Sign Out
Logged In:
Tab Image

Challenges of operational research in agricultural value chains

This article introduces operational research work in the agricultural sector. It serves as an introduction to the JORS paper Challenges of operations research practice in agricultural value chains,  authored by A J Higgins, C J Miller, A A Archer, T Ton, C S Fletcher and R R J McAllister. The paper is available to OR Society members in the Journal of the Operational Research Society (2011) ) 61, 964–973. The paper is freely quoted in this article, and its authors have approved this note.

Operational research (OR) methodologies in optimization have been extensively applied to problems in different agricultural value chains in recent years. This paper takes a critical stock take of such applications to date, and reflects on their contribution to value chain sustainability and resilience. The stock take shows that the rate of industry or policy adoption has been limited, partly due to the complex interactions across the segments of agricultural value chains, and the mathematical representation being different to the way the decision maker understands the problem. OR practice in agriculture is also being asked to cover greater spatial scales and engage more stakeholders, and is required to embrace resilience and sustainability objectives. A single-minded focus on optimizing parts of these complex systems without considering the whole system is no longer adequate, and new methods and approaches are required. Complex systems science methods are being applied to analyse the dynamics of complex social–ecological systems, and are starting to find a home in industrial supply chain analysis. The relevance and utility of OR in ensuring the success of agricultural value chains into the future will require practitioners to understand and model value chains as complex adaptive systems.


The agricultural value chain is fundamental to the survival of human society, the growth or maintenance of regional and national economies, and the wealth and welfare of individual producers. Sustainable supply, particularly in a dynamic and uncertain environment, is a major challenge. Climate change, natural resource degradation, societal demands and global markets all place pressures on agricultural supply. Society increasingly expects agricultural production and supply to exist within a framework of social and environmental responsibility. A challenge for OR practitioners is to incorporate these conceptually and empirically complex issues. Many past applications of OR in agriculture have focused on decomposing the problem into components, after which mathematical techniques were applied to identify optimal solutions for each component. This approach has limitations if implemented in agricultural applications where there are complex interacting drivers in productivity, markets, the environment and people.

Addressing challenges for OR practice in agricultural chains

An agri-food value chain is typically mapped as a linear sequence of activities, from primary production through to the consumer and waste management . Many value chain opportunities are well suited to OR methods, particularly those that can be modelled as isolated activities with respect to others in the chain. These include transportation of the raw product from farms to production facilities and transportation to end customers, which have a low level of uncertainty and social complexity. Such an application in the dairy industry was considered by Wouda  while Gigler  developed a model for generic agricultural chains. (For references, see the main article). Tatsiopoulos and Tolis  developed a comprehensive model to optimize transport and distribution for cotton biomass chains, which was able to account for co-generation of electricity. Such value chain models will work more readily in an industry where there is single ownership or control across the farming, transport and processing segments, rather than the multiple ownerships that are more common in agricultural value chains. These types of applications have clearly defined objectives, minimal uncertainty in model inputs and can be implemented with minimal change management issues.

Application of OR methods to agricultural chains expanded rapidly in the mid-1990s as seen in the increase in publications in the OR and agricultural literature. However, various reviews suggest that the success of these models in improving system efficiency and optimization has been limited. Part of the problem is that industry members (and OR practitioners) often do not have a whole-of-chain perspective, and may have difficulty in perceiving the ultimate ends of the chains and their role in the functioning of the whole chain. For example, the production/harvest segments may consider the primary processors as their client, while the processing segments may focus on selling to retail distributors. Retailers or consumers may consider agricultural primary products as simple inputs to the chain. This tunnel view of the chain may occur as major chain participants (such as in production, processing or retailing) organize the immediate chain segments around themselves. OR practitioners have focused on a limited number of system components or chain segments in an attempt to reduce complexity of fitting a model to the system.

Whole-of-farm planning has been a very popular application in OR and economics. While they are usually not value chain focused, they provide insights to the limitations of OR in agriculture in the past, which are relevant at the value chain scale. Optimal selection of various crops or livestock over a multi-year planning horizon to maximize profitability has been studied quite extensively and models have accounted for capital and natural resource constraints such as water. OR methods have also been developed to assist farmers in the combined selection of farm machinery and labour activities. However, the sociological processes that influence decision-making by farmers were not considered in these OR applications, which is likely to have contributed to the poor adoption of these tools. Other models attempt to overcome limitations of OR models in whole-farm planning by extending the scope of the mathematical programming models to accommodate environmental objectives and uncertainty. These models have been more useful for simulating strategic planning and policy options (mainly for research purposes) rather than being implemented explicitly by farmers.

Multi-agent modelling

This approach can be used to examine how complex behaviours, system structures or system dynamics evolve or emerge over time from relatively simple local interactions between systems components. Multi-agent modelling was originally designed for reactive, reliable and reconfigurable operations management systems, and has been extensively applied to manufacturing and information systems. They have recently been applied to agile chains in manufacturing and sophisticated chains under dynamic conditions. The basic idea of multi-agent modelling is that it uses rules of behaviour of autonomous entities (called agents), as well as rules of their interaction to simulate the consequences of these rules across the network of entities. Multi-agent models have the capacity to represent how complex behaviours, system structures or dynamics change or emerge over time from a relatively simple ‘bottom-up’ set of rules for entities (or agents). Agents exhibit characteristics that allow them to behave individually and interact with each other to fulfil system-wide goals.

Dynamical systems models

The motivation behind dynamical systems modelling is that one cannot, in general, optimize a dynamically varying system by optimizing a sequence of static system configurations. Optimizing a system for a static configuration takes no account of the pathways of change and the rates of change possible within a system. However, the rate at which management decisions or internal stabilizing mechanisms respond to external variability or stimuli is often the limiting factor in determining the capacity of a system to survive an unexpected perturbation, or respond to variability.

In an agricultural context, for instance, if we knew the average environmental conditions for the coming year, such as rainfall and growth rates, we might be able to identify ‘static’ optimal management characteristics that would maximize the sustainable productivity of our land. In reality, environmental conditions will vary throughout the year, but for each set of environmental conditions we experience, we should be able to identify different optimal management characteristics. However, simply moving from one set of optimal management characteristics to the next, as environmental conditions change, will not guarantee optimal performance across the entire year: instead we need an optimization framework that can incorporate the rate at which management can detect and respond to changing and variable conditions, even when those conditions can not necessarily be predicted ahead of time, to provide optimal sustainable output over the entire year.

Dynamical systems modelling can be informed by the results of multi-agent models and of agricultural value chain network models. It can be used to assess the capacity of the entire network to persist in an external environment of variability and unexpected perturbations or, equally, to assess the capacity of a single component of the system to survive when it is driven by variable input supply and output demand as part of the variability of the larger system. In addition, the results of dynamical systems analysis can be fed back into the network analysis of the agricultural value chain to assess the capacity of the system to survive a sequence of cascading failures throughout the system. Dynamical systems modelling can also capture the impact of situations in which a component of the system impacts other parts of the system; a fundamentally dynamic process.

Conclusions and future research

We have undertaken a critical stock take of OR applications to date, and reflected on their capacity to address the complexity inherent in agriculture value chains, particularly where resilience and sustainability are key system goals. Our stock take not only highlighted a worrying lack of practical adoption of OR research findings in agriculture value chains, but also provided new insights as to why adoption was not achieved in most cases. A key conclusion from the stock take was that the scope of value chain analyses in agriculture have expanded significantly in recent years with a greater industry and community requirement for these chains to be resilient as well as efficient, and to provide products and services over and above the supply of cheap agricultural produce. Increasingly, the questions being asked of OR practice cover much larger spatial scales, and engage a greater range of stakeholders. Larger scale, multi-sectoral problems are also complex socially and environmentally.

Consequently OR in agriculture value chains needs to embrace resilience and sustainability objectives, which may be more critical practical drivers for market share than just efficiency of supply. Complex systems science methods (namely multi-agent modelling, dynamic systems and network theory) have been applied extensively to analyse the sustainability of natural landscapes, and are starting to find a home in industrial supply chain analysis. We have indicated how these can be used for representing value chains, their participants and associated behaviours, while observing emerging patterns arising from interactions among them. The ‘art’ side of modelling is to select among a mix of different OR models in an integrated environment. Using some recent examples, we showed how OR practitioners are beginning to apply complex systems science methods to answer some questions being asked of agriculture value chains in order to increase the future sustainability of these industries. The philosophical underpinnings embedded within the application of OR methods are as important to its success as its technical contributions. Our paper reflected on the thinking embedded within OR's new methodological tool kit and adds a qualitative assessment on these new directions.

There remain a number of important methodological and fundamental questions to be answered in the application of complex systems science methods to OR. These include: ‘How can we best evaluate value chain systems and their dynamics?’; What is the relationship between resilience and value chain function (ie how much resilience is good)?’; ‘Can we determine the most appropriate point of trade-off between efficiency and resilience in a dynamic environment?’; and ‘What aspects of the value chain (eg water resource utilization, production processes, rural social institutions) are critical for resilience and sustainability?’ The relevance of OR to the success of agricultural value chains will depend on the capacity of practitioners to mentally model the value chain as a complex adaptive system, and to represent the relevant dynamics and actors in these complex systems in their models.

Full version first published to members of the Operational Research Society in JORS June 2011