Methods to simplify diet and food life cycle inventories: Accuracy versus data-collection resources

Accepted Manuscript Methods to simplify diet and food life cycle inventories: accuracy versus datacollection resources Franck Pernollet, Carla R.V. Coelho, Hayo M.G. van der Werf PII:

S0959-6526(16)30788-0

DOI:

10.1016/j.jclepro.2016.06.111

Reference:

JCLP 7482

To appear in:

Journal of Cleaner Production

Received Date: 26 October 2015 Revised Date:

16 June 2016

Accepted Date: 19 June 2016

Please cite this article as: Pernollet F, Coelho CRV, van der Werf HMG, Methods to simplify diet and food life cycle inventories: accuracy versus data-collection resources, Journal of Cleaner Production (2016), doi: 10.1016/j.jclepro.2016.06.111. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Food at industry gate

Food at retailer gate

Inputs

Inputs

Fc-f: Full cradle to farm gate

Food purchased

M AN U

Fc-r: Full cradle to retailer entrance

RI PT

Food at farm gate

Avoidable, unavoidable and potentially avoidable waste at home

Waste at retail

Waste at industry

SC

Inputs

Allocation related to industry processing

AC C

EP

TE D

Fc-m: Full cradle to mouth

Food uncooked

Inputs

Cooking weight change

Food cooked, ingested

1

ACCEPTED MANUSCRIPT

Word count 7854

2 3

Methods to simplify diet and food life cycle

4

inventories: accuracy versus data-collection resources Franck Pernollet, Carla R. V. Coelho, Hayo M.G. van der Werf*

7

SAS, Agrocampus Ouest, INRA, 35000, Rennes, France

8

[email protected]

9 10

keywords: LCA, LCI, food, diet, method, data-collection

11

M AN U

12

SC

6

RI PT

5

ABSTRACT

14

The number of Life Cycle Assessment (LCA) studies on foods and diets steadily increases.

15

However, due to lack of data on food products as well as time and resource constraints, many of

16

these studies ignore part of the system (e.g. cooking and waste in the household), which may lead to

17

underestimating impacts greatly. This LCA study compared diets using six methods with different

18

system boundaries; three of these are simplified methods we developed. The aim was to identify

19

which method best optimizes data collection for life cycle inventories from cradle to human mouth

20

of food products and diets. The principle behind the three simplified methods was that, for many

21

foods and impact categories, the farm (or fishery) is the life cycle stage that contributes most to

22

impacts. One average, one healthy and one vegetarian diet, each composed of up to 105 foods, were

23

assessed. Climate change, cumulative energy demand, eutrophication, acidification and land

24

occupation impacts were estimated. Recommendations are given on which methods, depending on

25

study goals, offer the best trade-off among available resources (time, money, and knowledge), while

26

providing the required robustness of results. Compared to a full LCA, simplified LCA methods can

27

yield more accurate results at a lower cost of data collection.

AC C

EP

TE D

13

28 29

INTRODUCTION

30 31

Life Cycle Assessment (LCA) is a method that assesses environmental impacts of a product, a

32

service or a system with a specific function and considers all stages of its life cycle (ISO, 2006a). 1

ACCEPTED MANUSCRIPT

This method consists of four phases (ISO, 2006a, b): goal and scope, life cycle inventory (LCI), life

34

cycle impact assessment, and interpretation. The LCI phase is dedicated to data collection and is

35

often the most resource consuming phase (Rebitzer et al., 2004). These issues become amplified for

36

LCA of larger systems. For example, LCA practitioners in civil engineering began studying the

37

building-material level and now study buildings and even districts within a town. The first LCAs in

38

agriculture were at the field level and now may focus on farm or farming-region levels (Loiseau et

39

al., 2012). To address data-collection issues, solutions have been developed (Lasvaux et al., 2014;

40

Loiseau et al., 2014) to bridge data gaps and/or to simplify LCIs by finding a balance between

41

resource availability and the consideration of sufficient detail to achieve acceptable uncertainty

42

(Rebitzer et al., 2004).

An increasing number of LCA studies on food have been published in recent years. Many authors

SC

43

RI PT

33

have focused on the cradle-to-farm-gate stage or on a particular stage of a food product e.g.

45

packaging (Banar and Cokaygil, 2009; Kang et al., 2013), while fewer studies have considered the

46

full life cycle, including distribution, retail, consumption and waste (DEFRA, 2011). To provide a

47

global view of human-nutrition impacts, focus is shifting from single foods to the diet level (Munoz

48

et al., 2010; Scarborough et al., 2014), which amplifies data-collection issues.

49

M AN U

44

To address data-collection issues, LCA practitioners studying diets have used a variety of simplifications such as reducing the number of foods considered or using a proxy to represent a

51

group of foods, i.e. using LCI of product A for product B (Munoz et al., 2010). Due to the

52

complexity of data collection required when investigating the full life cycle, some studies limit their

53

boundaries to the farm gate (Tilman and Clark, 2014) or to the retailer (Berners-Lee et al., 2012;

54

Meier and Christen, 2013; Saxe et al., 2013; Scarborough et al., 2014). Others limit the number of

55

impact categories, considering only the climate change (CC) impact, for which there are more data

56

available than for other impacts (Berners-Lee et al., 2012; Saxe et al., 2013). Methods have been

57

developed to bridge data gaps and simplify data collection for agricultural and food LCIs. Milà i

58

Canals et al. (2011) described two ways to bridge data gaps at the farm level: proxies and

59

extrapolated data (i.e. modify LCI of product A with specific parameters of product B). They also

60

recommended methods to use as a function of the data robustness required. Sanjuan et al. (2014)

61

offered solutions for approximating the energy demand of industrial food processing for several unit

62

operations. They evaluated the error in estimating CC by aggregating unit operations instead of

63

doing a fully detailed LCI. Doublet et al. (2014) developed a method to simplify LCIs of food

64

products from farm to industry gate. They highlighted which LCI data are most relevant for

65

different impacts and life cycle stages. Methods from these three studies help simplify LCI data

66

collection from cradle to farm gate, for industry, and from cradle to industry gate. However, to our

AC C

EP

TE D

50

2

ACCEPTED MANUSCRIPT

67

knowledge, there is no study that explores ways to simplify LCI data collection from cradle to

68

human mouth, addressing lack of data at different life cycle stages and quantifying errors made by

69

these simplifications.

70

We investigated to what extent simplifications in LCAs of foods and diets affect the accuracy of impact results. Three diets were assessed using different system boundaries and using simplified

72

methods. The objective was to establish recommendations on which methods to use, depending on

73

study goals, to obtain the best trade-off between result accuracy and available resources.

RI PT

71

74 MATERIALS AND METHODS

76

1-Full LCA framework

77

ISO-compliant LCA was performed for 105 foods. These foods represent the main food products

SC

75

eaten in French households. This study aimed to identify the contribution of life cycle stages to

79

several environmental impacts and develop a method to simplify the construction of LCIs of foods

80

and diets. This paper presents LCA results at both the food and diet levels. Consequently, two

81

functional units (FU) were used:,1) at the food level, the FU was “one kg of ingested food product

82

in a French two-person urban household”; and 2) at the diet level, the FU was “daily food ingestion

83

in a French two-person urban household”. Current average daily food intake (beverages not

84

included) in France is approximately 1.4 kg/person. The system boundaries were from cradle to

85

human mouth and considered the following stages: farm or fishery (hereafter, “farm” will stand for

86

“farm or fishery”, for simplicity), transport to industry, industrial processing, packaging, transport

87

to retailer through distribution platforms, storage at retailer, retailer waste and treatment, transport

88

to home, home storage, domestic waste, waste treatment, and preparation and cooking at home. The

89

assumption that not all of the food prepared is consumed led to the cradle-to-mouth system

90

boundaries. LCIs for each product represented the product consumed in France. For products whose

91

quantities varied greatly among diets (e.g. meat, meat substitute), additional efforts were made to

92

improve reliability, completeness and representativeness of LCI data. The main simplifications are

93

presented in Table 1, and further detail on data sources for the food products, country of origin for

94

imported products and cooking technology assumptions are listed in the Supplementary Material

95

(SM Table 1, SM Table 2, SM Table 3).The ecoinvent v2.2 database was used for background data.

96

Our study excluded impacts that were assumed to change little among diets (e.g. cleaning agents,

97

cooking tools, cutlery and plates). These assumptions are considered to be in line with cut-off

98

criteria for LCA.

AC C

EP

TE D

M AN U

78

99

Impacts estimated were CC (using 100-y GWP in kg CO2eq.), acidification (AC, in g SO2eq.),

100

and eutrophication (EU, in g PO4eq.) according to CML-IA baseline v4.2; land occupation (LO, in 3

ACCEPTED MANUSCRIPT

101

m2.year) according to CML-IA non-baseline v4.1; and Total Cumulative Energy Demand (CED, in

102

MJeq.) according to CED v1.08 from ecoinvent (renewable and non-renewable, excluding gross

103

calorific energy in biomass). Impacts due to direct and indirect land-use changes were not

104

considered.

105 2-Diets studied

107

Three diets were studied: Average, Healthy and Vegetarian. Each diet represents 15 days of food

RI PT

106

ingestion (excluding alcoholic beverages). The Average diet was adapted from survey data from

109

2010 on Nutritional Behavior and Food Consumption in France (Comportement et Consommation

110

Alimentaire en France) to approximate food consumption of an adult French man (CREDOC,

111

2015). Compared to the survey data, the Average diet supplied approximately the same energy and

112

macronutrients but included only the foods most consumed, for simplification. The Healthy diet

113

resulted from modifying the Average diet to adhere to French nutritional recommendations for

114

macronutrients according to Martin (2001) . The quantity of fruits, vegetables, starchy foods and

115

dairy products increased, and the quantity of meat and pastries decreased. The Healthy diet was

116

modified to obtain a Healthy Vegetarian diet (hereafter called Vegetarian diet): fish and meat were

117

replaced by eggs, dairy products, pulses, vegetables, tofu and mung bean sprouts. Figure 1 shows

118

the composition by mass of each diet by food category. Table 2 shows the main nutritional

119

characteristics of each diet (for further details on the quantities of food products in each diet, see

120

SM Table 4)

121

TE D

M AN U

SC

108

3-Method description

123

For each food product and diet, impacts were calculated according to six methods with different

124

system boundaries (Figure 2, Figure 3, Table 3). System boundaries of the first three, Fc-f, Fc-r and

125

Fc-m, correspond to standard system boundaries of a full LCA from cradle to farm gate, cradle to

126

retailer entrance and cradle to mouth, respectively. The fourth method, S (“simplified”), also has

127

system boundaries from cradle to mouth but does not include all of the processes of a full LCA.

128

Instead, it multiplies impacts of one kg of product from the full cradle-to-farm-gate LCA by the kg

129

of product at the farm gate necessary to obtain one kg of ingested product (based on waste at

130

industry, retailer and home; allocation factors for industry and the cooking weight-change due to

131

rehydration or dehydration during cooking), plus impacts of waste treatment (eq.1). The fifth

132

method, Sc, equals S plus impacts of home cooking (eq.2). Finally, method Sc-t equals Sc plus

133

impacts of transport from farms to industry, from industry to retailers and from retailers to homes

134

(eq.3). The life cycle stages ignored in the simplified methods are described in Table 3.

AC C

EP

122

4

ACCEPTED MANUSCRIPT

135 136

S: Ik,a,mouth = Ik,a,farm × Cind,a × Cret,a × Ch,a × Ccook,a + Ik,wt × Cind,a × (Cret,a × Ch,a - 1)

(eq. 1)

137

Sc: Ik,a,mouth =S + Ik,z × Tcook / Mportion

(eq. 2)

138

Sc-t: Ik,a,mouth =Sc + [(It,k,ind × Cind,a + It,k,ret) × Cret,a + Itk,h ] × Ch,a × Ccook,a

(eq. 3)

139 where

141

Cind,a is the scaled industry-processing coefficient of product ‘a’

142

= (mass of raw farm product/mass of processed product) × allocation factor

143

Cret,a is the retailer-waste coefficient of product ‘a’

144

= 1/(1- proportion of retailer waste of product ‘a’)

145

Ch,a is the home-waste coefficient of product ‘a’

146

= 1/(1- proportion of food wasted at home of product ‘a’)

147

Ccook,a is the cooking-weight-change coefficient

148

= (mass of uncooked product ‘a’)/(mass of cooked product ‘a’)

149

Ik,a,mouth is the impact k of 1 kg of ingested product ‘a’, from cradle to mouth

150

Ik,a,farm is the impact k from cradle to farm gate of 1 kg of raw material used to make product ‘a’

151

Ik,wt is the impact k of waste treatment of 1 kg of product

152

Ik,z is the impact k of cooking with technology z (oven, hot plate, etc.)

153

Itk,ind is the impact k of transport of 1 kg of product from farm to industry

154

Itk,ret is the impact k of transport of 1 kg of product from industry to retailer

155

Itk,h is the impact k of transport of 1 kg of product from retailer to home

156

Mportion is the mass of one cooked portion in kg

157

Tcook is the cooking time in minutes

SC

M AN U

TE D

EP

158

The S methods require data for the industry-processing coefficients Cind, waste coefficients Cret

AC C

159

RI PT

140

160

and Ch, and the cooking weight-change coefficient Ccook (SM Table 5). Cind were found in the

161

literature (SM Table 1). Retailer-waste coefficients (Cret) came from (Beretta et al., 2013;

162

Gustavsson et al., 2011). Cooking weight-change (Ccook) coefficients came from (Bellemans and

163

Maeyer, 2005).

164 165

4-Contribution analysis

166

Life cycle contribution analysis was performed for each food. The farm stage represents the

167

impact of the raw material necessary to provide 1 kg of product at the industry gate. We represented 5

ACCEPTED MANUSCRIPT

168

impacts of food waste as separate processes. The processes “food waste at home” and “retailer

169

waste” included waste treatment.

170 5-Food categories and average foods

172

The 105 foods studied were grouped into 16 categories (SM Table 5); in the figures some food

173

categories are merged. For each food category, an “average food” was calculated as the arithmetic

174

mean of the foods within the category. Since the number of foods per food category was small, a

175

few foods with atypical profiles of contribution of the main life cycle stages to impacts were

176

excluded when calculating the mean. This was the case for 16 foods, e.g. canned pâté and canned

177

tuna (due to the higher impact of packaging), margarine (because it is a mix of many raw materials)

178

and baguette and artisanal bread (due to the higher impact of the industry stage) (for details see SM

179

Table 5). Note that these foods were excluded only when assessing food categories; they were

180

included when assessing diets.

M AN U

181 182 183

RESULTS

184

186

In this study, method Fc-m was considered the reference method; therefore, when comparisons were made, the method whose results were closest to those of Fc-m was considered the best.

TE D

185

SC

RI PT

171

187 188

1-Comparison to literature results

189

In the literature, 15 studies were found that compared the CC impact of diets that were similar to our Average, Healthy and Vegetarian diets (Figure 4). Literature estimates for CC impact due to

191

diet varied more than six-fold (from less than 0.5 to 3 t CO2eq. per person per year). Two of these

192

studies relied on cradle-to-farm gate data to assess impacts of diets, while seven relied on cradle-to-

193

retailer data. Our data allowed us to determine that using cradle-to-retailer and cradle-to-farm-gate

194

system boundaries instead of cradle-to-mouth boundaries resulted in underestimates of impact of

195

approximately 30% and 70%, respectively (Figure 4).

AC C

196

EP

190

For CC, relative differences for Average, Healthy, and Vegetarian diets in our study were similar

197

to those from the literature, but our absolute values were in the lower range of those found in the

198

literature. The main differences and similarities in these studies compared to ours are presented in

199

the supplementary material (SM Table 6).

200 201

2-Comparison of methods 6

202

ACCEPTED MANUSCRIPT

Estimated environmental impacts of diets using simplified and full LCAs with different system boundaries relative to the full cradle-to-mouth LCA are presented in Figure 5. Since the Fc-f

204

system boundary yielded much lower impacts than the other boundaries, its results are presented

205

only in Figure 5 and Table 4. Results of method Sc-t were the closest to reference results (Figure

206

5). For all impacts except CED (for all diets) and CC (for the Vegetarian diet), the simplified

207

methods had results closer to those of the reference method than Fc-r did.

208

Fc-f results were significantly lower than those of the reference method. For all other methods,

209

depending on the diet assessed, estimates of CC were 62-88% of the value of the reference method;

210

estimates for AC, EU and LO varied from 74-96% of the values of the reference method; and CED

211

differed the most, at 27-70% of the reference value (Figure 5). For all impact categories except

212

CED, at least one method estimated impact values within 20% of those estimated by the reference

213

method.

SC

RI PT

203

For all methods, the Average diet had higher impacts than the Healthy diet, which had higher

215

impacts than the Vegetarian diet (Figure 6 and Table 4). Relative differences between diets varied

216

by method, however, sometimes differing greatly from differences estimated by the reference

217

method. For example, with method S, CED of the Vegetarian diet was 43% lower than that of the

218

Average diet, while the reference method estimated it as 16% lower (a difference of 27 percentage

219

points).

M AN U

214

For CC, Fc-r and Sc-t performed best; their estimate of the difference between the Average and

221

Vegetarian diets was within 5 percentage points of the estimate by the reference method. For AC,

222

EU and LO, Sc and Sc-t performed best in estimating the difference between the Average and

223

Vegetarian diets, lying within 6, 3 and 3 percentage points, respectively, of the values estimated by

224

the reference method. For CED, Fc-r performed best, lying within 6 percentage points.

EP

225

TE D

220

3-Contribution analysis

227

Contribution analysis of the average diet (Figure 7) showed that the farm stage contributed nearly

AC C

226

228

57% of CC, around 75% of LO, 72% of AC, 70% of EU, and 29% of CED. Food waste at home

229

contributed 17-22% of all impacts.

230 231

4-Comparison of food categories

232

Figure 8 shows results for the food categories that had the highest contribution to impacts of diets

233

and the food categories that had the largest differences depending on the method used. The

234

performance of most methods varied by food category. Overall, Sc-t did best for all food categories,

235

except for CED of dairy and egg, for which Fc-r had better results (Figure 8). Fc-r significantly 7

ACCEPTED MANUSCRIPT

236

overestimated impacts of wheat and wheat-based products (since it ignores that the water absorbed

237

by these products during cooking increases their weight by 150%) and underestimated impacts of

238

cooked vegetables and potato.

239 240

DISCUSSION

241 1-Comparison with other studies

243

That the CC impact of diets varied more than six-fold in the literature is probably due in part to

244

using data based on unsuitable system boundaries, such as cradle-to-farm-gate and cradle-to retailer

245

(Figure 4). Applying these boundaries to our diets underestimated CC by 30% (cradle to retailer)

246

and 70% (cradle to farm gate). These underestimates had two main causes: 1) impacts of life cycle

247

stages excluded from system boundaries are ignored, and 2) waste from and consequences of

248

(economic) allocation of these stages are also ignored, since product losses and differences in

249

values of co-products are ignored. These observations highlight the importance of using appropriate

250

system boundaries in diet studies and suggest the need for methods that allow a good trade-off

251

between available resources and accurate results.

M AN U

SC

RI PT

242

Our results for CC for the three diets with cradle-to-farm-gate, cradle-to-retailer and cradle-to-

253

mouth system boundaries lay at the lower end of the range of results in the literature, in particular

254

when compared to studies that used cradle-to-farm-gate and cradle-to-retailer boundaries. Factors

255

besides differences in system boundaries may have contributed to the variability observed in

256

literature results, such as differences in food origins, transport mode and distance, production

257

systems, diet compositions, foreground and background data, methods (e.g. allocation), and the

258

consideration and estimated percentages of waste (SM Table 6). This variability complicates

259

comparison of our results with those in the literature; however, relative impacts of our diets agreed

260

with those in the literature. According to a recent review (Hallström et al., 2015), CC and LO

261

decreased by 0-35% and 15-50%, respectively, when changing from an average diet to a healthy

262

diet. Our study found corresponding reductions of 18% and 25%. Hallström et al. (2015) found that

263

moving from an average diet to a vegetarian diet decreased CC by 20-35% and LO by 30-50%. In

264

our study, corresponding reductions were 39% and 46% (Table 4); the larger decrease in CC can be

265

explained by the fact that the vegetarian diet was also a healthy diet.

AC C

EP

TE D

252

266 267

2-Accuracy versus resources required for methods

268

Implementing a full cradle-to-farm-gate LCA (Fc-f) is resource-demanding; however, farm

269

impacts generally cannot be ignored. Fortunately, cradle-to-farm-gate LCIs are becoming 8

ACCEPTED MANUSCRIPT

270

increasingly available, and simplified LCI methods can be used, such as those of Milà i Canals et al.

271

(2011). S methods require a full Fc-f LCA to be implemented. They also require additional data for

273

industry-processing coefficients Cind, waste coefficients Cret and Ch and cooking weight-change

274

coefficients Ccook. Values for these coefficients can be taken from this study (SM Table 5), or

275

literature data can be used if judged more appropriate. Cind coefficients vary in particular when

276

several co-products with different economic values exist (e.g. slaughterhouse output) or when

277

several processing processes with different yields coexist (e.g. whole fish versus fileted fish).

278

Retailer-waste coefficients (Cret) can be found in the literature (Beretta et al., 2013; Gustavsson et

279

al., 2011) and depend on retailer size, location and storage management (Buurman and Velghe,

280

2013). Studies on waste in the household (Ch) are increasingly available in the literature (Beretta et

281

al., 2013; Quested et al., 2013; Quested and Murphy, 2014). Ch is a crucial parameter and must be

282

calculated carefully. When comparing foods or diets, the same reference should be used to calculate

283

the waste generated at home. Cooking weight-change (Ccook) is often available in the literature

284

(Bellemans and Maeyer, 2005) for a specific food and cooking method, but it can be harder to find

285

for products composed of multiple ingredients (e.g. cake, pizza). Overall, data required to move

286

from Fc-f to S are available.

M AN U

SC

RI PT

272

Moving from S to Sc requires few additional resources and decreases impact underestimates,

288

since values for cooking time and mass of cooked portions are included. Moving from Sc to Sc-t

289

requires additional resources but provides better results when compared to the reference method.

290

TE D

287

It is important to note that transport data (Itk,ind,Itk,ret,Itk,h) depend on country, product and season (Milà i Canals et al., 2007; Rizet et al., 2008). Consequently, using local data is recommended.

292

Approximations can be made by using transport of similar products or import-export databases such

293

as FAOSTAT (2014) or Eurostat (2015). In this study, supply chains varied greatly by food

294

category; therefore we suggest using data specific at least to the product category.

AC C

EP

291

295

Fc-r is far more resource-consuming than S, Sc and Sc-t, since a full LCI from cradle to retailer

296

entrance is needed. In particular, industry data are often difficult to obtain for individual products,

297

and data are often confidential. All simplified methods developed here (S, Sc and Sc-t )

298

underestimated impacts compared to the full cradle-to-mouth method because they exclude several

299

life cycle stages (i.e. transport, packaging, industry, cooking). Fc-r demands more resources than Sc-t,

300

but it was generally less accurate. Since Fc-f and Fc-r do not consider waste after the industry gate

301

and dehydration or rehydration during cooking, they may significantly overestimate or

302

underestimate the impact value.

303 9

ACCEPTED MANUSCRIPT

2-a) Suitability of methods for product categories

305

For all food categories, the Fc-r, S, Sc, and Sc-t methods were considered suitable for a given

306

impact when its impact value differed less than 20% from the value obtained with the reference

307

method for the average product of the food category, and when the standard deviation of the mean

308

in the category was less than 20%. This definition is arbitrary, and a method’s suitability depends

309

on the data robustness required for a given study. The system boundary Fc-r is only suitable for

310

products with low rates of waste generated at home and that do not change weight during cooking

311

(Table 5). Based on this observation, this method can be used to assess highly processed products

312

such as dairy products, sugar-based products and oil. Compared to the reference method, Fc-r

313

estimates relatively similar impacts for these products.

RI PT

304

Methods S and Sc can be used for products with a high farm impact and low impacts for cooking

315

and industrial transformation (Table 5). These methods are suitable for products such as meat, dairy

316

products and fish. Method Sc-t is suitable for products with high farm and cooking impacts and low

317

industrial transformation impacts. It is suitable for products such as homemade dishes (e.g.

318

shepherd’s pie, quiche, and pizza, which are made of multiple food products), cooked vegetables,

319

fruit, pulses, meat, dairy products and fish.

M AN U

SC

314

Some foods were excluded when calculating the “average food” of their food categories because

321

their impacts differed greatly from those of similar products. This was the case, amongst others, for

322

products in high-impact packaging (e.g. canned tuna, canned pâté, jam in glass jars). Packaging

323

should be included in the LCI for such products. Certain food categories, such as pulses, require a

324

relatively long cooking time; for these we recommend using Sc rather than S for CED and CC. For

325

imported foods, such as certain fruits and vegetables, transport is often non-negligible; therefore, we

326

recommend using Sc-t rather than S or Sc for CC, CED and AC.

EP

327

TE D

320

2-b) Suitability of methods for specific impacts

329

As presented in the contribution analysis (Figure 7), EU results mainly from the farm stage;

AC C

328

330

consequently, methods S, Sc and Sc-t are almost always suitable. Since AC is due mainly to the farm

331

stage and to transport to the retailer, S and Sc are suitable for AC for products with little transport to

332

the retailer, and Sc-t is suitable for products whose transport impacts up to retailer cannot be

333

excluded (e.g. vegetables, fruit, pulses). CC results from all stages from cradle to mouth, and those

334

that contribute most vary and depend on food category. Consequently, for CC it is not possible to

335

identify a method that is most suitable.

336

LO results mainly from the farm stage but also from the packaging stage, since wood is used in

337

cardboard. Thus, methods S, Sc, and Sc-t are suitable for LO of all foods except those with cardboard 10

ACCEPTED MANUSCRIPT

packaging (e.g. rice, pasta, couscous, pre-cooked durum wheat). For CED of products that are

339

highly processed (e.g. dairy products, wheat-based products, oil, sugar-based products), methods S,

340

Sc, and Sc-t are not suitable (Table 5). When products have negligible cooking or home-waste

341

components (e.g. sugar-based products), Fc-r may be suitable for estimating CED. When both the

342

processing stage, as well as cooking weight-change or waste generated at home, are non-negligible

343

(e.g. wheat-based and rice products), the simplified methods, Fc-f and Fc-r, are not suitable for

344

estimating CED. In this case, the method of Sanjuan et al. (2014) can be used to supplement Sc-t.

RI PT

338

345 3- Method biases when comparing diets

347

In this study, meat products had shorter transport distances than vegetables and fruits. Because

348

Healthy and Vegetarian diets contain less meat and more fruits and vegetables than the Average diet

349

(Figure 1), methods S and Sc (which exclude transport) estimated greater differences in CC (Figure

350

5), EU and AC (Table 4) between the Average and Vegetarian diets than the reference method. For

351

CED, S and Sc estimated greater differences between diets than the reference method because they

352

do not consider energy use for transport and industry. LO was due mainly to the farm stage, but

353

cardboard for packaging also contributed. In this study, plastic packaging was assumed for meat,

354

and corrugated cardboard packaging was assumed for fruits and vegetables. Because S and Sc

355

exclude packaging, they estimated greater differences in LO than the reference method.

M AN U

Because system boundaries of Fc-r extend up to the retailer and exclude food waste and cooking at

TE D

356

SC

346

home, it estimated higher relative impacts than the reference method for products that rehydrate

358

during cooking (e.g. couscous, rice, pasta, pre-cooked durum wheat, pulses) but estimated lower

359

relative impacts for products with high household waste, such as fruit and vegetables. In our study,

360

the Vegetarian diet contained more wheat-based products and pulses than the Average diet (Table

361

1), so Fc-r estimated smaller differences in impacts between these diets than the reference method.

362

Recent studies (Berners-Lee et al., 2012; Meier and Christen, 2013) that used Fc-r to compare

363

omnivorous and vegetarian diets may have thus underestimated or overestimated (depending on the

364

meat substitutes used) differences between these diets.

AC C

365

EP

357

Because Sc-t excludes industrial processing, compared to the reference method, it estimated lower

366

impacts for highly processed products (e.g. wheat-based food, tofu, coffee, dairy products) and

367

higher relative impacts for less-processed products (e.g. vegetables, fruits, meat, fish). Method Sc-t

368

also estimated larger differences in impacts between Average and Healthy or Vegetarian diets than

369

the reference method, but this is difficult to explain since there were contradictory effects. For

370

example, Sc-t estimated lower relative CED impact for tofu but higher relative CED impact for fruits

371

and vegetables. 11

ACCEPTED MANUSCRIPT

372 373

4-Method choice depends on impact categories assessed for foods and diets

374

For both food and diets, if resources preclude a full LCA, a simplified method can be used to

375

represent a cradle-to-mouth LCA. Study results have been summarized in a flow chart that can be

376

used to guide future LCAs of foods or diets, depending on the goal and available resources (Figure

377

9). If the goal is to assess food, the most suitable method can be selected (Table 5) depending on the

RI PT

378

impacts studied. If several methods are suitable, the least resource-consuming method is

380

recommended. For example, to assess EU and CC impacts of a chocolate cake (a homemade

381

dessert), the S, Sc and Sc-t methods are recommended for EU and Sc-t is recommended for CC (Table

382

5); consequently, Sc-t can be used. If the goal is to calculate absolute impacts of a given diet, we

383

recommend first estimating impacts of each food category within the diet as the category’s mass in

384

the diet × the average impact of 1 kg of food in the category; the latter can be estimated using an

385

average value or, if data are unavailable, a proxy. Once impacts of each food category are

386

estimated, food categories can be ranked to identify those with the highest impacts. For these food

387

categories, Table 5 can be used to decide which methods are most suitable. Finally, the method

388

suitable for most of the high-impact food categories can be selected.

M AN U

389

SC

379

If the goal is to compare relative impacts of several diets, Sc-t is recommended in most cases. However, if resources are limited, Sc can be used for AC, EU, and LO impacts. If there is access to

391

a database with cradle-to-retailer boundaries, Fc-r can be used to compare CC and CED impacts of

392

diets.

395 396 397

5-Study limits and perspectives

EP

394

• Simplified methods were compared to the detailed reference (Fc-m) method, which involved some simplifications, e.g. proxies were used for agricultural production of certain food products.

AC C

393

TE D

390

• This study focused on home-cooked dishes. Industrially produced dishes were not included,

398

although their consumption may not be negligible. However, these two types of dishes have been

399

shown to differ little in their impacts (Heller et al., 2013).

400

• Human excretions were not included due to lack of data; their impacts are small for CED, AC

401

and CC (Munoz et al., 2010), but not for EU. Impacts of treating human excretions depend on

402

diet composition, in particular nitrogen and phosphorous contents (Munoz et al., 2010), which

403

may slightly influence differences in impacts in the diets analyzed, since diets differed in

404

nitrogen contents (highest in Average and lowest in Vegetarian). 12

405 406 407

ACCEPTED MANUSCRIPT

• Uncertainties due to data quality were not assessed due to lack of appropriate data in the AGRIBALYSE LCI database. • Due to a lack of specific inventory data, we limited this study to the five impacts most studied in

408

agri-food systems. It would be interesting in future studies to assess other impacts that are

409

relevant to the food sector, such as water depletion and biodiversity loss due to land use.

410

• Data were collected to represent a French household, but French households have high variability (e.g. distance to retailer, energy consumption of household refrigerator, percentage of

412

waste). Future studies should assess whether our results hold for other countries, since diets and

413

food-supply chains differ and farm impacts vary considerably among countries (Doublet et al.,

414

2014).

RI PT

411

• The average diet was based on food consumption of an adult French man. Since men’s dietary

416

composition may differ from that of women, our results may not fully represent the average

417

French population.

M AN U

SC

415

418 6-Conclusion

420

Many LCA studies of diets use cradle-to-retailer or cradle-to-farm-gate system boundaries and

421

consequently ignore major life cycle stages and tend to underestimate impacts. Compared to a full

422

LCA with inadequate system boundaries, the simplified LCA methods developed can yield more

423

accurate results at a lower cost of data collection. This study developed an approach that provides

424

guidance for obtaining the best trade-off between available resources and the robustness of LCA

425

results.

426

TE D

419

AUTHOR INFORMATION

428

Corresponding author

429

*E-mail: [email protected]

430

AC C

EP

427

431

Notes

432

The authors declare no competing financial interests.

433 434

ACKNOWLEDGMENT

435

This work was supported by the AGRALID project (ANR-12-ALID-0003). We thank Geneviève

436

Gésan-Guiziou from INRA Rennes, UMR STLO, and employees from Terrena Group for supplying

437

data. We thank Michael Corson and three anonymous reviewers for their insightful comments.

438 13

ACCEPTED MANUSCRIPT

REFERENCES

440 441

Banar, M., Cokaygil, Z., 2009. A Life Cycle Comparison of Alternative Cheese Packages. CleanSoil Air Water 37, 136-141.

442 443

Bellemans, M., Maeyer, M., 2005. Poids et mesures - Manuel de quantification standardisée des denrées alimentaires 2nd ed. Conseil supérieur d'hygiène, Brussels, p. 274.

444 445

Beretta, C., Stoessel, F., Baier, U., Hellweg, S., 2013. Quantifying food losses and the potential for reduction in Switzerland - Supplementary Information. Waste Manage. Res. 33, 32.

446 447

Berners-Lee, M., Hoolohan, C., Cammack, H., Hewitt, C.N., 2012. The relative greenhouse gas impacts of realistic dietary choices. Energy Policy 43, 184-190.

448 449

Blake, L., 2014. People, plate and Planet- The impact of dietary choices on health, greenhouse gas emissions and land use. Centre for alternative technology, UK, p. 46.

450 451

Buurman, R., Velghe, J., 2013. Les supermarchés et le gaspillage alimentaire. Centre de recherche et d'information des consommateurs (CRIOC), Brussels, p. 58.

452 453

CREDOC, 2015. Centre de recherche pour l'étude et l'observation des conditions de vie, http://www.credoc.fr/, last accessed 16 October 2015.

454 455

DEFRA, 2011. Evidence to define the sustainability of a healthy diet. Department of Environment, Food and Rural Affairs, United Kingdom, p. 27.

456 457

Delamaire, C., 2011. La santé vient en mangeant. Le guide alimentaire pour tous. Programme National Nutrition Santé (PNNS), France, p. 130.

458 459 460 461

Doublet, G., Ingólfsdóttir, G.M., Yngvadóttir, E., Landquist, B., Jungbluth, N., Aronsson, A., Ramos, S., Keller, R., Ólafsdóttir, G., 2014. Key Environmental Performance Indicators for a simplified LCA in food supply chains, 9th International Conference LCA of Food, San Francisco, USA, p. 10.

462

Eurostat, 2015. Eurostat, http://ec.europa.eu/eurostat, last accessed 18 September 2015.

463 464

FAOSTAT, 2014. Food and Agriculture Organization of the United Nations - Statistics Division, http://faostat.fao.org/, last accessed 23 February 2015.

465 466

Fazeni, K., Steinmüller, H., 2011. Impact of changes in diet on the availability of land, energy demand, and greenhouse gas emissions of agriculture. Energy, Sustainability and Society 1, 6.

467 468

Gustavsson, J., Cederberg, C., Sonesson, U., Van Otterdijk, R., Meybeck, A., 2011. Global food losses and food waste. Food and Agriculture Organization of the United Nations, Rome, p. 38.

469 470

Hallström, E., Carlsson-Kanyama, A., Börjesson, P., 2015. Environmental impact of dietary change: a systematic review. J. Clean Prod. 91, 1-11.

471 472

Hoolohan, C., Berners-Lee, M., McKinstry-West, J., Hewitt, C.N., 2013. Mitigating the greenhouse gas emissions embodied in food through realistic consumer choices. Energy Policy 63, 1065-1074.

AC C

EP

TE D

M AN U

SC

RI PT

439

14

ACCEPTED MANUSCRIPT

ISO, 2006a. Environmental Management - Life Cycle Assessment - Principles and Framework, Geneva, Switzerland.

475 476

ISO, 2006b. Environmental Management - Life Cycle Assessment - Requirements and guidelines, Geneva, Switzerland.

477 478

Kang, D.H., Sgriccia, N., Selke, S., Auras, R., 2013. Comparison of bacon packaging on a life cycle basis: a case study. J. Clean Prod. 54, 142-149.

479 480

Koch, P., Salou, T., 2014. AGRIBALYSE: Methodology, Version 1.1. Ed ADEME, Angers, France, p. 384.

481 482 483

Lasvaux, S., Schiopu, N., Habert, G., Chevalier, J., Peuportier, B., 2014. Influence of simplification of life cycle inventories on the accuracy of impact assessment: application to construction products. J. Clean Prod. 79, 142-151.

484 485

Loiseau, E., Junqua, G., Roux, P., Bellon-Maurel, V., 2012. Environmental assessment of a territory: an overview of existing tools and methods. J. Environ. Manage. 112, 213-225.

486 487 488

Loiseau, E., Roux, P., Junqua, G., Maurel, P., Bellon-Maurel, V., 2014. Implementation of an adapted LCA framework to environmental assessment of a territory: important learning points from a French Mediterranean case study. J. Clean Prod. 80, 17-29.

489 490 491

Macdiarmid, J.I., Kyle, J., Horgan, G.W., Loe, J., Fyfe, C., Johnstone, A., McNeill, G., 2012. Sustainable diets for the future: Can we contribute to reducing greenhouse gas emissions by eating a healthy diet? Am. J. Clin. Nutr. 96, 632-639.

492 493

Martin, A., 2001. Apports nutritionnels conseillés pour la population française, 3 ed. Tec & Doc, Lavoisier, Paris.

494 495 496

Masset, G., Vieux, F., Verger, E.O., Soler, L.G., Touazi, D., Darmon, N., 2014. Reducing energy intake and energy density for a sustainable diet: a study based on self-selected diets in French adults. Am. J. Clin. Nutr. 99, 1460-1469.

497 498

Meier, T., Christen, O., 2013. Environmental Impacts of Dietary Recommendations and Dietary Styles: Germany As an Example. Environ. Sci. Technol. 47, 877-888.

499 500 501

Milà i Canals, L., Azapagic, A., Doka, G., Jefferies, D., King, H., Mutel, C., Nemecek, T., Roches, A., Sim, S., Stichnothe, H., Thoma, G., Williams, A., 2011. Approaches for Addressing Life Cycle Assessment Data Gaps for Bio-based Products. J. Ind. Ecol. 15, 707-725.

502 503

Milà i Canals, L., Cowell, S.J., Sim, S., Basson, L., 2007. Comparing domestic versus imported apples: A focus on energy use. Environ. Sci. Pollut. Res. 14, 338-344.

504 505

Munoz, I., Canals, L.M.I., Fernandez-Alba, A.R., 2010. Life cycle assessment of the average Spanish diet including human excretion. Int. J. Life Cycle Assess. 15, 794-805.

506 507

Quested, T., Ingle, R., Parry, A., 2013. Household Food and Drink Waste in the United Kingdom. WRAP, United Kingdom, p. 135.

AC C

EP

TE D

M AN U

SC

RI PT

473 474

15

ACCEPTED MANUSCRIPT

Quested, T., Murphy, L., 2014. Household Food and Drink Waste: A product focus, 2014 ed. WRAP, United Kingdom, p. 171.

510 511 512

Rebitzer, G., Ekvall, T., Frischknecht, R., Hunkeler, D., Norris, G., Rydberg, T., Schmidt, W.P., Suh, S., Weidema, B.P., Pennington, D.W., 2004. Life cycle assessment part 1: framework, goal and scope definition, inventory analysis, and applications. Environ. Int. 30, 701-720.

513 514 515

Rizet, C., Browne, M., Léonardi, J., Allen, J., Piotrowska, M., 2008. Chaînes logistiques et Consommation d’énergie : cas des meubles et des fruits & légumes. Institut français des sciences et technologies des transports, de l'aménagement et des réseaux (INRETS), France, p. 179.

516 517

Röös, E., Karlsson, H., Witthöft, C., Sundberg, C., 2015. Evaluating the sustainability of diets– combining environmental and nutritional aspects. Environmental Science & Policy 47, 157-166.

518 519

Sanjuan, N., Stoessel, F., Hellweg, S., 2014. Closing Data Gaps for LCA of Food Products: Estimating the Energy Demand of Food Processing. Environ. Sci. Technol. 48, 1132-1140.

520 521

Saxe, H., Larsen, T.M., Mogensen, L., 2013. The global warming potential of two healthy Nordic diets compared with the average Danish diet. Clim. Change 116, 249-262.

522 523 524

Scarborough, P., Appleby, P., Mizdrak, A., Briggs, A.M., Travis, R., Bradbury, K., Key, T., 2014. Dietary greenhouse gas emissions of meat-eaters, fish-eaters, vegetarians and vegans in the UK. Clim. Change 125, 179-192.

525 526

Tilman, D., Clark, M., 2014. Global diets link environmental sustainability and human health. Nature 515, 518-522.

527 528 529

Tukker, A., Goldbohm, R.A., de Koning, A., Verheijden, M., Kleijn, R., Wolf, O., PérezDomínguez, I., Rueda-Cantuche, J.M., 2011. Environmental impacts of changes to healthier diets in Europe. Ecol. Econ. 70, 1776-1788.

530 531 532

Van Dooren, C., Marinussen, M., Blonk, H., Aiking, H., Vellinga, P., 2014. Exploring dietary guidelines based on ecological and nutritional values: A comparison of six dietary patterns. Food Policy 44, 36-46.

533 534

Vieux, F., Darmon, N., Touazi, D., Soler, L.G., 2012. Greenhouse gas emissions of self-selected individual diets in France: Changing the diet structure or consuming less? Ecol. Econ. 75, 91-101.

SC

M AN U

TE D

EP

AC C

535

RI PT

508 509

16

536

ACCEPTED MANUSCRIPT

FIGURES 100%

Coffee Tofu and mung bean sprout 80% Homemade dishes Sugar-based products 60%

RI PT

Fruit Vegetables and potato Pulses

40%

Wheat-based products and rice Oil and margerine 20%

SC

Dairy and eggs Fish

Meat

0% Healthy

538 539

Vegetarian

M AN U

Average

537

Figure 1. Contributions of food categories (by mass) to Average, Healthy and Vegetarian diets. Homemade dishes include meatbased and vegetarian dishes as well as desserts; vegetables and potato include raw and cooked vegetables.

540

542 543 544 545 546 547 548 549

AC C

EP

TE D

541

Figure 2. Unit processes and system boundaries for methods to calculate diet impacts. Text in italics represents parameters included in all simplified LCA methods. Full LCA methods Fc-f, Fc-r and Fc-m have system boundaries from cradle to farm gate, cradle to retailer entrance and cradle to mouth, respectively. Simplified LCA method S multiplies impacts of 1 kg of product from the full cradle-to-farm-gate LCA by the kg of product at the farm gate necessary to obtain one kg of ingested product (based on allocation factors for industry, waste at industry, retailer and home, and cooking weight-change), plus impacts of waste treatment. Simplified method Sc equals S plus impacts of home cooking; simplified method Sc-t equals Sc plus impacts of all transport stages.

17

ACCEPTED MANUSCRIPT

550 Figure 3. Sankey diagram representing chicken fillet as an example to illustrate the simplified methods, based on product flows from farm gate to food ingested at home. Method S multiplies impacts of one kg of product from the full cradle-to-farm-gate LCA by the kg of product at the farm gate necessary to obtain one kg of ingested product, based on waste at industry, retailer and home, allocation factors for industry, and the cooking weight-change due to rehydration or dehydration during cooking.

RI PT

551 552 553 554

556 557 558

AC C

EP

TE D

M AN U

SC

555

Figure 4. Annual climate change impact for one person for Average, Healthy and Vegetarian diets for studies with different boundaries: cradle to farm gate, cradle to retailer, cradle to mouth.

18

Figure 5. Impacts of climate change (CC), cumulative energy demand (CED), acidification (AC), eutrophication (EU) and land occupation (LO) for three diets (Average, Healthy and Vegetarian) relative to the Fc-m full cradle-to-mouth LCA according to five methods: S simplified method; Sc simplified method plus cooking impact; Sc-t simplified method plus cooking and transport impact; Fc-r full cradle-to-retailer LCA, Fc-f full cradle-to-farm-gate LCA.

SC

559 560 561 562 563 564

RI PT

ACCEPTED MANUSCRIPT

Coffee

Fc-m

S

Sc

Sc-t

5

79%

4

75%

76%

84%

TE D

3

2

EP

50%

0

567 568 569 570 571

61%

Homemade dishes Sugar-based products Fruit Vegetables and potato Pulses

57%

1

565 566

82%

Tofu and mung bean sprout

65%

Wheat-based products and rice

50%

Oil and margarine Dairy and eggs Fish Meat

AC C

kg CO2eq/person/day

6

Fc-r

M AN U

7

Figure 6. Contribution of food categories to climate change impact of Average, Healthy and Vegetarian diets according to five methods. Homemade dishes include meat-based and vegetarian dishes as well as desserts; vegetable and potato include raw and cooked vegetables. Percentages indicate relative impacts of Healthy and Vegetarian diets compared to the Average diet. Fc-m full cradle-to-mouth LCA; S simplified method; Sc simplified method plus cooking; Sc-t simplified method plus cooking and transport; Fc-r full cradle-to-retailer LCA.

572 573 19

ACCEPTED MANUSCRIPT

574 575 100%

Cooking Food waste generated at home

80%

Transport from retailer to home

RI PT

Food waste at retailer 60%

Total storage Packaging

40%

SC

Transport up to retailer Industry

20%

0% CC

CED

AC

576 577

EU

LO

EP

TE D

Figure 7. Contribution analysis of the Average diet for the full cradle-to-mouth method (Fc-m). Impacts are climate change (CC), cumulative energy demand (CED), acidification (AC), eutrophication (EU) and land occupation (LO).

AC C

578 579

M AN U

Farm

20

ACCEPTED MANUSCRIPT

580

Meat (11 products)

Dairy and egg (12 products) 1.0

0.5

0.5 LO

CED

CED

0.0

AC

EU

Homemade vegetarian dishes (9 products)

AC

Legend

CC

CC

1.5

3

CE D

LO

AC C

1

EP

1.0

2

0

EU

CED

EU

AC

TE D

Wheat-based and rice products (4 products)

0.5 LO 0.0

M AN U

0.0

EU

CC 1.0

SC

LO

LO

RI PT

CC

CC 1.0

Cooked vegetables and potato (7 products)

AC

0.5

CED

0.0

EU

AC

21

581 582 583 584 585 586 587

ACCEPTED MANUSCRIPT

Figure 8. Comparison of impacts of average products relative to Fc-m for different food categories according to five methods. Impacts are climate change (CC), cumulative energy demand (CED), acidification (AC), eutrophication (EU) and land occupation (LO). A suitable boundary is shown at 20% of the reference value. A method was considered suitable for a given impact when its impact value differed less than 20% from the value obtained with the reference method for the average product of the food category, and when the standard deviation of the mean in the category was less than 20%. Fc-m full cradle-to-mouth LCA; S simplified method; Sc simplified method plus cooking; Sc-t simplified method plus cooking and transport; Fc-r full cradle-to-retailer LCA.

AC C

EP

TE D

M AN U

SC

RI PT

588

22

ACCEPTED MANUSCRIPT

Food or diet?

Food assessment

RI PT

Diet assessment

M AN U

Obtain absolute impacts of one diet

U8lize recommended methods for the food category concerned

Compare rela8ve impacts of several diets

Use Sc-t method

Separate diet into food categories

TE D

Es8mate impacts of each food category using proxies or food average

Rank food categories to iden8fy those that have the highest impacts

• In

case of very limited resources Sc can be used for AC, EU, LO impact

available, Fc-r can be used for CC and CED

• If

589 590

AC C

EP

Choose the method most suitable for impacts concerned

SC

objec8ve

Examine recommended methods for the selected food categories and choose the most suitable method

591 592

Figure 9. Flowchart describing methods to use when resources are not sufficient for a full cradle-to-mouth LCA. See also Table 5.

593

23

ACCEPTED MANUSCRIPT

594 595

Table 1. Data sources and main simplifications made for the full cradle-to-mouth life cycle inventory (LCI).

Stage

Data sources and simplifications Most data came from the AGRIBALYSE LCI database, version 1.11; other data came from the unpublished INRA UMR SAS (Rennes, France) LCI database. Proxies were used for some foods that, according to a literature search, had low greenhouse gas emissions (Error! Reference source not found.). Production variability (e.g. greenhouse versus grown outdoors) was

RI PT

Farm

considered only when data were available. For some imported foods for which no regional LCI data were available, only transport data were adapted. When co-products were produced, allocation rules of Koch and Salou (2014) were used.

SC

A single processing technology was considered for each product and was, as much as possible, the dominant technology used in France. Economic allocation was used between food and non-food products, e.g. when allocating impacts between meat and other coIndustry

M AN U

products such as hide, bones, and blood. Mass allocation was used to allocate among different fish and meat cuts. Milk-solid-based mass allocation was used for dairy products. Data came from the literature or from industry.

Most data came from the literature or from industry. Proxies were used when data were missing. For example, pasta packaging was used for semolina packaging. Only one type of Packaging

TE D

packaging was considered per food category; for example, vegetable packaging was assumed to be corrugated cardboard rather than tins or crates. For transport from farm to industry of a given food product, distance was calculated as a weighted average of the farm-to-industry distance for the food produced and transported in France and that for the two main countries exporting the product to France, using FAOSTAT

EP

Transport

(2014) or Eurostat (2015). For transport from industry to retailer, data came from the

AC C

literature or from industry. This stage included storage and waste at retailer, transport to home, home storage, packaging and food waste generated at home, household waste treatment, and preparation and cooking at home. To simplify, waste at home was considered to occur before cooking.

Storage and waste

Packaging waste was considered to be partly recycled (aluminum 9%, plastic 20%, paper and cardboard 34%, steel and iron 45%, glass 65%). Unrecycled packaging and food waste was assumed to be incinerated (53%) or landfilled (47%). Waste disposal processes, percentages for recycling rates, incineration and landfill were from ecoinvent.

596

24

ACCEPTED MANUSCRIPT

Table 2. Main nutritional characteristics for Average, Healthy and Vegetarian diets.

Nutritional characteristic

Average

Healthy

Vegetarian

Energy (kcal/day)

2299

2284

2174

Protein (g/day)

103.8

84.5

73.3

Fat (g/day)

95.2

93.5

84.0

Carbohydrates (g/day)

256.6

276.2

598

SC

Abbreviation

Name

Fc-f

Full LCA from cradle to farm gate

Fc-r

Full LCA from cradle to retailer

Fc-m

Full LCA from cradle to mouth Simplified method

Sc

Sc-t

Main stages ignored in the boundaries

Cradle to farm gate

Cradle to retailer

Cradle to mouth

Cradle to mouth

Transport, industrial processing, packaging, storage, cooking

Simplified method plus cooking impact

Cradle to mouth

Transport, industrial processing, packaging, storage

Simplified method plus cooking and transport impacts

Cradle to mouth

Industrial processing, packaging, storage

EP

TE D

S

Boundaries

AC C

600 601

281.1

Table 3. Abbreviations, names, system boundaries and main stages ignored for the six methods studied.

M AN U

599

RI PT

597

25

ACCEPTED MANUSCRIPT

Table 4 Relative differences (in %) in impact of Healthy and Vegetarian diets compared to the Average diet according to six methods for climate change (CC), cumulative energy demand (CED), acidification (AC), eutrophication (EU) and land occupation (LO). Shaded cells indicate values for the two methods closest to the Fc-m value; the method closest to the Fc-m value is identified with a
. Values in parentheses are differences (in percentage points) from results of the Fc-m method. Fc-m full cradle-to-mouth LCA; Fc-f full cradle to farm LCA; Fc-r full cradle-to-retailer LCA; S simplified method; Sc simplified method plus cooking impact; Sc-t simplified method plus cooking and transport impact.

CED 12.7 1.6

AC 30.4 18.5

EU 20.8 3.4

LO 25.3 11.8

Fc-f

(-8.0)

(-11.1)

(-11.9)

(-17.4)

(-13.5)

(-12.9)

(-3.9)

(-25.2)

(-33.8)

(-33.0)

16.5

7.6

26.1

15.2

21.6

34.9

9.7

44.0

30.0

Fc-r

37.1

(-7.6)

(-10.5)

(-8.5)

57.5

43.7

S

48.1

Sc Sc-t

608

(-2.0)

(-5.1)

(-4.3)

(-5.6)

(-3.7)

(-3.9)

(-6.4)

24.5

17.6

33.9

22.4

27.4

50.3

42.9

(+5.9)

(+4.9)

(+3.6)

(+1.6)

(+2.1)

(+11.5)

(+26.7)

(+5.9)

(+3.2)

(+2.5)

24.9

20.6

33.8

22.4

27.4

49.9

38.2

57.2

43.5

48.1

(+6.4)

(+7.8)

(+3.5)

(+1.6)

(+2.1)

(+11.2)

(+22.1)

(+5.6)

(+3.0)

(+2.5)

20.9

13.8

31.5

21.5

27.3

43.7

28.7

53.6

42.2

48.0

(+2.4)

(+1.1)

(+1.1)

(+0.7)

(+2.0)

(+4.9)

(+12.6)

(+2.0)

(+1.6)

(+2.4)

609

AC C

EP

TE D

610

M AN U

Fc-m

RI PT

CC 18.5 10.5

Relative impact difference between Average and Vegetarian diet (in % and percentual points) CC CED AC EU LO 38.8 16.1 51.6 40.5 45.6 25.9 12.2 26.4 6.7 12.6

Relative impact difference between Average and Healthy diet (in % and percentual points)

SC

602 603 604 605 606 607

26

ACCEPTED MANUSCRIPT

611

Table 5 Methods considered suitable for impact assessment according to food category and impact for climate change (CC), cumulative energy demand (CED), acidification (AC), eutrophication (EU) and land occupation (LO). (Sc-t simplified method plus cooking and transport; Sc simplified method plus cooking; S simplified method; Fc-r full cradle-to-retailer LCA).

CC

CED S

Meat










Fish










Dairy and eggs










Oil




Wheat-based products and rice




Pulses










Cooked vegetables and potato










Raw vegetables







Fruit







S

































































































Homemade desserts







Sc-t

Sc

S











































Fc-r

LO

Sc

TE D

Homemade meatbased dishes

Fc-r

Fc-r

Sc

S




Fc-r





























































Sc-t























































































EP

616




S

EU

Sc-t




Homemade vegetarian dishes

Sc

AC C

615

Sc-t

RI PT

Sc

Sugar-based products

Fc-r

AC

Sc-t

SC

Food category

M AN U

612 613 614

27