Resilience in agri-food supply chains: a critical analysis of the literature and synthesis of a novel framework

2.1 Step 1: review question formulation

The first step in an SLR is the formulation of a specific, purposeful, review question to determine the scope and focus of the review. The aim of this review is to comprehensively identify definitions, elements and strategies for resilience and to develop a holistic framework for how they apply to AFSCs. Hence, this review aims to address Q1.

To help structure the answer to this question, three sub-questions have been identified as follows:

2.2 Step 2: locating relevant literature

The purpose of this phase is to design search criteria in such a way as to ensure the identified literature is comprehensive enough to capture all salient points relevant to the review question (Denyer and Tranfield, 2009). One of the key research gaps driving this review was the need to cover a variety of fields relevant to AFSCs, not simply SCM, and therefore avoiding bias in selection was vital. Therefore, the following multi-database, cross-disciplinary online citation services were used: Google Scholar, Web of Science, ProQuest, Science Direct, Wiley Online, Emerald and Scopus. Consistent with a number of other SLRs in the area of resilience, this paper used a number of defined keywords as search criteria as summarised in Table I. The search was performed in December 2016, and the search for keywords was restricted to title and abstract. Keywords were initially selected based on the authors’ collective knowledge of the field which enabled them to draw up a long list of terms commonly associated with resilience in the literature. Following standard SLR practice (Hohenstein et al., 2015; Kamalahmadi and Parast, 2016), these were critiqued and validated through consultation with other research colleagues allowing us to arrive at the shortlist presented in Table I.

Search strings were composed of primary keywords and secondary keywords. The primary search phrase used in all databases was either “Community”, “Socio-Ecological System” or “Supply Chain”. Each primary search phrase was accompanied by AND “resilience/resiliency”. In addition, each search involved a secondary keyword which was one of either: “Risk/Risk Management”, “OR Vulnerability”, “OR Volatility”, “OR Security”, “OR Mitigation” or “OR Business Continuity”. These variations were run exhaustively, e.g. “Community” AND “Resilience” AND “Security”.

2.3 Literature selection and evaluation

From the initial search criteria, this review sourced a total of 1,270 articles. To maintain transparency and to ensure fit of identified material to the review question, stringent selection criteria were applied to this initial search pool. While material was not limited by publication date, materials were restricted to those published in the English language. Additionally, in line with other SLRs in the area of resilience (Hohenstein et al., 2015; Kamalahmadi and Parast, 2016), material was limited to peer reviewed publications as an indicator of the academic rigour of identified literature (Light and Pillemer, 1986). Once duplicates, non-peer reviewed results and non-English publications were excluded, and the remaining pool numbered 239 articles. Scanning of Introductions and Conclusions provided a better understanding of the fit of the material to the review question and its associated sub-questions. At this stage, 104 articles were excluded due to either being inaccessible (six articles), or being beyond the scope of AFSC-relevant resilience definitions, elements and strategies. Work cited in all accepted articles was also scanned for titles that matched the keyword criteria. In total, this provided a final review size of 137 articles (Figure 2).

2.4 Step 4: analysis and synthesis

The objective of this stage was to analyse and synthesise the final literature pool of 137 articles to identify new knowledge about the multi-disciplinary concept of food SCRES that would not have been apparent from reading each of the papers individually. Analysis was conducted using a Microsoft Excel spreadsheet to record summaries of the positions of each of the 137 articles regarding the key resilience concepts of definition, elements and implementation strategies. Synthesis was achieved via an integrative approach which compared multi-disciplinary works for convergent, divergent and co-evolving understandings of the aforementioned resilience concepts and used the results to build a synergistic conceptual framework of food SCRES. This was chosen over alternative approaches to synthesis, such as aggregative approaches as evidence suggests it better suits heterogeneous source material (Rousseau et al., 2008).

2.5 Step 5: reporting and using the findings

In this stage of an SLR, the findings from the analysis of the entire review pool, the relationships between salient concepts and the extent to which this information has addressed the review questions are reported (Denyer and Tranfield, 2009). Typically, synthesised findings can also be applied in a novel context to help generate new understandings of the relationships between concepts that may have been studied in isolation in the literature. In the context of this paper, Section 3 reports the findings of the review in relation to the review question and sub-questions. It proceeds to then synthesise and apply the findings in the form of a holistic framework that models resilience in AFSCs.

3.1 Descriptive analysis

Table II highlights that 40 per cent of the final 137 works reviewed originated in one of the leading seven journals of which Supply Chain Management: An International Journal and International Journal of Production Economics were the most popular. All of the aforementioned journals represent either the fields of SCM or operations management, in which the priority of resilience efforts tended to focus on business continuity and particularly competitiveness of individual actors (Hohenstein et al., 2015; Pereira et al., 2014; Elleuch et al., 2016a, 2016b). Indeed, when all publication sources are considered, 75 per cent of all articles considered in this review have an SCM or operations management origin. However, less common but still important sources of resilience literature were found in journals from a range of other disciplines which included ecological systems, social systems and engineering/physical systems as outlined in Figure 3. These alternative disciplines are an important source of resilience research, particularly publications with a focus on social systems, where the priority of resilience tends to be on the adaptive capacity of complex systems (Tendall et al., 2015; Milestad and Darnhofer, 2003). The authors feel that this supports the previous contention that existing works which explore resilience from an SCM and/or operations management perspective, with their focus on individual business continuity and competitive advantage, are not always readily transferrable to the topic of AFSCs.

Another notable observation is that all of the articles reviewed were published post-2000 with 65 per cent being published post-2010, suggesting that interest in the application of resilience as a concept is recent and growing phenomena. Evidence suggests that this is in response to a number of wide ranging and unexpected disruptions including Hurricane Katrina, the Icelandic eruptions at Eyjafjallajökull in 2010, the Fukushima nuclear disaster, as well as major terrorist incidents such as the 9/11 attacks in America and the 7/7 attacks in the UK (Kinsey et al., 2007; Sheffi, 2001; Scholten et al., 2014; Christopher and Lee, 2004).

Figure 4 analyses the literature according to its adopted methodology. Methodology is classified according to four categories borrowed from Natarajarathinam et al. (2009): conceptual/theoretical, analytical, empirical and applied. The term conceptual/theoretical refers to works which synthesise or develop existing understanding of SCRES but which are not supported by any empirical work. Literature reviews are classed within this category. Works involving substantial simulation or mathematical modelling of a real-world supply chain issue with specified parameters fall within the analytical category. Articles that involve the collection of real world data and its evaluation are classed as empirical. Finally, case studies, interviews and other forms of gathering thoughts and opinions are classed as applied. Ultimately, the most common form of methodological approach was conceptual/theoretical which accounted for 52 (38 per cent) of the reviewed articles.

The authors of this review concur with Hohenstein et al. (2015) that this represents the importance of theory building in what is still a relatively new research area. Encouragingly, in recent years, there have been an increasing number of empirical works, case-specific applied works and mathematical analysis-based works which suggests that the focus is moving away from defining resilience towards trying to understand what its functional “elements” are. However, a large number of such works attempt to measure or model resilience based on a very small number of commonly cited “elements”, particularly flexibility and redundancy; for example, Braunscheidel and Suresh (2009), Skipper and Hanna (2009) and Datta et al. (2007). While such works are highly valuable in the sense that there is a real need to empirically validate emerging resilience theories, there is a risk that resilience as a practical target could be oversimplified. This descriptive analysis of the resilience literature will now be used as a base from which to explore each of the review sub-questions individually.

3.2 Addressing Q1.1

This section addresses review sub-question one by exploring how resilience has been defined as a concept by different research fields (Figure 3). Resilience can best be thought of as an umbrella term for a range of linked factors that help ensure continuity in the face of disruption (Tendall et al., 2015). Before exploring the concept in more detail, it is important to provide clarity on the relationship between resilience and the similar terms of “sustainability” and “robustness” which investigation suggests are sometimes used interchangeably. Using the definition of sustainability outlined in the Brundtland Report: “meeting the needs of the present without compromising the ability of future generations to meet their own needs”, sustainability can be described as a normative measurement for assessing long-term performance against ideal environmental, economic and social standards (Derissen et al., 2011).

By contrast, resilience is more of a descriptive methodology concerning short-term ability to withstand and/or adapt to disturbance (Tendall et al., 2015). As such, it is a key attribute for any organisation with long-term sustainability goals in complex systems with ever-changing drivers. Thus, an organisation can be resilient and unsustainable, but not sustainable without the presence of resilience, as it would be too susceptible to short-term derailment from excessive exposure to disruption. Robustness is another term which is related to resilience and frequently used interchangeably. However, the two are separate terms, with robustness prioritising strength to withstand disturbances, whereas resilient systems include flexibility to adapt to disturbance (Asbjornslett, 1999; Jüttner et al., 2003). In this way, it is possible to see robustness as a component of resilience, and in turn, resilience as a short-term enabler of long-term sustainability (McDaniels et al., 2008). To summarise, while these terms would therefore appear to be synergistic, it is erroneous to use them interchangeably (Redman, 2014).

Moving on to focus on resilience, while a relatively new addition in the context of SCM and AFSCs in specific, it is by no means a new concept. The term has Latin origins, stemming from the word “resi-lire”, meaning to spring back and was first used by physicists to describe the stability of materials and their ability to resist external shocks (Manyena, 2006). It entered popular use in the field of Ecology in the 1960s and from there began to be translated to a range of new subject fields aided by a seminal article by Crawford Stanley Holling in 1973 (Holling, 1973). This article divided resilience into two distinct definitions that are commonly used today: engineering resilience and ecological resilience.

In the engineering definition, resistance to disturbance and the speed by which the system returns to a state of equilibrium are the marks of resilience. The phrase “a state of equilibrium” refers to the notion of optimal day to day operations (Rose, 2011). Heavy emphasis is placed on return time, efficiency, constancy and predictability, which it is claimed are the marks of a sound engineering design and hence the name (Holling, 1996). In the ecological definition, resilience is also measured by resistance to disturbance and speed of return to a state of equilibrium, but this definition also accepts that there are multiple possible equilibriums that the system could flip into depending on the magnitude of the disturbance.

It has been pointed out that a major shortcoming of both the engineering and ecological definitions of resilience is that they presume closed systems within which different actors can establish states of equilibrium. This is clearly not the case in something as complex as a food system where intertwined social, environmental, economic and political factors drive constant change across key operating parameters. In response to this, several authors have proposed a third definition of resilience which has been termed “Evolutionary” or “Adaptive” Resilience (Walker et al., 2006; King, 2008; Folke, 2006; Ambulkar et al., 2015). For consistency, we use the term adaptive resilience from now onwards in this review.

Adaptive resilience describes complex social–ecological systems where the interactions between different scales (for example, from individual species, to forests, to entire ecosystems), periods (referred to as temporal scales) and geographic distances (referred to as spatial scales) are all considered vital for overall system resilience. As such, there cannot be a “state of equilibrium” because external interference is continuous. Instead, resilience is something that is cyclical and cumulatively developed by a continual process of adaptation and learning from ongoing disturbances. It has been proposed that this continuous adaptive cycle has four distinct stages: exploitation, conservation, release and reorganisation as shown in Figure 5 (Allen et al., 2014; Walker et al., 2006).

The first phase is exploitation, which in the context of a business, is marked by use of readily available resources to form structure and core business priorities. An example might be that of a new start-up company with a novel product and market dominance. However, as an organisation grows, it will eventually reach a point where its size binds ever larger quantities of resources and its connectivity increases cross-scale interactions, known as the conservation phase. The existence of the phase is supported by evidence collected by Peck (2005) in multi-sectorial supply chain interviews. An example view expressed by a consultant in Electronics Manufacturing is:

It’s when the supply chain is supposed to be in the established steady state that it is most vulnerable, because that’s the point when it’s most susceptible to external effects. That’s when most people are trying to optimise and reduce control limits to reduce the variability of the process, but external risks may have changed the original scenario. (Peck, 2005).

In AFSCs in specific, this phase has been likened to contemporary drives towards intensification of agriculture and centralisation of factories and distribution centres, representing accumulation of capital and growing interconnectivity. Other assets bound up in AFSCs include significant amounts of land, water, carbon and other nutrients embodied in food (Fraser et al., 2005).

Whether from the perspective of an entire food system or a single business within the existing system, this phase is where susceptibility to disturbance is at its highest because so many assets are tied up in the current way of doing things and connectivity means exposure is at its highest. There is the potential for significant loss of resources if a big enough disturbance occurs, and this is known as the “Release” phase. This does not necessarily comprise pure financial loss, but might also concern loss of resources bound up in no longer tenable business structures. The business does not necessarily collapse at this point, but there will need to be some sort of adaptation (the Reorganisation Phase) at which point the cycle begins again (Davoudi et al., 2012).

Influencing the adaptive cycle are three components: “resilience” (capacity to absorb change), “adaptability” (capacity to evolve a given form of operation) and “transformability” (ability to completely change an untenable system of operation). These are effectively control mechanisms with which an organisation can influence the adaptive cycle stages (Figure 5). The adaptive cycle also differs from the engineering and ecological definitions of resilience by its underlying consideration of “Panarchy” (Allen et al., 2014). This represents complexity in a system where disruptions do not necessarily have to originate within the same period or geographic proximity as the focal organisation. This means that the relationships between cause and effect of a disturbance do not necessarily have to be linear. As such, small influences such as the input of single staff members in the face of disruption can have just as much, or even more, impact than large scale interventions. Such unpredictability challenges the adequacy of conventional risk management tools, such as extrapolation of past trends as a way of forecasting future events (Trkman and McCormack, 2009). The key differences between the engineering, ecological and evolutionary definitions of resilience are summarised in Table III.

In Table IV, the review pool is analysed according to which of the three definitions authors adopt. In total, 48 of the 137 articles being reviewed offered a definition for resilience. As Q1.1 concerns identifying suitable definitions of resilience for AFSCs, literature definitions were compared on whether they were from articles considering AFSCs in specific, or from different perspectives on resilience. Thirteen of the articles offering definitions considered AFSCs in specific (although this specificity was not always obvious in the definition provided) and 35 were more general in focus. The broader research contexts of the review articles were also compared to identify if certain research fields prioritise a specific type of definition.

Engineering definitions were distinguished by their focus solely on resisting and recovering rapidly from external disturbances with minimal impact on system deliverables. Ecological definitions, on the other hand, focussed on the amount of change a system can endure and recover from, possibly involving moving to a new equilibrium, while maintaining core functions. Adaptive definitions made no mention of states of equilibrium but instead focussed on adaptive change to volatile external operating environments. As such, in addition to mention of ability to “resist” and “recover”, the ability to “adapt” or “reorganise”, whether in response to, or in anticipation of a disruption was common in such definitions (Wu et al., 2013; King, 2008; Cimellaro et al., 2010).

It was identified that overall there was a slight preference for the adaptive definition of resilience. This is particularly true in works that were AFSC specific in focus, many of which hailed from a social systems perspective. Such works frequently considered resilience at community and societal scales and prioritised a system’s ability to continue providing food, rather than economic viability of individual businesses within the chain. Here, end consumers and the different AFSCs that feed them are considered within the sphere of the wider natural world, where change is constant and control over that change by any given actor is small. For example, as complex social-ecological systems, AFSCs are dependent on a number of ecosystem services to produce food, and significant social-economic factors to manufacture and transport food. Each of these is exposed to vulnerabilities, for example, in the form of policy interventions, consumer demand and environmental management.

A breakdown in any one of these areas can lead to harvests failing, transport links breaking and consumer demands and tastes changing (Milman and Short, 2008; Yang and Xu, 2015). Therefore logically, to be resilient in such a world is to prioritise constant adaptation and reorganisation. As such, the complexity of vulnerability sources is much broader than an individual organisation might consider from a risk management perspective, and this would explain the observed preference for the adaptive definition. Key features of adaptive food definitions included the ability to maintain “function” as well as the ability of systems to adapt rather than to return to existing states of equilibrium. Tendall et al. (2015) advance the field by linking “function” with the United Nations Food and Agricultural Organisation definition of food security which concerns the four pillars of availability, access, utilisation and stability of food to end consumers (Fao, 2012; Tendall et al., 2015).

In comparison, non-AFSC-specific works saw greater contribution from organisational management and engineering/physical systems approaches. In these contexts, resilience consideration often takes place within an enclosed system, for example, a factory, and vulnerabilities tend to be more controllable and predictable (for example, machine faults, staff illness, etc.), thus encouraging pursuit of a single optimal management strategy (Vlajic et al., 2013; Berle et al., 2011). This can be seen as analogous to a “state of equilibrium” and would explain the preference in such works for an ecological definition of resilience where the focus is on a particular organisation’s competitive advantage, specifically, minimising the time and cost of a disruption and exploiting competitor weaknesses (Pereira et al., 2014; Kim et al., 2015; Todo et al., 2015).

Articles in the area of SCM, regardless of whether they are AFSC focussed or not, have shown a growing shift away from engineering definitions of resilience towards adaptive definitions in recent years. There is evidence that this transition is linked to increasing awareness of the importance of constantly changing operating environments, in particular, the evolving challenges and opportunities of outsourcing to low-cost countries (Tang and Musa, 2011).

Moving forward, a number of definitions in Table IV refer to one or more of the following abilities: to “Resist”, to “Recover” and/or “Adapt”(Soni et al., 2014; Fahimnia and Jabbarzadeh, 2016; Leat and Revoredo-Giha, 2013; Annarelli and Nonino, 2016). Ponomarov and Holcomb (2009) categorised these into the distinct phases of readiness, response and recovery. Readiness refers to an organisation’s ability to anticipate disruption and either prepare for it or avoid it. Response refers to either innate or pre-planned elements that mitigate the impact of a disruption, as it happens. Recovery refers to the ability of an organisation to repair losses caused by a disruption and return to meeting core priorities. Hohenstein et al. (2015) add the fourth phase of “Growth” which concerns learning from and adapting core priorities post disruption so that competitiveness actually improves compared to pre-disruption levels. However, it has been noted that many articles overwhelmingly see disruption in light of the reactive and recovery phases only (Higgins et al., 2010; Hohenstein et al., 2015).

Therefore, to summarise findings in relation to review Q1.1, it has been identified that resilience of AFSCs is frequently equated with the ability not only to resist disruption but also particularly to maintain the core function of supplying food to end consumers. The priority of resilience in AFSCs can therefore be described as the food security of end consumers. AFSCs are also incredibly complex systems involving myriad bio-geophysical, social, economic and political drivers and feedbacks that must be managed holistically to enhance resilience. Therefore, any definition of AFSC resilience must include the ability to adapt in line with changing operating environments as well as to prioritise availability, access, suitability and stability of food supply. To do so, it must consider more than the traditional phases of resisting and recovering from disruption and also include anticipation and post-disruption learning. Therefore, the authors of this paper propose the following definition of AFSC resilience:

The collective ability of Agri-food supply chain stakeholders to ensure acceptable, sufficient and stable food supplies, at the required times and locations, via accurate anticipation of disruptions and the use of strategies which delay impact, aid rapid recovery and allow cumulative learning post-disruption.

This definition builds on existing adaptive definitions of food-related resilience by incorporating the priority of food security rather than individual organisational competitiveness. By nature, it implies that resilience strategies must consider how resilience strategies implemented by one actor impact overall SCRES. Furthermore, by incorporating the fourth food security pillar of stability, the synergistic relationship between resilience and sustainability is highlighted.

A key component of this definition is the word “mechanisms” and to explore what this practically entails; this review now moves on to Q1.2 to identify AFSC relevant resilience “elements” and “strategies”.

3.3 Addressing Q1.2

There have been a number of works which propose strategies for manipulating an actor’s resilience, many of which fall within the SCM discipline. Such strategies frequently rely on the assumption that resilience can be controlled by a portfolio of variously named “antecedents”, “attributes”, “capabilities”, “elements” and “enhancers” which are management tools to counteract specific vulnerabilities (Hohenstein et al., 2015; Pettit et al., 2010; Pereira et al., 2014; Kamalahmadi and Parast, 2016). For consistency, and in line with Christopher and Peck (2004), Hohenstein et al. (2015) and Kamalahmadi and Parast (2016), the term “elements” is used from now onwards.

In total, 61 articles proposed one or more key elements for resilience. From these, this review identified 40 unique resilience elements. This breadth of resilience elements has, to the author’s knowledge, not been attempted previously in the literature. These elements varied significantly in terms of “scope”. This refers to whether resilience elements were applicable in response to disruptions within an individual organisation (for example, machinery faults) or within a supply chain (for example, loss of a specific supplier), in which case, elements addressed ways in which the supply chain could collectively adapt. The list of identified elements and their respective scope and publication sources are given in Table V. It should be noted that some elements appear in both the intra-organisational and intra-supply chain columns albeit with different contexts. For example, redundancy at an organisational level refers to spare capacity and inventory, but at a supply chain level describes alternative transport routes between stages or backup infrastructure. When ranked according to the number of papers mentioning a specific element, flexibility, risk aware culture, redundancy and early warning detection systems were the most commonly cited elements at an organisational level. At a supply chain level, collaboration, flexibility, agility, visibility and adaptability were, respectively, the most commonly cited elements.

Despite there being a number of highly cited resilience elements, the overwhelming majority of elements identified appeared in less than 10 per cent of papers reviewed. This suggests that there is poor consensus on what elements are the most important for resilience. For example, Fiksel (2003) proposes four elements: diversity, efficiency, adaptability and cohesion. Pettit et al. (2010) on the other hand identifies 14 different elements. Without empirical validation, it is difficult to be sure that just because a resilience element is cited more frequently, that it is more significant for resilience than a less commonly cited capability. In particular, many of the less commonly cited elements are from research fields that are less active in the area of resilience, such as ecological and social systems. Such elements concern interactions and relations between organisations, communities and the natural environment as well as their ability to adapt, which are of major significance to “adaptive resilience” in AFSCs. Therefore, there is a need to capture the relationship between such elements and the more commonly cited elements.

The authors of this review identified that of the 40 resilience elements, some were broad in scope and some were much narrower, referring to specific aspects of the broader elements. These are referred to as “Core” and “Supporting” elements, respectively. For example, at a supply chain level, the authors of this review propose that flexibility is a “Core” resilience element, concerning supply chain wide alternative options of responding to a disruption. The resilience elements of “I016 Knowledge Management” and “IO16 Market Position” for example, while enabling resilience in their own right, are often cited as aspects of “IO1 Flexibility” (Carvalho et al., 2012a, 2012b; Pettit et al., 2010; Tomlin, 2006). Therefore, while IO15 and IO16 are not duplicates of IO1, they can be seen as “Supporting” resilience elements. This novel method of categorising resilience elements of relevance to AFSCs is shown in Figure 6. Categorising resilience elements in this way is useful for application to AFSCs because elements that represent communities and ecosystem services can be more easily recognised as supporters of more commonly cited elements. The proposed “Core” elements at an organisational level are now described in more detail.

3.4 Addressing Q1.3

This paper has so far analysed the multi-disciplinary definitions, elements and strategies concerning resilience and identified aspects that are of importance to AFSCs. In addressing Q1.3, this paper will now synthesise the identified multidisciplinary aspects of resilience into a conceptual framework of AFSC resilience.

As identified in addressing Q1.1, AFSCs are complex socio-ecological systems with interactions occurring across different scales, distances and periods, all of which must be assessed together to accurately model resilience. This review has identified a number of unique food system challenges, summarised as follows (Diabat et al., 2012; Higgins et al., 2010; Taylor and Fearne, 2006):

1. A network of potentially thousands of participants, in stark contrast to the widely accepted view of a linear buyer–seller chain reaching from farm to consumer.

   • It is important to appreciate that a vast range of supporting dependencies such as equipment suppliers, fuel infrastructure, financial services and logistics, among others, enable food to reach end consumers.

2. Strong social drivers, such as health, lifestyle, the need to protect the natural world, as well as economic goals.

3. Strong genetic, environmental and climatic variability.

   • Food products are naturally variable in colour, shape and size even before the effect of the growing environment and particularly climate change are considered in terms of their effect on yield.

4. Low-value end products.

   • Food is typically purchased frequently and represents a low proportion of household expenditure in relation to other consumer goods such as electronics (although it is accepted that proportion of household expenditure can vary significantly depending on location).

5. Declining margins.

   • A range of factors, in addition to those described in challenge (d), including globalisation and the increasing dominance of large multiple food retailers, are driving ever lower margins in AFSCs.

Figure 8 explores how these challenges manifest as unique risk sources for each of the traditional AFSC stages: primary production, processing, distribution, catering, retail and consumption (Elleuch et al., 2016a, 2016b).

Primary producers face a range of natural stressors which put yield and quality at risk, such as disease and bad weather, as well as anthropogenic damage to natural capital such as pollination, soil fertility and water access. Historically, they have also faced major downstream pressure from buyers which has squeezed their margins, often driving smaller farmers, and thus production diversity, out of business.

Food processors, who historically held much more supply chain power, are similarly facing downwards pressure from large retailers, favouring “lean” approaches which reduce non-value adding activities, thus reducing flexibility and redundancy (Van der Vorst and Beulens, 2002). Increasingly, viability is dependent on brand differentiation, a gap which retailers are fast closing with their own “private labels”.

Wholesalers are traditional stock holders in AFSCs and major risks stem from a reduction in customer base as smaller “cash and carry” buyers are being replaced by large supermarkets. Catering is commonly the biggest source of value in modern AFSCs. Key strengths include customer responsiveness and diversity, although there is some risk from market concentration.

Retailers themselves are often described as the gateways of modern AFSCs due to their market share and proximity to end consumers. Yet, the “Just in Time” models which enable them to offer high variety and value leaves them at risk of supply disruption. Their proximity to consumers also means that they can have less time to react to changing consumer demands.

The resilience of each stage described so far is vital in ensuring food security, or in more specifically, that food is physically available (ready for consumption in principle), accessible (somewhere the consumer can access it), acceptable (in a form that is culturally acceptable), safe and reliable.

In addition to their own unique risks, all of the stages together are influenced by a number of overarching risk sources in the wider social, political, environmental and economic spheres (Colicchia et al., 2010; Vlajic et al., 2013). These risk sources can often be separated by significant distance and even periods from a given organisation or supply chain and their impacts are not linear (Vlajic et al., 2012). For example, recent extreme weather in key regions of Spain and Italy decreased production by as much as 60 per cent. Due to retailer sourcing policies across the continent, many initiated decades ago, where focus was placed on a relatively small number of large-scale intensive producers, often purely for economic reasons, large sections of Europe suffered severe vegetable shortages in the winter of 2016-2017. Due to the growing times of crops, and length of buyer contracts, such disruptions can take many months to resolve (Food Navigator, 2017). Thus, it is vital that distance, time and scale are considered together.

One approach to addressing this issue is to break down AFSCs into constituent stages and optimise them based on average operating conditions, perhaps by identifying resilience elements and strategies as has been attempted at a farm, processing, retail and community level in the literature (Milestad and Darnhofer, 2003; King, 2008; Leat and Revoredo-Giha, 2013; Macfadyen et al., 2015). However, optimising individual stages of a supply chain in this way does not necessarily allow them to adapt to novel situations, and it is possible that optimising one stage may be detrimental to upstream or downstream stages which is unacceptable if the end goal is a more reliable food system overall.

In response, the authors of this review propose that the adaptive definition of resilience is an important lens through which any understanding of AFSC resilience must be built. The adaptive definition prioritises the role of cross-scale system component interactions to the point that external volatility is presumed to be a permanent feature and as such, rather than being a one-off fix, resilience must be seen as a cyclical process of “conservation”, “release”, “reorganization” and “exploitation”. In particular, similarities between the drive towards concentration of assets and connectivity in today’s global AFSCs and the “conservation” phase suggest vulnerability to major disruptions.

In addressing Q1.2, this paper identified the importance of capturing multidisciplinary “Supporting” elements of resilience, which reflect the role of social and environmental components of AFSCs rather than the traditional economic buyer–seller relations described in many supply chain works. These resilience elements are vital in addressing the unique AFSC risk sources identified previously. This review has also identified the importance of phase-based strategies of identifying which “Core” and “Supporting” resilience elements should be used and when. Therefore, the framework of AFSC resilience proposed is a synergistic one, combining the ecological science understanding of adaptive systems and “panarchy”, with resilience elements and strategies originating from SCM. A descriptive example of this framework can be found in Figure 9.

The framework proposes that parallels can be drawn between the four stages of the adaptive cycle (conservative, release, reorganisation and exploitation) and the four phases of a disruption, respectively (readiness, response, recovery and adaptation). It is proposed that there is similarity between the readiness phase of a disruption and the conservation stage of the adaptive cycle due to both considering the relative preparedness of a system before a disruption. There are also similarities between the response phase of a disruption and the release stage of the adaptive cycle, as both focus on the effects of a disruption. Similarly, there are overlaps between the recovery phase of a disruption and the reorganisation stage of the adaptive cycle as both concern regaining functionality. Finally, overlaps also exist between the adaptive phase of a disruption and the exploitation stage of the adaptive cycle, as both involve growth potential as a result of adaptation to previous disruptions.

These relations are exemplified in Figure 9 from the perspectives of an organisation, in this case a food processor, and the overarching food supply chain. Each faces a unique example risk from those categorised in Figure 8; the food processor a novel food product launched by a competitor unexpectedly and the supply chain, a serious regional natural disaster.

By dividing the disruption into phases, the food processor and the broader supply chain are able to assign bespoke mitigating resilience “elements” from those categorised in Figure 6. To better reflect characteristics of AFSCs, such as their importance to end consumer food security and diverse range of stakeholders, the actions for each resilience “element” are divided into social, environmental and economic indicators. This distinguishes them from previous works which have applied resilience elements for the purpose of organisational competitiveness. Not only do these three indicators represent the broad dependencies of food systems but they are also commonly used as the three pillars of sustainability; thus, the framework underpins the synergistic relationship between resilience and sustainability identified by others. Using the example of the supply chain, actions at the response phase where collaboration was identified as a suitable “Core” element include the need to work together as a supply chain to ensure food is available, safe and accessible to consumers. At an environmental level, the caveat is added that efforts to get food to consumers, perhaps by using alternative logistics, do not come at the cost of excessive pollution.

Economically, it is vital that organisations do their best not to exploit competitive advantage and drive food price inflation for end consumers. This, of course, is highly idealised, and the reality is that actions by individual organisations, in this example the food processor, to an earlier threat such as product competition, may actually preclude them from working collaboratively at a supply chain level. This represents a major advantage of using the adaptive model because it can explore the cross-scale interactions that can take place over great geographical and temporal distances.

A further key advantage of using the adaptive cycle as a basis for an AFSC framework is that it is cyclical in nature. In other words, there are no optimised “states of equilibrium” for an organisation to work towards, and this makes it inherently better suited to describing volatile operating environments, where disruptions are continuous, such as food systems, as illustrated in Figure 10. As such, the emphasis is on ingraining resilience across all activities, rather than as a one-off tool to address individual disruption risks, and in doing so, resilience becomes cumulative. In this way, a resilient food system is more of a safe-fail system rather than a fail-safe system (Anderies et al., 2013).