Journal Search Engine
Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 1598-7248 (Print)
ISSN : 2234-6473 (Online)
Industrial Engineering & Management Systems Vol.16 No.3 pp.330-341
DOI : https://doi.org/10.7232/iems.2017.16.3.330

Hybrid Simulation Model for the Upstream Rubber Supply Chain

Chawalit Manisri*, Juta Pichitlamken
Department of Industrial Engineering, Faculty of Engineering, Kasetsart University, Bangkok, Thailand
Corresponding Author, chawalit.ma@spu.ac.th
January 28, 2017 July 17, 2017 July 27, 2017

ABSTRACT

Thailand has been one of the largest natural rubber (NR) exporters in the world since 1991. However, the upstream rubber supply chain suffers from problems such as oversupply and decreasing price. This work proposed a decisionsupport model that combines System Dynamics and Agent-Based Modeling simulation while taking into account the relationship between NR price and planting rate per year, as well as between NR price and crude oil price. The input data between 2002 and 2015 included area sizes, tapping yield, planting rates, growth rates, and time to start or end tapping. The upstream rubber supply chain models were simulated from 2002 to 2021. The hybrid models in this study were validated by comparing between the simulated NR volume and actual values. The average Mean Absolute Percentage Error was 8.6%. The results of this work may be used by stakeholders in NR supply chains, such as in government agencies, to support their policy making.


초록


    1.INTRODUCTION

    Rubber has had the highest total export value among agricultural products of Thailand for the past 10 years. In 2015, rubber and rubber products were valued at up to US$ 11,452 million, as shown in Figure 1 (Information Technology and Communication Centre, 2016). In 2011, NR’s price was so high that farmers were motivated to expand their planting areas without country-wide planning.

    Rubber trees take around 6 to 7 years to harvest (or tap). Thus, NR output was greatly oversupplied when rubber trees were ready to be tapped. At the same time, NR price decreased continuously; the price fell from US$ 3.80 per kilogram in 2011 to US$ 1.30 per kilogram in 2015, reducing by 66% from its highest value (see Figure 2). Low NR price took a heavy toll on rubber farmers, resulting in farmers’ debt that led to social unrest.

    Because of the oversupply and low price of NR, the Thai government asked researchers for solutions that could increase Thai NR supply chain competitiveness. Singkarin (2012) and Tosanguan (2012) studied the relationships between stakeholders within the NR supply chain. Boonkwang (2007), Liangphansakul (2007) and Chanpuypetch and Singkarin (2009) considered the NR distribution channels. Some papers have analyzed the entire NR supply chain from upstream to downstream. Phanishsarn (2009) proposed value-adding approaches for NR. However, most research work is qualitative and cannot sufficiently support policy makers in decision-making for the NR supply chain such as for decreasing planting areas by cutting down rubber trees and guaranteeing the NR price.

    As a result, we propose a quantitative model for the NR supply chain in each geographical region of Thailand. Our simulation model takes historical input data, such as area sizes, tapping yield, planting rates, growth rates, and time to start or end tapping, from 2002 to 2015. The model’s output is the NR supply volume. System Dynamics (SD) and Agent-Based Modeling (ABM) are integrated so that it can provide both quantitative information and qualitative insights. The hybrid models also consider the relationships between NR price and planting rate per year, as well as NR price and crude oil price.

    It is notable that fossil oil is a raw material for synthetic rubber, a substitute for NR. The model can help stakeholders to support their decision-making in the upstream rubber supply chain because the model can show both qualitative and quantitative information. The information is based on the relationship between the key factors that impact the upstream element.

    This paper has been organized as follows: Section 1 shows the importance of NR, the causes of NR supply chain problems, scope, and research approaches. Section 2 introduces an overview of past research on NR supply chain, system dynamics and agent-based models, preferably on agricultural applications, and how to integrate SD and ABM. Section 3 provides details on our proposed hybrid model development. Section 4 describes the parameter relationship, model validation, and experimental results. Finally, Section 5 summarizes our research.

    2.PRELIMINARIES AND LITERATURE REVIEWS

    We provide an overview of the NR supply chains in section 2.1. System Dynamics and Agent-Based Modeling are introduced in section 2.2.

    2.1.Rubber Supply Chains

    Many studies on management aproaches endeavor to increase competitiveness in the NR supply chain because rubber is an important commodity corp. Singkarin (2013) and Somboonwiwat and Chanclay (2009) described key stakeholders in each process of the rubber supply chain, such as farmers, middlemen (traders, markets, and cooperatives), processing plants, rubber product manufacturers, and customers (Figure 3).

    Based on an overview of past research, literature concerning the NR supply chain can be classified into three dimensions (Singkarin, 2013).

    2.1.1.Supply Chain Dimension

    Research that focuses on the study of relationships in the supply chain. This may include relationships between the parties or relationships between groups, from two groups (individuals) or more, to cover the entire supply chain.

    Singkarin (2012) considers the potential of integrated logistics and supply chain management of the rubber industry in terms of the relationship between farmers, traders, and the midstream plant. The rubber industry has four issues: the rubber value chain, promoting the use of natural rubber in Thailand, the rubber logistics system, and the lack of integration for state policy. The proposed solutions include:

    • Analyze the demand, supply and yield of NR in Thailand, and rubber value chain management,

    • Promote the government’s rules for use of NR,

    • Analyze and create a decision-support system for rubber logistics from sources to different destinations, and

    • Find approaches for the integration of the government's policies by analyzing the role of the national rubber council and revising the Rubber Act.

    Tongchim and Rassameethes (2012) studied the relationship between the three processes of supply chain management (Planning supply chain managemenr, delivering management, and customer relationship management). Its performan is divided into two dimenssions, including competitive advantage (cost, quality, and time) and the performance of the organization. The managerial structure at the early level and mid-level is modeled. Research shows that supply chain management has resulted in competitive advantage and the enhanced performance of the firm. Tosanguan (2012) developed a supply chain and value chain for rubber and the rubber products industry in northern Thailand in addition to providing operators with guidelines for measuring performance and continuously improving their enterprise.

    Some research studies of the rubber supply chain have been conducted by separating upstream, midstream, and downstream chains. Rattanawong (2011) presented an overview of the upstream chain, which included five components as follows: rubber plantations, rubber tapping, collecting latex, maintaining latex, and the determination of dry rubber. The points to add value and improve performance were visualized. Sometimes, research may focus on supply chain management for specific rubber products. By focusing on the tire industry, Gabkom (2010) showed the supply of natural rubber from each regional source to a midstream plant and products from a midstream plant to a tire factory in the eastern and central part of Thailand. Moreover, forecasted demand for natural rubber, the capacity of the natural rubber plantation, area plantation, and potential locations for new tire factories to meet the demand for tires and natural rubber in Thailand were also presented in 2014.

    2.1.2.Logistics Dimension

    Research that focuses on the study of export channels by geographical area.

    Some researchers consider the rubber supply chain by geographical location. For example, the Rubber Research Institute of Thailand (2011) examined the rubber supply chain in eastern and northern Thailand. It also, studied rubber logistics management in the southern part of Thailand. The channels for logistics management from upstream to downstream were determined. The performance of the northern and eastern parts and the southern export channel were evaluated. Farmers, middlemen, plants, and logistics entrepreneurs were also suggested. Chanpuypetch and Singkarin (2009) considered the decision support system to select multimodal transport for exporting rubber from Thailand. They utilized the Fuzzy Analytic Hierarchy Process (FAHP) to evaluate factors in four dimensions including transportation, economics, port/border trade, and other environmental issues. This research helps rubber exporters to receive data necessary for decision support by selecting the appropriate channel for exporting rubber based on many factors. Liangphansakul (2007) studied the efficiency of rubber exports to southern China through the Chiang Saen Port. This work presents the structure of logistical costs, the quality of logistics management, and the efficiency of logistics costs for these channels. Boonkwang (2007) analyzed rubber trading in Ban Huak, Phayao, in two aspects: transportation cost, and customer response. Reliability was measured using a supply chain operations reference (SCOR) model to measure the performance of 3 routes to Ban Huak. The research shows the efficiency of each route and the adequacy of the route for each operation.

    2.1.3.Value Chain Dimension

    Research that focuses on two topics, with the first being the direction for creating value-added products in the downstream industry and the second being directions of value adding by considering the overall economic direction of Thailand.

    The most recent research that has impacted the overall rubber industry fits into this dimension. For example, Phruttikitti et al. (2007) studied alternatives to the systematization of rubber-farming organizations in Phatthalung, Trang and Na Korn Sri Thammarat Provinces for value chain creation by analyzing the efficiency of the rubber farming organizations and rubber estate organizations. In the end, they proposed alternatives to coordination between the rubber-farming organizations and rubber estate organizations for rubber processing, logistics, and stock management for domestic and international markets. Phanishsarn (2009) proposed approaches to creating value chains from rubber processing in Thailand to end users in China by increasing investment for rubber processing and the value of rubber products within Thailand before export to China. Nong Khai was suggested as a “rubber city” for the export of rubber products to Chongqing in western China due to it being one of the major production bases for the automotive industry in China. Rubber products from Guangxi to Chongqing are transported by rail through Laos and Vietnam. Somboonwiwat and Chanclay (2009) studied various means to create value addition by focusing on Thailand’s national targets. They considered the relationship and impact on the upstream and midstream industry by national target. Singkarin and Somboonwiwat (2010) proposed major forecasting models to predict demand for automotive tires because they comprise the main rubber products that involve operations and preparation from upstream through downstream chains.

    The research and academic articles on rubber supply chain management in various countries, especially major exporters in the world, are also considered. Indonesia and Malaysia are the second and third biggest rubber exporters in the world, respectively. Arifin (2005) examined the supply chain of natural rubber in Indonesia and assessed the propagation of prices to rubber growers in order to propose a suitable scheme that ensures high production standards and a sustainable return for natural rubber production. The results show that the roles of middlemen are extremely important in advancing natural rubber products from the village level to urban areas, where traders and brokers expect the NR slabs to be forwarded directly to crumb-rubber factories. Profits from changes in world prices are accumulated by traders and rubber factories, but not transferred down to the rubber farmers. Peramune (2007) considered a value chain assessment for the rubber industry in Indonesia. The country lacks technical support, while the monitoring system is poor. Thus, high-quality rubber trees are not distributed to farmers, with many low-quality rubber trees planted as a result. Poor tapping techniques shorten lifetime by 50%. Farmers lack awareness of the environmental benefits of systematic rubber plantation. The solution is to establish an operational unit to prepare high-quality rubber, provide training for farmer leaders to disseminate knowledge among farmers and provide technical knowledge concerning the tapping of rubber. Karim (2010) proposed a sustainability plan for the rubber industry in Malaysia from economic and social perspectives by setting up strategies and initiatives in two areas. The economic transformation program and strategies are fit for Malaysian rubber from 2010 to 2020 as a means to transform the rubber industry by restructuring the rubber smallholdings and increasing supply chain efficiency by improving efficiency in the marketing chain.

    In the next section, we discuss simulation models for NR supply chains.

    2.2.Simulation Models for Agricultural and NR Supply Chains

    There are three main simulation approaches such as discrete-event simulation (DES), system dynamics (SD), and agent-based modeling (ABM). DES may be the most applied simulation technique. DES models have three main elements. Entity is the object that moves through the system. Process is the object that the entities pass through. Resources - Objects which are needed to trigger events. SD, the system elements are represented by “Stock” (such as materials, money, and people), “Flow” (such as rate) between stock, and information for the calculation of flow rate, which is suitable for studying the behavior of complex systems that change over time, as shown by the graph. ABM is “a set of elements (agents) characterized by some attributes, which interact with each other through the definition of appropriate rules in a given environment” (Barbati et al., 2012). ABM focuses on three key components such as “agent,” “interact,” and “environment.” “Agent” is a discrete entity with its own goals and behaviors. “Interact” is based on “local” interaction between agents or agent relationships and the methods of interaction that define how and with whom agents interact. Finally, “environment” means the agents live in and interact with their environment in addition to other agents. Borshchev and Filippov (2004) compared three simulation approaches. DES applies operational and tactical levels of discrete time. SD applies a strategic level of continuous process. ABM applies operational, tactical and strategic levels of discrete time. For this research, the rubber supply chain problems cover all levels, from operational to strategic. The SD model determines the rate of change between the groups in the model. ABM considers the behavioral changes of the system. Thus, we provide the simulation models that will be implemented in our NR supply chain model; system dynamics and agent-based modeling.

    Although System Dynamics or SD (Forrester, 1961) has been applied in many other types of supply chains for more than 50 years, it is the only simulation approach that has been applied to the rubber supply chain (Angerhofer and Angelides, 2000). SD modeling is implemented in several areas of agricultural industry. For example, Bantz and Deaton (2006) used STELLA to model the growth of the biodiesel industry in the U.S. Hongling and Zhan (2007) studied the corn-processing industry in northeast China to analyze and propose ways to develop the industry. Cui et al. (2009) applied SD based on the orthogonal experimental design to find the most suitable crop combination for each pattern. Kibira and Shao (2011) modeled the growth of the U.S. corn industry with SD.

    In our past work, SD helped to analyze the rubber supply chain research, both qualitatively and quantitatively. Manisri et al. (2013) showed a conceptual model for the rubber supply chain in the northeast of Thailand by using SD. They applied the causal loop diagram (CLD) to analyze the relationship between the components and variables that affect the complex rubber supply chain. The CLD helps assess the impact of different factors or variables. Manisri and Pichitlamken (2015) created 4 models for each geographical region of the upstream rubber supply chains in Thailand by stock and flow diagram. The experimental results are the NR volume which is then compared with the actual data by the Mean Absolute Percentage Error (MAPE). The MAPE values are 4.09%, 2.35%, 9.28%, and 8.21% for the South, the East, the North, and the Northeast, respectively. However, Manisri and Pichitlamken (2015) did not consider important factors such as crude oil price and NR price that affect farmers’ decisions.

    ABM allows a functional relationship between the factors above. It focuses on three key components: “Agent,” “Interact,” and “Environment.” Agent is a discrete entity with its own goals and behaviors. It is autonomous with a capability to adapt and modify its behaviors. Examples of agents are farmers, areas, groups, organizations, robots, social insects, and swarms. Interact is based on local interaction among agents or agent relationships and methods of interaction that define how and with whom agents interact. Finally, Environment means the agents live in and interact with their environment as well as other agents (Macal and North, 2006).

    Some examples of ABM applications include Hilletofth and Lattila (2012), who showed the benefits and barriers of ABM. It can support decision-making in the supply chain management context. Lu and Wang (2008) applied ABM to the supply chain framework based on network economy. This research shows the information transfer process and cooperation between agents in supply chain operations as well as the network economy structure.

    If we integrate SD and ABM to model the NR supply chain model, the resulting model can be more practical. The concept of an integrated SD and ABM started only 15 years ago. Scholl (2001) proposed an overview of both approaches, described their areas of applicability, and discussed their relative strengths and weaknesses in an effort to identify overlapping complementary areas. Schieritz (2002) presented an approach to integrate SD and ABM by using SD at the micro level with the agents’ internal cognitive structure, and using ABM at the macro level. Kieckhafer et al. (2009) proposed a conceptual framework for integrating SD and ABM to support product strategy decisions in the automotive industry in the European Union. Later, Kieckhafer et al. (2012) actually constructed a simulation model of their earlier ideas. Manisri and Manisri (2013) proposed a conceptual model for the demand evaluation of mass transit systems from the capital to other cities by integrating SD and ABM, where passengers are agents.

    Borshchev (2013) presented three approaches to creating a hybrid simulation model between SD and ABM. The concepts are a SD model in agents of an ABM model, an agent of ABM model in a SD model, and a separated SD model and ABM model. Our hybrid simulation model belongs to the last approach: a separated SD model and an ABM model.

    After reviewing the NR supply chain literature, we found that it broadly covers every group. However, there is little quantitative research available. Our work aims to answer the need for numerical results. Quantitative research is comprised of 2 approaches: regression and simulation. With regression, Somboonwiwat and Chanclay (2009) studied approaches to support the players operations and forecast status of each player if the automotive tire export volumes are on target for the next 5 years. Gabkom (2010) used the regression model to forecast the export value of automotive tires. Both articles focus on only one product champion. For simulation, Manisri et al. (2013) showed a conceptual model for the rubber supply chain northeast Thailand with CLDs. Manisri and Pichitlamken (2015) used only the SD approach to create 4 models for each geographical region of the upstream rubber supply chains in Thailand.

    In this article, our hybrid model of SD and ABM was created on AnyLogic® simulation software, which has main features that are similar to many other wellknown simulation packages such as Vensim, Powersim (Borshchev, 2013) or NetLogo. This software can be modified with Java to allow more flexibility in modeling. It can also be integrated with other simulation approaches such as discrete-event or agent-based modeling. Integration is advantageous for expanding the model in the future and to fully consider all levels of simulation such as strategic, tactical, and operational.

    3.PROPOSED MODEL

    Figure 4 shows a conceptual idea for the hybrid simulation approach: the main model uses SD for calculating NR volume, while ABM is employed to simulate planting behaviors by rubber farmers in the sub-model. The upstream section begins when the farmer plants new rubber trees in the potential planting areas. Farmers will start to tap or harvest when the young rubber trees reach 6 to 7 years old. We model the time to begin harvesting as a uniform random variable with a minimum of 6 and maximum of 7 years. However, the rubber trees may be tapped before the 6th year if the price is high. The tapping yield ranges from 395 to 635 kg/acre/year depending on climate, geography, and time to tapping in each region (Office of Agricultural Economics, 2016). Total NR volumes are derived from the size of the tapping area as well as the tapping yield for each region. The south gives the highest yield and the biggest NR volume, as shown in Table 1. Rubber trees can be harvested for 20 to 25 years before they are cut down. We model the age of rubber trees as a uniform random variable with a minimum of 20 and maximum of 25 years.

    In the main model, our SD consists of two sections: NR volume and existing planting area. In the first part, NR volume is assigned as Stock, while the tapping yield is assigned as Flow. The NR volume is calculated from the yield and number of existing tapping areas. The new tapping area is assigned as an Interaction between SD and ABM: the new tapping area sends values from the ABM section to the SD section.

    In the second part, the existing planting areas start with a young tree that grows up to become an adult tree by the growing-up rate. Subsequently, it moves to accumulate in the existing tapping area. The rubber trees will be harvested for a delay time at the end-of-life rate. The existing tapping area is then returned to the stock of the potential area in the ABM section.

    The new planting areas (farmers) are modeled with ABM. Each area (acre) is modeled individually as an Agent. Agents of new areas decide to plant new rubber trees by considering the price and potential areas. For the new planting areas section, the potential area is decreased annually by planting rate, adding to the cumulative young rubber trees in the new planting area. Young trees get delayed in the new planting area by growing up rate before they move to accumulate in the new tapping area. When farmers harvest latex for the length of time equal to the end-of-life rate similar to the existing planting areas section, the new tapping area of the rubber tree will be returned to accumulate again in the potential area.

    We transform the conceptual model into a computer simulation model via AnyLogic® simulation software, as shown in Figure 5, and Figure 6. The next section describes our simulation experiment and its results.

    4.NUMERICAL EXPERIMENT

    We provide details concerning our parameter setting in Section 4.1 and analyze the simulation results in Section 4.2.

    4.1.Parameter Settings

    Simulation inputs are divided into three groups: initial parameters, random variables, and time-dependent parameters. The initial parameters are starting values for potential planting areas and already-existent tapping areas, as shown in Table 2 (Office of Agricultural Economics, 2016; Rubber Research Institute of Thailand, 2012; Somboonwiwat and Chanclay, 2009).

    Table 3 presents probability distribution parameters for random variables such as “time to begin harvest” and “age to cut down.” We estimate all of these parameters that are taken from data in annual rubber reports (Rubber Research Institute of Thailand, 2012; Office of Agricultural Economics, 2016) and a related research paper (Somboonwiwat and Chanclay, 2009). They show that the southern farmers usually start to tap 6 to 7 year after rubber trees are planted. On the other hand, farmers in other regions begin harvesting sooner, at around 5 to 6 years. Rubber farmers throughout Thailand cut down rubber trees after 20 to 25 years. The specific cut-down age depends on the condition and yield of each rubber tree.

    Time-dependent parameters consist of crude oil prices, NR prices, planting rates, growth rates (only in existing area), and tapping rates, as shown in Table 4 (Knoema, 2016; Office of Agricultural Economics, 2016; Rubber Research Institute of Thailand, 2012; Somboonwiwat and Chanclay, 2009). Growth rates in existing areas are de- fined as the area sizes of just-ready-to-harvest trees being added to the pool of existing planting areas. The growth rates beyond the 5th or 6th year are zero because the assumption is that not many new rubber trees are planted. The north region is the newest planting area in Thailand. The growth rates in 2002-2004 were zero because the young trees were not ready to be harvested.

    The relationship between NR prices and crude oil prices, as shown in Figure 2, shows a relatively strong correlation between crude oil price and NR price for the actual data during 2002-2015. We use the regression model that the statement for the main hypothesis (H0), which is the constant regression coefficient. NR prices are regressed with crude oil prices using the actual data from 2002 to 2015. The NR price (PNR) depends on the crude oil price (POil) as specified in Equation (1) below:

    P N R = 0.334 + 0.02129 P o i l
    (1)

    The adjusted coefficient of determination (R2adj) for Equation (1) is 0.5548, so the regression model can explain NR price at around 55.48%. The p-value is 0.000 that rejects H0, the relationship between NR price and the crude oil price is statistically significant (p-value < 0.05). Finally, a positive correlation (r = 0.76) indicates that NR price tends to increase when the crude oil price increase.

    The relationship between planting rate and NR price is considered because past data showed that the plantation area also tends to increase when NR price increase. The statement for the main hypothesis (H0) is the constant regression coefficient. If the p-value of regression models is less than 0.05, the planting rate and NR price have a relationship. The regression equations are shown in Table 5, as derived from historical data between 2002 and 2015.

    We can see that the relationship between NR prices and planting rates are the strongest in the south with the R2adj of 0.961 and the p -value of 0.000. Because rubber trees have been planted in the south for the past 50-60 years and rubber is the main agricultural resource for this region, the planting rate has stabilized. The other regions are new plantation areas and rubber is not the main agricultural resource. They typically plant other crops such as rice, cane, cone, and canvas. Although the farmers in the east and the north are new, the planting rate has stabilized because they have few potential rubber plantation areas, and rubber trees were gradually planted. The regression model can explain NR price around at 53.50% and 59.90%, respectively. The p-value of both regions is lower than α = 0.05; the relationship between NR prices and planting rates is statistically significant as well. However, the northeast is another new rubber planting area, so planting rates fluctuate more than in other regions since the northeast has many potential areas for rubber trees. Rubber trees offer a new hope for farmers to earn higher incomes. The region has a statistically significant relationship between NR prices and planting rates under α = 0.10. The regression model can explain NR price at around 42.56%.

    We create 4 different models for each geographical region of Thailand to experiment with the upstream rubber supply chain using input from 14 years of data. Each model is simulated with 10 replications for 20 years to get NR volumes that are then compared with the actual values for the first 14 years and the forecasts for the remaining years up to 2021.

    4.2.Discussion

    Figure 7 shows a comparison between the simulated and actual NR volumes for the 4 geographical regions of Thailand. The simulated values for the south have slight deviations from the actual data because of stable plantations and tapping. On the contrary, rubber planting is new and uses mixed plantation with other agricultural crops in the east, the north and the northeast. Therefore, NR supplies from these regions are not as predictable. Historical data for input modeling is also scarce for these new rubber planting areas. This lack of data impacts the accuracy of functional relationships between planting rates and NR prices (Table 5). As a result, forecasts for the east, the north and the northeast regions have larger deviations from the actual data than those of the south.

    The simulated NR volume is numerically compared to the actual data by the Mean Absolute Percentage Error (MAPE). This measure for forecasting errors is computed as the average difference (et) between actual (At) and forecasted value (Ft) at time t, as shown in Equation (2):

    MAPE= 1 n t = 1 n | e t A t |
    (2)

    where the number of time series is n.

    The average MAPE values of the south, the east, the north, and the northeast regions are 4.77%, 5.33%, 16.04%, and 8.28%, respectively, as shown in Table 6. The average MAPE value for all regions is 8.61%.

    5.CONCLUSIONS

    Our hybrid simulation model can simulate the upstream rubber supply chain for each region of Thailand. The average MAPE is 8.6%, which shows a relatively small error for the simulated values from actual data. The stable rubber operations and quantity of data have an impact on the analytical relationships and accuracy of the results. The separated region models support this conclusion. The south has been planting rubber trees for many years and has strong relationship factors, so slight deviation is the result. For other regions, they have low relationship levels that lead to high deviation results. Because there are new plantation areas and rubber is not the main agricultural resource, it affects to the limited amount of input data and low-fidelity functional relationships between NR prices and planting rates as well as between NR prices and crude oil prices.

    The hybrid simulation model can generate numerical data to support stakeholders’ decision-making. However, the hybrid simulation model for the upstream rubber supply chain needs more consideration of the relationships such as NR price and GDP, or GDP and export volume. In future research, the hybrid model should be expanded to other parts of the rubber supply chain, such as midstream and downstream, as well as focusing on the key rubber products. The rubber supply chain needs holistic analysis that describes the effects along the entire the supply chain.

    ACKNOWLEDGEMENTS

    This research is supported by Grant No.60.58 from Kasetsart University Research and Development Institute (KURDI). The first author would also like to thank Sripatum University for doctorate scholarship.

    Figure

    IEMS-16-330_F1.gif

    Export values of rubber and rubber products between 2002 and 2015.

    IEMS-16-330_F2.gif

    Crude oil price and NR price (Knoema, 2016; Rubber Research Institute of Thailand, 2016).

    IEMS-16-330_F3.gif

    The NR supply chain in Thailand.

    IEMS-16-330_F4.gif

    Conceptual framework for our hybrid model for NR supply chain.

    IEMS-16-330_F5.gif

    The system dynamics section of the hybrid model.

    IEMS-16-330_F6.gif

    The agent-based modeling section of the hybrid simulation model.

    IEMS-16-330_F7.gif

    Comparison of the simulated and actual NR volumes.

    Table

    Yield, tapping area, and NR volume for each region in Thailand 2015

    Initial-parameter inputs

    Random variables inputs

    Time-dependent parameters inputs

    Functional relationship between planting rate (Y) and NR price (PNR) for each region in Thailand (sample size = 14)

    The average MAPE and standard deviation of 10 replications

    REFERENCES

    1. Angerhofer B. , Angelide M. (2000) System dynamics modeling in supply chain management: Research review , Proceedings of the 2000 Winter Simulation Conference, ; pp.342-352
    2. Arifinn B. (2005) Supply-chain of natural rubber in Indonesia , Journal of Management and Agribusiness, Vol.2 (1) ; pp.1-16
    3. Bantz S.G. , Deaton M.L. (2006) Understanding U.S. biodiesel industry growth using system dynamics modeling , Proceedings of 2006 IEEE Systems and Information Engineering Design Symposium, ; pp.156-161
    4. Barbati M. , Bruno G. , Genovese A. (2012) Applications of agent-based models for optimization problems: A literature review , Expert Syst. Appl, Vol.39 (5) ; pp.6020-6028
    5. Boonkwang A. (2007) Analyze the logistics system of rubber trade via the Ban Huak temporarily permitted area, Phayao (Unpublished master’s thesis), Chiang Mai University,
    6. Borshchev A. (2013) The Big Book of Simulation Modeling, Multimethod Modeling with Anylogic 6, Any- Logic North America,
    7. Borshchev A. , Filippov A. (2004) From system dynamics and discrete event to practical agent based modeling: reasons, techniques, tools, Available from: http://www.systemdynamics.org/conferences/2004/SDS_2004/PAPERS/381BORSH.pdf
    8. Chanpuypetch W. , Singkarin D. (2009) Decision support system design for route selection to rubber export of Thailand , Proceedings of the 9th Thai VMCL, ; pp.331-342
    9. Cui J. , Huang G. , Tang Y. (2009) System dynamic simulation of three crops per year in paddy field , World Journal of Modelling and Simulation, Vol.5 (2) ; pp.144-150
    10. Forrester J.W. (1961) Industrial dynamics, MIT Press,
    11. Gabkom N. (2010) The study of natural rubber supply chain management of tire industry (Unpublished master’s thesis), King Mongkut’s University of Technology Thonburi,
    12. Hilletofth P. , Lattila L. (2012) Agent based decision support in the supply chain context , Ind. Manage. Data Syst, Vol.112 ; pp.1217-1235
    13. Hongling S. , Zhan G. (2007) The system dynamics analysis on the development of corn processing industry in northeast China , Proceedings of 2007 international conference on management science and engineering, ; pp.899-903
    14. (2016) Thailand trading report, Available from: http://www2.ops3.moc.go.th
    15. Karim M.T. (2010) Available from: http://www.theirrdb.org/irrdb/irc2010/SEC6/MSIA_sustain_econ_161010_FINALE%20%202003.pdf
    16. Kibira D. , Shao G. (2011) Modeling of U.S. corn ethanol industrial growth , Proceedings of the ASME 2011 5th international conference on energy sustainability, ; pp.1-10
    17. Kieckhafer K. , Walther G. , Axmann J. , Spengler T. (2012) Model-based decision support for future OEM power-train portfolios: Academic solutions for practical requirements , Available from: http://www.systemdynamics.org/conferences/2012/proceed/papers/P1127.pdf.
    18. Kieckhafer K. , Walther G. , Spengler T. , Axmann J. (2009) Integrating agent based simulation and system dynamics to support product strategy decisions in the automotive industry , Proceedings of the 2009 Winter Simulation Conference, ; pp.1433-1443
    19. Knoema (2016) Crude oil price forecast: Long term 2016 to 2025: Data and charts , Available from: https://knoema.com/yxptpab/crude-oil-price-forecast-longterm-2016-to-2025-data-and-charts
    20. Liangphansakul O. (2007) Efficiency of the logistics management for rubber export to southern of China through Chiang Saen port (Unpublished master’s thesis), Chiang Mai University,
    21. Lu L. , Wang G. (2008) A study on multi-agent supply chain framework based on network economy , Comput. Ind. Eng, Vol.54 (2) ; pp.288-300
    22. Macal C.M. , North M.J. (2006) Introduction to agent-based modeling and simulation, Available from: http://www.mcs.anl.gov/~leyffer/listn/slides-6/MacalNorth.pdf
    23. Manisri C. , Manisri T. (2013) The conceptual model for the demand evaluation of mass transit systems from the capital to province by integrated simulation techniques , Sripatum Review of Science and Technology, Vol.5 ; pp.122-128
    24. Manisri C. , Pichitlamken J. (2015) An upstream rubber supply chain model of thailand by system dynamics , Proceedings of the international conference on management, business, and economics and the 3rd international conference on tourism, transport, and logistics, ; pp.427-436
    25. Manisri C. , Wangwatcharakul W. , Pichitlamken J. (2013) A conceptual model for rubber supply chain in the northeast of thailand by system dynamics , Journal of Industrial Technology Ubon Ratchathani Rajabhat University, Vol.3 ; pp.34-40
    26. (2016) Agricultural production data, Available from: http://www.oae.go.th/download/prcai/farmcrop/rubber.pdf
    27. Peramune M.R. (2007) A value chain assessment of the rubber industry in Indonesia, Development Alternatives,
    28. Phanishsarn A. (2009) Value chain analysis of exporting Thai processed rubber to China and neighboring countries, The sixth national conference of economists , Ramkhamhaeng University,
    29. Phruttikitti B. , Somboonsuke B. , Dhammasaccakarn W. , Nitirungjarus K. (2007) Study alternatives to systematization the rubber farmer organization in Phatthalung, Trang and Na Korn Sri Thammarat province for value chain creation with rubber estate organization, The Thailand Research Fund,
    30. Rattanawong W. (2011) Upstream rubber supply chain in Thailand, Available from: http://www.logistics.go.th/index.php/news-information/bol-article/878-1-gujranwala-25
    31. (2011) Supply Chain of Rubber in the Eastern Part of Thailand, Annual Full Research Report 2011 Rubber Research Institute of Thailand, Department of Agriculture, Ministry of Agriculture and Cooperatives,
    32. (2012) Rubber data 2012, Department of Agriculture, Ministry of Agriculture and Cooperatives,
    33. (2016) Thai rubber statistical, Available from: http://www.rubberthai.com/statistic/stat_index.htm
    34. Schieritz N. (2002) Integrating system dynamics and agent-based modeling, Available from: http://www.systemdynamics.org/conferences/2002/proceed/papers/Schieri1.pdf
    35. Scholl H.J. (2001) Agent-based and system dynamics modeling: A call for cross study and joint research , Proceedings of the 34th Annual Hawaii International Conference on System Sciences, ; pp.3003-3010
    36. Singkarin D. (2012) The Evaluation of Integrated Industrial Logistics System and Supply Chain Management in Thailand, The Thailand Research Fund,
    37. Singkarin D. (2013) Rubber logistics: The integration and the fragmented of research, Available from: http://cai.ku.ac.th/article/Logistic_110653.pdf
    38. Singkarin D. , Somboonwiwat T. (2010) Where is value added of thailand rubber come from? , Thailand Engineering Journal, Vol.63 (1) ; pp.91-97
    39. Somboonwiwat T. , Chanclay K. (2009) Supply chain of Thailand rubber value added creation , Proceedings of the 9th Thai Vmcl, ; pp.64-75
    40. Tongchim P. , Rassameethes B. (2012) Relationship of supply chain management and performance measurement in rubber supply chains , WMS Journal of Management, Vol.1 (1) ; pp.58-66
    41. Tosanguan P. (2012) Supply chain of rubber industry and rubber products in northern Thailand , Panyapiwat Journal, Vol.3 ; pp.1-25