The life cycle EEA of beef was performed using the BASF methodology (BASF SE, Ludwigshafen, Germany), which follows ISO 14040:2006a and 14044:2006a standards for LCAs and ISO 14045:2012 for EEA. The analysis was validated and verified by NSF International under Protocol P352 Parts A (BASF 2013a) and B (BASF 2015), respectively. Data sources included reported and IFSM-simulated cattle production data (Rotz et al. 2013), life cycle inventory databases, public databases, expert opinion, and primary processing and use data. Environmental impact was measured by the following metrics: consumption of non-renewable raw materials or abiotic depletion potential (ADP), cumulative energy demand (CED), consumptive water use (CWU), land use, and human toxicity potential (HTP) of materials used. Air, solid waste, and water emissions were also quantified. Air emissions included acidification potential (AP), global warming potential (GWP), ozone depletion potential (ODP), and photochemical ozone creation potential (POCP). Particulate matter was not included in the current study due to lack of industry data, complexity in characterization, and resulting unavailability of standard LCA procedures. Figure 1 shows a schematic of the main components of the life cycle impact analysis. Summaries of the data and sources are provided in related tables and the Supplementary Information.
Figure 1 shows a schematic of the system boundaries of the life cycle impact analysis. Cattle originating as cull animals from the dairy industry as well as beef imports and exports were not included in the production system. Feed imports were also excluded as all feed was produced within the USMARC production system. Of the total national major feed grains supply reported by the USDA-ERS (2017) which included use for other livestock species as well as food and industrial uses, less than 12% was imported at the time of this study. Capital, equipment, buildings, infrastructure, and materials for repairs and maintenance were excluded from the study as they are not usually included in LCAs of agricultural commodities, goods, and services due to their insignificant effects. Impacts that contributed < 1% individually as specific components or unit operations or < 3% in total as a group of components or processes to the overall value chain were generally excluded from the LCA. Among these were office and administrative impacts, employee commutes, cattle veterinary medicines, and retailer and consumer use of cleaning compounds.
Data sources and input information
Required inventory data included resource use and emissions for feed and cattle production and packing, case-ready, retail, restaurant, and consumer phases. Feed and cattle primary data were developed through reports of the USMARC and simulation of its production system (Rotz et al. 2013). This research facility provided high-quality and extensive management data that were difficult to obtain otherwise from the industry. The crop, feed, and animal management practices of this facility were characteristic of those practiced in the Great Plains (this region stretches north to south of the central USA and encompasses western Texas and eastern New Mexico through North Dakota and eastern Montana) where most cattle are produced. An exception to this similarity was greater irrigation use at the USMARC to grow feed crops and pasture compared to the rest of the industry. Another difference, albeit negligible when considering environmental impact, was a greater labor involvement in production at this federal research facility. Due to the region-specific nature of production practices across the USA resulting from differences in climate, available resources, and culture, the USMARC facility is not meant to represent all beef production systems in the nation. It does, though, provide a general representation of beef cattle production in the USA.
Biological and physical processes of the beef cattle production system were simulated with the IFSM, and the predicted performance of the operation was found to agree well with production records of the research center (Rotz et al. 2013). The full operation was simulated as three components: feed production, cow-calf production, and feedlot finishing. The accuracy of the IFSM predictions for feed production and use, energy use, and production costs (within 1% of reported values in 2011) justified its use for this study (Rotz et al. 2013). The USMARC and IFSM simulations primarily provided resource and farm input data as well as direct emissions in feed and cattle production from which life cycle inventories were derived.
The post-farm gate assessment consisting of the packing, case-ready, retail, restaurant, and consumer phases, used primary data from industry and information from public databases, literature review, and when necessary expert opinion. Various data were collected between 2011 and 2013. All feed, cattle, and packing phase data were representative of 2011 management practices, while primary retail and restaurant data were from 2013. Case-ready data from study partners were from 2011 and 2013. Specific details of the sources of input data are given in the relevant sections below. Impacts of resource use and emission outputs were quantified using life cycle inventories of inputs, processes, and outputs of each phase. The inventories were compiled from primary industry data, the BASF life cycle inventory database, Boustead Model Version 5.1.2600.2180 (Boustead 2005), and Ecoinvent Version 2.2 (Frischknecht et al. 2005).
All primary data obtained from packing plants, case-ready, and restaurants were obtained by providing each study partner with spreadsheets listing all inputs going into the system for which they provided the input values. The details of all primary data cannot be listed for confidentiality reasons. Neither can the extensive inventory of materials and resources from BASF’s database used in this study because this would be impractical and this also has propriety concerns. This paper’s intention is to provide sufficient detail and to adhere to quality standards required for dissemination of findings to stakeholders and the interested public. Therefore, the study underwent critical review by a third party, NSF International, and is reported as Protocol P352 Parts A (BASF 2013a) and B (BASF 2015). Through this peer review process, the study was verified and approved as valid.
The feed production phase studied the life cycle of feed produced and purchased for cattle consumption at the USMARC. Based on year 2011 records, the USMARC produced feed crops on 2108 ha of irrigated land and maintained 9713 ha of unirrigated pasture to feed the cattle produced. Feed crops included alfalfa/grass hay and silage, corn silage, high moisture corn, and dry corn grain. Feed purchased to supplement the farm’s production included 1790 t (dry matter) of wet distiller’s grain (WDGS). Crop growth, feed utilization, and nutrient cycling were simulated with IFSM using daily weather conditions on the crop farm during the study period (Rotz et al. 2013). Resource utilization, operation timeliness, crop losses, and feed quality were predicted based on reported tillage, planting, harvesting, and storage practices. Nutrient flows traced through the farm predicted environmental losses including reactive nitrogen (NH3, NO3−, and N2O), phosphorous, and net greenhouse gas (CO2, CH4, and N2O) emissions. Emission factors obtained from IFSM simulations directly related to crop production and additional field emissions derived using recommendations of the Intergovernmental Panel on Climate Change (IPCC 2006) are presented in Table 2. Cattle manure used to fertilize the farm and pastureland was generated within the USMARC and therefore had no associated pre-chain impacts. Rotz et al. (2013) provide a detailed description of the USMARC facility and IFSM simulations. Land use change impacts on greenhouse gas emissions can be substantial in areas where forests are transformed into farms for feed crop production. As cattle feed in the USA are not typically sourced from such areas, land use change was not considered as an important contributor to the GWP in this study.
Purchased corn for WDGS was ascribed Ecoinvent profiles (Table S1, Electronic Supplementary Material) representative of the Iowa corn-belt region which unlike the USMARC-cultivated corn was not irrigated. The average transportation roundtrip distances assumed for purchased corn to the distillery for WDGS production were 400 km and that of WDGS from the distillery to the USMARC was 32 km. As WDGS is a by-product of corn-based bioethanol distillation, an economic allocation was used to determine its impacts. This was done by deducting the drying energy from the life cycle inventory of 1 kg of dried distiller grains with solubles (DDGS) and multiplying the weight of the DDGS by 1.55 (Bonnardeaux 2007) to reflect the weight of WDGS. The bioethanol profile was adjusted to represent Iowa corn production yields, and the WDGS profile was created by allocating 21% of the distillation process and pre-chain impacts based on economic allocation using ethanol, WDGS, and DDGS pricing which were US$0.54/kg, US$0.09/kg, and US$0.24/kg at the time of the study.
The chemical oxygen demand (COD) for pesticides (Table S1, Electronic Supplementary Material) was calculated using their chemical formula (C, O, N, and H stoichiometry). Runoff and leaching emissions of heavy metals, including those from fertilizer application, were obtained using the Swiss Agricultural Life Cycle Assessment (SALCA) Heavy Metals calculator (Freiermuth 2006). As SALCA did not include options for selecting US specific soil characteristics, German values were substituted to simulate representative heavy metal dynamics (such as heavy metal percolation, deposition, and leaching rates) in the soil. This assumption was not seen to have a meaningful impact in the results as the representative soil type used was similar to US agricultural soils.
Irrigation was used to produce all feed crops and some pasture on the USMARC operation. Irrigation-associated impacts included CWU sourced from on-farm wells with a small amount (1%) from surface water. Electricity and natural gas were both used to pump water through center pivot irrigation systems, and pre-chain impacts for their production and use were included.
Cattle production, consisting of cow-calf and feedlot operations, followed the life cycle of cattle from birth to harvest. In 2011, USMARC maintained 5050 calves, 5498 cows, 285 bulls, and 1180 replacement heifers in the cow-calf operation and 3724 cattle were finished in the feedlot (Rotz et al. 2013). In the cow-calf operation, cattle were grazed on pasture (a small part of which received irrigation) and fed hay and silage during winter. Weaned calves were moved to a feedlot where they received a high-forage backgrounding diet of hay and distiller’s grain for 3 months and then were put on a high-grain finishing diet of corn grain, corn silage, and distiller’s grain for another 7 months. At 16 months of age, cattle were harvested with an average weight of 581 kg. Also included in the harvest were cull cows and bulls from the cow-calf phase of the operation (Rotz et al. 2013).
Supplementary feed was accounted for in this phase, whereas all grazed and harvested forage and grains were included in the feed production phase. Enteric CH4 and CH4, N2O, and NH3 emissions from excreted feces and urine and phosphorous and nitrogen runoff losses from pastureland were simulated with the cattle operations. Drinking water was supplied by USMARC wells, which accounted for the consumptive water used by the livestock. Energy used for pumping drinking water and its associated pre-chain impacts was also accounted for in this phase.
Transportation of calves and cows within USMARC operations was minimal with no effect on the LCA’s results. Generally, transportation impacts have been found to be relatively low in beef cattle systems in the USA even when cattle were transported over long distances (Rotz et al. 2015). The minor effect of cattle transportation to the packing plant was included in the packing phase.
Enteric CH4 emitted by cattle was the only form of biogenic carbon included in the analysis. The carbon in the enteric CH4 emitted was assumed to be assimilated from CO2 in the atmosphere during crop growth. To account for the assimilation and avoid double counting, a 1 CO2 eq credit was applied to the global warming potential (GWP) factor of CH4 (thus utilizing a GWP of 24 CO2 eq for methane as opposed to the standard factor of 25 CO2 eq). The major inputs for the cattle phase’s life cycle inventory are shown in Table S2 (Electronic Supplementary Material).
The packing phase of the value chain processes live animals into edible beef. Primary and calculated input data describing operational emissions and waste (Table S3, Electronic Supplementary Material) were obtained from site visits and interviews with three packers who collectively processed 60% of the US beef harvested annually. For equity in representation, the packers studied represented both small and large-scale operations and their weighted averages (based on weight of beef processed by each size of operation) were used. Primary transportation data for cattle, packaging (paper and plastics), liquid CO2, and wastes of the packing phase were used, and their average transport distance of 2033 km was assigned to all other raw materials and supplies. The by-products of beef (including hides, offal, blood, tallow, bones, and bone meal) received an economic allocation of 11.7% of the production and packing impacts based on primary sales data obtained from the collaborating packers. The COD emissions (Table S3, Electronic Supplementary Material) were calculated using the C, O, N, and H stoichiometry of the relevant compounds.
End-of-life impacts of 96% of each packaging material (plastic and corrugated cardboard packaging containing 30% recycled fiber) were assigned to the case-ready or retail phase as these materials entered either phase directly from the packing phase whose waste profile received the remaining 4% of the packaging plastics’ end-of-life impacts. The remaining 4% of corrugated cardboard was recycled and had no further impacts attributed to it following cutoff allocation rules.
The case-ready phase further processes primal cuts of beef into consumer-ready packaged cuts. Based on data from the case-ready study partners, 63% of US beef was assumed to be packaged in the case-ready phase. Primary data were obtained from a packer with a case-ready operation and a stand-alone case-ready operation (Table S3, Electronic Supplementary Material). Input data were also received from a collaborating packer with case-ready operations where energy input, packaging, consumable items, and waste were directly reported. For water use and cleaning chemicals, input values were assumed to be 10% of the reported use of the packing facilities based on knowledge of this facility’s operators and industry experts. End-of-life impacts of 96.5% of the packaging material were attributed to the retail or consumer phases. The study boundary did not include the 3.5% of corrugated cardboard that was recycled (following cutoff allocation rules).
The retail phase represented distribution and marketing of beef to the consumer. Primary data (Table S4, Electronic Supplementary Material) were obtained from three retailers ranging in size and representing 8% of the total retailed US beef. As the retail operations included sales other than beef, an economic allocation was performed based on the ratio of beef to total store sales. For refrigerated sales, allocations of refrigerant leakage and electricity use were refined using averages of ICF (2005), USEPA (2011, 2012), and FMI (2012) data.
The consumer phase included consumer impacts from transportation to the retail store and beef consumption at home. Transportation considered the ratio of beef to total supermarket purchases, and cooking was based on the average beef cooking preference of the consumer in relation to the average energy required to cook 1 kg of consumed beef while considering the percentages of electric or gas stoves. Public information and data gathered from literature were used to calculate national averages of consumer-phase inventory inputs (Table S4, Electronic Supplementary Material) related to consumer repackaging of beef, transportation (USDOT 2011), refrigeration electrical energy consumption (AHAM 2011), cooking energy (USEIA 2005), and beef waste (USDA-ERS 2012b). Volumetric allocation based on the typical US consumer’s diet (considered to consist of 12.7% meat by volume) followed by the application of economic purchasing factors to derive the beef portion of meat consumed was used to allocate impacts related to consumer beef refrigeration (USDA-ERS 2005).
The restaurant phase studied impacts of beef sold to consumers in restaurants and used primary data from quick-service and casual sit-down restaurants representing nearly 6% of US beef sold in restaurants. As restaurants typically sell other consumables in addition to beef, an economic allocation factor calculated as the ratio of beef sold to total restaurant sales was used. It was also assumed that 53 and 47% of US beef consumption were in restaurants and at home, respectively, based on industry data (Meat Solutions, Inc. 2014). System inputs in the restaurant phase are listed in Table S5 (Electronic Supplementary Material).
Recycling and waste
Packaging of beef was assumed to be done either in the case-ready phase (for sale at retail) or by the retailer. Footprints of the 30% recycled fiber in the corrugated cardboard used by most of the post-farm facilities surveyed were obtained from pre-defined profiles in Ecoinvent which also took into consideration the item’s production processes and appropriate allocations. Of post-farm packaging other than cardboard, disposal was assumed to be 82 and 18% in landfills and incinerators, respectively, with energy recovered from that incinerated (USEPA 2010). Additionally, a modified Ecoinvent profile was given to waste packaging disposed in municipal and solid waste landfills to remove accounting for contaminants such as heavy metals that were originally part of the municipal waste Ecoinvent profile but not found in beef packaging waste. While all direct waste generated at each phase of the value chain was evaluated based on its final fate and degradation after emissions to water and air, solid waste associated with the production of resource inputs was assessed based on its final disposal (i.e., recycling, incineration, or landfilling). Finally, the cutoff method was applied for impacts from incineration, as the impacts from incineration were assumed to be attributed to the energy consumer (i.e., purchaser of the electricity generated from incineration).
Environmental impact metrics
Abiotic depletion potential (kg Ag eq/CB) measured the effects of raw material use on the availability of natural reserves. As described by Uhlman and Saling (2010), the mass of basic raw materials from each resource needed to manufacture a product was weighted by a factor consisting of the quantity of each material’s geologic reserves and life span as defined by the USGS (Guinée et al. 2002) given current global extraction rates for all uses. Thus, materials with lower reserves and/or higher consumption rates received higher weightings. Sustainably managed renewable resources had a weighting factor of “0” to indicate infinite reserves. Information on demand and available reserves were obtained from national mineral commodity statistics (USGS 2012) and fossil energy reserve data (BP 2012). Table S6 (Electronic Supplementary Material) provides a list of essential raw materials considered, their global reserves, and assigned weighting factors.
Cumulative energy demand (MJ/CB) was the sum of all energy needed for the production, use, and disposal of a product as well as the energy content of the product. All individual energy sources (e.g., biomass, coal, lignite, natural gas, nuclear, oil, and wind) measured in MJ/CB were summed to obtain the CED per CB. As electricity companies use an assortment of fuels to generate electricity, a life cycle inventory was assembled for each type contributing to the national grid. The contribution of each fuel type to the national grid was based on the 2011 US Energy Information Administration’s data on electricity generation by energy source. Losses during the conversion of electrical energy to steam and line losses of electricity were included. The gross bioenergy content of feedstock (energy released through combustion of feed biomass) was included but not the solar energy required for its production. The gross bioenergy for corn silage, corn grain, alfalfa silage and hay, and grass from managed and natural pastures ranged from 17.8 to 18.6 MJ/kg dry matter (Ecoinvent Version 2.2; Frischknecht et al. 2005). No weighting factors were applied to the CED.
Consumptive water use (L eq/CB) quantified withdrawn freshwater lost from the watershed of origin in a product’s life cycle via evaporation, absorption into products and waste, or transfer out of the watershed (Pfister et al. 2009). Water stress indices (WSIs) obtained from Pfister et al. (2009) and served as midpoint characterization factors applied to the volume of absolute CWU (L abs/CB) to obtain the CWU (L eq/CB). Coefficients defining absolute CWU represented the consumptive fraction of the water used in a given process and were taken as midpoint ranges from USGS consumptive water data (Solley et al. 1998). These included cropping (70%), livestock (55%), industrial (25%), and thermoelectric power (50%). For example, in crop production, 100 L of water used for irrigation was 70 L absolute consumed (70% loss to evapotranspiration and runoff from the watershed) multiplied by the water stress index of 0.499 giving an equivalent of 35 L eq/CB.
Human toxicity potential (dimensionless) quantified possible toxic effects of material exposure on human health pre-chain, along the value chain, and material disposals within study boundaries (Landsiedel and Saling 2002). The production, use, and disposal of all materials relevant to the beef value chain were inventoried and assigned scores between 0 and 1000 based on hazard statements (H-phrases) from safety data sheets according to the Globally Harmonized System of Classification and Labelling of Chemicals (Table S7, Electronic Supplementary Material; adapted from Landsiedel and Saling 2002). Based on expert opinion, the scores were further modified, normalized, and weighted considering exposure conditions, substance’s persistence, and exposure risk as described in BASF (2013a). The sum of the products of the quantities of each substance used and their calculated scores was found. Finally, materials were assigned weightings of 20, 70, and 10%, based on the possibility of exposure at the production, use, and disposal phases, respectively. Pre-chain H-phrases were considered in the production phase of resource inputs only. The heaviest weighting of 70% was assigned to “use” as it is at this stage that the highest risk of personal exposure was expected. When a substance’s use occurred as part of its production and no disposal occurred because it was used up with the value chain, an integration of the weighting factors yielded a value “100”; thus, weighting was irrelevant.
Land use (m2a eq/CB, where a = time in years) was considered in terms of both occupation and change (transformation). A series of occupations taking place at different time periods made up the transformation. Land use was calculated as the total area of land and the degree of development required to provide the CB. The degree of land development was based on the ecosystem damage potential (EDP) which was assigned based on the level of occupation or transformation and provides indicators of biodiversity as estimated from the species richness of vascular plants (Koellner and Scholz 2008). Categories of land use occupations and transformations from one use type to another were described by the EDP (Table S8, Electronic Supplementary Material; Frischknecht et al. 2005) following an established BASF methodology. The land use classes included native vegetation, arable, permanent crop, pasture and meadow, urban, industrial, mineral extraction site, traffic areas, dump sites, and water areas.
Air emissions had four main categories (Table S9, Electronic Supplementary Material). Acidification potential (kg SO2 eq) included SOx, NOx, NH3, and HCl emissions (Saling et al. 2002). Global warming potential (kg CO2 eq) included anthropogenic CO2, CH4, N2O, and halocarbons (HC), with each gas adjusted by their 100-year GWP (Forster et al. 2007; IPCC 2006). Ozone depletion potential of HC was reported as kg CFC eq. Photochemical ozone creation potential considered those emissions responsible for ground-level ozone, including non-methane volatile organic compounds (NM-VOC) and CH4 measured in kg C2H2 eq (Heijungs et al. 1992).
Solid waste impact (kg municipal waste eq/CB) considered materials disposed in a landfill or incinerated. These materials were placed in five categories based on their potential environmental effects. The categories were municipal waste, hazardous waste (as defined by the Resource Conservation and Recovery Act), construction waste (non-hazardous waste materials generated during building or demolition), mining (non-hazardous earth or overburden generated during raw materials extraction), and radioactive waste (as defined by the International Atomic Energy Agency). Existing life cycle inventories provided waste information for the production of resource inputs, while the inventory for the value chain was developed from the primary waste profile data provided by industry partners. As there were no standardized assessment criteria at the time of this study, individual impacts were weighted by the normalized average disposal cost of each waste category in a landfill compiled internally by BASF (Table S10, Electronic Supplementary Material) and then summed to an overall solid waste impact. Any special waste categories from mining raw material inputs were treated according to the specific category’s requirements, while the non-hazardous waste (mainly earth) stayed on site for fillings and therefore were assigned a disposal cost of zero (Table S10, Electronic Supplementary Material).
Water emission (L diluted water eq/CB) was based on critical volumes defined as the contaminant concentration multiplied by a dilution factor (the reciprocal of a regional regulatory maximum emission concentration) for each contaminant (Table S11, Electronic Supplementary Material). Total volumes of water discharged from wastewater treatment plants and directly to surface waters were considered. The contaminants included in the BASF EEA method were NH4+, total-N, and PO43−, as well as heavy metals, hydrocarbons (including detergents and oils), and SO42−. Also included were Cl−, adsorbable organically bound halogens, and COD. In order to determine total water emissions, the sum of all critical volumes was found and then normalized to aid aggregation of all parameters into a single value (Saling et al. 2002).
Qualitative uncertainty and sensitivity analyses
All data sources used for this study were ranked as high (primary data) to medium (literature review and industry average) quality. The feed, cow-calf, and finishing phases were analyzed with primary data obtained from USMARC and IFSM simulations and were ranked as high to medium quality. Although observed farm environmental data were unavailable with which to make model comparisons of impact metrics, proper modeling of production systems in IFSM has been shown in previous studies to produce accurate predictions of emissions (Rotz et al. 2006, 2010; Stackhouse-Lawson et al. 2012). The IFSM simulations of feed production over local weather conditions, energy use, and production costs fell within 1% of reported values (Rotz et al. 2013).
Both the packing and case-ready phase data were ranked as high quality. The retail and restaurant data were primary; however, economic allocations were done resulting in a high to medium-quality classification. Data for the consumer phase was described as medium quality having been taken from literature and industry reports. A review of the system inputs showed data to be complete and representative of current industry practices; thus, no critical uncertainties were identified so as to impact the study’s results and conclusions.
Sensitivity analyses were done to account for specific integrated processes along the value chain. Three alternative scenarios were studied independently and compared with the base analysis. For two scenarios, analysis of wet distiller’s grains by mass allocation (scenario 1) and energy allocation (scenario 2) was compared to the economic allocation used in the base analysis. In a third scenario, analysis of consumer refrigeration by economic allocation (scenario 3) was compared to the volumetric allocation of the base analysis.