Rethinking the food system: an Operations Research approach

Rohmer, Sonja U.K.


The food system is a complex global structure, comprising an intrinsic web of inter-related supply chain and consumption activities. As such, it is deeply embedded in our society, contributing significantly to our economy and well-being. However, its current setup also leaves a considerable environmental footprint, by depleting valuable resources and polluting the planet, thus threatening the food security of future generations. A growing population and increasing standard of living further contribute to these environmental threats, while unhealthy consumption behaviour causes a rise in obesity and non-communicable diseases.

This thesis shows how Operations Research approaches can contribute to finding solutions for a more sustainable food system. By applying mathematical optimisation and solution techniques, the research reconsiders the system's setup and evaluates possible alternative scenarios in order to address the current challenges. In order to provide a holistic view of the system and consider the perspective of different decision makers, different decision levels are presented and investigated in this thesis.

In Chapter 2, the food system is considered from a network perspective, taking into account relations between consumption and supply chain decisions. In this context, a network flow problem is proposed to investigate the shifting towards a more plant-based dietary consumption on the basis of a number of alternative scenarios. The problem includes several echelons and interlinkages between different food supply chains by integrating sourcing, production and transportation decisions within a common framework. Consumption decisions are incorporated in the form of different types of consumer demands, maintaining a sufficient dietary intake level for the Dutch population. The problem is illustrated, with the help of real-life LCA data, on the basis of a case study and solved for different objectives using a linear programming approach. A multi-objective analysis, based on the epsilon-method and compromise programming, provides further insights into the existing trade-offs between the investigated environmental and economic objectives. The findings show that a plant-based dietary consumption holds the largest potential to reduce the environmental impact of the food system, while indicating the implications of such a shift for the supply chain configuration. Moreover, insights are provided on the allocation and shifting of burdens in the system depending on the chosen impact indicator.

Chapter 3 continues the investigation at the network level from a more nutritional perspective. Building on the modelling approach of Chapter 2, the research is more restrictive in terms of dietary intake choices and applies tighter nutritional bounds. Minimising several environmental impact indicators, the resulting consumption alternatives are compared with regards to environmental footprint, product mix and the underlying supply chain configuration. Given the nutritional emphasis, the comparison also includes the effect of different alternatives on the overall dietary intake. The findings indicate benefits of shifting towards a more plant-based consumption both from a health perspective as well as from an environmental standpoint. Highlighting the connection between meat and dairy products, the research also shows the importance of taking product relations into account.

Chapter 4 shifts the focus to operational aspects in the system, by addressing inventory management and routing decisions in the context of innovative last mile distribution concepts for perishable products. Assuming a two-echelon framework, the considered inventory-routing problem consists of a supplier, an intermediary depot and individual customer locations. The supplier delivers products to the depot, where storage may occur and from which they are then delivered by smaller vehicles to the customer locations. Storage at the depot incurs a holding costs, while customer preferences and availability for delivery are specified in the form of customer delivery patterns. Minimising total transportation and holding cost, the problem is formulated as a mixed-integer program. Given the complexity of the problem, a two-stage matheuristic is proposed for finding solutions on the basis of an adaptive large neighbourhood search and a reduced version of the problem. Three variants of the heuristic are compared in terms of their computational performance on a variety of randomly generated instances. Focusing on computational aspects, the findings highlight the importance of taking the cost structure into account when choosing the most suitable solution approach.

Another last mile delivery concept for the distribution of fresh products is considered in Chapter 5, investigating the effect of alternative delivery locations, in the form of customer pick-up points, on daily routing operations. Due to the existence of customer pick-up points, customers can either be delivered directly at the customer location, or indirectly through a pick-up point, where products are stored until pick-up occurs. Customer pick-up points allow for more flexibility, as direct delivery is restricted by tight time windows. However, storage is capacitated and requires cooling, resulting in an additional cost to operate the facility. Minimising total transport and storage cost, the location-routing problem is formulated using a mixed-integer program and solved by means of an adaptive large neighbourhood search. The heuristic is tested on a set of benchmark instances. The results from these experiments indicate the potential of incorporating customer pick-up stations in last mile distribution systems for fresh products to save costs and make delivery operations more efficient.

Zooming further into consumer plates, Chapter 6 looks at individual product concepts and how to design more sustainable alternatives to currently consumed products. Revisiting the shifting towards a more plant-based dietary consumption, the study focuses on the design of meat replacers with an equivalent nutritional contribution as chicken or beef, with regards to a set of key nutrients. Particular attention is given to protein quality and iron absorbability. Minimising different environmental impact indicators, a number of alternatives are proposed, as solutions to the linear programming based blending problem. Environmental impacts of ingredients are quantified through life-cycle assessment (LCA) data. The findings show that the largest impact reduction can be achieved through a vegan replacement, except for water use where the best result is provided by an insect-based replacement. The results further indicate the potential benefits of soy as an ingredient, due to its favourable amino acid composition.

Chapter 7 presents a general discussion and conclusion following from the main findings of this thesis.

The thesis highlights the multifaceted nature of challenges in the current food system and demonstrates the ability of Operations Research approaches to contribute to decision making on different levels in the system. At the same time, synergies between Operations Research and other food related disciplines give rise to new optimisation problems with practical implications, providing insights into different application areas.