Design and control of carbon aware supply chains

In this dissertation the impact of carbon emissions on the design and control of supply chains is studied. Increasing awareness for global warming and the role of greenhouse gasses in this has made companies more aware of carbon dioxide emissions caused by supply chains. As a result of this awareness, carbon emission regulations have been developed enforcing companies to incorporate a carbon cost (for certain activities in certain regions). Moreover, companies are voluntarily restricting their carbon emissions by specifying emission reduction targets, as a response to pressure from customers and stakeholders. In this dissertation we develop models with emission regulation and also with voluntary emission targets. We study well-known trade-offs in the field of operations management, such as between inventory and transport costs, by incorporating a carbon emission component, historically often neglected, and investigate the impact of the emissions on decisions. It is important for companies to take carbon emissions explicitly into account in decision making as carbon related costs are expected to increase in the future. Carbon emissions can be reduced to a certain extent by taking efficiency measures that both reduce emissions and costs. As companies can also invest in these measures from a pure cost perspective, we do not consider them in this dissertation. Moreover, it is likely that these measures yield insufficient emission reductions to achieve global emission targets. Hence, to achieve substantial emission reductions, measures that require investments, or increase operational costs, might be necessary. We explore several strategies for companies to reduce carbon emissions and investigate when a certain strategy is cost-effective. Examples of emission reduction strategies are to switch transportation to a mode with lower emissions, or to invest in production technology or off-shore production capacity. The focus of the research is on production companies and their carbon emissions generated during production and transportation activities, either to facilities of the same company or from suppliers or to customers. When considering emissions from transportation, we assume that transport is executed by a third party logistics service provider, as is often seen in practice. As a result, the control of the production company over the transport (emissions) is limited. The optimization of the load of the vehicle, and the traveled route is outside the control of the production company. However, the production company can decide which transport mode, or combination of modes be used, which determines the emissions to a large extent. In Chapters 2, 3, and 4, this emission reduction opportunity is studied in settings with one or multiple products and imposing the use of one or two modes. Then, in Chapter 5, the focus is extended to include emissions from production. We consider a company facing emission regulation for production and consider the possibility to invest in cleaner technology or to offshore production to a location without emission regulation. We next present a summary of the models and results presented in Chapters 2 through 5. First, in Chapter 2, we study the transport mode selection decision for a single product subject to emission regulation. We investigate the impact of different types of emission regulations and investigate under what circumstances a transport mode switch may occur. A transport switch implies that the selected transport mode in a setting with emission regulation differs from the selected mode in absence of emission regulation. The tradeoff under consideration is that a fast mode results in low inventory costs but in high transportation costs and emissions (costs), and vice versa. In a setting with stochastic demand we consider an order-up-to inventory policy including an emission cost. To accurately estimate the carbon emissions from transportation, we use a carbon emission measurement methodology based on real-life data and incorporate it into an inventory model. We observe that not the emission cost but the product characteristics, such as weight, density, and value, mainly determine which transport mode is selected. Consequently, a switch to a less polluting transport mode only results for a very high emission cost or if a product has a low weight or density or a high value. We find that even though large emission reductions can be obtained by switching to a different mode, the actual decision depends on the regulation and non-monetary considerations, such as lead time variability. Then, in Chapter 3, we consider a multi-item setting in which a self-imposed emission reduction target is set for a group of items. One item represents a combination of a particular product and a particular customer for which regular shipments occur, which determines the demand, product characteristics and the distance to be traveled. As the choice of transport mode (and corresponding transport costs) is up to the production company, the quoted price to the customer is also a decision variable. Since a single emission constraint is set for a group of items, the model is a constrained multi-item deterministic problem which can be solved using Lagrangian relaxation. Setting an emission target for a group of items allows for taking advantage of the portfolio effect: reducing emissions first where it is overall less costly. For a fixed emission target the transport mode that minimizes the total logistics cost is selected. If a range of emission targets are considered and we compare the cost-minimizing solutions, then it appears that two opportunities exist for the producer to reduce emissions: first of all, to select a mode that results in lower emissions per product shipped, and secondly to select a slightly higher sales price which results in lower demand and hence lower emissions. In a case study, we apply our model (with fixed sales price) to a business unit of Cargill and observe that emissions can be reduced by 10% at virtually no cost increase. Emissions can be reduced by at most 27% which results in a 30% cost increase. In an extension in which the sales price can be set, we observe that the portfolio effect results in at most 20% profit savings, a value which is relatively robust to price-sensitivity of demand. As in this case study road transport is the most polluting mode, larger emission reductions can be expected when air transport is used for shipments. Next, in Chapter 4 we examine the possibility to use two supply modes for a given product simultaneously, which is referred to as dual sourcing in inventory literature, in a multi-item emission-constrained setting with stochastic demand. By using two supply modes, a fast and a slow, one can combine the low transport costs and emissions (the slow mode) with being highly responsive (the fast mode) when required, i.e. in case of a stock out situation. As has been investigated in the literature using dual sourcing may result in lower expected period costs than using only a single mode. From an emission perspective using dual sourcing is beneficial compared to single sourcing since emission reductions can be achieved on a continuous scale. In some situations switching all shipments to a less polluting mode is too costly. Dual sourcing may then provide a large part of the emission reduction at a lower cost than using only the slow mode. We assume that a so-called single-index policy is used, which specifies two order-up-to levels: one for each mode. As a result of this policy, the fast mode is used when the demand in a certain period exceeds a certain value. Making use of a special case with exponentially distributed demand, we provide structural insights for a single product model. Then we extend these results to a model with two products and an aggregate emission constraint which provides insight into the more general situation with n products. In a numerical study we observe that if dual sourcing results in a cost decrease, then emissions can be reduced to a large extent without increasing the costs compared to using only a single mode. For a two-product setting we study if setting an emission constraint for a group of items is more or less beneficial if the products are more similar with respect to the value for one variable. We observe that the demand variability, and not so much for product weight and the penalty cost factor, has a large impact on how beneficial dual sourcing is, i.e. less similar products benefit less from dual sourcing. Lastly, we study the investments of a production company in production technology and capacity under asymmetric and uncertain emission regulation in Chapter 5. Asymmetric emission regulation refers to the fact that in different regions of the world different, or no, emission regulations exist and as result the emission price differs from region to region. We consider a producer of an energy-intensive good which incurs an emission cost for emissions generated during production. The company is deciding how much to invest in production technology in the regulated market, and how much capacity to build in a location with no emission regulation, the unregulated market. As emission regulation may result in off-shoring production and an increase in total emissions, regulators can implement measures to combat this undesirable effect. We refer to these measures as anti-leakage policies and study for each policy how it affects the company’s investment decisions and ultimately global emissions. We consider three different anti-leakage policies: Border Tax, which imposes a cost for products imported into the regulated region, Output-based allocation, which reimburses a certain emission cost per product produced in the regulated market, and Grandfathering, which reimburses a lump sum of emission cost, provided actual emissions exceed the amount. We consider four scenarios, one without an anti-leakage policy (baseline scenario) and three for the anti-leakage policies just described, and determine the optimal investment strategy and also the production strategy, which specifies how much to produce in each location given an emission cost realization, and the global emissions. We have observed that four possible strategies exist, two of which are to invest and produce only in one market (either the regulated or the unregulated) and two involve investment in both markets. When an anti-leakage policy is implemented and we compare the investments to the baseline scenario two effects may occur. First of all, less capacity may be built in the unregulated market, while not changing the production strategy. Secondly, it may result in the selection of a strategy with more production in the regulated market. We have applied our model to a data set based on a European-based cement producer and conducted a full factorial study for several important parameters. Overall we have observed that the grandfathering policy is preferred from both the company’s and regulator’s perspective. It is however, important to set the reimbursement not too low or too high. Finally, in Chapter 6 we present the conclusions of the research presented in this dissertation and provide directions for future research.