A conceptual framework for modelling the role of livestock systems in sustainable diets and a sustainable planet

Modelling economic sustainability

Economic assessment of sustainability is the process of applying accounting frameworks to biophysical systems. This domain considers financial metrics such as profitability, efficiency, productivity, as well as diversification, risk, and durability (Lebacq et al. 2013). Many farm/system financial calculators exist, as do input–output models, although they vary in scope and metrics used. An example of this is the Model of an Integrated Dryland Agricultural System (MIDAS) model (Kingwell and Pannell 1987; Young et al. 2014), a whole-farm model used for modelling of economic sustainability in Australian farms, including the impacts of climate change on whole-farm performance (Thamo et al. 2017). The Australian farm-optimisation model (Young and Young 2022; Young 2023) builds on the MIDAS model by including environmental impacts from on-farm activities, allowing for economic and environmental sustainability assessment. Some other economic models also consider environmental sustainability, but at a more limited scope than economic sustainability, such as in the BIO model (Grealis and O’Donoghue 2015; Tsakiridis et al. 2020). The BIO model uses an input–output approach for assessment of economic and environmental impacts, but uses carbon footprint as its sole indicator of environmental sustainability, and does not analyse the impact of aquaculture or wild-caught fish on biodiversity or other marine ecosystem indicators. One key limitation of this approach is that economic input–output approaches work for sectors that produce homogenous goods, but are not appropriate for comparing different products within the same sector (Tsakiridis et al. 2020), such as differences in product quality among breeds of livestock.

Modelling environmental sustainability

Assessing environmental sustainability can be thought of as applying accounting methods to environmental impacts instead of financial impacts. Input–output analysis has been used to compare and analysis systems with regard to metrics such as greenhouse-gas (GHG) emissions (Hendrie et al. 2016) and water scarcity (Ridoutt et al. 2018). To be successful, this requires that input–output data are available at the desired level and resolution. Life-cycle assessment (LCA) is a common approach that examines the environmental impacts of a product by mapping out its life cycle (creation, processing, use, disposal, recycling) and tracing the flows of materials in and out of the product system, as well as the environmental impacts associated with these flows and stages (de Vries and de Boer 2010). Life-cycle assessments are conducted according to a set of international standards (ISO 2006a, 2006b, 2018a). Building on the ISO framework, the LEAP framework of FAO is a set of best practices for LCA of livestock production (FAO 2016a, 2016b, 2016c).

Life-cycle assessment can be combined with economic frameworks to create combined economic and environmental input–output LCA, or can be process-based, using life-cycle inventory data and biophysical modelling to represent all the flows involved in the system under study (Tsakiridis et al. 2020). Many LCA studies use a software designed for such a purpose, such as SimaPro (PRé Sustainability 2023), OpenLCA (GreenDelta 2023), and GaBi (Sphera 2023). Some industries have their own industry-specific software, such as the Pig Production Environmental Footprint Calculator (National Pork Board 2017) or the Vit’LCA program for viticulture (Renouf and Renaud-Gentié 2023). However, other assessments use existing biophysical models that simulate the metabolism and production of the species under consideration. The environmental impacts can then be calculated from within the biophysical model itself, or the outputs of the biophysical model can then be combined with those obtained using other software. One advantage of this approach is the flexibility it provides, allowing for clear consideration of the animal-level impacts of a wide range of management strategies as long as sufficient data are available (Rotz 2018; Dougherty et al. 2019a, 2019b; Van Amburgh et al. 2019; Thomas et al. 2021).

Sustainability assessment should begin with clearly defining the question being explored, and choosing metrics that directly address this question (McClelland et al. 2018). Metrics must also be identified and compared at the right level of aggregation. For example, GHG emissions have global impacts. However, impacts such as eutrophication potential, biodiversity impacts, and land/water scarcity have impacts at a more regional scale and have inherently regional boundaries (Picasso et al. 2014; Ridoutt et al. 2018; Lemke 2021). Carbon footprint, one of the most popular metrics, relates the environmental impacts of GHG relative to that of carbon dioxide, allowing for reporting as a single metric (Forster et al. 2021). This metric has become a default metric of reporting environmental impacts and sustainability, although such a myopic focus on carbon at the expense of examining other metrics has the potential to lead to reduction in GHG at the expense of other sustainability dimensions such as water use, nutrient cycling, and land use (Harrison et al. 2021).

Metrics such as resource use must be also reported in the context of local scarcity and/or sufficiency. ISO (2014) standards mandate that water footprints cannot solely consider total water usage; they must also contain region-specific assessments of the impacts of water use in terms of specific metrics such as water scarcity and/or eutrophication potential. Such weighting is vital to both the interpretation of results as well as the implementation of water-saving strategies. For example, the use of irrigation for crops will have different impacts depending on the water scarcity of the region they are produced in, leading to differences in recommended cropping strategies and mitigation methods (Ridoutt et al. 2018). Similar frameworks apply to the weighting and reporting of land usage, weighting by its contribution to relative cropland scarcity and the environmental impacts of other land uses.

Many of these principles also apply to modelling at the dietary/human food-supply level. Multiple different indices have been developed to rank the nutrient quality of foods and diets, and combining nutritional quality with LCA, i.e. ‘n-LCA’ (Green et al. 2020, 2021), requires clear definition of both the functional unit of the assessment and the criteria on which decisions are being made, particularly when GHG emissions are expressed relative to units of a specific nutrient. Because the objective of many health eating/dietary health-scoring systems (Trichopoulou et al. 2003; Shams-White et al. 2023) is primarily to identify patterns of dietary intake associated with mortality risk, chronic disease prevention or chronic disease outcome management, rather than nutrient adequacy or direct measures of sustainability, their scope of healthy diet patterns still may not accommodate other dimensions of health and sustainability, depending on personal/cultural requirements or food accessibility. For example, the planetary health-diet index uses GHG emissions as its sole metric of sustainability (Cacau et al. 2021), while the sustainable-healthy-diet index utilises a questionnaire to measure sustainable dietary consumption, including questions about use of local food sources, reducing food waste, bottled water use, recycling, and composting, with the goal of capturing sociocultural aspects of sustainability (Tepper et al. 2021). Additionally, few nutritional indices or studies examine differences in nutrient bioavailability among foods, which also affects results (Ridoutt 2021b). Similarly, Cifelli et al. (2022) examined the impacts of replacing nutrients in dairy with other foods and found that it was possible to do so while optimising for either cost, energy intake, or quantity of food required, but that improvements in one metric led to significant increases in the other metrics. Recommendations from nutritional LCAs are dependent on the choice of a functional unit and a nutritional scoring system (Ridoutt 2021a), and assessing the sustainability of foods without considering their nutritional quality and contribution to meeting human requirements will inherently bias results.

Modelling social sustainability

There are fewer frameworks for assessing social sustainability than economic or environmental sustainability (Lemke 2021). The United Nations (UNEP/SETAC 2009, 2020) developed a framework for applying LCA to social sustainability that is compliant with ISO standards. This framework currently serves as the primary best-practice framework for modelling and assessing social sustainability across multiple sectors. However, application of this framework to food production is still an emerging area; a recent review of social sustainability assessment in the agrifood sector (Tragnone et al. 2022) found that social LCA was hindered by a lack of clarity about boundaries and functional units, poor data availability, a lack of transparent methodological choices, and poor inclusion of stakeholders in definition of the social themes being assessed. Similarly, de Olde et al. (2018) reviewed trade-offs between comprehensiveness and implementation in social sustainability-assessment frameworks and tools, and recommended harmonising metrics as much as possible with current best practices, while still ensuring that they are flexible enough to capture differences among systems and provide actionable information. An example of this is the work of Rivera-Huerta et al. (2019), which synthesised indicators from multiple sources to develop a comprehensive set of metrics that were then used to assess the social sustainability of livestock production in Mexico. Aranda et al. (2021) applied a similar holistic social sustainability-assessment framework to pork production in Spain. These studies are examples of adapting the more general UNEP/SETAC framework to the particular needs and concerns of the livestock sector, and the methods of these studies could be used to further inform and develop social sustainability frameworks specifically for the livestock sector.

Increasing the prevalence of multi-metric, multi-scale assessments

Despite the recognised need for multi-metric, multi-scale assessments, many studies consider a limited number of sustainability metrics, primarily environmental (de Vries and de Boer 2010; Gerber et al. 2014; Picasso et al. 2014; Ridoutt et al. 2017; McClelland et al. 2018; Harrison et al. 2021). A review by McClelland et al. (2018) found that 98% of LCA studies of livestock production included climate change as an impact category, and 28% of studies looked only at that category. Such a narrow focus is a key limitation in how sustainability is assessed and reported.

In contrast, multi-metric studies are vital for helping ensure that progress in one area does not come at the cost of worse performance in another, and correlations among different metrics vary depending on the scale being assessed. For example, there is a poor correlation between carbon and water footprints when comparing different systems producing the same kind of livestock (Dougherty et al. 2019b; Klopatek et al. 2022). However, intensification of livestock production by moving away from grazing-based systems lowers carbon footprint at the cost of higher fuel and energy use, greater cropland and mining land use, and increased risks of eutrophication owing to fertiliser use (Picasso et al. 2014; Dougherty et al. 2019b; Klopatek et al. 2022). When different foods are compared, Ridoutt and Navarro Garcia (2020) found that plant-based foods such as cocoa, coffee, tea, nuts, and oilseeds contributed more to cropland scarcity than did ruminant products, particularly when assessed in terms of biomass productivity, malnutrition risk, and biodiversity. Similarly, Ridoutt et al. (2018, 2019, 2021a) found that optimising diet patterns to reduce climate impact led to an increase in water-scarcity impacts, and that plant-based foods had a higher water-scarcity footprint than did livestock products, either when assessed on a commodity or whole-diet basis. Due to correlations between metrics and the complexity of production systems, it may be impossible in some circumstances to find a solution that is optimal across all metrics and scales. Therefore, multi-metric studies are vital to ensure that progress in one area does not come with unexpected consequences down the road.

Incorporating planetary health perspectives into models of livestock systems

Sustainable food production and systems require a shift in thought from ‘what is sustainable for humans?’ to ‘what is sustainable for the planet?’ One example of this is the provision of ecosystem services in grasslands. Rangelands in areas such as North America, Central Asia, and Africa evolved under grazing pressure from large herbivores, and such gazing pressure has been a key ecological feature of these areas (Price et al. 2022; Thompson et al. 2023). The specific effects of grazing on biodiversity and ecosystem health are highly variable by site, and depend on the ecosystem’s evolutionary history with grazing pressure (Price et al. 2022), but where appropriate, carefully managed grazing can be used to provide a range of ecosystem services, such as the following:

Two-thirds of the world’s agricultural land is either permanent pastures, meadows, or land that is too marginal to grow row crops on because of limitations of climate, terrain, soil characteristics, or other factors (Mottet et al. 2017; FAO 2020a). While some of this land has historically been converted to cropland, it comes with a reduction in soil quality and carbon sequestration potential (Zhang et al. 2021) and such conversion is a leading cause of habitat and biodiversity loss (McClelland et al. 2018; Lark et al. 2020). The IPCC (2022) noted that preventing degradation of existing rangelands and restoring previously converted land are key pathways to reducing emissions and sequestering carbon, and that animal management is a key tool towards this goal.

While the incorporation of ecosystem services into sustainability assessment frameworks is an emerging area, there are existing methods for modelling soil organic carbon and biodiversity that could be incorporated into this framework, as reviewed by Thomas et al. (2023). Fewer frameworks are currently available to address the issue of avoided emissions, largely due to a lack of methods to quantify the emissions reductions from novel management practices. The use of such frameworks and/or models would be dependent on sufficient data to robustly parameterise, independently validate, and implement these methods. However, among-site and within-site variation may be high enough that such methods may be feasible only for more generalised or regional assessments rather than site-specific analyses. Further exploration of data and methods available could be used to determine which of the above frameworks would be most feasible to implement and/or most useful to stakeholders and would aid in the inclusion of ecosystem services and their planet-level/ecosystem-level benefits into holistic modelling of livestock production.

Representing economic and social multifunctionality in livestock production systems

Livestock play multiple roles in human society and can contribute positively to economic and social sustainability. In addition to being a ‘battery of nutrients’ (Wyngaarden et al. 2020), livestock can serve as a ‘living bank’, being both a source of income but also a form of risk management against future economic insecurity, particularly in low- or middle-income countries (LMIC) and in lower-income households (Herrero et al. 2013; Alders et al. 2021a, 2021b). Livestock also have a strong synergy with cropping systems; manure accounts for half of world nitrogen fertiliser use and one-third of nitrogen fertiliser applied to crops (Gao and Cabrera Serrenho 2023). This latter point is especially the key, given that ammonia production for use as synthetic N fertiliser uses 1–2% of the world’s total energy supply, primarily from fossil fuels, and 3–5% of the world’s annual natural gas production, in addition to other environmental effects (Ghavam et al. 2021).

Livestock have historically been a key source of draught labour, a role that continues to this day in LMIC and rural areas worldwide; although understudied, this role is key to successful farming in many areas, and, depending on the region the value of draught services and manure, may exceed that of milk and meat, although the value of draught power is not often incorporated in sustainability assessments (Mota-Rojas et al. 2021). The rearing and sale of livestock and their products are a key pathway to food and financial security for women, (Herrero et al. 2013; Alders et al. 2021b), and greater support for women livestock producers, either individually or through the formation and promotion of women-led agricultural cooperatives (Oliver 2016; Lecoutere 2017), has been identified as a key pathway to resilience and sustainability by groups such as the FAO (2020b) and ILRI (2023). Incorporating these aspects into a social sustainability-assessment framework would help better reflect the varied reasons people keep livestock, and their important role in economic and social response to changes and longer-term sustainability.

Better incorporation of role of consumer choice and human diversity in sustainable diets

Discussion around sustainable diets, whether in scientific literature or the broader media, has been primarily framed in terms of environmental impacts. However, this is too narrow a perspective; diets must be evaluated from a triple bottom-line framework, while also accounting for variation in human needs and consumer choice. Many nutritional studies define optimal diets primarily in terms of meeting specific nutrient requirements, omitting the effects of regional and cultural differences in what foods are consumed (Steenson and Buttriss 2020; Ridoutt 2021a, 2021b; Biesbroek et al. 2023). Similarly, medical conditions such as food allergies, chronic kidney disease, or celiac disease require specialised diets, leading to differences in recommended nutrient intakes, higher than average food costs (Leiman et al. 2022; Scurlock et al. 2022), and therefore different optimal recommendations when evaluated from a triple bottom-line perspective. The lack of regional variation in dietary sustainability studies is especially concerning for LMIC (Reider et al. 2020; Tam et al. 2020), which have a high prevalence of nutrient-poor diets, affecting growth and development, and increasing risk of malnutrition and iron-deficiency anaemia (Keats et al. 2018; Khan et al. 2022). Therefore, certain areas, such as many LMIC, may benefit from higher proportions of animal foods, as rich sources of vitamin B12, iron, calcium, and zinc, to help combat prevalent nutrient deficiencies (Beal et al. 2023), contrary to current widespread recommendations of reducing meat and other animal-product consumption. Analyses must also consider regional variation in prices and availability of different foods, as well as substitution of one food for another on the basis of differences in cost and/or consumer income. All of these issues must be factored into assessments of sustainability at the dietary level, or risk excluding large sectors of the population; however, literature around sustainable diets does not often adequately consider these factors, and they have yet to be incorporated into many modelling frameworks.

Proposed dietary interventions must also be realistic in terms of reflecting human behaviour and consumption patterns. For example, US dietary guidelines recommend that ≤15% of calories come from discretionary foods such as sweetened beverages, candies, and alcohol; however, actual intake is often higher. Although reducing discretionary intake could significantly reduce environmental impacts, consumption trends suggest limited reduction feasibility and therefore such discretionary intake must be factored into modelled diets if they are to be realistic representations of how people eat (Hendrie et al. 2016; US DHHS and USDA 2020; Ridoutt et al. 2021b). Similarly, when people eliminate one type of food from their diet, they may not adequately replace lost nutrient sources. For example, individuals that exclude dairy products from their diet, whether due to personal preference or medical necessity, may replace dairy intake with dairy alternatives that vary in their ability to replace the nutrients previously present in dairy, particularly vitamin D and calcium (Ridoutt et al. 2020; Clegg et al. 2021). This is of particular concern for human health, as people in many countries already struggle to meet calcium and vitamin D intake recommendations, increasing the risk of negative health outcomes (Shlisky et al. 2022; Cui et al. 2023).

Multi-metric scoring systems have great potential to help explore the trade-offs among different diets from a net impact perspective, as long as the diets modelled are realistic in terms of variation in human choices and resources available. For example, Ridoutt et al. (2021a) found that optimising diet patterns to reduce climate impact led to an increase in water-scarcity impacts. Similarly, Cifelli et al. (2022) examined the impacts of replacing nutrients in dairy with other foods and found that it was possible to do so while optimising for either cost, energy intake, or quantity of food required, but that improvements in one metric led to significant increases in the other metrics. Recommendations from nutritional LCAs are dependent on the choice of functional unit and nutritional scoring system (Ridoutt 2021a) and assessing the sustainability of foods without considering their nutritional quality and contribution to meeting human requirements will inherently bias results.

Worldwide, animal-sourced foods provide 39% of per capita protein intake (FAOSTAT 2023) and are a key source of bioavailable nutrients such as iron, vitamin D, folic acid, and essential fatty acids (White and Hall 2017; Adesogan et al. 2020). As such, animal-sourced foods can be a key source of nutrients to prevent deficiencies in at-risk populations. It is therefore important to consider the role and efficacy of animal-derived nutrients in ensuring accessibility to nutrient-dense foods, especially where micronutrient supplementation and fortification programs continue to face barriers of access or coverage to healthcare services, adequate education, cultural acceptance and community engagement, and program sustainability (Zavaleta 2017; Stelle et al. 2019), and draws attention to the risk of ‘one size fits all’ dietary patterns exacerbating existing marginalisation, an issue that must be more addressed by sustainability science going forward. Incorporating regional and cultural variation in dietary patterns and substitution would enhance the social sustainability domain of a triple bottom-line assessment, and prevent the ‘one size fits all’ myopia seen elsewhere in the literature. This could then be a part of a multi-metric assessment framework that considered the impacts across nutritional, economic, and environmental scales.

The need for representation of the impacts of livestock consumption of crop by-products and food waste on triple bottom-line sustainability and the impact on human food supply

Livestock also play a vital role in upcycling human-inedible crop by-products into food, fuel, and fibre, adding nutrients contained in these products back into the human food chain and reducing waste. However, the net impacts of such circularity among different sectors of food production are not always captured in biophysical models and sustainability frameworks. Modelling these effects at the food production-system scale is key to exploration and assessment of the broader impacts of food production. The majority of biomass produced from crop production is not human-edible; every kilogram of common human-edible crops grow yields of roughly 3 kg of human-inedible biomass in the form of crop residues and process by-products such as canola meal, beet pulp, and straw. These by-products account for one-third of the mass of feed fed to livestock worldwide (Mottet et al. 2017, 2018; Thompson et al. 2023). At the farm scale, nutrients remain trapped in post-harvest residues such as corn stover and wheat stubble, hindering their use as soil amendments, although the stubble can still serve as an important nutrient source to later crops (Stavi et al. 2016; Li et al. 2021; Thompson et al. 2023). This residue can also be grazed by livestock, returning nutrients to the soil in a more bioavailable from as manure, increasing later crop yields (Hunt et al. 2016; Stavi et al. 2016; Li et al. 2021), and reducing net farm-level losses of carbon, nitrogen, sulfur and potassium from the farm when compared with removal of stubble via burning (Agriculture Victoria 2023).

Worldwide, 10–13% of feed fed to livestock across all species comes from the human-inedible by-products of crops grown primarily for human consumption, such as oilseed cake, beet pulp, spent grains, and distiller’s grains from ethanol production (Mottet et al. 2017, 2018; Sandström et al. 2022). However, the environmental and economic importance of this often goes unreported in sustainability assessments; most such assessments are cradle-to-farm gate, excluding further crop-processing stages and their associated by-products and co-products. One key example of this is dried distiller’s grains with solubles (DDGs), a by-product of ethanol production. Increasing world production of ethanol for biofuel has led to concurrent increases in the supply of DDGs, which could be used to replace almost half of cereals in swine and poultry diets, or even higher levels in cattle diets, with largely neutral-to-positive impacts on productivity depending on the inclusion level (Buenavista et al. 2021; Sandström et al. 2022). This improves whole-of-system efficiency and reduces feed–food competition (van Hal et al. 2019; van Zanten et al. 2019).

Similarly, growing per capita food consumption and the association of higher-quality diets with higher amounts of food waste mean that reducing food waste is a key sustainability issue (Australian DCCEEW 2018; Conrad et al. 2018; FIAL 2021), one which careful feeding of food waste to livestock could be used to address within a multi-scale sustainability framework. Food waste can be safely used in diets for livestock such as cattle, swine, and prawns, replacing human-edible ingredients without sacrificing performance (Pinotti et al. 2021; Truong et al. 2021). One key example of this is the EcoFeed program in Japan and South Korea; this process uses a combination of heat treatment and ensiling to turn food waste into pelleted stockfeed, decreasing food waste and feed–food competition, reducing biosecurity risks from the use of food waste as stockfeed and redirecting grain from livestock fodder to direct human consumption (Sugiura et al. 2009; zu Ermgassen et al. 2016; Salemdeeb et al. 2017).

The use of these by-products and waste represents a net upgrading of nutrients into the human food chain, as livestock take human-inedible nutrients and ‘upcycle’ them into highly digestible, high-quality food while reducing total resources required for food production, and increasing the net nutrient contribution to the human food supply. Mottet et al. (2017, 2018) reported that ruminants produce 67% more human-edible protein than they consume, with similar or even higher numbers when protein quality (digestibility and amino acid balance) is considered (Ertl et al. 2015; Baber et al. 2018; Thomas et al. 2021; van Barneveld et al. 2023). Because these effects span multiple biophysical scales, the true impacts of such changes are best realised within a multi-scale, multi-metric modelling framework. Such a framework could then be used to further explore livestock’s role as ‘upcyclers’, balancing effects on animal-scale productivity with system-scale impacts on nutrient capture, food loss, and waste reduction.

Filling data gaps by increasing collection of region-specific data and variation in livestock systems and food production systems

Scientists have a great opportunity to improve the applicability and utility of sustainability assessments by collecting additional data on regional variation in food production systems, specifically in understudied regions and types of production system. However, existing sustainability literature is imbalanced in terms of regions studied; Europe and North America combined contain only ~15% of the world’s population (United Nations 2022) and produce one-third of the world’s animal protein by mass (FAO GLEAM 2023), yet the majority of life-cycle inventory (LCI) data used in sustainability assessments come from these regions, as do the majority of agrifood sustainability studies (McClelland et al. 2018; Alhashim et al. 2021; United Nations 2023). This is an especially pressing issue for the assessment of sustainable livestock production, as world per capita demand for animal protein is expected to increase by 14% between 2020 and 2050, primarily in Africa and Asia, where demand is expected to increase 55% and 49% respectively (Komarek et al. 2021).

Data for many of these regions is sparse; what exists may not use common units or methods, or may be out of date, incomplete, or otherwise not usable in analyses. A report by Eady et al. (2014) identified ‘a deficiency in accurate and consistent local information’ as a key issue with conducting LCA in Australia and other poorly represented regions, noting that substantial differences in regions, yields, and production practices make the use of European/North American data outside their respective regions of origin a known source of error. There have been efforts to address this by creating region-specific databases in countries such as Australia (Eady et al. 2014), Japan (ILCAJ 2023), and Brazil (Shirosaki Marçal de Souza et al. 2021), but such efforts are constrained by data availability issues and a lack of local practitioners (Valdivia et al. 2017), as well as lack of funds for systemic data-collection and data-sharing initiatives. This is also a priority under the United Nations Sustainable Development Goals (United Nations 2023), which highlight the imbalances in data and resources that limit the progress of LMIC in developing region-specific solutions for sustainability.

As part of filling data gaps in underrepresented regions, particular care needs to be paid to representing a wide range of variation in production systems and their corresponding biophysical factors. Many LCI databases treat food products as more or less interchangeable bulk commodities, but in livestock systems the quality and source of a product can substantially effect environmental and economic impacts. Livestock breeds and management vary widely among and within regions, but this variation may be poorly represented in the literature or other data sources (Emerenciano et al. 2022). Just as North America and Europe are over-represented in current LCA data, so are their preferred breeds of livestock, while more tropically adapted breeds such as Nellore cattle are under-represented, further hindering sustainability efforts. Native breeds are also undervalued in terms of their role in sustainability, despite their adaptation to existing environments, particularly regarding heat tolerance and disease resistance (Mwai et al. 2015; Sejian et al. 2018; Son et al. 2019). Between-breed variation in resource needs and product quality can also substantially affect the environmental impacts of food production systems. For example, cheese produced from Jersey milk has lower environmental impacts and resource use than that produced from Holstein milk due to breed differences in milk composition and energy metabolism (Capper and Cady 2012; Kristensen et al. 2022), while Herron et al. (2021) found that genetic selection for maturity rate and carcass traits had significant effects on the environmental impacts of beef production, particularly when differences in carcass quality were included. This latter point is especially key, as such product differentiation (i.e. carcass quality and yield) also drives differences in producer payments and therefore economic sustainability.

Biophysical models can be used to represent among-breed differences in resource use, product quality, and environmental impact. Many models already incorporate bred-specific growth parameters for Bos indicus and Bos taurus cattle, such as GrassGro (Freer et al. 2012) and the NASEM beef model (NASEM 2016). Additional breeds can be represented, provided sufficient data are available for parameterisation. These differences should be included by using biophysical models to underpin LCA analyses, and among-breed variation in impacts should be included in LCA databases to reflect these differences and better capture the effects of different systems of livestock production. This would allow for broader representation of among-breed differences in product quality and environmental impacts, improving the specificity and applicability of analyses. To address these issues, regional data collection efforts should be prioritised for support to enable clear, representative assessment of regional variations in livestock production and sustainability, with a focus on historically under-represented areas. There is also a need for greater resourcing to develop and/or upgrade parameters.

Improve coordination of data collection and organisation to improve efficacy of data use across disciplines

Data can come from many sources such as research, industry, producer groups, government, all with their own goals, rules, and preferences regarding the collection and use of data. The framework proposed above requires the collection of additional data to ensure proper parameterisation, testing, and independent validation. It is true that advances in technology mean that more data are being collected than ever before, but actually accessing such data can be trickier; it may be siloed in proprietary software, vary in format, under existing intellectual property agreements, and vary in scope and scale (Osinga et al. 2022). Producers often have mixed feelings towards use of their data in ‘smart farming’ or broader data-collection initiatives, such as concerns about privacy, control over how the data will be used, data security. A survey of Australian farmer attitudes towards smart farming and data (Wiseman et al. 2019) found that the majority of respondents felt uncomfortable about service providers turning a profit from data collected on-farm, and trust in data security and control over data was low, similar to results found in other countries.

Therefore, this issue must be addressed from multiple perspectives, including the following:

Similar data collection and use agreements already exist in human-studies research conducted by governments and researchers, but private industry is not bound by such rules, contributing to the erosion of trust discussed above. Government agencies can play a key role in the collation and organisation of databases from multiple sources, but this requires specific support staff to ensure that data are managed appropriately and that data collected meets criteria for quality and format; such databases must be maintained and updated in the long term.

The need for clear, transparent standards and methods to define best practices for environmental sustainability assessment

Even within existing frameworks of best practices in sustainability assessment, such as the ISO and LEAP frameworks, there are still areas where frameworks require additional development. In particular, lack of consensus about allocation methods, system boundaries, and modelling of avoided impacts are key issues that should be addressed as part of developing clearer, more transparent assessment frameworks.

Uncertainty regarding modelling of avoided impacts is partly due to lack of consensus in the best methods for modelling impact allocation and system boundaries. Impact allocation refers to the divvying of impacts among multiple outputs of the same system, such as meat, dairy, and manure. Current best-practice frameworks such as ISO and LEAP (ISO 2006a, 2006b, 2014, 2018a, 2018b; FAO 2016a, 2016b, 2016c) state that allocation among livestock products should be based on biophysical relationships and/or properties where possible; otherwise, economic allocation can be used. However, allocation ratios vary substantially among study, species, and method; Wiedemann et al. (2015) found that the wool:meat allocation ratio in Australian sheep farming varied from 4:96 to 52:48, depending on the farm and the allocation method; similar variability has been found in other species (Chen et al. 2017; Dougherty et al. 2019b; Kyttä et al. 2022; Mazzetto et al. 2022).

System boundaries define the scope and scale of assessment. While it is impractical to model everything, many assessments of food production are cradle-to-farm-gate, omitting the processing and/or retail phases. Although data for these phases may be limited in many cases, not accounting for the resources and waste produced in the processing and delivery of food to the consumer will significantly affect allocation among products, and can lead to biased estimation of the impacts of a product or system (Kyttä et al. 2022; Sun and Ruiz-Carrascal 2023). For example, edible offal might account for only a small proportion of carcass weight, but it is rich in protein and micronutrients. Depending on the functional unit (i.e. weight, protein, contribution to human nutrient requirements), the choice of which products to model and the impacts and resources used in their processing can substantially change the perceived impacts and benefits of the entire system (Weidemann and Yan 2014; Wingett and Alders 2023).

A lack of consensus regarding allocation methods and optimal system boundaries has led to further uncertainty in how best to model the impacts of feeding crop residues and by-products to livestock. The most boundaries used in LCA of food production are cradle-to-farmgate. However, these boundaries do not include impacts from crop by-products or processing, leading to limited reporting and exploration of the impacts of these stages. This contributes to an inherent mismatch in boundaries between the product as it exits the farm and the product as it plays a role in human diets, as well as a lack of consistency in how to account for the wastes and co-products produced at different stages. In the LEAP framework, the distinction among by-product, residue, and waste can be either biophysical or based on revenue (FAO 2016a, 2016b, 2016c) as shown in Table 1.

However, these definitions cause problems in practice, in that such methods are context-specific and sensitive to market fluctuations, or because changes in production methods or value-adding alter the possible uses and prices commanded by a product. At what point is the economic value no longer ‘negligible’? In the LEAP framework for large ruminants, if manure from a dairy farm is sold or otherwise considered a ‘valuable output of the farm’, biophysical allocation should be used on the basis of the proportion of feed energy used for milk, meat, and manure, with example allocation factors of 77.5%, 7.2%, and 15.3% respectively (FAO 2016a). However, if manure is classified as a residual, it receives no allocation of emissions. This is a methodological inconsistency in current best-practice frameworks that leads to variation in reported results, making it trickier to use these results can be used to inform decisions, and therefore is a key issue to address.

Such inconsistency is particularly key in the context of using crop wastes/by-products as livestock feed, and current best-practice frameworks regarding the use of food waste and crop by-products by livestock could be considered to be ‘billing the garbage truck for trash collection’. Crops such as almonds, sugarcane, grains, etc. are grown for the value of their primary products to the human food supply. Livestock consumption of the remnant biomass (almond hulls, bagasse, distiller’s grains, etc.) serves as active waste prevention by repurposing items rejected for human consumption. As an example, if a biscuit company disposes of burnt/rejected biscuits (a waste) by selling them for a nominal price to a local swine producer, the unsuitable biscuits become a residual. From an LCA perspective, the biscuits disappear from the LCA of the biscuit company and appear on that of the swine producer, with no emissions allocated to the biscuits except for those from their transport to the farm. Under such methods, the biscuit company is able to avoid emissions associated with waste disposal, the swine producer is able to get a ‘lower-emissions’ feed product, and the emissions associated with the avoided waste are not accounted for or credited to either side. However, if the rejected biscuits were to become more economically valuable over time, they would then instead incur a share of emissions during production and use, effectively ‘billing’ the end user for what had previously been a way of avoiding emissions associated with waste disposal.

The use of these products also has the potential to reduce ‘feed–food competition’ by reducing the human-edible fractions of livestock diets, while simultaneously repurposing human-inedible crop by-products (Van Zanten et al. 2019). For example, approximately 30% of human-edible crops fed to livestock are fed to poultry (Sandström et al. 2022). Approximately 70% of poultry diets are edible by humans, but replacing human-edible feed with human-inedible biomass such as oilseed meal, distiller’s grains, etc. may lead to a decline in performance, which would be economically infeasible despite its environmental benefits (Sandström et al. 2022). The environmental benefits of such replacements can be quite substantial; study van Hal et al. (2019) noted that factoring such reductions in feed–food competition into sustainability assessment reduced the carbon footprint of egg production by 57% and opportunity cost-weighted land use by 88%, without factoring in further avoided emissions or potential economic impacts.

Just as producers should be credited for improvement, it is equally important to give credit to those who are already doing the right thing, even if doing the right thing environmentally may be less economically feasible (Herrero et al. 2016; Harrison et al. 2021; Thomas et al. 2023). This is both an opportunity and a serious need that must be addressed for holistic assessment of the impact of food waste/crop by-products as food from a holistic perspective. Such benefits and tradeoffs would be explicitly considered under the holistic modelling framework presented in this review, but to properly accomplish this, there must be clearer frameworks in place for how to conduct these assessments. Once these clearer methods are in place, it would then be possible to develop and implement programs to make these reductions in feed–food competition more economically sustainable via the use of carbon credits, biodiversity credentialling programs, or other initiatives, which could then help stakeholders make better decisions.