Diet-related greenhouse gas emissions and major food contributors among Japanese adults: comparison of different calculation methods

Objective: To develop a greenhouse gas emissions (GHGE) database for Japanese foods using three different approaches, compare the results of estimated diet-related GHGE and determine major food contributors among Japanese adults. Design: Cross-sectional. Three GHGE databases were developed: (1) a literature-based method including a literature review of life cycle assessment studies of Japanese foods and (2) production- and (3) consumption-based input–output tables (IOT)-applied methods using the Japanese IOT. All databases were linked to the Japanese food composition table and food consumption data. Diet-related GHGE was estimated based on each database and the 4-d dietary record data. Diet-related GHGE were compared in both total and food group level between the databases. Setting: Japan. Participants: 392 healthy adults aged 20–69 years. Results: The mean diet-related GHGE significantly differed according to the calculation methods: 4145 g CO2-equivalent (CO2-eq)/d by the literature-based method, 4031 g CO2-eq/d by the production-based method and 7392 g CO2-eq/d by the consumption-based IOT-applied methods. It significantly differed in food group level as well. Spearman’s correlation coefficients between three methods ranged from 0·82 to 0·86. Irrespective of the methods, the top contributor to GHGE was meat (19·7–28·8 %) followed by fish and seafood (13·8–18·3 %). Conclusions: Although the identified major food contributors to GHGE were comparable between the three methods, the estimated GHGE values significantly differed by calculation methods. This finding suggested that caution must be taken when interpreting the estimated diet-related GHGE values obtained using the different calculation methods of GHGE.


Overview
This material provides detailed description on the development of the greenhouse gas emission (GHGE) databases for food products consumed in Japan with the following primary purposes: to estimate the diet-related GHGE, to identify the major food contributor among Japanese, and to further investigate the relationship between diet-related GHGE and other variables such as demographic variables, food intake, and diet quality. GHGE databases were developed based on life cycle assessment (LCA), the reference method in evaluating the environmental impact of products including foods, in three different ways. One was developed using a literature-based method, in which the results of the previous LCA studies on GHGEs of food products consumed in Japan were extracted. Two were developed based on the global link input-output (GLIO) model (2) , which describes the relationship between the production and consumption systems of Japan and other countries through of international trade. GHGE value of each food was obtained from the literature review or GLIO model, then linked to the food items in the STFCJ 2015 (1) , which were selected from frequently consumed food among Japanese children and adults. Supplemental Figure 1 summarized the method of database development described below.

Literature-based method
In the literature-based method, GHGE database was developed by a literature review of existing LCA studies for foods consumed in Japan. When there was no LCA study regarding a particular food from the literature review, LCA data from other countries were also used.

Literature search
In the literature-based method, GHGE database was developed by literature review. The systematic literature search for LCA studies that focused on foods consumed in Japan was completed in July 2018 with the following three types of literature: peer-reviewed journal papers, conference proceedings and grey report. Searches for peer-reviewed journal papers and conference proceedings in English were completed in MEDLINE (PubMed), Web of Science, Environmental Science Database -ProQuest, Ebsco, and Google Scholar using the keywords ("life cycle assessment" OR "life cycle analys*s" OR "LCA" OR "life cycle") AND ("greenhouse gas*" OR "GHG*" OR "carbon dioxide" OR CO2 OR "global warming potential" OR GWP) AND (Japan*) AND "food name (both plural and singular form)." Searches for literature in Japanese were performed using CiNii and Google scholar. Moreover, two Japanese journals ("Journal of Life cycle assessment, Japan" and "Journal of the Japanese Agricultural Systems Society") were manually searched. Inclusion criteria and exclusion criteria were as follows.

Inclusion criteria:
Process LCA studies for the food which is included in the STFCJ 2015 and produced in Japan or imported to Japan.
LCA studies including at least one "cradle to farm-gate." The GHGE value that should be reported in carbon dioxide equivalent (CO2-eq) or individually for the following three main gases-CO2, nitrogen oxide (N2O), and methane (CH4).
The study that discloses system boundary, functional unit, and location of production.
GHGE value that was calculated as Carbon footprint or global warming potential (GWP) in a 100-year horizon (GWP 100).
Exclusion criteria: GHGE values that are not presented as CO2-eq or CO2-eq value is not able to be estimated because all three main individual gases (CO2, N2O and CH4) are not presented.
Details about the method of calculation including considered system boundary are not provided.
Study that is not available in the English or Japanese language.
Study that presents results for farms or areas as opposed to a functional unit of a food type (e.g. kg of food as consumed).
LCA values that are reported on feed for livestock. GHGE value that was not calculated according to the other indicators (i.e. GWP in a 20year horizon).
After duplicate articles were removed, title and abstract were screened according to the predefined eligibility criteria. The following data were extracted: author name, study year, publication year, food name, original system boundary, GHGE value, geographic location of the study, LCA approach utilized, farming methods (e.g. conservation or organic) type of literature (e.g. peer-reviewed article, conference report, grey report, etc.), GHGE value or values of the three main greenhouse gases, original functional unit, version of GWP. A total of 47 reports (32 peer-reviewed articles, 9 conference proceeding, 5 reports, 1 University Bulletin; 24 in English and 23 in Japanese) were found (Supplemental Figure 2). The article information and assessed food items are summarized in Supplemental Table 1  GHGE values from other literature or data sources were additionally extracted for some foods because the number of food items with GHGE values obtained from a systematic review were too few to cover the major 310 foods.
As for domestic production food; Step 1: When the value was reported in Barilla Center for Food & Nutrition (BCFN) data sheet (4,5) and its system boundary included production stage, processing stage, and transportation stage, that value was extracted.
Step 2: When there was no value meeting the criteria in Step 1, a further literature search was conducted in Web of Science. In this step, we did not limit the literature with the country or region. If the reported GHGE value was obtained from the literature search, that value was used.
Step 3: When no data were obtained from Step 1 to Step 2, the value was substituted by the value of other foods.
As for imported food: Step 1: Existing data in BCFN data sheet (4) were searched within the country or region which was the primary producer of the targeted food and was ranked in the top three country as importing country of the product to Japan. When the value was reported in BCFN data sheet (4) , and its system boundary included production stage, processing stage, and transportation stage, that value was extracted.
Step 2: When there was no value meeting the criteria in Step 1, a further literature search was conducted in the Web of Science with the country name of the top three importers. If there was more than one report found, the value reported was extracted.
Step 3: When no report was found in Step 2, we used the value in BCFN data sheet (4) without consideration of the country of origin.
Step 4: When there were no data in BCFN data sheet (4) meeting Step 3, further literature search was conducted in Web of Science without consideration of the country name.
Step 5: When no data was obtained from step 1 to 4, the value was substituted by the value of other food.

Determination of greenhouse gas emission (GHGE) value for each food
The functional unit of the original GHGE value extracted from the articles was standardized as "g CO2-eq/g food." When the original functional unit was represented as "per farm area" and the yield per area was also described in the article, per mass GHGE value was calculated.
System boundary was also standardized to "from farm to the regional distribution center or retail." When the original literature included only the production stage as system boundary, GHGE from the post-farm stage was added to the original value according to the previous literature (see Supplemental Table 3). Transport from retail to consumer's home, cooking at home, and the management of food waste and the waste of package were excluded from the system boundary due to the lack of data regarding these processes corresponding to dietary data and LCA study. Furthermore, between-person variation of distance from the retail to the consumer's home, means of transportation, cooking method and equipment made it difficult to take these processes into account. However, the cooking method of rice was almost the same in any household, and the emission from the rice cooking was considered according to the existing LCA data (6) .
After the functional unit and system boundaries were standardized, the sample means of the GHGE value accounting each food were calculated. Different cultivation or feeding method (i.e. conservation and organic) were treated in the same manner. As a result, GHGE values for 163 foods were obtained. These 163 GHGE values were assigned to 2231 food items including 2229 food items in the Standard Tables of Food Composition in Japan 2015 (STFCJ 2015) (1) and additional two food items "water for cooking" and "water for drinking," which is not originally included in STFCJ. The values were determined according to the following five-step method.
Step 1: When GHGE values were available from the literatures, that value was assigned to the same food or different form of the same food in the STFCJ 2015 (n=910). For example, GHGE value for "white potato" was assigned to both "white potato, raw" (STFCJ 2015 food code: 2017) and "white potato, boiled" (STFCJ 2015 food code: 2019). GHGE values from literature were also assigned to "water for cooking" and "water for drinking." Step 2: When GHGE value was not available in step 1 but the GHGE values for the food which was made with a similar production method to the food in STFCJ 2015 or for the food being the major ingredients of the food in STFCJ 2015 were available from the literatures, that available values were assigned (n=637). For example, GHGE value for "tomato" was assigned to "tomato, canned, without salt" (STFCJ 2015 food code: 6184).
Step 3: When GHGE value was not available in step 2 for the food items in STFCJ 2015 composed of single ingredients, the mean value of the GHGE of the same food group was assigned (n=508). For example, the mean value of GHGE values for vegetables was assigned to "celery" (STFCJ 2015 food code: 6119).
Step 4: When GHGE value was not available in step 2 for the food items in STFCJ 2015 composed of several ingredients (e.g. "carry, retort," STFCJ 2015 food code: 18001), the value was calculated according to the recipe data provided in STFCJ 2015 and nutrition composition data of products. (n=69) Step 5: When GHGE value was not available in the above step, GHGE value for tap water was assigned. Especially, there was no GHGE value for seaweed from the literature, the GHGE value for seaweed and seaweed products were assigned with the value for "tap water." (n=106) Step 6: GHGE values for "breast milk" was assumes as "0." (n=1)

Weight basis adjustments
To take into account the weight change during cooking and wastage, GHGE values were adjusted by the wastage rate and weight change rate with STFCJ 2015 as needed.

Production-based Input-Output Tables-applied method
In the Input-Output Table (IOTs)-applied method, GHGE database was developed based on emission intensity value from the production-based GLIO model (2,7) and price data of each food item.

Data sources
Production-based greenhouse gas emission (GHGEP) for each food item was calculated by multiplying the production costs by GHGE intensities based on the producer price. In this study, GHGE intensities were determined using the GLIO model (2,7) . GHGE intensities were expressed as per standard monetary unit (e.g. t CO2-eq per million Japanese yen; [M-JPY]) for each sector. The production value, production volume, and unit prices (yen per product weight [PW] or volume) of each commodity except for some agricultural products or seafood products included in the sectors could be obtained from the " Table of  where, GHGEP (X) is the producer price-based GHGE per PW for X, Sk is the sector k, Ck,i is the food commodity i included in Sk, and UPP (X) is the unit price for X. This equation assumed that the environmental burden generated from the commodity is in proportion to its price, i.e., the production cost. (2,7) .

Data complement
TDP represents sector name, sub-sector name, production volume, production value, and unit price for each commodity. However, production volume and unit price were not described in TDP 2005 for several agricultural products, seafood products, and alcoholic beverages. When the unit price for Ck,i was described in TDP 2005, that unit price was used to calculate GHGE value for product i.
When the commodity name for a product in TDP 2005 was available, with neither its production volume nor its unit price, the unit price was complemented by using the National Statistics. In this case, production volume for each commodity was retrieved from the National Statistics. Then, unit price was calculated as the quotient of production value per production volume. The National Statistics used for this complement are shown in Supplemental Table 4. The production volume on the National Statistics was cited only when the products in the National statistics were identified as similar with the product described in the TDP and the production volume in that National Statistics for 2000 was same as the production volume in TDP for the year 2000 (TDP 2000).
Note that the number of missing data for unit price and production volume was smaller in TDP 2000 than in TDP 2005. Thus, production volume in TDP 2000 was used to check the validity of using the National Statistics for data complement. For salt and alcoholic beverages, we could not find any National Statistics describing the production volumes consistent with the production volume listed in TDP 2000. Therefore, the shipment unit price (M-JPY/t) calculated as the quotient of the shipment value per shipment quantity which was both obtained from the Census of Manufactures ("Kogyo-Toukei," Ministry of Internal Affairs and Communications) was used as a substitute value. The shipment quantity in the Census of Manufactures and the production volume of the TDP were different but had the same order figures. As for some vegetables and fruits (strawberry, watermelon, melon, cucumber, tomato, eggplant, pumpkin, lettuce, and bell pepper), production values were described using cultivation method (open ground and house) in TDP. On the contrary, the Crop Survey (Ministry of Agriculture, Forestry and Fisheries) described the production volume of these vegetables and fruits in total volume (i.e., production volume by open ground plus production by volume house). Thus, production values in TDP were summed by the type of vegetables or fruits (production value by open ground plus production by value house) and divided by total production volume to obtain the unit price.
The values of the production volume were standardized to "per production weight." For several commodities expressing the values as "per volume" OR "per slice," "per production weight" was calculated using density values or reference amount reported in the STFCJ 2015 or FAO/INFOODS Density Database (8) . If the density value was not available for the food from the previous report, density was calculated assuming 1 ml = 1 g. After data complement, the GHGE value for each TDP food commodity (t CO2-eq/t) was obtained by multiplying the unit price for each commodity (M-JPY/t) by the embodied GHG value for each sector from the GLIO model (t CO2-eq/million yen) including the commodity.
There were no production volume and unit price in TDP for imported food because TDP was the statistical table for the domestic production. Mainly imported foods, such as banana and avocado, were not described in TDP, even their commodity name. To take into these imported foods, we calculated the unit price by dividing imported price by imported volume obtained from the National Trade Census 2005 (Ministry of Finance). The process of this data complement was partly described in Supplemental Figure 3.

Determination of GHGE value for each food
Production-based GHGE for each food item was obtained by multiplying production-based intensity value by sector in GLIO model and food price data of each food items. Consequently, GHGE values for 354 food items were obtained. GHGE values for 2231 food items including 2229 food items in the STFCJ 2015 (1) and additional two food items "water for cooking" and "water for drinking," which is not originally included in STFCJ 2015. The values were determined according to the following eight-step method.
Step 1: When only one identical food commodity existed in TDP for the STFCJ 2015 food items, that TDP food commodity assigned. Food commodity "mineral water" in TDP was assigned to "water for drinking." (in total, n=1564) Step 2: When more than two commodities existed in TDP for one food item in STFCJ 2015, and these TDP commodities were included in the same sectors in TDP, embodied values were calculated and were assigned as follows: where GHGEP(X) is the producer based GHGE for X, Fm is the food m, Sk is the sector k, Ck,i is the product value of commodity Ck,i, n is number of TDP commodities identified, and PVL(X) and PWT(X) are the product value and volume for product X. For example, for "tomato, raw" (STFCJ 2015 code = 6182), three commodity items included in the sector "Fruits", "tomato, open-field culture" (TDP commodity code = 113010107), and "tomato, greenhouse culture" (TDP commodity code = 113020107) were identified. In this case, embodied unit price was calculated as the summed production value divided by the summed production volume of three commodities. Then, GHGE value for "tomato, raw" was calculated as GHGE value for sector "Fruits" multiplied by the embodied unit price. (n=51) Step3: When more than two food commodities in TDP were identified and these commodities were included in different sectors, their mean values were selected as the GHGE values. For example, two commodities "salmon/ Sea surface fishery sector" (TDP commodity code= 1710111051) and "salmon/ Inner water surface fishery sector" (TDP commodity code= 312010001) were identified for "chum salmon, raw" (STFCJ 2015 code= 10132). Thus, mean GHGE value of "salmon/ Sea surface fishery sector" and "salmon/ Inner water surface fishery sector" was used as the GHGE value for "chum salmon, raw." As for tea and coffee, two commodities in different forms were identified. In this case, GHGE value was adjusted to the weight of the beverage form, then their mean value were selected. For example, "green tea, leaves/ Tea and coffee sector" (TDP commodity code=1129011101) and "green tea beverage/ Beverage sector" (TDP commodity code=1129021301) were identified as "green tea, sencha, infusion" (STFCJ 2015 code=16036). Assuming that tea infusion was made by 10g tea and 430 ml hot water according to the STFCJ 2015, GHGE value of "green tea, infusion" was obtained by multiplying the GHGE value for "green tea, leaves" by 10/440. Then, GHGE value for "green tea, sencha, infusion" was calculated as the mean value of "green tea, infusion" and "green tea beverage" (n=41).
Step 4: When identical food commodity was not available in TDP but was available in the National Trade Census 2005, the GHGE value calculated by multiplying embodied GHGE value from GLIO model by unit price obtained from the imported value and the imported value was assigned. (n=4) Step 5: When no identical food commodity was available in TDP, a similar commodity with comparable producing or processing process was assigned. (n=316) Step 6: When there was no similar item available in TDP, and the food was composed of single ingredient, embodied GHGE value obtained from the commodities included in the same sector was assigned (n=230).
Step 7: When there was no similar item available in TDP, and the food was composed of the multiple ingredients, the values were calculated from the recipe from Food Commodity Intake Database because there was no standard recipe database in Japan. (n=23) Step 8: GHGE values for "water" was assumed "0" due to lack of appropriate data for tap water to calculate unit price. GHGE values for "breast milk" was also assumes as "0." (n=2) Japanese self-sufficiency was low except for rice, vegetables, potatoes, some fruits, milk and egg, and many foods were imported from abroad. Unfortunately, there were no detailed data on the self-sufficiency of each food. Additionally, the food classification of TDP and National Trade Census was inconsistent. Therefore, we could not take into account all imported foods.
Unit price from TDP was assigned to calculate GHGE value for STFCJ 2015 food items, regardless of the proportion of the food that was imported from abroad except for a few cases described in Step 4. For example, unit price from TDP was used to calculate the GHGE value for soybeans whose self-sufficiency rate was 51% assuming that the unit price of the imported soybeans was similar to that of the domestic soybeans.

Weight basis adjustments
The crude GHGE value obtained by multiplying the embodied GHGE from GLIO for each sector by the cost of each commodity was based on the uncooked products including inedible parts such as vegetable skin, seed, and fish born. To consider the weight change during cooking and from wastage, GHGE values were adjusted by the wastage rate and weight change rate with STFCJ 2015. Tables-applied method IOTs-applied method, GHGE database was developed based on emission intensity value from the consumption-based GLIO model (2,7) and price data of each food item.

Data sources
Consumption-based greenhouse gas emission (GHGEC) for each food item was calculated by multiplying the costs by GHG emission intensities based on the purchase price for household consumption expenditure. GHG emission intensities were also determined using GLIO model (7) .  (8) was used.

Data complements
For food items selected as the mainly consumed food but whose price was not identified in the NRP 2005, prices were taken from the websites of the nationally distributed supermarket (Seiyu, AEON, and Ito-Yokado Japan). However, the price of vegetables, fruits, and seafood could not be obtained from the website because of the seasonality. As a result, food prices of 12 food item were obtained from the websites. (STFCJ 2015) (1) and additional two food items "water for cooking" and "water for drinking," which is not originally included in STFCJ were determined according to the following threestep method. The values were determined by assigning the food commodities in NRP 2005 and additional food price data were obtained above.
Step 1: When there was only one identical food commodity in the NRP 2005 or the complemented price data with the website, that UPc was assigned. (n=1009).
Step 2: When no identical food was available in the NRP 2005 or the complemented price data, UPc of the similar commodity with comparable producing or processing process was assigned (n=650).
Step 3: When there was no similar item available in the NRP 2005 or the complemented price data, mean value of UPc of the same food group was assigned (n=547).
Step 4: For foods composed of several ingredients, the value was calculated according to the recipe data provided in STFCJ 2015 and nutrition composition data of the products (n=23).
Step 5: GHGE values for "water for coking" was assumed "0" due to lack of appropriate data for tap water to calculate unit price. GHGE values for "breast milk" was also assumes as "0" (n=2).

Weight basis adjustments
The crude GHGE value obtained by multiplying the embodied GHGE from GLIO for each sector by the cost of each commodity was based on the products at the discount sale. To consider the weight change during cooking and from wastage, GHGE values were adjusted by the wastage rate and weight change rate with STFCJ 2015as needed.
Supplemental Figure 1. Summary of method to develop the greenhouse gas emission databases IOTs, Input-Output Tables; GHGE, greenhouse gas emissions; GLIO, the global link input-output model (2) ; TDP, Table of  Full text articles included review (n=161) Full text articles included review (n=59) Duplicate (n=5) Unit is not per mass or no data to calculate per product mass GHGE (n=34) Food consumed in other country (n=1) Only CH4 was assessed (n=1) Only CO2 was assessed (n=46) GHG was not assessed (n=1) Production stage was not included (n=9) Product was not human food (n=4) Method description is insufficient (n=4) Hand searching from reference list (n=15) Supplemental Figure 3. The method of data complement in production-based Input-Output Table-applied method.
GHGE, greenhouse gas emissions; JPY, Japanese yen; M-JPY, million JPY; GLIO, the global link input-output model (2) ; TDP, Table of Domestic Products by Sector and Commodity (Bumonbetsu-Hinmokubetsu Kingaku-hyo, Ministry of Internal Affairs and Communications). Note; Pattern 1, when unit value existed in TDP 2005, GHGE for the food item was calculated using the unit value from TDP 2005 and intensity value from GLIO model. Pattern2, when unit value and production value did not exist in TDP 2005, the unit value was calculated using the other National Statistics with three steps. First, search the National Statistics describing the production data for the targeted food. Then, the production volume for 2000 in TDP 2000 and other National Statistics (e.g., Crop survey) were compared to check the validity of using the National Statistics. If these production volumes were the same value, the production volume in 2005 were extracted from that National Statistics. Next, unit price for 2005 was calculated by dividing production value from TDP by production volume from the National Statistics. Lastly, GHGE for the food item was calculated using the calculated unit value and intensity value from GLIO model.