Contents lists available at ScienceDirect
Global Environmental Change
journal homepage: www.elsevier.com/locate/gloenvcha
Country-speciﬁc dietary shifts to mitigate climate and water crises
Brent F. Kim
, Raychel E. Santo
, Allysan P. Scatterday
, Jillian P. Fry
, Colleen M. Synk
Shannon R. Cebron
, Mesﬁn M. Mekonnen
, Arjen Y. Hoekstra
, Saskia de Pee
Martin W. Bloem
, Roni A. Neﬀ
, Keeve E. Nachman
Johns Hopkins Center for a Livable Future, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, 21202, United States
Department of Environmental Health & Engineering, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, 21205, United States
Department of Health, Behavior and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, 21205, United States
Department of Health Sciences, Towson University, Towson, MD, 21252, United States
Robert B. Daugherty Water for Food Global Institute, University of Nebraska, Lincoln, NE, 68508, United States
University of Twente, 7522 NB, Enschede, Netherlands
Lee Kuan Yew School of Public Policy, National University of Singapore, Singapore, 259772, Singapore
United Nations World Food Programme, Rome, 00148, Italy
Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, 21205, United States
Risk Sciences and Public Policy Institute, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, 21205, United States
Greenhouse gas emissions
Undernutrition, obesity, climate change, and freshwater depletion share food and agricultural systems as an
underlying driver. Eﬀorts to more closely align dietary patterns with sustainability and health goals could be
better informed with data covering the spectrum of countries characterized by over- and undernutrition. Here,
we model the greenhouse gas (GHG) and water footprints of nine increasingly plant-forward diets, aligned with
criteria for a healthy diet, speciﬁc to 140 countries. Results varied widely by country due to diﬀerences in:
nutritional adjustments, baseline consumption patterns from which modeled diets were derived, import patterns,
and the GHG- and water-intensities of foods by country of origin. Relative to exclusively plant-based (vegan)
diets, diets comprised of plant foods with modest amounts of low-food chain animals (i.e., forage ﬁsh, bivalve
mollusks, insects) had comparably small GHG and water footprints. In 95 percent of countries, diets that only
included animal products for one meal per day were less GHG-intensive than lacto-ovo vegetarian diets (in
which terrestrial and aquatic meats were eliminated entirely) in part due to the GHG-intensity of dairy foods.
The relatively optimal choices among modeled diets otherwise varied across countries, in part due to con-
tributions from deforestation (e.g., for feed production and grazing lands) and highly freshwater-intensive forms
of aquaculture. Globally, modest plant-forward shifts (e.g., to low red meat diets) were oﬀset by modeled in-
creases in protein and caloric intake among undernourished populations, resulting in net increases in GHG and
water footprints. These and other ﬁndings highlight the importance of trade, culture, and nutrition in diet
footprint analyses. The country-speciﬁc results presented here could provide nutritionally-viable pathways for
high-meat consuming countries as well as transitioning countries that might otherwise adopt the Western dietary
Undernutrition, obesity, and climate change have been described as
a synergy of pandemics (Swinburn et al., 2019). Together with fresh-
water depletion and other related ecological harms, these intersecting
global challenges share food and agricultural systems as an underlying
driver. Leveraging those patterns presents an opportunity to address
multiple challenges in tandem, with an eye toward avoiding the
unintended consequences of making progress in some areas at the ex-
pense of others. For many low- and middle-income countries, for ex-
ample, messaging about sustainable diets is complicated by a persistent
high prevalence of all forms of undernutrition (Development Initiatives,
2018). Accounting for these and other factors at a country-speciﬁc level
could help inform eﬀorts among high-meat consuming countries to
better align diets with public health and ecological goals, while pro-
viding nutritionally-viable strategies for transitioning countries that
Received 13 June 2018; Received in revised form 14 May 2019; Accepted 19 May 2019
Corresponding authors at: Johns Hopkins Center for a Livable Future, 111 Market Place, Suite 840, Baltimore, MD, 21202, United States.
E-mail addresses: rneﬀ1@jhu.edu (R.A. Neﬀ), email@example.com (K.E. Nachman).
Global Environmental Change xxx (xxxx) xxxx
0959-3780/ © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
Please cite this article as: Brent F. Kim, et al., Global Environmental Change, https://doi.org/10.1016/j.gloenvcha.2019.05.010
might otherwise adopt the Western dietary pattern, particularly among
their urban population.
Shifts toward plant-forward diets are essential for meeting climate
change mitigation targets (Bajzelj et al., 2014;Bryngelsson et al., 2016;
Hedenus et al., 2014) and remaining within planetary boundaries
(Willett et al., 2019). These and other concerns have fueled eﬀort-
s—proposed and enacted—to reduce animal product consumption
through approaches including behavior change campaigns (de Boer
et al., 2014d;Morris et al., 2014), environmental impact labeling
(Leach et al., 2016), dietary recommendations (Fischer and Garnett,
2016), and taxes (Säll and Gren, 2015;Springmann et al., 2017;
Wirsenius et al., 2011). At the same time, animals raised for food can
provide a range of agro-economic beneﬁts, including converting in-
edible crop residues and by-products into human-edible food, and uti-
lizing the share of grassland unsuitable for crop production (Mottet
et al., 2017). Furthermore, animal-source foods are a valuable source of
protein and bioavailable micronutrients, especially for young children
(de Pee and Bloem, 2009d;Semba, 2016;Swinburn et al., 2019).
Policy and behavioral interventions aimed at promoting sustainable
diets could be better informed with evidence about where they could
oﬀer the greatest potential beneﬁts, the nutritional status of diﬀerent
populations, and the relative environmental impacts of each diet in
each country. Previous studies documenting ecological impacts of
dietary scenarios have called for greater geographic speciﬁcity
(Aleksandrowicz et al., 2016;Jones et al., 2016), as most have ex-
amined only one or a few—almost exclusively industrialized—
countries, or a regional or global aggregate (Appendix A, Table A1).
To help address these gaps, we modeled the greenhouse gas (GHG)
footprint and blue and green water footprint (WF) of baseline con-
sumption patterns and nine increasingly plant-forward diets with
varying levels of animal products for 140 individual countries and
territories (henceforth: “countries”). Diets were modeled in accordance
with health criteria, oﬀering nutritionally-viable scenarios (to the ex-
tent possible without accounting for micronutrients) that adjust for
over- and under-consumption. We account for blue water (surface and
groundwater, e.g., for irrigation) and green water (soil moisture from
precipitation); the latter is often excluded from similar studies on the
rationale that it does not directly impact water scarcity (e.g., by de-
pleting aquifers). Green water accounting is important, however, be-
cause eﬃcient use of green water in rainfed agriculture can lessen re-
liance on blue water elsewhere. In an internationally-traded economy,
one cannot be considered independently of the other, and both are part
of an increasingly scarce global pool (Hoekstra, 2016;Schyns et al.,
2019). We also incorporate footprints of aquatic animals, nuts, and
seeds—common protein alternatives to terrestrial animal products—-
which most prior studies excluded or only narrowly considered (Ap-
pendix A, Table A1).
By accounting for import patterns and associated diﬀerences in the
GHG and water footprints of food items based on the production
practices unique to items’ countries of origin (COO), the study model
satisﬁes recent appeals (Heller and Keoleian, 2015;Wellesley et al.,
2015) to incorporate trade ﬂows when measuring the environmental
impacts associated with national consumption patterns. Moreover, in-
ternational accounting systems commonly attribute environmental
impacts associated with imported foods to producing countries rather
than the countries in which they are consumed, thereby displacing
accountability away from the populations responsible for changing
demand (Dario et al., 2014;de Ruiter et al., 2016;Peters and Hertwich,
This research identiﬁes a range of country-speciﬁc scenarios in
which dietary patterns could better align with climate change mitiga-
tion, freshwater conservation, and nutrition guidelines.
We developed a model to estimate the annual per capita and whole
country GHG, blue water, and green water footprints for baseline
consumption patterns and nine increasingly plant-forward diets speciﬁc
to 140 countries. We also estimate the per-serving, per-kilocalorie, per-
gram of protein, and per-kilogram edible weight footprints of common
food groups. The model was developed in Python version 3.6. Model
input and output are available in Mendeley Data (Kim et al., 2019).
2.1. Baseline consumption patterns
To characterize baseline consumption patterns for each country, we
averaged data over the 2011–2013 Food and Agriculture Organization
of the United Nations (FAO) food balance sheets (FBS) (FAO, 2017a),
which provide estimates of per capita domestic food supplies after ac-
counting for imports, exports, losses (where data are available), animal
feed, and other non-food uses (FAO, 2001). Quantities reported in FBS
reﬂect food availability and thus overestimate quantities actually con-
sumed. Bovine meat supplies, for example, are reported in dressed
carcass weight, which includes bones and other parts typically con-
sidered inedible. These data are appropriate for diet footprint modeling,
however, because they reﬂect the amount of production involved in
feeding populations (e.g., we measure the footprint of the carcass re-
quired to produce the edible portion of beef in the diet). Food balance
sheets are also well-suited for comparing consumption patterns across
countries (Fehrenbach et al., 2016) and have precedent in the literature
for measuring diet footprints across regions (Hedenus et al., 2014;Popp
et al., 2010;Pradhan et al., 2013;Tukker et al., 2011) and globally
(Bajzelj et al., 2014;Stehfest et al., 2009;Tilman and Clark, 2014).
2.2. Food losses and waste
For some items in some countries, where suﬃcient data were
available, FBS subtracted supply chain losses from food supply esti-
mates. We added these quantities back in to food supplies for two
reasons: First, estimates of diet footprints should reﬂect the fact that
some amount of waste inevitably occurs between the producer and the
consumer, thus for footprint modeling purposes we needed the original
quantities of FBS items prior to supply chain losses. Second, in cases
where it was appropriate to subtract supply chain losses—i.e., when
dealing with amounts of calories or nutrients actually consumed
(Section 2.4)—we used a more comprehensive source for food losses
and waste (Gustavsson et al., 2011); combining this with FBS estimates
would have resulted in double-counting. Detailed methods for esti-
mating food losses and waste are provided in Appendix B.1.
2.3. Food items
Study diets were comprised of 74 items in FBS (Mendeley Data
input/item_parameters). Twenty-four additional FBS items were ex-
cluded due to the small quantities in which they are typically consumed
(e.g., spices), limited footprint data (e.g., game meats), and/or because
they are not typically considered food (e.g., alcohols, cottonseed). Most
FBS items are expressed in terms of primary equivalents, i.e., the
quantity of a raw commodity required to produce a given quantity of
processed goods. For example, wheat products (e.g., wheat ﬂour and
bread) are quantiﬁed in terms of the unprocessed wheat required for
their production, and dairy products, except for butter and cream, are
quantiﬁed as whole milk equivalents (FAO, 2001;2017b). FBS items
range from speciﬁc (e.g., bananas) to broad (e.g., freshwater ﬁsh).
Other model inputs, including trade data and item footprints, were
expressed in terms of speciﬁc items (e.g., walnuts), so we developed
schemas to match them to the associated FBS items (e.g., nuts and
For modeling purposes, we added several custom items to represent
foods either not included in FBS (e.g., edible insects) or more speciﬁc
than those in FBS. The custom item for forage ﬁsh, for example, in-
cludes small, schooling pelagic ﬁsh such as sardines and herring that
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
are prey for larger species, and unlike the FBS item “Pelagic Fish” it
does not include larger species such as tuna. In Mendeley Data, custom
items are identiﬁable by an FBS item code of 9000 or greater.
2.4. Modeled diets
For each of the 140 study countries, we modeled nine increasingly
plant-forward diets that adhered to parameters for a healthy diet
(summarized in Fig. 1; see also Mendeley Data input/item_parameters).
Each diet used the country’s baseline consumption pattern as the
starting point. In all steps where groups of FBS items (e.g., protein
foods) were scaled up or down, the relative proportions of items within
each group were preserved, reﬂecting each country’s unique dietary
pattern. For example, the residents of South Korea consume relatively
little dairy, so if they removed red meat from their diet we would not
expect milk products to be a popular protein substitute. When com-
paring FBS item quantities with nutritional criteria (e.g., the target for
caloric intake described below), we ﬁrst subtracted region- and food
group-speciﬁc losses occurring during processing and packaging, dis-
tribution, and consumption (Gustavsson et al., 2011). This step ensures
that criteria are met based on quantities that are closer to amounts
actually consumed, versus quantities in the food supply.
Diets were modeled as follows. First, to adjust for over- and under-
consumption, the baseline pattern was scaled to 2300 kilocalories—the
upper bound of average per capita energy requirements calculated by
Springmann et al. (2016). We held caloric intake constant across all
modeled diets for consistency when making cross-country comparisons.
In the steps described below (e.g., removing animal foods), the caloric
content of the diet underwent further changes and subsequently had to
be adjusted back to 2300, but performing this step ﬁrst kept the relative
proportions of FBS items closer to the baseline. Following the initial
adjustment for caloric intake, amounts of nuts, seeds, and oils were held
constant for all diets.
Where applicable, selected animal foods were removed (Fig. 1); e.g.,
terrestrial and aquatic meats were removed from the lacto-ovo vege-
tarian diet. Modeled diets were then adjusted to meet two health
guidelines from the World Health Organization and FAO (2003): Fruits
and vegetables (excluding starchy roots, e.g., potatoes, yams) were
scaled up to a ﬂoor of 400 g per day, or approximately ﬁve servings; and
added sugars were capped to contribute no more than 10% of total
energy intake. For diets in which meat was eliminated, the fruit and
vegetable ﬂoor was raised to six or seven servings per day (Fig. 1),
based on the rationale that healthy vegetarian and vegan dietary pat-
terns tend to include more of these items (Springmann et al., 2016).
Note that we use the term “vegan” to refer to exclusively plant-based
diets, without reference to other behaviors sometimes associated with
the term, such as avoidance of leather products.
The low red meat diet additionally included a cap on red meat (i.e.,
bovine, sheep, goat, pig) of 350 g cooked weight per week, or roughly
three servings, as per recommendations (World Cancer Research Fund
and American Institute for Cancer Research, 2018). We converted the
350 g cap from cooked to raw weight (467 g) using the same conversion
factors we used for per-serving footprints (Section 2.8, Mendeley Data
input/per_unit_serving_sizes), and from raw weight to carcass weight
(648 g) using the average of FAO extraction rates for bovine and pig
meat (FAO, 2017). Taken together with adjustments for added sugars,
fruits and vegetables, calories, and protein, this diet is intended to ap-
proximate the adoption of dietary recommendations.
For the low food chain diet, protein from insects replaced 10% of
the protein from terrestrial animal products, and protein from forage
ﬁsh and bivalve mollusks replaced 70% and 30%, respectively, of the
protein from aquatic animals. Insects are not included in FBS, so nu-
tritional content was derived from Payne et al. (2016). Forage ﬁsh and
bivalve mollusks are included in FBS but grouped with other items (e.g.,
“Molluscs, Other” includes snails), so nutrient content was derived from
the United States Department of Agriculture (USDA) food composition
database (USDA, 2017). See Mendeley Data input/nu-
trient_comp_custom_items for details.
Following these adjustments, selected energy staples, i.e., FBS items
in the grains and starchy roots groups, were scaled up or down to return
to the 2300 kilocalorie target. Selected protein groups (Fig. 1) were
then scaled up as needed to meet a protein ﬂoor of 69 g per day—12%
of total energy intake, within the recommended range of 10–15%
(World Health Organization and FAO, 2003). To hold calories constant
while scaling up protein, caloric increases from protein foods were
counter-balanced with commensurate reductions in calories from en-
ergy staples. The equation for this step is provided in Appendix B.2.
We also modeled an adjusted variant of the baseline pattern, scaled
to 2300 kcal and the protein ﬂoor (Figs. 1,5b, 6). When comparing
plant-forward modeled diets with baseline consumption patterns, the
adjusted baseline allows for isolating the eﬀects of food substitutions
independent of adjustments for over- and under-consumption.
The meatless day and two-thirds vegan diets were modeled as
combinations of two diets. Meatless day was patterned after behavior
change campaigns promoting one day of the week without meat (e.g.,
Meatless Monday) and assumes a lacto-ovo vegetarian diet for one day
per week and the adjusted baseline for the other six days. We included
this diet because it can serve as an entry point toward more plant-for-
ward diets. Two-thirds vegan was patterned loosely after “Vegan Before
6” (Bittman, 2013) and assumes a vegan diet for two out of three meals
per day and the adjusted baseline for the third, with each meal pro-
viding equal caloric content. This approach does not account for the
possibility that people in some countries may consume more animal
products at dinner, for example, compared to breakfast and lunch.
Fig. 1. Parameters for study diets. Partial shading indicates food groups that
were included only on selected days/meals, e.g., meat was included in six of
seven days for meatless day, and in one of three meals for two-thirds vegan.
Red meat includes bovine, sheep, goat, and pig meat.
When dairy products were scaled to meet the protein ﬂoor, only the FBS item
“Milk, Excluding Butter” (which also includes some milk-derived products such
as cheese and yogurt) was scaled. The FBS items “Butter, Ghee” and “Cream”
were not scaled for protein.
The fruits and vegetables ﬂoor and added sugars cap for meatless day were
only applied for one day of the week, reﬂecting one day of the lacto-ovo ve-
getarian diet and six days of the adjusted baseline.
The 2/3 vegan diet reﬂects the vegan diet for two out of three meals per day
and the adjusted baseline for the third. The fruits and vegetables ﬂoor and
added sugars cap were only applied to the two vegan meals.
For the low-food chain diet, protein from insects replaced 10% of the protein
from terrestrial animal products, and protein from forage ﬁsh and bivalve
mollusks replaced 70% and 30%, respectively, of the protein from aquatic an-
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
We also included a hypothetical scenario in which all study coun-
tries adopt the average baseline consumption pattern of high-income
OECD countries (The World Bank, 2018; Figs. 1,6and 8a–d), illus-
trating potential outcomes of the Western diet becoming more wide-
spread. Furthermore, by holding diet composition constant across
countries, this scenario isolates the eﬀects of import patterns and COOs
on GHG and water footprints.
We ran the model for the 140 countries with suﬃciently robust
trade and food supply data for inclusion in the 2011–2013 FAO detailed
trade matrices and FBS (FAO, 2017a).
2.6. Import patterns and countries of origin
An item’s footprint varies based on the conditions and practices
speciﬁc to its COOs (e.g., Figs. 3 and 4). To account for these diﬀer-
ences, for each country and diet, we traced the supply of each FBS item
back to the countries in which it was produced. Of Japan’s pig meat
supply, for example, 48% was produced domestically over 2011–2013,
22% was imported from the United States (US), 10% from Canada, 7%
from Denmark, and so on. For total imports by importing country and
FBS item, we used trade data averaged over 2011–2013 FBS, and to
allocate the share of total imports among COOs, we used 2011–2013
FAO detailed trade matrices (FAO, 2017a). Detailed methods are pro-
vided in Appendix B.3. Note that for this study, COOs were only re-
levant in cases where suﬃcient country-speciﬁc item footprint data
2.7. Diet footprints
Contributions of FBS items to diet footprints were modeled using
two approaches. The ﬁrst method used country-speciﬁc footprints, i.e.,
for the items consumed in a given country, the GHG and water foot-
prints were speciﬁc to the COOs from which each item was imported.
Since we did not have suﬃcient country-speciﬁc data to apply this
method in all cases, it was limited to the GHG and water footprints of
terrestrial animal products (excluding insects), WFs of plant foods, and
all land use change (LUC) CO
footprints. After adapting country-spe-
ciﬁc footprint data to FBS items, this method yielded 16 009 footprint
data points (available in Mendeley Data input/item_footprints_by_coo).
These were then multiplied by the corresponding quantities of each
item, allocated over COOs, in each country-diet combination. This
method and the associated data sources are described in Sections
2.7.2–2.7.4 with technical details covered in Appendix B.4.
The second method was used in cases where we did not have suf-
ﬁcient country-speciﬁc data to diﬀerentiate footprints by COO, i.e., for
the GHG and water footprints of aquatic animals and insects, and the
GHG footprints of plant foods. For this method we performed a litera-
ture search and adapted results from 114 peer-reviewed studies,
yielding 764 data points (available in Mendeley Data input/item_-
footprints_distributions). For these item-footprint pairs, we used a
bootstrapping approach to reﬂect the heterogeneity across the countries
and production systems examined in the 114 studies. The bootstrapping
approach is described in Sections 2.7.5–2.7.6, with the literature search
described in Appendix B.5.
All results reﬂect cradle-to-farm gate activities only, and thus do not
account for GHG and water footprints associated with processing,
transportation, retail and preparation. This limitation is discussed in
While most FBS items are expressed in terms of primary equivalents,
there were some cases where we needed to allocate shares of GHG and
water footprints among processed items originating from the same root
product, e.g., butter and cream from milk. We adapted the economic
allocation method described in Hoekstra et al. (2011). The method and
how it was applied in each case are described in Appendix B.6.
2.7.2. GHG and land-use change CO
footprints of terrestrial animal
products, by COO
For GHG footprints of terrestrial animal products (excluding in-
sects), we adapted data from FAO’s Global Livestock Environmental
Assessment Model GLEAM-i tool (FAO, 2017c). The tool applies a
consistent, transparent approach to quantifying production data and
GHG emissions associated with terrestrial animal production speciﬁc to
235 diﬀerent countries, accounting for diﬀerences in feed composition,
feed conversion ratios, manure management techniques, and other
parameters associated with the various species and production systems
(e.g., grasslands cattle, feedlot cattle, broiler chickens, layer chickens)
unique to each setting. The level of granularity provided by GLEAM-i
further allowed us to report CO
emissions from deforestation-driven
LUC separately from other emissions sources. These qualities made
GLEAM-i a robust choice for diﬀerentiating GHG footprints based on
Although GLEAM-i accounts for soil carbon ﬂuxes associated with
land use change, e.g., conversion from forest to grassland, it does not
account for the eﬀects of livestock management practices on soil carbon
losses or sequestration—an important limitation that should be ad-
dressed in future research (see Section 3.3). Furthermore, GLEAM-i
does not allocate GHG emissions to oﬀals and other slaughter by-
products, thus overestimating the GHG footprints of meat and under-
estimating those of oﬀals (see Appendix B.6).
With the exception of oﬀals, the GLEAM-i tool allocates GHG
emissions from each production system among the associated animal
products (e.g., cattle meat and milk from grassland systems in Brazil)
based on protein content. The GHG footprints of these items, as re-
ported by GLEAM-i, are speciﬁc to country, production system, and
item but are not speciﬁc to the emissions source (i.e., LUC for soy feed,
LUC for palm kernel cake feed, LUC for pasture expansion, and all other
sources of GHG emissions). One of our study aims was to highlight the
contributions of deforestation to GHG footprints. To this end, we allo-
cated the GHG footprints of items among emissions sources based on
the assumption that within a given a country and production system,
the relative shares of source-speciﬁc GHG emissions among the items
from that system is the same as the relative shares of total GHG emis-
sions among those items, which was provided by GLEAM-i. For ex-
ample, for United Kingdom (UK) layer systems, based on GLEAM-i data,
82% of the total GHG footprint was allocated to eggs and 18% was
allocated to poultry meat. Thus, we applied the same percentages to
allocate LUC CO
emissions from the use of soy feed in UK layer systems
(also reported by GLEAM-i) between eggs and meat. The equations for
this method are detailed in Appendix B.4.
Since GLEAM-i reports GHG footprints per kilogram of protein, we
converted to per-kilogram primary weight footprints (e.g., carcass
weight for meat, whole milk for dairy) as follows. For each GLEAM-i
item gproduced in country c, the primary weight GHG footprint GHG
was calculated as
= ×GHG GHGP PP
c g c g
where GHGP is the GHG footprint per kilogram of protein, PP is the
annual production in kilograms of protein, and Pis the annual pro-
duction in kilograms primary weight.
Footprints of GLEAM-i items (e.g., buﬀalo meat, cattle meat) then
needed to be translated to footprints of FBS items (e.g., bovine meat).
We developed schemas matching GLEAM-i countries and items to those
used in FBS. For each FBS item fproduced in country c, we then cal-
culated the primary weight GHG footprint as the average footprint of
the associated GLEAM-i item(s) gproduced in c, weighted by the ton-
nages produced P:
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
g c c g c g
g c c g
in , ,
If there were no GLEAM-i footprint data for an FBS item in a given
country, we used a regional average, weighted by the tonnage of the
FBS item produced in each country (FAO, 2017a), as follows:
GHG GHG P
r f c r c f c f
c r c f
Finally, if there were no footprint data for fin r, a weighted global
average was used.
2.7.3. Land-use change CO
footprints of soy and palm oils intended for
human consumption, by COO
Soybeans, soybean oil, palm oil, and palm kernel oil reported in FBS
food supply data reﬂect uses for human consumption; GHG footprints of
soy and palm as animal feed are described in Section 2.7.2. Land-use
footprints for the former items were adapted from FAO
GLEAM documentation (FAO, 2017d), which provides per-hectare LUC
footprints associated with soy and palm production for 92 (soy)
and 14 (palm) countries. Per-hectare footprints were converted to per-
kilogram footprints using country-speciﬁc crop yields from FAOSTAT,
averaged over 2011–2013. The LUC CO
footprints of soy and palm oils
were then derived from their root products using the economic allo-
cation method described in Appendix B.6. If there were no LUC CO
footprint data associated with soy or palm production in a given
country, the LUC CO
footprint was assumed to be zero.
2.7.4. Water footprints of plant foods and terrestrial animal products, by
We adapted data from literature quantifying the blue and green WFs
of plant foods (Mekonnen and Hoekstra, 2010a) and terrestrial animal
products (Mekonnen and Hoekstra, 2010b) speciﬁc to over 200 coun-
tries. We developed schemas matching countries and items from these
datasets to their FBS counterparts. Parallel to our approach for GHG
footprints, for each FBS item fproduced in country c, we calculated the
WFs as the average footprint of the associated water dataset item(s) w
produced in c, weighted by the tonnages produced P(FAO, 2017a):
WF WF P
c f w c c w c w
w c c w
,in , ,
If there were no country production data for an item w, an un-
weighted country average was used. If there were no WF data matching
FBS item fproduced in country c, a weighted regional or global average
footprint was used, following the method described above for GLEAM-i.
One FBS item (honey) had no associated WF data and was thus
excluded from WF calculations. Mekonnen and Hoekstra’s datasets did
not include insects, so the WF of insects was taken from Miglietta et al.
(2015) and used for insect production in all countries.
Note that this method does not account for levels of water scarcity
in countries of origin. While we acknowledge that there are diﬀering
perspectives regarding the need for scarcity-weighted WFs, our ap-
proach is informed by Hoekstra (2016), which argues that WFs have
implications for freshwater conservation wherever withdrawal occurs.
In an internationally-traded economy, all freshwater is part of an in-
creasingly scarce global pool. Even in regions with abundant freshwater
availability, if water is used ineﬃciently in agriculture or aquaculture,
wasted water is water that could have otherwise been used to produce
more food—thus lessening the need for other, potentially water-scarce,
regions to produce as much.
2.7.5. Bootstrapping approach for GHG footprints of plant foods, aquatic
animals, and insects
In contrast to the datasets used for footprints by COO—which used
uniform methods across FBS items and countries—plant food, aquatic
animal, and insect GHG footprints from the literature search reﬂected a
diversity of studies with varied methods, and represented some coun-
tries more than others. To maximize consistency across studies and with
the country-speciﬁc data describe above, we applied strict inclusion/
exclusion criteria and standardized results to the degree possible (de-
scribed in Appendix B.5); however, the practices under study still
varied greatly, e.g., by fertilizer and pesticide application rates, use of
organic practices, irrigation method, crop rotations, use of protected
cultivation (e.g., greenhouses), ﬁsh stocking density, and ﬁshing
method (e.g., long-lining, trawling). These may not be representative of
the prevailing practices for a given country-item combination.
To account for this heterogeneity, we create a weighted probability
distribution for each FBS item’s footprint observations. When a study
provided results for multiple scenarios involving the production of the
same item in the same country, e.g., for ﬁve GHG footprint observations
for Spanish wheat with varying levels of nitrogen fertilizer inputs, we
assigned a weight to each observation equal to the reciprocal of the
number of observations, e.g., 1/5, preventing studies with multiple
observations from being overrepresented. If there were no observations
for an FBS item, proxies were used, e.g., a distribution of all grains
footprints was used for sorghum and products, and a distribution of all
citrus fruit footprints was used for grapefruit and products. All item
footprint distributions used in the model are provided in Mendeley Data
To calculate the contributions of plant foods, aquatic animals, and
insects to the GHG footprint of a country-diet combination, we used a
bootstrapping approach designed to capture the distribution of item
footprint values from the literature. The weighted distribution of GHG
footprint values for tomatoes, for example, was skewed right; simply
using the median or average would ignore this important detail. For our
approach, we 1) selected 10 000 random samples from each FBS item
footprint distribution, e.g., 10 000 samples from 23 weighted GHG
footprint values (kg CO
e/kg) for barley; 2) multiplied each sampled
footprint value by the corresponding quantity of the FBS item in the
diet, e.g., 46 kg barley/capita/year in the Moroccan vegetarian diet;
and 3) summed the resulting values for FBS items within the same
group, e.g., resulting in a distribution of 10 000 values for the kg CO
capita/year associated with grains in the Moroccan vegetarian diet.
Summing the median value from each distribution with results by COO
(Sections 2.7.2–2.7.4) yielded the total per capita footprint of a given
country diet. We also present interquartile ranges (error bars in Fig. 7,
also provided in Mendeley Data output) to convey variations across
bootstrapped outputs. Note that these ranges apply only to items for
which bootstrapping was used, as the COO-speciﬁc method does not
account for uncertainty and is deterministic, returning a single footprint
value for each permutation of inputs (e.g., FBS item, diet, country, and
2.7.6. Bootstrapping approach for water footprints of aquatic animals
Aquatic animal WFs were limited to farmed species and accounted
for blue and green WFs associated with feed production and, where
applicable, blue water used to replace evaporative losses from fresh-
water ponds and to dilute seawater in brackish production. Water
footprints of wild-caught aquatic animals were assumed to be negli-
For feed-associated WFs, we created a distribution of WF values
adapted from Pahlow et al. (2015) for each FBS item associated with
farmed species. We did not have information about the share consumed
in a given country that was farmed versus wild-caught, so we made
assumptions based on 2014 global production patterns, e.g., 79% of
harvests associated with the FBS item “Freshwater Fish” were from
aquaculture (FAO, 2017e), so when this item was included in diets, we
only applied the feed-associated WF to 79% of the amount consumed
regardless of the country.
For freshwater pond aquaculture, we created a distribution of blue
WF values for each of the FBS items “Freshwater Fish” and
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
“Crustaceans” (Gephart et al., 2017;Henriksson et al., 2017;Verdegem
and Bosma, 2009). For “Crustaceans” we created an additional dis-
tribution of blue WF values for brackish water pond aquaculture
(Henriksson et al., 2017;Verdegem and Bosma, 2009). Both distribu-
tions were weighted using the method described in Section 2.7.5, ex-
cept for the 31 values for freshwater production in China from Gephart
et al. (2017), which were weighted by the percentage of Chinese
freshwater production represented by each data point. We did not have
information about the shares consumed in a given country that were
from freshwater or brackish ponds, so as per our method for feed-as-
sociated WFs, we made assumptions based on 2014 global production
patterns (FAO, 2017e; Mendeley Data input/aquaculture_-
Contributions of aquatic animals to country-diet WFs were calcu-
lated as follows, using the bootstrapping approach described in Section
2.7.5. We (1) selected 10 000 random samples from each FBS item-
footprint distribution, e.g., for “Crustaceans” we selected 10 000 sam-
ples each from the distributions for feed blue WF, feed green WF,
freshwater pond blue WF, and brackish water pond blue WF; (2) mul-
tiplied each sampled footprint value by the corresponding quantity of
the FBS item in the diet; and (3) summed the resulting values for FBS
items within the same group, i.e., “Aquatic animals,” keeping results for
each water footprint type separate.
2.8. Footprints of individual food items
In addition to calculating diet footprints, we presented per-serving,
per-kilocalorie, per-gram of protein, and per-kilogram edible weight
footprints associated with grouped FBS items (Figs. 2, S1–S3). For per-
kilogram footprints, we converted carcass weight and whole aquatic
animal footprints of terrestrial and aquatic meats to edible weight
equivalents (FAO, 1989, n.d.;Nijdam et al., 2012;Waterman, 2001).
Where nut footprints were expressed in terms of in-shell, we converted
them to shelled. Although the model handled dairy products in terms of
whole milk equivalents (except for butter and cream), for comparative
purposes we added the footprints of cheese and yogurt, derived from
milk using economic allocation (see Appendix B.6). Per-kilogram edible
weight footprints were then converted to per-serving footprints using
US standards (U.S. Food and Drug Administration, 2016). Serving sizes
and conversion factors are provided in Mendeley Data input/per_-
In addition to presenting the median and interquartile range for
each group footprint, for groups with footprints speciﬁc to COO, we
calculated global averages weighted by the mass produced in each
country. For groups with footprints from our literature search, averages
were weighted by the reciprocal of the number of observations from
each study to prevent studies with multiple observations from being
overrepresented (consistent with the weighting method described in
3. Results and discussion
3.1. Footprints of individual food items
Our study model incorporated 3850 GHG, 5402 blue water, and
7521 green water data points (Mendeley Data input/item_-
footprints_by_coo, input/item_footprints_distributions) reﬂecting
cradle-to-farm gate footprints of the individual food items comprising
diets, spanning diverse production practices and conditions unique to
COO. These are presented per serving (Fig. 2), per kilocalorie (Fig. S1),
per gram of protein (Fig. S2) and per kilogram edible weight (Fig. S3) as
global averages weighted by the tonnage produced in each country
(where suﬃcient country-speciﬁc data were available). These ﬁgures
show footprint values aggregated over common food groups (e.g.,
grains), whereas the study model handled items with greater speciﬁcity
(e.g., maize, millet, barley).
Whether by serving, energy content, protein, or mass, ruminant
meats (i.e., bovine, sheep, goat) were by far the most GHG-intensive
items. Per serving, bovine meat (weighted average: 6.54 kg CO
ving) was 316, 115, and 40 times more GHG-intensive than pulses, nuts
Fig. 2. Average per serving (a) GHG, (b) blue water, and (c) green water item
footprints. For items with suﬃcient country-speciﬁc footprint data (i.e., GHG
and water footprints of terrestrial animal products excluding insects, WFs of
plant foods, and LUC CO
footprints), footprints were averaged across countries
and weighted by the tonnage produced in each country. For all other items (i.e.,
from the literature search), see Section 2.7.5 for how averages were weighted.
Most items shown here are grouped (e.g., grains); footprints associated with
speciﬁc items used in the study model (e.g., maize, millet, barley) are provided
in Mendeley Data input. Diamonds represent medians and error bars show in-
terquartile ranges. See Mendeley Data input/per_unit_serving_sizes for primary
weight to serving size conversions.
† Forage ﬁsh GHG footprints are based on sardines and herring. Pond-raised
WFs largely reﬂect tilapia, carp and catﬁsh. Blue WFs for brackish pond
aquaculture reﬂect freshwater used to dilute seawater. Water footprints of wild-
caught aquatic animals were assumed to be negligible.
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
and seeds, and soy, respectively. Insects (e.g., mealworms, crickets) and
forage ﬁsh (e.g., sardines, herring) were among the more climate-
friendly animal products, much more so than dairy. Plant foods were
generally the least GHG-intensive overall, often by an order of magni-
tude, even after accounting for GHGs associated with deforestation for
palm oils and soy.
Blue WFs of pond-raised ﬁsh (e.g., carp, tilapia, catﬁsh; weighted
average: 698 L/serving) and farmed crustaceans (e.g., shrimp, prawns,
crayﬁsh; weighted average: 1184 L/serving) exceeded those of other
item groups by an order of magnitude. Our model accounted for water
used in production ponds and crop production for aquaculture feed. Re-
ﬁlling ponds to replace evaporative losses, together with freshwater
used to dilute seawater in brackish production, accounted for 94.7%
and 95.1% of the blue WFs for pond-raised ﬁsh and farmed crustaceans,
Bovine meat was the only item group for which the weighted
average blue WF was greater than the 75
percentile blue WF. This
suggests that most bovine meat production occurs in countries where
blue water use for bovine meat is particularly high.
The wide interquartile ranges of country-speciﬁc item footprints
(error bars in Figs. 2, S1–S3; see also Figs. 3 and 4) illustrate variations
in the conditions and practices unique to where items are produced.
The per-kilogram GHG footprints of bovine meat from Paraguay and
Brazil, for example, were 17 and ﬁve times higher, respectively, than
that of Danish bovine meat (Fig. 3). These diﬀerences were largely at-
tributable to deforestation for grazing lands and higher methane
emissions from ruminant eructation (belching). While there were in-
suﬃcient data to account for COO in all cases, we did so for most of the
items with the greatest magnitude and variance in footprints, e.g., GHG
footprints of terrestrial animal products (excluding insects).
3.2. Footprints of whole diets
We modeled scenarios illustrating the potential per capita and
whole-country footprints of nine plant-forward diets. These in part re-
ﬂect modeling choices; they represent potential outcomes for con-
sideration and may not reﬂect actual consumption behaviors. Scenarios
involving country-wide shifts to a particular diet, for example, are un-
likely to occur, but can reveal opportunities where policy and beha-
vioral interventions could have the broadest eﬀect, particularly in po-
pulous countries (Figs. 6b, 8c and d).
3.2.1. Global implications of adopting the OECD diet
In a scenario in which all 140 study countries adopted the average
consumption pattern of high-income OECD countries, per capita diet-
related GHG and consumptive (blue plus green) water footprints in-
creased by an average of 135 and 47 percent, respectively, relative to
the baseline (shown for selected countries in Figs. 6,8a–d). These
ﬁndings echo prior literature (e.g., Bajzelj et al., 2014;Willett et al.,
2019) on the climate implications of rising meat and dairy intake, and
the importance of both reducing animal-product intake in high-con-
suming countries and providing viable plant-forward strategies for
3.2.2. Global implications of adjusting for under-consumption
We modeled scenarios in which dietary patterns could better align
with ecological goals alongside nutrition guidelines—while also iden-
tifying some of the challenges in doing so. For example, baseline protein
and caloric availability were below recommended levels (Section 2.4)
in 49 and 36 percent of countries, respectively. The resulting adjust-
ments for under-consumption attenuated—and in some cases com-
pletely oﬀset—the GHG and water footprint reductions associated with
dietary shifts. For a scenario in which all 140 study countries adopted
either the low red meat or meatless day diet, our model projected an
average net increase in diet-related GHG, blue water, and green water
footprints relative to the baseline (Fig. 5a). Populous countries char-
acterized by under-consumption were the largest contributors to this
phenomenon, namely India and to a lesser degree Pakistan and In-
donesia (Figs. 6–8); loss-adjusted baseline protein availability in these
Fig. 3. Per-kilogram GHG footprints of bovine meat, by producing country, shown for countries that produced over 100 000 metric tons in 2011–2013.
Fig. 4. Per-kilogram blue and green WFs of rice, by producing country, shown
for countries that produced over 1 000 000 metric tons (1 megaton) in
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
countries was 14, 9, and 12 g below the recommended minimum of
69 g, respectively. Thus, interventions that aim to address both sus-
tainability and health goals must ensure plant-forward shifts are am-
bitious enough to oﬀset the potential ecological burdens associated
with providing adequate nutrition.
By contrast, if we hold caloric intake constant—that is, independent
of adjustments for over- and under-consumption (i.e., relative to an
adjusted variant of the baseline pattern, scaled to 2300 kcal and the
protein ﬂoor)—shifting to the low red meat or meatless day diets re-
sulted in an average net 4% or 3% reduction in diet-related GHG
footprints, respectively (Fig. 5b). Regardless of their eﬀectiveness in
climate change mitigation, these modest shifts may oﬀer an accessible
starting point toward more plant-forward dietary patterns.
3.2.3. Importance of country-speciﬁc analyses, trade, and countries of
The global aggregates shown in Fig. 5 are limited insofar as they
obscure the considerable variation across countries, illustrated by the
interquartile ranges. This variation was attributable to diﬀerences in
food supply composition (e.g., the degree to which the aquatic animals
group is comprised of pond-raised species), how animal products are
replaced when shifting diets, adjustments for over- and under-con-
sumption, and import patterns and the associated production practices
(e.g., pasture-based vs. intensive; irrigated vs. rainfed) and climatic
conditions (e.g., precipitation, evapotranspiration) unique to COOs. A
country-speciﬁc analysis reveals, for example, that shifting to the
meatless day diet reduced GHG and water footprints in 47% and 57% of
study countries, respectively—with some of the greatest per capita re-
ductions in Paraguay, Israel, and Brazil—even though the average net
eﬀect was an increase in footprints. Fig. 7 further illustrates the degree
to which the relative environmental beneﬁts among diets varied across
countries, along with the relative contributions of diﬀerent food groups.
Notably, of the 140 individual countries examined in this study, mos-
t—including those identiﬁed as having the most GHG- and water-
Fig. 5. Potential per capita changes in diet-related GHG, blue water, and green water footprints across all 140 study countries, calculated as the average Δfootprint
weighted by the population of each country. Shown for the nine modeled diets relative to (a) baseline consumption patterns and (b) an adjusted variant of each
country’s baseline, scaled to 2300 kcal with a 69g/capita/day protein ﬂoor. The adjusted baseline allows for comparisons between plant-forward diets and baseline
patterns independent of adjustments for over- and under-consumption, isolating the eﬀects of food substitutions. Diamonds represent medians and error bars show
Fig. 6. Greenhouse gas footprints for selected diets, by country, (a) per capita and (b) for whole country populations. Countries are sorted by baseline footprint. Due
to space constraints, of the 140 study countries, only the following are shown here: (a) the 59 countries above the 58th percentile for whole country baseline
footprint, and (b) the 11 countries above the 92nd percentile for whole country baseline footprint.
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
Fig. 7. Per capita diet-related GHG footprints by country, diet, and food group. Shown for the top four countries with the largest whole-country diet-related baseline
GHG footprints: (1 st) mainland China, (2nd) India, (3rd) Brazil and (4th) the United States. Indonesia, ranked 7th for whole-country footprint, is also shown as an
example of a country with high consumption of aquatic animals. Most items shown here are broadly grouped (e.g., plant foods); diet footprints are provided with
greater speciﬁcity in Mendeley Data output. Error bars show interquartile ranges and apply only to items for which bootstrapping was used, i.e., plant foods, aquatic
animals, and insects (see Section 2.7.5).
Fig. 8. Water footprints by country (a) per capita, blue WF only; (b) per capita, combined blue plus green WFs; (c) for whole countries, blue WF only; (d) for whole
countries, combined blue plus green WFs; and (e) per capita, for baseline diets only, separated by blue and green WF. Countries are sorted by (a–d) baseline footprint
or (e) blue WF. Due to space constraints, of the 140 study countries, only the following are shown here: (a, b, e) the 35 countries above the 75th percentile for whole
country baseline footprint, and (c, d) the 14 countries above the 90th percentile for whole country baseline footprint.
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
intensive diets—have been vastly underrepresented in the literature
(Appendix A, Table A1).
The scenario in which countries adopt the average baseline con-
sumption pattern of high-income OECD countries (Figs. 6,8a–d) iso-
lates the eﬀects of import patterns and COO on GHG and water foot-
prints. Holding diet composition constant across the 140 study
countries, the GHG and consumptive (blue plus green) water footprints
associated with this scenario showed substantial variation (averaging
2.5 ± 0.9 metric tons CO
e/capita/year and 1.5 ± 0.5 megaliters/
A number of country governments, including Brazil (Ministry of
Health of Brazil, 2014) and more recently Canada (Health Canada,
2019), have put forth dietary guidelines emphasizing predominantly
plant-based foods. While this is a critical step toward aligning domestic
consumption patterns with public health and ecological goals, coun-
tries’ production and export patterns merit additional attention. Brazil,
for example, was the top exporter of bovine meat (based on an average
of 2011–2013 data) and was in the top quartile for GHG-intensity of
bovine meat production (Fig. 3). Together with other major GHG-in-
tensive exporters such as India and Paraguay, Brazilian bovine meat
exports contributed to the large GHG footprints of diets in importing
countries like Chile, Hong Kong, Kuwait, Venezuela, and Israel. In a
hypothetical scenario in which the share of Hong Kong’s bovine meat
imports from Brazil came from Denmark instead, Hong Kong’s per ca-
pita GHG footprint for the baseline pattern was 18% lower. While not
necessarily feasible or desirable, this scenario further illustrates the
importance of accounting for trade patterns and COO.
3.2.4. Per capita GHG footprints of whole diets
The countries with the most GHG-intensive baseline consumption
patterns (Fig. 6)—and the greatest potential GHG reductions from
shifting toward plant-forward diets—included those with the highest
per capita intake of bovine meat (Argentina, Brazil, Australia), the most
GHG-intensive bovine meat production (Paraguay, Chile; Fig. 3), and
the greatest contributions of deforestation to the GHG footprints of diets
(Paraguay, Chile, Brazil; Brazil is shown in Fig. 7). Deforestation ac-
counted for 61% of the GHG footprint for the Paraguayan baseline
consumption pattern, and over 10% of the GHG footprints for 32
countries’ baseline patterns.
Over all 140 study countries, a theoretical shift to vegan diets re-
duced per capita diet-related GHG footprints by an average of 70%,
relative to the baseline (Fig. 5a). Vegan diets had the lowest per capita
GHG footprints in 97% of study countries. Given the low per-kilocalorie
GHG footprints of most plant foods (Fig. S1), even substantial increases
in consumption had only modest eﬀects on GHG emissions of diets. For
the US vegan diet, for example, scaling up plant foods recouped 100%
of the calories and protein from animal foods with only 16% of the GHG
emissions relative to the adjusted baseline (Fig. 7).
Relative to vegan diets, low-food chain diets (i.e., predominantly
plant-based plus forage ﬁsh, bivalve mollusks, and insects) oﬀer greater
ﬂexibility by allowing for modest animal product intake with compar-
able environmental beneﬁts (Fig. 5). Low-food chain diets also met the
recommended intake of vitamin B12 for adults (2.4 μg/day; Institute of
Medicine Food and Nutrition Board, 1998) in 49% of study countries,
illustrating that there may be ways to mitigate this potential limitation
of plant-forward diets even without supplementation, at least for the
Mostly plant-based diets were generally less GHG-intensive than
lacto-ovo vegetarian diets, in part due to the relatively high GHG
footprint of dairy (and eggs, depending on the basis of comparison;
Figs. 2, S1–S3) and the reliance on dairy as one of only three food
groups in the lacto-ovo vegetarian diet used to meet the protein ﬂoor
(Fig. 1). This phenomenon was particularly notable for India (Figs. 6
and 7). In 95% of countries, two-thirds vegan diets were less GHG-in-
tensive than lacto-ovo vegetarian (e.g., Figs. 6 and 7). Countries where
this was not the case included those with some of the most GHG-
intensive baseline consumption patterns (i.e., Paraguay, Chile, Argen-
tina), largely because of the GHG-intensity of ruminant meat in those
countries. In 64% of countries, the GHG footprints of no dairy diets
were lower than those of lacto-ovo vegetarian diets (e.g., India and
Indonesia, Fig. 7; also Fig. 6). In 91% of countries, the GHG footprints
of low-food chain diets were less than half those of lacto-ovo vegetarian
diets. These ﬁndings suggest populations could do far more to reduce
their climate impact by eating mostly plants with a modest amount of
low-impact meat than by eliminating meat entirely and replacing a
large share of the meat’s protein and calories with dairy.
3.2.5. Per capita water footprints of whole diets
Per capita blue WFs of diets (Fig. 8a, e) were in many cases largest
in countries with 1) low annual precipitation, increasing reliance on
irrigation for domestic crops; and 2) climatic factors such as high
temperatures that contribute to high evapotranspiration rates, and
thereby decrease crop water productivity (i.e., crop output per unit of
water consumed). These included Iran, Egypt, and Saudi Arabia. Do-
mestically-produced rice was among the top contributors in high-blue
WF countries, four of which (Kazakhstan, Afghanistan, Pakistan, Iran)
were also among the most blue water-intensive rice-producing coun-
tries (e.g., Fig. 4; rice WFs for all countries are provided in Mendeley
Data input/item_footprints_by_coo). For blue WF reductions, the most
impactful per capita dietary shifts were in Egypt, in part due to the high
blue water intensity of Egyptian bovine meat and dairy.
For baseline consumption patterns, the consumptive (blue plus
green) WF was highest for Niger (Fig. 8b, e), 98% of which was attri-
butable to green water. Domestically-grown millet was the largest
single contributor (40%) to the consumptive WF of the baseline con-
sumption pattern. Niger had by far the highest per capita millet supply
of any country, and was the 3rd largest producer and 8
intensive millet-producing country. The low water productivity of
millet in Niger was attributable to low edible yield and high evapo-
transpiration rates. Inedible millet crop residues, however, provide fuel,
construction materials, and livestock fodder (Sadras et al., 2009), il-
lustrating how sociocultural and economic provisions of agricultural
goods must be considered alongside ecological outcomes (see Section
Potential reductions in per capita consumptive WFs from shifting to
vegan diets were largest in Bolivia, Israel, and Brazil. Bovine meat,
poultry, and dairy together accounted for over half of the consumptive
WFs of the baseline consumption patterns in each of these countries. In
Israel, for example, the per capita consumptive WFs of the low-food
chain and vegan diets were 66% and 67% lower, respectively, than that
of the baseline consumption pattern. Bolivia was the most water-in-
tensive producer of bovine meat and the second for dairy, and most of
the country’s supply of these items was produced domestically. Bolivia
also has a high prevalence of anemia (Development Initiatives, 2018),
thus eﬀorts to mitigate high WFs through dietary interventions must
give this careful consideration.
For many countries, the blue WFs of low and no red meat, no dairy,
and pescetarian diets were higher than those of baseline consumption
patterns (Figs. 5a, 8a). These diets scaled up aquatic animals, of which
the FBS items “Freshwater Fish” and “Crustaceans” were highly blue
water-intensive when raised in ponds (Figs. 2, S1–S3). Contributions of
aquatic animals to the blue WFs of baseline, low red meat, and no red
meat diets exceeded those from terrestrial meat in 29%, 34%, and 69%
of countries, respectively. In mainland China and Indonesia, for ex-
ample, aquatic animals contributed 29% and 26%, respectively, to the
blue WFs of baseline consumption patterns. In both countries, a sub-
stantial share of domestic ﬁsh production was from aquaculture (72%
and 38%, respectively), predominantly for domestic consumption and
not export (Belton et al., 2018). Replacing water-intensive pond-raised
species with forage ﬁsh and bivalve mollusks, as in the low-food chain
diet, could reduce both water and GHG footprints (see Section 3.3 re-
garding limits to increased aquatic animal intake).
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
Note that we did not have information about the shares of fresh-
water ﬁsh and crustaceans consumed in a given country that were
farmed in ponds, so we made assumptions based on global production
patterns (see Section 2.7.6). This method overestimates blue WFs of
countries that source a large share of these species from wild ﬁsheries
or non-pond aquaculture, while underestimating blue WFs of countries
for which the converse is true.
3.2.6. Targeting dietary interventions and whole-country footprints of diets
All else being equal, optimal interventions would promote dietary
shifts in countries with large potential reductions in both per capita and
whole country GHG and water footprint (acknowledging that “optimal”
depends on a wide range of factors, including many not considered
here; see Section 3.3). Based on shifting to a two-thirds vegan diet for
purely illustrative purposes, only three countries—Brazil, the US, and
Australia—were in the highest quintile for all four of the following
criteria: greatest potential per capita and whole-country reductions in
both GHG and consumptive water footprints (Fig. S4).
3.3. Limitations and opportunities for future research
There is much variability and uncertainty in accounting for post-
farm gate activities (e.g., processing, transportation, retail) and soil
carbon ﬂuxes, and accordingly, they are rarely included in the scope of
item footprint studies. Both were thus excluded from this study. We do
not expect the former to aﬀect our overall conclusions, as the majority
(80–86%) of diet-related GHG emissions have been attributed to the
production stage (Vermeulen et al., 2012).
Accounting for soil carbon sequestration has been shown to lower
estimates of the GHG footprints of ruminant products, particularly
those from management-intensive grazing systems (e.g., Pelletier et al.,
2010;Tichenor et al., 2017). Further research is needed to measure the
potential for soil carbon sequestration to reduce ruminant GHG foot-
prints over a broad geographic and temporal scale, given it is time-
limited; reversible; and highly context-speciﬁc based in part on soil
composition, climate, and livestock management (Garnett et al., 2017).
Conversely, the potential for soil carbon losses (e.g., from overgrazing
or feed crop production) to increase ruminant GHG footprints should
also be considered. Regardless of the uncertain role of well-managed
grazing systems in carbon sequestration, the potential beneﬁts for soil
health, biodiversity, animal welfare, and other dimensions independent
of climate change should also be taken into consideration. Apart from
livestock production, carbon ﬂuxes in crop and polyculture systems
should also be further explored.
Aside from shifting consumption patterns, our study model holds
other factors constant over time, including climatic conditions, popu-
lation dynamics, food wastage, trade patterns, and the GHG- and water-
intensity of production. Over the gradual course of changing diets,
these factors will change in ways that are diﬃcult to anticipate, e.g., as
a result of rising incomes, evolving technology, changing trade policies,
and economic feedback eﬀects. Furthermore, we assume a proportional
relationship between shifting demand and supply-side impacts, whereas
the impact of dietary shifts on blue water conservation, for example,
may be limited without policies promoting sustainable withdrawal
rates (Weindl et al., 2017). Similarly, reducing animal product intake
cannot reverse CO
emissions from deforestation unless land is taken
out of production and reforested (Searchinger et al., 2018). Given their
uncertain potential, dietary shifts should be complemented with other
behavioral and policy interventions.
Further research is needed to examine dietary shifts in the context of
social, economic, ecological, and agronomic feasibility, particularly in
low- and middle-income countries (Kiﬀ et al., 2016), as well as the
eﬀects on other health, social, and ecological measures not considered
here (e.g., producers’ livelihoods, land availability, biodiversity, and
eutrophication potential). Shifts to plant-forward diets, for example,
must ensure target populations have suﬃcient physical and economic
access to a variety of nutrient-dense plant-based foods. Agricultural
systems would need to scale up production of fruits, vegetables, and
proteins to meet the nutritional needs of the current population (KC
et al., 2018), concurrent with a more equitable redistribution of
available food. Dietary scenarios that increase aquatic animal con-
sumption, meanwhile, raise concerns regarding depletion of wild stocks
and ecological issues associated with increasing production of certain
farmed species (Thurstan and Roberts, 2014). The feasibility of sus-
tainable diets may further depend on how well proposed eating patterns
align with historical and cultural context. Van Dooren and Aiking
(2016) demonstrate a method for balancing several of these domains by
simultaneously optimizing modeled diets for nutrition, climate change
mitigation, land use, and cultural acceptability. Our use of baseline
consumption patterns as a reference point helped to preserve countries’
eating patterns when modeling diets (Section 2.4); cultural receptivity
could be further reﬁned, however, by using national food-based dietary
guidelines (FBDGs) to deﬁne criteria for healthy diets for individual
countries, as in Vanham et al. (2018), rather than global re-
commendations (Section 2.4). Alternatively, or in cases where countries
do not have FBDGs, this research could help deﬁne FBDGs that are
healthy, sustainable, and culturally appropriate. Country-speciﬁc ana-
lyses that account for cultural acceptability could then be placed within
the context of the planetary boundaries for food systems proposed by
the EAT-Lancet Commission (Willett et al., 2019). The need to better
characterize the impacts of, viability of, and strategies for shifting to-
ward plant-forward diets, however, must be balanced against the pre-
ponderance of evidence calling for immediate action.
We evaluated nine plant-forward diets aligned with nutrition
guidelines, speciﬁc to 140 individual countries, for their potential roles
in climate change mitigation and freshwater conservation. Accounting
for country-speciﬁc diﬀerences in over- and under-consumption, trade
and baseline consumption patterns, and the GHG- and water-intensities
of foods by COO can help tailor policy and behavioral interventions.
Using this approach, we present a range of ﬂexible options for each
country that better align dietary patterns with public health and eco-
logical goals, including viable alternatives for low- and middle-income
countries that might otherwise adopt the consumption patterns of
Declaration of Interest Statement
B.F.K and S.R.C. developed the model with guidance and con-
tributions from all co-authors; J.P.F. provided guidance and expertise
on the modeling and analysis of aquatic animal footprints; M.M.M. and
A.Y.H. provided guidance and expertise on water footprints and co-
product allocation; S.D.P. and M.W.B. provided guidance and expertise
on modelling healthy diets; A.P.S., B.F.K., R.E.S., and C.M.S. performed
the search and standardization of item footprint studies; R.E.S. per-
formed the literature review of other diet footprint studies; B.F.K. and
R.E.S. wrote the manuscript; and K.E.N. and R.A.N. provided guidance
and expertise on all facets of and supervised the project. All authors
reviewed and contributed to manuscript drafts.
We thank Danielle Edwards and Emily Hu for research assistance;
Rebecca Ramsing, Alana Ridge, and Marie Spiker for general guidance
and discussions; Tomasz Filipczuk from the Crops, Livestock & Food
Statistics Team of the FAO Statistics Division for guidance on the use
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
and interpretation of FAO data; and Ruth Burrows, Bailey Evenson,
Carolyn Hricko, Shawn McKenzie, Matthew Kessler, Rebecca Ramsing,
Marie Spiker, and James Yager for comments on the manuscript. This
work was supported by the Columbus Foundation. The funders had no
role in study design; data collection, analysis, or interpretation; pre-
paration of the manuscript; or decision to publish.
Supplementary ﬁgures, tables, and appendices related to this article
can be found, in the online version, at doi:https://doi.org/10.1016/j.
gloenvcha.2019.05.010. Supplementary data are provided via
Mendeley Data (Kim et al., 2019).
Aleksandrowicz, L., Green, R., Joy, E.J.M., Smith, P., Haines, A., 2016. The impacts of
dietary change on greenhouse gas emissions, land use, water use, and health: a sys-
tematic review. PLoS One 11, e0165797. https://doi.org/10.1371/journal.pone.
Bajzelj, B., Richards, K.S., Allwood, J.M., Smith, P., Dennis, J.S., Curmi, E., Gilligan, C.A.,
2014. Importance of food-demand management for climate mitigation. Nat. Clim.
Change 4, 924–929. https://doi.org/10.1038/nclimate2353.
Belton, B., Bush, S.R., Little, D.C., 2018. Not just for the wealthy: rethinking farmed ﬁsh
consumption in the Global South. Glob. Food Secur. 16, 85–92. https://doi.org/10.
Bittman, M., 2013. VB6: Eat Vegan Before 6:00 to Lose Weight and Restore Your Health…
for Good. Clarkson Potter.
Bryngelsson, D., Wirsenius, S., Hedenus, F., Sonesson, U., 2016. How can the EU climate
targets be met? A combined analysis of technological and demand-side changes in
food and agriculture. Food Policy 59, 152–164. https://doi.org/10.1016/j.foodpol.
Dario, C., Anna, L., Steven, J.D., Simone, B., Ken, C., 2014. CH4 and N2O emissions
embodied in international trade of meat. Environ. Res. Lett. 9, 114005.
de Boer, J., Schösler, H., Aiking, H., 2014d. “Meatless days” or “less but better”?
Exploring strategies to adapt Western meat consumption to health and sustainability
challenges. Appetite 76, 120–128. https://doi.org/10.1016/j.appet.2014.02.002.
de Pee, S., Bloem, M.W., 2009d. Current and potential role of specially formulated foods
and food supplements for preventing malnutrition among 6- to 23-month-old chil-
dren and for treating moderate malnutrition among 6- to 59-month-old children.
Food Nutr. Bull. 30. https://doi.org/10.1177/15648265090303S305.
de Ruiter, H., Macdiarmid, J.I., Matthews, R.B., Kastner, T., Smith, P., 2016. Global
cropland and greenhouse gas impacts of UK food supply are increasingly located
overseas. J. R. Soc. Interface 13. https://doi.org/10.1098/rsif.2015.1001.
Development Initiatives, 2018. 2018 Global Nutrition Report: Shining a Light to Spur
Action on Nutrition. Bristol, UK. https://globalnutritionreport.org/reports/global-
Fehrenbach, K.S., Righter, A.C., Santo, R.E., 2016. A critical examination of the available
data sources for estimating meat and protein consumption in the USA. Publ. Health
Nutr. 19, 1358–1367.
Fischer, C.G., Garnett, T., 2016. Plates, Pyramids, Planet: Developments in National
Healthy and Sustainable Dietary Guidelines: A State of Play Assessment. United
Nations Food and Agriculture Organization and The Food Climate Research Network,
Food and Agriculture Organization of the United Nations, 1989. Yield and Nutritional
Value of the Commercially More Important Fish Species, FAO Fisheries Technical
Paper. FAO, Rome.
Food and Agriculture Organization of the United Nations, 2001. Food Balance Sheets: A
Handbook. Rome. http://www.fao.org/docrep/003/x9892e/X9892E00.htm#
Food and Agriculture Organization of the United Nations, 2017a. FAOSTAT. Accessed
December 2017. http://www.fao.org/faostat/en/.
Food and Agriculture Organization of the United Nations, 2017b. FAOSTAT Commodity
Deﬁnitions and Correspondences. Accessed December 2017. http://www.fao.org/
Food and Agriculture Organization of the United Nations, 2017c. GLEAM-i Version 2.0
Revision 3. http://www.fao.org/gleam/resources/en/.
Food and Agriculture Organization of the United Nations, 2017d. Global Livestock
Environmental Assessment Model, Version 2.0, Model Description Revision 4. Rome.
Food and Agriculture Organization of the United Nations, 2017e. Fishery Statistical
Collections: Global Production. Accessed December 2017. http://www.fao.org/
Food and Agriculture Organization of the United Nations, n.d. Technical Conversion
Factors for Agricultural Commodities. http://www.fao.org/ﬁleadmin/templates/ess/
documents/methodology/tcf.pdf. Accessed December 2017.
Garnett, T., Godde, Cc., Muller, A., Röös, E., Smith, P., de Boer, I., zu Ermgassen, E.,
Herrero, M., van Middelaar, C., Schader, C., van Zanten, H., 2017. Grazed and
Confused? Ruminating on Cattle, Grazing Systems, Methane, Nitrous Oxide, the Soil
Carbon Sequestration Question – and What it All Means for Greenhouse Gas
Emissions. Food Climate Research Network, Oxford.
Gephart, J.A., Troell, M., Henriksson, P.J.G., Beveridge, M.C.M., Verdegem, M., Metian,
M., Mateos, L.D., Deutsch, L., 2017. The ‘seafood gap’ in the food-water nexus lit-
erature—issues surrounding freshwater use in seafood production chains. Adv. Water
Resour. 110, 505–514. https://doi.org/10.1016/j.advwatres.2017.03.025.
Gustavsson, J., Cederberg, C., Sonesson, U., Otterdijk, Rv., Meybeck, A., 2011. Global
Food Losses and Food Waste: Extent, Causes and Prevention. Food and Agriculture
Organization of the United Nations, Rome.
Health Canada, 2019. Canada’s Dietary Guidelines. Ottawa.
Hedenus, F., Wirsenius, S., Johansson, D.J.A., 2014. The importance of reduced meat and
dairy consumption for meeting stringent climate change targets. Clim. Change 124,
Heller, M.C., Keoleian, G.A., 2015. Greenhouse gas emission estimates of U.S. dietary
choices and food loss. J. Ind. Ecol. 19, 391–401. https://doi.org/10.1111/jiec.12174.
Henriksson, P.J.G., Tran, N., Mohan, C.V., Chan, C.Y., Rodriguez, U.P., Suri, S., Mateos,
L.D., Utomo, N.B.P., Hall, S., Phillips, M.J., 2017. Indonesian aquaculture futures –
evaluating environmental and socioeconomic potentials and limitations. J. Clean.
Prod. 162, 1482–1490. https://doi.org/10.1016/j.jclepro.2017.06.133.
Hoekstra, A.Y., 2016. A critique on the water-scarcity weighted water footprint in LCA.
Ecol. Indic. 66, 564–573. https://doi.org/10.1016/j.ecolind.2016.02.026.
Hoekstra, A.Y., Chapagain, A.K., Aldaya, M.M., Mekonnen, M.M., 2011. The Water
Footprint Assessment Manual: Setting the Global Standard. Earthscan, London, UK.
Institute of Medicine Food and Nutrition Board, 1998. Dietary Reference Intakes (DRIs):
Thiamin, Riboﬂavin, Niacin, Vitamin B6, Folate, Vitamin B12, Pantothenic Acid,
Biotin, and Choline. National Academy Press, Washington, D.C.
Jones, A.D., Hoey, L., Blesh, J., Miller, L., Green, A., Shapiro, L.F., 2016. A systematic
review of the measurement of sustainable diets. Adv. Nutr. 7, 641–664. https://doi.
KC, K.B., Dias, G.M., Veeramani, A., Swanton, C.J., Fraser, D., Steinke, D., et al., 2018.
When too much isn’t enough: does current food production meet global nutritional
needs? PLoS One 13, e0205683. https://doi.org/10.1371/journal.pone.0205683.
Kiﬀ, L., Wilkes, A., Tennigkeit, T., 2016. The Technical Mitigation Potential of Demand-
Side Measures in the Agri-Food Sector: A Preliminary Assessment of Available
Measures. CGIAR Research Program on Climate Change, Agriculture and Food
Security, Copenhagen. https://cgspace.cgiar.org/handle/10568/77142.
Kim, B.F., Santo, R.E., Scatterday, A.P., Fry, J.F., Synk, C.M., Cebron, S.R., Mekonnen,
M.M., Hoekstra, A.Y., de Pee, S., Bloem, M.W., Neﬀ, R.A., Nachman, K.E., 2019. Data
for: Country-Speciﬁc Dietary Shifts to Mitigate Climate and Water Crises. Mendeley
Data, v2. https://doi.org/10.17632/g8n8w8snmj.1.
Leach, A.M., Emery, K.A., Gephart, J., Davis, K.F., Erisman, J.W., Leip, A., Pace, M.L.,
D’Odorico, P., Carr, J., Noll, L.C., Castner, E., Galloway, J.N., 2016. Environmental
impact food labels combining carbon, nitrogen, and water footprints. Food Policy 61,
Mekonnen, M.M., Hoekstra, A.Y., 2010a. The Green, Blue and Grey Water Footprint of
Crops and Derived Crop Products. UNESCO-IHE, Delft, the Netherlands. http://
Mekonnen, M.M., Hoekstra, A.Y., 2010b. The Green, Blue and Grey Water Footprint of
Farm Animals and Animal Products. UNESCO-IHE, Delft, the Netherlands. http://
Miglietta, P.P., De Leo, F., Ruberti, M., Massari, S., 2015. Mealworms for food: a water
footprint perspective. Water 7, 6190–6203.
Ministry of Health of Brazil, 2014. Dietary Guidelines for the Brazilian Population, 2nd
Morris, C., Kirwan, J., Lally, R., 2014. Less meat initiatives: an initial exploration of a
diet-focused social innovation in transitions to a more sustainable regime of meat
provisioning. Int. J. Sociol. Agric. Food 21, 189–208.
Mottet, A., de Haan, C., Falcucci, A., Tempio, G., Opio, C., Gerber, P., 2017. Livestock: on
our plates or eating at our table? A new analysis of the feed/food debate. Glob. Food
Secur. 14, 1–8. https://doi.org/10.1016/j.gfs.2017.01.001.
Nijdam, D., Rood, T., Westhoek, H., 2012. The price of protein: review of land use and
carbon footprints from life cycle assessments of animal food products and their
substitutes. Food Policy 37, 760–770. https://doi.org/10.1016/j.foodpol.2012.08.
Pahlow, M., van Oel, P.R., Mekonnen, M.M., Hoekstra, A.Y., 2015. Increasing pressure on
freshwater resources due to terrestrial feed ingredients for aquaculture production.
Sci. Total Environ. 536, 847–857. https://doi.org/10.1016/j.scitotenv.2015.07.124.
Payne, C.L., Scarborough, P., Rayner, M., Nonaka, K., 2016. A systematic review of nu-
trient composition data available for twelve commercially available edible insects,
and comparison with reference values. Trends Food Sci. Technol. 47, 69–77.
Pelletier, N., Pirog, R., Rasmussen, R., 2010. Comparative life cycle environmental im-
pacts of three beef production strategies in the Upper Midwestern United States.
Agric. Syst. 103, 380–389. https://doi.org/10.1016/j.agsy.2010.03.009.
Peters, G.P., Hertwich, E.G., 2008. CO2 embodied in international trade with implications
for global climate policy. Environ. Sci. Technol. 42, 1401–1407. https://doi.org/10.
Popp, A., Lotze-Campen, H., Bodirsky, B., 2010. Food consumption, diet shifts and as-
sociated non-CO2 greenhouse gases from agricultural production. Glob. Environ.
Change 20, 451–462. https://doi.org/10.1016/j.gloenvcha.2010.02.001.
Pradhan, P., Reusser, D.E., Kropp, J.P., 2013. Embodied greenhouse gas emissions in
diets. PLoS One 8, e62228.
Sadras, V.O., Grassini, P., Steduto, P., 2009. Status of Water Use Eﬃciency of Main Crops:
SOLAW Background Thematic Report - TR07. Food and Agriculture Organization of
the United Nations, Rome. http://www.fao.org/ﬁleadmin/templates/solaw/ﬁles/
Säll, S., Gren, I.-M., 2015. Eﬀects of an environmental tax on meat and dairy consumption
in Sweden. Food Policy 55, 41–53. https://doi.org/10.1016/j.foodpol.2015.05.008.
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx
Searchinger, T.D., Wirsenius, S., Beringer, T., Dumas, P., 2018. Assessing the eﬃciency of
changes in land use for mitigating climate change. Nature 564, 249–253. https://doi.
Schyns, J.F., Hoekstra, A.Y., Booij, M.J., Hogeboom, R.J., Mekonnen, M.M., 2019. Limits
to the world’s green water resources for food, feed, ﬁber, timber, and bioenergy. Proc.
Natl. Acad. Sci. U. S. A. 116, 4893–4898. https://doi.org/10.1073/pnas.
Semba, R.D., 2016. The rise and fall of protein malnutrition in global health. Ann. Nutr.
Metab. 69, 79–88. https://doi.org/10.1159/000449175.
Springmann, M., Godfray, H.C.J., Rayner, M., Scarborough, P., 2016. Analysis and va-
luation of the health and climate change cobeneﬁts of dietary change. Proc. Natl.
Acad. Sci. U. S. A. 113, 4146–4151. https://doi.org/10.1073/pnas.1523119113.
Springmann, M., Mason-D’Croz, D., Robinson, S., Wiebe, K., Godfray, H.C.J., Rayner, M.,
Scarborough, P., 2017. Mitigation potential and global health impacts from emissions
pricing of food commodities. Nat. Clim. Change 7, 69–74. https://doi.org/10.1038/
Stehfest, E., Bouwman, L., van Vuuren, D.P., den Elzen, M.G.J., Eickhout, B., Kabat, P.,
2009. Climate beneﬁts of changing diet. Clim. Change 95, 83–102. https://doi.org/
Swinburn, B.A., Kraak, V.I., Allendar, S., Atkins, V.J., Baker, P.I., Bogard, J.R., Brinsden,
H., Calvillo, A., De Schutter, O., Devarajan, R., et al., 2019. The global syndemic of
obesity, undernutrition, and climate change: the Lancet Commission report. Lancet
393, 791–846. https://doi.org/10.1016/S0140-6736(18)32822-8.
Thurstan, R.H., Roberts, C.M., 2014. The past and future of ﬁsh consumption: can sup-
plies meet healthy eating recommendations? Mar. Pollut. Bull. 89, 5–11. https://doi.
Tichenor, N.E., Peters, C.J., Norris, G.A., Thoma, G., Griﬃn, T.S., 2017. Life cycle en-
vironmental consequences of grass-fed and dairy beef production systems in the
Northeastern United States. J. Clean. Prod. 142, 1619–1628. https://doi.org/10.
Tilman, D., Clark, M., 2014. Global diets link environmental sustainability and human
health. Nature 515, 518–522. https://doi.org/10.1038/nature13959.
Tukker, A., Goldbohm, R.A., de Koning, A., Verheijden, M., Kleijn, R., Wolf, O., Pérez-
Domínguez, I., Rueda-Cantuche, J.M., 2011. Environmental impacts of changes to
healthier diets in Europe. Ecol. Econ. 70, 1776–1788. https://doi.org/10.1016/j.
U.S. Food and Drug Administration, 2016. CFR - Code of Federal Regulations Title 21.
United States Department of Agriculture, 2017. Food Composition Databases. Accessed
December 2017. https://ndb.nal.usda.gov/ndb/search/list.
Van Dooren, C., Aiking, H., 2016. Deﬁning a nutritionally healthy, environmentally
friendly, and culturally acceptable low lands diet. Int. J. Life Cycle Assess. 21,
Vanham, D., Comero, S., Gawlik, B.M., Bidoglio, G., 2018. The water footprint of diﬀerent
diets within European sub-national geographical entities. Nat. Sustain. 1, 518–525.
Verdegem, M.C.J., Bosma, R.H., 2009. Water withdrawal for brackish and inland aqua-
culture, and options to produce more ﬁsh in ponds with present water use. Water
Policy 11, 52–68. https://doi.org/10.2166/wp.2009.003.
Vermeulen, S.J., Campbell, B.M., Ingram, J.S.I., 2012. Climate change and food systems.
Annu. Rev. Environ. Resour. 37, 195–222. https://doi.org/10.1146/annurev-
Waterman, J.J., 2001. Measures, Stowage Rates and Yields of Fishery Products. FAO,
Weindl, I., Bodirsky, B.L., Rolinski, S., Biewald, A., Lotze-Campen, H., Müller, C., Dietrich,
J.P., Humpenöder, F., Stevanović, M., Schaphoﬀ, S., Popp, A., 2017. Livestock pro-
duction and the water challenge of future food supply: implications of agricultural
management and dietary choices. Glob. Environ. Change 47, 121–132. https://doi.
Wellesley, L., Happer, C., Froggatt, A., 2015. Changing Climate, Changing Diets:
Pathways to Lower Meat Consumption, London.
Willett, W.W., Rockström, J., Loken, B., Springmann, M., et al., 2019. Food in the an-
thropocene: the EAT-Lancet Commission on healthy diets from sustainable food
systems. Lancet 393, 447–492. https://doi.org/10.1016/S0140-6736(18)31788-4.
Wirsenius, S., Hedenus, F., Mohlin, K., 2011. Greenhouse gas taxes on animal food pro-
ducts: rationale, tax scheme and climate mitigation eﬀects. Clim. Change 108,
The World Bank, 2018. World Bank List of Economies (June 2018). Accessed December
World Cancer Research Fund and American Institute for Cancer Research, 2018. Diet,
Nutrition, Physical Activity and Cancer: A Global Perspective. Continuous Update
Project Expert Report 2018. https://www.wcrf.org/dietandcancer/resources-and-
World Health Organization and Food and Agriculture Organization of the United Nations,
2003. Diet, Nutrition, and the Prevention of Chronic Diseases. WHO Technical Report
Series 916. http://apps.who.int/iris/bitstream/handle/10665/42665/WHO_TRS_
B.F. Kim, et al. Global Environmental Change xxx (xxxx) xxxx