Introduction
Recently the United States Department of Agriculture (USDA), in collaboration with the Environmental Protection Agency, launched an initiative to reduce food waste throughout the supply chain (USDA, 2013). The USDA has estimated that 31% of food produced in the USA in 2011 was wasted, an amount equivalent to 59 million metric tons of food and estimated to be worth $(US) 161.6 billion, as purchased at retail prices (Buzby et al., Reference Buzby, Wells and Hyman2014). Globally, an estimated one-third of all food produced for human consumption is lost or wasted (FAO, 2013). As global population increases along with rising affluence (United Nations DESA, Population Division, 2013), minimizing food waste and its corresponding environmental impacts will be advantageous for both achieving food security and for striving toward economic, social and environmental sustainability.
It has been estimated that 95% of food waste generated in the USA enters a landfill (Buzby et al., Reference Buzby, Wells and Hyman2014). In addition to reducing the reliance on landfills as a waste destination, separating and treating food waste could also reduce methane gas release from landfills. Further, food waste offers the potential for recovery of energy and nutrients, if the food waste is removed from the waste stream entering the landfill and treated. Source reduction is the preferred option in the EPA Food Recovery Hierarchy; however, some amount of food waste is unavoidable, e.g., inedible fractions such as peelings from vegetables. As waste managers consider treatment options it will be valuable to estimate both the edible (i.e., potentially reducible) fraction of food waste and the inedible fraction to understand the potential range of capacity for designing treatment systems. As communities consider options for dealing with their food waste streams, an understanding of the volume, composition and variability of these streams is needed to inform the decision-making process and potentially justify the capital investments required for separation and treatment operations.
Understanding the origins and composition of food waste throughout the supply chain is important so that management and behavioral interventions can be optimized. As Koester (Reference Koester2013) discusses, the location at which food loss occurs in the supply chain significantly impacts the economic value of the lost food, as well as the extent to which the loss could have been avoided (and the related resources redirected toward other uses). Detailed information regarding the type, quantity, cost and chemical composition of food wasted at each stage in the supply chain is lacking. This sort of detailed information is necessary when evaluating management options from inventory decisions to selection of a waste treatment technology. In addition, it is unclear how much of this food waste stream is edible versus inedible to humans. Accordingly, the objective of this study is to provide such information for the food waste occurring at the retail level of the supply chain for an institutional food service provider. This information will be important for estimating how much food waste can be avoided through management and individual behavior compared with the unavoidable organic waste stream.
Institutional food service operations offer an opportunity both to obtain data from a controlled setting (as compared with households) and, since these operations are highly managed, to more readily facilitate interventions to reduce waste. A few studies have estimated the quantity of food waste generated at the institutional level (Buzby and Guthrie, Reference Buzby and Guthrie2002; Engström and Carlsson-Kanyama, Reference Engström and Carlsson-Kanyama2004), but these previous studies lack detail, particularly with regard to meat-based foods and the inedible portion of the food that is wasted. In this study, food waste was collected from four all-you-care-to-eat Campus Dining Services (CDS) facilities over 3 months in 2014. CDS is operated by the University of Missouri and approximately 75% of the food (by retail value) is supplied by a single major food distribution company. Pre-consumer food waste was separated into eight categories and post-consumer food waste was separated into 18 categories. In both instances the edible and inedible portions of the total food were inventoried.
When food is wasted, the upstream inputs are also wasted; an objective of this study was to demonstrate this point by estimating the greenhouse gas emissions (GHGs) embodied in the food waste observed. Agricultural production results in numerous negative environmental impacts, including: GHGs; nutrient, herbicide and pesticide pollution; decreased biodiversity; water resource depletion and soil erosion (Cassman et al., Reference Cassman, Wood, Choo, Cooper, Devendra, Dixon, Gaskell, Shabaz, Lal, Lipper, Pretty, Primavera, Ramankutty, Viglizzo, Wiebe, Kadungure, Kanbar, Khan, Leakey, Porter, Sebastian and Tharme2005). Since tracking began in 1990, the agricultural sector has consistently contributed between 6 and 7% of total GHGs in the USA based on the 100-yr global warming potential (GWP) conversions (EPA US, 2014). Venkat (Reference Venkat2011), estimated that from a life-cycle perspective, food waste is responsible for at least 113 million metric tons GHG (CO2e, 100-yr equivalents). The detailed inventory provided herein enables a deeper understanding about the specific types of food waste, which can be used to identify interventions to reduce the occurrence of food waste, and to estimate the environmental impacts embodied in food waste.
Definitions of food loss and waste in the literature
The Food and Agricultural Organization (FAO) (2013)defines six different phases of the food supply chain: agricultural production, post-harvest handling and storage, processing, distribution, consumption and end of life. This study is focused on the food services or consumption stages. Although food waste can occur at all phases in the food supply chain, the largest losses in industrialized countries are understood to occur at the last phase, i.e., the consumption phase (Kantor et al., Reference Kantor, Lipton, Manchester and Oliveria1997; Parfitt et al., Reference Parfitt, Barthel and Macnaughton2010). Kantor et al. (Reference Kantor, Lipton, Manchester and Oliveria1997) estimated that 25% of US edible food was lost in 1995 due to overpreparation, leftovers and extensive choices of menu. Accordingly, the consumption stage has been more thoroughly investigated than any other stage (Buzby and Guthrie, Reference Buzby and Guthrie2002; Engström and Carlsson-Kanyama, Reference Engström and Carlsson-Kanyama2004; Parfitt et al., Reference Parfitt, Barthel and Macnaughton2010).
There is no specific, agreed upon definition of food waste (Buzby and Hyman, Reference Buzby and Hyman2012). In the literature, two different terms are used: food waste and food loss. Buzby et al. (Reference Buzby, Wells and Hyman2014) define food loss as the amount of edible food that is available for human consumption and is not consumed for any reason and food waste as an edible item that goes unconsumed as a result of action or inaction of human activity. The FAO defines food loss as a decrease in mass or nutritional content of food and food waste as food appropriate for human consumption being discarded. Plate waste, a specific type of food waste, has been defined as edible food returned after being served to customers (Buzby and Guthrie, Reference Buzby and Guthrie2002).
The food waste literature also differentiates between lost or wasted food according to three other categories: avoidable, possibly avoidable and non-avoidable (WRAP, 2008; Parfitt et al., Reference Parfitt, Barthel and Macnaughton2010). Considerable amounts of food waste come from avoidable waste, such as: loss due to poor storage or management and/or lack of consumer awareness or action (WRAP, 2008; Al-Domi et al., Reference Al-Domi, Al-Rawajfeh, Aboyousif, Yaghi, Mashal and Fakhoury2011). Serving loss and preparation loss constitute non-avoidable food waste, while storage loss, overproduction, spoilage and plate waste are generally considered avoidable food waste. This study did not explicitly define what portions of the food waste inventoried were avoidable or unavoidable, but implicitly assumes that all edible food waste could be avoided, while inedible food waste is considered unavoidable. Spoiled foods were inventoried in this study as inedible.
Finally, food waste in the consumption stage is often categorized as being either pre-consumer or post-consumer waste. Storage loss, preparation loss, serving loss and overproduction constitute pre-consumer waste, whereas post-consumer waste is typically defined as food purchased by a consumer and subsequently not eaten. Overproduction could be conceptualized as post-consumer food waste since, despite not being sold to customers, it was prepared with that intention; however, it was not possible to separate these food waste categories given the method of collection and the volume of material collected. Most studies in the literature have focused on post-consumer plate waste (Buzby and Guthrie, Reference Buzby and Guthrie2002; Al-Domi et al., Reference Al-Domi, Al-Rawajfeh, Aboyousif, Yaghi, Mashal and Fakhoury2011; Whitehair et al., Reference Whitehair, Shanklin and Brannon2013). Few studies in the literature directly measure food waste and even fewer estimate food waste with sufficient detail to allow for estimation of embodied GHG emissions. For instance, the embodied GHG emissions for beef are an order of magnitude larger than any other meat-based food (de Vries and de Boer, Reference de Vries and de Boer2010; González et al., Reference González, Frostell and Carlsson-Kanyama2011); thus, it is important that weights for individual foods be collected. The range in GHG emissions for vegetable-, grain- and fruit-based foods is narrower, reducing the need for separation by individual food (González et al., Reference González, Frostell and Carlsson-Kanyama2011). Whitehair et al. (Reference Whitehair, Shanklin and Brannon2013) measured only total edible plate waste and Engström and Carlsson-Kanyama (Reference Engström and Carlsson-Kanyama2004) and Al-Domi et al. (Reference Al-Domi, Al-Rawajfeh, Aboyousif, Yaghi, Mashal and Fakhoury2011) consolidate all meats from collected plate waste into one category. Pre-consumer waste has received some attention, most notably in studies considering waste generated in kitchens during final food preparation (Engström and Carlsson-Kanyama, Reference Engström and Carlsson-Kanyama2004; Goonan et al., Reference Goonan, Mirosa and Spence2014). Note that since food often arrives somewhat or heavily processed, losses and waste occurring in upstream manufacturing stages are outside of the boundary of this and most existing analyses.
Many studies have only considered the edible fraction of food waste (Buzby and Guthrie, Reference Buzby and Guthrie2002; Al-Domi et al., Reference Al-Domi, Al-Rawajfeh, Aboyousif, Yaghi, Mashal and Fakhoury2011; Williams and Walton, Reference Williams and Walton2011), as the primary consideration of those studies was the loss of nutritional and economic value. However, when considering treatment and disposal options for lost or wasted food, the inedible fraction is also of interest. In fact, even if all efforts to reduce the loss of food waste across the supply chain are achieved, this inedible fraction will still need to be treated. Therefore, this study has inventoried this fraction of the food waste stream as well. Definitions used in this study are as follows:
-
• Pre-consumer food waste includes kitchen waste, spoiled food and excess food. Kitchen food waste includes any organic material that is discarded during normal food preparation activities, e.g., peels, fat, bones, ends of fruits or vegetables; these materials are considered inedible. Spoiled food waste is defined as food items past their expiration date or visibly decaying and is considered inedible. Waste due to overpreparation is defined as prepared but goes unconsumed; this food is considered edible.
-
• Post-consumer food waste includes food served to consumers that is not eaten (Williams and Walton, Reference Williams and Walton2011). The distinction between edible and inedible post-consumer food waste was based on typical eating behaviors, e.g., banana peels, chicken bones, apple cores, were categorized as inedible.
Methods
Food waste data collection
In the Spring 2014 semester, food waste was collected from four different CDS facilities at the University of Missouri (MU) in Columbia, Missouri. The four facilities are all-you-care-to-eat, tray-less operations and included Pavilion at (Dobbs), Rollins, Plaza 900 and The Mark. The study period covered the 3 months from February 17 to May 16, 2014. Food waste was collected at each facility, transported to another location, separated into different food categories and weighed.
Pre-consumer waste could only be collected from Dobbs, note that we collected all of the food waste generated at the facility over a given day. During the initial phase of the pre-consumer food waste collection process (February 17–April 29, totaling 48 unique collections), pre-consumer food waste was separated into two categories: edible and inedible wastes. During the later phase (April 30–May 16, totaling eight collections), the edible and inedible pre-consumer food wastes were further classified into the following categories: grains, fruits and vegetables, meat and protein, and dairy, for a total of eight categories. Further distinction among the pre-consumer food waste was prohibited due to the aggregate nature and quantity (on average, 130 kg of food waste per day) of the material collected (see photographs in the supplementary material for an impression of the size of the pre-consumer food waste collections). Student workers collected the waste from the kitchen, transported it to a compost facility, and sorted the food as described above. In all cases, qualitative descriptions of the contents of each of these categories were recorded to note the specific food items collected, e.g., lettuce, chicken breasts, etc., and the total weights of the edible and inedible portions were recorded. Examples are provided in the supplementary material.
Post-consumer food waste was collected from all four dining facilities on campus. Student employees collected a total of 42 post-consumer food waste samples each consisting of 100 large and small plates at each facility once per week during the study period. Of the 42 samples, 21 were lunch samples, 16 were dinner samples and 5 were breakfast samples. The food waste was placed into 5-gallon buckets, covered with a snap-on lid and placed into a refrigerator until it was sorted. Students sorted the food waste into the following edible and inedible categories: beef, poultry, pork, dairy, eggs, fish, grains, fruits and vegetables, for a total of 18 individual categories. Each category was weighed separately and the specific food items in each category were recorded.
GHG estimate for food waste
GHG emissions associated with edible food waste only were estimated using average life-cycle emissions reported by González et al. (Reference González, Frostell and Carlsson-Kanyama2011). Estimation for the edible fraction only is consistent with other Life Cycle Assessment (LCA) estimates for foods (de Vries and de Boer, Reference de Vries and de Boer2010; González et al., Reference González, Frostell and Carlsson-Kanyama2011; Hamerschlag and Venkat, Reference Hamerschlag and Venkat2011). González et al. (Reference González, Frostell and Carlsson-Kanyama2011)determined the average GHG emissions for the functional unit of 1 kg for 84 individual foods derived from production through delivery to a Swedish port. The study of González et al. (Reference González, Frostell and Carlsson-Kanyama2011) was utilized because very few LCA studies report such a large, comprehensive range of food products with consistent assumptions and boundaries. This is critical as boundaries across studies vary substantially, making it difficult to ensure a common baseline for comparison across studies (Reap et al., Reference Reap, Roman, Duncan and Bras2008; Roy et al., Reference Roy, Nei, Orikasa, Xu, Okadome, Nakamura and Shiina2009; de Vries and de Boer, Reference de Vries and de Boer2010). Life-cycle stages included in the study boundary are: primary agriculture and processing data, fertilizer manufacturing, emissions associated with machinery, transportation and electricity at the farm (González et al., Reference González, Frostell and Carlsson-Kanyama2011). Packaging, distribution from the port to retail outlets and to household, additional manufacturing of food products, preparation and disposal were not included. Inclusion of these additional stages is infeasible due to lack of a consistent and publically available dataset. Studies have shown that the majority of GHGs occur in the production stages, i.e., prior to leaving the farm gate (Weber and Matthews, Reference Weber and Matthews2008; Hamerschlag and Venkat, Reference Hamerschlag and Venkat2011). Inclusion of these additional stages would modestly increase these GHG estimates. However, these estimates provide a comparable baseline to demonstrate that considerable amounts of GHGs are embodied in food waste and represent the relative differences across food types. Estimates for meat products were made assuming bone-free carcass weight.
Additional assumptions were required in order to apply the GHG estimates to the food waste dataset compiled. First, not all of the fruits and vegetables found in the food waste collected from CDS were included in the study of González et al. (Reference González, Frostell and Carlsson-Kanyama2011), nor were the weights of individual fruits and vegetables recorded in our study. Therefore, the average of the fruits (0.3, 0.3 and 0.38 kg CO2e per kg apple, orange, strawberries, respectively) and vegetables (0.4, 0.1, 0.12, 0.08, 0.2, 0.1, 0.19, 0.3, 0.09 kg CO2e per kg broccoli, cabbage, carrots, cucumber, lettuce, onions, potato, tomatoes, winter squash, respectively) included in González et al. (Reference González, Frostell and Carlsson-Kanyama2011) was calculated and applied to the fruit, vegetables, and fruit and vegetables categories in this study. Secondly, the majority of the food products observed in the ‘grains’ category were wheat-based bread products and additional grain-based foods were not included in the analysis. While rice was observed occasionally, it was impractical to separate it from the collected food waste, given that students collected plate waste into one bucket and hand-sorted. In order to apply the GHG estimate from González et al. (Reference González, Frostell and Carlsson-Kanyama2011) for wheat on a dry-weight basis to the mass recorded in the grain category the mass was multiplied by a factor to convert bread products to whole wheat flour [0.67 kg wheat: 1 kg wheat flour (Meisterling et al., Reference Meisterling, Samaras and Schweizer2009)] and a conversion to dry-weight equivalent [88.5% (Hong et al., Reference Hong, Swaney and Howarth2011)]. Finally, foods observed in the pre-consumer meat and protein category included beef, chicken, pork, seafood and beans. As noted above, further categorization of the pre-consumer food waste was infeasible. In order to create a conservative estimate for the pre-consumer food waste, a weighted-average of GHG estimates was calculated using the ratios from post-consumer observations, Figure 3, and estimates from González et al. (Reference González, Frostell and Carlsson-Kanyama2011)including beef (29 kg CO2e per kg), pork (8.2 kg CO2e per kg), chicken (4.7 kg CO2e per kg), fish (3.1 kg CO2e), egg (3 kg CO2e) and beans (0.86 kg CO2e).
Results
Data analysis
The University of Missouri's CDS provided access to their production and inventory management data, which is tracked using the CBORD software system (www.cbord.com). From inventory and sales data contained within CBORD we obtained the total number of customers served and the weight of total food served over the interval of February 17 to May 16, 2014, across the four CDS dining facilities, for each of the breakfast, lunch and dinner meals; these data appear in Table 1. During the study period approximately 246.3 t of food reached the retail level at the four facilities. 232.4 t of this food was served and 13.9 t (10.1 t of edible and 3.8 t of inedible) were lost as pre-consumer waste. Over the same time period, an estimated 26.4 t of post-consumer food waste was generated at these facilities, composed of 21.2 t of edible and 5.3 t of inedible. As illustrated in Figure 1, overall, 5.6% of food reaching the retail level was lost at the pre-consumer stage and 10.7% was lost at the post-consumer stage.
Figure 1. Materials flow diagram for a representative ton of food entering campus dining operations.
Table 1. Total customers and amount of food served over the study period (February 17–May 16, 2014) by facility and meal.
Pre-consumer food waste
Over this time interval, all pre-consumer food waste was collected from one facility (Dobbs). The total weight of this pre-consumer food waste was 4210 kg, of which 1140 kg were identified as being inedible, with the remaining 3070 kg classified as edible (supplementary table S4). Recall from Table 1 that 70,500 total kg of food was served at Dobbs over this timeframe. This suggests that 74,700 total kg of food reached the retail level at Dobbs (defined here as the total food served plus the total pre-consumer food waste). Thus, we estimate that 5.6% of the food reaching the retail level at this facility was lost as pre-consumer food waste. Pre-consumer food waste per customer totaled 24.6 g per customer (6.7 g inedible, predominately fruit peels; and 17.9 g edible, predominately grains 9.9 g and fruits 6.3 g per customer, respectively; Fig. 2, Table 3).
Figure 2. Weight of pre- and post-consumer food waste by food category, edible and inedible fraction, and GHGs estimated for the edible fraction of wasted food.
In order to determine if there were statistically significant differences between the amounts of food waste identified in each of the eight categories described above across different days, a Kruskal–Wallis test (all statistical analyses were performed using IBM SPSS Statistics 22; Tamhane and Dunlop, Reference Tamhane and Dunlop2000) was performed on the pre-consumer food waste collected during the study's later phase. This analysis found no statistically significant difference across different days, at this one facility, for each of these eight food waste categories, at a 0.05 level of significance. Thus, we concluded that the composition of pre-consumer food waste that was observed over the later phase of the study could justifiably be applied to the total amounts of edible and inedible pre-consumer food waste collected in the study's initial phase.
Note that we did not obtain similar pre-consumer food waste data for the other three facilities that were examined during the post-consumer phase of our study. However, the primary factor that influences pre-consumer food waste is the management function that determines ordering and production decisions at each facility. Assuming that this management function is performed in a similar manner across these four facilities, that facilities serve similar menus, and that ratios of post-consumer food waste were statistically similar across facilities, one would expect to observe similar pre-consumer food waste amounts and composition relative to the total amount of food served. Thus, in order to generate an estimate of the total pre-consumer food waste generated across these four facilities, we applied the factor observed at Dobbs (5.6% of the food reaching the retail level at this facility lost as pre-consumer food waste) to the total amount of food served across all four facilities. This would suggest a total of 13,820 kg of pre-consumer food waste generated over this timeframe. If we further apply the relative percentages of the pre-consumer food waste observed at Dobbs, across the eight food waste categories, to this total 13,820 kg of pre-consumer food waste, we obtain estimates for all four facilities.
Post-consumer food waste
As described above, post-consumer plate waste was collected from four CDS facilities. A Kruskal–Wallis test (Tamhane and Dunlop, Reference Tamhane and Dunlop2000) was performed in order to determine if there were statistically significant differences between the amounts of post-consumer food waste identified in each of the 18 categories described above (i) across different sampling days, for a constant facility and meal; and (ii) across different facilities, over all days, for a constant meal. This analysis found no statistically significant differences across different days, for a constant facility and meal, for each of the 18 food waste categories, at a 0.05 level of significance. Further, no statistically significant differences were observed across different facilities, for each of the 18 food waste categories, for a constant meal, at a 0.05 level of significance. Thus, we concluded that the per-customer estimates of post-consumer food waste generated at each meal, divided into the 18 categories discussed above, could justifiably be applied to the total number of customers served at these four facilities over the entire timeframe of our analysis (Fig. 2).
Using this information, we estimated plate waste per customer, in each category, for each meal; we present the mean, 25th percentile, and 75th percentile values observed across all facilities in our data, shown in Table 2. The largest weights recorded across all meals were in the grain category. Edible fruit was the second largest, followed by vegetables. Depending on the meal, vegetable food waste was larger than fruit; fruit food waste was larger at breakfast and vegetable food waste larger at lunch and dinner. Poultry is the fourth largest category by weight over all meals, followed by beef and eggs (the bulk of which occurred at breakfast), pork, dairy and, finally, fish. Additional detail regarding the meat-based food waste composition is provided in Figure 3; the difference in weight versus embodied GHG emissions is clearly shown.
Figure 3. (a) Weight and corresponding GHG emissions associated with the edible portion of post-consumer food waste in the meat and protein category; (b) percent contribution of individual meats to total weight and GHG emissions associated with the edible portion of post-consumer food waste in the meat and protein category.
Table 2 Estimated post-consumer food waste, per customer.
Note: Total is computed as sum of all values corresponding to a meal/edible versus inedible categorization/statistic.
Table 3 presents our best single-point estimates of the food waste generated per customer at these four facilities over this timeframe, separated into edible and inedible components, for each of pre- and post-consumer waste, across four food types. While the food-type categories were different in the pre- and post-consumer analyses, observe that the post-consumer categories were simply more disaggregated, and could be easily aggregated up to the pre-consumer categories, assuming beef, poultry, pork, fish and eggs (from the post-consumer analysis) taken together comprise the ‘meat and protein’ category (from the pre-consumer analysis). Given that most of the meat-based foods arrive at CDS de-boned, very little inedible food waste was observed in these categories. The primary items observed were chicken wing bones and shrimp shells. The largest quantity of inedible food waste was observed in the fruit category; examples include banana and orange peels, apple cores, melon rinds, etc.
Table 3. Per customer estimates of food waste mass by category and GHGs for edible food waste; and food lost by category as a percentage of food entering the dining facility.
Liquid or semi-liquid items, particularly dairy products such as butter, melted cheese, yogurt and ice cream, were not well captured by the collection method. Cheese was the predominate food product weighed in the dairy category in the post-consumer portion of the analysis and was often melted onto a meat or bread food product, limiting the ability to isolate the material for weighing. Cow's milk, juices and soft drinks were collected and weighed in aggregate and not included in this analysis. Space and time restrictions within the kitchens during collection did not allow for further separation of these liquid items.
GHG results
Over the study period an estimated 26.4 g CO2e per customer were embodied in the pre-consumer food waste and 99.8 g per customer in post-consumer food waste for a total of 124.5 g per customer. Recall that GHG estimates were applied to the edible portions of estimated food waste. The meat and protein category represents the largest embodiment of GHG emissions in both the pre- (17.2 g CO2e per customer) and post-consumer (87.6 g CO2e per customer) categories despite ranking fourth and below in total weight. Beef represents the largest contribution to post-consumer GHG emissions embodied in food waste with an estimated 60.7 g CO2e per customer followed by 13.8 g CO2e per customer for poultry, 8.7 g CO2e per customer for pork, 3.5 g CO2e per customer for eggs and 1.1 g CO2e per customer for fish (Figure 3a and b). GHGs embodied in the pre-consumer meat and protein food waste category were estimated at 17.2 g per customer, although further breakdown is not possible as described above. However, in aggregate, this category is still the largest source of CO2e in the pre-consumer category. The meat and protein category ranks only third in terms of mass alone in both pre- and post-consumer categories, reflecting the relative GHG-intensity of animal-sourced foods. The grain category is the second largest contribution of GHGs in both the pre- and post-consumer categories; embodied GHGs are estimated at 3.4 and 5.7 g CO2e per customer, respectively. The GHGs embodied in food waste in the fruit and vegetables category are estimated at 1.4 g CO2e per customer for pre-consumer and 3.3 g CO2e per customer for post-consumer categories. Finally, the dairy category is estimated to have embodied 2.7 (pre-consumer) and 3.2 (post-consumer) g CO2e per customer.
Discussion
Food wasted as a percentage of the total amount of food entering the dining operations is similar or slightly lower to that found in other studies (Buzby and Guthrie, Reference Buzby and Guthrie2002; Engström and Carlsson-Kanyama, Reference Engström and Carlsson-Kanyama2004). Recognizing that the estimates provided by the Food and Agriculture Organization (FAO, 2013) and the United Nations Department of Economic and Social Affairs (United Nations DESA, Population Division, 2013) include food losses across all stages in the food supply chain, results were compared against studies that focused on food wasted in the retail and consumption stages. Engström and Carlsson-Kanyama (Reference Engström and Carlsson-Kanyama2004) found that on average 20% of the food delivered to Swedish food service institutions, including the inedible portions, was lost. Results from this study are slightly lower, with 16.3% loss for food entering the retail level, perhaps reflecting existing efforts by CDS to reduce losses. CDS has implemented efforts to reduce food waste over multiple years, including removing trays from all facilities and posting food waste information and the weight of plate waste observed.
Plate waste losses between 9 and 11% of food entering the institution found in Engström and Carlsson-Kanyama (Reference Engström and Carlsson-Kanyama2004) overlap with 10.7% observed plate waste rate found in this study. Whitehair et al. (Reference Whitehair, Shanklin and Brannon2013) found an average of 57 g of edible food waste per tray, while our study found approximately 40 g of edible food waste per customer. The smaller estimate found in our study may reflect the lack of trays at CDS operations, the removal of which has been found to reduce plate waste by up to 25 to 30% (Aramark, 2008). Engström and Carlsson-Kayama (Reference Engström and Carlsson-Kanyama2004) also found that meat-based foods were the smallest food waste category by weight, with grains or vegetables contributing the most. Buzby and Guthrie (Reference Buzby and Guthrie2002) reviewed multiple studies that found fruit and vegetables were the most wasted items in the plate waste category. In our study, grain-based products constituted the largest food waste category by weight, closely followed by fruits and vegetables.
The GHG estimates suggest that consideration of more than the quantity of solid waste generated via food waste should be taken into consideration when evaluating the environmental impact of food waste in relation to sustainability. Additional metrics could be explored from this lens, such as nutrient or pesticide runoff or land use. Institutions concerned with their GHG footprint may want to consider the GHGs embodied in their food waste. In particular, the additional detail within the meat and protein category from this study provides much needed data for evaluating embodied environmental impacts, which have been demonstrated to vary by an order of magnitude across animal-based food products (de Vries and de Boer, Reference de Vries and de Boer2010) and up to two orders of magnitude between animal-based and vegetable/grain-based food products (González et al., Reference González, Frostell and Carlsson-Kanyama2011).
The GHG data applied to the edible portion of wasted food included emissions only to the farm gate, a common boundary in food LCA studies. There are a number of additional sources of GHGs associated with the manufacture, transportation and preparation of foods. For example, the energy and corresponding GHGs due to our predominately fossil-based energy system, required to slaughter and process chicken to manufacture chicken nuggets, and the energy consumed by the ovens or embodied in the packaging materials containing them, are not included in this analysis. Quantification of these GHGs is in the nascent stages of research development due to the myriad of manufacturing and cooking possibilities. While inclusion of these additional stages would increase overall emissions, it is unlikely that the overall conclusions drawn herein would be reversed. For example, in order for a chicken-based food product to supersede a beef-based product for the same amount, the manufacturing and cooking emissions of chicken would have to be over six times greater than the manufacturing and cooking emissions associated with beef.
Consistent with the comments of Koester (Reference Koester2013): (i) this analysis focuses solely on items at a single level (consumption) of the food supply chain, although a distinction is made between waste occurring at the pre- and post-consumer stages; and (ii) the total food losses have been disaggregated into their edible and inedible portions, for four (pre-consumer) or nine (post-consumer) different food types. Focusing on the consumption level prevents the aggregation of waste for a common food type (e.g., an apple) across different levels of the supply chain, which is problematic since, e.g., an apple consumed at a farm and an apple transported to a consumer's home have significantly different economic value and embodied resource consumptions. Disaggregating between edible and inedible portions of food waste allows for more realistic reduction goals to be identified, since some amount of food waste is not avoidable, even at the consumption level.
This level of disaggregation allows for a distinction to be made between the food types that generate the most kg of food waste (fruits and vegetables, if considering all waste; grains, if considering only edible waste) and the food types that generate the most kg CO2e associated with food waste (meat and protein). This distinction is relevant when considering alternative treatment options for food waste. If the desire is solely to reduce disposal and landfill costs (which are generally incurred on a weight-basis) by diverting food waste from the trash stream, efforts should focus on reducing the waste from fruits and vegetables and from grains. However, were the objective of such alternative treatment options to reduce GHG emissions, efforts to reduce the waste from meat and protein would likely be most effective, despite the relatively small masses of such wastes (and, correspondingly, relatively small opportunity to decrease disposal and landfill costs via such reductions). Further, in the EPA's food waste hierarchy, reduction is the most desirable outcome; however, in the event that these efforts are highly successful the inedible fraction of the food waste stream will still occur. Therefore, inclusion of and distinctions between both the edible and inedible fractions is important for evaluating the most preferable treatment option.
This is a question with significant practical importance: food service managers need to plan their production in an uncertain demand environment; inventory theory (Hillier and Lieberman, Reference Hillier and Lieberman2015), as well as common sense, dictate that the production level will not be set to an average demand level, due to concerns about production shortfalls below demand. But which food types should be overproduced (on average) to accommodate such demand uncertainties? Such decisions are informed by considerations of the cost to procure, produce, store and dispose of food; such cost factors could be potentially extended to include environmental considerations. Any policy intended to capture these environmental costs in disposal and landfill costs would face the challenge of requiring an accounting for the composition of food waste by food category. A more easily implemented approach would be to include such emissions costs in the food purchase costs, which would likely reduce the served amount of these high-relative-emissions foods (like meats and proteins) and increase the served amount of low-relative-emissions foods, and indirectly reduce total food waste-derived emissions. Such an extension of inventory theory to address these considerations relative to food waste, whether implemented at the procurement or at the disposal level, would help food service managers to identify economically rational food loss, and is an intended subject of future research. Moreover, all of the facilities that were examined in this study were all-you-care-to-eat facilities. This fact likely influences the amount of plate waste generated, since diners are not paying any marginal cost for additional food, and thus might be viewed as having less incentive to avoid taking excess food. What is perhaps less intuitive is that this fact also likely influences the amount of pre-consumer waste, since food service providers do not receive marginal revenues for additional food items sold, and thus do not have the typical lost sales incentive to avoid underproduction of items. Because CDS (and many other university food providers) operates both à la carte and all-you-care-to-eat facilities, an examination of these effects is also intended for future research.
Supplementary Material
For supplementary material accompanying this paper, visit http://dx.doi.org/10.1017/S1742170514000209
Acknowledgements
The authors would like to thank our funding sources: the Richard Wallace Faculty Incentive Grant and the Mizzou Advantage Initiative. We thank the undergraduate students who sorted through tons of food waste: Nicholas Boshonek, Bennett Harman, Trevion McGhaw, Allison Pittman, Sami Tellatin, Sara Wasinger, and Thomas Welby. Many thanks to the Campus Dining Services employees, Nancy Monteer and Eric Cartwright in particular, for their support and assistance in conducting this work. We also thank the three anonymous reviewers who greatly improved the quality of the manuscript. And finally, we thank Margaret Killjoy for editing the manuscript to improve readability.
