Part I. Systems-based operation parameters of FWTTs
FWTT database. Literature search and selection criteria: The primary source of data was the Web of ScienceÒ Core Collection. Search keywords in various combinations [“food loss” OR “food waste”, AND “anaerobic digestion” OR “composting” OR “feed” OR “landfill”, AND “life cycle assessment (LCA)” OR “life cycle inventory (LCI)”] were used for the initial search. Additional references were identified by cross searching the lists of literature cited in some of the primary publications. English language publications were used for the most part, plus publications in Chinese language. The initial search results are screened for inclusion with subsequent data extraction based on the following criteria: (i) the reports must be for real-life actual food waste treatment operations, i.e. not laboratory experiments or simulation studies; (ii) the treatment must be for food waste materials solely, i.e. no co-digestion or co-composting; (iii) the study was conducted with system boundaries as defined in Fig. 1, covering from food waste collection to transport to treatment and to the generation of product and byproduct; (iv) at lease one parameter on system operation or sustainability is reported.
Data extraction and inclusion: From studies selected, an array of data was extracted and entered into the database, including two data categories. The first is system operation metrics, including system input (energy, water), output (product, byproduct), and LCA-based assessment of global warming potential (GWP), acidification potential (AP), and eutrophication potential (EP), as reported in original studies. These parameters were normalized for the base unit of 1000 kg (1 metric ton) ton raw food waste material treated. The second data category includes information on treatment type (AC, AD, DF, WF, etc.), treatment capacity or size of operation, where (in what country) the study was performed and when (year of publication). Once selected for inclusion in the database, little filtering (e.g. outlier identification and exclusion) was applied to the original data as extracted from individual studies. The final FWTT database consisted of >470 data entries from 93 studies with global distribution (Fig. 2).
Data analysis. Descriptive statistics (mean, standard error of the mean) was performed using Excel 2013. Note that as case studies, the original reports had little or no information regarding variance of the data. Aggregated results were graphically displayed in Extended Data Fig. 1, for system input, some of the environmental effect, per base unit treated.
Due to the lack of repeated measures and variance estimates in the studies included, meta-analysis was conducted using bootstrapping procedure. The bootstrapped mean and 95% confidence interval were estimated for core parameters, including GWP and product/byproduct yields of AC, AD, or Re-Feed per base unit treated. These parameters are used in various analyses subsequently. With the bootstrapping procedure, all estimation was conducted using STATA 17MP, StataCorp LLC, College Station TX. Estimation of the means and 95% confidence interval for the core parameters was performed using a bootstrapped simulation with 1000 replicates for linear regression model. To permit for small departures from normality, a robust estimation of the variance was used. All marginal (model adjusted) means were reported with their respective 95% confidence interval (95% CI).
Part II. Global significance of FWTTs, recovery-treatment scenario analysis
FLW database. Since the landmark publication in 2011 on global food loss and waste11, many studies have emerged with additional information on food loss/waste parameters, providing a more granular and broader coverage. We thus constructed a FLW database with literature search, study screening and selection, and data extraction. Briefly, “food waste” OR “food loss” are used as the keyword to search the peer-reviewed literatures from Web of Science and China National Knowledge Infrastructure up to October 2021. To further ensure the relevance of selected publications, we screen out articles that contained data on FLW for (i) at least one food commodity (e. g. cereals, oilseeds, fruits and vegetables, meat, fish, milk), one food supply stage, and one region or country; (ii) provided explicit, or calculable, loss factors for a food commodity at some stage along the supply chain.
Data extracted from selected studies were organized for food loss and waste factors along the food supply chain in seven sector/subsectors: P1 for at farm harvest loss, P2 for postharvest handling and storage loss, P3 for manufacturing loss, P4 for distribution loss, P5 for retailing waste, P6a for consumer waste at home, and P6b for consumer waste out-of-home. The seven sector/subsectors are in line with FAO33. Information on the origin of the country or region for which the data were derived was also entered into our database. The final FLW database consisted of 1,135 data point entries from 117 studies.
From the FLW database, we derived food loss/waste parameters specific for each of the seven global-regions (Fig. 2). Aggregated data are presented in Extended Data Table 2.
Global-region FLW amount. Employing the calculation methods of Gustavsson et al. (2011), global-region FLW amounts P1 through P6 were calculated based on food availability data (FAO-FBS19; averaged 2017-2019) multiplied by FLW parameters obtained above (Extended Data Table 2). Results are summarized in Extended Data Table 3, and graphically illustrated in Extended Data Fig. 2.
Food waste recovery scenarios and treatment schemes. Food wasted at consumer-level (P6a, P6b) is scattered in countless homes and foodservice places, resembling non-point sources; food lost in sectors P2-P5 generally concentrate in fewer places, resembling point sources. We presumed that food waste recovery would be lower for the non-point sources than point sources. Excluding food recovery for P1 (in-field food loss to remain in the field), we designated food loss/waste recovery rates to be 60, 70, and 80% recovery for P2-P5 (post-harvest through retail/wholesale loss and waste); 35, 50, 65% for P6a (consumer at-home waste), and 50, 65, 80% for P6b (consumer out-of-home waste). Consequently, across-sector recovery scenarios would be 60-35-50% (low), 70-50-65% (medium), and 80-65-80% (high).
Two treatment schemes were considered: (i) 100% of food waste recovered to be treated via AC or AD or Re-Feed (the individual treatment scheme), (ii) recovered food waste to be treated via AC and AD and Re-Feed, 1/3 each (the combined scheme). The individual scheme would allow us to estimate the resource and climate mitigation capacities of AC, AD, or Re-Feed solely. The combined scheme would be more flexible and adaptive, as food waste is not created equal20; high quality food waste would be suited for feed-making whereas poor quality food waste might be best treated via AC or AD.
Food waste recoveries are then calculated by multiplying FLW amounts (Extended Data Table 3) with the rates under low, medium, and high scenarios. Subsequently, GWP reduction and resource recovery capacities were calculated, described below.
Global warming potential. For each of the seven global-regions, GWP for different FWTTs is calculated as:
where i represents AC, or AD, Re-Feed or LF; represents the GWP of a given FWTT; represents the parameters of GWP for different FWTTs (Fig. 2b); denotes the amount of food waste recovered under the low, medium, high scenarios with two treatment schemes. Per global-region results are in Extended Data Table 5; Aggregated results as global sums are presented in Table 1.
Resource recovery capacity of FWTTs. The amounts of FWTT products (compost, biogas and digestate, novel feeds) per global-regions are calculated as follows:
where i represents a given FWTT product/byproduct (compost, biogas and digestate, novel feed); represents yield of products (compost, digestate, biogas, novel feeds) as shown in Fig. 2c; and represents the amount of food waste recovered under the low, medium, high scenarios and treated via the two schemes.
Macro-nutrients of total N, P and K contained in the product (compost or digestate obtained above) and the energy equivalence of biogas are then calculated as below:
where i stands for type of product (compost or digestate); j for N or P or K; for the concentration of the nutrient in the given product (Extended Data Table 1); and for water content of the product; for heating value of biogas with 65% volume of methane of 22 MJ m-3 23. Results on nutrient and biogas recoveries per global-region are in Extended Data Table 4. Aggregated global sums are in Table 1.
Estimation of feed grain replacement. Maize and soybeans are major feed grains used in modern-day livestock diets. Crude protein and gross energy are critical nutritional attributes in feedstuffs pertaining to animal nutrition and feeding. Marta et al. (2018), analyzing feed samples generated from restaurant food waste via contemporary Re-Feed technologies, reported crude proteins averaging 24% and ether extract 7.94%34. Using the equation of Son and Kim (2017)35 (below),
we calculated the gross energy to be 4888 kcal kg-1 DM or 20.5 MJ kg-1 DM for the feed samples of Marta et al. These values (crude protein 24% DM basis, gross energy 20.45 MJ kg-1 DM) are comparable to those of global primary data36 on raw consumer food waste (crude protein averaging 19.7% DM and gross energy 20.2 MJ kg-1 DM). Therefore, we proceeded to calculate quantities of maize and soy that could be replaced with novel feeds on the basis of matching the crude protein vs. matching gross energy content, respectively, as below:
where i represents crude protein or gross energy; represents feed production coefficient (i.e. product yield of Re-Feed; 152 kg feed per base unit; Fig. 2); represents novel feed dry matter content; represents novel feed nutritional (crude protein or gross energy) content; and represent the substation amount of maize and soybean, respectively; and represent the dry matter content of maize and soybean; and represent the nutritional (crude protein or gross energy) content of maize and soybean (Extended Data Table 1); the average ratio of global maize and soybean feed consumption in 2017-2019 is 0.72 : 0.2819.
The grain replacement amounts under the low, medium, high food waste recovery scenarios with the individual vs. combined treatment scheme are summarized for global total, U.S., and China in Extended Data Table 6.
Part III. Cascading effect through avoidance analysis
The cascading impact (i.e. grain replacement dividends) from novel feeds replacing maize and soy was evaluated for U.S. and China. The two countries were selected mainly for the availability of data needed in various calculations.
Land, fertilizer, and herbicide spared. Yields of maize and soybeans along with application rates of fertilizers (N, P, K) and herbicides for U.S. and China were obtained from national statistics25,37 (Extended Data Table 7). Acreage of land and amounts of fertilizer and herbicides spared (no longer needed) due to novel feeds replacing maize and soy are calculated as below:
Where , , stand for total, maize and soy acreage, respectively; and for maize and soy yield for the U.S. (or China), respectively; for total fertilizer amounts (including N, P and K) spared; and for relevant application rate of the fertilizer (Extended Data Table 7); for total herbicides amounts spared; and for relevant application rates of herbicide (Extended Data Table 7).
Fuel and water spared. A sub-database of fuel and water consumption was constructed for U.S. and China maize and soybean production systems from literature data-mining. Peer-reviewed relevant publications were identified via Web of Science databases and China National Knowledge Infrastructure. Search key words included “maize” OR “corn” OR “soybean” AND “yield” AND “irrigation” OR “fuel consumption” AND “US” OR “China” in the abstract and key words. Selection criteria included: (i) the data must be measured in field experiment; experiments under rain-proof shelter conditions and model simulations were excluded; (ii) crop yield and water productivity were reported; (iii) fuel use or energy consumption for agricultural inputs; (iv) experimental sites were described. From 841 published studies we collected, normalized and aggregated water and fuel consumption parameters were derived (Supplementary Table 3). Then, amounts of fuel and water spared are calculated as below:
where denotes energy consumption now spared; and represent energy consumption factors for maize and soybean, respectively, Supplementary Table 3; and represent irrigation water use (i.e. excluding rain-fed maize or soy production) for maize and soybean, respectively. We estimated irrigation water use by dividing the crop yield of maize, and soybean by the rate of irrigation water productivity (IWP). A meta-analysis was performed to quantify IWP of maize, and soybean in China and U.S.
Nr loss avoided. Another sub-database was constructed to document Nr losses in U.S. and China maize and soybean production systems. Literature search was based on Web of Science databases and China National Knowledge Infrastructure for the period of 1990 to 2017. Search keywords included different combinations of “N2O” OR “NO3-” OR “NH3” AND “maize” OR “corn” OR “soybean” AND “China” OR “US”. Selection criteria were: (i) the N applied was in the form of urea or ammonium or nitrate; studies using slow release or controlled release fertilizer or organic materials such as manure or compost were excluded; (ii) At least one form of Nr loss (N2O emission, NO3- leaching, or NH3 volatilization) was determined and reported; the measurement of N loss must be performed during field operations and throughout the crop growing season, (iii) N2O emission was measured using closed static chamber, or acetylene inhibition methods; NH3 volatilization were measured by continuous air flow chamber method, venting method, Drager-Tube method, or microclimate and wind tunnel method; NO3- leaching was determined using suction cup and lysimeters or from soil samples, or soil water hydrological modeling. The final sub-database consisted of 930 observations from 186 published studies. Normalized and aggregated results are presented in Extended Data Table 7. Subsequently, avoided Nr losses are calculated as below:
where the , and represent avoided N2O emission, NH3 volatilization and NO3- leaching, respectively; and represent the N input to maize and soybean production; EFs represent the ratios of N2O emission, NH3 volatilization and NO3- leaching to total N input for maize and soybean production (see Extended Data Table 7).
GHG emissions avoided. The novel feeds replace the maize and soybean demand, thus the GHG emissions associated with agricultural inputs (N, P and K fertilizers, seeds, herbicide, insecticide and irrigation water), direct Nr losses from cropland plus fuel consumption can be avoided. We established a GHG emissions database related to U.S. and China maize and soybean production systems based on literature data-mining. Search key words included “maize” OR “corn” OR “soybean” AND “GHG emission” AND “US” OR “China” in the abstract and key words. Normalized and aggregated GHG emission parameters are shown in Supplementary Table 4 and the avoided GHG emission is calculated as below:
where stand for avoided GHG emissions; i for the feed grains (maize and soy); j for the seven agricultural inputs (above); for parameters of agricultural inputs (Extended Data Table 7); for spared acreage of maize and soybean; and for GHG emission parameters of the various agricultural inputs in maize and soy production systems.
33 Food and Agriculture Organization of the United Nations (FAO), 2014. Definitional Framework of Food Loss; FAO: Rome, Italy.
34 Marta, C. et al. Pet Food as the Most Concrete Strategy for Using Food Waste as Feedstuff within the European Context: A Feasibility Study. Sustainability-Basel 10, 2035, doi:10.3390/su10062035 (2018).
35 Son, A. R. & Kim, B. G. Prediction equations for gross energy in feed ingredients. FASEB Journal (2017).
36 Dou, Z. X., Toth, J. D. Global primary data on consumer food waste: Rate and characteristics – A review. Resources, Conservation and Recycling 168, 105332, doi:10.1016/j.resconrec.2020.105332 (2020).
37 USDA Census of Agriculture (United States Department of Agriculture, accessed 5 January 2022); https://www.nass.usda.gov/Publications/AgCensus/2017/Full_Report/Volume_1,_Chapter_1_US/usv1.pdf.
38 Lee, S. H., Choi, K. I., Osako, M. & Dong, J. I. Evaluation of environmental burdens caused by changes of food waste management systems in Seoul, Korea. The Science of the total environment 387, 42-53, doi:10.1016/j.scitotenv.2007.06.037 (2007).
39 Aye, L. & Widjaya, E. R. Environmental and economic analyses of waste disposal options for traditional markets in Indonesia. Waste management 26, 1180-1191, doi:10.1016/j.wasman.2005.09.010 (2006).
40 Blengini, G. A. Using LCA to evaluate impacts and resources conservation potential of composting: A case study of the Asti District in Italy. Resources, Conservation and Recycling 52, 1373-1381, doi:10.1016/j.resconrec.2008.08.002 (2008).
41 Takata, M. et al. The effects of recycling loops in food waste management in Japan: based on the environmental and economic evaluation of food recycling. The Science of the total environment 432, 309-317, doi:10.1016/j.scitotenv.2012.05.049 (2012).
42 Boldrin, A., Andersen, J.K., Møller, J., Favoino, E. & Christensen, T.H. Composting and compost utilization: accounting of greenhouse gases and global warming contributions. Waste Management & Research 27, 800-812, doi:10.1177/0734242X09345275” (2009).
43 Martinez-Blanco, J. et al. The use of life cycle assessment for the comparison of biowaste composting at home and full scale. Waste management 30, 983-994, doi:10.1016/j.wasman.2010.02.023 (2010).
44 Oldfield, T. L., White, E. & Holden, N. M. The implications of stakeholder perspective for LCA of wasted food and green waste. Journal of Cleaner Production 170, 1554-1564, doi:10.1016/j.jclepro.2017.09.239 (2018).
45 Jensen, M. B., Moller, J. & Scheutz, C. Assessment of a combined dry anaerobic digestion and post-composting treatment facility for source-separated organic household waste, using material and substance flow analysis and life cycle inventory. Waste management 66, 23-35, doi:10.1016/j.wasman.2017.03.029 (2017).
46 Colón, J. et al. Environmental assessment of home composting. Resources, Conservation and Recycling 54, 893-904, doi:10.1016/j.resconrec.2010.01.008 (2010).
47 Quirós, R. et al. Environmental assessment of two home composts with high and low gaseous emissions of the composting process. Resources, Conservation and Recycling 90, 9-20, doi:10.1016/j.resconrec.2014.05.008 (2014).
48 Ma, H. M., Gao, B., Han, W., Chen, Z. X., Li, B. Investigation and analysis on current status of soil utilization of kitchen waste anaerobic digestate. China Biogas 39, 27-34 (2021).
49 Righi, S., Oliviero, L., Pedrini, M., Buscaroli, A. & Della Casa, C. Life Cycle Assessment of management systems for sewage sludge and food waste: centralized and decentralized approaches. Journal of Cleaner Production 44, 8-17, doi:10.1016/j.jclepro.2012.12.004 (2013).
50 Heuzé V., T. G., Lebas F. Maize grain. Feedipedia, a programme by INRAE, CIRAD, AFZ and FAO. https://feedipedia.org/node/556 Last updated on September 7, 2017, 14:29. (2017).
51 Heuzé V., T. G. Soybean (general). Feedipedia, a programme by INRAE, CIRAD, AFZ and FAO. https://www.feedipedia.org/node/753 Last updated on August 3, 2016, 10:37. (2016).