The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains

Journal of Cleaner Production xxx (2015) 1e14

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains Riccardo Accorsi a, *, Susan Cholette b, Riccardo Manzini a, Chiara Pini a, Stefano Penazzi a a b

Department of Industrial Engineering, University of Bologna Alma Mater Studiorum, Bologna, Italy College of Business, San Francisco State University, California, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 27 June 2014 Received in revised form 15 June 2015 Accepted 17 June 2015 Available online xxx

Global food demand will double by 2050 and strain agro-food supply chains. The increasing relevance of non-agrarian activities within the food supply chain mandates a systemic perspective for addressing sustainability. We consider the food supply chain as an ecosystem and define more inclusive boundaries. We present a design framework that supports strategic decision-making on agriculture and food distribution issues while addressing climate stability. We describe the methodology used to construct the framework, which entails a multi-disciplinary approach. An original land-network problem merges localized and large-scaled decisions as land-use allocation and location-allocation problems in an agrofood network. A linear programming model optimizes infrastructure, agriculture, and logistics costs and also balances carbon emissions within the agro-food ecosystem. A regional potato supply chain illustrates the effectiveness of the proposed model. Findings show the interdependency between infrastructure, production, distribution, and environmental resources. Results highlight the consequences of unbalanced planning focused solely on cost efficiency. In conclusion we identify enabling conditions, drivers and metrics for the design of cost effective and carbon balanced agro-food ecosystems. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Food supply chain Sustainability Land-use Carbon footprint Linear programming Climate change

1. Introduction Humanity is demanding more and richer food. Population growth and rising per capita consumption of animal products will double global food demand by 2050 (Koning and Van Ittersum, 2009). The rapid development of new economy countries propels demand not only in absolute quantities but also with respect to year round availability. Global food supply chains are expanding to match seasonal food production to demand (Ahumada and Villalobos, 2009; World Bank, 2011). These trends support considering the food supply chain as a whole: not only cultivating and processing but also packaging, storing, and distributing foods as well as handling by-products and waste. These processes result in more complex supply networks, increasing distances between stages and decreasing consumer awareness of what is required to get from farm-to-fork (Manzini and Accorsi, 2013). Even the collection of by-products and waste has significant ramifications (Ruggeri et al., 2009). Food supply chain actors typically focus on economic sustainability: cost

* Corresponding author. Tel.: þ39 051 2090468. E-mail address: [email protected] (R. Accorsi).

reduction. Environmental sustainability receives less attention, and environmental externalities are neither accounted for nor assigned to any actors. This negligence results in the agro-food sector contributing greatly to climate change (Desjardins et al., 2007). Growing demand is forcing the conversation on reconciling economic growth with environmental sustainability. Tilman et al. (2002) consider the tradeoffs between intensive global food production based on economies of scale and sustainable localized, smaller farming models. Others document the diverse environmental impacts of intensive farming and production upon natural resources and ecosystems (Tukker et al., 2008), as well as the energy requirements and resultant greenhouse gas (GHG) emissions
i Canals et al., 2007; from food storage and transportation (Mila Cholette and Venkat, 2009). Some studies challenge the perceived “common knowledge” about food systems (McCown, 2002a,b; Hinrichs, 2003; Nilsson, 2004; Partidario et al., 2007; Weber and Mathews, 2008; Massoud et al., 2010) and identify challenges and research agendas (Soussana, 2014) for mitigating the environmental impacts of the food sector. Sustainable agro-food systems must address food security, promote environmentally sustainable development, and balance food production with ecosystem services and biodiversity (FACCE-JPI, 2011). These ambitious challenges require holistic tools for analyzing the entire food supply

http://dx.doi.org/10.1016/j.jclepro.2015.06.082 0959-6526/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

2

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

chain, coupling cost control with natural resource management and emissions reduction (Akkerman and Van Donk, 2010; FAO, 2012; Leduc et al., 2013). This paper builds upon extant agro-industrial ecology research on creating and maintaining equilibria within observed ecosystems (Varga and Kuehr, 2007). UNEP (2009) provides the theoretical background for the paper as follows. While agro-food ecosystems provide nourishment, they are also fundamentally connected to climate regulation by acting as either carbon sources or sinks. Canadell et al. (2007) recommend that an ecosystem's absorption of GHGs balance the ecosystem's natural emissions and any additional human induced emissions. This paper studies the role of the agro-food ecosystem as a whole in reducing the carbon cycle imbalance. We redefine the system boundaries to be more inclusive. Fig. 1 shows how cultivation, packaging, storage and distribution become parts of the same ecosystem along with carbon sinks to mitigate resultant emissions. This original framework merges both local and global scaled strategic decisions. It breaks down the geographic boundaries of food production and distribution and accounts for both costs and environmental impacts. Local scaled decisions include but are not limited to the following: farm design, crop allocation, adoption of equipment and infrastructure, landscape planning, and groundwater management. Such decisions greatly affect land-use and are responsible for much of human-induced emissions (Ovando and Caparros, 2009; Ponsioen and Blonk, 2012). Emissions goals at the local scale are reflected in the tradeoff between food crops and other land uses, such as forestry or biofuel production. At the agriculture level, modelling the spatially constrained system is suited for operation research techniques such as linear programming (LP). Singh (2012) provides a comprehensive survey of relevant operation research applications to agro-food system problems solved at a local scale. Singh (2012) identifies groundwater management, irrigation scheduling, crop yield enhancement, land planning, and resource and waste management as active areas of interest. Additional studies show land-use allocation (LUA) problems may consider a wide set of goals and objectives, including returns and costs (Singh and Panda, 2012), crop yields (Zeng et al., 2010), and watermanagement regimes (Yang et al., 2009; Dai and Li, 2013). Global scaled decisions involve designing supply chain infrastructures, namely selecting growers/suppliers, creating multiechelon logistics networks (Apaiah and Hendrix, 2005), and distributing appropriately (Ahumada and Villalobos, 2011). Such logistics decisions can be modeled and solved via linear

programming. The classic location-allocation (LA) problem selects from a set of potential nodes and minimizes distribution costs while meeting demands (Azarmand and Neishabouri, 2009; Manzini et al., 2014). Global net emissions goals are reached through proper network design and mitigation of environmental externalities. By representing the agro-food supply chain as an inclusive ecosystem (see Fig. 1), the proposed framework integrates local and global scaled decisions to provide sufficient food and balance net emissions. Three local scaled decisions are considered. The first locates crops, maximizing crop yields on the basis of climate, soil and water availability conditions for a given area. The second places facilities for the following operations: consolidating raw materials, processing, packaging, and storing finished food products. The third allocates renewable energy sources, solar arrays or wind farms, to power the agro-food supply chain plants and services. The global scaled decisions concern network design and route selection for distributing food over the larger region. Since the supply chain produces emissions through cultivation, packaging, storage and transport processes, a more inclusive agro-food ecosystem should offset these emissions. To this purpose, the framework includes a land use for sequestering carbon. The economic and environmental trade-offs associated with ecosystembased carbon mitigation strategies are embedded in the landnetwork (LN) problem. The LN problem models the agro-food supply chain as a closed ecosystem comprised of integrated networks of lands allocated to multiple uses and services: crops, facilities, energy sources. The agro-food ecosystem must manage the available land and energy resources to satisfy food demand and reduce net emissions. From a modeling perspective, the LN problem integrates LUA and LA problems and balances food costs with climate stability goals. The proposed framework contributes to the literature by considering the agro-food supply chain as a whole ecosystem. This holistic approach integrates strategic long-term decisions on production, and distribution processes, exposing the underlying interdependencies between agriculture and logistics in sustainable agro-food ecosystems. Furthermore, the model incorporates environmental externalities by constraining the food supply chain activities in accordance with climate stability and ecosystem conservation goals. To this purpose, the framework involves alternative land uses, carbon plantings and renewable energy fields, not commonly included in supply chain analyses but necessary to guarantee that demand is fulfilled efficiently in presence of climate stability constraints.

Fig. 1. An inclusive definition of the agro-food ecosystem: agro-food system processes across the food supply chain and climate mitigation strategies.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

The remainder of the paper is organized as follows. Section 2 presents the framework of the LN problem and guides the decision-maker through the steps needed to apply the framework. Section 3 illustrations such an application with a regional potato supply chain. Economic and environmental impacts resulting from solving the example are discussed in Section 4. Finally, Section 5 states the framework's simplifications and concludes with proposals for further improvements. 2. The land-network framework Addressing environmental sustainability in agro-food systems requires an inter-disciplinary approach to recognize the

3

interdependency between agriculture, supply chains and ecological issues (Meinke et al., 2009; Notarnicola et al., 2012). The proposed framework implements a multi-step methodology for planning sustainable, cost-effective agro-food ecosystems that provide two fundamental services: (1) fulfillment of food demand and (2) GHG mitigation towards a zero-emissions ecosystem. Fig. 2 depicts these 5 steps. 1. 2. 3. 4. 5.

Defining ecosystem boundaries; Collecting a spatial data inventory; Performing a land-use assessment; Determining land-use benefits and costs; Solving the land-network model.

Fig. 2. Land-network framework.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

4

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

The following sections describe the steps in greater detail. Decision-making must be integrated across information types to capture the interdependency between actors, activities, and resources in the agro-food sector and to account for the economic costs and environmental impacts associated with ecosystem services. 2.1. Ecosystem boundaries The decision-maker first identifies the land areas of interest. The size of these areas should scale with the accuracy of spatial information. The strength of the analysis depends upon the robustness, quality, and resolution of this information (Ostendorf, 2011). Since climatic and soil characteristics are crucial for accurate modelling of the LUA sub-problem, use of a fine grid, such as a one hectare scale, is preferred. Fig. 2 depicts a schematic example of ecosystem boundaries represented by a cluster of lands, each characterized by a number of hectares and geographic coordinates. The ecosystem boundaries can include heterogeneous lands, even those separated by vast distances, as candidates for land-use changes.

2.2. Spatial data inventory Applications of agricultural spatial information include solving the location problem (Lucas and Chhajed, 2004) to evaluate landuse alternatives, boost crop yields, schedule operations and labor, and minimize costs or environmental impacts (Pacini et al., 2004; Ahumada et al., 2009; Bryan et al., 2011). Since agriculture is the sector most influenced by the environment, an assessment of agrofood ecosystems must consider soil, groundwater, climate, and other inter-dependent environmental factors (FAO, 1979). Robust data collected over many growing cycles enables crop simulation models to better allocate land (Soltani and Hoogenboom, 2007). In logistical operations, the application of spatial information supports decision making at both strategic and tactical levels: establishing production and distribution networks, routing, scheduling shipments and loading vehicles. The LN framework integrates agriculture and supply chain processes within the same agro-food ecosystem, factoring climate and soil data into strategic supply chain decisions. Indeed, the location of crops affects the establishment of food production systems and warehouses with tradeoffs between economic and

Fig. 3. Data collection domains.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

environmental costs. The number of processes involved necessitates a multi-disciplinary approach: collecting, integrating, and managing data from different fields. Fig. 3 illustrates a subset of spatial information required by the LN framework. It includes data from different informative layers: the land layer, the soil layer, the climate layer, and the supply chain layer. Given a set of selected areas, the land layer includes geographic coordinates, available hectares, population density, and distances between lands. The soil layer and the climate layer provide parameters necessary for forecasting agricultural production, plant growth, and soil erosion (FAO, 1979; Van Ittersum et al., 2003; Jones et al., 2003). A land's water availability is presently calculated from the total rainfall within the time horizon of analysis, as constrained by the soil permeability. The spatial data inventory does not currently include a layer to collect exogenous water inputs, a feature planned for future development. The supply chain layer covers food production and logistics operations and includes the locations, capacities, and operative costs/impact of the existing logistics mid-points (MPs), demand points and quantities, and economic and environmental characteristics of available transport modes. Successful management of strategic ecosystem decisions requires a wide range of information. The LN framework supports the linkage between disciplines and identifies relationships between

5

climate, soil, environmental resources, energy sources, and supply chain information through the top-down approach shown in Fig. 2. The dearth of robust data sources spanning disciplines impedes collecting the appropriate information. Improving such datagathering networks and information management systems is crucial to ensuring accurate management of agro-food ecosystems. 2.3. Land-use assessment The land-use assessment step considers alternative land-uses for the provision of agro-food ecosystem services. Land-use decisions are considered not only at the farm, but also at downstream supply chain levels, where interactions have been undervalued (Witlox, 2005). The current framework includes the following five uses: croplands, processing facilities, warehouses, renewable energy generation, and forests. Crops produce food that is then processed and packaged. Food is next stored in warehousing facilities that enable product conservation and, ultimately, delivery to retailers. These processes can be fueled by either endogenous renewable energy sources or exogenously produced fossil fuels. While the use of land to generate renewable energy decreases the land available for food production, the benefit is a decreased reliance on traditional fossil fuels and lower emissions. Emissions resulting from burning fossil fuels can

Fig. 4. Land-network uses for a sustainable ecosystem.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

6

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

also be offset through forestation. Although crops can sequester carbon, forests require less water and fewer nutrients to do so, resulting in cheaper and longer term emissions mitigation as well as biodiversity conservation (Strengers et al., 2008; Brussard et al., 2010; West et al., 2010; Seguin, 2011). By including supply chain links (transportation) and nodes (facilities and demand points) the observed agro-food ecosystem exceeds local boundaries, although the framework considers the distance food travels. While the competition between renewable and non-renewable energy is not a core issue of the framework as per Fischer et al. (2010), deciding between energy sources affects land availability as well as the need for plantings to offset emissions.

climate, soil) and endogenous data (i.e. nutrients, water inputs) to forecast crop yields. Data from the spatial climate layer determines a land's potential for renewable energy generation, selecting whether it is better suited for solar or wind installations. Additionally, the benefits and costs of logistics land-use are quantified based upon facilities' infrastructure, technologies for packaging, storage, and shipping, transportation modes and geographic distance between supply chain stages. The goal of this step is to identify and measure inputs, outputs and externalities associated with all ecosystem activities, generating the following required sets and parameters for the LN problem.

Sets l c uεU

1,..,L lands 1,..,C Points of food Demand, PoD {crop (cr), facility (pr), warehouse (wh), energy (en), plantings (pl)}

Parameters dll dlc tcuu tcuc CO2lu pCO2lu tCO2uu' tCO2uc pwlu wlu tw uu tw uc nl1 fcu cpul eu fdc

geographic/routing distance between lands l and l’ (km) geographic/routing distance between land l and PoD c (km) transport cost between use u and use u' (V/km) transport cost between use u and PoD c (V/km) carbon emissions/absorption resulting from use u in land l (ton CO2eq/hectare year) carbon emission resulting by processing a flow of product on land l by use u (ton CO2eq/ton) carbon emissions for transport between use u and u' (ton CO2eq/km ton) carbon emissions for transport between use u and PoD c (ton CO2eq/km ton) energy required to process a flow of product with use u on land l (kwh/ton) energy required to establish use u on land l (kwh/hectare year) energy for transport between use u and u' (kwh/km ton) energy for transport between use u and PoD c (kwh/km ton) number of hectares per land fixed costs to establish use u (V/hectare) capacity of use u (per hectare) allocated to land l conversion factor for carbon emission associated with power generation (ton CO2eq/kWh) demand for food by PoD c

The assessment of land-use alternatives illustrates the interdependency between agriculture, logistics, energy and the environment. Researchers (Tilman et al., 2002; Sheerr and McNeely, 2008; Spiertz, 2010) have emphasized the huge environmental impacts associated with agriculture, including emissions, pesticides, biodiversity losses, and soil degradation. The framework expands the dimension of the problem to a holistic perspective. Even though reducing emissions is not the only facet of environmental sustainability, the framework tackles one of the key goals of international agreements to mitigate human-induced climate change. Fig. 4 illustrates the five land uses and the food and carbon cycles within the inclusive ecosystem. Three flows are traced: food as it travels through the supply chain, power from renewable energy generation, and emissions being sequestered by carbon plantings. 2.4. Land-use benefits and costs Based upon the data collected the decision-maker next quantifies the potential of the lands for providing ecosystem services. A crop-forecast model similar to those found in other agricultural studies (FAO, 1979; Bouman et al., 1996) determines food production potentials. Such models use both spatial exogenous (i.e.

xlu ylul'u' ycluc

2.5. Land-network model The use of optimization in rural planning and supply network designs is well documented (Riveira and Maseda, 2006; Ahumada and Villalobos, 2009). The LN framework is formulated as a strategic single period linear programming (LP) model, addressing the objective of cost minimization for growers and supply chain actors while including carbon and energy balance constraints to support policy makers' climate goals. Although operational food distribution planning typically involves time-dependent optimization, the strategic perspective of land-use planning fits within a single period model with an inherently long time horizon. Allocating land to a specific use requires substantial time and cost, so switching between uses during the implied time horizon would not be optimal. Therefore, the LN model handles the strategic planning of the agro-food ecosystem, identifying the long-term enabling conditions for sustainable economic and climate stability. Given the sets and parameters proposed in Section 2.4, the LN model is formulated as follows. 2.5.1. Decision variables

number of hectares of land l devoted to use u (hectare) product flow between land l and l' devoted respectively to use u and u' (ton) product flow between land l devoted to use u and Point of food Demand (PoD) c (ton)

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

2.5.2. Objective function

min

L X X

fcu xlu þ

þ

X

dll0 ylul0 u0 tcuu0

l;l0 ¼1 u;u0 2U

l¼1 u2U L X X C X

L X

(1)

dlc ycluc tcu0 c

7

Constraints (6) and (7) balance the food flows within the agrofood ecosystem, subject to potential downstream losses. Constraint (6) forces the amount of food processed and warehoused to be at least that found at the warehouse and retail (consumer) level. Constraint (7) ensures food set to the processors from croplands is at least as much as the processed food sent to the warehouses.

l¼1 u2U c¼1

The objective function (1) minimizes the total fixed and variable costs of the agro-food ecosystem. The fixed costs represent the establishment of food crops and facilities, renewable energy generation, and carbon plantings. The variable costs currently tallied are those associated with transport activities, but costs associated with agriculture, packaging and storage could later be included. The costs, emissions, and energy requirements for transport are dependent upon supply chain stage. For example, packaged food may have differing energy requirements for transport than it does prior to processing and packaging. Thus, the LN model includes also the decision to process and package onsite, close to crops.

2.5.6. Land availability constraints

X

xlu  nll cl : 1; ::; L

Constraint (8) manages the cluster of lands as a finite resource within the ecosystem. 2.5.7. Environmental constraints

L X

2.5.3. Demand constraints

(8)

u2U

X

ylul0 u0 pco2lu þ

l;l0 ¼1 u;u0 :cr;pr;wh L X X

ycluc  fdc cc : 1; ::; C

(2)

þ

2.5.4. Capacity constraints

C X

L X

X

ylul0 u0 tco2uu0 dll0 þ

X

xlu co2lu 

(3)

L X X

ylul0 u0  cpul xlu cl : 1; ::; :L; u : pr

(4)

u0 :wh

ycluc tco2lc dlc

L X X

eu cpul xlu

l¼1 u:en

(9) X

ylul0 u0 pwlu þ

þ

L X

C X L X X

þ

L X

X X

l¼1 u:cr;pr;wh

ycluc pwlc

c¼1 l¼1 u:wh

ylul0 u0 twuu0 dll0 þ

l;l0 ¼1 u;u0 :cr;pr;wh

L X X

cpul xlu þ

l¼1 u:pl

l;l0 ¼1 u;u0 :cr;pr;wh

ycluc  cpul xlu cl : 1; :::L; u : wh

C X L X X c¼1 l¼1 u:wh

l¼1 u:cr;pr;wh;en

L X

c¼1

l0 ¼1

þ

ycluc pco2lc

c¼1 l¼1 u:wh

l;l0 ¼1 u;u0 :cr;pr;wh

l¼1 u:wh

The set of linear constraints includes constraints from both the LUA and LA problems. Constraint (2) guarantees that consumers' total food demand is met.

L X

C X L X X

xlu wlu 

C X L X X

ycluc twlc dlc

(10)

c¼1 l¼1 u:wh L X X

cpul xlu

l¼1 u:en

Constraints from (3) to (5) respectively limit the food flows from warehouses, packaging facilities, and croplands in accordance with storage capacity, packing-line throughput and crop yield. Constraint 5 shows crop yields depend on climate and soil conditions denoted in the spatial data inventory and the water supply, which is limited by a land's total rainfall over the time horizon. The framework assumes appropriate irrigation balances soil evapotranspiration.

Constraints 9 and 10 allow the model to address environmental externalities. The economic and environmental trade-offs can be analyzed by comparing scenarios where these constraints are either enforced or relaxed. Constraint (9) enforces a zero-carbon ecosystem; overall emissions associated with crops and logistics activities must either be offset through sequestration from forestation (use pl) or renewable energy usage. Constraint (10) strictly limits the supply of energy to the renewable energy sources (use en) established within the ecosystem. When constraint 10 is enforced, the total amount of fossil fuels needed to power the food supply chain is matched the production of renewable from either solar fields or wind farms.

2.5.5. Transfer flow constraints

xlu ; yluc ; ylul0 u0  0 cl : 1; ::L; u2U; c : 1; ::; C

L X X l0 ¼1 u0 :pr

L X

ylul0 u0  cpul xlu cl : 1; …L; u : cr

X

yl0 u0 lu 

l0 ¼1 u0 :pr;wh

L X X l0 ¼1 u0 :wh

ylul0 u0 þ

C X

(5)

ycluc cl : 1; ::L; u : wh

c¼1

(6) L X X l0 ¼1 u0 :cr

yl0 u0 lu 

L X X l0 ¼1 u0 :wh

ylul0 u0 cl : 1; :::L; u : pr

(7)

(11)

Finally, constraint (11) enforces non-negativity for the decision variables. Fig. 5 illustrates land-use decisions and their impacts on the agro-food ecosystem. The logistics network integrates clusters of lands within the ecosystem and shares inputs, outputs, and externalities. Food demand is met through a combination of local and global crops and logistics facilities to build the food pipeline from farm-to-fork. Transport-intensive networks result in higher emissions, thus requiring more land set aside for offsets either through renewable energy generation or forestation.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

8

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

Fig. 5. Land-network models: parameters and sets (see Section 2.4).

3. On the design of a regional scaled potato supply chain We verify the effectiveness of the proposed framework by designing a sustainable agro-food supply chain for Bologna Potatoes. Emilia-Romagna produces 2.2 million tons of potatoes annually with a turnover of about 50 million euro. High quality standards are reflected in its protected designation of origin (PDO), granted in 2012 by the European Commission. The Bologna Potato has a traditional place in the regional diet for its nutritional properties that are enhanced by the local soil and climate. Italians annually consume 21 million tons of potatoes, but Italy produces only 15 million tons, importing the remainder mostly from France. The demand for PDO Bologna Potatoes has been increasing not just within Emilia-Romagna but across Europe. In order to fulfill growing demand, production and distribution activities are shifting from small scale operations to intensive agrofood models, thereby affecting the long-term local environmental conditions (i.e. air, water, and soil quality) that contribute to the sine qua non of this potato. The LN framework supports strategic decisions in planning a new sustainable Bologna Potato supply chain that would provide economic and environmental benefits. We place croplands, processing facilities, and warehouses and then optimize the flows within the resulting land-network as well as establish renewable energy installations and carbon plantings. Results from the analysis could show agro-food actors (i.e. farmers, agro-food entrepreneurs, logistics providers, and policy makers) how to balance economic goals with environmental externalities in a more efficient, fair and sustainable agro-food system. Strategic planning for a new sustainable Bologna Potato supply chain incorporates the existing network of demand points: several retailer shops and other markets within the observed region. The collected spatial data inventory includes twenty owners of potential land clusters within the Bologna province. Each of these clusters has the following data: available hectares per each cluster, fixed and variable costs for the adopted technology and equipment, energy requirements and resultant emissions associated with each process, climate and soil of the observed geographic regions, and product properties, such as water and nutrient requirements.

Fig. 6 shows how climate characteristics vary in the observed lands. While climate conditions are similar within the observed region, Table 1 illustrates the diversity of soil characteristics between 16 of the 20 lands. The framework calculates land-use benefits based on these land-dependent climate and soil conditions. Fig. 7 depicts the different potentials from establishing food crops and renewable energy sources in the various lands. The benefits from allocating lands to processing, distribution and storage depend not on environmental conditions but on the available infrastructure and equipment, which are assumed to be identical across the observed region for PDO Potatoes. The demand points are depicted as blue spots in the map adjacent to Table 1. Table 2 reports land-network characteristics used as inputs: the land clusters that comprise the agro-food ecosystem, the markets and demand, the transport modes and costs, infrastructure costs (i.e. land-use allocation), energy requirements and related emissions. For crops, renewable energy generation, and carbon plantings capacities depend on climate and soil conditions. Capacity for processing and warehousing facilities depends on the available equipment. The FAO (1979) consolidated crop-forecast model calculates crop yields as illustrated in Fig. 7. We consider different transport modes between supply stages. The model allows for shipments between warehouses, but these are recommended only for the adoption of intermodal transport modes such as truck-to-train that reduce the environmental impact of distribution activities. We define three separate scenarios. The first scenario allows for unlimited use of imported fossil fuels to power the agro-food supply chain activities from farm-to-fork and makes no requirements for mitigating resultant emissions. By relaxing constraints (9) and (10) this business-as-usual scenario depicts the pure economic optimization of agro-food facility location and flow allocation across the supply chain. The second scenario requires net carbon neutrality of the agro-food supply chain by enforcing constraint (9). Finally, the third scenario models the design of a selfsufficient agro-food ecosystem that supplies renewable energy sufficient to power the food supply chain. This scenario is enacted by enforcing both constraints (9) and (10). Note that we do not need to assume that solar and wind farms will directly power vehicles,

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

9

Fig. 6. Average monthly climate conditions (i.e., temperature, solar radiation, rainfall, humidity) experienced by the observed lands.

scenario 2 plants forests to offset the emissions from fossil fuels, resulting in extensive land use. Conversely, scenario 3 is forced to generate renewable energy sufficient to offset agro-food supply chain emissions. An almost negligible amount land in scenario 3 is allocated for plantings to sink carbon generated from non-food related activities, specifically the establishment of the solar and wind farms. However, the overall land usage is much smaller than in scenario 2. Scenario 3's requirement for the endogenous production of energy to support the agro-food ecosystem raises the price of food only slightly (<1%) from Scenario 2, but results in a dramatically different solution considering other KPIs such as land use. The illustrated scenarios show the interdependency between infrastructure, food production and distribution, and environmental resources, highlighting the consequences of unbalanced planning. The LN model quantifies the opportunity costs of addressing environmental objectives in land-network design and considers both economic and environmental leverage. Applying this framework to another set of lands and their particular characteristics would likely yield differing scenario results and subsequent recommendations. For example, the cost difference of enforcing

but rather the renewable energy feeds into the regional electric grid, displacing the burning of fossil fuels for other uses outside the ecosystem's services, such as powering homes. The LN model is modeled in AMPL and solved with Gurobi. For each of these three scenarios, Gurobi takes less than 1 s to find the optimum solution on a computer configured with Intel® Quad Core 2.4 GHz processors and 8 GB of RAM. This speed suggests that the framework could be expanded to address larger and more complex problems in the future. The results are summarized in Table 3, along with a radar graph that depicts economic and environmental KPIs for each of the 3 scenarios. These four KPIs follow: the cost per supplied product, net mitigated emissions, and the land use and the energy use impact per kilogram of food production. Scenario 1 supplies cheap but carbon-intensive food, costing only 0.12 V/kg but emitting 1.83 kg CO2eq/kg. As the next two scenarios incorporate environmental externalities, resultant food prices increase by approximately 150% from the business-as-usual scenario. While some slight energy reduction is achieved through changes to the food supply chain, most of the emissions require offsets generated either through either plantings or renewable energy. Rather than invest in renewable energy generation,

Table 1 Soil features are measured for each land. The brown squares represent the available lands, and the blue circles represent demand points. Map of the land across the regional area

Land

Coarse Clay Silt Sand pH H2O pH CaCl2 OC

CaCO3 N

P

K

CEC

Land01 Land02 Land03 Land04 Land05 Land06 Land07 Land08 Land09 Land10 Land11 Land12 Land13 Land14 Land15 Land16

18 9 10 3 10 7 5 7 3 10 20 12 13 17 11 7

77 192 148 240 247 165 13 156 0 12 97 15 166 116 85 42

22.4 76.1 10.7 14.5 0 34.5 0 25.8 0 19.7 13 44 11.3 15.7 0 29.1

299.9 849.1 243.4 307.8 206.4 358.5 269.6 409.7 552.9 115.6 182.7 381.1 255 221.6 532.2 224.9

27.2 23.9 24.4 27.1 23.1 15.4 76.7 26.6 32.4 21.6 25.2 39.5 16.3 22 34.1 21.2

38 33 37 52 40 23 68 54 59 32 42 54 28 36 62 33

49 58 58 47 53 58 29 44 33 47 48 36 58 37 37 52

13 10 6 1 7 19 3 2 8 22 10 10 14 27 2 15

7.89 7.44 7.75 7.99 8.18 8.07 7.05 8.2 7.56 7.98 7.71 7.55 7.74 8.19 7.92 7.93

7.34 7.05 7.22 7.18 7.36 7.42 6.87 7.49 6.92 7.47 7.29 7.16 7.18 7.48 7.25 7.29

8.8 25.4 10.8 13.5 9.3 11.7 160.3 10.9 12.6 7.5 12.9 14.2 13.5 9.8 13.3 11.9

1.1 2.5 1.3 1.5 1.1 1.2 9.5 1.3 1.5 1.1 1.5 1.7 1.6 1.2 1.6 1.2

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

10

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

Solar Energy (kWhE-1/hc year)

128400 128200

128000 127800 127600 127400 127200

Land01 Land02 Land03 Land04 Land05 Land06 Land07 Land08 Land09 Land10 Land11 Land12 Land13 Land14 Land15 Land16 Land17 Land18 Land19 Land20

Crop Yield (kg/hc year)

120000 115000 110000 105000 100000 95000 90000

Land01 Land02 Land03 Land04 Land05 Land06 Land07 Land08 Land09 Land10 Land11 Land12 Land13 Land14 Land15 Land16 Land17 Land18 Land19 Land20

cp pl,Land02

cpcr,Land02

Crop Yield (kg/hc year) Processing Throughput (ton/hc year)

cpen,Land02

Land02 cp pr,Land02

Land12

Storage Capacity (ton/hc year) Solar Energy (kWh·10/hc year)

cpwh,Land02

Carbon Sink (kg CO2eq/hc year)

Fig. 7. Different land-use benefits/potentials (i.e., cpu,l) calculated per each land (e.g., Land02, Land06, and Land12): crop yield and retrievable solar energy.

endogenous production of renewable energy might have been more significant were renewable costs higher or yields lower. Conversely, if renewable generation became more cost effective or perhaps land availability decreased, renewable energy generation could become more attractive than carbon sink plantings. In that case Constraint 10 would be non-binding at the optimum solution, and the results from scenarios 2 and 3 would effectively be identical. A larger ecosystem comprised of lands with diverse climate and soil characteristics might result in scenario 2's zero net emissions goal being met through a more balanced mixture of both forestation and renewable energy generation. In short, the importance of using finely scaled and accurate data as provided by the LUA portion of the framework can be seen by considering that slight revisions to the underlying data may have large effects. For example, decreasing the relative productivity of land for carbon plantings compared to renewable power generation could dramatically shift results and subsequent policy recommendations at a more global scale. The model would be valuable in determining the optimal mix of renewable energy generation and

carbon sink plantings for a particular geographic region's agro-food system to meet climate stability goals. Although only potatoes are considered in this example, the framework could handle supply chains associated with other food products and could even be extended to simultaneously model multiple products. 4. Discussion on the usage of the framework The LN framework provides a useful tool for policy makers and food industry practitioners responsible for the design of costeffective and sustainable agro-food systems. First, the framework is underpinned by the collection of robust spatial data inventories that integrate information across disciplines such as agronomy, logistics, and environmental engineering. Few planners today consider regional climate or soil properties when designing logistics services. Yet the framework demonstrates that climate and soil conditions affect the optimal allocation of the crops and the resulting supply chain, even in the compact region covered by the potato supply chain. Different latitudes' varying environmental

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

11

Table 2 Land-network characteristics and inputs. Consumers

Land-network

Demand Nodes (e.g., retailers) Potato Demand per Consumer (tons) Annual Total Demand (tons)

10 1500 15,000

Cluster of Lands Hectares per Cluster Total Hectares

20 25 500

From Land-use

To Land-use

Transport Mode

Cost (V/kgkm)

GHG Emission (gCO2eq/kgkm)

Energy (kWh/kgkm)

Transport Costs Crops Facility Warehouse Warehouse

Facility Warehouse Consumer Warehouse

Truck Reefer Truck Reefer Truck Reefer Train

0.00016 0.00020 0.00020 0.00013

0.042 0.050 0.050 0.011

0.00017 0.00020 0.00020 0.00004

Land-use

Infrastructure Costs (V/ha/year)

Capacity

Crops Facility Warehouse Energy Carbon Plantings

6500 300,000 200,000 75,000 11,000

Climate/Soil-dependent 9,000,000 10,000,000 Climate-dependent 110,000

conditions can affect supply chain performance in terms of costs and associated environmental impacts, including emissions and land-use. Thus, the application of the framework to planning globalized agro-food production and distribution systems has great potential. The framework addresses the struggle between the deterritorialization and re-territorialization of agro-food systems (Morgan et al., 2006). De-territorialization is promoted by globalindustrial food systems based on concentrated monoculture, spatial homogeneity and intensive use of fossil fuels for cultivation, processing and transport. Re-territorialization aims to restore the link between land and consumers, attempting to reduce environmental impacts associated with the food ecosystem. The LN framework optimizes the supply chain by allocating land and shortening the pipeline from farm-to-fork when appropriate but is not a blind advocate for local food models (Edwards-Jones et al., 2008; Coley et al., 2009). Instead, it highlights the inter-

Crop yield Processing throughput Storage Renewable energy Carbon sink

(kg/ha/year) (kg/ha/year) (kg/ha/year) (kwh/ha/year) (kg CO2eq/ha/year)

dependency between multiple inputs, modelling the agro-food supply chain as a whole ecosystem that exists to fulfill food demand fulfillment and create economic value for supply chain actors while satisfying emissions goals. Thus, the design of sustainable food processing and distribution networks largely depends on localized factors such as an area's crop yield and potential for renewable energy collection. The enabling conditions of the ecosystem services depend on the intensity of human-based activities and exogenous environment properties illustrated in Section 2.2. Through offsetting emissions and generating renewable energy, the LN framework enables decision-makers to better balance long term objectives of ecosystem preservation. The framework identifies trade-offs between agro-food ecosystem services and environmental sustainability objectives, a concern of Gunasekaran and Kobu (2007). Cost-efficiency couples with land-use, energy-use and emissions, providing a panel of economic and environmental KPIs. The framework incorporates

Table 3 Land-network results and KPIs comparison. Results

GHG Impact of Energy Supply (kg CO2eq/kWh) GHG Impact of Transport (kg CO2eq/kWh) Total Cost (MV) Fixed Costs (MV) Transportation Costs (kV) GHG Impact (ton CO2eq) GHG Sink (ton CO2eq) Fossil Fuel (MWh) Renewable Energy (MWh) Crops Land (ha) Facility Land (ha) Warehousing Land (ha) Energy Land (ha) Plantings Land (ha) Total Land (ha) Food Cost (V/kg) Food Carbon Impact (kg CO2eq/ kg) Food Land-use Impact (m2/kg) Food Energy-use Impact (kWh/ kg)

Scenario 1

Scenario 2

Scenario 3

No requirements on energy or CO2e offsets required CO2e offsets

CO2e offsets, use of renewables required

0.4455

0.4455

0

0.247

0.247

0.247

1.857 1.765 91.35 27,386.00 0 61,517.70 0 125.5 1.67 1.5 0 0 128.67 0.1238 1.83

4.594 4.503 91.6 27,373.00 27,373.00 61,488.60 0 125.5 1.67 1.5 0 248.85 377.52 0.3063 0

4.63 4.538 91.6 23.12 23.12 61,488.60 61,488.60 125.5 1.67 1.5 36.94 0.21 165.82 0.3087 0

0.09 4.1

0.25 4.1

0.11 0

Land-networks KPIs comparison

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

12

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

quantitative metrics (see Table 2) for measuring the efficacy of sustainable agro-food systems in their ability to address systemic issues such as food security or land grabbing in response to climate change (Cotula et al., 2009; World Bank, 2011; Popp et al., 2014). These metrics can serve as drivers for the design of new sustainable supply chains, rather than just the assessment of the existing ones, and they can be applied across different geographic areas and foods. First, food cost (see Table 3) is a recognized KPI that drives agro-food entrepreneurs to invest in crop equipment, packaging facilities or warehouses, and highlights how each step of the supply chain (i.e. agriculture, packaging, storage, transport) contributes to the overall cost. An analysis of this metric's cost components informs supply chain actors on what aspects merit improvement. The framework also quantifies other factors that will impact KPIs such as perceived product value, logistics and transport costs, and return on investment. The second proposed metric, food carbon impact, tallies the overall emissions from agro-food supply chain processes and enables policy makers to make holistic decisions to combat climate change. As societies and ecosystems are vulnerable to even modest levels of climate change (UNEP, 2009), huge benefits may exist for regulating manmade emissions at the government level: improving air quality, reducing healthcare costs, decreasing disaster risk, tackling social and environmental disruptions. The food carbon impact quantifies the agro-food ecosystem's role and provides policy makers options to regulate the sector to comply with environmental policy (Stern, 2006). For example, in the presence of carbon taxes the LN framework calculates the cost of shifting from the business-as-usual scenario to a carbon-balanced scenario, quantifying for each agro-food actor the economic benefit of implementing environmental strategies instead of paying taxes. Conversely, in presence of cap-and-trade regulations the framework serves as a tool for policy makers to measure the carbon impacts of existing agro-food systems and thereby impose realistic boundaries upon emissions induced by the agro-food industries. The third metric, land-use impact, directs policy makers, geographers, and land planners toward more conscious territory management, considering the agro-food ecosystem services and environmental constraints of the observed region. The last metric, energy-use impact, enables both policy makers and supply chain actors to scrutinize their fossil fuel dependency. Aside from environmental considerations, processes that rely on fossil fuels are also less resilient in energy crises and are more susceptible to fuel price shocks. As battery powered trucks are not yet a realistic option for most transport routes, other strategies might be carried out across the supply chain to reduce hydrocarbon dependency. These include solar or wind fields to power storage facilities, packing lines, and hybrid crop equipment. This last KPI guides decision-makers towards a more sustainable and resilient management of the agrofood system. The combination of multiple indicators within a dashboard allows for vast quantities of economic and environmental supply chain data to be summarized and compared. By equalizing indicators' weights, the panel allows for different supply chains to be benchmarked and for the most cost-effective strategies to achieve environmental goals to be found, per (Jasch, 2000). Policy makers should note the inherent tradeoffs between food cost and environmental policies, as emphasized by Gunasekaran and Kobu (2007). Section 3's potato case study illustrates how to design the best network from growers to consumers through use of the LN framework. Since this supply chain has been designed from a green field, no benchmarks are available to compare the resulting scenarios with the actual supply chain. However, the LN framework quantifies the environmental externalities in terms of both fossil

fuel use and emissions. In scenario 2, the climate stability of the agro-food system is achieved through carbon plantings, dramatically increasing the overall land-use impact KPI and costs. Scenario 3 yields an alternative carbon reduction land use strategy: generating renewable energy. In this last scenario the land usage is much smaller than in scenario 2, but the cost is slightly higher. Of course, KPIs depend on the region's potential of renewable energy collection compared to sequestration. The findings of the case study highlight the role of the LN framework in identifying the optimal enabling conditions for an agro-food supply chain. Despite policy makers' interests, most agro-food actors are currently more concerned with costs than carbon footprints, and incentive barriers and technological constraints impede the implementation of mitigation options in the food sector (Smith et al., 2007). The LN framework aims to minimize the infrastructure and, ultimately, the operative costs of the agro-food supply chain, and distributes the responsibility for environmental impacts among growers, processors, carriers, and consumers. Indeed, another objective of this model is to promote more sustainable choices at each supply chain stage (e.g., crop placement for farmers, warehouse location for distributors, energy infrastructure for producers' packaging plants), through increased visibility of the induced environmental impacts. Inspired by industrial-ecology patterns (Ometto et al., 2007; Motet et al., 2007), the LN framework develops an original holistic methodology to analyze the life cycle assessment (LCA) of agrofood systems from farm-to-fork, observed as a hotspot of climate change (Andersson, 2000; Amate and De Molina, 2013). Notwithstanding the importance of assessing ex-post the environmental impacts induced by food systems (Roy et al., 2009; Virtanen et al., 2011; Van der Werf et al., 2014), the LN framework supports exante decision-making on agro-food ecosystem using both economic and environmental criteria. As the LN model considers both the local and global dimensions of agro-food systems, it can ultimately provide insight into the trade-offs inherent in choosing between localized small food systems and globalized intensive agriculture models. Taking into account the economic and environmental drivers of Table 3, the decision-maker can weigh the lower transportation emissions from local supply chains against the lower production costs from crops grown in less developed countries. Currently, this balancing between small production and intensive agriculture does not consider greater crop yields or other efficiencies generated from combining adjacent lands for homogeneous production. Thus, one planned improvement will be to capture economies of scale effects for alternative land uses. 5. Conclusions Although this is not the first time the environmental impacts of agro-food supply chains have been studied, our proposed landnetwork (LN) framework goes beyond current planning methodologies for agro-food systems. The LN framework involves a multistep methodology for planning sustainable agro-food supply chains according to climate stability and ecosystem conservation purposes. The framework is underpinned by a multi-disciplinary approach balancing agricultural, logistical and ecological issues and integrates spatial data inventories to better understand the interdependency between actors and activities in the agro-food sector and to rationalize the economic and environmental costs for the provision of ecosystem services given fixed resources. This paper contributes in four ways. First, the proposed framework models the agro-food supply chain and all associated processes as a unique closed ecosystem. Secondly, it includes costbased efficiencies and environmental externalities in a strategic

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

long-term support decision model. Thirdly, it merges and harmonizes network location-allocation and land-use constraints within a unique linear programming (LP) model to minimize costs for growers and supply chain actors while meeting constraints imposed by policy makers. Lastly, since cost-efficiency is coupled with land-use, energy-use and emissions, it quantifies a panel of comparable and concise metrics that drive decision makers toward more sustainable strategies in agro-food systems. This framework designs a sustainable and cost-effective ecosystem that provides food services while satisfying environmental goals. While the environmental goals currently focus upon emissions mitigation, the framework could be easily extended to consider other environmental factors through the addition of further limiting constraints. For example, the inclusion of soil and water impacts into a more comprehensive panel of KPIs would be a potential area for further development. The wide multidisciplinary dataset (see Fig. 3) required represents a critical limitation to the extended applicability of the framework. A detailed awareness of the observed agro-food system includes the costs, throughputs, capacities, and energy consumption of facilities, in addition to agricultural equipment, transport means, and climate and soil properties and related agronomic parameters for each food variety and crop. Thus, as far the findings of the framework support the decisions of any supply chain actor (e.g., farmers, food producers and carriers, and even consumers), the spatial data inventory should be organized by policy makers through eliciting partner collaboration and information sharing. When data are available, the proposed framework can be used to design a new sustainable supply chain from a green field. For instance, this framework could be applied in less developed countries by farmers' associations to identify the optimal conditions for establishing an entire agro-food ecosystem and answer questions such as: where to grow what crops? where and how to pack, store and transport food? Food producers from developed countries could implement the framework over a large scaled geography to evaluate the most cost-effective supply chain from farm-to-fork in presence of environmental constraints. The panel of the indicators could be used as an educational tool for those consumers concerned with environmental impacts associated with their foods and could even shift purchasing behaviors. Finally, it would provide a tool for policy makers to assess and identify best practices and new drivers for sustainable agro-food systems design. The LN framework is tested with a case study inspired by a PDO potato supply chain in Italy. Although the example is local in scale, the framework as developed could support a larger and more geographically diverse set of lands, subject only to limitations on data availability. Thus, a more substantive collection of robust spatial data inventories will be necessary to support the design of sustainable agro-food supply chains on a grander scale. Furthermore, economies of scale should be considered and added to the LN framework so that an appropriate balance between small, local food businesses and intensive global agriculture models can be determined. Further framework developments entail widening the ecosystem boundaries and including other land uses and flows. For example, incorporating the production and supply of packaging materials and modelling the disposal of food waste according to a farm-to-farm cycle are both priority enhancements. A what-if multi-scenario analysis may help to identify the right balance of small and global intensive agro-food systems and assess the factors (e.g., network infrastructure, agriculture technologies, climate and soil characteristics) affecting the economic and environmental sustainability of food supply chains. Finally, the authors do not intend that the LN framework be viewed as an exhaustive approach to address every issue facing agro-food systems, but rather as a call for research across disciplines to provide tools for

13

the design of cost-effective agro-food supply chains that proactively consider environmental externalities. Acknowledgments The authors would like to sincerely thank the anonymous reviewers for their valuable and fruitful comments. These contributed significantly to improving the manuscript and to highlighting its scientific value. References Ahumada, O., Villalobos, R., 2009. Application of planning models in the agro-food supply chain: a review. Eur. J. Oper. Res. 195, 1e20. Ahumada, O., Villalobos, R., 2011. A tactical model for planning the production and distribution of fresh produce. Ann. Operation Res. 190, 339e358. Akkerman, R., Van Donk, D.P., 2010. Balancing environmental and economic performance in the food processing industry. Int. J. Entrepreneursh. Innov. Manag. 11 (3), 330e340. Amate, J., Gonzalez de Molina, M., 2013. 'Sustainable de-growth' in agriculture and food: an agro-ecological perspective on Spain's agro-food system (year 2000). J. Clean. Prod. 38, 27e35. Andersson, K., 2000. LCA of food products and production systems. Int. J. Life Cycle Assess. 5 (4), 239e248. Apaiah, R.K., Hendrix, E.M.T., 2005. Design of a supply chain network for pea-based novel protein foods. J. Food Eng. 70 (3), 383e391. Azarmand, Z., Neishabouri, E., 2009. Location Allocation Problem. In: Facility Location: Concepts, Models, Algorithms and Case Studies Contributions to Management Science. Springer, pp. 93e109. ISBN 3790821519, 9783790821512. Bouman, B.A., Van Keulen, H., Van Laar, H., Rabbinge, R., 1996. The school of de wit crop growth simulation models: a pedigree and historical overview. Agric. Syst. 52, 171e198. Brussard, L., Caron, P., Campbell, B., Lipper, L., Mainka, S., Rabbinge, R., Babin, D., Pulleman, M., 2010. Reconciling biodiversity conservation and food security: scientific challenges for a new agriculture. Curr. Opin. Environ. Sustain. 2, 34e42. Bryan, B., King, D., Ward, J.R., 2011. Modelling and mapping agricultural opportunity costs to guide landscape planning for natural resource management. Ecol. Indic. 11, 199e208. Canadell, J.G., Le Quires, C., Raupach, M., Field, C., Buitenhuis, E., Ciais, P., Conway, T., Gillet, N., Houghton, R.A., Marland, G., 2007. Contributions to accelerating atmospheric CO2 growth from economic activity, carbon intensity, and efficiency of natural sinks. PNAS 104 (47), 18866e18870. http://dx.doi.org/10.1073/ pnas.0702737104. Cholette, S., Venkat, K., 2009. The energy and carbon intensity of wine distribution: a study of logistical options for delivering wine to consumers. J. Clean. Prod. 17 (16), 1401e1413. Coley, D., Howard, M., Winter, M., 2009. Local food, food miles and carbon emissions: a comparison of farm shop and mass distribution approaches. Food Policy 34, 150e155. Cotula, L., Vermeulen, S., Leonard, R., Keeley, J., 2009. Land Grab or Development Opportunity? Agricultural Investment and International Land Deals in Africa. IIED/FAO/IFAD, London/Rome. ISBN: 978-1-84369-741-1. Dai, Z.Y., Li, Y.P., 2013. A multistage irrigation water allocation model for agricultural land-use planning under uncertainty. Agric. Water Manag. 129, 69e79. Desjardins, R.L., Sivakumar, M.V.K., De Kimpe, C., 2007. The contribution of agriculture on the state of climate: workshop summary and recommendations. Agric. For. Meteorol. 142, 314e324. Edwards-Jones, G., Mil
a i Canals, L., Hounsome, N., Truninger, M., Koerber, G., Cross, P., York, E.H., Hospido, A., Plassmann, K., Harris, I., Day, G., Tomos, A.D., Cowell, S.J., Jones, D.L., 2008. Trends Food Sci. Technol. 19, 265e274. FACCE-JPI, 2011. Scientific Research Agenda of the ‘Agriculture, Food Security and Climate Change. Joint Programming Research Initiative. FAO, 1979. Yield Response to Water. In: FAO Irrigation and Drainage Paper, vol. 33. FAO, Rome. FAO, 2012. Sustainability Assessment of Food and Agriculture Systems (SAFA). FAO, Rome. Fischer, G., Prieler, S., Van Velthuizen, H., Berndes, G., Faaij, A., Londo, M., De Wit, M., 2010. Biofuel production potentials in Europe: sustainable use of cultivated land and pastures, part II: land use scenarios. Biomass Bioenergy 34, 173e187. Gunasekaran, A., Kobu, B., 2007. Performance measures and metrics in logistics and supply chain management: a review of recent literature (1995e2004) for research and applications. Int. J. Prod. Res. 45 (12), 2819e2840. Hinrichs, C., 2003. The practice and politics of food system localization. J. Rural Stud. 19 (1), 33e45. Jasch, C., 2000. Environmental performance evaluation and indicators. J. Clean. Prod. 8, 79e88. Jones, J.W., Hoogenboom, G., Porter, C.H., Boote, K.J., Batchelor, W.D., Hunt, L.A., Wilkens, P.W., Singh, U., Gijsman, A.J., Ritchie, J.T., 2003. The DSSAT cropping system model. Eur. J. Agron. 18, 235e265.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082

14

R. Accorsi et al. / Journal of Cleaner Production xxx (2015) 1e14

Koning, N., Van Ittersum, M., 2009. Will the world have enough to eat? Curr. Opin. Environ. Sustain. 1, 77e82. Leduc, W., Van Kann, F., 2013. Spatial planning based on urban energy harvesting toward productive urban regions. J. Clean. Prod. 39, 180e190. Lucas, M., Chhajed, D., 2004. Applications of location analysis in agriculture: a survey. J. Oper. Res. Soc. 55, 561e578. Manzini, R., Accorsi, R., 2013. The new conceptual framework for food supply chain assessment. J. Food Eng. 115, 251e263. Manzini, R., Accorsi, R., Bortolini, M., 2014. Operational planning models for distribution networks. Int. J. Prod. Res. 52, 89e116. Massoud, M., Fayad, R., El-Fadel, M., Kamleh, R., 2010. Drivers, barriers and incentives to implementing environmental management systems in the food industry: a case of Lebanon. J. Clean. Prod. 18, 200e209. McCown, R.L., 2002a. Changing systems for supporting farmers' decisions: problems, paradigms, and prospects. Agric. Syst. 74, 179e220. McCown, R.L., 2002b. Locating agricultural decision support systems in the troubled past and socio-technical complexity of 'models for management'. Agric. Syst. 74, 11e25. Meinke, H., Howden, M., Struik, P., Nelson, R., Rodriguez, D., Chapman, S., 2009. Adaption science for agriculture and natural resource management e urgency and theoretical basis. Curr. Opin. Environ. Sustain. 1, 69e76.
i Canals, L., Cowell, S.J., Sim, S., Basson, L., 2007. Comparing domestic versus Mila imported apples: a focus on energy use. Environ. Sci. Pollut. Res. 14 (5), 338e344. Morgan, K., Marsden, T., Murdoch, J., 2006. Worlds of Food. Place, Power and Provenance in the Food Chain. Oxford University Press, Oxford, UK, p. 225. Motet, A.R., Ramos, P., Lombardi, G., 2007. The benefits of a Brazilian agro-industrial symbiosis system and the strategies to make it happen. J. Clean. Prod. 15, 1253e1258. Nilsson, H., 2004. What are the possible influences affecting the future environmental agricultural policy in the European Union? An investigation into the main factors. J. Clean. Prod 12, 461e468. Notarnicola, B., Hayashi, K., Curran, M.A., Huisingh, D., 2012. Progress in working towards a more sustainable agro-food industry. J. Clean. Prod. 28, 1e8. Ometto, A.R., Ramos, P., Lombardi, G., 2007. The benefits of a Brazilian agroindustrial symbiosis system and the strategies to make it happen. J. Clean. Prod. 15, 1253e1258. Ostendorf, B., 2011. Overview: spatial information and indicators for sustainable management of natural resources. Ecol. Indic. 11, 97e102. Ovando, P., Caparros, A., 2009. Land use and carbon mitigation in Europe: a survey of the potentials of different alternatives. Energy Policy 37, 992e1003. Pacini, C., Wossink, A., Giesen, G., Huirne, R., 2004. Ecological-economic modelling to support multi-objective policy making: a farming systems approach implemented for Tuscany. Agric. Ecosyst. Environ. 102, 349e364. Partidario, P., Lambert, J., Evans, S., 2007. Building more sustainable solutions in production-consumption systems: the case of food for people with reduces access. J. Clean. Prod. 15, 513e524. Ponsioen, T.C., Blonk, T.J., 2012. Calculating land use change in carbon footprints of agricultural products as an impact of current land use. J. Clean. Prod. 28, 120e126. Popp, A., Rose, S., Calvin, K., Van Vuuren, D., Dietrich, J., Wise, M., Stehfest, E., € der, F., Kyle, P., Van Vliet, J., Bauer, N., Lotze-Campen, H., Klein, D., Humpeno Kriegler, E., 2014. Land-use transition for bioenergy and climate stabilization: model comparison of drivers, impacts and interactions with other land use based mitigation options. Clim. Change 123 (3:4), 495e509. Riveira, I.S., Maseda, R.F., 2006. A review of rural land use planning models. Environ. Plan. B Plan. Des. 33, 165e183. Roy, P., Nei, D., Orikasa, T., Xu, Q., Okadome, H., Nakamura, N., Shiina, T., 2009. A review of life cycle assessment (LCA) on some food products. J. Food Eng. 90, 1e10. Ruggeri, L., Cadena, E., Martinez-Blanco, J., Gasol, C., Rieradevall, J., Gabarrell, X., 2009. Recovery of organic wastes in the Spanish wine industry: technical,

economic and environmental analyses of the composting process. J. Clean. Prod. 17 (9), 830e838. Seguin, A., 2011. How could forest trees play an important role as feedstock for bioenergy production? Curr. Opin. Environ. Sustain. 3, 90e94. Sheerr, S., McNeely, J., 2008. Biodiversity conservation and agricultural sustainability: towards a new paradigm of eco-agriculture landscapes. Phil. Trans. R. Soc. B 363, 477e494. Singh, A., 2012. An overview of the optimization modelling applications. J. Hydrol. 466e467, 167e182. Singh, A., Panda, S.N., 2012. Development and application of an optimization model for the maximization of net agricultural return. Agric. Water Manag. 115, 267e275. Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H., Kumar, P., McCarl, B., Ogle, S., O'Mara, F., Rice, C., Scholes, B., Sirotenko, O., Howden, M., McAllister, T., Genxing, P., Romanenkov, V., Schneider, U., Towprayoon, S., 2007. Policy and technological constraints to implementation of greenhouse gas mitigation options in agriculture. Agric. Ecosyst. Environ. 118, 6e28. Soltani, A., Hoogenboom, G., 2007. Assessing crop management options with crop simulation models based on generated weather data. Field Crops Res. 103, 198e207. Soussana, J.-F., 2014. Research priorities for sustainable agro-food systems and life cycle assessment. J. Clean. Prod. 73, 19e23. Spiertz, H., 2010. Food production, crops and sustainability: restoring confidence in science and technology. Curr. Opin. Environ. Sustain. 2, 439e443. Stern, N., 2006. The Stern Review: the Economics of Climate Change. HM Treasury, London. Strengers, B.J., Van Minnen, J., Eickhout, B., 2008. The role of carbon plantations in mitigating climate change: potentials and costs. Clim. Change 88, 343e366. Tilman, D., Cassman, G., Matson, P., Naylor, R., Polasky, S., 2002. Agricultural sustainability and intensive production practices. Nature 418, 2002. Tukker, A., Emmert, S., Charter, M., Vezzoli, C., Sto, E., Andersen, M.M., Geerken, T., Tischner, U., Lahlou, S., 2008. Fostering change to sustainable consumption and production: an evidence based view. J. Clean. Prod. 16, 1218e1225. United Nation Environment Programme (UNEP), 2009. Ecosystem Management: Part of the Climate Change Solution. UNEP Research Brief. Van der Werf, H.M.G., Garnett, T., Corson, M., Hayashi, K., Huisingh, D., Cederberg, C., 2014. Toward eco-efficient agriculture and food systems: theory, praxis and future challenges. J. Clean. Prod. 73, 1e9. Van Ittersum, M.K., Leffelaar, P.A., Van Keulen, H., Kropff, M.J., Bastiaans, L., Goudriaan, J., 2003. On approaches and applications of the Wageningen crop models. Eur. J. Agron 18, 201e234. Varga, M., Kuehr, R., 2007. Integrative approaches towards zero emissions regional planning: synergies of concepts. J. Clean. Prod. 15, 1373e1381. Virtanen, Y., Kurppa, S., Saarinen, M., Katajajuuri, J., Usva, K., Maenpaa, I., Makela, J., Gronroos, J., Nissinen, A., 2011. Carbon footprint of food e approaches from national input-output statistics and a LCA of a food portion. J. Clean. Prod. 19, 1849e1856. Weber, C.L., Matthews, H.S., 2008. Food-miles and the relative climate impacts of food choices in the United States. Environ. Sci. Technol. 42 (10), 3508e3513. West, P., Gibb, H., Monfreda, C., Wagner, J., Barford, C., Carpenter, S., Foley, J., 2010. Trading carbon for food: global comparison of carbon stocks vs. crop yields on agricultural land. Proc. Natl. Acad. Sci. U. S. A. (PNAS) 107 (46), 19645e19648. Witlox, F., 2005. Expert systems in land-use planning: an overview. Expert Syst. Appl. 29, 437e445. World Bank, 2011. Rising Global Interest in Farmland: Can it Yield Sustainable and Equitable?. Cured by Deininger, K., Byerlee, D., Lindsay, J., Norton, A., Selod, H., Stickler, M. ISBN: 978-0-8213-8592-0. Yang, C.-C., Chang, L.C., Chen, C.S., Yeh, M.S., 2009. Multi-objective planning for conjunctive use of surface and subsurface water using genetic algorithm and dynamics programming. Water Resour. Manag. 23, 417e437. Zeng, X., Kang, S., Li, F., Zhang, L., Guo, P., 2010. Fuzzy multi-objective linear programming applying to crop area planning. Agric. Water Manag. 98, 134e142.

Please cite this article in press as: Accorsi, R., et al., The land-network problem: ecosystem carbon balance in planning sustainable agro-food supply chains, Journal of Cleaner Production (2015), http://dx.doi.org/10.1016/j.jclepro.2015.06.082