it also has  23 The actual Recovery Factor (randomly drawn from a uniform distribution in the beginning of the simulation) is 227.113651. The results of the calculations are based on this unrounded number. 24 The calculated demand for high quality timber in the model is based on the unrounded recovery factor and is slightly different than the result in Equation (5.1).  Chapter 5:Model Validation and Verification  100 enough high value logs to utilize all of its production capacity, no low value lumber is manufactured in this time step. Only when low quality log inventories reach a specified limit (30% of total production capacity for this example), does the sawmill starts processing them. According to the reverse of Equation ( 5.1), total manufactured lumber is expected to be 32000 mbf, which is what the sawmill produces in the model. Total volume of pulp logs from the harvested timber can be calculated from Equation ( 5.4). Please note that the assumption of the model is that only the timber that is going to be processed into lumber is sorted in each step. Therefore, since no low quality lumber is to be manufactured in this period, only high quality timber is sorted into pulp logs and saw logs by the mill. Total Pulp Log Volume= (Pulp Log Ratio× High Quality Timber Volume) ൌ 0.2× 176,123 ؆ 35,224 m3 ( 5.4)  Total waste material volume (sawdust and chips from the manufacturing process) can also be calculated from the difference of the timber input volume and the lumber output volume. However, since lumber output is measured in mbf, it has to first be converted back to m3 using the volume conversion factor25 of 2.359737 m3/mbf. Total Waste Volumeൌ Total Processed Timber Volume – Total Pulp Log Volume   ሺTotal Volume of Manufactured Products ×Conversion Factor) ൌ 176,123-35,224-32000×2.359737 ൌ 65,387 ݉ଷ ( 5.5)  The manufacturing cost is calculated using the variable production cost in Table  5.1. Variable Production Costൌ Total Volume of Manufactured Products ൈ Variable Production Costs ൌ 32000×$260/mbf ൌ $8,320,000 ( 5.6)  Unit production cost for the lumber output of sawmill is calculated based on all harvesting (for high quality timber only), manufacturing and fixed operating costs.  25  This is a standard volume conversion factor, not a recovery factor dependent on sawmill equipment.  Chapter 5:Model Validation and Verification  101 Unit Production Costൌ ሺHarvesting Cost + Variable Production Cost ൅  Fixed Operating Costሻ/Production Volume ൌ ሺ$8,020,773 ൅ $8,320,00 + $1,280,000ሻ/32000mbf؆550 $/mbf ( 5.7)  This production cost and an agent-specific profit target are used to determine a reserve price. In this period the sawmill’s profit target is 100 $/mbf. This profit target changes every period based on the realized market price of the last period and the forecasted market price for the current period. If the agent forecasts the price will not be high enough to generate the desired profit target, it reduces the profit target accordingly. The reserve price is then calculated as the sum of production cost and the profit target, which is approximately 650 $/mbf in this example. During the market trading step, the final market price is determined by the highest reserve price among the offered products, as long as the highest reserve price is within the allowed fluctuation limit of the target price (which is externally modified and enforced, 700 $/mbf for this period in this example). The highest reserve price for high value lumber in this period is 700 $/mbf which is within the allowable range from the target price. Therefore, all agents, including the sawmill in this example, trade their products at this realized price. Total sales revenue for lumber products, pulp logs, and waste material is calculated below. The pulp logs are sold at the price of 35$/mbf and waste material are sold at 40$/mbf (approximately 17 $/m3). Total Lumber Sales Revenueൌ Production Volume ൈ Market Price ൌ 32000mbf ൈ $700/mbfൌ $22,400,000 ( 5.8)  Total Pulp Log Sales Revenueൌ Pulp Log Volume ൈ Market Price ൌ 35,224mbf ൈ $35/mbfൌ$1,232,861 ( 5.9)  Total Waste Material Sales Revenue=Waste Volume × Market Price ൌ 65,387m3 × $17/m3=$1,111,580 ( 5.10)  Final value for the working capital of the agent is expected to be equal to the output of Equation ( 5.11).  Chapter 5:Model Validation and Verification  102 Final Working Capital= Initial Working Capital – Costs + Revenues ൌ $29,472,000 – $17,620,773൅ $24,744,440 ൌ $ 36,595,667 ( 5.11)  The final value in the model output is slightly higher than this value ($36,596,076) which is because of rounding errors (the conversion factors and the market prices) in this chapter. Therefore, it is confirmed that the material flow through the simulation from forest to agents and then to the market does not leave any volume unaccounted for. Additionally, the profits and costs are calculated correctly based on market prices and cost parameters of agents. 5.3 Validation In order to ensure that the agents in CAMBIUM 2.0 are behaving similar to actual firms in a forest sector supply chain, simple tests are performed to study their reactions to changes in their environment. For example, an increase in production costs is expected to reduce the sawmill profits and cause more firm insolvencies in the model. Some of these behavior changes can be observed in Chapter 3 where cyclic prices change the growth and expansion trend of agents and impact their profits and production volumes. Some additional tests are performed in this chapter: 1) increasing the sawmill production costs, 2) increasing the product prices (no cyclic changes), 3) Decreasing the log prices, 4) lowering the log import limit, and 5) limiting the log export volume. The resulting change in model behavior is then investigated to see if it has changed as expected. All results are compared with the “base case” presented in Chapter 3 (scenario I, with the presence of a sort yard). Increasing Sawmill Production Costs For the first test, the variable production cost of all sawmills is increased by 50%. Based on the results presented for scenario one in Chapter 3, the commodity mills become insolvent in all of the simulation runs while all the high value mills remain profitable. It is expected that when the costs are increased, more agents have trouble keeping a positive cash flow and therefore become insolvent.  Chapter 5:Model Validation and Verification  103 Figure  5.1 shows the average number of agents and their observed minimum and maximum values for 50 simulation runs for the base case scenario and the scenario with increased production costs. As expected, Figure  5.1(b) shows that on average, two additional agents become insolvent compared to the base case scenario. Furthermore, sawmill profits are expected to decrease and mills are not expected to expand as quickly as they did in the base case. Figure  5.2 and Figure  5.3 show these expected trends. Sawmill profits are significantly lower when production costs are increased, as shown in Figure  5.2 (b). Since the sawmills can no longer expand, the sort yard has more access to timber and can expand its capacity and earn significantly higher profits through exports compared to the base case. Figure  5.3(b) shows that total production capacity of sawmills decreases in the first few periods, most probably because of sawmills becoming insolvent and then increases moderately throughout the simulation.  Figure  5.1 (a) Average number of active agents, base case, (b) Average number of active agents, increased production costs. Error bars show minimum and maximum observed number of agents. (a) (b) 3 4 5 6 7 8 9 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 N um be r o f A ge nt s Time (years) 3 4 5 6 7 8 9 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 N um be r o f A ge nt s Time (years)  Chapter 5:Model Validation and Verification  104  Figure  5.2  (a) Average profits of supply chain members (േ1SD), base case, (b) Average profits of supply chain members (േ1SD), increased production costs  Figure  5.3 (a) Average lumber production of sawmills (േ1SD), base case, (b) Average lumber production of sawmills (േ1SD), increased production costs Therefore, it is seen that the model behaves according to expectations with regards to production costs. Increasing Market Prices In order to test model response to market prices, the prices for all products are increased by 50% compared to the base case. Total agent profits are expected to increase and the agents are also expected to grow more rapidly. Combined Sawmill Profits Sort Yard Profits Total Profits (a) (b) -150 -100 -50 0 50 100 150 200 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) -150 -100 -50 0 50 100 150 200 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) High Value Lumber Production Commodity Lumber Production Total Lumber Production Capacity (a) (b) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m bf ) Time (years) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m bf ) Time (years)  Chapter 5:Model Validation and Verification  105 Figure  5.4(a) shows that the average number of agents is approximately the same as the base case, however the minimum observed number of agents during the later years of the simulation is higher compared to the base case. This shows that, on average, agents stay in the simulation longer compared to the base case. This is reasonable since higher profit means that agents will have less cash flow problems. Higher prices mean agents are more profitable and grow more rapidly, as shown in Figure  5.4(b). Since timber resources are scarce and the agent capacities are larger, the competition for timber is more intense compared to the base case. It is also expected that higher prices encourage the mills and the sort yard to harvest more timber and import more logs. Comparing the harvest levels of this scenario with the base case shows higher harvest levels in the first 25 years. However after year 25, when the harvestable timber becomes scarce, the harvest level in this scenario is approximately the same as the harvest volume in the base case. Log imports are also slightly higher compared to the base case but are limited by the upper bound of 400,000 m3. Since the difference between the two scenarios is very small, the graphs for harvest and log imports are not presented here.  Figure  5.4 (a) Average number of active agents, increased prices. Error bars show minimum and maximum observed number of agents, (b) Average lumber production of sawmills (േ1SD), increased prices Figure  5.5 shows that the profit changes as expected, with sawmill profits growing at a much higher rate compared to the base case. (a) (b) 3 4 5 6 7 8 9 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 N um be r o f A ge nt s Time (years) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m bf ) Time (years) High Value Lumber Production Commodity Lumber Production Total Lumber Production Capacity  Chapter 5:Model Validation and Verification  106  Figure  5.5 Average profits of supply chain members (േ1SD), increased prices Decreasing the Log Prices In order to test the impact of lower log purchasing costs on the behavior of sawmills and sort yard, the model was run with reduced log prices (a 30% reduction). It is expected that this lower price would encourage sawmills to buy logs from the sort yard. As can be seen from Figure  5.6(a) the export ratio is noticeably lower compared to the base case, meaning that the sort yard sells a higher proportion of its log output to the sawmills. Additionally, since the sawmills now have access to the logs that would have been exported in the base case, their production levels are higher, as shown in Figure  5.6(b). Harvest levels (not shown here) remain very close to the base case. 0 100 200 300 400 500 600 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Pr of its  (m ill io n $) Time (years) Combined Sawmill Profits Sort Yard Profits Total Profits  Chapter 5:Model Validation and Verification  107  Figure  5.6 (a) Average log export ratio (േ1SD) of the sort yard for the base case and the case with lower log prices, (b) Average production capacity (േ1SD) of the sort yard for the base case and the case with lower log prices Lowering the Log Import Limit Since the mills depend on log imports for meeting their timber demand when harvestable timber is scarce, it is expected that limiting the import volume impacts their profits and production levels negatively. Additionally it may be expected that lowering the import limit pushes the mills to buy logs from the sort yard. In order to see if the model behaves according to these expectations, the import limit is reduced by 50% (to 200,000 m3per year). The harvest and profit levels are very similar to the base case, while production levels are slightly lower as shown in Figure  5.7. The reason for lower production level when the import limit is reduced is the lower availability of logs to be processed into lumber products. The reason for similar profits levels is that although the mills prefer to import high quality logs and manufacture specialty lumber products, purchasing these imported logs is more expensive compared to harvesting. Therefore, the lost revenue from not producing higher volume of specialty lumber products is offset by the lower cost of purchasing the logs and processing them. Additionally, the lower production capacity compared to the base case means lower fixed costs which also partially offset that lost revenue. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Lo g E xp or t r at io  (% ) Time (years) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m bf ) Time (years) Lower Log Profits Base Case (a) (b)  Chapter 5:Model Validation and Verification  108  Figure  5.7 Average production volume (േ1SD) of high value lumber for the base case and the case with reduced import limit Limiting the Log Export Volume The final test is performed to see what happens if there was a limit on sort yard’s log exports. Since the decision of the mills to buy logs from the sort yard is only affected by the total purchase cost, limiting the export would not impact the volume of the logs purchased by the mills. However, it is expected that lower levels of export would negatively impact the sort yard since in the base case the sort yard always has the option of selling the logs to external destinations. Limiting the export volume would result in lower export revenue and since no additional domestic revenue is anticipated, the sort yard would not be able to expand its capacity as much as the base case. In order to implement this scenario, an export limit of 200,000 m3 was enforced on both high and low quality logs, since they are modelled as two different products. Figure  5.8(a) shows the profit of the sort yard with and without the export limit. As expected, profit is much lower when the exports are limited. Figure  5.8(b) shows that when the exports are limited, the log output of the sort yard remains stable at a much lower level compared to the base case since it does not generate enough profits to expand its capacity. The harvest levels are very similar to the base case, but the production capacity of the mills is higher and does not show the declining trend seen in the base case (graph not shown here). This is because the sort yard does not expand as much as it does in the 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 V ol um e (m ill io n m 3 )  Time (years) Lower Log Import Limit Base Case  Chapter 5:Model Validation and Verification  109 base case. This means higher availability of timber for the sawmills which help them expand their production capacities.  Figure  5.8 (a) Average sort yard profits (േ1SD) for the base case and the case with limited exports, (b) Average log output (േ1SD) of the sort yard for the base case and the case with limited exports Comparison to Forest Products Industry on the BC Coast Since the cost and operational data for the coastal BC forest products supply chain in this dissertation were not acquired directly from the companies, and considering the resource inventory data were projected from aggregate information rather than actual volumes on the ground, it is not possible to test the model outcomes against actual profit or harvest values of existing companies. However, CAMBIUM 2.0 clearly reflects the general conditions of the forest sector on the BC Coast which has seen many commodity mills shutting down over the past decade (Pearse, 2001). Additionally, it has previously been suggested that high value wood products (including specialty lumber products modelled in CAMBIUM 2.0) have a promising outlook on the BC Coast (International Wood Markets Group, 2007). Furthermore, a shortage of timber resources on the coast, which arises in the simulation model, has been pointed out (Marshall, 2002; The Working Forest, 2011).  Therefore, it is reasonable to believe that CAMBIUM 2.0 correctly represents the operations of a typical forest products supply chain on the coast. 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e  (m ill io n m 3) Time (years) 0 10 20 30 40 50 60 70 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) Limited Exports Base Case (a) (b) 110 Chapter 6: Conclusions  Chapter 6. Conclusions   6.1 Conclusions In British Columbia (BC), the forest industry is a large part of the economy, but it faces many challenges from decreasing value and volume of manufacturing shipments and a declining employment trend (Statistics Canada, 2011a, 2011b). Lack of capital investments and outdated manufacturing equipment, particularly in the coastal BC forest sector, has resulted in high production costs and low profitability of the forest industry in this region. Investment in new operating strategies and modifying management policies may improve the sector’s performance in the long run. Supply chain management can be a helpful tool in evaluating various investment decisions and operating strategies by identifying best practices and providing answers to “what-if” questions. This thesis aimed to develop a decision support tool that could be tailored to the case of the forest sector on the BC Coast and would help policy makers and managers in examining the impact of different configurations of the supply chain and changing policies. CAMBIUM 2.0 was developed based on CAMBIUM, a multi-agent forest sector model (Schwab, 2008). In this thesis, CAMBIUM 2.0 was used to answer the following three research questions: 1) What would be the economic impact of establishing a new log sorting facility within the existing supply chain of the coastal BC forest industry? 2) How would a change in harvest policies impact the performance of the supply chain members? 3) How would a change in harvest policies impact the state of forest resources in the region?  Chapter 6: Conclusions  111 In CAMBIUM 2.0, agents interact with each other under predefined rules and use the available forest resources in order to manufacture products and compete in the market. At each time step of the simulation, agents perform planning, harvesting, manufacturing and distribution activities. They also choose an investment strategy in each period, learning from their past experiences and adjusting their strategy choices accordingly. Their learning behavior was modelled using the EWA-Lite learning algorithm (Ho et al., 2001). The model has been programmed in Java, using the Recursive Porous Agent Simulation Toolkit for Java (Repast J) and is a grid-based spatial model where grids contain information on either forest resources, or active economic agents. Major improvements in CAMBIUM 2.0 compared to the previous version have been: (1) lumber product differentiation based on log quality, (2) sawmill differentiation based on the lumber product output, (3) adding a facility location module to allow new agent entry, (4) modelling sort yard operations as a new agent type, and (5) allowing log import and export activities. The contributions of the thesis were presented in three chapters. The second chapter presented and discussed the formulation and algorithm for integrating the facility location module within the agent-based simulation model. The facility location optimization problem was a mixed integer program, solved in every time step of the simulation until the expected net present value of establishing the new facility would be positive and the new agent would enter the simulation.  CAMBIUM 2.0 was applied to the case of a forest supply chain on the BC Coast - with eight sawmills and six Tree Farm Licenses (TFLs) - in order to establish a new sort yard in one of the three potential locations on Vancouver Island. After the new agent was located, its location decision and the optimal profits and the flow of logs and products in the solution were compared to the observed profits and material flow during the simulation. It was seen that access to correct market forecasts significantly improved the predictive abilities of the facility location module, while access to competitor cost structure did not improve the results as much. The location decision in this specific case study was not sensitive to the level of available information. It was also observed that the optimization problem represented a simplified version of the supply chain operations and could not predict the capacity expansions or occasional fluctuations of the profits that would happen during the simulation.  Chapter 6: Conclusions  112 Log sort yards have been recommended for BC as the means for diversifying and strengthening small wood-manufacturing businesses, and bridging the gap between wood suppliers and wood users (Sunderman, 2003). In order to answer the first research question of this thesis, the third chapter evaluates the impact of establishing a log sort yard on the BC Coast. Two market price scenarios were considered to investigate the performance of the supply chain under different conditions, with and without the presence of a sort yard. Confirming the findings of previous research, the addition of a sort yard did not improve the profitability of the supply chain as a whole. The main reason for this was identified to be the limited forest resources that all agents needed to access. The timber shortage meant that adding a new agent would increase the competition for this limited resource and would intensify the impact of the shortage on the existing agents. Another observation from this study was that low stumpage prices, combined with high costs of purchasing the logs from the sort yard caused sawmills to prefer harvesting their timber rather than buying logs from the sort yard. Therefore, the sort yard exported most of its log output to external destinations which may not be a desirable outcome for local communities and businesses. In order to ensure the accuracy of the model, steps were taken to verify and validate it by testing different scenarios and observing the change in model behavior. These steps are presented in Chapter 5. Although efforts are being made by the BC government to meet the public’s growing demand for sustainable management of timber and non-timber forest values, some argue that current policies, especially in setting Annual Allowable Cut (AAC) levels, are leading to rapid exhaustion of the province’s natural timber reserve. Since it is expected that timber supply of the province will decline over the short term, both the forest industry and the government must be prepared to face this shortage and try to design and implement policies that would efficiently make use of available timber resources. Efforts must be made to ensure that future forest stands can support long term timber needs of the province, regarding both the resource volume and value. This may be done through different policy instruments such as adjusting AAC levels or implementing different harvest priorities. The second and third research questions of this thesis were addressed in the fourth chapter and investigated the impact of changes to resource availability and harvest priorities as well as the sustainability of the timber resource. The results showed that AAC reductions would impact the industry  Chapter 6: Conclusions  113 negatively while causing an increase in the volume and value of the remaining timber in the forest. It may be inevitable for the industry to downsize during the timber shortage period since the increasing cost of accessing the desired timber will force some forest companies out of business. It was also seen that prioritizing older forest stands for harvesting impacts the industry negatively in the near future, but results in long run benefits because of higher timber availability in the future. Prioritizing higher value stands over lower values ones results in immediate benefits to the industry, but causes a period of downsizing in the future. However, this scenario still benefits the remaining value of the forest more than the status quo scenario (harvesting the closest stands first). Removing the requirement that mills only harvest specific log qualities forced the mills to harvest a mix of forest stands that generated different levels of profits (both high and low). This noticeably improved the sustainability of the timber resource compared to status quo and allowed the unit value of the remaining timber to grow and return to its initial level as the simulation progressed. The analyses performed in this thesis show the potential of CAMBIUM 2.0 for further investigating supply chain investment decisions and regulatory changes. The interaction of agents creates the structure of the supply chain from the bottom up and allows the decision makers to easily observe the impact of their policies on each individual economic player, as well as the entire system. 6.2 Limitations The findings of this research are impacted by the dataset of the available forest inventory in the selected areas. This dataset was constructed from aggregate data that was publicly available. Some of these public reports had not been updated in two or three years and therefore did not correctly represent the current state of the forest. Additionally, while the inventory projection algorithm generated a landscape that matched the aggregate inventory data, the layout of the forest tiles was only one realisation from a large number of potential layouts. A different layout of the forest tiles would generate different transportation costs from the forest to the mills, especially when harvest priorities are changed in Chapter 4. The same logic applies to estimated costs and product recovery factors of the agents. The profitability of the agents is dependent on these assumptions, although the impact of these estimations is not as extensive as the forest inventory data.  Chapter 6: Conclusions  114 Road network maps of the modelled area were not available for the purpose of this research and the distances between different tiles in the model were represented with Euclidean distances. Including the existing road network in the region could greatly impact the calculated transportation costs and the final outcome of the model should be evaluated considering this issue. Additionally, increasing the number of potential sites for the new agent (or allowing a continuous search of the modelled area) could identify areas that are most appropriate for establishing the new facility which can be refined later based on managerial preferences and feasibility issues. As discussed in Chapter 2, the optimization problem solved by the candidate agent is a simplified version of what goes on in each simulation time step. Introducing the dynamic and stochastic elements of the simulation (e.g. potential growth of agents, successful implementation of business strategies, and random order of accessing the forest tiles) into the optimization problem would generate improvements in the predictive abilities of the candidate agent and could impact the final location decision. However, this may cause computational complications in finding a feasible and optimal solution for the problem in reasonable time, which may need to be addressed as well by using heuristic solution approaches. The log import assumption of the model currently forces a limit on the total volume of the imported logs (which are assumed to come from adjacent forest areas). However, it may be more realistic to assign a portion of this import limit to high value logs and the rest to low value logs. This way, the mills could not use their entire import allowance to purchase high value logs from external sources, which would mean that the adjacent forest areas also have limited availability of high value logs. 6.3 Future Work CAMBIUM 2.0 can be further extended and improved in a number of ways. The integrated optimization problem for locating new facilities can be extended to include more information on the expansion and closing options of agents which could improve the predictive abilities of the new agent that is entering the simulation. However, this may cause computational  Chapter 6: Conclusions  115 complications in finding a feasible and optimal solution for the problem in reasonable time, which may need to be addressed as well by using heuristic solution approaches. Improving the quality of the input data, especially the resource inventory dataset, could greatly increase the applicability of the results for specific forest companies. Modelling the correct location of old growth stands for example could affect the cost of accessing these stands which in turn would impact the profitability of the agents. A more accurate representation of the market demand for each type of product (instead of the current linear relationship between price and quantity), especially at the regional level could also improve the results. CAMBIUM 2.0 can be used to investigate the impact of adding new agent types to the supply chain, such as bio-energy plants, similar to the work presented in Chapter 3. Also, further details in representing lumber products (different species or dimensions) can identify the products that have promising future economic outlooks based on the availability of resources and expected changes in the market. Regarding the state of the forest resources, it may be beneficial to investigate different forest regeneration strategies. For example, currently it is assumed that the forest tiles are regenerated using the same initial species. What would happen if different species were planted (depending on the geographical location) that were more desirable economically? Different issues regarding the behaviour of agents can be addressed with CAMBIUM 2.0. For example, what would happen if the mills were also allowed to export logs? How would changes in the log export and import limits impact the performance of the supply chain? Or how would increasing the stumpage costs impact the overall structure of the industry? Finally, additional strategic research questions arise from the results of this dissertation. What policies should be implemented to encourage sawmills to purchase logs from the sort yard? When is an appropriate time to introduce a new mill into the supply chain, considering the state of the remaining forest? What would be the impact of the sort yard on local value added businesses in the region? By changing assumptions on the interaction of agents and imposing different restrictions on their activities, CAMBIUM 2.0 may be used to address these and other strategic questions  Chapter 6: Conclusions  116 regarding the performance of the forest industry supply chains on the BC Coast, or in other regions.  117 References  Adams, D. M., & Haynes, R. W. (1980). The 1980 softwood timber assessment market model: Structure, projections, and policy simulations. Forest Science (Monograph 22), 26(3), a0001-z0001. Angerhofer, B. J., & Angelides, M. C. (2000). System dynamics modelling in supply chain management: Research review. Paper presented at the 32nd Conference on Winter Simulation, Orlando, Florida. Baesler, F. F., Araya, E., Ramis, F. J., & Sepulveda, J. A. (2004). The use of simulation and design of experiments for productivity improvement in the sawmill industry. Paper presented at the 36th Conference on Winter Simulation, Washington, D.C. BC Competition Council. (2006). Wood products industry advisory commitee report to the council. Vancouver, BC: BC Competition Council. BC Ministry of Forests, Lands and Natural Resource Operations. (2011a). Current allowable annual cut (aac) for tree farm licences (tfl). Retrieved 1 March 2011, from http://www.for.gov.bc.ca/hts/tfls.htm BC Ministry of Forests, Lands and Natural Resource Operations. (2011b). Forest analysis and inventory. Retrieved 10 March 2011, from http://www.for.gov.bc.ca/ hts/analysis.html BC Ministry of Forests, Lands and Natural Resource Operations. (1997). Glossary of statistical reporting terms. Retrieved 9 July 2011, from http://www.for.gov.bc.ca/hts/ inventory/reports/glossary/index.html#t.htm BC Ministry of Forests, Lands and Natural Resource Operations. (2002). Growth and yield modelling models: Models decision tree flow chart. Retrieved 21 January 2011, from http://www.for.gov.bc.ca/hre/gymodels/GY-Model/flochart.htm BC Ministry of Forests, Lands and Natural Resource Operations. (2003). British Columbia's forests: A geographical snapshot. Retrieved 1 Aug 2011, from http://www.for.gov.bc.ca/hfd/pubs/docs/mr/mr112/pdf.htm BC Ministry of Forests, Lands and Natural Resource Operations. (2006). Timber tenures in British Columbia: Managing public forests in the public interest. Retrieved 1 Aug 2011, from http://www.for.gov.bc.ca/ftp/hth/external/!publish/web/timber-tenures/ timber-tenures-2006.pdf BC Ministry of Forests, Lands and Natural Resource Operations. (2010). The state of British Columbia's forests (third edition). Retrieved 1 Aug 2011, from http:// www.for.gov.bc.ca/hfp/sof/#2010_report BC Ministry of Forests, Lands and Natural Resource Operations. (2011c). Log market reports, coast. Retrieved 1 July 2011, from http://www.for.gov.bc.ca/hva/ logreports.htm References  118 BC Ministry of Forests, Lands and Natural Resource Operations. (2011d). Market pricing system (mps) parameters, coast: July 2011. Retrieved 1 July 2011, from http://www.for.gov.bc.ca/hva/parameters.htm Beamon, B. M. (1998). Supply chain design and analysis: Models and methods. International Journal of Production Economics, 55(3), 281-294. Beaudoin, D., LeBel, L., & Frayret, J.-M. (2007). Tactical supply chain planning in the forest products industry through optimization and scenario-based analysis. Canadian Journal of Forest Research, 37(1), 128-140. Benders, J. F. (1962). Partitioning procedures for solving mixed variables programming problems. Numersiche Mathematik, 4(1), 238-252. Bilek, T. (2004). Lsy: Documentation for a spreadsheet tool to evaluate log-sort yard economics (General Technical Report No. FPL–GTR–184). Madison, WI: USDA Forest Service, Forest Products Laboratoty. Binkley, C. S., Percy, M., Thompson, W. A., & Vertinsky, I. B. (1994). A general equilibrium analysis of the economic impact of a reduction in harvest levels in British Columbia. The Forestry Chronicle, 70(4), 449-454. Borges, J. G., Hoganson, H. M., & Rose, D. W. (1999). Combining a decomposition strategy with dynamic programming to solve spatially constrained forest management scheduling problems. Forest Science, 45(2), 201-212. Borshchev, A., & Filippov, A. (2004). From system dynamics and discrete event to practical agent based modeling: Reasons, techniques, tools. Paper presented at the 22nd International Conference of the System Dynamics Society, Oxford, England. Bredstrom, D., Lundgren, J. T., Ronnqvist, M., Carlsson, D., & Mason, A. (2004). Supply chain optimization in the pulp mill industry - ip models, column generation and novel constraint branches. European Journal of Operational Research, 156(1), 2-22. Broad, L. R. (1989). Note on log sort yard location problems. Forest Science, 35(2), 640-645. Buongiorno, J., Zhu, S., Zhang, D., Turner, J., & Tomberlin, D. (2003). The global forest products model: Structure, estimation, and applications. San Diego, CA: Academic Press. Burda, C., Gale, F., & M'gonigle, M. (1998). Ecoforestry versus the state(us) quo: Or why innovative forestry is neither contemplated nor permitted within the state structure of British Columbia. BC Studies: The British Columbian Quarterly, 119, 45-86. Cashore, B., Hoberg, G., Hewlett, M., Rayner, J., & Wilson, J. (2000). Fine tuning the settings: The timber supply review (chapter 6). In In search of sustainability: British columbia forest policy in the 1990s. Vancouver, BC: UBC Press. Canadian Council of Forest Ministers (CCFM). (1997). Criteria and indicators of sustainable forest management in canada: Technical report. Retrieved 30 July 2011, from http://cfs.nrcan.gc.ca/pubwarehouse/pdfs/10350.pdf References  119 Canadian Council of Forest Ministers (CCFM). (2003). Defining sustainable forest management in canada: Criteria and indicators. Retrieved 30 July 2011, from http://www.ccfm.org/pdf/CI_Booklet_e.pdf Chan, F. T. S., & Chan, H. K. (2005). The future trend on system-wide modelling in supply chain studies. The International Journal of Advanced Manufacturing Technology, 25(7), 820-832. Chauhan, S. S., Frayret, J.-M., & LeBel, L. (2009). Multi-commodity supply network planning in the forest supply chain. European Journal of Operational Research, 196(2), 688-696. Church, R. (2007). Tactical-level forest management models. In Andres Weintraub, Carlos Romero, Trond Bjørndal, Rafael Epstein & J. Miranda (Eds.), Handbook of operations research in natural resources (Vol. 99, pp. 343-363). New York, NY: Springer. Church, R., Murray, A. T., & Weintraub, A. (1998). Locational issues in forest management. Location Science, 6(1-4), 137-153. Coast Forest Products Association. (2003). BC coast report. Retrieved 25 July 2011, from http://www.coastforest.org/pdf/bc_coast_report.pdf Cohen, M. A., & Lee, H. L. (1989). Resource deployment analysis of global manufacturing and distribution networks. Journal of Manufacturing and Operations Management, 2, 81-104. TimberWest Forest Corp. (2010). Annual report. Retrieved 4 July 2011, from http://www.timberwest.com/documents/Investors/SEDAR%202010%20ANNUAL% 20REPORT.pdf Current, J., Min, H., & Schilling, D. (1990). Multiobjective analysis of facility location decisions. European Journal of Operational Research, 49(3), 295-307. D'Amours, S., Ronnqvist, M., & Weintraub, A. (2008). Using operational research for supply chain planning in the forest products industry. INFOR: Information Systems and Operational Research, 46(4), 265-281. Daskin, M. S., Snyder, L. V., & Berger, R. T. (2005). Facility location in supply chain design. In A. Langevin & D. Riopel (Eds.), Logistics systems: Design and optimization (pp. 39-65). New York, NY: Springer. Daugherty, P. J., & Fried, J. S. (2007). Jointly optimizing selection of fuel treatments and siting of forest biomass-based energy production facilities for landscape-scale fire hazard reduction. INFOR: Information Systems and Operational Research, 45, Number(1), 17-30. Dramm, J. R., Govett, R., Bilek, T., & Jackson, G. L. (2004). Log sort yard economics, planning, and feasibility (General Technical Report No. FPL−GTR−146). Madison, WI: USDA Forest Service, Forest Products Laboratoty. Dramm, J. R., Jackson, G. L., & Wong, J. (2002). Review of log sort yards (General Technical Report No. FPL-GTR-132). Madison, WI: USDA Forest Service, Forest Products Laboratoty. References  120 Drezner, Z., & Hamacher, H. (Eds.). (2002). Facility location: Applications and theory. Berlin, Germany: Springer-Verlag. Dudek, G., & Stadtler, H. (2005). Negotiation-based collaborative planning between supply chains partners. European Journal of Operational Research, 163(3), 668-687. Eiselt, H. A., Laporte, G., & Thisse, J. F. (1993). Competitive location models: A framework and bibliography. Transportation Science, 27(1), 44-54. Fjeld, D. E. (2001). The wood supply game as an educational application for simulationg dynamics in the forest sector. In K. Sjostrom & L.-O. Rask (Eds.), Supply chain management for paper and timber industries (pp. 241-251). Växjö (Sweden). Forest Practices Board. (2009). High retention harvesting and timber sustainability on the British Columbia coast (Special Investigation Report No. FPB/SIR/20). Victoria, BC: Forest Practices Board. Forget, P., D'Amours, S., & Frayret, J.-M. (2008). Multi-behavior agent model for planning in supply chains: An application to the lumber industry. Robotics and Computer- Integrated Manufacturing, 24(5), 664-679. Forget, P., D'Amours, S., Frayret, J.-M., & Gaudreault, J. (2009). Study of the performance of multi-behaviour agents for supply chain planning. Computers in Industry, 60(9), 698-708. Frayret, J. M., D’Amours, S., Rousseau, A., Harvey, S., & Gaudreault, J. (2007). Agent- based supply-chain planning in the forest products industry. International Journal of Flexible Manufacturing Systems, 19(4), 358-391. Fu, M. C. (2002). Optimization for simulation: Theory vs.Practice. INFORMS Journal on Computing, 14(3), 192-215. Gautier, A., Lamond, B. F., Pare, D., & Rouleau, F. (2000). The quebec ministry of natural resources uses linear programming to to understand the wood-fiber market. Interfaces, 30(6), 32-48. Gibson, R., Poon, J., Nikolakis, W., & Cohen, D. (2009). Market fit analysis of coastal hemlock: A comparison between coastal British Columbia, washington and oregon (Technical Report No. P.02). Vancouver, BC: FPInnovations. Gjerdrum, J., Shah, N., & Papageorgiou, L. G. (2001). A combined optimization and agent- based approach to supply chain modelling and performance assesment. Production Planning and Control, 12(1), 81-88. Gunn, E. A. (2007). Models for strategic forest management. In A. Weintraub, C. Romero, T. Bjørndal, R. Epstein & J. Miranda (Eds.), Handbook of operations research in natural resources (Vol. 99, pp. 317-341). New York, NY: Springer. Gunnarsson, H., Rönnqvist, M., & Carlsson, D. (2006). A combined terminal location and ship routing problem. Journal of the Operational Research Society, 57, 928–938. Gurobi Optimization. (2011). Gurobi optimizer, java library interface (Version 3.0.3). Houston, TX: Gurobi Optimization. References  121 Gutierrez, G. J., Kouvelis, P., & Kurawala, A. A. (1996). A robustness approach to uncapacitated network design problems. European Journal of Operational Research, 94(2), 362-376. Haartveit, E. Y., & Fjeld, D. E. (2002). Experimenting with industrial dynamics in the forest sector – a beer game application. Paper presented at the Symposium on Systems and Models in Forestry, Punta de Tralca, Chile. Hamilton, G. (2011, 12 May). Forest sector profitable again, but remains on life support. Vancouver Sun. Han, H.-S., Bilek, T., Dramm, R., Loeffler, D., & Calkin, D. (2009). Financial feasibility of a log sort yard handling small-diameter logs. Paper presented at the 32nd Annual Meeting of the Council on Forest Engineering (COFE): environmentally sound forest operations. June 15-18, 2009, Kings Beach, California. Hickey, G. M., & Innes, J. L. (2008). Indicators for demonstrating sustainable forest management in British Columbia, canada: An international review. Ecological Indicators, 8(2), 131-140. Ho, T. H., Camerer, C. F., & Chong, J. K. (2001). Economic value of EWA Lite: A functional theory of learning in games (Social Science Working Paper No. 1122). Pasadena, CA: California Institute of Technology. Horne, G., Paul, N., & Riley., D. (1991). The provincial economic impacts of a supply reduction in the British Columbia forest sector. In Forest resource commission background papers, volume 4. Victoria, BC: Planning and Statistics Division, Ministry of Finance and Corporate Relations. Hultqvist, D., & Olsson, L. (2004). Demand based tactical planning of the roundwood supply chain with respect to stochastic disturbances (Report No. FSCN R-03-44). Sundsvall, Sweden: Fiber Science and Communication Network (FSCN). International Wood Markets Group. (2007). B.C. Coast strategic options: Current business, future opportunities and outlook to 2020. Vancouver, BC: Coast Forest Products Association. Jones, A., Seville, D., & Meadows, D. (2002). Resource sustainability in commodity systems: The sawmill industry in the northern forest. System Dynamics Review, 18(2), 171-204. Jung, J. Y., Blau, G., Pekny, J. F., Reklaitis, G. V., & Eversdyk, D. (2004). A simulation based optimization approach to supply chain management under demand uncertainty. Computers & Chemical Engineering, 28(10), 2087-2106. Kaihara, T. (2001). Supply chain management with market economics. International Journal of Production Economics, 73(1), 5-14. Kaihara, T. (2003). Multi-agent based supply chain modelling with dynamic environment. International Journal of Production Economics, 85(2), 263-269. Kallio, A. M., Dykstra, D. P., & Binkley, C. S. (Eds.). (1987). The global forest sector: An analytical perspective. New York, NY: John Wiley & Sons. References  122 Kallio, A. M., Moiseyev, A., & Solberg, B. (2004). The global forest sector model EFI- GTM-the model structure (Internal report No. 15). Joensuu, Finland: European Forest Institute (EFI). Kiss, L. (2010). More policy. Coastal Clarion, Coast Forest Products Association Newsletter, 6(2), 3-4. Kleijnen, J. P. C. (2005). Supply chain simulation tools and techniques: A survey. International Journal of Simulation & Process Modelling, 1(1/2), 82-89. Klose, A., & Drexl, A. (2005). Facility location models for distribution system design. European Journal of Operational Research, 162(1), 4-29. Lax, L., & Parker, J. (1992). The economic impacts of selected timber harvest scenarios in British Columbia. In Forest resource commission background papers, volume 4. Victoria, BC: Planning and Statistics Division, Ministry of Finance and Corporate Relations. Lee, W. B., & Lau, H. C. W. (1999). Multi-agent modeling of dispersed manufacturing networks. Expert Systems with Applications, 16(3), 297-306. Lee, Y. H., Cho, M. K., Kim, S. J., & Kim, Y. B. (2002). Supply chain simulation with discrete-continuous combined modeling. Computers & Industrial Engineering, 43(1- 2), 375-392. Lemieux, S., D’Amours, S., Gaudreault, J., & Frayret, J. M. (2009). Agent-based simulation to anticipate impacts of tactical supply chain decision-making in the lumber industry (Working Paper No. CIRRELT-2009-51). Montreal, QC: Interuniversity Research Centre on Enterprise Networks, Logistics and Transportation (CIRRELT). Lin, F. R., Tan, G. W., & Shaw, M. J. (1998). Modeling supply chain networks by a multi- agent system, Thirty-First Annual Hawaii International Conference on System Sciences. Kohala Coast, HI, USA. Lin, W., Kline, D. E., Araman, P. A., & Wiedenbeck, J. K. (1995). Design and evaluation of log-to-dimension manufacturing systems using system simulation. Forest Products Journal, 45(3), 37-44. Lonnstedt, L. (1986). A dynamic forest sector model with a swedish case. Forest Science, 32(2), 377-397. Louveaux, F. V. (1993). Stochastic location analysis. Location Science, 1(2), 127-154. Luckert, M. K., & Haley, D. (1995). The allowable cut effect as a policy instrument in canadian forestry. Canadian Journal of Forest Research, 25(11), 1821-1829. Maness, T. C., & Adams, D. M. (1991). The combined optmization of log bucking and sawing strategies. Wood and Fiber Science, 23(2), 296-314. Marshall, D. (2002). Down the value chain: The politics and economics of raw log exports (Policy Brief). Vancouver, BC: Canadian Center for Policy Alternatives (CCPA). Masui, T. (2005). Policy evaluations under environmental constraints using a computable general equilibrium model. European Journal of Operational Research, 166(3), 843- 855. References  123 McDill, M. E., Rebain, S. A., & Braze, J. (2002). Harvest scheduling with area-based adjacency constraints. Forest Science, 48(4), 631-642. Meixell, M. J., & Gargeya, V. B. (2005). Global supply chain design: A literature review and critique. Transportation Research Part E: Logistics and Transportation Review, 41(6), 531-550. Mele, F., Guillen, G., Espuna, A., & Puigjaner, L. (2006). A simulation-based optimization framework for parameter optimization of supply-chain networks. Industrial and Engineering Chemistry Research, 45(9), 3133-3148. Melo, M. T., Nickel, S., & Saldanha-da-Gama, F. (2009). Facility location and supply chain management - a review. European Journal of Operational Research, 196(2), 401- 412. Mendoza, G. A., Meimban, R. J., Luppold, W. G., & Araman, P. A. (1991). Combining simulation and optimization models for hardwood lumber production. Paper presented at the SAP National Convention, August 4-7, San Francisco, CA. Meng, Q., Huang, Y., & Cheu, R. L. (2009). Competitive facility location on decentralized supply chains. European Journal of Operational Research, 196(2), 487-499. Min, H., & Zhou, G. (2002). Supply chain modeling: Past, present and future. Computers & Industrial Engineering, 43(1-2), 231-249. Mirhassani, S. A., Lucas, C., Mitra, G., Messina, E., & Poojari, C. A. (2000). Computational solution of capacity planning models under uncertainty. Parallel Computing, 26(5), 511-538. Mitra, S., & Webster, S. (2008). Competition in remanufacturing and the effects of government subsidies. International Journal of Production Economics, 111(2), 287- 298. Monostori, L., Váncza, J., & Kumara, S. R. T. (2006). Agent-based systems for manufacturing. CIRP Annals - Manufacturing Technology, 55(2), 697-720. Montreal Process. (1995). Criteria and indicators for the conservation and sustainable management of temperate and boreal forests. Retrieved 1 Aug 2011, from http://www.rinya.maff.go.jp/mpci/rep-pub/1995/santiago_e.html#1 Moyaux, T., Chaib-draa, B., & D'Amours, S. (2004). An agent simulation model for the quebec forest supply chain. Paper presented at the International Workshop on Cooperative Information Agents No. 8, Erfurt, Germany. Natural Resources Canada. (2007). Canada's forests - glossary. Retrieved 1 Aug 2011, from http://canadaforests.nrcan.gc.ca/glossary/ North, M. J., Collier, N. T., & Vos, J. R. (2006). Experiences creating three implementations of the repast agent modeling toolkit. ACM Transactions on Modeling and Computer Simulation, 16(1), 1-25. Owen, S. H., & Daskin, M. S. (1998). Strategic facility location: A review. European Journal of Operational Research, 111(3), 423-447. References  124 Parunak, H. V. D., Savit, R., & Riolo, R. L. (1998). Agent-based modeling vs. Equation- based modeling: A case study and users’ guide. In J. S. Sichman, R. Conte & N. Gilbert (Eds.), Proceedings of the first international workshop on multi-agent systems and agent-based simulation (pp. 10-25). Berlin, Germany: Springer Verlag. Pearse, P. H. (1976). Timber rights and forest policy in British Columbia: Report of the royal commission on forest resources. Queen’s Printer, Victoria, BC. Pearse, P. H. (2001). Ready for change: Crisis and opportunity in the coast forest industry. Vancouver, BC: BC Ministry of Forests. Pedersen, Larry. (2003). Allowable annual cuts in British Columbia:The agony and the ecstasy. Retrieved 20 July 2011, from http://www.for.gov.bc.ca/hts/ pubs/jubilee_ubc.pdf Plastria, F. (2001). Static competitive facility location: An overview of optimisation approaches. European Journal of Operational Research, 129(3), 461-470. Prudham, S. (2007). Sustaining sustained yield: Class, politics, and post-war forest regulation in British Columbia. Environment and Planning D: Society and Space, 25(2), 258- 283. Randhawa, S. U., Brunner, C. C., Funck, J. W., & Zhang, G. (1994). A discrete-event object- oriented modeling environment for sawmill simulation. Simulation, 62(2), 119-130. Random Lengths. (2011). Monthly composite prices. Retrieved 10 July 2011, from http://www.randomlengths.com/base.asp?s1=In_Depth&s2=Useful_Data&s3=Monthl y_Composite_Prices Ronnqvist, M. (2003). Optimization in forestry. Mathematical Programming, 97(1), 267-284. Ronnqvist, M., & Ryan, D. (1995). Solving truck despatch problem in real time. Paper presented at the 31st Annual Conference of the Operational Research Society of New Zealand, August 31-September 1, Wellington, New Zealand. Santoso, T., Ahmed, S., Goetschalckx, M., & Shapiro, A. (2005). A stochastic programming approach for supply chain network design under uncertainty. European Journal of Operational Research, 167(1), 96-115. Schadendorf, C., & Chiocca, O. R. (1997). Do log sort yards benefit the value-added forest products industry in BC? (Forest Community Economic Development Program Project Report No. 124). Nelson, BC: Timberland Consultants, Forest Renewal BC, and Science Council of BC. Schadendorf, C., & Eddy, L. B. (1995). A log sort yard for the boundary area: A feasibility study (Forest Community Economic Development Program Project Report No. 60). Nelson, BC: Timberland Consultants, Forest Renewal BC, and BC Ministry of Employment and Investment. Schwab, O. (2008). An agent-based forest sector modeling approach to analyzing the economic effects of natural disturbances (Ph.D. Thesis). University of British Columbia, Vancouver. References  125 Schwab, O., & Maness, T. (2010). Rebuilding spatial forest inventories from aggregated data. Journal of Forestry, 108(6), 274-281. Schwab, O., Maness, T., Bull, G., & Roberts, D. (2009). Modeling the effect of changing market conditions on mountain pine beetle salvage harvesting and structural changes in the British Columbia forest products industry. Canadian Journal of Forest Research, 39(10), 1806-1820. Schwarzbauer, P., & Rametsteiner, E. (2001). The impact of SFM-certification on forest product markets in western europe -- an analysis using a forest sector simulation model. Forest Policy and Economics, 2(3-4), 241-256. Sessions, J., Boston, K., Hill, R., & Stewart, R. (2005). Log sorting location decisions under uncertainty. Forest Products Journal, 55(12), 53-57. Sessions, J., & Paredes, G. (1987). A solution procedure for the sort yard location problem in forest operations. Forest Science, 33(3), 750-762. Shapiro, J. F. (2001). Modeling the supply chain. Pacific Grove, CA: Wadsworth Group. Shen, W., Hao, Q., Yoon, H. J., & Norrie, D. H. (2006). Applications of agent-based systems in intelligent manufacturing: An updated review. Advanced Engineering Informatics, 20(4), 415-431. Shen, Z.-J. M. (2006). A profit-maximizing supply chain network design model with demand choice flexibility. Operations Research Letters, 34(6), 673-682. Singer, M., & Donoso, P. (2007). Internal supply chain management in the chilean sawmill industry. International Journal of Operations and Production Management, 27(5), 524-542. Sloan, G. M. (1945). Report of the commissioner, the honourable Gordon McG. Sloan, chief justice of British Columbia relating to the forest resources of British Columbia. King’s Printer, Victoria, BC. Snyder, L. V. (2006). Facility location under uncertainty: A review. IIE Transactions, 38(7), 547-564. Spelter, H., & Alderman, M. (2005). Profile 2005: Softwood sawmills in the united states and canada (Research Paper No. FPL-RP-630). Madison, WI: USDA Forest Service, Forest Products Laboratory. Srivastava, S. K. (2007). Green supply-chain management: A state-of-the-art literature review. International Journal of Management Reviews, 9(1), 53-80. Stadtler, H. (2005). Supply chain mangement and advanced planning- basics, overview and challenges. European Journal of Operational Research, 163(3), 575-588. Statistics Canada. (2011a). British Columbia employment by detailed industry, annual averages (prepared by BC stats). Retrieved 5 July 2011, from http://www.bcstats.gov.bc.ca/data/dd/handout/naicsann.pdf Statistics Canada. (2011b). Manufacturing sales, by North American industry classification system (NAICS) and province (table 3040015). Retrieved 15 July 2011, from http://datacenter2.chass.utoronto.ca/cgi-bin/cansimts/getArray.pl?a=3040015 References  126 Statistics Canada. (2011c). Sawn lumber production and shipments (table 3030009). Retrieved 5 July 2011, from http://datacenter2.chass.utoronto.ca/cgi-bin/cansimts/ getArray.pl?a=3030009 Sterling Wood Group Inc. (2002). Dry-land and water-based log sorts (Reference paper). Victoria, BC: BC Ministry of Forests, Lands and Natural Resource Operations, Strategic Land Policy and Legislation Branch. Sterman, J. D. (1984). Instructions for running the beer game, MIT System Dynamics Group Working Paper D-3679. Cambridge, MA: MIT Sloan School of Management. Sterman, J. D. (1989). Modeling managerial behavior: Misperceptions of feedback in a dynamic decision making experiment. Management Science, 35(3), 321-339. Sterman, J. D. (2000). Business dynamics: Systems thinking and modeling for a complex world. New York, NY: McGraw Hill. Sunderman, R. (2003). Establishment of the creston log sort yard: Case study. BC Journal of Ecosystems and Management, 3(1), 28-33. Swaminathan, J. M., Smith, S. F., & Sadeh, N. M. (1998). Modeling supply chain dynamics: A multiagent approach. Decision Scineces, 29(3), 607-632. Terzi, S., & Cavalieri, S. (2004). Simulation in the supply chain context: A survey. Computers in Industry, 53(1), 3-16. The Blue Denim Consulting Group. (2007). Variable density yield prediction (vdyp 7) (Version 7.27h). Victoria, BC: BC Ministry of Forests, Land, and Natural Resource Operations. The Working Forest. (2011). Two vancouver island sawmills close due to lack of logs. Retrieved 1 April 2011, from http://www.workingforest.com/two-vacouver-island- sawmills-close-due-lack-logs Thielmann, T., & Tollefson, C. (2009). Tears from an onion: Layering, exhaustion and conversion in British Columbia land use planning policy. Policy and Society, 28, 111- 124. Thomas, D. J., & Griffin, P. M. (1996). Coordinated supply chain management. European Journal of Operational Research, 94(1), 1-15. Todoroki, C., & Ronnqvist, M. (2002). Dynamic control of timber production at a sawmill with log sawing optimization. Scandinavian Journal of Forest Research, 17, 79-89. Towill, D. R., Naim, M. M., & Wikner, J. (1992). Industrial dynamics simulation models in the design of supply chains. International Journal of Physical Distribution & Logistics Management, 22(5), 3-13. Troncoso, J. J., & Garrido, R. A. (2005). Forestry production and logistics planning: An analysis using mixed-integer programming. Forest Policy and Economics, 7(4), 625- 633. Venn, T. J., Chung, W., Loeffler, D. R., Jones, J. G., Han, H.-S., & Calkin, D. E. (2009). Assessment of the potential for log sort yards to facilitate forest health restoration and fuel reduction treatments. Paper presented at the 32nd Annual Meeting of the References  127 Council on Forest Engineering (COFE): environmentally sound forest operations. June 15-18, 2009, Kings Beach, California. Vidal, C. J., & Goetschalckx, M. (1997). Strategic production-distribution models: A critical review with emphasis on global supply chain models. European Journal of Operational Research, 98(1), 1-18. Vila, D., Beauregard, R., & Martel, A. (2009). The strategic design of forest industry supply chains. INFOR: Information Systems and Operational Research, 47(3), 185-202. Vila, D., Martel, A., & Beauregard, R. (2006). Designing logistics networks in divergent process industries: A methodology and its application to the lumber industry. International Journal of Production Economics, 102(2), 358-378. Vila, D., Martel, A., & Beauregard, R. (2007). Taking market forces into account in the design of production-distribution networks: A positioning by anticipation approach. The Journal of Industrial and Management Optimization, 3(1), 29–50. Weber, A. (1929). Uber den standort der industrien (theory of the location of industries): University of Chicago Press. Weintraub, A., Barahona, F., & Epstein, R. (1994). A column generation algorithm for solving general forest planning problems with adjacency constraints. Forest Science, 40(1), 142-161. Weintraub, A., Jones, G., Meacham, M., Magendzo, A., & Malchuk, D. (1995). Heuristic procedures for solving mixed-integer harvest scheduling–transportation planning models. Canadian Journal of Forest Research, 25(10), 1618-1626. Weintraub, A., & Romero, C. (2006). Operations research models and the management of agricultural and forestry resources: A review and comparison. Interfaces, 36(5), 446- 457. Western Forest Products Inc. (2010). Annual report. Retrieved 10 March 2011, from http://www.westernforest.com/files/annual_reports/2010_Annual_Report_144.pdf Western Forest Products Inc. (2011). Annual information form. Retrieved 10 March 2011, from http://www.westernforest.com/files/annual_reports/Annual_Information_ Form_2010_147.pdf Wijewardana, D. (2008). Criteria and indicators for sustainable forest management: The road travelled and the way ahead. Ecological Indicators, 8(2), 115-122. Wilson, J. (1987). Forest conservation in British Columbia, 1935-85: Reflections on a barren political debate. BC Studies: The British Columbian Quarterly, 76, 3-32. Wilson, J. (1998). Talk and log: Wilderness politics in British Columbia, 1967–96. Vancouver, BC: UBC Press. Winston, W. L., & Venkataramanan, M. (2003). Introduction to mathematical programming: Applications and algorithms (4th ed. Vol. 1). Pacific Grove, CA, USA: Duxbury Press. References  128 Woodbridge Associates. (2009). Opportunity BC 2020, bc's forest industry: Moving from a volume focus to a value perspective. Retrieved 15 July 2011, from http://www.bcbc.com/Documents/2020_200910_Woodbridge.pdf    129 Appendices Appendix A. Additional Graphs for Chapter 2 A.1 Average Agents Profit in the Optimal Solution of Facility Location Problem 0 20 40 60 80 100 120 140 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed  p ro fit  o f e xi st in g ag en ts   ( m ill io n$ ) Time (years) 8.0 8.5 9.0 9.5 10.0 10.5 11.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed   p ro fit  o f ne w   a ge nt  (m illi on  $ ) Time (years) (a) (b)  Figure A.1 (a) Average profits (േ1SD)of existing agents - scenario 1, (b) Average profit (േ1SD)of new agent, scenario 1 0 10 20 30 40 50 60 70 80 90 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed  p ro fit  o f e xi st in g ag en ts   ( m illi on $) Time (years) 8 8.5 9 9.5 10 10.5 11 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed   p ro fit  o f ne w   a ge nt  (m illi on  $ ) Time (years) (a) (b)  Figure A.2 (a) Average profits (േ1SD) of existing agents - scenario 2, (b) Average profit (േ1SD) of new agent, scenario 2  Appendix A  130 0 20 40 60 80 100 120 140 160 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 P re di ct ed  p ro fit  o f e xi st in g ag en ts   ( m illi on $) Time (years) 0 2 4 6 8 10 12 14 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Pr ed ic te d  p ro fit  o f ne w   a ge nt  (m illi on  $ ) Time (years) (a) (b)  Figure A.33 (a) Average profits (േ1SD) of existing agents - scenario 3, (b) Average profit (േ1SD) of new agent, scenario 3    Appendix B  131 Appendix B . Additional Graphs for Chapter 3 B.1 Average Total Harvest (a) (b) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3) Time (years) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Figure B.14 (a) Average Total harvest volume (േ1SD): scenario I - without a sort yard, (b) Average Total harvest volume (േ1SD): scenario I - with a sort yard (a) (b) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3) Time (years) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Figure B.25 (a) Average Total harvest volume (േ1SD): scenario II - without a sort yard, (b) Average Total harvest volume (േ1SD): scenario II - with a sort yard  Appendix B  132 B.2 Average Total Saw Log Volume Purchased by Sawmills from the Sort Yard 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m illi on  m 3 ) Time (years) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (m ill io n m 3 ) Time (years) (a) (b)  Figure B.36 (a) Average Total purchased logs from sort yard (േ1SD): scenario I, (b) Average Total purchased logs from sort yard (േ1SD): scenario II   Appendix C  133 Appendix C. Additional Graphs for Chapter 4 C.1 Base Case 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (m ill io n m 3) Time (years) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) (a) (b)  Figure C.17 (a) Average total harvest volume of sawmills (±1SD) – Base case, (b) Average total log import volume of sawmills (±1SD) – Base case 0 100 200 300 400 500 600 700 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 V ol um e (th ou sa nd  m bf ) Time (years) -80 -60 -40 -20 0 20 40 60 80 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) (a) (b)  Figure C.28 (a) Average total profit of sawmills (±1SD) – Base case, (b) Average total production capacity of sawmills (±1SD) – Base case  Appendix C  134  C.2 Scenarios I and II  Figure C.39 (a) Average total harvest volume of sawmills (±1SD) - Scenario I, (b) Average total harvest volume of sawmills (±1SD) – Scenario II C.3 Scenarios III and IV 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (m illi on  m 3 ) Time (years) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 Vo lu m e (m illi on  m 3 ) Time (years) (a) (b) Scenario IVScenario III Figure C.410 (a) Average total harvest of sawmills (±1SD) - Scenario III, (b) Average total harvest of sawmills (±1SD), scenario IV 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Reduced AAC(20%) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) Reduced AAC(40%) Scenario IIScenario I (a) (b)  Appendix C  135 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (m ill io n m 3 ) Time (years) 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 V ol um e (m ill io n m 3 ) Time (years) Scenario IVScenario III (a) (b)  Figure C. 611(a) Average total log import volume of sawmills (±1SD), scenario III, (b) Average total log import volume of sawmills (±1SD), scenario IV 350 400 450 500 550 600 650 700 750 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (th ou sa nd  m bf ) Time (years) 350 400 450 500 550 600 650 700 750 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 V ol um e (th ou sa nd  m bf ) Time (years) Scenario IVScenario III (a) (b)  Figure C.612(a) Average total production capacity of sawmills (±1SD), scenario three, (b) Average total production capacity of sawmills (±1SD), scenario four  Appendix C  136 -140 -120 -100 -80 -60 -40 -20 0 20 40 60 80 1 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 P ro fit s (m ill io n $) Time (years) -140 -120 -100 -80 -60 -40 -20 0 20 40 60 80 1 6 8 10 20 30 40 50 60 70 80 90 10 0 11 0 12 0 13 0 14 0 15 0 P ro fit s (m ill io n $) Time (years) Scenario IVScenario III (a) (b)  Figure C.713(a) Average total profits of sawmills (±1SD) - scenario III, (b) Average total profits of sawmills (±1SD) - scenario IV C.4 Age Distribution of Timber Harvesting Land Base (THLB) for All Scenarios  Figure C.814 Average THLB area (±1SD) by age group for the base case 0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 A re a (th ou sa nd  h a) Time (years) 0-39 yrs 40-79 yrs 80-119 yrs 120+ yrs  Appendix C  137  Figure C.915 (a) Average THLB area (±1SD) by age group for scenario I, (b) Average THLB area (±1SD) by age group for scenario II   Figure C.1016(a) Average THLB area (±1SD) by age group for scenario III, (b) Average THLB area (±1SD) by age group for scenario IV 0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 A re a (th ou sa nd  h a) Time (years) 0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 A re a (th ou sa nd  h a) Time (years) (a) (b) 0-39 yrs 40-79 yrs 80-119 yrs 120+ yrs 0 50 100 150 200 250 300 350 400 450 0 50 100 150 A re a (th ou sa nd  h a) Time (years) 0 50 100 150 200 250 300 350 400 450 0 50 100 150 A re a (th ou sa nd  h a) Time (years) (a) (b) 0-39 yrs 40-79 yrs 80-119 yrs 120+ yrs  Appendix C  138  Figure C.1117 Average THLB area (±1SD) by age group for scenario V 0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 A re a (th ou sa nd  h a) Time (years) 0-39 yrs 40-79 yrs 80-119 yrs 120+ yrs

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0072459/manifest

Comment

Related Items