Extending the scope of eco -labelling in the food industry to drive [629296]

Journal of Environmental
Management
Manuscript Draft

Manuscript Number: JEMA -D-16-03948

Title: Extending the scope of eco -labelling in the food industry to drive
change beyond sustainable agriculture practices

Article Type: SI:Impact & decision

Keywords: Green supply chain network design; food industry; eco -labels;
organic labels; multi -objective optimization

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 Extending the scope of eco -labelling in the food industry to drive change beyond sustainable
agriculture practices
Marco A. Miranda -Ackermana* & Catherine Azzaro -Pantelb
a Cátedra CONACYT -El Colegio de Michoacán, Sede La Piedad: Cerro de Nahuatzen 85, Fracc.
Jardines del Cerro Grande, C.P. 59370, La Piedad, Michoacán, México
b Université de Toulouse, Laboratoire de Génie Chimique U.M.R. 5503 CNRS/INP/UPS. 4 allée
Emile Monso, 31432 Toulouse Cedex 4, France
Abstract:
New consumer awareness is shifting industry towards more sustainable practices, creating a virtuous
cycle between producers and consumers enabled by eco -labelling . Eco -labelling informs consumers of
specific characterist ics of products and has been used to market greener products. Eco -labelling in the
food industry has yet been mostly focused on promoting organic farming, limiting the scope to the
agricultural stage of the supply chain, while carbon labelling informs on t he carbon footprint
throughout the life cycle of the product. These labelling strategies help value products in the eyes of
the consumer. Because of this, decision makers are motivated to adopt more sustainable models. In the
food industry , this has led to important environmental impact improvements at the agricultural stage,
while most other stages in the Food Supply Chain (FSC) have continued to be designed inefficiently.
The objective of this work is to define a framework showing how carbon labelling can be integrated
into the design process of the FSC. For this purpose, the concept of Green Supply Chain Network
Design (GSCND) focusing on the strategic decision making for location and allocation of resources
and production capacity is developed consideri ng operational, financial and environmental (CO 2
emissions) issues along key stages in the product life cycle. A multi -objective optimization strategy
implemented by use of a genetic algorithm is applied to a case study on orange juice production.
The results show that the consideration of CO 2 emission minimization as an objective function during
the GSCND process together with techno -economic criteria produces improved FSC environmental
performance compared to both organic and conventional orange juice production. Typical results thus
highlight the importance that carbon emissions optimization and labelling may have to improve FSC
beyond organic labelling . Finally, CO 2 emission -oriented labelling could be an important tool to
improve the effects eco -labelling has on food product environmental impact going forward.
Highlights:
 A framework for the design process of a food supply chain (FSC) that takes into account
carbon labelling is proposed
 A multi -objective optimization strategy is implemented by use of a genetic algorithm for a an
orange juice cluster
 CO 2 emission minimization and techno -economic criteria are considered simultaneously
during the GSCND process
 The FSC environmental performance is improved compared to both organic and conventional
orange juice production
Keywords:
Green supply chain network design, food industry, eco -labels, organic labels, multi -objective
optimization
Manuscript
Click here to view linked References

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 1. Introduction
New consumer awareness and behavio ur favouring greener products and services is shifting industry
towards more environmentally sustainable production systems. Eco-labelling influences the market
force of consumer by incentivising the producer to provide greener products that consumers value
differently than conventional ones .. Eco -labelling is a means to inform consumers of specific
characteristics of products and has been used to target how client preferences for greener products
change the value of a product based on the green attribute . It has somewhat recently been used to
intro duce information on environmental performance of products and the production systems they
come from in more detail. Depending on the product and key environmental damage , product eco –
labels inform the consumer on measures taken by the producers to minimize environmental impact .
One example of a product would be paper coming from a managed forest, in the case of a service ,
airlines market carbon emissions offsetting services as an added service to transport (i.e. planting a
tree with your flight) . One type of eco -label that has gain traction is the organic eco-labels for food
products. This type of labelling focuses on promoting organic farming, mainly targeting the
agricultural stage of the food supply chain. A second one is carbon labelling that is used to inform
consumer of the carbon footprint produced due to the production and consumption of products and
services. These two labelling strategies help consumers and producers set the value of the products in
a different way than that adopted for conventional products. By using this strategy , producers are
incentivized to adopt innovative and more sustainable practices in order to gain access to these
consumer markets . The effects on overall performance of supply chains have just started to be stud ied
(Beske et al., 2014; Brindley and Oxborrow, 2014) .
Agrofood supply chains have all the stages and characteristics of any consumer product supply chain.
It is made up of suppliers, focal companies, clients, distribution routes and centres. Key differences are
that the products are consumed by humans and animals, and that the raw materials are grown through
agricultural practices and land use. But while many supply chains for different products may be
studied and improved , in order t o use eco -labelling strategies, food products have restrictions.
Depending on the region or country, these restrictions focus on different aspects of the

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 product/production life cycle of food products. In a globalized economy many food products are
global ly sources and processed. This is due to many reasons, one key issue is the environmental
conditions that allow for the efficient production of some food products. Favourable climates for some
cultures are limited to specific regions of the globe. This in term makes the agrofood supply chain one
that is globally distributed , where many steps for getting food from orchard to plate are not only
related to agriculture but also to processing, manufacturing and transportation.
Environmental impact of agrofood p roduction is thus not limited to the initial stage of production,
where organic labelling applies a market pressure for improvement, but also extends to stages farther
downstream. Green supply chain management paradigm provides a framework to study the ful l life
cycle of product or service and integrates operational, economic and environmental indicators , with
the aim at improving overall performance. In particular Green Supply Chain Network Design is a
process which facilitates strategic decision making on issues related to the location, installation, and
allocation of resources and production capacity, through the scope of GSCM paradigm (Eskandarpour
et al., 2015) . Through this scope , measurement on, for example, CO 2 emissions along key stages in the
life cycle of a product can be captured and integrated into a decision framework. This allows the
decision makers (e.g. managers and executives) to improve performance and allows for the use of eco –
labelling strategies targeting demanding consumers. It allows the marketing departments to take
advantage of new consumer awareness (e.g. consum ers having a good idea of what “CO 2
emissions/unit of product” means) in order to differentiate and add value in a commodity driven
market.
This paper presents the development and deployment of a GSCND strategy that targets economic and
environmental objec tives through a Multiobjective Optimization formulation and solved through the
use of Genetic Algorithms. The approach is applied to an orange juice supply chain case study. The
finding of the study shows that Organic labelling can be complemented with Carbon labelling in order
to improve key emissions hotspots outside of the scope of Organic labelling. The results of the
optimized results of the supply chain network are compared with those of s ome reference values taken
from LCA studies on conventional and organic orange juice supply chains.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
2. Background
Eco-labels are defined by the International Organization for Standardization (ISO) as: “… a voluntary
method of environmental performance certi fication and labelling that is practiced around the world.
An "ecolabel" is a label which identifies overall, proven environmental preference of a product or
service within a specific product/service category”. The goal of Eco -labelling is to promote
susta inability managed production and consumption , categorize d in three types:
Type I – a voluntary, multiple -criteria based, third party program that awards a license that authorizes
the use of environmental labels on products in dicating overall environmental preference of a product
within a particular product category based on life cycle considerations.
Type II – informative environmental self -declaration claims
Type II I – voluntary program that provide quantified environmental data of a product, under pre -set
categories of parameters set by a qualified third party and based on life cycle assessment, and verified
by that or another qualified third party.
The scope of th e case study (that is presented further down) is geographically defined by the regions
that make up the SC, mainly the raw materials sourcing region (i.e. Mexico and Brazil in Latin
America) and consuming regions (i.e. France and Germany in the European Un ion). As eco -labelling
is intended to inform the consumer – the marketed region is the determinant in what labelling policies
apply. In the case study these fall within the European Union (EU) policy structure.
In the EU there is a distinction between E co-labels and Organic labels. The EU Eco -label scheme was
launched in 1992 to promote the production and consumption of products that have a reduced
environmental impact in comparison to existing products on the market. Through these labels
transparency, r eliability and scientific credibility is guaranteed to the customer without the need of any
technical understanding to interpret the label. This allows the consumer to make environmentally
friendly choices when purchasing products, and by this , promoting the product providers to adhere to
this standard to maintain competitive stands . While Eco -labels (under the EU definition) can be

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 applied to different product groups (e.g. cosmetics, hygiene, cleaning, clothing, paints, electronics
equipment, building ma terials, household appliances, etc.) it does not apply to food and feed products.
According to the EU Eco -label website referencing a report commissioned by the EU (Oakdene
Hollins Research and Consulting, 2011) on the feasibility of developing Eco -label for food and feed
products with very interesting and important conclusions.
“…the Commission is not intending to develop Ecolabe l criteria for food and feed products at
this time. The Commission could, however, revisit this question at some point in the future
considering the possible role of the EU Ecolabel within the framework of the development of
any wider EU food strategy, in particular in light of developments in methodologies, and other
tools, for measuring the environmental impact (including by, for example, environmental
footprinting) of products.”
Two main points are to be noted from this statement. First and foremost is t hat food products are
outside of the scope of Eco -labels in the EU under their definition . The second is that, this could
change, and there is a suggestion of taking (organizational) environmental footprinting (OEF) as a
candidate strategy.
In (Pelletier e t al., 2013) , some OEFs are compared in terms of four criteria that define the European
Commission Organization Environmental Footprinting (EC OEF) scheme. These are: (1) multi –
criteria, (2) life cycle -based approach that considers all organizational an d related activities across the
supply chain, (3) provides for reproducibility and comparability over flexibility, and (4) ensures
physically realistic modelling. According to (Pelletier et al., 2013) only Global Resource In itiative
(GRI) takes a broad scope of environmental impact, and states that all other methods (included in the
study: ISO 14064, Greenhouse Gas Protocol, Bilan Carbone ADEME 2007, DEFRA 2009, Carbon
Disclosure Project and Global Reporting Initiative) refer to single impact categories. While the EC
OEF proposes a multiple criteria approach.
This paper assumes the possibility of the inclusion of Food category within the EU Eco -label scheme.,
To illustrate the incorporation of the multiple criteria approach pr oposed in EC OEF the case study is

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 presented with to product streams Organic and Carbon footprint minimized orange juice . In addition,
we adhere to the current trend to use single impact categories (i.e. GWP), as the environmental
footprinting category wit hin the context of the EU Commission prospective on eco -labelling and the
GSCM paradigm .
2.1. Organic production
Indeed, environmental labe lling does exist in the EU for food products under the category of Organic
or Biological product labelling. The EU Organic Certification label scheme has the aim at improving
environmental impact of production and consumption of agricultural products. According to “EC
Council regulation No 834/2007 on organic production and labelling of organic products a nd
repealing Regulation ‘EEC) No 20029/91 ” states that “Organic production is an overall system of
farm management and food production that combi nes best environmental practice , …and a production
method in line with the preference of certain consumers for products produced using natural
substances and processes. ”(EC Council regulation No 834/2007 -Art. 1 ). And goes on to define
Organic Production as “… the use of the production method compliant with the rules established in
this Regulation, at all stages of production, preparation and distribution ……and including its storage,
processing, transport, sale or supply to the final consumer, and where relevant labelling, advertising,
import, export and subcontracting activities ” (EC Council regulation No 834/2007 -Art.2)
Continues to add specification applied to processing of Organic Food stating “… the production of
processed organic food shall be based on … organic agricultural ingredients … the restriction of the
use of food additi ves, of non -organic ingredients… the exclusion of substances and processing
methods that might be misleading regarding the true nature of the product… the processing of food
with care …”(EC Council regulation No 834/2007 -Art.6 )
Following the principals and norms presented before, this s tudy assume s all necessary requirements to
access Organic certification are achieved in Organic product flow in the simulated case study. Herein
the term Eco -label relates to the achievement and use of the Organic certification label that uses the
logos il lustrated for Germany and France in Figure 1 .

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
FIGURE 1 ORGANIC CERTFICATION ECO-LABEL FOR EU COUNTRI ES GERMANY AND FRANC E
2.2. GSCND
Supply Chains are viewed as networks of elements that involve suppliers, manufacturers, distributors
among other stakeholders and reflect materials, information and economic flows. They are physically
constructed of natural resource extraction facilities, processing facilities, manufacturing plants, trucks,
sea vessels, warehouses, etc …that are located in different locations around the world. Supply Chain
Network Design (SCND) involves a decision and model framework that searches “through one or a
variety of metrics, for the “best” configuration and operation of all of these (SC network) elements.”
(Garcia and You, 2015) .
Some of the most important challenges that SCND holds reflect the issues that complex real systems
face including for example decisions at multiple scales, multiple levels, multiple periods, multiple
objectives and undoubtedly multiple stakeholders.
SCND consists in formulating the SC network as nodes and arcs that connect, featured in layers for
each echelon that construct the SC of interest ( illustrated in Figure s 2). In each layer , differe nt
alternatives are presented that can represent differences in modes of transport, technologies used,
geographical locations of sites, among many other choices, while the arcs may represent attributes and
criteria of interest such as distances, costs, tim e periods, etc. The process of optimizing the SCND is to
find the best configuration of the network, this is to say, the best route of arcs and nodes that fulfil the
single or multiple objectives that are of interest to the decision maker.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
FIGURE 2 SUPPLY CHAIN NETWORK EXAMPLE

3. Problem definition
The research question is formulated in terms of the case study:
1. Is Organic certification a sufficient driver to minimize the environmental footprint of
agrofood taking into account the full life cycle i.e. from raw material to delivery of end
product to market?
2. Developing an optimal design approach for an orange juice supply chain network in order to
minimize CO 2 emission at all the stages of the supply chain, wi th two product streams, one for
organic and one for conventional – can overall performance be equal of better than that of a SC
that only targets the Organic label objectives?
In order to answer these questions a mathematical model reflecting a globally di stributed orange juice
supply chain is formulated with the following variables to determine the following items :
– Design of production and transformation systems
• Agro practice selection
• Technology/equipment selection
– Composition Production « Mix »:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 • Organic vs. conventional
• From concentrate or not?
– Location and allocation:
• Where to install the transformation units?
• In which region/country?
– Supplying:
• Number of suppliers (contract farming policy)?
• Which supplier to select?
– Sales price policy
• Price fixing
• Premium or not?
Historical and published data are used to define operational, economic and environmental parameters
and variables. The model is designed and solved as a multi -objective optimization problem in order to
find trade -off solutions from antagonistic objectives (Miranda -Ackerman et al., 2014) . The two main
objectives targeted are Net Present Value (NPV), reflecting the preferences and objectives of the
compa ny, and Global Warming Potential (GWP in kg CO 2-eq emissions), reflecting the preferences of
society and the environment. The model is solved through a Genetic Algorithm approach in order to
obtain the best compromise solutions. Different scenarios and co nfiguration are optimized and
analy sed in order to highlight the performance of the supply chain in terms of the criteria being studied
in relation to the research questions. The results are compared in terms of CO 2-eq emission to those of
the current sci entific literature .
4. Case study
Before going further presenting the different optimization schemes, let us define the key elements of
the case study. It illustrates a globally distributed orange juice supply chain as represented in Figure 3.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65

Energy
Water
Energy
Water
Organic
juice
Organic
orchards
Pasteurization
Concentration
Bulk
transportation
Bottling
Bottling
Bulk
transportation
Final product
transportation

Conventional
orchards

Organic orange
Conventional
orange
Conventional
juice
Final product
transportation
Organic
concentrate
Conventional
concentrate
Organic
juice
Conventional
juice
Organic
concentra
te
Eco-labelled
bottled FCOJ
Conventional label
bottled FCOJ
Conventional
concentrate
Conventional
juice
Organic
juice
Conventional
bottled NFCOJ
Eco-label bottled
NFCOJ
Electricity
Methane
Electricity
Methane
Fuel
Electricity
Water
Container
Fuel
Electricity
Water
Container
Fuel
Fuel

Fertilizers
Pesticides
Fertilizers
Pesticides FIGURE 3 MATERIALS AND RESOUR CE FLOWCHART FOR THE CASE STUDY
The Focal Company that manages this chain needs to select a project to increase capacity. In the SCM
paradigm as in the GrSCM, a central or focal company (FC) as p roposed in (Seuring and Muller,
2008) is characterised by being the designer or owner of the product or service offered, governing the
supply chain, and having contact with all SC stakeholders including the customers. The FC can also
some times be the processing or manufacturing company, as in the case study.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 The FC is considered to be the integrator firm within the context of contract farming as described by
(Rehber, 2000) , under a Management and Income Guaranteeing contract, (Richard and Koh ls, 1998)
also known as Production management contract (PMC) (Minot, 1986) .
The potential market demand is assumed to be known. The main assumptions are the following ones:
1. Two potential raw material supplying re gions are considered, i.e., Mexico and Brazil, to meet
raw material requirements.
2. Only one region has to be selected, from which a set of suppliers are contracted in order to
satisfy the capacity level as required by the demand and the quality of oranges .
3. The oranges will be processed at a plant located near the supplier. A selection of technologies
and capacities has to be carried out to best satisfy market needs.
4. The final products are of four types, combining the label attribute ( organic label led1 and
conventionally labe lled) and the processing attribute (from concentrate and not from
concentrate).
5. The market target is composed of ten principal cities in two countries (France and Germany).
6. A set of 6 potential sites to locate a bottling/distribu tion site for each country is considered.
The decisions under the scope of the modelling and optimization framework can be synthetized in the
following :
Supplier: Raw material selection, supplier region selection, supplier selection, agro practice selection,
land surface to be contracted (agricultural production capacity).
Transformation: Selection of bacteriological stabilization technology (i.e; pasteurization e quivalent,
such as Pulse Electric Field or High hydrostatic pressure processing), Concentration technology
selection (e.g. evaporators, freeze concentration, reverse osmosis).
Packaging/bottling : Bottling plant location, bottling technology selection (e.g. glass bottles, carton,
PET containers).

1 Recalling that eco -label is used to refer to product that can hold the “organic” or “bio” certification labelling
under the EU regulation on food and beverage labelling , except where explicitl y stated not to.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 Market: Demand coverage (i.e. market to be covered by planed production capacity), product mix (i.e.
quantity of each type of product based on organic or conventional raw materials use, and if
concentration or not f rom concentrate juice is bottled).
The parameter values use d for this case study, which can be found in (Miranda -Ackerman, 2015) , are
taken from relevant literature and adapted to this example.
5. Methods and tools
The solution approach, is based on the coupling of Multi -objective Genetic Algorithms (Dietz et al.,
2006; Gomez et al., 2010) and Multi ple C riteria Decision Making (MCDM) (Ho et al., 2010) to model
the complex supply chain system based on interconnected networks from suppliers to consumers ( see
Figure 4). The solution strategy is flexible enough to allow the mode ller to evaluate different strategies
based on the specification of the food system under consideration.

FIGURE 4 MOO AND MCD M WORKFLOW DIAGRAM
The choice of an evolutionary algorithm (EA) as a multi -objective optimization procedure is mainly
influenced by the following items that make them preferable over classical optimization strategies: a)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 considerations for convexity, concavity, and/or continuity of functions are not necessary in EAs; b)
their potential of finding multiple Pareto -optimal solutions in a single simulation run ; c) Nonlinear
constraints and criteria can be tackled by such algorithms; d) they are known to be efficient to tackle
combinatorial problems. In the supply chain design problem encountered in this work, integer
variables are consid ered representing the decisional choices relative to the existence or absence of a
node in the network as well as the operational variables of the supply chain.
The use of NSGA -II as the stochastic search algorithm is thus justified. Table 1 summarizes the values
used for the parameters required by the algorithm . They are fixed based on both empirical trial -and-
error experience and on the sensitivity analysis that is not detailed here (Dietz et al., 2006) . In this
study a set of scenarios are described in detail and analysed in order to evaluate different modelling
strategies, because of this different parameters are used when using the GA. The higher number of
individuals in the population associated with a higher number of generations used for scenario 1
compared to that used for scenarios 2 -6 (i.e. a double value) is used to overcome the difficulties
encountered in stochastic search method s involving equality constraints. It must be highlighted that a
relatively high value for mutation rate (i.e. 0.5) was adopted which can be considered inconsistent
compared to what occurs in natural evolution. This phenomenon was already observed in mixed
integer problems similar to the pure integer problem treated in this work (Dietz et al., 2006; Gomez et
al., 2010) .
TABLE 1 PARAMETER SET FOR MU LTI-OBJECTIVE GA
Scenario 1 Scenario 2 -6
Population size 200 400
Number of generations 400 800
Cross -over rate 0.9 0.9
Mutation rate 0.5 0.5

The results are presented in terms of the Pareto front solutions produced by the GA.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 At the final step of the strategy, a multiple criteria decision making (MCDM strategy ) provides a way
to find a solution in the diversity of the solution space represented by the Pareto front. It allows the
decision maker to rank solutions with the flexibility to reflect different values and preferences among
the best solutions that were id entified by the optimization procedure. In this work, the M -TOPSIS
method (Ren et al., 20 10) is used. It has a set of weight parameters that can be used to assign
importance to each criterion. Unless explicitly mentioned, the same weight is allocated to each
criterion. It must be yet highlighted that different values can also be used reflec ting the preference of a
stakeholder in real world decision -making environment. All the optimization strategies that are
proposed are carried out following the Life Cycle Optimization process following the guidelines
proposed in (Yue et al., 2014) and in (Ouattara et al., 2012) .
5.1. Optimization Model
The model provides a means to represent the behavio ur of the food supply chain. The
mathematical formulation of the supply chain model takes into account materials flows and demand
satisfaction. The model ling strategy is used as a generalized model for each scenario instance (see
section 5.1.6 ). It is conceptually constructed in three sets of constraints that are described in what
follows.
5.1.1. Mass Balance and Demand Constraints
Materials flow throughout the network of suppliers, production plants and markets are reflected in a
subset of constraints so tha t production capacities at each level in the supply chain can meet market
demand requirements. The amount of raw materials from the Supplier Echelon interfaces as the input
for the Processing Echelon, this last itself interfacing with the Market Echelon . The quantity of final
product is restricted to be equal or higher than the targeted demand quantity.
5.1.2. Operational and Economic Functions
Using variable values on materials quantities and requirements along with parameters related to costs
at different sta ges in the supply chain model, a subset of operational and economic functions are
constructed. The goal is to reflect the relationship between operational capacities, economic costs and

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 market demand drivers. These integrate into the economic objective fun ction of the multiobjective
optimization model.
In this subset of functions , a key component is the Bottling Echelon that interfaces with market
demand , with a range of markets that can be satisfied with different types of products based on
processing attributes (i.e. from concentrate orange juice vs Non from concentrate orange juice) and by
raw materials attribute (i.e. organic oranges vs conventional oranges as raw materials). Each type of
product has a different price based on those attributes, and influence s the results obtained both in
terms of the objectives and the decision variables results.
5.1.3. Environmental Impact Functions
The same basic modelling structure is used for the definition of the environmental impact functions as
of that of operational and economic functions , reflecting the e nvironmental impact in global warming
potential as expressed in kgCO 2eq/kg based on the parameters and decision var iable values .
This part of the model feeds directly into the environmental impact objective function, and captures
the effects of making difference design choices related to location, allocation of resources,
technologies being used, and market demands th at are targeted, among many other choices reflected
by the model (for example , choosing different types of equipment with different energy inputs (e.g.
gas, electricity) , selecting the country or region for the installation of processing or bottling plants with
some countries produc ing electricity from less pollutant sources of energy than others ).
5.1.4. Transportation Functions
The transportation activities involved through the supply chain have an economic and environmental
cost. The four intermediate product types, i.e., pasteurized single strength (NFCOJ) organic and
conventional orange juice, and concentrated multiple streng th (FCOJ) organic and conventional
orange juice differ from their production cost, related to their operations but share the same
transportation cost in terms of kilogram.kilometer (kg.km) per mode of transport. These intermediate
products are transported in bulk by different modes and route; for our case study, transport is limited
to sea freight transport from raw materials production region to consumer regions . Within each market

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 region, a set of markets (10 in the case study) made up of the most popul ated cities. Economic costs
are reflected in the different quantities sold and different regions chosen given the quantities and
distances (i.e. $/kg.km); In a similar way environmental impact of each transport trajectory measured
in kg of CO 2 eq / kg.k re flects the effect of choosing one supply chain configuration over another
among other decisions that affect performance and the factors related to transportation .
It is through these functions that the effects of choices related to location and allocation of
resources and production targets that define the environmental and economic costs of transport are
reflected onto both economic and environmental objective functions.
5.1.5. Objective Functions
In order to evaluate the performance of the supply chain netwo rk, different criteria are developed.
Initially one needs to empirically or through an “objectives and preferences study” choose a set of
criteria of interest, which reflect the economic and environmental performance of the SC. The model
considers four pos sible objectives : NPV, GWP, average VUC and I.
5.1.5.1. Net Present Value (NPV) and Investment (I)
One of the most widely used KPIs is the net present value (NPV) of a project. The advantage of this
indicator is that it looks at the long -term plan taking into cons ideration the effect of time. Additionally,
it considers the operational and the fixed capital cost within a single framework in contrast to single
facets of a project such as Sales Revenue, Project Cost, among others KPIs. Investment is reflect ed by
the equipment cost via Lang factor (fL) for the type of production system .
In order to calculate the NPV and Investment output , a number of variables need to be defined and/or
calculates such as: sales revenue, sales price (SP), variable unit cost, sales m argin, etc. that are
reflected by the constraints and functions that conform the model.
5.1.5.2. Global Warming Potential (GWP)
Concurrently the environmental impact measurements are also determined for each optimization
instance. GWP is used and defined as the sum of the environmental impact output per unit given the
type of product and market to which it is transported to (i.e. each of the 20 market destinations

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 demanding the 4 types of products, 80 unique Unit Environmental Impact ) times the number of
product s produced to cover each demand .
5.1.5.3. Average Variable unit cost (AVUC)
AVUC is defined by th e sum of the product of each variable unit cost times the quantity that is
produced ( total planed production ) for each type of product given its label (i.e. organic or
conventional) , fabrication steps (i.e. extraction, concentration, bottling) and the marketed production
output planned for all products to all markets gives the average variable cost.
5.1.6. Generalized MOO model
In summ ary the generalized form of the multiobje ctive optimization model is used to describe each
instance of the different scenarios that where studies. It is formulated to capture the interrelation
between the decisions variables, the model variables and the parameters that describe th at system as a
set of restrictions and their influence on the objective functions results, as follows:
min [f 1(x,y), f 2(x,y), …,f n(x,y)] ; s.t. g(x,y) ≤0 ; h(x,y) = 0 ; x ∈ Zn, y ∈ {0,1}m Equation 1

This formulation involves a set of objective functions (f) from 1 to n to minimize, subject to a set of
inequality constraints (g), a set of equality constraints (h), and the variables are defined as (x) for
integer and (y) for binary.
6. Scenario analysis

The approach that was developed to model the Agrofood Supply Chain (SC), takes into account the
different perspectives and preferences of the principal stakeholders, mainly suppliers, focal company,
customers and natural environment. This approach to supply chain does not consider a chain of
businesses with one-to-one, business -to-business relationships, but a network of multiple businesses
and relationships. The Green Supply Chain Network Design (GrSCND) approach allows the mode ller
to use different techniques to formulate, experiment, evaluate and analy se the types of problems that
are related to the supply chain issue. Different optimization strategies, based on the supply chain

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 design model, where applied following different scenarios that reflect the specific targets of the
interconnected stakeholders. Fo r this purpose, three main optimization strategies are proposed:
(1) Sequential Optimization Scheme, involving a two -stage optimization process first reflecting
customer aims for cheaper and more environmentally friendly product, and then followed by
company’s aims related to profitability and environmental performance using the breakeven
point deduced from the first step.
(2) Concurrent Optimization Scheme, based on an integrated optimization where the objectives of
the main stakeholder are simultaneously optimize d, in order to find SC networks that produce
environmentally friendly and profitable products.
(3) Differentiated -Product Optimization Scheme encompassing an integrated optimization
approach that similarly considers not only the main stakeholders’ objectives, but also the
added value of organic eco-labels and the sales price of the final products.
Each optimization scheme was evaluated evaluating different configurations of the model: objectives
and restrictions. The set of scenarios evaluated is presented in Table 2 .
6.1. Scenario Results
The results are also summarized in Table 2 with key formulation differences in the Model column and
the results are presented with a description of the findings for each scenario in the Description & Key
results column.
TABLE 1 SUMMARY OF SCENARIOS USED TO EVALUATE MOD ELING CONFIGURATIONS
Optimization
Scheme Scenario Model Description & Key
results
Sequential (Sc1)

Fixing NPV to zero to
find minimum Variable
Unit Cost at lowest GWP
output in order to reflect
the customers’
preference; also used to
estimate a base Sales
Price to be used in other
Scenarios.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 (Sc2)

Integrating fixed Sales
Price for all products to
the value found in
Scenario 1 while
maximizing NPV and
minimizing global GWP.
Used as a baseline
model.
(Sc3)

Adding the Investment
cost as a minimization
objective function to
consider a second
economic criterion to
favour project initiation
phase.
Sc 3 produces the best
trade -off results yet.
(Sc4)

Poor performing
solutions compared to
scenarios 2 and 3.
Concurrent (Sc 5)

The objective functions
used is the same as for
Scenario 3. Sales Price is
a variable dependent on
the AVUC for each
product, with a 25% sale
margin. The outco me
was poor, compared to
the solution found using
the Sequential
Optimization Scheme.
(Sc 6)

The objective functions
evaluated are the same as
in Scenario 4. The same
variable sales price
policy used for Scenario
5 is used. No
improvement from
Scenario 5 was achieved.
Differentiated –
Product (Sc 7)

Objective functions
focus on investment. A
sales price premium is
attributed to organic
products. NPV outcome
is greatly improved,
while AVUC reaches
comparable values with

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 Sc6.
(Sc 8)

Same pricing strategy as
on Sc7. NPV and
Investment perform
similarly to Sc7. AVUC
is not significantly
improved unexpectedly.

Aggregating the results obtained in each scenario gives a clearer understanding of the advantages and
disadvantages of each strategy. The M -TOPSIS solution for each scenario is considered in Figure 5
where the values for the main objective functions, main ly NPV, GWP, AVUC and Investment are
indicated. The left Y axis is used to measure the NPV (blue bar), GWP (red bar) and Investment
(purple line with exes); the right Y axis is used to measure the average AVUC (green line with
triangles). Looking only at t he NPV bars, Sc8 is the best performing. The worst performing scenario in
terms of NPV is Sc5. Looking only at GWP, Sc7 is the best performing, while Sc6 is the poorest. In
terms of AVUC Sc3 followed closely by Sc4 is the best performing, while Sc7 is th e worst. The main
idea to take away from these observations is that the results are mixed and a clear trade -off solution is
not evident. The most promising solution strategy is Sc3 that provides a compromise between all three
criteria while finding the bes t AVUC values overall. A second important observation that can be made
is that, even though the scenarios are performed under different conditions, there is a clear relation
between the Investment cost and the Variable Unit Cost. Scenarios 2, 3 and 4 perfo rm better in
relation to AVUC, but have higher Investment costs; and the contrary is true for scenarios 5, 6, 7 and
8. These last four clearly have much higher AVUC costs.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
FIGURE 5 RESULTS FOR THE MAIN CRITERIA PER SCENARI O
In the current literature, th e most common approach to multi -objective GSC ND is to only consider
NPV and GWP as the objective functions. We will follow this approach in order to illustrate the main
finding and emphasis on the goal of the study.
Focusing more attention on the environmental issue, Figure 6 presents the environmental impact
measuring GWP in kg CO 2 equivalent emissions per litre of juice (y -axis) – assigned by product type
and market region (x -axis). Eight reference values taken from related literature on life cycle
assessment of orange juice production are also shown.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
FIGURE 6 GWP PER UNIT SUMMARY DEVIDED BY COUNTRY A ND PRODUCT TYPE WITH REFERENCE VALUES (DO UBLET ET AL. 2013)
Many observations and conclusions can be drawn by comparing the different values and behavio ur
obtained from the scenarios given the mode lling and optimization approach proposed against those
provided by the literature (Beccali et al., 2009; Doublet et al., 2013; Dwivedi et al., 2012 ; Landquist et
al., 2013) . Firstly, comparing the NFC with FC for both Organic and Conventional products one sees

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 that GWP can be higher for NFC that FC. This can be counterintuitive given that less processing is
made to NFC orange juice; this behavio ur is explained through the efficiency lost due to transportation
and “last mile” refrigeration (i.e. bottling plant to market) for NFC orange juice ; it is also partially
explained by the smaller quantity produced given less agrochemicals producing an averag e
environmental impact at the farming system that is not much different for conventional vs organic
farming (Meier et al., 2015) . This phenomenon can be clearly seen for most scenarios for the France
market, as well as, in general for the reference values.
Secondly, Ref.6 proposed by (Knudsen et al., 2011) exhibits the lowest GWP value since it does not
take into account the bottling’s impact. Most other reference values are within 0.6 to 1 kgCO 2 eq/L
slightly higher than that obtained in the case study . The GWP levels obtained with the modeling and
optimization approach are explained by two main factors. The first is that the SC is optimized while
the reference values are based on case studies focused on measuring the SC and not on imp rovement
of its performance. Secondly, it must be yet emphasized that the modeling approach does not entail a
full LCA for each SC network evaluated. It only takes into account the effects of using agrochemicals,
energy and water through out the production and transportation processes, and thus the environmental
impact is lower than that if a detailed LCA is performed. One main observation is that all product
types, no matter the label or processing used on average fall beneath most reference values. In the case
of German region it is clear that scenarios 7 and 8, because they use the price premium for organic
eco-labelled products, have better performing SC network systems in terms of GWP. On average, Sc7
and Sc8 find a trade -off between regions, this is to say, while it is the best performing in the German
market region it is a poor performer in the France market region; but for both regions these scenarios
insure that GWP performance is as good or better than the reference values excluding Ref. 6 that , as
noted before, does not consider the bottling process. By developing the model to this level of detail
and proposing the Differentiated -Product Optimization Scheme , globally environmentally efficient SC
networks can be found.
Lastly, comparing the difference between organic and conventionally labelled products, there is not
much difference between scenarios within each region. This is contrary to popular belief that organic

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 product globally outperform conventionally labelled products. If one where to take s uppliers echelon
in isolation environmental performance may be improved by using less agrochemicals, but in terms of
the global supply chain strategy that is proposed, the agro practice used during raw materials
production (i.e. oranges) is less important than that of the other stages (e.g. processing, transportation,
bottling, etc.). This can be observed through references 7 and 8 that follow the opposite pattern, this is
to say, organic product is outperformed by conventionally labelled product. To furth er illustrate this
phenomenon let us compare the LCA results presented in (Doublet et al., 2013) shown in Figure 7
with an example taken from Sc3 M -TOPSIS best compromise solution for product (of all four types)
destined for Market 1 in Germany shown in Figure 8.
Figure 7 provides a detailed allocation of the sources of GWP emissions throughout the product life
cycle. In addition to the classification provided by the autho r a set of reference clustering through
brackets are proposed. Using this arrangement to more closely resemble the level of detail used for
the case study shown in Figure 8. While the reference LCA does provide more detail by dealing GWP
in terms of more sub process, there is little emphasis on the transportation stages during the products
life cycle. The example taken from Sc3 one sees that the steps are more aggregated but emphasis is
given to the SC echelons and their interfaces. Nonetheless similar dis tribution of the sources of GWP
in the different stages is appreciated. And more importantly, and looking back to the point previously
developed in relation to the effect organic production has over GWP outcome, one sees that for both
LCAs the main source is the bottling process while orange raw material production is far behind.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
FIGURE 7 LCA OUTPUT IN TERMS OF GWP PER KG (~1L) OF ORANGE JUICE IN P ET BOTTLE (DOUBLET E T AL. 2013)

FIGURE 8 LCA OUTPUT IN TERMS OF GWP PER L OF ORAN GE JUICE IN PET BOTT LE FROM SC3 M -TOPSIS SOLUTION MARK ET 1 IN
GERMANY

Bottling
Transport to bottling
Pasteurization (NFCOJ)
Conventional orange production

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 Furthermore one can appreciate the importance of the tranportation stages for the Sc3 example as
being the second most important source of GWP emissions. This is in contrast to the results presented
in the ref erence example that indeed uses a more optimistic approach of modeling transportation. This
leads us to conclude that while most literature in relation to environmental impact focus on measuring
and in evaluating different technique, a more holistic approa ch provides better insight and a way to
take advantage of the scope provided by framing the problem as one of green supply chain network
design. This in term highlighted the need to consider carbon emissions minimized processes and
products, and the potent ial that carbon footprinting and respective ecolabelling could have in order to
improve agrofood supply chain beyond the farming stage.
7. Discussion
During each stage of this research work different questions arose that fell near the edge of the scope of
the work but could not be covered. These questions and observations remain outstanding and could
motivate future research.
Water impact mode lling: it must be highlighted on the one hand that water consumption was
included within the model ling scope for both the Green Supplier Selection Problem and the Green
Supply Chain Design problem; on the other hand, eutrophication and acidification of water were
included as environmental impact criteria in the Green Supplier Selection problem formulation. Yet,
these water centric environmental impact criteria were not included in the Green Supply Chain Design
problem formulation. Furthermore, other important issue s, like irrigation systems, were included in a
very limited way within the scope of the case study. This is not a problem for seasonal agro food
products and agricultural systems that depend on the natural rain fall. But for other food products that
are he avily depend on irrigation systems this issue could require additional attention.
Furthermore , the case study limited the scope of the processing step, excluding the initial washing
stages of production that are pervasive for most fresh fruit and fruit-derived food products. In some
cases , this can be considered negligible or inexistent, but there are cases where water consumption is
very important. Related to this point, another issue is that given that many food production unit
operations are in batch form, cleaning of silos, containers, hoppers, feeders, etc. may also require

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 important quantities of water and cleaning products as well as chemicals , that consume water and
may pollute water runoff. These could also be further detailed in future work depe nding on the focus
and product being studied.
Land use: one very important issue that was considered in the model through the measurement of
yields is the land use . While it was considered directly in the model formulation, its environmental
impact was not quantified nor was included as an explicit optimization objective. Land use and yield
are a very important issue given that food security and demographic growth have justified until now
the rampant change of land use. Deforestation and erosion of many nat ural landscapes that should be
protected must be also considered. A focus on the value obtained by limiting the changes in use of
land could be an important branch of research within the Green Supply Chain domain.
Waste is another issue that fell outside of the scope of this work but is highly related to the objective
being considered. W aste byproducts are produced in different stages in the product lifecycle. In the
first stage a sorting operation is usually necessary for food products, where some residue s or non –
conforming products are discarded. These waste materials can be treated as solid waste (to be
discarded) or could be used by other entities as a raw material. In the developed case study, the
potential to consider the biomass from the extraction and concentration processes as a byproduct for
the production of animal feed (Lanuzza et al., 2014) or more recently biogas (Wikandari et al., 2015)
constitute a pot ential pathway of improvement for supply chain modeling and for product valorization.
This type of reuse of waste materials has been treated in literature in different ways, some of the most
popular ones are Industrial Ecology and the Closed -Loop Supply Ch ain Logistics . These approaches
could be explored as potential additions or extensions of the method here proposed.
The consideration of the packaging materials at the end of life stage could be also taken into account
in more detail considering Reverse Logistics . It relates to the recovery of materials that can be treated
and reused or repurposed. In the case of the food beverage industry bottles are used that can be
recovered. Each country has their own policies in place to sort and recover valuable mat erials and
innovative solutions to recycle and recover unavoidable waste are receiving a lot of attention: for
instance, in the case of Germany for example, plastic and glass bottles are recycled by incentivizing

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 the consumer to sort and bring back the mat erial to places of purchase by paying for the recovery
service. Reverse logistics is not new but had been left aside for many years due to the efficiencies
gained at producing very cheap packaging materials, but has started to become more important given
the renewed awareness of the potential of limiting externalities of food consumption related to
packaging. This could be considered in future work where trajectory vectors could be added to the
network model to accommodate for reverse logistics. This could be very interesting given that the
finding of this work and other research papers show that one of the main contributors to the
environmental impact of non -alcoholic beverages like orange juice comes from the bottle.
Additionally the scope of the work was limited to Greening the supply chain by using the Life Cycle
Assessment method in order to measure and improve the environmental performance of the supply
chain. Recent works have extended it to include the social aspect through the so -called Social Life
Cycle Assessment (SLCA). In this approach, the aspect related to labor, social benefit, job creation,
community development among other things is also measured and targeted for improvement. In this
work, the social element was limited to the collaboration and contract schemes that are proposed in
chapter 3 through Partnership for Sustainability. This could be extended in order to evaluate the social
benefits of decentralizing suppliers, process plants, de -mechanizing processes in order to produce
more jobs for instance in addition to the new social measurements that little by little are starting to be
included within this new SLCA paradigm.
From a methodological point of view, some important perspectives could be incorporated into future
research. The inclu sion of uncertainty into the model framework could be important to overcome
many of the random events and fluctuations inherent in agro food supply chains related to the
volatility of the weather, global markets, consumer behavior among many other uncertai nties. We
collaborated in developing some systems related to this during the PhD work that resulted in two
publications (Fernandez Lambert et al., 2015, 2014) . But it would be interesting to incorporate
uncertainty measurements and variability within the framework presented in this work . Connected to
this issue is also the possibility of the inclusion of a dynamical systems approach where changes that
occur in time could be integrated into the framework such as the yield per tree based on the age of the

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 trees, soil erosion, soil nutrient replenishment, and other time dependent phenomena that could
provide better descriptions of the system in order to make better decisions.
In summary, the contributions from this research have paved the way to extend the base model and
methodology for greening the agro food supply chain and improving the integration of tool s to
overcome the technical challenges of developing future sustainable production systems.
8. Conclusion
Three optimization schemes to the green supply chain network problem were presented. Each has
different advantages and weaknesses. First, Sequential Opti mization Scheme , a base optimization
scenario is carried out to obtain the best solution from the customers’ perspective. This base scenario
is then used to set the Sales Price for each product based on the type (e.g. organic, conventional, FC or
NFC) and market that it will be sold and distributed to in subsequent scenarios (i.e. scenarios 2, 3 &
4). By fixing the Sales Price – the scheme proposes solutions that are evaluated during the GA
optimization process that are competitive in terms of GWP and AVUC (and thus price). In the
subsequent scenarios, different objective functions are used to model the focal company prerogative to
be profitable. Using KPIs such as NPV, investment cost and Variable Unit Cost, the optimization
process is driven to search for solutions that minimize the investment, operations and transport cost
incurred by the focal company during the production and distribution process. By evaluating different
objective functions in each scenario, the Pareto front solutions can be iteratively improved in relative
terms, providing the best set of alternatives to the decision maker.
In the Concurrent Optimization Scheme, different criteria were evaluated simultaneously. The fixed
pricing strategy used in the Sequential Optimization Scheme was cha nged to a variable pricing
strategy. In this scenario, a 25% price margin cost is added to the Variable Unit Cost of the product to
fix the Sales Price (SP). Because no threshold was established for the SP – different solution
alternatives were found. Unex pectedly but justifiably the solutions were dominated by those found in
the Sequential Optimization Scheme for the reasons presented in the result section.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 Lastly, a Differentiated -Product Optimization strategy was evaluated. This approach takes into
account the premium price that a customer is willing to pay for higher quality organic eco-labelled
food products. This is particularly sound because the differentiation helps counteract part of the
additio nal cost that may be incurred when producing products under an environmentally conscious SC
network design. The optimization search process explores solution spaces that would not otherwise be
considered. This approach takes into account the preferences of the consumer by attaching a variable
Sales Price based on the AVUC that is minimized. It also takes the focal company objective into
consideration through the NPV criteria, while being environmentally conscious through the GWP
minimization objective funct ion.
The main finding from this part of the research lies in three main points. First, the method proves to be
not only feasible but efficient at mode lling and finding optimal trade -off solutions that would
otherwise be impossible to find. Secondly, the d ifferent objective functions and pricing strategies that
are proposed and studied, provide insight on the importance of choosing the best approach to agro
food supply chain problems. Indeed, the main contribution was corroborating that, while organic
certification of products in order to add value through eco -labels at the same time as improving
environmental performance is useful, the use of more general eco -labelling that reflects the full supply
chain could be more suitable and effective. In particular, the case study showed in the final results that
the main contributors to one of the main pollution indicators, mainly GWP, come from other stages in
the supply chain, e.g. transportation and bottling. By focusing on the agricultural stages of the supply
chain, important attention that should be directed at these operations is misrepresented in the current
organic eco-labelling policy.
The contribution of this work lies in proposing an integrated and holistic approach to greening the
agrofood cluster supply chain network design process. Through the case study we provided an
illustrative example of its potential use. Furthermore this example allowed us to find insight into the
specific case of the orange juice supply chain. The results show that each step in t he supply chain
holds opportunities to improve environmental performance equal or greater than that of only looking
at the agriculture stage of the food supply chain. Because of this, the application and adaptation of this

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 approach to other food products m ay provide a better design and improvement method for supply
chain practitioners. Finally, a wider more inclusive scheme, such as the one proposed in this work can
be adopted in mainstream industry and consumers in order to promote better and more effectiv e
production systems and greener consumption.
Acknowledgment
The authors would like to thank the Mexican science council CONACYT and ministry of education
SEP for their participation in funding this project.
Bibliography
Beccali, M., Cellura, M., Iudicello, M., Mistretta, M., 2009. Resource Consumption and
Environmental Impacts of the Agrofood Sector: Life Cycle Assessment of Italian Citrus -Based
Products. Environ. Manage. 43, 707 –724. doi:10.1007/s00267 -008-9251 -y
Beske, P., Land, A., Seuring, S., 2014. Sustainable supply chain management practices and dynamic
capabilities in the food industry: A critical analysis of the literature. Int. J. Prod. Econ.,
Sustainable Food Sup ply Chain Management 152, 131 –143. doi:10.1016/j.ijpe.2013.12.026
Brindley, C., Oxborrow, L., 2014. Aligning the sustainable supply chain to green marketing needs: A
case study. Ind. Mark. Manag. 43, 45 –55. doi:10.1016/j.indmarman.2013.08.003
Dietz, A., Az zaro-Pantel, C., Pibouleau, L., Domenech, S., 2006. Multiobjective optimization for
multiproduct batch plant design under economic and environmental considerations. Comput.
Chem. Eng. 30, 599 –613. doi:10.1016/j.compchemeng.2005.10.017
Doublet, G., Jungblut h, N., Flury, K., Stucki, M., Schori, S., 2013. Life cycle assessment of orange
juice. SENSE -Harmonised Environmental Sustainability in the European food and drink
chain, Seventh Framework Programme: Project no. 288974. Funded by EC. Deliverable D 2.1
ESU -services Ltd.: Zürich.
Dwivedi, P., Spreen, T., Goodrich -Schneider, R., 2012. Global warming impact of Florida’s Not –
From -Concentrate (NFC) orange juice. Agric. Syst. 108, 104 –111.
doi:10.1016/j.agsy.2012.01.006
Eskandarpour, M., Dejax, P., Miemczyk, J., P éton, O., 2015. Sustainable supply chain network design:
An optimization -oriented review. Omega 54, 11 –32. doi:10.1016/j.omega.2015.01.006
Fernandez Lambert, G., Aguilar Lasserre, A., Azzaro -Pantel, C., Miranda -Ackerman, M.A., Purroy
Vázquez, R., del Rosar io Pérez Salazar, M., 2015. Behavior patterns related to the agricultural
practices in the production of Persian lime (Citrus latifolia tanaka) in the seasonal orchard.
Comput. Electron. Agric. 116, 162 –172. doi:10.1016/j.compag.2015.06.007
Fernandez Lambe rt, G.F., Aguilar Lasserre, A.A.A., Miranda -Ackerman, M.A., Sánchez, C.G.M.,
Rivera, B.O.I., Azzaro -Pantel, C., 2014. An expert system for predicting orchard yield and
fruit quality and its impact on the Persian lime supply chain. Eng. Appl. Artif. Intell. 33, 21 –
30. doi:10.1016/j.engappai.2014.03.013
Garcia, D.J., You, F., 2015. Supply chain design and optimization: Challenges and opportunities.
Comput. Chem. Eng. doi:10.1016/j.compchemeng.2015.03.015
Gomez, A., Pibouleau, L., Azzaro -Pantel, C., Domenech, S., Latgé, C., Haubensack, D., 2010.
Multiobjective genetic algorithm strategies for electricity production from generation IV
nuclear technology. Energy Convers. Manag. 51, 859 –871.
Ho, W., Xu, X., Dey, P.K., 2010. Multi -criteria decision making approache s for supplier evaluation
and selection: A literature review. Eur. J. Oper. Res. 202, 16 –24.
doi:10.1016/j.ejor.2009.05.009

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65 Knudsen, M.T., Almeida, G.F. de, Langer, V., Abreu, L.S. de, Halberg, N., 2011. Environmental
assessment of organic juice imported t o Denmark: a case study on oranges (Citrus sinensis)
from Brazil. Org. Agric. 1, 167 –185. doi:10.1007/s13165 -011-0014 -3
Landquist, B., Ingólfsdóttir, G.M., Yngvadóttir, E., Jungbluth, N., Doublet, G., Esturo, A., Ramos, S.,
Ólafsdóttir, G., 2013. Set of en vironmental performance indicators for the food and drink
chain SENSE -Harmonised Environmental Sustainability in the European food and drink
chain, Seventh Framework Programme: Project no. 288974. Funded EC Deliv. D 2.
Lanuzza, F., Mondello, F., Tripodo, M .M., 2014. Studies About the Utilization of Citrus Wastes in
View of Environment Protection, in: Salomone, R., Saija, G. (Eds.), Pathways to
Environmental Sustainability. Springer International Publishing, pp. 147 –156.
Meier, M.S., Stoessel, F., Jungbluth, N., Juraske, R., Schader, C., Stolze, M., 2015. Environmental
impacts of organic and conventional agricultural products – Are the differences captured by
life cycle assessment? J. Environ. Manage. 149, 193 –208. doi:10.1016/j.jenvman.2014.10.006
Minot, N., 1986. Contract farming and its effect on small farmers in less developed countries.
Michigan State University, Department of Agricultural, Food, and Resource Economics.
Miranda -Ackerman, M.A., 2015. Optimisation multi -objectif pour la gestion et la concep tion d’une
chaine logistique verte: application au cas de la filière agroalimentaire du jus d’orange [WWW
Document]. http://www.theses.fr. URL http://www.theses.fr/s139158 (accessed 1.22.16).
Miranda -Ackerman, M.A., Fernández -Lambert, G., Azzaro -Pantel , C., Aguilar -Lasserre, A.A., 2014.
A Multi -Objective Modelling and Optimization Framework for Operations Management of a
Fresh Fruit Supply Chain: A Case Study on a Mexican Lime Company, in: Valadi, J., Siarry,
P. (Eds.), Applications of Metaheuristics in Process Engineering. Springer International
Publishing, pp. 373 –394.
Oakdene Hollins Research and Consulting, 2011. EU Ecolabel for food and feed products (Feasibility
study No. ENV.C.1/ETU/2010/0025). DG Environment, European Commission.
Ouattara, A., Pi bouleau, L., Azzaro -Pantel, C., Domenech, S., Baudet, P., Yao, B., 2012. Economic
and environmental strategies for process design. Comput. Chem. Eng. 36, 174 –188.
doi:10.1016/j.compchemeng.2011.09.016
Pelletier, N., Allacker, K., Pant, R., Manfredi, S., 20 13. The European Commission Organisation
Environmental Footprint method: comparison with other methods, and rationales for key
requirements. Int. J. Life Cycle Assess. 19, 387 –404. doi:10.1007/s11367 -013-0609 -x
Rehber, E., 2000. Vertical Coordination In Th e Agro -Food Industry And Contract Farming: A
Comparative Study Of Turkey And The Usa (Food Marketing Policy Center Research Reports
No. 052). University of Connecticut, Department of Agricultural and Resource Economics,
Charles J. Zwick Center for Food and Resource Policy.
Ren, L., Zhang, Y., Wang, Y., Sun, Z., 2010. Comparative Analysis of a Novel M -TOPSIS Method
and TOPSIS. Appl. Math. Res. EXpress. doi:10.1093/amrx/abm005
Richard, L., Kohls, J.N.U., 1998. Marketing of agricultural products. New Jersey: Prentice Hall, Upper
Saddle River.
Seuring, S., Muller, M., 2008. From a literature review to a conceptual framework for sustainable
supply chain management. J. Clean. Prod. 16, 1699 –1710.
Wikandari, R., Nguyen, H., Millati, R., Niklasson, C., Taherzadeh, M.J., 2015. Improvement of
Biogas Production from Orange Peel Waste by Leaching of Limonene. BioMed Res. Int.
2015, e494182. doi:10.1155/2015/494182
Yue, D., You, F., Snyder, S.W., 2014. Biomass -to-bioenergy and biofuel supply chain optimizati on:
Overview, key issues and challenges. Comput. Chem. Eng., Selected papers from ESCAPE -23
(European Symposium on Computer Aided Process Engineering – 23), 9 -12 June 2013,
Lappeenranta, Finland 66, 36 –56. doi:10.1016/j.compchemeng.2013.11.016