Unravelling life cycle impacts of coffee: Why do results differ so much among studies?

Full text

Sustainable Production and Consumption 47 (2024) 251–266
Available online 8 April 2024
2352-5509/© 2024 The Author(s). Published by Elsevier Ltd on behalf of Institution of Chemical Engineers. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Research Article
Unravelling life cycle impacts of coffee: Why do results differ so much
among studies?
C. Ch´
eron-Bessou
a
,
b
,
*
, I. Acosta-Alba
c
, J. Boissy
d
, S. Payen
a
, C. Rigal
a
,
e
, A.A.R. Setiawan
f
,
M. Sevenster
g
, T. Tran
h
, A. Azapagic
i
a
CIRAD, UMR ABSys, Elsa Group, Av. Agropolis, F-34398 Montpellier, France
b
James Cook University, PO Box 6811, Cairns, QLD 4870, Australia
c
EvaLivo, Independent Consultant, F-02100 Saint Quentin, France
d
Agro-transfert Ressources et Territoires, 80200 Estr´
ees-Mons, France
e
ICRAF, Vietnam ofce, Hanoi, Viet Nam
f
Research Center for Sustainable Production System and Life Cycle Assessment, National Research and Innovation Agency (BRIN), 15314 Tangerang, Selatan, Indonesia
g
CSIRO, Black Mountain Science and Innovation Park, ACT 2601, Canberra, Australia
h
CIRAD, UMR Qualisud, Elsa Group,73 rue JF Breton, F-34398 Montpellier, France
i
Sustainable Industrial Systems, Department of Chemical Engineering, The University of Manchester, M139PL Manchester, UK
ARTICLE INFO
Editor: Prof. Shabbir Gheewala
Keywords:
Agriculture
Carbon footprint
Coffee
Environmental impacts
Life cycle assessment
ABSTRACT
Coffee beans are a major agricultural product and coffee is one of the most widely traded commodities and
consumed beverages globally. Supply chains and cropping systems are very diverse, with contrasted potentials
and performance, as well as environmental impacts. Life Cycle Assessment (LCA) studies are needed to inform on
reduction in impacts, but there is a lack of comprehensive understanding of the variability of existing LCA results
and impacts of the cropping systems and their trade-offs along the supply chains. In an attempt to address this
knowledge gap, the paper presents a systematic literature review of coffee LCA, considering a total of 34 studies
covering 234 coffee systems. Global warming potential (GWP) was the impact category most reported in the
literature, but the results varied greatly at both the farm and drink levels. For the former, the GWP values ranged
from 0.15 to 14.5 (median: 3.6) kg CO
2
eq./kg green coffee beans and for the latter the values ranged from 2 to
23 (median: 8.8) kg CO
2
eq./kg consumed coffee in drinks. Main contributors to the GWP of production of green
coffee beans were land use change (LUC), fertilisers and wet processing. However, there were great in-
consistencies across studies in terms of LUC accounting, eld emissions and wet process modelling. Green coffee
beans production was also the main contributor to the GWP of coffee consumed, followed by brewing and coffee
cup washing. Some studies covered other impacts, in addition to GWP. At both the farm and drink levels, fer-
tilisers and pesticides were the main contributors to eutrophication and acidication, and to ecotoxicity,
respectively. Brewing was the second main contributor at the drink level, in some cases the top contributor for
energy-related indicators. Assumptions on packaging, cup washing and waste disposal were highly variable
across studies. Water impact indicators were hardly comparable due to the system variability and method in-
consistencies. Given the large diversity of coffee cropping systems worldwide, but also the diversity of possible
coffee drinks, we recommend that LCA studies be standardised with respect to the denition of the functional
unit, including consistent quality aspects for both green coffee beans (moisture) and coffee drinks (organoleptic
properties). They should also be more thorough in detailing processes at all stages. More attention should be paid
to the farming system complexity and a mass balance should be ensured when assessing biomass ows con-
cerning LUC, co-products and residue emissions. Finally, more primary data would be needed to decipher the
cropping system diversity, as well as to characterise emissions from all inputs to the eld and bean processing,
notably for wet and semi-wet processing.
* Corresponding author at: CIRAD, UMR ABSys, Elsa Group, Av. Agropolis, F-34398 Montpellier, France.
E-mail address: cecile.bessou@cirad.fr (C. Ch´
eron-Bessou).
Contents lists available at ScienceDirect
Sustainable Production and Consumption
journal homepage: www.elsevier.com/locate/spc
https://doi.org/10.1016/j.spc.2024.04.005
Received 28 December 2023; Received in revised form 3 April 2024; Accepted 3 April 2024

Sustainable Production and Consumption 47 (2024) 251–266
252
1. Introduction
Coffee is one of the most widely consumed beverages and one of the
most traded commodities in the world (FAO, 2023). It is a “typical
example of a global commodity” (Viere et al., 2011). Over the past ten
years, the global coffee production has continuously increased by
1.1–2.4 % annually (Statistica, 2023), catching up with the long-term
average growth rate of coffee consumption worldwide of 2.3 % over
the period 1990–2018 (ICO, 2023). To meet the growing global demand,
coffee production is expected to double by 2050 (Conservation Inter-
national, 2020), potentially driving land use change (LUC) and defor-
estation and impacting on biodiversity and climate change. In its recent
policies against “imported” deforestation (EU, 2023), the European
Commission targeted coffee, among other global commodities, that pose
such risks. Although the environmental concerns have pushed a rapid
development of sustainability initiatives among coffee sector stake-
holders (Noponen et al., 2012), it is still not clear how these initiatives
help to reduce the impacts of coffee in practice.
Coffee is grown in the tropics and consumed all around the world,
and in particular in Europe (54 %), Asia and Pacic (46 %), and North
America (31 %) (ICO, 2023). Western Europe concentrates the coffee
roasting industry, which produces roasted coffee consumed locally or
exported to other regions (Hejna, 2021). The great diversity of agri-
cultural systems in the tropics and the various trade routes give rise to
very diverse supply chains with contrasted potentials and performance.
Coffee can notably be grown in agroforestry plots, whose potential
triggers interests in the application of Climate Smart Agriculture stra-
tegies to coffee production (Djufry et al., 2022; Gabiri et al., 2022). On
the other hand, several studies have shown the climate sensitivity of
coffee and the variable impact of climate change on coffee suitability,
yield and farmers' livelihoods (Alemu and Dufera, 2017; Grüter et al.,
2022; Rahn et al., 2014). Both mitigation and adaptation strategies
require quantifying the performance and improvement opportunities,
while accounting for the diversity of the production systems.
In this context, Life Cycle Assessment (LCA) studies of coffee prod-
ucts are needed to provide information on impact contributions and
improvement pathways. LCA is a widely used methodology for quanti-
fying environmental impacts as its holistic approach covers the whole
supply chain and a number of environmental impacts. However, given
the variability in coffee production systems, as well as in the LCA
studies, the results vary signicantly. Consequently, there is still a lack
of comprehensive understanding of the impacts of various management
systems and their trade-offs along the supply chains. Therefore, there is a
need for an in-depth review of existing LCA studies, disentangling
methodological aspects from the inherent variability of coffee systems.
This article presents a systematic review of coffee LCA literature,
investigating rst the diverse supply chains and system boundaries, then
the main impact drivers for the various system boundaries. The goal of
the study is two-fold: i) to dissect the intrinsic system variability and its
inuence on the results, as well as to understand better the need for
more knowledge and data for coffee LCA; and ii) to provide insights on
how to increase comparability between coffee LCA studies and harmo-
nise LCA practices for coffee and perennial cropping systems at large.
2. Methods
2.1. Literature review
We conducted a systematic review of coffee LCA studies available in
the literature. The search was carried out on August 14, 2023 using the
search strings: “coffee (Topic)” AND lca OR “life cycle a*” (Topic) with
no further restriction on language. The Web of Science and Scopus
yielded 147 and 172 outputs, respectively. The search on Google Scholar
yielded 285 outputs despite being more restricted to avoid too many
false outputs, using the search strings: “coffee lca” OR “coffee life cycle
assessment” OR “coffee life cycle analysis” OR “lca of coffee” OR “life
cycle assessment of coffee” OR “life cycle of coffee” OR “life cycle
analysis of coffee” OR “life cycle analyses of coffee” in English only and
without including references. The least Google relevant pages, i.e. the
second half of output pages (Jansen and Spink, 2006), were ltered
manually.
We further added studies dedicated to carbon footprint analyses (i.e.
92 outputs from Google Scholar on August 14, 2023 using the search
strings “coffee carbon footprint” OR “carbon footprint of coffee”).
Although we originally aimed at reviewing LCA studies only, carbon
footprint studies were also relevant since i) they mostly are LCA-based, i.
e. partial LCA studies; ii) they are more numerous as many studies focus
on climate change issues only; and iii) they could provide signicant
insights on how this impact was calculated, providing further clues on
data or methodological bottlenecks. On the other hand, we did not
specically add partial studies on water footprint since, contrary to
carbon footprint, there are too many diverging methodologies poten-
tially involved behind the “water footprint” term, including mostly non-
LCA-based approaches.
Checks on search errors and duplicates led to a consistent corpus of
227 papers and reports. Then, publications were rst ltered according
to their goal and scope, and studies eliciting no specic system bound-
aries or coffee LCA results were discarded (76 %). Most of those dis-
carded studies i) did not present LCA coffee results but rather
inventories, sustainability assessment indicators, LCA-based water
footprints, and so on (21 %); ii) focused on technologies, processing or
packaging only (18 %); iii) were out of scope, such as reviews on
biomass or LCA recommendations (15 %); and iv) considered recycling
processes for coffee waste that entered the system with no environ-
mental burden, i.e. not accounting for coffee production and processing
(15 %). Spent ground coffee, in particular, was the focus of many recent
publications, occupying about a half of the coffee-related LCA studies
published in 2022–2023. The rest of the studies were related to chemical
analyses of coffee waste (3 %), socio-economic aspects including con-
sumers' views on LCA results (3 %), or were inaccessible (1 %).
An in-depth review resulted in further 18 studies being discarded
because they were either partially inconsistent or redundant. The most
common source of error or uncertainty in the paper quality was the lack
of explicit eld emissions modelling. In case of any doubt, we wrote to
the authors to seek clarication. When sufcient clarications were
given, studies were kept in the nal corpus.
In the case of theses (PhD and MSc), whose parts were also published
as articles or book chapters, we consolidated all needed information
from the various sources and only kept a unique reference associated
with a given dataset to avoid any redundancy. We did the same in the
case of papers published in conference proceedings which were further
published as journal articles, or articles from the same authors providing
complementary information on unique LCA studies. As a result of the
various lters and consolidated information, the nal corpus consisted
of 34 examined studies: 29 journal articles, three public reports that had
undergone an external peer review, one non-peer reviewed report, and
one PhD thesis (Table S1). Altogether, roughly 76 % of the studies were
published in the last ten years.
2.2. Data collection and analysis
The data collection included metadata on the studied countries,
coffee species, farming systems and processing types. Impact values
were recorded per functional units and sub-systems. Where necessary,
results on impacts and contributing stages were extracted from gures
using a free online tool (https://apps.automeris.io/wpd/). In some
cases, we also re-calculated some results to harmonise the functional
units, i.e. to convert acre- into hectare-based results or coffee drink- into
volume or coffee weight-based results (see further comments in Section
3).
The analysis was straightforward based on simple descriptive sta-
tistics. Data exploration was carried out with R v.4.2.1 on R studio
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
253
v.2023.06.2. (RStudio Team, 2023). We rst analysed the consistency
and reliability of studies in methodological terms, notably, regarding the
data representativeness and scopes of the studies. Then, we investigated
further the impact values. How detailed the systems were, and whether
results were disaggregated or not, varied widely across studies. Finally,
combining reection on the methods and results, we made some rec-
ommendations in order to make better use of coffee LCA studies, as well
as to how to improve future studies.
3. Results and discussion
The following sections rst provide an overview of the goal and
scope considered across the studies, starting with coffee origins and
cropping systems, followed by details on the system boundaries and
impact assessments (Section 3.1). Then, impact results are presented and
discussed in Section 3.2 for the global warming impact and in Section
3.3. for the other most reported impact categories encountered across
the studies.
3.1. Goal and scope of the reviewed coffee LCA studies
3.1.1. Coffee origins and cropping systems
The great majority of reviewed studies (25, covering 76 % of the
studied coffee systems) investigated Coffea arabica sp.; four studies
looked at Coffea canephora sp. Robusta (7 % of the studied systems); the
remaining studies (5) considered both or did not specify (17 % of the
studied systems); and none investigated Coffea liberica sp. (Fig. 1a).
Although the dominance of arabica was relevant in the past, robusta's
global share is getting close to half nowadays, i.e. 44 % of total coffee
production in 2023 (ICO, 2023). Hence, robusta and other species were
underrepresented in the studies. Central and South America was the
most represented region with 72 % of all studied systems (Fig. 1b),
including 18 % and 16 % for Colombia and Brazil alone, respectively
(Fig. S1). This is aligned with this region representing about 70 % of the
global coffee production (Rega and Ferranti, 2019). Costa Rica and
Vietnam were also considered in various studies. From the main current
producing countries, only Ethiopia and Uganda were not represented in
the studies reviewed here. In some cases, the indicated countries of
origin were not actually investigated. In Humbert et al. (2009), for
instance, the focus was to compare the impacts of various coffee prep-
arations while mixing coffee from Brazil, Colombia and Vietnam as the
main producers. The theoretical systems relied on secondary data for the
farm stage, whereby Brazil was assumed as a proxy for all considered
coffee producing countries. In fact, only transport routes were different
depending on the origin, which may be misleading in terms of links
between origins and country-specic cropping systems. Likewise, Has-
sard et al. (2014) used proxy data from Nicaragua to model their theo-
retical supply from Guatemala. Overall, a third of the studies did not use
primary data for the farm stage and relied on existing published data-
sets; those were mostly the ones on Brazil by Coltro et al. (2006), with
43 % of studies relying on secondary data from this single reference. In
these secondary data-based studies, the uncertainty of results was
greater due to potential uncovered discrepancies between theoretical
concatenated systems and actual practices in the eld.
In terms of the cropping system complexity, four main categories
were covered: complex agroforest (27 %); simple agroforest (22 %) - also
called by some authors shaded monoculture; “full-sun” monoculture (27
%); and “full-sun” polyculture (1 %). The rest of the studied systems (23
%) were either mixes of various systems or did not provide enough de-
tails on the cropping system types; those studies mostly used secondary
data for the farm stage. By denition, agroforestry is a broad concept
whose baseline is the combination of crops, which can be both annuals
or perennials, and trees. However, there are critical differences among
various agroforestry systems. In a traditional coffee agroforestry system,
also called “rustic” (Van Rikxoort et al., 2014), natural forest is only
partially cleared in order to keep existing native trees within the plot. On
the contrary, in commercial polyculture, new trees are usually planted
together with coffee trees in order to provide specic benets. Hence,
both the density and the type of associated trees matter when analysing
agroforest diversity, as both imply different practices and overall plot
performance.
In this review, we categorised simple agroforestry systems as those
encompassing coffee trees under a single shade tree species. The more
complex systems, with several associated annual or tree species, were
categorised as complex agroforestry systems. This category encompasses
traditional polyculture, commercial polyculture with several shade tree
species, and unspecied agroforestry coffee plantations. “Full-sun”
polyculture differs from agroforestry systems due to coffee trees not
being tall enough to create an actual vertical stratication as symp-
tomatic for agroforestry systems. “Full-sun” polyculture may be associ-
ations of coffee with maize or banana, for instance, which may be rather
common in some countries but were not much investigated in the
reviewed corpus. Various typologies of coffee systems exist, which were
not reviewed as that was beyond the scope of the LCA review. The simple
typology discussed above and adopted here was aimed at helping with
the result analysis and did not reect further on other potential
typologies.
In the reviewed studies, associated trees were mostly the focus of
standing biomass estimation (see Section 3) and not much attention was
paid to the interactions among crops and trees and how to account for
associated ecosystem services and allocation issues within LCA, which is
discussed further later on. Only a few studies explicitly accounted for
allocation ratios among associated crops (e.g. Basavalingaiah et al.,
2022; Enveritas, 2023) or investigated complex outputs from agrofor-
estry systems (e.g. Acosta-Alba et al., 2020, 2019).
For the rst processing stage, the great majority of studies (55 %)
Fig. 1. Overview of the 234 studied coffee systems and origins (% of the total number of coffee system studied).
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
254
investigated wet processing only or together with dry processing (26 %);
19 % of the studies investigated dry processing only. Whether this
processing was taking place on- or off-farm was not systematically
specied but could be deduced from specic transportation details.
Semi-wet processing was not explored in any reviewed study.
Very few studies considered the perennial cycle of coffee trees in the
modelling of agricultural practices and related input-output uxes. Most
only gathered data for one year of productive coffee plantations. As
shown in the literature, given the complex pluri-annual functioning of
perennial plantations and delays in responses to environmental stresses
or management practices, it is paramount to consider several consecu-
tive years of production and also to integrate other development stages
to average the performance and impacts (Bessou et al., 2016, 2013;
Cerutti et al., 2013). Bias in results may be critical, particularly in
studies relying on data collected in the eld for a reduced number of
plantations and over short periods of time (e.g. Quack et al., 2009), as
well as in those comparing contrasted cropping systems (e.g. Brenes-
Peralta et al., 2022). Such a bias would be less critical in studies covering
several years (e.g. Noponen et al., 2012) or covering large samples,
whereby regional disparities in practices and performance along the
crop cycle may be geographically averaged (e.g. Enveritas, 2023).
There were a few exceptions, though, with some more systematic and
holistic studies including the nursery stage, various productive years, or
the full cycle (e.g. Acosta-Alba et al., 2020; Rahmah et al., 2023). Brenes-
Peralta et al. (2022) included the nursery stage but then relied on data
for just one harvest. In Trinh et al. (2020), data inventory was detailed
by plantation stage. When focusing on yields, which directly inuenced
the impact values due to being the functional-unit common denomina-
tor, it was interesting to note that average yields of green coffee beans
(t/ha) were 18–21 % higher when computed over the 21 years of full
productivity compared to the average yields computed over the whole
cycle of 30 years, with the latter including no or less productive years.
Indeed, yields during the rst six years of production initiation and the
last three years of coffee aging were respectively 2.4 and 1.8 times lower
on average than those during full productive years and the averaged
cycle. Depending on the year for the data collection and the age of the
plantation, those differences in yields over the whole crop cycle would
have affected the LCA results if calculated without integrating the whole
cycle. Moreover, those differences might vary across compared systems,
leading to a consequent bias in the results. In the exemplied study,
variability in yields along the whole cycle was smaller in the conven-
tional intensive system compared to the two others, i.e. conventional
moderate and organic intensive (Trinh et al., 2020).
Representativeness is a key data quality attribute in LCA because it
denes how well-suited the data are to full the study objectives. In
agricultural systems, all dimensions of representativeness matter,
including the geographical, temporal and technological ones, since
management practices are highly dependent on the local contexts and
can vary greatly. Table 1 lists some of the main agronomic parameters
gathered from the reviewed studies. High standard deviations indicated
a great variability in all parameters within the sample, in particular for
nitrogen (N)-based fertilisers and green coffee bean yields. Seven studies
considered irrigation in coffee plantations. More systems may have
required some irrigation but the information was missing in many
studies and irrigation practices were globally poorly detailed.
The least detailed practices at the farm level were related to crop
residues and organic soil amendments. Crop residues on coffee farms
may come from two main sources: the coffee itself and the associated
crops or trees. Coffee residues consist of both leaf litter and pruning
residues within the plantation and coffee residues from processing
(coffee pulp, husks and parchment). Leaf litter and pruning residues may
account for 5–12 t/ha depending on both coffee and shade trees den-
sities (Van Rikxoort et al., 2013). Further crop residues may be brought
from other plots or farms as organic amendments. Likewise, coffee
plantation residues may be exported or recycled outside of the plot (such
as pruning wood used for fences or fuel wood). However, those scenarios
were not discussed in the coffee studies, apart from the two bioenergy-
dedicated studies by the same authors (Aristiz´
abal-Marulanda et al.,
2021a, 2021b). The amounts and management types of crop residues
and derived organic fertilisers may signicantly inuence coffee per-
formance, both in terms of agronomic outputs and environmental im-
pacts (Haggar et al., 2011; Van Rikxoort et al., 2013; Youkhana and Idol,
2016). Depending on the type of organic matter and how it may degrade,
ferment or be stabilised (e.g. by composting), emission types and
amounts will vary. According to the IPCC (2006) as implemented in the
Cool Farm Tool, residues left in heaps or pits would emit 33 times more
CO
2
eq. than when used as mulch in the eld due to anaerobic conditions
leading to emissions of potent greenhouse gases (GHG), such as N
2
O and
CH
4
, while biogenic CO
2
from aerobic decomposition is considered as
carbon-neutral. There are many possible emission intensities along this
33-fold span being determined by the various combination of co-
products, residues and their managements. Nevertheless, very few
studies recorded precise information on residues and none of them in-
ventoried all potential residues and their fate. For instance, Enveritas
(2023) inventoried coffee leaf litter and husks only, while Rahmah et al.
(2023) compared scenarios with and without eld application of coffee
pulp. In the latter, however, the actual emissions of degrading coffee
pulp were not explicitly accounted for.
3.1.2. Coffee system boundaries
LCA and carbon footprint studies aimed to assess either the impacts
of coffee as an agricultural commodity, more or less processed, or the
impacts of coffee drinks. All reviewed studies applied the attributional
LCA approach. Details on the overarching methodologies applied are
listed in Table S2. The supply chain from plantation up to the con-
sumption of coffee as considered in the literature is summarised in
Fig. 2, together with key information on details for the main stages and
inputs. Across the reviewed studies, various plantations and processing
routes were covered, except for semi-wet (also called honey) coffee,
hence not displayed in the gure. As for the farming stage, about one
third of the studies did not use primary data for the processing stages.
The wet processing route was most represented across the studies (67 %
investigating either wet processing only or both wet and dry processing),
in accordance with arabica being the most studied species and the one
mostly processed in this way. Primary processing is dened as the pro-
cessing of cherries into green coffee beans; it includes several stages to
separate the beans from the outer layers, then sort out the market-
quality beans. Secondary processing involves a sequence of several
processing stages further down the supply chain to convert green coffee
beans into ground or instant coffee; it notably includes roasting and
grinding but also packaging and, in some cases, further processing, such
as instant coffee freeze or spray drying.
In terms of the system boundaries, about a third of the studied sys-
tems included consumption of coffee drinks, mostly comparing at least
three types. Moreover, some studies presented results both at farm or
processing-plant gate and post-consumption, which provided results for
Table 1
Variations in key agronomic parameters across reviewed studies (both coffee
species were considered together and results are displayed as they appeared in
the studies, hence there is no linear relationship between outputs in the table).
Parameter Mean Median Min Max
Planting density (coffee trees/ha) 4,067 (±41 %) 4,500 150 10,000
Nitrogen fertilisers
a
(kg/ha) 215 (±72 %) 177 0 1152
Fresh coffee cherry yield (kg/ha) 5,288 (±51 %) 4,800 628 13,605
Coffee parchment yield (kg/ha) 1,094 (±56 %) 1,032 126 2,387
Green coffee beans yield (kg/ha) 1,419 (±79 %) 1,064 373 5,386
Irrigation water (m
3
/ha)
b
3,103 (±45 %) 3,458 1,124 4,940
a
Not all studies displayed the detailed amount for each fertiliser types nor the
N content of organic amendments applied. Total N fertiliser estimates are likely
underestimated. Standard deviations to the means are given in brackets.
b
Focusing on irrigated systems only (studied systems n =14).
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
255
234 coffee systems in total. Surprisingly, three studies dened the
functional unit as “1 kg of green coffee beans”, although they included
secondary processing and coffee consumption (Birkenberg and Birner,
2018; Killian et al., 2013; Nab and Maslin, 2020). These results could be
misleading, especially if extracted from the studied contexts and
compared on the same functional-unit basis but with different system
boundaries. The use of the “green coffee beans per hectare” or “hectare”
metrics when including primary processing of coffee cherries (i.e. at
primary processing gate and not farm gate per se, even though “primary
processing gate” maybe within the farm) may also be confusing as a
hectare most commonly refers to outputs from the eld without
including any processing. Finally, none of the reviewed papers consid-
ered “the moisture in the coffee” functional unit. At the global level, the
moisture of green coffee beans only varies between 10 and 14 % as this is
standardised internationally. However, variations in this parameter
could matter when comparing studies since cumulative losses in weight
(through moisture) or actual quantities along the chain would affect
linearly the impacts per unit output.
At the consumption level (cradle-to-grave), assumptions on coffee
dilution and coffee waste varied across the type of the drink and serving
and could lead to some confusion when comparing coffee drinks and
their impacts. Some studies presented results for two functional units:
“per serving” (with various volumes) and “per coffee volume”, which
made it possible to limit result differences strictly due to the dilution
effect. To avoid confusion with the dilution effect, when analysing
further the coffee drinks studies, we harmonised the results according to
the actual coffee content. Note that coffee drinks including sugar or milk
were not included in the review due to the added impacts from those
components not related directly to the coffee itself.
Despite focusing on differences in the type of coffee drinks, none of
the studies included organoleptic criteria within the functional unit. For
instance, espresso or ltered coffee was mostly compared on a volume
basis without any consideration of differences in strength or taste. As
consumer taste preferences might be the main driver for the coffee
preparation type, which in turn may inuence the coffee impacts, it
would be justied to consider some organoleptic properties. The only
exception was the study by Gosalvitr et al. (2023), where the authors
compared coffee drinks on the basis of a common amount of caffeine
provided (100 mg). Future studies could further investigate organoleptic
properties associated with both the type of coffee and its preparation,
and adjust the LCA calculations to the actual expected function of coffee
(i.e. more focused on the strength, taste or other coffee properties).
Studies investigating coffee drinks composition depending on both
coffee types (Mussatto et al., 2011; Vignoli et al., 2014) and drink
preparation types (e.g. Gobbi et al., 2023) could help to dene such a
properties-based functional unit for coffee drinks. The details on the
coffee drinks composition and how it may affect their taste and con-
sumer choices could be useful to dene a taste or properties-based
functional unit for coffee drinks as it is done, for instance, when ac-
counting for fat and protein content in milk with the fat- and protein-
corrected functional unit in LCA of milk products.
Capital goods were generally excluded from the studies, which is in
line with commonly used guidelines for agricultural or horticultural
production such as PAS2050-1 (BSI, 2012), as justied by some authors.
Capital goods were only included in two cradle-to-grave studies (Chayer
and Kicak, 2015; Humbert et al., 2009), although they were excluded in
background data for the farming stage. These studies highlighted some
contributions of the manufacture of coffee brewer and the dishwasher,
Fig. 2. Supply chain from plantation up to the consumption stage as explored in the reviewed literature, with corresponding numbers of studies according to the
system boundaries.
Credits: Scooter by Draftphic; Tractor and trailer by Azam Ishaq; Ship by Jordan Ivey; Truck by Jonathan Li; Trolley by Saifurrijal – Noun Project CCBY3.0.
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
256
although more signicantly for the water impact indicators. This sug-
gests that it might be relevant to investigate further the discrepancies in
impacts between coffee drinks prepared using different coffee machines.
A great majority of the studies did not consider or mention any co-
product allocation. Among the remaining studies, there were i) three
studies without primary data for the farming stage and relying on
background data, including system expansion for waste management
and energy recovery (Brommer et al., 2011; Gosalvitr, 2021; Hassard
et al., 2014); ii) two studies (but by the same authors) focusing on
downstream energy production from cut stems by applying mass allo-
cation between coffee and stems, then substitution (Aristiz´
abal-Mar-
ulanda et al., 2021a, 2021b); and iii) two studies applying economic
allocation between coffee and associated crops in the same plot (Basa-
valingaiah et al., 2022; Enveritas, 2023). Given the diversity of systems,
including within agroforestry types, as well as the diversity of coffee co-
products, the lack of an in-depth investigation on co-products revealed
potential gaps in accounting for the specicities and discrepancies
across coffee supply chains.
3.1.3. Impact categories and assessment methods
About 60 % of the studies covered more than just GWP (a.k.a. climate
change or carbon footprint). Half of those were full LCA mostly relying
on various versions of the ReCiPe method (seven studies), with a small
number using CML 2001 (two), and ILCD and TRACI (one each). The
remaining studies looked at GWP and either water consumption or en-
ergy related impacts. In a few LCA studies, the focus on GWP plus one or
a few more indicators did not align well with the ISO 14040 (2006)
requirement to select a comprehensive set of impacts. Even when more
indicators were selected, the choices were typically justied only
partially compared to the ISO 14040 requirement. While it is recognised
that studies are often limited in resources, better discussion of the lim-
itations of the impact assessment method used would improve the
interpretation of results.
Concerning the GWP, not all the studies relied on the same charac-
terisation factors. Most single-impact or full LCA studies based on the
ReCiPe 2008, CML 2001, TRACI 2008 and ILCD 2011 methods, as well
as those using Cool Farm Tool v1., relied on the 100-year GWP values
from the IPCC Fourth Assessment Report (IPCC, 2007). The other full
LCA studies based on ReCiPe 2016 and Usva et al. (2020), estimated the
100-year GWP values based on the Fifth Assessment Report (IPCC,
2014). The discrepancies among the two versions would be mostly
critical for biogenic methane emissions from wastewater treatment,
since GWP varies from 25 to 34 kg CO
2
eq./kg CH
4
with feedback. No
study used the IPCC Sixth Assessment Report from 2021, where dis-
crepancies for the GWP of N
2
O would have also mattered.
Besides the GWP, ILCD- and CML-based studies mostly focused on
ve to seven categories: eutrophication (including freshwater, marine or
terrestrial eutrophication), acidication, ozone later depletion, non-
renewable cumulative energy demand, human toxicity, abiotic deple-
tion and water depletion. Contrary to the ILCD- and CML-based studies,
those following the ReCiPe method focused more on the green coffee
beans production and relied slightly more on primary data for the farm
stage. Across these studies, not all ReCiPe indicators were considered or
discussed in detail. One study applied the IMPACT 2002+method
(Humbert et al., 2009). Finally, one study reported eight other impact
categories of TRACI to compare three brewing methods (Hicks, 2018).
Given differences in the applied impact assessment methods across
all studies and impact categories, we could not carry out a quantitative
analysis of all individual impacts. Instead, in the next section we focus
on the GWP that was the most systematically investigated in the liter-
ature (Section 3.2). However, in a subsequent section we also provide
some insights on the main other common midpoint indicators consid-
ered in the bulk of studies: terrestrial acidication, various eutrophi-
cation indicators, ecotoxicity and ozone depletion (Section 3.3.1).
Finally, we discuss in more detail energy-, mineral resource- (Section
3.3.2) and water-related impacts (Section 3.3.3).
3.2. Global warming potential of coffee reported in LCA studies
3.2.1. A general overview
The results for the GWP varied greatly across studies depending on
the system boundaries, but also for similar system boundaries. A sum-
mary of key results is given in Table 2. For different system boundaries,
both originally published results and those adjusted in this study to
enable comparisons are listed (see Table S3 for details). The adjusted
results combined ndings for the parchment and green coffee beans for
the cradle-to-primary-processing gate system boundary, with or without
LUC, and adjusted results per g of coffee for the cradle-to-grave
boundary. For the latter, results expressed per “kg green coffee beans”
were not included in the adjusted results range due to too many possi-
bilities and uncertainties in the conversion ratios for nal ground and
consumed coffee.
3.2.2. Overview of cradle-to-grave results
For the cradle-to-grave system boundary, when adjusting the results
to “per g of coffee”, the results for GWP varied by an order of magnitude,
from 0.002 to 0.04 kg CO
2
eq. (Table 2). The adjusted range was 25
times lower than that of the published results, given the large differences
across compared drink types and volumes, and even quantities in kg (80
times lower). Comparing on the same quantities of coffee stressed the
variability due to actual differences in the supply chains (from coffee
farming to brewing type and waste disposal), smoothing out the dilution
effect. Comparing on a similar volume basis with different coffee dilu-
tion ratios would not be suitable as long as the quality of the drink is not
investigated. One study investigating differences in coffee drinks im-
pacts based on a common caffeine content highlighted that such a unit
would further reduce the variability range (Gosalvitr et al., 2023).
Beyond the actual process differences across studied chains, the
choice of the functional unit added further variability in the results. This
variability was then linearly exacerbated by the varying assumptions on
conversion ratios for the various processes along the supply chain; the
cherry-to-green coffee ratio was particularly variable across studies
(Table 2). Conversion ratios were not systematically reported in the
studies, nor was the moisture content of the various coffee products (not
even for the functional units), despite their potential inuence on both
the output ows and the nal coffee quality.
The main impact contributors were the production of green coffee
beans
1
(median 63 %), brewing, cup washing and waste management,
each accounting for about 18 % (median), packaging (median 9–18 %)
and roasting and grinding (median 8 %). The packaging contribution
differed greatly in the case of single-pod or capsule use (18 %) compared
to all the other systems (9 %). However, this contribution was associated
with a great uncertainty as not all studies potentially using single-pods
necessarily specied it. In particular, some studies investigating
espresso, where the packaging contribution was signicantly above the
median, could be related to espresso single-pods.
Only one study (Hassard et al., 2014) provided the distinct contri-
bution of instant-coffee processing, which was 14 % in the exemplied
supply chain. In that study, instant coffee had much higher impact due
to both the added process stage and the higher amount of green coffee
beans needed per unit of instant coffee (Table 2), hence, enhancing the
contribution of the farm stage. Three other studies compared instant
coffee to other drinks (Büsser and Jungbluth, 2009; Gosalvitr, 2021;
Humbert et al., 2009). In contrast to the above-mentioned study, the
contribution of instant coffee processing was not detailed and the nal
impact of instant coffee was lower compared to the other coffee drinks.
In Büsser and Jungbluth (2009), the amount of instant coffee used was
1
Results from one study (Nab and Maslin, 2020), were left out of the average
calculations due to a possible aw in the theoretical modelling of supply chains
based on secondary data on the green coffee modied from De Marco et al.
(2018), so that the farm stage barely led to any emissions.
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
257
3.5 times lower than that of ground coffee, compared to a 1–2.2 factor
across all coffee drinks in Hassard et al. (2014). In Humbert et al. (2009),
the amount of instant coffee used was 3.25–6.75 times lower than that of
ground coffee, and the amount of green coffee beans needed to produce
instant coffee was 1.8 times higher than to produce ground coffee.
Not all the studies distinguished all the various contributors to GWP.
Brewing and cup washing were sometimes grouped together in a single
contributor, with or without waste management, or all were grouped
into a “use” contributor. Hence, the median prole for all contributors
did not sum up to 100 % and only provided an approximation of the
relative order of magnitude for the various contributions (Fig. 3). The
contributions of brewing and washing stages were related to the amount
of energy used and varied depending on assumptions related to coffee
waste, including the energy for keeping the coffee warm. When looking
at energy or water use indicators, the contributions of these stages were
even greater (see Section 3.4). The variability in practices and studied
details regarding the waste considered and their management increased
the results variability and uncertainty. Standard deviations to the mean
for all contributions were above 50 %, except for the green coffee pro-
duction stage (32 %), which was consistently the main contributor. At
the retail level, when consumption was not included, contributions of
the other stages were proportionally higher. Median contribution of
green coffee beans production reached 85 %. The key role of the farm
stage up to green coffee production stressed the need to compare studies
based on a similar coffee content.
About one third of the studies included coffee import, of which nine
studies to European countries (mostly Germany, then Finland, UK and
Italy), two to North America, one to Japan, and one to several of them. In
the great majority of cases, coffee was imported from Latin America.
Across these studies, the GWP of the import transport varied signi-
cantly from negligible in the case of ship transportation, up to >73 % in
the case of airfreight. On average, when shipping was considered, the
contribution of transport accounted for a few percentage points (up to
12 %, with the median of ~3 %) mostly correlated to the relative
contribution of the green coffee production. None of the cradle-to-grave
studies considered any LUC at the farm stage. Some investigated the
Table 2
Number of studied systems for which at least one impact result was provided in the reviewed studies and an overview of the results for global warming potential.
Total number of systems: 234 Cradle-to-farm gate
without any processing
Cradle-to-primary-processing
gate (on- or off-farm)
Cradle-to-secondary-
processing gate
Cradle-to-grave (including coffee
consumption)
Studied systems count (n) 76 65 19 74
Studied system count by functional unit 1 ha⋅yr: 42
1 acre⋅yr: 3
1 kg coffee cherry: 31
1 kg green coffee: 45
1 kg coffee parchment: 14
1000USD ha-outputs
(although processed): 3
1 ha⋅yr
(although processed): 3
1 kg ground coffee: 10
1 kg instant coffee: 4
1 kg decaf blend coffee:
1
1 MJ ethanol/
electricity: 4
Drip/lter coffee: 30
Espresso coffee: 8
Instant coffee: 6
Pressed coffee: 7
Single-pod coffee: 10
Ground coffee: 2
1 kg “green coffee” (although
consumed as ground coffee): 6
Various: 5
Averaged coffee product ratios Cherry kg/ha: 5288
(±51 %)
Cherry/parchment: 5
cherry/green: 5.7 (±19 %)
parchment/green: 1.25
Green/ground: 1.20
(±6 %)
green/instant: 2.5 (±4
%)
Various drinks with various coffee
contents
Published GWP range: min-max (kg CO
2
eq.) Per ha⋅yr (with and
without LUC): −9960 to
102,330
per kg fresh cherry: 0.03
to 1.82
Per kg green coffee beans
(with and without LUC): 0.15
to 10.52
per kg coffee parchment: 3.10
to 11.61
per 1000USD ha-outputs:
1500 to 3500
per ha⋅yr: 6400 to 8700
Per kg ground coffee:
0.53 to 8.50
per kg instant coffee:
15.2 to 17
per kg decaffeinated
coffee blend: 3.29
per MJ ethanol/
electricity: −0.005 to
0.24
Drip/lter coffee (various functional
units): 0.01 to 0.80
Espresso coffee (various functional
units): 0.03 to 5.10
Instant coffee (various functional
units): 0.035 to 0.20
Pressed coffee (various functional
units): 0.01 to 0.06
Single pod coffee: 0.03 to 1.2
Ground coffee: 0.09 to 0.13
1 k g “green coffee” (although
consumed as ground coffee): 3.02 to
16.04
Overall (various functional units
covering a range from g to kg coffee):
0.01 to 16.04
Adjusted functional unit-GWP range (including
differentiation between with or without LUC
a
):
min-max (kg CO
2
eq.)
Per ha⋅yr (with LUC):
−9960 to 102,330
per ha⋅yr (without LUC):
109 to 10,220
per kg fresh cherry (with
LUC): none
per kg fresh cherry
(without LUC): 0.03 to
1.82
Per kg green coffee beans
b
(with LUC): 1.63 to 10.52
per kg green coffee beans
b
(without LUC): 0.15 to 14.51
Per kg ground coffee:
0.53 to 8.50
per kg instant coffee:
15.2 to 17
per kg decaf blend
coffee:
3.29
per MJ ethanol/
electricity:
−0.005 to 0.24
Drip/lter coffee (per g coffee
consumed): 0.002 to 0.02
Espresso coffee (per g coffee
consumed): 0.002 to 0.01
Instant coffee (per g coffee
consumed): 0.007 to 0.04
Pressed coffee (per g coffee
consumed): 0.002 to 0.01
Single pod coffee (per g coffee
consumed): 0.003 to 0.02
Ground coffee and various (per g
coffee consumed): 0.01 to 0.02
Overall per g coffee drunk: 0.002 to
0.04
Overall per 100 mg caffeine: 0.07 to
0.16
a
LUC: land use change.
b
Conversion of parchment into green coffee embodied uncertainty related to both the ratio and an underestimation of added potential impact from parchment
hulling. Results expressed at the primary-processing gate per ha⋅yr and USD⋅ha were also available per kg green coffee and are therefore only provided in this latter
functional unit to avoid redundancy.
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
258
potential inuence of LUC but did not include it in the cradle-to-grave
results.
3.2.3. Overview of cradle-to-primary processing gate results for green coffee
beans
3.2.3.1. Global warming potential and contributors. Focusing on the
green coffee beans production (with the system boundary from cradle-
to-primary processing gate), the discrepancy across results could be
critical, leading to either net positive or net negative GWP, depending on
the considerations of biogenic carbon storage and LUC. Overall, four
studies investigated direct LUC
2
(Enveritas, 2023; Noponen et al., 2013;
Ruben et al., 2018; Usva et al., 2020) and four others considered some
biogenic carbon storage in the coffee plantations without modelling any
LUC (Basavalingaiah et al., 2022; Jaramillo et al., 2017; Maina et al.,
2016; Van Rikxoort et al., 2013). Biogenic carbon stored within coffee or
other trees at the plantations should not be included in the GWP unless
considered within a proper long-term land use and LUC modelling, as
specied by various guidelines (e.g. IPCC (2006) and PAS2050 (BSI,
2011)). Carbon storage in any stand may be accounted for only in
relative quantities compared to previous stands and providing that a
consistent time frame is aligned with a minimum time-averaged storage
(at least over 20 years according to the IPCC (2006) recommendations).
Across the reviewed studies, net negative GWP, such as in Noponen et al.
(2013), might have resulted from distorted LUC modelling or inconsis-
tent biogenic carbon accounting. Distortion might be due to varying
choices across studies in terms of time parameters. Inconsistent biogenic
carbon accounting might be due to aws in the extrapolation of carbon
stock changes or imbalanced accounting for carbon storage and release
in LUC contexts. In Noponen et al. (2013), for instance, carbon stocks
were estimated and amortised over nine years due to experimental
constraints. The consensual time frame for carbon estimates is at least
20 years and short-term storage in waste should not be accounted for
(IPCC, 2006). Therefore, the LUC modelling in Noponen et al. (2013)
might be distorted and is not discussed further here.
According to the remaining studies, the GWP of green coffee beans
following LUC to establish the plantations varied from 1.63 to 10.52 kg
CO
2
eq./kg green coffee beans, based on the IPCC Tier 1 for LUC
modelling (IPCC, 2006). The LUC contribution to the GWP ranged from
1 % to 75 % (Fig. 3), hence leading up to a four-fold increase in the
impact. In the 1 % contribution case, LUC contribution was averaged
over a whole region including thousands of farmers, among whom very
few would mention any LUC. The authors specied that the modelled
LUC was very likely underestimated and would require a more in-depth
investigation (Enveritas, 2023). LUC contribution is usually quite crit-
ical in agriculture-related LCA, particularly in the tropics where rain-
forest may be converted into agricultural land (Gibbs et al., 2008). It can
hence lead to signicant differences between coffee systems given
contrasted local development contexts and LUC history. Taking all the
studies into account, with and without LUC, the total GWP of green
coffee beans varied between 0.15 and 14.5 kg CO
2
eq./kg green coffee
beans, with a median value at 3.6 kg CO
2
eq./kg green coffee beans
(Fig. 3).
LUC apart (studied systems n =51), the main contributors to the
GWP of green coffee beans production were synthetic fertilisers with a
total median contribution of 66 % for both fertilisers upstream (manu-
facture and transport) and downstream (eld emissions; Fig. 3). In
studies providing disaggregated information, GHG emissions from fer-
tilisers upstream contributed 8–49 % to the impact (median value: 20 %)
and those from downstream emissions 18–58 % (median value: 37 %).
The second main contributor was wet processing due to the energy
used for processing and emissions from anaerobic treatment of waste-
water. Fermentation emissions related to wet processing were not
consistently modelled across studies, which raised a critical issue as they
were quite signicant contributors, from 27 % to 66 % of the total
impact (e.g. Killian et al., 2013; Maina et al., 2016; Van Rikxoort et al.,
2014). This contribution was highly variable and uncertain across
Fig. 3. Summary of global warming potential and main contributors; left: cradle to grave (studied systems n =141) and right: cradle-to-primary-processing stage
(studied systems n =68). [Export and instant coffee are excluded from the cradle-to-grave contributors and values, and the contribution of packaging would be
higher in the case of a single-serve pod. The value for green coffee beans is averaged for both dry and wet processing routes. Arrows indicate where overlapping and
higher uncertainty in contributions would be most critical. These overlapping explain why the total contributions exceed 100%, since not all studies disaggregated
packaging, brewing and washing & waste, some double-counting may be embedded in the displayed disaggregated contributions. Translucent overlapping lines
indicate standard deviations to the sample means by contributor.]
Credits: Coffee icons by Dong Gyu Yang (cup) and Muhammad Nur Auliady Pamungkas (beans), Noun Project CCBY3.0.
2
None of the four studies included indirect LUC. These studies referred to
IPCC (2006) to calculate LUC emissions but only one study mentioned the soil
organic carbon and did not provide any further details.
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
259
studies. Some studies did not include emissions from wet processing due
to a lack of data or provided no explanation (e.g. Brommer et al., 2011;
Quack et al., 2009; Vera-Acevedo et al., 2016; Rahmah et al., 2023). In
comparison, GHG emissions from dry processing and other post-harvest
operations up to green coffee beans were negligible. Drying was mostly
in the sun and hulling only contributed to 1–2 % when disaggregated
from other post-harvest energy-related contributors.
The third main contributor was N
2
O emissions from residues, albeit
those emissions were not systematically considered and led to quite
contrasted contributions. Overall, only nine studies considered some
emissions from crop residue decomposition in the eld and contribution
varied from 4 % in Maina et al. (2016) up to 90 % in Jaramillo et al.
(2017). In Noponen et al. (2012, 2013), emissions related to residue
decomposition contributed 9–42 % to the GWP across systems. Crop
residues in Maina et al. (2016) were not detailed. In the other eight
studies, all except one study considered emissions from coffee leaf
waste, but only ve studies also considered those from coffee pruning
and from litter and/or pruning from associated trees (Noponen et al.,
2012; Van Rikxoort et al., 2014; Van Rikxoort et al., 2013; Acosta-Alba
et al., 2020). The median contributions for those crop residues across
these studies were 11 % and 14 % at the mill and farm gate (no pro-
cessing), respectively. Finally, only one study also accounted for emis-
sions related to coffee husk decomposition (Enveritas, 2023). Some
other studies mentioned the application of coffee processing residues,
such as waste from de-pulping but without making it clear whether and
how related eld emissions were accounted for. Only one study explic-
itly mentioned emissions related to coffee waste-based compost pro-
duction and application (Acosta-Alba et al., 2019).
The contributions of other stages were less signicant. This includes
fuel for transport (median contribution of 3 % up to the primary pro-
cessing stage) or eld operations (median: 2 % up to the primary pro-
cessing stage), pesticides upstream emissions (median: 5 % at the farm
gate) and irrigation. Only 15 % of the studied systems (seven studies)
included irrigation and only three provided some detail on its GWP
contribution, which was highly variable (4–31 % up to the primary
processing stage). The results on irrigation contributions are unlikely to
be representative of contrasted types and intensities of irrigation across
coffee farming systems and are not further discussed here.
3.2.3.2. Main uncertainties in the estimations of global warming potential
of green coffee beans. At the plantation level, despite some mention of
quite complex coffee systems, such as agroforestry plots, little attention
was paid to this complexity and all potential ows. As indicated in
Fig. S2 in Supplementary information, there was no clear difference in
the GWP between different cropping system type, with large variabilities
within each type. This may be partly due to the fact that the dened
cropping farming types were not consistently discriminated against
fertiliser inputs that were highly variable across all systems and the main
contributor to the GWP. Some studies comparing extensive versus
intensive cropping systems based on different fertiliser strategies
showed a more contrasted impact across the systems (e.g. Basava-
lingaiah et al., 2022; Trinh et al., 2020; Noponen et al., 2012). However,
part of the discrepancies in the impact among the systems may have also
be missed due to incomplete descriptions and modelling of the diverse
system structures and functioning. For instance, some studies estimated
carbon stock in associated shade trees but did not investigate how
competitions for resources may affect inputs and outputs among crops
and trees and whether potential allocation issues would arise. Some
other studies mentioned the potential importance of ecosystem services
provided and how “shade system can also inuence production (yield,
quality and input efciency), environmental indicators and production
cost”, but that it was not accounted for (Brenes-Peralta et al., 2022).
Apart from a few studies mentioned before, there was a lack of a clear
systemic delineation between coffee and associated plants in the case of
agroforestry plots.
Although fertilisers-related eld emissions were explicitly modelled
across the studies (mostly based on IPCC (2006) and derivatives), there
was still a lack of transparency and details. Apart from a few studies (e.g.
Maina et al., 2016; Noponen et al., 2012), most studies did not specify if
indirect N
2
O emissions or CO
2
eld emissions related to urea and liming
practices were accounted for. Moreover, not all studies provided a
detailed inventory of inputs, notably of fertiliser types and amounts, nor
did they differentiate systematically between synthetic and organic
ones. The variability among studies with no details available spanned a
range as large as that of the other studies (Fig. 4). Hence, the lack of
transparency and details hampered a clear analysis of correlations. We
tried to disentangle the main contributing factors, analysing GWP by
considering different rates of nitrogen (N) application and splitting re-
sults based on primary processing type and inclusion or not of emissions
from residues. The N-rate classes were dened in order to yield com-
parable sample sizes across classes. However, we could not identify any
clear fertiliser-based tendency. Fig. 4 rst shows a large variability
across N-rate classes and no clear delineation in impacts between fer-
tiliser management. Some low-input systems had large emissions and
vice versa. Fertiliser management embeds many factors that could not be
disentangled and fully discriminated against due to the lack of details
available in the published studies. In particular, the difference between
organic versus synthetic fertilisers played a key role in some comparisons
of systems as upstream emissions from organic fertilisers were much
lower than those of synthetic ones, while fertiliser upstream emissions
were signicant contributors (e.g. Acosta-Alba et al., 2019; Noponen
et al., 2012). On the other hand, emissions from crop residues had a
signicant impact (Fig. 4a), so that more details on all residues or other
organic inputs to the eld and their emission proles would be para-
mount to assess fully the impact of different fertiliser types. The lack of
details on compost emissions, both up- and downstream, might be
particularly critical where conventional and organic cropping systems
were compared, with the latter mostly relying on compost instead of
synthetic mineral fertilisers (e.g. Trinh et al., 2020). Finally, large dis-
crepancies in emissions from wet processing may have also smoothed
out part of the comparative results across the N-rate classes. For
instance, in the “>334 N” class the wet-process coffee supply chains had
a much lower impact than both the dry process coffee supply chains
within the same N-rate class and the wet-process coffee supply chains
within lower N-rate classes (Fig. 4b).
Most of the studies did not consider emissions from coffee cultivation
residues and potentially other associated crops or trees, if the latter was
considered within the system boundary. It was mostly implicit, but in
some cases, authors specied that those emissions “were excluded
because of insufcient data” (e.g. Trinh et al., 2020). As previously
detailed, emissions from crop residues were identied as the third main
contributor to GWP; hence, their inclusion or exclusion clearly affected
the results (Figs. 4a and 5a). Also, depending on the processing chain,
further residues might be brought to the eld or wasted and lead to
further emissions in both cases. More attention should be paid to
quantifying on-farm or off-farm residue decomposition and emission
proles properly, so as to make sure that the quantication of emissions
is complete, as well as to check whether synthetic fertilisers were or
could be substituted. Studies focusing on coffee co-products or waste (e.
g. Catalan et al., 2019; Cruz, 2014; Dadi et al., 2019) could have pro-
vided insights on quantities and properties of those residues to enable a
more systematic accounting.
Finally, emissions from primary processing were also highly variable
and questionable (Fig. 5b). The clear split between dry processing for
arabica or robusta may relate to differences at the farm stage as dry
processing would not be signicantly different per se depending on the
coffee species. The clear split between dry and wet processing for
arabica is likely due to both large variabilities and uncertainties in the
modelling of both coffee supply chains. It stresses the likely underesti-
mation of emissions from the wet process, as not all studies included
wet-processing emissions nor used the most updated characterisation
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
260
factor for biogenic CH
4
. Emissions from wet processing depend on many
factors that can be highly variable, but mostly depend on the amount of
water used for washing and fermenting. The origin of water may also
affect the energy-related emissions for pumping. In the end, the amount
of emitted CH
4
is related to the wastewater amount and treatment that
differ widely among geographical contexts and applied processes.
Traditional full washing processes use up to four times as much water
compared to processes that reuse water (Van Rikxoort, 2011). CH
4
emissions, when included, were based on the IPCC (2006) (Chapter 5)
coefcients for wastewater treatment. But there was still a lack of
information on the overall process; i.e. detailing the origin of water (e.g.
energy impact for pumping), the amount of water used (dilution effect),
the duration of the whole process which inuences fermentation out-
puts, and the type and duration of wastewater treatments. More data and
knowledge would be needed to decipher the proper emission proles of
wet processes according to their specicities. Moreover, when waste-
water is not treated, which reduces the CH
4
emissions linked to the
treatment itself, other pollutants in the wastewater, such as reactive
organic compounds, may also become an environmental threat (Beyene
et al., 2012; Blinov´
a et al., 2017; Chanakya and De Alwis, 2004).
Fig. 4. Global warming potential of green coffee beans by applied N-rate classes and depending on the accounting for emissions from: (a) crop residues: no/yes; and
(b) the type of primary processing: dry/wet [Cradle-to-primary-processing gate, no LUC considered, studied systems n =50: one study is not displayed due to the rst
processing type not being discriminated. NA: details on applied N-rate not available].
Fig. 5. Global warming potential of green coffee beans depending on the accounting for emissions from: (a) crop residues (studied systems n =50: one study is not
displayed due to the rst processing type not being discriminated); and (b) type of primary processing: wet/dry (studied systems n =50) [Cradle-to-primary-
processing gate, no LUC considered].
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
261
However, downstream impact of wastewater discharged with or without
pre-treatment was not investigated in the reviewed studies.
3.3. Other impact categories
3.3.1. Terrestrial acidication, eutrophication, ecotoxicity and ozone
depletion
3.3.1.1. Overview of cradle-to-grave results. At the cradle-to-grave level,
comparison across studies was hampered by both varying functional
units and various impact assessment methods. In the two CML-based
studies (Büsser and Jungbluth, 2009; de Figueiredo Tavares and
Mourad, 2020), 12 coffee drinks were investigated with only two com-
mon ones, i.e. espresso and black coffee. Neither study used the same
functional unit nor considered the same level of detail, in particular
regarding the amount of consumed ground coffee. Common indicators
among these studies were energy use (see Section 3.3.2), eutrophication
and acidication. For the two last, the green coffee beans production
was the main contributor (40–99 %) across the scenarios considered.
The use of pesticides at the farm stage also contributed signicantly to
freshwater ecotoxicity and human toxicity, but as for the other in-
dicators, no comparison or conclusion was possible since no discussion
or details were provided in the papers. Overall, press-based and instant
coffee drinks tend to have lower impacts compared to pod coffee due to
waste management and relative to espresso due to the energy required
by the coffee machine.
The other two studies (Gosalvitr, 2021; Humbert et al., 2009) also
demonstrated the dominant contribution of the green coffee beans
production across the various impact categories. In the ReCiPe-based
study up to and including coffee consumption (Gosalvitr, 2021), green
coffee beans production contributed 85–99 % to the impact categories
across the drink types considered. The only exception was ionising ra-
diation which was mainly due to the consumption stage, related to nu-
clear energy in the electricity mix. The importance of the green coffee
stage was further emphasised when comparing roasting intensities and
black coffee drinks
3
, which affected the amount of coffee needed.
Overall, coffee transportation had negligible impacts and packaging
only mattered in the case of coffee pods and for other impact categories.
In the TRACI-based study (Hicks, 2018), the green coffee beans
production was also the main contributor to eutrophication, ecotoxicity,
and acidication. There were no signicant differences in impact con-
tributions across the three assessed drinks, i.e. drip lter, French press
and coffee pods, except for the added impacts related to the plastic cup
only used for the pod coffee. When focusing on eutrophication, the
impacts increased with the amount of ground coffee, conrming the
signicant contribution of the farm stage. Brewing was the second main
contributor across the impact categories and particularly to ozone
depletion and smog.
Despite the great variability across the studies in both the goals and
scopes and the applied impact assessment methods, the production of
green coffee beans remained a major contributor to coffee drink impacts,
notably eutrophication, acidication and ecotoxicity. Beyond variations
related to the drink type and the amount of ground coffee used, main
variations among studies concerned assumptions on waste, in terms of
both consumption patterns (e.g. amount of water or coffee waste, the use
of a cup) and disposal treatment (e.g. packaging). At the farm stage, as
for GWP, fertilisers were the main contributors to these impacts, except
for ecotoxicity which was mostly related to pesticide use, where applied.
When including primary processing, emissions from wet milling and
wastewater could add signicantly to eutrophication and ecotoxicity,
but there was an overall lack of details on this stage across the studies.
3.3.1.2. Overview of cradle-to-primary processing gate results. For the
cradle-to-primary-processing gate system boundary, across the ILCD-
and CML-based studies, the farm stage contributed in particular to
eutrophication and acidication. In the only study with primary data for
the farm stage (Acosta-Alba et al., 2020), the main impact sources were
the use of fertilisers and their manufacturing. Post-harvest operations
(wet mill located at the farm) had overall a lower contribution (median
contribution across acidication, terrestrial, freshwater and marine
eutrophication: 12 %), except for the less intensive system, for which
orders of magnitude of post-harvest contributions were similar to those
of upstream emissions from inputs. In-eld emissions from inputs were
the main contributors across impact indicators and cropping systems.
Their median contribution was 55 % (min: 3 % to max: 99 %). In
comparison, within on-farm impacts, weed management, compost use
and pesticides had no signicant impacts. The higher contribution of
manufacturing inputs came from nitrogen mineral fertilisers (median
contribution 17 %) for almost all categories and, in particular, terrestrial
acidication (up to 25 %) and ecotoxicity (up to 43 %). Within post-
harvest activities, the use of diesel for pulping machines contributed
from 1 % up to 36 % for the ILCD indicators (median 12 %). Fig. 6 shows
a qualitative synthesis of main contributors for this system boundary.
On-farm emissions and fertiliser manufacture also dominated the
ReCiPe-based results for terrestrial acidication, freshwater eutrophi-
cation and ne particulate matter formation (Basavalingaiah et al.,
2022), except for the organic farming system. The latter had very low
Fig. 6. Simplied overview of contributors to terrestrial acidication, eutro-
phication, ozone depletion, and ecotoxicity for the cradle-to-primary-processing
gate system boundary.
[Surface areas reect trends across systems and impact assessment methods
regarding main contributors and median contributions but are not quantita-
tively proportional to exact values, given the heterogeneity in details across
studies. Field operations (tractor icon) and wastewater pollution from pro-
cessing (water discharge icon) were the most variable contributors across
studies; therefore, their circle sizes are not proportional to the other lled or
hatched circles and could be larger as indicated by the question mark in the
outer circle. Field operations were signicant for ozone depletion in one study
(Hicks, 2018), whereas this impact was not investigated in Acosta-Alba et al.
(2020), for instance. Wastewater pollution due to primary processing water
discharge could matter, but was barely investigated and detailed across studies.
Hatched circles show that the contributions from upstream emissions during
production and transportation of inputs and in-eld emissions decreased, and
other contributors relatively increased, in the case of integrated or organic
coffee systems, whose synthetic inputs were limited].
Credits: Acid rain icon by Bartama Graphic; Polluted water drop for eutrophi-
cation by Nawicon; Ozone depletion icon by Good Wife; Skull from the ak by
zafdesign; Tractor by IronSV; Water discharge by Shashank Singh – Noun
Project CCBY3.0.
3
Drinks including milk were not considered here due to the added impacts
from the milk.
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
262
impacts for these indicators (per ha), although eutrophication from on-
farm emissions was higher than for the conventional and integrated
systems. The ecotoxicity results were dominated by fertiliser manufac-
ture, again with the exception of the organic system. Another study
applying ReCiPe (Trinh et al., 2020) had overall larger impacts from
conventional than from organic coffee cropping systems. Differences
were due to the lack of synthetic mineral fertilisers and “manual pest
control” in organic systems, leading to signicantly lower eutrophica-
tion, acidication and terrestrial ecotoxicity impacts. However, more
details on compost emissions and impact contributions would be needed
in that study as compost proved to create some trade-offs across impact
categories in Acosta-Alba et al. (2019). The dominant contribution of the
farm stage to freshwater eutrophication, freshwater ecotoxicity and
terrestrial acidication was also found in two other studies (Brenes-
Peralta et al., 2022; Ruben et al., 2018). In Brenes-Peralta et al. (2022),
contribution of the primary processing appeared not to be negligible,
contributing 5 % to eutrophication and 8 % to ecotoxicity. Those con-
tributions are likely due to the wastewater emissions from the wet mill
processing; the authors provided inventory data on wastewater (i.e.
biological oxygen demand and chemical oxygen demand) but did not
actually discuss the impacts.
In the TRACI-based study (Hicks, 2018) considering the production
of green coffee beans, fertilisers also played a major part in all the impact
categories but two, in particular terrestrial acidication for N-fertilisers
and eutrophication for P-fertilisers. Pesticide use in the coffee planta-
tions almost entirely dominated the ecotoxicity impacts. Fuel use
contributed to a few other impact categories, most signicantly to ozone
depletion (up to 88 %), although the study did not specify whether the
split between diesel and gasoline was related to the farm stage and the
primary processing. Overall, across all cradle-to-primary-processing
studies, i.e. up to and including green coffee beans production, contri-
bution of the primary processing to the impacts such as eutrophication
was detailed and discussed. Impacts linked to the energy used for pro-
cessing were sometimes detailed (i.e. Acosta-Alba et al., 2020), but
further impacts due to wastewater discharge in the wet process, for
instance, were not (Fig. 6).
3.3.2. Energy, mineral and fossil resource use impact indicators
Cumulative energy or primary energy demand were investigated in a
few studies as part of a multi-criteria LCA or as a complementary indi-
cator to global warming. Cumulative energy demand is widely used in
LCA but there have been various conceptual approaches (Frischknecht
et al., 2015). In some studies reviewed here, this energy indicator was a
life cycle inventory ow rather than an impact indicator; it could
sometimes be found in the life cycle inventory details of a study rather
than as a midpoint impact (which is arguably a correct approach,
although energy demand is widely reported as an “impact”). In other
cases, the impact assessment methods weighted the energy inventory
ows, depending on efciency conversions or energy sources, such as
renewables. Without any harmonised impact characterisation across
studies, the energy indicators could hardly be compared. In an Indone-
sian study, in particular, human labour was included in the cumulative
energy demand and was a main driver together with fertilisers (Rahmah
et al., 2023). Human labour, as well as animal traction, are commonly
excluded from LCA, which makes comparison with the other studies
difcult. There were no further studies detailing energy consumption
drivers at the farm level.
For the cradle-to-grave system boundary, results were extremely
variable. First, approaches varied and were not systematically detailed.
In one study, authors relied on CML 2001 but only provided life cycle
inventory-based indicators (de Figueiredo Tavares and Mourad, 2020).
Second, various drinks were studied and the system boundaries were not
all harmonised. For instance, the farm stage was not systematically fully
included; e.g. energy accounting started with cherry processing in Has-
sard et al. (2014). Hence, results varied from 0.02 to 0.12 MJ/g coffee in
drinks (mean: 0.05 MJ/g) among three studies (de Figueiredo Tavares
and Mourad, 2020; Domínguez-Pati˜
no et al., 2014; Hassard et al., 2014)
providing a life cycle inventory-based indicator and from 0.09 to 0.45
Non-renewable MJ eq./g coffee in drinks in the study based on CML
2001 (Büsser and Jungbluth, 2009) or 0.19 to 0.63 Non-renewable MJ
eq./g coffee in drinks in the study based on IMPACT 2002+(Humbert
et al., 2009). Here too we only looked at black coffee and espresso in
order not to add variability due to the impact of milk added to the
drinks. Main contributors were quite variable across studies. In studies
fully including the farm stage within the green coffee production, the
median contribution of green coffee production was around 40 %, with a
large variability (38 %–86 %) due to the system discrepancies and the
lack of details on post-harvest processing on- or off-farm. Nevertheless,
in all the studies green coffee production was among the three main
contributors. Energy used to heat the water or brew the coffee was either
the rst or second contributor and its contribution varied drastically
depending on the type of drinks (11–72 %), notably whether a coffee
machine or a kettle was used and how much hot water was used. In one
study, wet milling-based primary processing was the main contributor to
energy demand before brewing (Hassard et al., 2014).
Coffee roasting and packaging were not systematically dis-
aggregated. Overall, when disaggregated, packaging had a lower
impact, except in cases of pod coffee (e.g. 41–74 % in de Figueiredo
Tavares and Mourad (2020); 70 % in Büsser and Jungbluth (2009); 35 %
in Humbert et al. (2009)) or canned coffee (55 % in Hassard et al.
(2014)). As for GWP, the contribution of instant coffee processing to the
overall energy demand varied with assumptions on the amount of both
green and instant coffee compared to ground-coffee drinks (up to 35 %
in Hassard et al. (2014), where the instant processing was dis-
aggregated). There was not any further energy-demand impact assess-
ment for the cradle-to-primary-processing gate system boundary.
Mineral and fossil resource depletion indicators were included in
studies applying ReCiPe. These studies mostly ended with green coffee
production without much detail on the contribution proles (e.g. Trinh
et al., 2020), except for De Marco et al. (2018) and Gosalvitr (2021).
However, De Marco et al. focused on decaffeinated coffee so those re-
sults are not comparable to the other studies. Nevertheless, their study
exemplied the relative contribution of the decaffeination process,
whereby caffeine extraction and separation caused 30 % and 45 % of
mineral and fossil resource depletion, respectively, per kg of decaffein-
ated coffee. In Gosalvitr (2021), who considered both mineral and fossil
resource depletion and primary energy demand, green coffee was the
main contributor, followed by consumption across all drinks; packaging
was the second highest contributor to mineral resource depletion in the
case of pod coffee. The study by Gosalvitr also provided a detailed
analysis at the gate-to-gate level, investigating the disaggregated im-
pacts from freeze-drying and various roasting intensities. Those details
may be useful given the lack of details on these across the other studies.
It showed a 1.6-fold difference in energy use from light to dark roasting;
a 10-fold difference in energy used between roasting (higher) and
grinding (lower); and 6 to 9-fold difference in energy use between
roasting and freeze-drying.
For the three indicators considered in this section, the variability
across the studies in both the goal and scope and the applied impact
assessment methods had more inuence on how dominant the green
coffee production was. There were overall more details regarding energy
and resource used beyond the farm stage. Like the contribution analysis
for GWP, little information was provided on mechanised eld opera-
tions, such as weed control or harvesting, that would notably contribute
to fossil resource depletion. Manual harvesting is typical in many
countries (Illy and Viani, 2005), but there has also been a move towards
mechanisation (Adams and Ghaly, 2007). Besides differences in the
functional unit across studies, impacts related to energy use were also
dependent on varying background assumptions on electricity mixes,
electricity consumption of coffee machines and washing up practices
(Gosalvitr, 2021).
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
263
3.3.3. Water impact indicators
Water footprint was the most covered impact category after global
warming. Twelve studies investigated water ows or impacts, but
applying contrasted approaches. They covered various cropping systems
from various origins. Five studies included the coffee drink assessment, i.
e. for the cradle-to-grave system boundary. Although the ISO 14046
standard (ISO 14046, 2014) and its practical guide (ISO 14046, 2017)
provide guidance and clear denitions of all aspects related to water
footprint, a lot of studies did not use the correct terminology. In
particular, authors mostly referred to water “use” without specifying if it
was “consumption” or “withdrawal”. This lack of clarity led to some mis-
interpretation: e.g. an amount of water withdrawn for coffee processing
(Coltro et al., 2006) was interpreted as water consumption in some ar-
ticles (e.g. De Marco et al., 2018; De Figueiredo Tavares and Mourad,
2020). However, this amount of process water was actually fully
released as wastewater and was not consumed. This distinction is crucial
because only water consumption (i.e. evaporated, incorporated in the
product or transferred to another watershed) should be considered in the
impact assessment. According to the ISO standard, the withdrawn water
that is then released back in the same environment should not be part of
any water-scarcity impact category. It may, however, contribute to
water quality related impacts, such as eutrophication. Such a distinction
affects agricultural LCA results, notably in the case of irrigated systems
(Payen et al., 2018). In some cases (e.g. Ratchawat et al., 2020; Ruben
et al., 2018), the distinction between water withdrawal and consump-
tion was clearly made but the assumptions underlying the estimation of
the actual water consumed were unclear or not provided. Overall, there
was a great variability in primary data on water due to variabilities
across both the covered systems and the water modelling assumptions.
There was also a lack of clarity regarding the scope of the life cycle in-
ventory for water ows: water consumption for background activities (e.
g. farm input production) were likely not included in several studies.
When focusing on water consumption, irrigation was dominating at
both cradle-to-primary processing gate and cradle-to-grave levels,
despite the water consumed in the drinks (Fig. 7). Irrigation-water
amounts were highly variable, ranging from 293 to 5825 m
3
/t green
coffee across the four studies, including irrigated plantations. The sec-
ond contributor was wet processing and the actual amount of water
consumed varied signicantly depending on the process efciency.
Unless specied otherwise in the studies, we considered water con-
sumption as the difference between total water use and water dis-
charged. Wet-process water consumption ranged from 0 to 100 m
3
/t
green coffee (mean: 22.6) and wet-process water withdrawal ranged
from 11.4 to 15.2 m
3
/t green coffee (mean: 13.7). As observed in Coltro
et al. (2006), primary data from various sites showed a large variability
in the amount of wastewater from coffee washing during wet processing.
Depending on the process, large amount of this water may be either
discharged or recycled. Owing to the high volume of water potentially
used during wet processing, a clear water inventory for coffee processing
is crucial.
The few cradle-to-grave studies with detailed information on water
ows showed that, in between irrigation and wet processing, water use
in the consumption stage might not be negligible but was too variable to
be conclusive. However, even at the drink level, the main contributor
was related to the green coffee production and depended on the amount
of green coffee used per drink, which is directly related to the dilution
effect. As discussed for the GWP, when comparing studies, the functional
unit should account for organoleptic properties of coffee drinks or rely
on an equivalent amount of coffee used. This is even more critical for
water impact indicators, since irrigation (correlated with the amount of
coffee used) is the key driver. Also, assumptions on water used to pre-
pare drinks were quite contrasted across studies, varying from 39 to
8800 ml/functional unit or 1 to 37 ml/ml coffee (Chayer and Kicak,
2015; de Figueiredo Tavares and Mourad, 2020; Hassard et al., 2014;
Humbert et al., 2009). Discrepancies in water use and wastewater in the
coffee preparation stage, but also water used to wash coffee machines or
even cups may inuence the overall water use with contrasted distri-
butions between water consumption and withdrawal.
Characterising impacts from water consumption requires going
beyond a simple volumetric measure (i.e. an inventory ow) by
including relevant geographical and temporal dimensions to reect the
pressure on water resources. Only three studies characterised impacts by
accounting for local water scarcity: Humbert et al. (2009) and Acosta-
Alba et al. (2020), using the EcoScarcity method (2006), and Usva et al.
(2020) using the AWARE method (Boulay et al., 2018), the latter being
notably recommended by the Global Guidance on Environmental Life
Cycle Impact Assessment Indicators (UNEP and SETAC, 2019). Contri-
bution analysis showed that irrigation dominated impacts, followed by
coffee making and/or washing (Humbert et al., 2009; Usva et al., 2020).
When there was no irrigation, background processes, such as fertiliser
manufacture, were the main contributors (Acosta-Alba et al., 2020; Usva
et al., 2020). Across these studies, wet processing water did not matter as
it was considered only withdrawn and not consumed. Considering the
water scarcity level may change the relative contribution of life cycle
stages in comparison to the inventory data on water. For instance, the
water consumed in the use phase for coffee making had an increased
share of the overall impacts compared to the cultivation and processing
stages in Humbert et al. (2009). This was due to the fact that the level of
water stress was on average lower in the coffee producing countries
compared to the consuming countries.
It is also worth mentioning that LCA databases use different in-
terpretations of the water footprint indicator and that data are highly
variable. Therefore, using different databases will also affect this impact
signicantly. This, together with the above discussion, reinforces the
need for harmonisation of the water-related data and the estimation of
water impact indicators.
3.4. Further remarks
For all coffee impacts, a considerable range of results was found in
the literature, most notably for GWP, which varied by a factor of ten.
Fig. 7. Contributors to water consumption for coffee drinks (life cycle in-
ventory indicators) based on ten studies assessing water impacts with full or
partial data up to and including green coffee beans production (irrigation:
studied systems n =4; wet processing: studied systems n =7) and four studies
up to and including coffee drinks. [Water consumption ows are related to one
tonne of ground coffee consumed as drinks, hence including the irrigation
water, wet processing and water use at the consumption level. Based on the
included studies, we considered a green-to-ground coffee ratio of 1.24:1. Dis-
played values are median values across the ten studies, including two systems
with no irrigation compared to other systems within the same study. At the
plantation level, water consumption in the nursery stage was not included. At
the primary processing level, water withdrawal was not included (its amount
was similar to the water consumption). For each stage, water ows were highly
variable but only the median values were displayed due to graphical
constraints].
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
264
This is similar to the ndings in Poore and Nemecek (2018) who found
that impacts can vary by up to a factor of 50 for certain products. This is
largely due to differences in production practices, but also due to the
way LCA practitioners deal with more complex methodological issues,
such as allocation, LUC accounting and emission modelling related to
residues. The range for the other impacts was further inuenced by the
various impact assessment methods applied.
Overall, it was not possible to discriminate the quantied shares in
the variability of results that are due either to the intrinsic system
variability or to methodological discrepancies, given that both vari-
ability sources may have overlapped or could not be systematically
disentangled across studies. Nevertheless, we can assume that both
variability sources are of the same order of magnitude; for instance,
when looking at Fig. 5. Both sources of variability may compound to
split systems further apart (e.g. comparing rst two systems with crop
residues along the impact gradient due to methodological discrepancies,
and then comparing a system without crop residues to a system with
crop residues in the higher range of the impact in Fig. 5a) or compensate
(e.g. within the span of impact ranges among the systems with and
without crop residues in Fig. 5a). Also, looking at just one year of a
perennial crop cycle instead of accounting for the whole perennial cycle
would be a methodology-related variability, but nal difference in re-
sults would also be inuenced by the intrinsic variability within the
system due to changes in practices or climate conditions over time.
From an ISO 14040/14044 perspective, we noted that the choice of
the functional unit was not made sufciently specic. For the cradle-to-
grave boundary, functional units were problematic as they mostly did
not account for organoleptic properties. Most studies compared very
different coffee drinks in terms of coffee content and other ingredients,
such as sugar and milk, but without considering their different tastes or
potential effect as stimulant or other functions. The dilution effect,
which was inuential, was also not accounted for in the functional unit,
except for a few studies looking at various functional units. For the
cradle-to-farm-gate boundary, more than half of the studies used area
(ha⋅yr), which is not a functional unit as intended by the ISO standards.
Although a unit of area can “provide a reference to which the input and
output data are normalised” (ISO 14040, 2006), it does not reect the
function of the system and it does not allow for discriminating land-use
impacts in the impact assessment as they are correlated to the surface
area used, which is not justiable for “comprehensive” agriculture-
related LCA. In the case of more or less complex agroforestry coffee
systems in particular, the functional unit should be discussed in light of
the structure and functioning of the complete ecosystems. Surface area-
based functional units are often used in those cases to overcome allo-
cation issues and the lack of harmonised framework to account for
various functions at once, but it is a mere expedient. At least one study
referred to coffee-equivalent yield for coffee-pepper co-production sys-
tems but results for that unit were not provided, nor were the economic
values and the allocation fractions used to calculate it (Basavalingaiah
et al., 2022). Another one used a xed monetary output as the functional
unit, which would be worth developing further based on contrasting
economic and societal values of various ecosystem services (Acosta-Alba
et al., 2020). Finally, the lack of precision in the functional unit some-
times hid critical assumptions, such as whether primary processing
occurred on- or off-farm, the incremental processing and loss ratios, or
the moisture content of the nal product.
Also, in general, reporting of study assumptions, indicator choices
and results was not fully transparent, with impact results sometimes
only shown in a graphical form, or discussed partly in the text. It could
be argued that this is another breach of ISO 14044 (2006) requirements
that state “results […] shall be transparent and presented in sufcient
detail to allow the reader to comprehend the complexities and trade-offs
inherent in the LCA”. In this review, it prevented in some cases in-depth
analyses of results and comparability across studies.
4. Conclusions
Overall, the variability across coffee LCA studies reected the great
diversity in coffee systems, the diverging assumptions, and data quality
levels. In order to improve the robustness and accuracy of LCA of green
coffee, we recommend:
i) to consistently apply the IPCC (2006) guidelines for land use and
LUC accounting, i.e. clearly differentiating between long-term
storage of biogenic carbon, over at least 20 years, and short-
term biogenic carbon turnover, analysing transparently all car-
bon pools, including soil organic carbon;
ii) to model properly the perennial crop cycle, accounting for a
weighted average of inputs and outputs along the cycle,
depending on the various development stages;
iii) to quantify thoroughly all direct and indirect emissions in the
eld, including all amendments, mineral and organic, but also
crop residues; and
iv) to check the mass balance along the supply chain, also beyond the
plantation, in order to ensure that all co-products or wastes are
considered and their emissions from treatment, recycling or
disposal can be tracked.
At the coffee plantation level, more primary data would be needed in
order to i) to account better for the cropping system complexity and
interactions among crops within agroforestry systems; and ii) to char-
acterise better the emission proles from organic fertilisers, in particular
those derived from coffee co-products, such as husk-based co-compost.
At the primary processing level, more studies would be needed to
investigate the various processing routes, especially to uncover the po-
tential great diversity from small-scale artisanal up to industrial large-
scale processing for all three (wet, semi-wet and dry) routes. In partic-
ular, there is a critical lack of information and data to characterise all
potential impacts of wet processing, depending on the processing scale,
the fermentation duration, the amount of wastewater and the duration
and efciency of the treatment before discharge. More site- and process-
specic primary data would be needed to characterise better the GWP
but also other impacts, notably water scarcity impacts.
Impacts of coffee drinks primarily depend on the impacts of green
coffee. Hence the quality of the LCA of coffee drinks will mostly depend
on the inventory data used to characterise the green coffee impacts (at
least in the case of black coffee drinks without any added sugar or milk).
Therefore, even at the coffee drink level, it is highly recommended to use
primary inventory data for the cradle-to-primary-processing gate system
boundary and to avoid using too many proxies for the green coffee
suppliers based only on the country of origin and not considering the
technical specicities of the coffee plantations. However, we recognise
that such data are not widely available. Finally, the type of coffee drink
will also inuence the nal impacts. Therefore, the consumers' choices
may count notably in terms of relative impacts of the brewing method
and packaging. This stresses the need to account for organoleptic
properties within the functional unit, as those are the ones that ulti-
mately drive consumer choices. Organoleptic properties could be related
further to other aspects along the supply chains, from the coffee type and
origin up to the quality of preservation related to the packaging.
Beyond coffee LCA, it might be worth reecting on LCA practices in
general. Despite the harmonised norms and detailed guidelines, there
was overall a lack of transparency and details on the studied systems and
assumptions made, which may be due, at least partly, to space con-
straints in scientic publications. We would recommend that all details,
as required by the ISO standards, be provided in supplementary infor-
mation, if not in the paper. We also believe that data quality and asso-
ciated uncertainties should be assessed and discussed more
systematically as LCA results are no more meaningful than identifying
how improving knowledge and data quality might affect them. Scenario
testing can help to explore sources of uncertainty and the robustness of
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
265
results. However, it is paramount that scenarios be as realistic as
possible. At the cradle-to-grave level in particular, given the length of
the supply chain and the resulting large need for data, scenarios may
combine primary and secondary data sets that are not consistent. Ex-
amples include a proxy for a farming system that is too different from the
actual system, or a scenario that does not correlate changes in inputs and
outputs at the plot level. Given that the agricultural stage often con-
tributes signicantly to many impacts, even at the cradle-to-grave level,
we should bear in mind that agricultural systems are complex living
ecosystems, whose congurations may be extremely numerous and
proper characterisations lie in the detail.
CRediT authorship contribution statement
C. Ch´
eron-Bessou: Data curation, Formal analysis, Funding acqui-
sition, Investigation, Methodology, Project administration, Supervision,
Validation, Visualization, Writing – original draft, Writing – review &
editing. I. Acosta-Alba: Data curation, Investigation, Writing – original
draft. J. Boissy: Investigation. S. Payen: Data curation, Investigation,
Writing – original draft. C. Rigal: Data curation, Investigation, Writing –
original draft. A.A.R. Setiawan: Data curation, Investigation, Writing –
original draft. M. Sevenster: Data curation, Investigation, Writing –
original draft. T. Tran: Investigation. A. Azapagic: Investigation,
Writing – review & editing.
Declaration of competing interest
The authors declare the following nancial interests/personal re-
lationships which may be considered as potential competing interests:
C´
ecile Ch´
eron-Bessou reports nancial support was provided by Euro-
pean Commission and the Institute for Scientic Information on Coffee.
If there are other authors, they declare that they have no known
competing nancial interests or personal relationships that could have
appeared to inuence the work reported in this paper.
Acknowledgements
This work is part of the project Soil quality Assessment in Agriculture
For life cycle assessment-based Eco-design (SAAFE), funded by the Eu-
ropean Commission under the Global Individual Marie Skłodowska-
Curie Fellowship Grant Number: 843845. This work was co-funded by
ISIC, the Institute for Scientic Information on Coffee.
The authors warmly thank the three reviewers and the editor for
their very thorough and valuable comments on the manuscript. They
greatly contributed to improve the quality of this article.
Finally, the authors would like to pay a tribute to their colleague and
friend, Joachim Boissy, co-author of this work. Joachim was a beloved
beautiful person, he will not be forgotten.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.spc.2024.04.005.
References
Acosta-Alba, I., Chia, E., Andrieu, N., 2019. The LCA4CSA framework: using life cycle
assessment to strengthen environmental sustainability analysis of climate smart
agriculture options at farm and crop system levels. Agr. Syst. https://doi.org/
10.1016/j.agsy.2019.02.001.
Acosta-Alba, I., Boissy, J., Chia, E., Andrieu, N., 2020. Integrating diversity of
smallholder coffee cropping systems in environmental analysis. International
Journal of Life Cycle Assessment. https://doi.org/10.1007/s11367-019-01689-5.
Adams, M., Ghaly, A.E., 2007. Maximizing sustainability of the Costa Rican coffee
industry. Journal of Cleaner Production, From Material Flow Analysis to Material
Flow Management 15, 1716–1729. https://doi.org/10.1016/j.jclepro.2006.08.013.
Alemu, A., Dufera, E., 2017. Climate smart coffee (coffea arabica) production. American
Journal of Data Mining and Knowledge Discovery 2 (2), 62–68.
Aristiz´
abal-Marulanda, V., Garcia-Velasquez, C.A., Cardona Alzate, C.A., 2021a.
Environmental assessment of energy-driven bioreneries: the case of the coffee cut-
stems (CCS) in Colombia. International Journal of Life Cycle Assessment. https://
doi.org/10.1007/s11367-020-01855-0.
Aristiz´
abal-Marulanda, V., Solarte-Toro, J.C., Cardona Alzate, C.A., 2021b. Study of
bioreneries based on experimental data: production of bioethanol, biogas, syngas,
and electricity using coffee-cut stems as raw material. Environ. Sci. Pollut. Res. 28,
24590–24604. https://doi.org/10.1007/s11356-020-09804-y.
Basavalingaiah, K., Paramesh, V., Parajuli, R., Girisha, H.C., Shivaprasad, M.,
Vidyashree, G.V., Thoma, G., Hanumanthappa, M., Yogesh, G.S., Misra, S.D.,
Bhat, S., Irfan, M.M., Rajanna, G.A., 2022. Energy ow and life cycle impact
assessment of coffee-pepper production systems: an evaluation of conventional,
integrated and organic farms in India. Environ. Impact Assess. Rev. 92, 106687
https://doi.org/10.1016/j.eiar.2021.106687.
Bessou, C., Basset-Mens, C., Tran, T., Benoist, A., 2013. LCA applied to perennial
cropping systems: a review focused on the farm stage. Int. J. Life Cycle Assess. 18,
340–361. https://doi.org/10.1007/s11367-012-0502-z.
Bessou, C., Basset-Mens, C., Latunussa, C., V´
elu, A., Heitz, H., Vanni`
ere, H., Caliman, J.-
P., 2016. Partial modelling of the perennial crop cycle misleads LCA results in two
contrasted case studies. International Journal of Life Cycle Assessment 21, 297–310.
https://doi.org/10.1007/s11367-016-1030-z.
Beyene, A., Kassahun, Y., Addis, T., Assefa, F., Amsalu, A., Legesse, W., Kloos, H.,
Triest, L., 2012. The impact of traditional coffee processing on river water quality in
Ethiopia and the urgency of adopting sound environmental practices. Environ.
Monit. Assess. 184, 7053–7063. https://doi.org/10.1007/s10661-011-2479-7.
Birkenberg, A., Birner, R., 2018. The world’s rst carbon neutral coffee: Lessons on
certication and innovation from a pioneer case in Costa Rica. J. Clean. Prod. 189,
485–501. https://doi.org/10.1016/j.jclepro.2018.03.226.
Blinov´
a, L., Sirotiak, M., Bartoˇ
sov´
a, A., Sold´
an, M., 2017. Review: utilization of waste
from coffee production. Research Papers Faculty of Materials Science and
Technology Slovak University of Technology 25, 91–101. https://doi.org/10.1515/
rput-2017-0011.
Boulay, A.-M., Bare, J., Benini, L., Berger, M., Lathuilli`
ere, M.J., et al., 2018. The WULCA
consensus characterization model for water scarcity footprints: assessing impacts of
water consumption based on available water remaining (AWARE). Int. J. Life Cycle
Assess. 23 (2), 368–378. https://doi.org/10.1007/s11367-017-1333-8.
Brenes-Peralta, L., De Menna, F., Vittuari, M., 2022. Interlinked driving factors for
decision-making in sustainable coffee production. Environ. Dev. Sustain. https://doi.
org/10.1007/s10668-022-02821-6.
Brommer, E., Stratmann, B., Quack, D., 2011. Environmental impacts of different
methods of coffee preparation. Int. J. Consum. Stud. 35, 212–220. https://doi.org/
10.1111/j.1470-6431.2010.00971.x.
BSI, 2012. PAS 2050-1:2012 Assessment of life Cycle Greenhouse Gas Emissions From
Horticultural Products, 46.
BSI (British Standards Institution), 2011. The Guide to PAS 2050:2011: How to Carbon
Footprint Your Products, Identify Hotspots and Reduce Emissions in Your Supply
Chain. London, UK, p. 79 (ISBN 978-0-580-77432-4).
Büsser, S., Jungbluth, N., 2009. The role of exible packaging in the life cycle of coffee
and butter. International Journal of Life Cycle Assessment 14, 80–91. https://doi.
org/10.1007/s11367-008-0056-2.
Catalan, E., Komilis, D., Sanchez, A., 2019. Environmental impact of cellulase production
from coffee husks by solid-state fermentation: a life-cycle assessment. J. Clean. Prod.
https://doi.org/10.1016/j.jclepro.2019.06.100.
Cerutti, A.K., Beccaro, G.L., Bruun, S., Bosco, S., Donno, D., Notarnicola, B., Bounous, G.,
2013. LCA application in the fruit sector: state of the art and recommendations for
environmental declarations of fruit products. J. Clean. Prod. https://doi.org/
10.1016/j.jclepro.2013.09.017.
Chanakya, H.N., De Alwis, A.A.P., 2004. Environmental issues and management in
primary coffee processing. Process Saf. Environ. Prot. 82, 291–300. https://doi.org/
10.1205/095758204323162319.
Chayer, J.-A., Kicak, K., 2015. Life Cycle Assessment of Coffee Consumption: Comparison
of Single-Serve Coffee and Bulk Coffee Brewing. Quantis, Montreal, Canada.
Coltro, L., Mourad, A., Oliveira, P., Baddini, J., Kletecke, R., 2006. Environmental prole
of Brazilian green coffee. Int. J. Life Cycle Assess. https://doi.org/10.1065/
lca2006.01.230.
Conservation International, 2020. The Sustainable Coffee Challenge Sets Ambitious 2050
Climate Goal. retrieved from webpage: https://www.conservation.org/press-releas
es/2020/12/21/the-sustainable-coffee-challenge-sets-ambitious-2050-climate-goal.
Cruz, R., 2014. Coffee by-products: Sustainable Agro-Industrial Recovery and Impact on
Vegetables Quality. Facultade de Farmacia, Master II, Universidade de Porto, Porto
(117p).
Dadi, D., Daba, G., Beyene, A., Luis, P., Van der Bruggen, B., 2019. Composting and co-
composting of coffee husk and pulp with source-separated municipal solid waste: a
breakthrough in valorization of coffee waste. Int J Recycl Org Waste Agricult 8,
263–277. https://doi.org/10.1007/s40093-019-0256-8.
De Figueiredo Tavares, M.P., Mourad, A.L., 2020. Coffee beverage preparation by
different methods from an environmental perspective. International Journal of Life
Cycle Assessment 25, 1356–1367. https://doi.org/10.1007/s11367-019-01719-2.
De Marco, I., Riemma, S., Iannone, R., 2018. Life cycle assessment of supercritical CO2
extraction of caffeine from coffee beans. J. Supercrit. Fluids. https://doi.org/
10.1016/j.supu.2017.11.005.
Djufry, F., Wulandari, S., Villano, R., 2022. Climate smart agriculture implementation on
coffee smallholders in Indonesia and strategy to accelerate. Land 11, 1112. https://
doi.org/10.3390/land11071112.
C. Ch´
eron-Bessou et al.

Sustainable Production and Consumption 47 (2024) 251–266
266
Domínguez-Pati˜
no, J., Martínez, A.R., Romero, R.J., Orozco, I.H., 2014. Life cycle
assessment on real time in a coffee machine. J. Chem. Eng. https://doi.org/
10.17265/1934-7375/2014.12.007.
Enveritas, 2023. Establishing carbon footprint baselines for Robusta coffee production in
two origins in Southeast Asia: Central Highlands in Vietnam and Southern Sumatra
in Indonesia. In: Lead Implementing Partner. Enveritas.
EU, 2023. Position of the European Parliament adopted at rst reading on 19 April 2023
with a view to the adoption of Regulation (EU) 2023/1115 of the European
Parliament and of the Council on the making available on the Union market and the
export from the Union of certain commodities and products associated with
deforestation and forest degradation and repealing Regulation (EU) No 995/2010.
FAO, 2023. Markets and Trade, Commodities, Coffee. webpage. https://www.fao.org/m
arkets-and-trade/commodities/coffee/en/.
Frischknecht, R., Wyss, F., Büsser Kn¨
opfel, S., Lützkendorf, T., Balouktsi, M., 2015.
Cumulative energy demand in LCA: the energy harvested approach. International
Journal of Life Cycle Assessment 20, 957–969. https://doi.org/10.1007/s11367-
015-0897-4.
Gabiri, G., Luswata, K.C., Sebuliba, E., Nampijja, J., 2022. Climate Smart Agriculture in
Uganda (Report). Accelerating Impacts of CGIAR Climate Research for Africa.
Gibbs, H.K., Johnston, M., Foley, J.A., Holloway, T., Monfreda, C., Ramankutty, N.,
Zaks, D., 2008. Carbon payback times for crop-based biofuel expansion in the
tropics: the effects of changing yield and technology. Environ. Res. Lett. 3, 034001
https://doi.org/10.1088/1748-9326/3/3/034001.
Gobbi, L., Maddaloni, L., Prencipe, S., Vinci, G., 2023. Bioactive compounds in different
coffee beverages for quality and sustainability assessment. Beverages 9. https://doi.
org/10.3390/beverages9010003.
Gosalvitr, P., 2021. Life Cycle Environmental and Economic Sustainability of Energy and
Resource Recovery Options in the Food and Drink Sector in the UK. PhD thesis,
University of Manchester, Manchester, p. 430.
Gosalvitr, P., Cuellar-Franca, R., Smith, R., Azapagic, A., 2023. An environmental and
economic sustainability assessment of coffee production in the UK. Chem. Eng. J.
465 https://doi.org/10.1016/j.cej.2023.142793.
Grüter, R., Trachsel, T., Laube, P., Jaisli, I., 2022. Expected global suitability of coffee,
cashew and avocado due to climate change. PloS One 17, e0261976. https://doi.
org/10.1371/journal.pone.0261976.
Haggar, J., Barrios, M., Bola˜
nos, M., Merlo, M., Moraga, P., Munguia, R., Ponce, A.,
Romero, S., Soto, G., Staver, C., de M. F. Virginio, E., 2011. Coffee agroecosystem
performance under full sun, shade, conventional and organic management regimes
in Central America. Agroforestry Systems 82, 285–301. https://doi.org/10.1007/
s10457-011-9392-5.
Hassard, H.A., Couch, M.H., Techa-erawan, T., McLellan, B.C., 2014. Product carbon
footprint and energy analysis of alternative coffee products in Japan. Journal of
Cleaner Production 73, 310–321. https://doi.org/10.1016/j.jclepro.2014.02.006.
Towards eco-efcient agriculture and food systems: Selected papers from the Life
Cycle Assessment (LCA) Food Conference, 2012, in Saint Malo, France.
Hejna, A., 2021. Potential applications of by-products from the coffee industry in
polymer technology – current state and perspectives. Waste Manag. 121, 296–330.
https://doi.org/10.1016/j.wasman.2020.12.018.
Hicks, A.L., 2018. Environmental implications of consumer convenience: coffee as a case
study. J. Ind. Ecol. 22, 79–91. https://doi.org/10.1111/jiec.12487.
Humbert, S., Loerincik, Y., Rossi, V., Margni, M., Jolliet, O., 2009. Life cycle assessment
of spray dried soluble coffee and comparison with alternatives (drip lter and
capsule espresso). J. Clean. Prod. https://doi.org/10.1016/j.jclepro.2009.04.011.
ICO, 2023. Coffee Report and Outlook. International Coffee Organization. December
2023. (43p.).
Illy, A., Viani, R., 2005. Espresso Coffee: The Science of Quality. Academic Press.
IPCC, 2006. In: Eggleston, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006
IPCC Guidelines for National Greenhouse Gas Inventories, Prepared by the National
Greenhouse Gas Inventories Programme. IGES, Japan. Published.
IPCC, 2007. In: Core Writing Team, Pachauri, R.K., Reisinger, A. (Eds.), Climate Change
2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change. IPCC,
Geneva, Switzerland (104 pp.).
IPCC, 2014. In: Core Writing Team, Pachauri, R.K., Meyer, L.A. (Eds.), Climate Change
2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth
Assessment Report of the Intergovernmental Panel on Climate Change. IPCC,
Geneva, Switzerland (151 pp.).
ISO 14040, 2006. Environmental Management – Life Cycle Assessment – Principles and
Framework. International Organization for Standardization, Geneva.
ISO 14044, 2006. Environmental Management – Life Cycle Assessment – Requirements
and Guidelines. International Organisation for Standardisation, Geneva.
ISO 14046, 2014. Environmental Management—Water Footprint—Principles,
Requirements and Guidelines. ISO, Geneva, Switzerland.
ISO 14046, 2017. Environmental Management—Water Footprint—A Practical Guide for
SMEs. ISO, Geneva, Switzerland.
Jansen, B.J., Spink, A., 2006. How are we searching the World Wide Web? A comparison
of nine search engine transaction logs. Inf. Process. Manag. 42, 248–263.
Jaramillo, S., Saraz, J.A.O., Correa, G., 2017. Emission and xation of greenhouse gases
in potential specialty coffee production zones in Antioquia -Colombia. Revista
Facultad Nacional de Agronomía Medellín 70, 8341–8349. https://doi.org/
10.15446/rfna.v70n3.62639.
Killian, B., Rivera, L., Soto, M., Navichoc, D., 2013. Carbon Footprint across the Coffee
Supply Chain: The Case of Costa Rican Coffee.
Maina, J.J., Mutwiwa, U.N., Engineering, G.M.K., M.G. 1Department of A. and B.E.S. of
B. systems and E., Technology, J.K.U. of A. and, Works, P.O.B. 62000-00200 N.
2Wildlife, Voi, P.O.B. 310-80300, 2016. Evaluation of Greenhouse Gas Emissions
along the Small-Holder Coffee Supply Chain in Kenya.
Mussatto, S.I., Machado, E.M.S., Martins, S., Teixeira, J.A., 2011. Production,
composition, and application of coffee and its industrial residues. Food Bioprocess
Technology 4, 661–672. https://doi.org/10.1007/s11947-011-0565-z.
Nab, C., Maslin, M., 2020. Life cycle assessment synthesis of the carbon footprint of
Arabica coffee: case study of Brazil and Vietnam conventional and sustainable coffee
production and export to the United Kingdom. Geo: Geography and Environment 7,
e00096. https://doi.org/10.1002/geo2.96.
Noponen, M.R.A., Edwards-Jones, G., Haggar, J.P., Soto, G., Attarzadeh, N., Healey, J.R.,
2012. Greenhouse gas emissions in coffee grown with differing input levels under
conventional and organic management. Agr. Ecosyst. Environ. 151, 6–15. https://
doi.org/10.1016/j.agee.2012.01.019.
Noponen, M.R.A., Haggar, J.P., Edwards-Jones, G., Healey, J.R., 2013. Intensication of
coffee systems can increase the effectiveness of REDD mechanisms. Agr. Syst. 119,
1–9. https://doi.org/10.1016/j.agsy.2013.03.006.
Payen, S., Falconer, S., Ledgard, S.F., 2018. Water scarcity footprint of dairy milk
production in New Zealand – a comparison of methods and spatio-temporal
resolution. Sci. Total Environ. 639, 504–515. https://doi.org/10.1016/j.
scitotenv.2018.05.125.
Poore, J., Nemecek, T., 2018. Reducing food’s environmental impacts through producers
and consumers. Science 360, 987–992. https://doi.org/10.1126/science.aaq0216.
Quack, D., Eberle, U., Liu, R., Stratmann, B., 2009. Case Study Tchibo Privat Kaffee
Rarity Machare by Tchibo Gmbh. PCF Pilot Project Germany, Berlin.
Rahmah, D., Mardawati, E., Kastaman, R., Pujianto, T., Pramulya, R., 2023. Coffee pulp
biomass utilization on coffee production and its impact on energy saving, CO2
emission reduction, and economic value added to promote green lean practice in
agriculture production. Agronomy-Basel 13. https://doi.org/10.3390/
agronomy13030904.
Rahn, E., L¨
aderach, P., Baca, M., Cressy, C., Schroth, G., Malin, D., Van Rikxoort, H.,
Shriver, J., 2014. Climate change adaptation, mitigation and livelihood benets in
coffee production: where are the synergies? Mitig Adapt Strateg Glob Change 19,
1119–1137. https://doi.org/10.1007/s11027-013-9467-x.
Ratchawat, T., Panyatona, S., Nopchinwong, P., Chidthaisong, A., Chiarakorn, S., 2020.
Carbon and water footprint of Robusta coffee through its production chains in
Thailand. Environ. Dev. Sustain. 22, 2415–2429. https://doi.org/10.1007/s10668-
018-0299-4.
Rega, F.V., Ferranti, P., 2019. Life cycle assessment of coffee production in time of global
change. Encyclopedia of Food Security and Sustainability. 497–502.
RStudio Team, 2023. Developer: Posit, Joseph J. Allaire, License: GNU Affero General
Public License v3. https://posit.co/products/open-source/rstudio/.
Ruben, R., Allen, C., Boureima, F., Mhando, D.G., Dijkxhoorn, Y., 2018. Coffee value
chain analysis in the southern highlands of Tanzania. In: Report for the European
Commission, DG-DEVCO. Value Chain Analysis for Development Project (VCA4D
CTR 2016/375-804), p. 135. +annexes.
Statistica, 2023. Coffee production worldwide from 2003/04 to 2021/22 (in million 60
kilogram bags). In: Data for Coffee Production by All Exporting Countries Retrieved
From the Webpage. https://www.statista.com/statistics/263311/worldwide-pro
duction-of-coffee/ (viewed on 2023.12.12).
Trinh, L.T.K., Hu, A.H., Lan, Y.C., Chen, Z.H., 2020. Comparative life cycle assessment
for conventional and organic coffee cultivation in Vietnam. Int. J. Environ. Sci.
Technol. https://doi.org/10.1007/s13762-019-02539-5.
UNEP and SETAC, 2019. In: Frischknecht, R., Jolliet, O. (Eds.), Global Guidance on
Environmental Life Cycle Impact Assessment Indicators, 2, p. 202.
Usva, K., Sinkko, T., Silvenius, F., Riipi, I., Heusala, H., 2020. Carbon and water footprint
of coffee consumed in Finland-life cycle assessment. International Journal of Life
Cycle Assessment. https://doi.org/10.1007/s11367-020-01799-5.
Van Rikxoort, H., 2011. The Potential of Mesoamerican Coffee Production Systems to
Mitigate Climate Change. In: Climate Change Management. Springer Berlin
Heidelberg, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-31110-9_43.
Van Rikxoort, H., L¨
aderach, P., Van Hal, J., 2013. The potential of Latin American coffee
production systems to mitigate climate change. In: Leal Filho, W. (Ed.), Climate
Change and Disaster Risk Management, Climate Change Management. Springer
Berlin Heidelberg, Berlin, Heidelberg, pp. 655–679. https://doi.org/10.1007/978-3-
642-31110-9_43.
Van Rikxoort, H., Schroth, G., L¨
aderach, P., Rodríguez-S´
anchez, B., 2014. Carbon
footprints and carbon stocks reveal climate-friendly coffee production. Agronomy
and Sustainable Development. 34, 887–897. https://doi.org/10.1007/s13593-014-
0223-8.
Vera-Acevedo, L.D., V´
elez-Henao, J.A., Marulanda-Grisales, N., 2016. Assessment of the
environmental impact of three types of fertilizers on the cultivation of coffee at the
Las Delicias indigenous reservation (Cauca) starting from the life cycle assessment.
Revista Facultad de Ingeniería Universidad de Antioquia 93–101. https://doi.org/
10.17533/udea.redin.n81a09.
Viere, T., von Enden, J., Schaltegger, S., 2011. Life cycle and supply chain information in
environmental management accounting: a coffee case study. Environmental
Management Accounting and Supply Chain Management, Eco-Efciency in Industry
and Science. https://doi.org/10.1007/978-94-007-1390-1_2.
Vignoli, J.A., Viegas, M.C., Bassoli, D.G., Benassi, M.D.T., 2014. Roasting process affects
differently the bioactive compounds and the antioxidant activity of arabica and
robusta coffees. Food Res. Int. 61, 279–285. https://doi.org/10.1016/j.
foodres.2013.06.006.
Youkhana, A.H., Idol, T.W., 2016. Leucaena-KX2 mulch additions increase growth, yield
and soil C and N in a managed full-sun coffee system in Hawaii. Agrofor. Syst. 90,
325–337. https://doi.org/10.1007/s10457-015-9857-z.
C. Ch´
eron-Bessou et al.