The Sankey diagram in Fig. 1 illustrates the mass flows of all food commodities, in millions of metric tonnes (MMT), along the U.S. FSC in 2016. The detailed calculation procedure used to generate it can be found in Supplementary Notes 4–9 and the results can be found in Supplementary Data 1. The total FLW from the five FSC stages is 343.2 MMT, though due to 7.8 MMT FLW reentering the FSC and consumed via food donation, the total FSC FLW is 335.4 MMT. among the 335.4 MMT of FLW, 19% (63.7 MMT) of FLW is the mass lost via evaporation or other processes for the manufacturing of dried products (e.g., powdered milk and dried fruit). While it is mass lost in the manufacturing stage, and therefore FLW, it was considered a distinct FLW mass flow and kept separate from the remaining manufacturing FLW for the analyses. As this is specifically water removed via drying techniques that do not require wastewater disposal (e.g., evaporation, boiling) it was not included in the analysis to allocate FLW to different management pathways. The mass remains part of the mass flow before the manufacturing stage, as it still consumes energy for growth and transport to and within the manufacturing facilities. The management of on-farm animal-related FLW (8.9 MMT; 2.7%) is described as unknown due to data limitations. In addition, most of the rest of the FLW is either recovered or recycled (185.5 MMT, 55.3%). The remainder of the FLW (77.3 MMT, 23.1%) is disposed by landfilling, incineration, or wastewater treatment. In the following descriptions of FLW composition and management, we will be discussing FLW on a per stage basis, (i.e., a total of 343.2 MMT from all stages), and thus food donation is still considered FLW and as one of the management pathways.
Mass flows and FLW generation by FSC stage
As demonstrated by Fig. 1, a total of 888.7 MMT of agriculture materials were harvested in the U.S. in 2016 (including materials not intended for human consumption), while 40.8 MMT were planted but left unharvested and considered FLW. Among all harvested, 315.9 MMT was used for non-food purposes (e.g., seeding, animal feed manufacturing, or biofuel and alcohol production). That left 572.9 MMT of agricultural materials available for manufacturing, but 12.8 MMT failed to enter the food market (i.e., became unsold FLW) caused by poor quality, damage during harvesting operation, and on-farm storage. In addition, 159.4 MMT of agricultural products left the U.S. as international trade, and a net 18.1 MMT was added to the agricultural stock (difference between the 2016 initial stock and ending stock). Supplementary Note 4 provides more details on the data and calculations for this stage.
While 382.6 MMT of U.S. and 29.5 MMT of imported agricultural materials entered the manufacturing stage in 2016 (for a total of 412.1 MMT), only 203.6 MMT of food products were produced by U.S. food manufacturers (Supplementary Note 5), including 11.2 MMT exported. At the same time, 14.8 MMT of manufactured food products were imported, leading to a total of 207.2 MMT of food entering the distribution stage. With a total of 208.5 MMT of manufacturing FLW (Fig. 1), the U.S. food manufacturing sector appears to be extremely inefficient. However, the estimated manufacturing FLW does not indicate wastefulness, but is simply the mass difference from the beginning and end of the stage. FLW in the manufacturing stage is mostly caused by the necessary separation of edible food from the uncommonly eaten (e.g., offal, organs, cartilage, whey, oil meal), inedible parts (e.g., bones, cores, seeds, shells, germ, and bran), and water (e.g., evaporation during the manufacturing of sugar and some dairy products). Furthermore, most of the manufacturing FLW (140.2 MMT) is recycled or recovered through several pathways (such as animal feed manufacturing), which will be discussed further later.
With 207.2 MMT of food products being distributed and 196.9 MMT entering W&R in 2016 (Fig. 1), the distribution stage accounts for a relatively low amount of FLW (10.4 MMT). This efficiency is likely driven by advanced logistic practices (e.g., cold chains). From the W&R stage, 175.6 MMT arrived at the consumption stage (i.e., food services and households), leaving 21.3 MMT of W&R FLW. The consumption stage also has an input of 7.8 MMT of donated food from the FSC that gets consumed. At this stage, only 133.9 MMT food (including 126.1 MMT of the purchased food and 7.8 MMT donated food29) was consumed, leaving 49.4 MMT as consumption FLW (49.4 from households and food services and 0.38 MMT from food banks). More details regarding the determination of the food donation streams can be found below and in Supplementary Note 8.
Mass flows and FLW generation by food commodity
The mass flows and FLW generation at different FSC stages disaggregated by food commodity are illustrated in Figs. 2 and 3, respectively. As shown in Fig. 2, on-farm production activities of animal products (i.e., M&P, eggs, seafood, and dairy) result in relatively smaller amounts of FLW (2.3 MMT for M&P, 0.6 MMT for seafood, 1.7 MMT for eggs, and 4.3 MMT for dairy) compared to crops (e.g., 9.2 MMT for vegetables and 25.9 MMT for grains). However, considering the relatively high resource demand and GHG emissions from raising animals30,31,32, any animal-related FLW reductions would be impactful. Grains, fruits, and vegetables, as illustrated by Fig. 3, are the highest contributors to the on-farm FLW, but grains are also the highest volume food product at this stage. The ratio between the masses of on-farm FLW and agricultural materials harvested (for both human food and other purposes; details in Supplementary Note 4) shows that grains have a lower loss factor (5.6%) compared to vegetables (16%) and fruits (20%). Seafood, eggs, and nuts (0.5 MMT) have the smallest volumes of on-farm FLW, likely caused by their lower demand.
While several commodities have little mass lost at the food manufacturing stage (e.g., fruit, nuts, and eggs with <20% of loss in their mass flows), others are dominated by FLW (e.g., oil with 81% FLW and dairy with 57% FLW) or water evaporation (e.g., 86% for sugar), as shown by Fig. 2 and Supplementary Note 5 with more details. Figure 3 shows that the largest contributors to manufacturing FLW, are dairy (24.6%), sugar (22.7%), and oil (22.6%). Besides, grain (20%), vegetable (20%), and fruit (14%) add a large amount of FLW at this stage, partially driven by the high demand for these commodities. Conversely, nuts, seafood, and eggs contribute a very small portion of FLW at the manufacturing stage and others, again, because of their relatively low demand.
Five commodities contribute over 85% of the FLW from later stages (distribution, W&R and consumption): vegetables (19%, 17% and 26%), fruit (15%, 17% and 18%), dairy (18%, 22% and 14%), grain products (19%, 21% and 16%) and M&P (14%, 6.6%, and 11.2%), as shown in Fig. 3 and Supplementary Note 6. This is understandable as fresh vegetables, fruits, meat, and poultry, as well as bakery and dairy products are both highly perishable and in high demand. The high FLW generation from perishable products at these stages reveals the need for efforts to prevent FLW by prolonging product shelf-life, which will be further discussed below.
In addition to knowing which stages have the highest FLW, knowing which commodities produce the most FLW (Figs. 2–4) can benefit FLW reduction. Again, while sugar and oil products have a high contribution to total FLW (15.4% and 16.2%), this FLW is mainly caused by the unavoidable removal of inedible (but still recyclable) portions or water during manufacturing and is not caused by spoilage or FSC inefficiency. After oil and sugar, dairy products make up the largest portion of the FLW (20%), followed by grains (17.8%), vegetables (13.8%), fruits (7.1%), all of which include products with high demand and perishability. These commodities’ FLW is largely generated on-farm and during manufacturing but FLW from those stages, and as shown below, is mostly recycled or recovered, whereas the FLW generated in the later FSC stages is still considerable and mostly disposed of through landfill, wastewater, or incineration (LWI). This observation once again shows the importance of technologies or actions that can prolong product shelf-life or optimize storage and transportation conditions. M&P also contributes a large amount of FLW at each FSC stage, except for on-farm production. The large amount of M&P FLW generated at the distribution, W&R, and consumption stages is, again, caused by the high demand and perishability. Conversely, the manufacturing M&P FLW is mainly unavoidable (inedible or undesirable) and can only be reduced by lower demand (e.g., diet change or reducing FLW at downstream stages). Even though on-farm production does not generate a large amount of M&P FLW, the high costs, resource demands, and GHG emissions relating to animal farming make the minimization of on-farm M&P FLW a continued priority. The other food commodities (e.g., nuts, eggs, seafood) contribute a very small portion (2.7%) of U.S FLW, due to their overall low demand.
FLW by management pathway
While Fig. 1 provides a general picture of the management of U.S. FLW from each FSC stage, Fig. 5 expands on this and provides more details on the used FLW management pathways throughout the FSC. The FLW management in the on-farm stage has not been well studied, so this study is only able to separate the crop-based on-farm FLW into three categories: LWI (11.2 MMT, i.e., landfill and incineration) and recycled (33.6 MMT) without knowing the specific channels (i.e., recycled other). Additionally, no reliable data were found for the management breakdown of animal-based on-farm FLW (8.9 MMT; Supplementary Note 8), so they were designated to have unknown management. Finally, as stated above, some of the manufacturing FLW is water removed from product processing and is not disposed of via the FLW management pathways. As the FLW management analysis uses factors to assign values of FLW to the different disposal pathways, the manufacturing FLW used in this part of the analysis (144.8 MMT) excludes the evaporated water (63.7 MMT; see Supplementary Note 5).
As illustrated by Fig. 5, manufacturing ends up being the most efficient stage when considering FLW management; only 4.1 MMT (2.8%), are diverted to landfill and 0.5 MMT (0.3%) are incinerated. Most of the manufacturing FLW (109.6 MMT; 75.7%) is used for animal feeding, followed by land application (19.3 MMT; 13.3%).
The data sources used for the FLW management analysis do not distinguish between the distribution and W&R stages. As they have similar FLW drivers, this study uses the same factors for FLW management pathways for these two stages. Distribution generates 10.4 MMT of FLW and W&R generates 21.3 MMT. Of this, 45.1% (4.7 MMT and 9.6 MMT) is diverted from waste streams to some form of recycling and 54.9% (5.7 MMT and 11.7 MMT) is sent to landfill or incinerated. Most of the 49.4 MMT of consumption FLW (43.8 MMT, 88.7 %) is washed down the drain (and sent to wastewater treatment), landfilled, or incinerated, with very little (5.6 MMT, 11.3 %) recycled.
For the whole FSC, animal feeding (116.6 MMT) is the largest FLW management pathway, and manufacturing contributes the most to this pathway (109.6 MMT, 94% of FLW managed through animal feeding). The third and fourth preferred FLW management pathways (i.e., land application/composting and other industrial uses) only received 22.7 MMT and 14.4 MMT of FLW, respectively. Moreover, in 2016, 77.3 MMT of FLW from all five stages was diverted to LWI, which are the least preferred pathways. Unfortunately, despite being the most desirable FLW management pathway, only 8.1 MMT of FLW attempted to reenter the FSC for possible human consumption through food donation, and 0.4 MMT of this reentering amount was not consumed and ended up with other pathways.
This study conducts uncertainty analyses by disturbing each coefficient. Since this study estimated mass flows and FLW generation following the accounting approach adopted by Calderia et al.17, we adopted the same method for uncertainty analysis. Under the uniform distribution assumption, a Monte Carlo simulation yields the 95% confidence intervals of the amounts of FLW by FSC stages and food commodities, illustrated by Fig. 6. The figure shows that our estimations of FLW by food commodities and stages fall within the relatively small 95% confidence intervals.
To further demonstrate the impact of each coefficient, Fig. 7 shows the difference between the estimated FLW amount and the upper bound, the lower bound, and the mean value of the confidence intervals by FSC stages and food commodities. The FLW of each food commodity demonstrates a higher level of uncertainty at the distribution stage since the limited source of the waste factor at this stage. Moreover, the on-farm FLW of grains, dairy, animal products, fruits, and vegetables also demonstrate a high level of uncertainty, since the waste factors used for these food commodities are not the U.S. alone (see Supplementary Note 4). Finally, at the food manufacturing stage, grains, vegetables, fruit, and eggs also demonstrate a higher level of uncertainty since this study derived the factor for each type of food (e.g., dried vegetables), instead of using coefficient for each specific food product (e.g., dried potatoes).
Compared with several previous studies that focus on U.S. FLW quantification (listed in Table 1), this study has similar FLW estimations for later FSC stages (i.e., distribution, W&R, and consumption). However, other studies have either no or lower estimations for upstream stages. The detail is provided in Supplementary Note 9.