Linking Sustainability with Geographical Proximity in Food Supply Chains . An Indicator Selection Framework

Despite policymakers’ promotion of food relocalization strategies for burden mitigation, the assumption that local food chains are more sustainable than the global ones might not hold. This literature review tries to highlight a possible framework for exploratory analyses that aim at associating sustainability with the geographical proximity of food supply chains. The purpose of the article is identifying a set of communicative and information-dense indicators for use by evaluators. Bread is the selected test food, given its importance in human nutrition and the relevance of some of its life cycle phases for land use (cereal farming) and trade (cereal commercialization). Article searching (including keyword selection, explicit inclusion/exclusion criteria, and computer-assisted screening using the NVivo® software) was carried out over the Scopus, Web of Science, and Google Scholar databases, and returned 29 documents (refereed and non-refereed publications). The retrieved literature shows varied research focus, methods, and depth of analyses. The review highlighted 39 environmental, 36 economic, and 27 social indicators, along the food chain. Indicators’ reporting chains are heterogeneous; even the comparison of standard procedures, e.g., Life Cycle Assessment, is not straightforward. Holistic approaches are missing.


Introduction
Food supply chains are the sequences of the activities (steps) associated with food production, processing, distribution, consumption, and end-of-life disposal [1,2] (in this article, the terms "food supply chain" and "food chain" are used interchangeably).The geographical proximity of chain steps is a criterion for classifying food supply chains based on the distance among the firms that carry out those processes [3]; food supply chains can be classified as global, national, or local, when the involved steps are respectively located internationally, nationally, or subnationally [1,2].Since the early 2000s, the local-global dichotomy has informed the scientific debate around the sustainability of local and global food chains [4][5][6].Food (re)localization was born as a policy strategy to improve food chain sustainability via the mitigation of some of the drawbacks of global food chains [7].The governments of affluent countries (e.g., the United States, European Union) supported food relocalization as a sustainability-oriented strategy [3,7,8]; however, the assumption of the superior sustainability of local food chains may not hold [9].The local vs global dispute is still open [10,11].To the best of our knowledge, comprehensive evaluation frameworks are missing for use in comparative sustainability assessments of local and global food chains [10].To contribute to that debate, this article anchors to a worldwide staple food, i.e., bread, for four main reasons: (1) there is spatial variability in consumption and production patterns; (2) wheat (the main raw material) is a commodity subject to market price volatility; (3) unpredictable international events linked bread price with political change [12][13][14]; (4) information about wheat cropping location can be lost throughout the supply chain [15].Via a literature review, this article aims at providing evaluators with a set of communicative and information-dense sustainability indicators, that allow comparing food chains that differ for their geographical proximity.After this introduction, the article starts by reviewing the research approaches to sustainability assessment in agri-food chains (Section 2).Then, the research design is presented in Section 3, including keyword selection, explicit inclusion/exclusion criteria, and computer-assisted screening using the NVivo ® software (QRS International, Melbourne, Australia).The results of the literature review are presented and discussed, respectively, in Sections 4 and 5.

Literature Review
The evaluation of sustainability and related policies in the agri-food sector may involve monetary or non-monetary analyses [16].
In monetary analyses, the costs and benefits of the system under study are given monetary values, and their performance is evaluated using a benchmarking procedure, with the benchmark being provided by some wealth indexes of the measure of the private/public stream of social costs and benefits of alternative systems [17].Major monetary evaluations are generally based on cost-benefit and cost-effectiveness analyses.The former involves the comparative assessment of outcome variables, such as, e.g., social welfare, before and after policy implementation.The latter evaluates whether public money was appropriately invested by measuring a series of indexes, such as, e.g., social wellbeing [17].The intrinsic value of agricultural produce increases along the supply chain; then, monetary assessments may consider the extent to which the income of food chain players is subject to change, as well as how the distribution of value among those players is likely to change [18].Another strand of literature addresses consumers' willingness to pay for experience and credence attributes of local and global food.According to those authors, consumers' willingness to pay may be higher for local than for non-local food [19,20].This can be due to the general increase in social worries about food origin and the higher popularity of local food chains, which boosted origin labeling [21].Authors with higher interest in the environmental impact of food supply chains are more concerned with the total economic value of natural resources involved in the chain itself, and estimate the non-use value of those resources via the stated or revealed preferences of consumers [22].The benefit transfer is an additional research framework, relying on place and time-specific information to infer the economic value of goods and services in different places and times [23] (see also a special issue of Ecological Economics [24]).Disentangling the economic, environmental, and social impacts harden monetary evaluations of food chain sustainability; extending the scope of the assessment to consider the geographical proximity of food chain steps is even more complex.
Non-monetary assessments rely on sustainability indicators, i.e., measurable approximations of phenomena that occur at the economic, environmental, and social level, such as, e.g., social utility [25] or stakeholders' utility [26].Product-based indicators (e.g., calculated using the Life Cycle Assessment (LCA) tool) supplement sustainability assessments with the impact associated with the geographical extension of the food supply chain which, in turn, is approximated by the "farm-to-fork" supply chain [27].The set of indicators tend to vary across methodologies [28][29][30].Within this framework, multicriteria analysis is a widespread methodology that deals with the economic, environmental, and social impacts that are perceived, as such, by stakeholders in the food chain [31]; the collected stakeholder preferences are synthesized through modeling algorithms [25].When it comes to delivering general judgments around sustainability, cost-benefit analysis is superior to multicriteria analysis, considering population rather than stakeholder preferences [32].According to some authors, cost-benefit analysis allows a more impartial consideration of social preferences than multicriteria analysis [22].Both cost-benefit and multicriteria analyses struggle with the quantification of intangible benefits, such as, e.g., animal welfare [33], as well as with the disclosure of stakeholders' willingness to pay, and the distribution of costs and benefits along the chain.Indeed, multiple processing steps and reduced collaboration among stakeholders harden all assessments based on stakeholder preferences [34].Demand or supply side approaches to cut down on the environmental burden of agricultural production of food include the sustainable intensification of agriculture, the adoption of circular economy practices, e.g., for treating biowaste, and the promotion consumption-driven production shifts, towards the least-emitting foods or production methods/locations [35].The most popular methodology for assessing the sustainability of ecological innovation in agricultural practices or technologies is LCA (ISO 14040:2006, ISO 14044:2006).The method provides evaluators with data about material flows and many environmental impacts at each stage of the food chain, while targeting interventions towards the most impacting stages, or hotspots [36].Life cycle analyses are popular approaches to sustainability assessments, when evaluators need information about material flows and environmental impacts at each stage of the food supply chain [36].LCA results consider many impact categories (midpoint or endpoint) covering the depletion of natural resources and the damage to the environment and human health.LCA-based evaluations of single issues (e.g., global warming potential) are known as footprints; especially, carbon and water footprints are popular formalized methods (ISO 14064 and ISO 14046, respectively).LCA can be combined with economic (Life Cycle Costing) and social (Social LCA) assessments to cover the three dimensions of sustainability [37], provided that they share the same system boundaries [38].Ideally, the three methods should be based on the same data inventory [39].This is generally not an issue for the combination of LCA and Life Cycle Costing, the latter accounting for all the costs associated with all processes throughout the products' life cycle [38].However, Social LCA relies more on activity variables, such as, e.g., working hours [39].Social impacts are the consequences of behaviors, socioeconomic decisions, and human, social, and cultural capitals, occurring at the process level (production, processing, distribution, consumption, disposal), which generate positive or negative pressures on social wellbeing [39].The assessment is carried out at the stakeholder level (workers, local community, society, consumers, supply chain actors) and encompasses the following impact categories: human rights, working conditions, health and safety, cultural heritage, governance, and socioeconomic repercussions [39].The combination of the three-dimensional life cycle techniques in the Life Cycle Sustainability Assessment raised the interest of policymakers [40][41][42].Policy applications of data generated via LCA involve, e.g., extended producer responsibility or certification and labeling schemes to reduce information asymmetry business-to-consumer [43].
Identifying cause-effect relationships between the geographical proximity of food chain steps and their overall sustainability is a hard and risky task, with the comparative sustainability of global, national, or local food chains being the subject of a wide debate.For example, food miles may not greatly contribute to the overall economic and environmental impact of marketed food, compared to food production or processing methods [44,45].Moreover, the definition of the boundaries of the system under study is crucial.For example, the geographical proximity of food consumption depends of on product distribution channels (e.g., farmers' markets, direct online sales, retailers) and on consumer preferred channels [46].

Materials and Methods
This article is part of a broad case study research, included in an EU-funded research project, with the overarching objective of improving the scientific knowledge about the impact of food supply chains differing by their geographical proximity, to promote evidence-based public policies and private strategies for increasing food chain sustainability, while informing consumer purchases [47].Real-world food chains with different geographical proximity were selected across European countries, to cover the cereal, horticulture, fruit, dairy, and meat sectors [48].The present paragraph is based on the Italian case study.A global and two local bread supply chains were analyzed.The global chain refers to soft bread, available at retailers' stores; plastic seal packaging allows for around 30 days shelf life.One of the two local chains is a vertically integrated firm (farm to product sale); the other involves the production and distribution of bread under a geographical indication scheme; the final product is available at retailers' stores.The two local chains refer to freshly baked bread, sold in unsealed paper bags, with a few days shelf life.Other differences among the three chains involved wheat origin and cultivar, farming method, public support by the EU Common Agricultural Policy schemes, scale, governance, degree of integration, number of intermediaries, target consumer segment, and marketing strategy.A detailed description of the three chains is available from the case study report [49].Readers may refer to [50,51] for further insights.Besides the literature review, the case study research included interviews with key informants and an interdisciplinary collaboration with food scientists.The different research components of the case study research were carried out by different members of the research team, with the purpose of allowing the cross-fertilization among research components.Details about the EU project and full access to reports are available from [47,48].This section of the article provides essential information about the interviews and extensively describe the design and implementation of the literature review.

Interviews with Key Informants
The purpose of interviews was identifying case study-specific sustainability issues and proposed solutions by food chain stakeholders and experts.Based on that, we used a snowballing procedure to select key informants to cover the research, industry, and retail sectors.Twelve key informants agreed to participate in the project, including university professors, business operators, producers' associations, independent consultants, and company managers (Table 1).
Table 1.List of key informants.Source: authors' own elaboration based on [49].shared with all European case studies that drew on a wide theoretical framework, relying on media analysis and a Delphi survey [52].Each research team adapted the questionnaire based on the case studies, to address a set of food-specific critical issues, identified by the theoretical framework [52].A sample questionnaire is available from the case study report [49].
Besides basic information about the respondent, different questionnaires were administered to supply chain stakeholders and experts.Supply chain stakeholders were asked to map and discuss supply chain steps, including their geographical proximity, their importance to allow supply chain viability, and their criticalities.A set of questions aimed at describing firm scale, production methods, material and resource use, and end-of-life disposal.Other questions addressed environmental, economic, and sustainability issues associated with firm management and operations, and possible approaches to problem solving.Compared to supply chain stakeholders, the structured questionnaire administered to experts was more flexible, to bridge knowledge gaps.Experts were asked to give theoretical and practical explanations about the environmental, economic, and social issues (or benefits) associated with the practices and technologies adopted by the three supply chains.Addressed topics were as follows: seed selection, farming, milling and processing methods and machineries; marketing strategies and distribution channels, including the adoption of labeling schemes (environmental labeling, geographical indication); packaging material and technology; waste management options.
Semi-structured interviews were carried out, face-to-face, with the twelve key informants.Meetings were arranged with all respondents, and structured questionnaire templates were filled in on the spot by one of the researchers in charge of the interview.Additionally, all meetings were recorded to check for the accuracy of template filling during questionnaire analysis.
The main bread chain phases were wheat farming, kernel milling, and baking.Critical issues specific to the case study were biodiversity (via seed mix selection), technological innovation (to reduce impacts on the environment and human health, and improve product shelf life and quality), nutrition (e.g., through recipe and leavening process), information and communication (nutritional and health claims, environmental labeling, geographical indication), resource use and pollution (organic and low input farming, impact monitoring via LCA), and value creation and distribution (e.g., via contract farming, contracts with retailers, direct sale).

Design and Implementation of the Literature Review
Research literature reviews are a widespread tool for managing knowledge diversity, when trying to answer specific research questions [53], allowing for highlighting of knowledge gaps and to create directions for future research [54].Designing a research literature review implies making explicit, its objectives, concepts, and methods, including paper selection [38], to allow study replication and evaluation.The research question should drive papers' selection strategy, which, in turn, needs be validated.Implementing research steps at the research team level, or comparing the research design with refereed literature, are examples of validation methods [26].The search strategy is a stepwise procedure that involves identifying suitable bibliographic sources and exploring them using keywords associated with the research question, combining keywords into a string using Boolean operators, as well as creating and applying practical screening criteria for including or excluding papers, also with the aid of qualitative data analysis software [37].Many scientific articles, e.g., [53,54], and academic books, e.g., [55][56][57], guide researchers throughout the review process.Additional support is available from web sources, most notably, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses website [58].Conditional on research aims and scope, the existing literature review frameworks allow authors some flexibility, as long as they document the stepwise process [59].The reader may refer to the existing literature for methodological details.This paragraph reports on the design and implementation of this literature review, a process involving six consecutive steps.

Selection of Bibliographic Sources
For refereed literature, Elsevier Scopus ® and Web of Science TM are the source databases.For non-refereed literature, Google Scholar TM is the source database.

Keywords and Search
The string "(sustainability OR sustainable) AND (bread OR grain OR wheat OR food) AND (local OR global) AND chain" was run over the fields title, abstract, and keywords of academic databases; different combinations of string terms were run over Google Scholar.A total of 2229 records were collected via EndNote ® citation manager.

Creation and Application of Inclusion/Exclusion Criteria
No cut-off criteria were applied to research methods, quality, and time span.Only the academic database search was limited to journal articles and reviews written in English.Two exclusion and one inclusion criteria were consecutively applied.The first exclusion criterion involved journal selection: papers were excluded if journal aims and scope did not relate to agricultural economics, rural studies, or agricultural multidisciplinary.As an applied science, agricultural economics research covers heterogeneous fields, such as, e.g., farm management, production economics, environmental and resource economics, or food and consumer economics.The research area of rural studies is broad as well, being mainly concerned with the sociological aspects of access to environmental resources and vulnerability, including food and nutrition security.This category is linked with agricultural economics, especially when farm management and issues associated with food production and availability are the topics under study [60].Agricultural multidisciplinary is not an academic discipline, but a category of Web of Science TM , including 56 journals dealing with heterogenous topics, e.g., engineering, economics, genetics, food science, or consumer studies, among others.The exclusion criterion based on journal aims and scope was carried out manually, based on journal websites and personal experience.Journals in the field of medical sciences were not excluded a priori, being not many (around 40), as the related disciplines are often associated with food and consumers economics and food sociology, especially when dealing with life cycle impacts of food production and consumption [60].After the application of the first exclusion criterion, 1827 records were discarded.The second exclusion criterion aimed at the exclusion of double counts in EndNote ® , and allowed 112 records to be discarded.The inclusion criterion relied on the manual scanning of abstracts; papers were included if they addressed the sustainability of food supply chains.This criterion allowed the exclusion of 66 records.The availability of full texts through authors' university library system further limited the number of papers to 135.The full texts of those articles were retrieved and manually screened.Having highlighted the presence of very technical articles, especially agronomy or food science studies presenting field or laboratory experiments, we decided to implement a simple computer-assisted text analysis over the 135 full texts, using the Nvivo ® qualitative data analysis software.This step was carried out by three out of the four authors.The rationale behind the computer-assisted screening was highlighting the possibility of extracting sustainability indicators from the retrieved documents, by isolating constructs that incorporated relevant terms associated with the research question.A text search query was formulated to create paper-paper connections based on word proximity; specifically, i.e., the term "indicator" had to appeared within three words from the terms "environment/environmental" OR "economy/economic" OR "society/social" OR "sustainability/sustainable".Connected papers [29] were selected as the units of analysis and retrieved for review.Nvivo ® screening markedly reduced the number of documents under review; some of those documents might be relevant for the study.This is a drawback of the computer-assisted screening; manual screening may have led to different results.Using a different qualitative data analysis software may have led to different results, as well, each software having a different structure.We opted for the automated, rather than for the manual, text analysis to improve screening objectivity, thereby reducing researcher bias associated with this step.

Validation of the Review Process
The implemented process was compared with refereed literature, such as, e.g., [61,62].Early versions of the paper were shared at international conferences [28] and on research networking websites.Authors carried out each process at least in pairs.

Material Description and Evaluation
This is the focus of the results section.The retrieved papers are described based on key attributes and patterns [63].Attributes are paper aim, object, and country of analysis, research method, and addressed dimension(s) of sustainability.Patterns refer to the concept of geographical proximity and to system boundaries.Patterns are identified, deductively, based on the existing literature [63].The definitions for geographical proximity patterns were inductively created after paper review.Instead, system boundary definitions were based on the existing literature (deductive).The combination of deductive with inductive approaches ensures the external (the existing literature) and internal (the literature under study) validity of the identified categories [64].Table 2 shows the patterns and their relative definitions.The patterns for geographical proximity are four, i.e., global, national, local, and farm.The first three patterns were identified and defined based of the location of the activities involved in supply chain processes [1,2].The fourth pattern, i.e., farm, was inductively created for flagging those articles that were focusing on farm-level performance.The patterns for system boundary are three, i.e., cradle-to-grave, cradle-to-retail, and gate-to-gate, and based on the common LCA terminology, (see e.g., [65]).
Indicators used in the retrieved literature were extracted and classified under environment, economy, and society, based on their application in the original paper (type of research and researcher perspective).Based on the critical issues presented in the previous section, the indicators were clustered under biodiversity, technological innovation, nutrition, resource use and pollution, information, and communication.Though aiming at objectivity, the authors' judgement may have affected indicator allocation among sustainability dimensions and critical issues.
Sustainability can be represented and studied using quantitative, conceptual, standardizing, physical, or pictorial visualization approaches, differing for data requirements, level of integration among dimensions, type of outputs, and target audience [66].Based on the preliminary literature overview, the pictorial visualization approach was selected to classify the indicators proposed by the units of analysis based on the dimension(s) addressed by the proposed indicator(s).Different pictorial visualization tools are available, each of them having known advantages and limitations [67].The adopted pictorial visualization tool is the Venn diagram model.This step of the review is critical, given the risk of bias by researchers' view and approach to model selection; different research teams may opt for different models, thereby generating different research outputs.We adopted the integrated perspective of sustainability, and selected that specific model to classify the indicators extracted by each record, given the marked heterogeneity of the units of analysis, especially with respect to research fields, methodologies, and investigated systems.In addition, model simplicity and popularity offer immediate understanding to readers, irrespective of their background [68].The model is based on a three-overlapping circle symbolism, each circle representing a dimension of sustainability, namely environment (protection of natural resources and ecosystem health), economy (sector, business and people livelihood viability) and society (equity, health); sustainability may be achieved when all three circles overlap [69].Under this framework, tridimensional indicators were identified and selected.Venn diagrams were built for visualizing the extracted indicators, and the relative importance of each dimension and dimension combination.Despite the computer-assisted screening, not all papers under review explicitly flag the proposed indicators as owing to the environmental, economic, and/or social dimension(s) of sustainability.This is due to the specific text syntax.When the dimension was not explicit, we contextualized indicator allocation to one or more of the three dimensions.Our attempt was to be as objective and transparent as possible; however, a certain degree of subjectivity may have affected indicator allocation.

Interpretation
Results' interpretation is based on the outputs of the application of the Venn diagram model to the extracted indicators.This is the subject of the discussion paragraph, which is structured towards the identified set of indicators.

Material Description
The units of analysis cover 16 years (1998-2014) and are distributed among heterogeneous publication sources, most of them being represented by one paper (Appendix A, Figure A1).The most represented journals are the International Journal of Agricultural Sustainability, the International Journal of Life Cycle Assessment.Three papers are book chapters, and two are extracted from conference proceedings.Research aims vary considerably throughout the analyzed literature (Table 3).Three articles address food chain localization strategies, six deliver impact assessments of food chain steps, six focus on sustainability strategies at the business or organizational level, and four concentrate on ethical traceability.Additionally, two papers address life cycle impacts of diets, two focus on ecological aspects of different food chains, one concentrates on stakeholder perceptions of sustainability, and three more articles concentrate on methodological aspects of impact assessments.Concerning the addressed dimensions of sustainability, fifteen papers concentrate on environmental issues and seven add the economic and/or social dimension.Social issues are the focus of six papers.The economic dimension tends to be associated with other dimensions.Quantitative and qualitative researches have similar purposes, but differ for data gathering, handling, and reporting.Mainly, quantitative research designs (17 papers) rely on LCA-based techniques (6 papers), including carbon footprint and life cycle inventory, or on alternative life cycle impact calculation methods (7 papers).Qualitative studies propose literature reviews (6 papers) or analysis of interview-based information (5 papers).Nine documents focus on bread, twelve on primary or processed foodstuffs different from bread, and eight assess sustainability at the retailer, diet, meal, food chain, or farm level.Geographic settings are diverse, with the United Kingdom (7 papers) and Scandinavian states (5 papers) being the most investigated ones.Moving to geographical proximity patterns, most articles study national supply chains (20 papers), followed by local supply chains (8 papers); six papers analyze supply chains with different geographical proximity and two deliver farm-level evaluations.System boundaries are fairly evenly distributed among the three patterns, i.e., cradle-to-grave, cradle-to-retail, and gate-to-gate, with the latter accounting for a slightly larger number of articles.
Paper analysis returned 78 indicators addressing single and multiple dimensions of sustainability: environment ( A1).As far as possible, indicators with a similar purpose or definition were flagged under the same label, though proposed by different authors and used under different research frameworks.Five out of six critical issues are represented (Figure 1).
(5 papers).Nine documents focus on bread, twelve on primary or processed foodstuffs different from bread, and eight assess sustainability at the retailer, diet, meal, food chain, or farm level.Geographic settings are diverse, with the United Kingdom (7 papers) and Scandinavian states (5 papers) being the most investigated ones.Moving to geographical proximity patterns, most articles study national supply chains (20 papers), followed by local supply chains (8 papers); six papers analyze supply chains with different geographical proximity and two deliver farm-level evaluations.System boundaries are fairly evenly distributed among the three patterns, i.e., cradle-to-grave, cradle-toretail, and gate-to-gate, with the latter accounting for a slightly larger number of articles.
Paper analysis returned 78 indicators addressing single and multiple dimensions of sustainability: environment ( A1).As far as possible, indicators with a similar purpose or definition were flagged under the same label, though proposed by different authors and used under different research frameworks.Five out of six critical issues are represented (Figure 1).None of the identified indicators explicitly targets the impacts of technological innovation.Compared to resource use and pollution, and value creation and distribution, the number of indicators aimed at delivering measures about biodiversity, nutrition and information, and communication, is reduced.While some articles relate supply chain sustainability with geographical proximity, no single indicator explicitly provides a measure of that relationship.The identified indicators deliver (and are based on) qualitative and quantitative information, with heterogeneous measuring systems.This may prevent objective judgements.Environmental indicators include measures of impact on the physical or biologic environment; they are generally very specific and require detailed measures, especially when compared to social indicators.At times, information delivered by some environmental indicators tends to overlap (e.g., "land use" vs "total land requirements"), which may slow down the process of indicator selection for decision-making purposes.Mainly, economic indicators address firms' ability to deal with appropriate investments and to optimally allocate resources, while social indicators focus on human health environmental assessments generally follow standard data collection procedures, to allow for replicability.This may not be the case when gathering information form people via interviews or surveys, with increased risk of researcher bias.

Material Evaluation
This subsection draws on the application of the Venn diagram model to the retrieved literature.At the article level, the environmental dimension has the greatest importance, followed by the social (Figure 2).None of the identified indicators explicitly targets the impacts of technological innovation.Compared to resource use and pollution, and value creation and distribution, the number of indicators aimed at delivering measures about biodiversity, nutrition and information, and communication, is reduced.While some articles relate supply chain sustainability with geographical proximity, no single indicator explicitly provides a measure of that relationship.The identified indicators deliver (and are based on) qualitative and quantitative information, with heterogeneous measuring systems.This may prevent objective judgements.Environmental indicators include measures of impact on the physical or biologic environment; they are generally very specific and require detailed measures, especially when compared to social indicators.At times, information delivered by some environmental indicators tends to overlap (e.g., "land use" vs "total land requirements"), which may slow down the process of indicator selection for decision-making purposes.Mainly, economic indicators address firms' ability to deal with appropriate investments and to optimally allocate resources, while social indicators focus on human health environmental assessments generally follow standard data collection procedures, to allow for replicability.This may not be the case when gathering information form people via interviews or surveys, with increased risk of researcher bias.

Material Evaluation
This subsection draws on the application of the Venn diagram model to the retrieved literature.At the article level, the environmental dimension has the greatest importance, followed by the social (Figure 2).Stand-alone economic assessments are missing, being frequently combined with environmental analyses.This suggests a general trend towards the evaluation of the economic viability of the strategies for improving the environmental performance of food supply chains.For life cycle analyses, this trend could benefit from the similar data requirements of environmental and economic assessments.The integration of social aspects into multidimensional assessments is reduced.This may be due to the large use of qualitative data analyses by social scientists, especially when dealing with case study-specific data.This suggests the need for a greater research effort towards the application and the development three-dimensional assessment frameworks (such as, e.g., the Life Cycle Sustainability Assessment).
Moving to the identified indicators, the trade-off between the evaluation of environmental and economic issues is reduced (Figure 3).The reduced area of the social dimension suggests the need for more research towards the development of social impact indicators and of multidimensional indicators, which may facilitate the work of evaluators.Five indicators show up where all three circles overlap, i.e., breeding effectiveness (b1), product/process certifications (i2), water use intensity/savings (r32), waste generation/reduction (r31), and business uncertainty (v3) (Table 4).Stand-alone economic assessments are missing, being frequently combined with environmental analyses.This suggests a general trend towards the evaluation of the economic viability of the strategies for improving the environmental performance of food supply chains.For life cycle analyses, this trend could benefit from the similar data requirements of environmental and economic assessments.The integration of social aspects into multidimensional assessments is reduced.This may be due to the large use of qualitative data analyses by social scientists, especially when dealing with case study-specific data.This suggests the need for a greater research effort towards the application and the development three-dimensional assessment frameworks (such as, e.g., the Life Cycle Sustainability Assessment).
Moving to the identified indicators, the trade-off between the evaluation of environmental and economic issues is reduced (Figure 3).Stand-alone economic assessments are missing, being frequently combined with environmental analyses.This suggests a general trend towards the evaluation of the economic viability of the strategies for improving the environmental performance of food supply chains.For life cycle analyses, this trend could benefit from the similar data requirements of environmental and economic assessments.The integration of social aspects into multidimensional assessments is reduced.This may be due to the large use of qualitative data analyses by social scientists, especially when dealing with case study-specific data.This suggests the need for a greater research effort towards the application and the development three-dimensional assessment frameworks (such as, e.g., the Life Cycle Sustainability Assessment).
Moving to the identified indicators, the trade-off between the evaluation of environmental and economic issues is reduced (Figure 3).4).The five indicators were identified from eight papers and cover four critical issues, i.e., biodiversity, information and communication, resource use and pollution, and value creation and distribution (cf.Table A1).Three out of five indicators are used for the simultaneous assessment of multiple sustainability dimensions in the original paper.Instead, the two indicators addressing issues associated with resource use and pollution are sometimes used for purely environmental evaluations.In the original papers, indicator selection mainly relied on the existing literature, though sometimes authors supplemented the literature with stakeholder consultations.Just two out of eight papers measured the indicator using primary data sources.For quantitative applications, indicators are expressed by some physical property of object of analysis.For qualitative applications, the value of indicators is expressed in relative terms, by means of author-defined ranking scales or binary variables, such as, e.g., the presence/absence of a sustainability label.

Water Use Intensity/Savings
The LCA tool offers the only comprehensive indicator, i.e., the volume of water consumed over product life cycle per functional unit (L/kg bread ready for consumption at home).However, the cost of an LCA may be prohibitive for smaller food manufactures [43,85].Wheat farming, consumer, waste, and transport phases are not a water-intensive phase, contrary to wheat-to-bread processing [98].A measure of the rough annual water consumption (e.g., based on water bills) at the bakery level may provide a proxy for water use sustainability.Concerning chain's geographical extension, generally, local bakeries have a smaller scale compared to the global ones, which distribute bread internationally.Due to economies of scale, bigger facilities are expected to display more efficient water use that smaller ones.Water use efficiency indicators are expected to be lower for local bakeries.

Waste Generation/Reduction
Waste refers to unsold bread and packaging.Besides energy consumption, resource wastage on farm may encompass harvest losses due to adverse weather conditions or human error.Waste occurs at the mill phase as well.Generally, the baking phase shows the highest waste production in the wheat-to-bread chain [79].However, most waste is produced beyond the bakery gate, at the consumer and retail phases [99].The total weight of unsold bread, including packaging per unit area of retail stores and per household (e.g., using data form official statistics) may be useful for the sustainability assessment of bread supply chains.The shorter the shelf life, the greater the expected quantities of food and packaging waste and the higher the GHG emissions [79].Short shelf-life bread is likely to be produced, traded, and consumed locally.Local products tend to be characterized by cultural and community embeddedness.This may prevent the generation of food waste generation; e.g., the Italian gastronomy offers countless ways for cooking stale bread.The waste efficiency indicator may display poorer performance as chain's geographical proximity decreases.Energy and/or nutrients may be recovered from bread, being organic waste; packaging (paper or plastic) may undergo the recycling process.Plausible indicators may consider the ratio between the weight of bread diverted to either biomass-to-energy or composting plants, and the weight of bread sent to landfill per unit area of the retail store or per household.Again, the smaller the geographic extension of the chain, the better the expected chain's performance.

Breeding Effectiveness
Wheat breeds differ for their resistance to biological and environmental stresses, yield, and nutritional properties, especially gluten content and glycemic index.Healthier nutritional properties decrease with yield, which in turn, increases with mineral inputs [100].Large scale wheat cropping, for selling on the commodity market, generally rely on high yield and nutrient intensive seeds to ensure business stability.Productivity maximization needs (quasi-)optimal soil and climatic conditions, as well as wide land extension, which may limit the competitiveness of, e.g., hilly or mountain regions.Farmers producing for smaller scale markets in suboptimal farm conditions may benefit from locally adapted or heritage wheat varieties, given their hardiness.This may also help biodiversity conservation locally.Those hardy varieties are more suitable to low input farming than mainstream high-yield varieties, despite their higher variability in yield and technological quality of kernels [15].One of the strengths of food (re)localization strategies is giving "diverse" food a market [71].For example, a sustainability indicator may be the ratio between the income generated by cultivating (or baking with) non-mainstream and mainstream breeds, measured at the regional, national, and supernational levels.That indicator may capture the ability of those (re)localization strategies to allow the viability of "alternative" bread chains.

Product/Process Certifications
From a buyer perspective, product labeling (e.g., carbon footprint, water footprint, sustainable farming, fair trade) is the easiest way to know about product or process sustainability.From a food producer perspective, meeting certification requirements may constrain business performances, though, entrepreneurs acknowledge the marketing value added by sustainability labeling [71,95].Certifications generally involve traceability systems, which need collaboration among chain partners and fees, which tend to be a barrier for local food producers.The availability of sustainability labeling is a proxy for chains' sustainability, which may show more positive values for more geographically extended chains.

Business Uncertainty
Pursuing sustainability goals at the firm level-e.g., while deciding on business development or resource allocation-requires considering social, economic, and environmental aspects.Economic aspects are priorities [95].Indicators of the economic performance of a business are many, e.g., focusing on profitability, competitiveness, or risk.Just a few of those indicators can capture the differences associated with the geographical extension of bread supply chains, which, in turn, may depend on stakeholders' priorities [101].For example, boosting profitability while reducing uncertainty is the priority, when it comes to deciding to participate in a supply chain with a given geographical extension [95].On farms, decisions about the crop mix and farming practices depend on the degree of business uncertainty.Yield stability of low-input or hardier breeds and the availability of financial support for implementing low-input farming practices may approximate for reduced business uncertainty for farmers.Concerning downstream phases, marketing strategies dedicated to the promotion of flour and bread from low-input (or hardier) grains or originated from low-input farming may reduce business uncertainty.Given the general higher market shares, business uncertainty may be expected to be lower for bread companies with wider geographic extension.

Conclusions
Using the example of bread, this paper links sustainability with the geographical extension of the food chain, by means of a literature review.Results show that sustainability assessments rarely and simultaneously cover environmental, economic, and social dimensions, with the latter being the most disadvantaged dimension.A significant share of papers proposes LCAs.Those studies may not consider the opportunity cost of resource use or identify the connections and interdependences among chain phases and, thus, miss proposing solutions for improving resource use efficiency.
The task of synthetizing social preferences adds complexity to sustainability assessments of the food chain, because preferences depend on cultural factors, media, and marketing strategies, among others.Moreover, available information about the food chain is often asymmetric.Stakeholders' opportunistic and/or irrational behavior make the task even harder.Results suggest that reliable sustainability assessments of the food chain should highlight areas of consensus or disagreement, to allow the use of assessment's outcomes for reducing uncertainty in decision-making.The authors' research field is likely to influence method selection, as well as the prospected study outcomes.At times, different research approaches are used to address the same issues within the same dimensions of sustainability (overlapping).Decision-makers in charge of delivering judgements about food chain sustainability need synthetic information about food chains' complexity, and about existing balancing or reinforcing loops, and trade-offs among stakeholder preferences.Knowing the opportunity costs of allocating investments and distributing resources across chain stakeholders (and phases) is also required.The lack of comprehensive evaluations of the economic, environmental, and social sustainability of the food chain hinders the delivery of judgements.Besides raw indicator measures, decision-makers should know social preferences and perceptions, because path and context dependency, and the influence of media, can alter the value of available information.
To overcome the methodological limitations of this study, further research may focus more on the integration of manual with computer-assisted text analysis, especially when it comes to paper selection based on textual constructs.Further research may improve the understanding of the sustainability of different technological advancements over food life cycle, including single process sustainability, for example, by means of Life Cycle Sustainability Assessment models.Social and, especially, consumer preferences for foodstuffs belonging to supply chains with different geographical extension are worth being integrated in future assessments, as well as the opportunity cost of allocating financial resources to different chain phases for improving the life cycle sustainability performance.Regionalized Life Cycle Sustainability Assessment may help integrate the geographical dimension in sustainability indicators.Supplementing life cycle with systems dynamics studies is a possible way for considering feedback loops, when evaluating the sustainability of alternative food chains.

Figure 2 .
Figure 2. Venn diagram of sustainability dimensions per paper.Circle intersections show the relative distribution of single and multidimensional assessments within the retrieved literature; circle areas are proportional to indicator counts per single and multiple dimensions.Source: authors' own elaboration.

Figure 3 .
Figure 3. Venn diagram of indicators.Circle areas and intersections are proportional to indicator counts per single and multiple dimensions.b: biodiversity; n: nutrition; information and communication; r: resource use and pollution; v: value added creation and distribution.See TableA1for the complete coding.Source: authors' own elaboration.

Figure 2 .
Figure 2. Venn diagram of sustainability dimensions per paper.Circle intersections show the relative distribution of single and multidimensional assessments within the retrieved literature; circle areas are proportional to indicator counts per single and multiple dimensions.Source: authors' own elaboration.

Figure 2 .
Figure 2. Venn diagram of sustainability dimensions per paper.Circle intersections show the relative distribution of single and multidimensional assessments within the retrieved literature; circle areas are proportional to indicator counts per single and multiple dimensions.Source: authors' own elaboration.

Figure 3 .
Figure 3. Venn diagram of indicators.Circle areas and intersections are proportional to indicator counts per single and multiple dimensions.b: biodiversity; n: nutrition; information and communication; r: resource use and pollution; v: value added creation and distribution.See Table A1 for the complete coding.Source: authors' own elaboration.The reduced area of the social dimension suggests the need for more research towards the development of social impact indicators and of multidimensional indicators, which may facilitate the work of evaluators.Five indicators show up where all three circles overlap, i.e., breeding effectiveness (b1), product/process certifications (i2), water use intensity/savings (r32), waste generation/reduction (r31), and business uncertainty (v3) (Table4).

Figure 3 .
Figure 3. Venn diagram of indicators.Circle areas and intersections are proportional to indicator counts per single and multiple dimensions.b: biodiversity; n: nutrition; information and communication; r: resource use and pollution; v: value added creation and distribution.See Table A1 for the complete coding.Source: authors' own elaboration.

Table 2 .
Identified patterns and relative definitions for geographical proximity and system boundary.Source: authors' own elaboration.

Table 3 .
Overview of the retrieved papers.Source: Authors' own elaboration.
a Chapters of the same book.b Papers that share co-authors.c Papers that present different parts of the same research and do not describe the adopted indicators with detail.

Table 4 .
Overview of the five sustainability indicators highlighted via the application of the Venn diagram model.Source: authors' own elaboration.