Various devices have been developed that enable direct measurements of enteric methane emissions. These devices are now used by most research organisations ( Mesgaran et al., 2021), particularly in INRAE’s experimental units. These include respiratory chambers, which are considered the reference method, and sulphur hexafluoride (SF6), a tracer gas that is administered orally with a vector which allows for controlled, constant release into the animal’s rumen. Another device is GreenFeed®, which takes spot measurements of the methane emitted by animals when they visit an integrated automatic concentrate dispenser. Doreau et al. (2018) showed comparable average methane emission levels between the three measurement devices for groups of animals in European livestock systems and validated the reliability of GreenFeed® measurements on-farm. However, the workload involved in using GreenFeed®, the cost of acquiring and operating the device and the reliance on the company for data collection are major obstacles to its use on commercial farms. The applicability of these devices in tropical environments has also recently been discussed ( Gbenou et al., 2024a).

For several years, equations have been developed to predict enteric methane emissions, by establishing the relationships between measured emissions and various explanatory factors such as feed intake and diet composition, for different categories of animals (e.g. INRA, 2018; van Lingen et al., 2019; Belanche et al., 2023). Measurements in biological matrices obtained from the animals (milk and faeces) are also used for larger-scale estimates. Initially, the fatty acid (FA) composition of milk was used as an indicator of methane emissions in dairy cows fed maize silage containing extruded linseed ( Chilliard et al., 2009), and subsequently in cows fed a wide range of diets ( Bougouin et al., 2019). Determining the FA composition by gas chromatography is expensive and time-consuming. Therefore, equations based on the detailed FA composition are not suitable for measuring large numbers of animals. Conversely, the mid-infrared (MIR) absorption spectra of milk are now routinely collected for millions of cows to predict the milk composition for dairy inspection purposes. These spectra have been used to predict methane emissions in dairy cattle ( Vanlierde et al., 2021; Fresco et al., 2024b) and enable genomic evaluation to select low emitted dairy cattle (Part 3). Prediction equations have also been developed using near-infrared absorption spectra (NIRS) acquired from faeces, a preferred matrix for suckler herds and for growing dairy animals. Promising results have been obtained using a dataset of suckler cows for which methane emissions were measured using the GreenFeed® method ( Andueza et al., 2022).

However, the robustness of the various prediction equations must be assessed to accurately estimate the reductions that can be achieved by mitigation strategies at the level of the livestock farming systems ( Benaouda et al., 2019; Blondiaux et al., 2024).

Direct and indirect measurement tools should complement one another. This is essential for obtaining accurate predictors. The experimental infrastructures of INRAE and its research and development partners are acquiring new devices to complement those already in use (Box 2). These tools will be essential for future studies aimed at testing the most effective strategies for reducing enteric methane emissions in various farming environments. Portable Accumulation Chambers (PACs) are now available for small ruminants. They measure the gases emitted by flatulence over a period of 1 hour, as well as the methane emitted by belching and exhalation ( Goopy et al., 2011). Sniffers are usually located inside milking robots and use infrared technology to measure the animals’ methane emissions. The predictive potential of these indirect methods must be compared with the predictions based on MIR spectra in milk. Developing predictors adapted to farm constraints is indeed a priority. In France, sniffers will soon be installed on 50 commercial dairy farms, and PAC will be used on various commercial small ruminant farms to collect data. This will enable comparisons to be made between farming systems and the predictors to be evaluated in different contexts.

Improving the accuracy of already existing predictors is also essential for extending their use. Complementary research studies in cattle and small ruminants will be conducted to evaluate the potential of milk MIR spectra or faecal short-wave infrared (SPIR) spectroscopy to predict methane emissions. Other new predictors are under study, including plasma metabolites such as amino acids, glucose, carboxylic acids and glycerolipids ( Yanibada et al., 2020), the carbon isotopic abundance of milk ( Saro et al., 2025), microorganisms and hydrogen fluxes ( Bedoya Mazo et al., 2023).

Continuous data acquisition, supported by animal experimentation under national and international collaborations, remains essential. Combining the databases will enable us to develop more robust predictors and discriminating models of methane enteric emissions. It will also ensure that inter-individual variability is considered when evaluating farming systems in different pedoclimatic contexts (Part 5).

Manipulating the production and consumption of hydrogen is considered a key approach to mitigating enteric methane emissions ( Martin et al., 2010; Mackie et al., 2024). Researchers have focused on evaluating alternative hydrogen acceptors, to divert hydrogen away from methanogenesis. Guyader et al. (2017) demonstrated that supplementing diet with nitrates, an oxidant that can be reduced by hydrogen, can stimulate metabolic pathways that accept hydrogen, thereby competing with methanogenesis. However, the underlying microbial processes remain to be elucidated. Other alternative hydrogen acceptors such as phenolic compounds have been studied using in vitro approaches ( Huang et al., 2023; Romero et al., 2023). Phenolic compounds are secondary plant metabolites that can be metabolised by microbial populations into acetate, while consuming hydrogen. Of the 10 or so phenolic compounds tested at INRAE, gallic acid and phloroglucinol (which were derived here from the chemical industry, but are also found naturally in tree bark) increased gas production and total VFA production, particularly acetate, without stimulating methane production. The effect of phloroglucinol was evaluated in combination with 2-bromoethanesulfonate, a specific methanogen inhibitor. Methane production was completely inhibited, hydrogen accumulation was reduced and acetate production increased compared to treatment with 2-bromoethanesulfonate alone. Research programmes on hydrogen transactions are on-going at INRAE, with a particular focus on quantitative data within the rumen ecosystem.

After birth, the rumen is colonised sequentially by microbiota. Microorganisms are present from the second day of life and are characterised by the abundance of Pseudomonadota (formerly Proteobacteria) which accounts for nearly 70% of the total. The composition changes rapidly, with the proportion of Pseudomonadota falling to 30% before the ingestion of solid food. Meanwhile, the proportion of Bacteroidota, particularly Prevotella, increases to over 50%. If any induced changes in rumen microbiota persist in the long term, the plasticity of this microbiota at the beginning of an animal's life may provide opportunities to reduce enteric methane emissions. In dairy calves, administering 3-nitrooxypropanol (3-NOP), a methanogenesis-inhibiting additive, from birth until 14 weeks of age resulted in a lasting reduction in enteric methane production, which was still observable at 1 year of age ( Meale et al., 2021). The composition of bacteria, and to a lesser extent that of Archaea, was altered in the rumen, and this alteration persisted with some variations even after the administration of 3-NOP ceased. Similarly, by supplementing the diet of pregnant heifers - and then, their calves from birth to weaning - with 3-NOP, Martinez-Fernandez et al. (2024) reported a persistent reduction in methane emissions 28 weeks after stopping the treatment for at least one group of treated calves. The rumen microbiota of these calves was similar to that of the low-emitting adult cows found in Prim'Holstein INRAE herds (Ramayo-Caldas et al., 2020).

Such early modulation of the microbiota enables constraints associated with different cattle management practices, such as grazing, to be overcome. Indeed, when grass is mature and not particularly diverse, grazing increases methane emissions in ruminants compared to a diet based on energy-rich concentrates ( Roques et al., 2024).

Mechanistic modelling is a complementary approach to animal experimentation for representing and predicting rumen metabolism. Muñoz-Tamayo et al. (2016) proposed a model of the rumen microbiota organised into functional groups according to their utilisation of glucose (hexoses), amino acids, and hydrogen. This model includes glucose utilisation flux allocation factors and hydrogen dynamics. It is crucial to consider hydrogen in the model. However, our understanding of the microbial processes and thermodynamic rules that govern hydrogen transactions is incomplete, particularly for quantifying hydrogen fate and the impact of energy allocation on microbial fermentation ( Morgavi et al., 2023). Nevertheless, this model can accurately depict in vitro fermentation data from mixed diets containing different levels of concentrates ( Serment et al., 2016). It does not consider biological processes, such as VFA absorption rates and the passage of substrates. This model is currently being extended to incorporate these in vivo conditions, and preliminary results comparing simulated and experimental data from a small number of Nordic Red dairy cows are promising ( Muñoz-Tamayo et al., 2023a). In parallel, a modelling approach has been developed that links microbial data to VFA dynamics; this uses a semi-continuous in vitro system (RUSITEC) that mimics in vivo conditions, and data obtained from the rumen of dairy cows ( Davoudkhani et al., 2024).

The model developed by Muñoz-Tamayo et al. (2016) was also expanded to represent the effect of the macroalga Asparagopsis taxiformis on the dynamic profile of microbial fermentation measured in vitro. The inhibition of methane production by this macroalga leads to the accumulation of hydrogen, which exerts thermodynamic control over fermentation by directing the flux towards VFA production. The models proposed by INRAE ( Muñoz-Tamayo et al., 2021) and van Lingen et al. (2021) are the only two mechanistic models capable of representing the impact of anti-methanogenic additives. At the microbial level, modelling kinetics enabled the quantification of metabolic and energetic differences between three methanogen species ( Muñoz-Tamayo et al., 2019). The data suggest that, in addition to kinetic and thermodynamic factors, other forces such as microbial interactions (e.g. endosymbiosis with protozoa and microbial aggregation) and spatial variation (e.g. fractions of liquid and particles) contribute to the ecology of the methanogen community in the rumen, thereby promoting microbial diversity.

Integrating genomic information from the microbiota should improve the predictive accuracy of the models ( Muñoz-Tamayo et al., 2023b). The genomic sequence of the cellulolytic bacterium Fibrobacter succinogenes S85 can be used to reconstruct its metabolic network. Then, it is possible to develop a dynamic metabolic model that predicts the production of acetate, succinate and formate from glucose, cellobiose and cellulose metabolism ( Fakih et al., 2023). This approach will soon be extended to other rumen microorganisms, including methanogens.

A better understanding of how microbiota produce methane could lead to the development of new strategies reducing enteric methane emissions, by managing rumen hydrogen metabolism, modulating the microbiota during early life and using mechanistic modelling to address microbial complexity. This would help identify the trade-offs between reducing methane emissions, improving feed efficiency and production performance, and ensuring animal health and welfare.

Two main research perspectives have been identified to achieve this:

(i) Elucidating the dynamics of the rumen microbial ecosystem, particularly its resilience, and the interactions between microbial species. This would improve our understanding of the rumen microbiota and allow us to consider this dynamic more effectively in future strategies to reduce enteric methane emissions.

(ii) Develop a continuum of in vivo, in vitro and in silico studies. In vitro approaches enable the fermentation pathways associated with substrates and additives, as well as the related physicochemical parameters, to be evaluated. This simplifies the complexity observed in vivo. In silico approaches are based on in vitro and in vivo methods to predict the effectiveness of proposed strategies. The continuum operates in both directions: towards in silico for understanding and towards in vivo for application. This synergy in elucidating microbial mechanisms and proxies contributes to the principles of the 3Rs - replace, reduce, and refine - of using animals for scientific purposes.

Predicting methane emissions from MIR spectra enables large populations of phenotyped animals to be obtained. Indeed, MIR spectra are acquired in very large numbers, approximately 15 million per year from around 2.5 million dairy cows, and the marginal cost of predictions is low once the equations have been established. Two research and development consortia (Optimir and MethaBreed) have developed equations based on reference measurements from respiratory chambers or using the SF6 method (Fresco et al., 2021), or by using GreenFeed® in nine experimental conditions ( Fresco et al., 2024b). These equations can be used to predict various emission traits, such as per animal (g/day), per production unit (g/kg standardised milk) or per intake unit (g/kg dry matter intake). Daily methane production can also be calculated by multiplying methane emissions expressed as per kilogram milk, by the daily quantity of milk produced. These equations have moderate accuracy. This shortcoming is partially offset by the large amount of data, which significantly reduces loss of accuracy at the genetic level without eliminating bias.

Fresco et al. (2024a) subjected predictions to an in-depth genetic analysis, which showed that heritability is quite strong. This is not surprising, given that the absorbances of the most informative wavelengths in the MIR spectra are relatively heritable themselves. The genetic correlations between predicted traits are moderate, with the highest correlation (0.6) found between methane emissions in gram per day predicted by the two equations established by the two consortia. This moderate correlation can be explained by the fact that the two consortia used different measurement methods and calibration populations. With the exception of the reconstructed prediction, which is closely linked to milk yield, the correlations with production traits are moderate. Moreover, the reconstructed prediction is challenging to utilise, as a decrease in enteric methane emissions would result in selection against milk production.

The pilot genomic evaluations are available ( Fresco et al., 2024c). They were carried out using a single-step model, based on predictions from the official milk recording of several million cows, out of which several hundred thousand were genotyped. This method simultaneously analyses the performance of both genotyped and untyped animals. It implicitly estimates the genotype expectation of untyped animals based on the genotyping of their relatives. This increases accuracy and reduces bias. These genomic evaluations demonstrate the potential of young bulls in reducing enteric methane emissions. Furthermore, they demonstrate that the genetic potential of animals for methane emissions has been minimally impacted by recent selection objectives.

Genomic evaluations are expected to be officially launched by selection organisations in 2026. The methane trait (g/day) is the average of the two equation predictions established by the two consortia. This is a more robust approach than selecting just one of the two predictions when there is a lack of objective criteria on which method is superior. In the absence of direct economic value for the farmers, the weight given to methane emission reduction in the overall selection objective, as defined by selection organisations, may be based on the expected gain. Targeting 1% annual reduction in methane emissions (in g/day) means a weight of around 25% for the methane trait in the selection objective ( Fresco et al., 2025). This would moderately slow down the genetic progress for the other traits already under selection. Although no adverse effects have been observed to date, it will be necessary to confirm that selecting against methane does not affect the digestive processes of the ruminants. It is not feasible to use milk MIR spectra in suckler cows, as they are not milked. An approach based on SPIR spectra in faeces is currently under study. If this proves effective, then a measurement system will need to be developed before selection against methane emissions in suckler cows can be considered.

Predictions based on milk MIR spectra are moderately accurate. Many factors influencing methane emissions are not, or are only partially, considered in MIR predictions based on milk production. Therefore, to reduce methane emissions more effectively, it is also appropriate to address these factors in addition to direct genetic selection. Most of these factors depend on both farming systems and genetics. Therefore, the content presented in this section is relevant to the implementation of changes to farming systems. Improvements in breeding traits correlated with methane emissions, such as age at calving, fertility, health and body size, generally have a positive effect on the economic performance (reducing feed and veterinary costs) of the farms.

The key solutions are reducing losses, particularly mortality, and minimising unproductive periods, primarily by lowering the number of heifers.

One option is to reduce the age at first calving. In France, the age at first calving ranges from 2 to 3.5 years; more than 65% of the dairy and suckler herds calve at 3 years of age or older. Reducing this age of first calving at 2 years would save an annual cohort of 1 million near-adult-sized animals, each producing 250–300 g of methane per day. This would primarily affect the farming system, but selecting for precocity could also be helpful, especially for suckler breeds which tend to mature later. Although calving at 2 years of age may seem ambitious, any progress towards greater precocity is desirable despite the various obstacles. For suckler cows, it is also important to minimise the number of years without calving, which implies maximising fertility.

A second option is reducing the replacement rate. Although genetics is not the most important factor, improving longevity by selection is possible, with gains of +0.5 to +1.0 lactation being achieved. This would decrease the need for replacement heifers. Reducing the replacement rate by 25% per year would contribute to a 6% reduction in methane emissions from French dairy cattle with a first calving at two years of age. With first calving at three years of age, the reduction in methane emissions would be 10%. These results are illustrated by Dall-Orsoletta et al. (2019). Reducing the rate of involuntary culling would also lower the replacement rate. This could be achieved by improving health and fertility. Health problems result in production losses and unnecessary methane emissions. It is now possible to select for health, with multiple indicators available such as udder health, leg health, and susceptibility to diseases (paratuberculosis and ketosis). Thus, health and fertility are important traits in dairy selection objectives.

A third option is to reduce methane emissions related to maintenance requirements. French dairy and suckler cattle breeds are characterised by their large size. The bigger the animal, the more feed it consumes and the more methane it produces during the breeding period. Reducing adult weight by 100 kg is possible by selection. This would reduce maintenance requirements by 10%, and lead to a 3% to 5% decrease in methane emissions (depending on the relative importance of maintenance in total requirements). This would also have favourable effects on other traits, such as precocity and longevity. Selecting for a smaller size could be achieved without adversely affecting milk production, as there is only a very moderate genetic correlation between the two traits. Precocity in animals reared for meat, which is facilitated by the reduction in adult size, would also favour fat deposition; this would reduce the age at which the animals are slaughtered and the amount of concentrates needed.

Finally, in a given farming system, improving animal productivity is widely used as a GHG mitigation strategy ( Dall-Orsoletta et al., 2019): low-producing dairy cows produce more methane per kilogram of milk, and thus, should be avoided. Selecting animals with better feed efficiency and improved robustness (i.e. better adapted to the environment) is also an option.

There are many genetic options to reduce methane emissions, both directly and indirectly. One strategy, such as in the on-going project Methane 2030 (Box 3), is to construct a synthetic methane index that incorporates all traits (direct and indirect) associated with variations in methane emissions. The success will depend primarily on the selection pressure – that is, how much weight is given to methane in the selection objective. Farmers must be financially incentivised for their efforts before they agree on high selection pressure. This means that progress in reducing methane emissions can be measured and valued economically on markets. Therefore, voluntary public policies and/or industry initiatives are required. To reduce methane emissions by 1% per year, the selection objective must place a higher value on methane than the current market price of carbon. The value must reflect the situation in 8-10 years from now, since the effects of genetic selection only become apparent in the long term.

Livestock farming systems include different management practices in both temperate zones and under hot climates, such as free-range, grazing, rangeland, agroforestry and mixed rations combining fodder and concentrates, among others. There are also various types of feeds, including co-products, crop residues, industrial by-products, and tree and forage species. This results in a wide range of enteric methane emissions between systems. Notably, enteric methane emissions vary between temperate and tropical zones ( Archimède et al., 2011; Martin et al., 2021; Gbenou et al., 2024b). The types of feed used differ, as do the breeds. A better understanding of the physiology of some breeds is still needed.

Feeding systems provide valuable insights into feed composition, animal characteristics and feeding practices in temperate regions. For example, the latest version of INRA feeding system ( INRA, 2018) calculates methane emissions based on digestible organic matter, considering feed intake and the proportion of concentrates in the ration. It estimates a decrease in methane yield for each percentage increase in concentrates. However, feeding systems are less suitable for ruminant farming systems in hot areas. A literature review by Quantin et al. (2025) indicates that the methane emissions per kilogram of dry matter intake are 23.2 (range 16.9–29.1) for dairy cows and 27.8 (range 18.7–34.5) for heifers in sub-Saharan Africa, compared with an average of 21.0 in temperate zones. The values are 21.9 (range 16.8–26.5) and 31.7 (range 12.6–49.8) for sheep and goats, respectively, in sub-Saharan Africa, compared with an average of 18.0 in temperate zones.

Equations can be used to predict methane emissions and nitrogen excretion to quantify the emissions of all pollutants associated with livestock farming. However, they are subject to uncertainty ( Benaouda et al., 2019) and are not always applicable to plants that are rich in secondary bioactive compounds or to additives that have specific modes of action. Emission factor uncertainty and lack of metadata result in the inadequate consideration of the environmental impacts of livestock in GHG calculators and inventories ( Eugène et al., 2019; IPCC, 2019). Nevertheless, these issues can be addressed by dedicated methods, to improve the selection of instruments for the direct measurement of GHG emissions ( Hammond et al., 2016), as well as by relevant proxies ( Vanlierde et al., 2024) and models ( Blondiaux et al., 2024).

At the farming-system level, strategies to mitigate enteric methane emissions often involve increasing herd productivity, longevity and reproduction. The aim is to reduce emissions through dilution while maintaining economic viability. In temperate regions, highly specialised livestock farming systems that prioritise technical and economic performance can result in lower methane emissions per unit of milk or meat produced. However, these systems produce more methane per unit of surface area due to the high concentration of animals, and result in significant environmental damage through high nitrogen emissions into the air and water, as well as reduced biodiversity ( Chemineau et al., 2025). Conversely, agroecological systems increase soil carbon storage, promote biodiversity and reduce nitrous oxide emissions. However, they have little effect on methane emissions ( Blaix et al., 2026). Strategies to reduce enteric methane emissions could be introduced, provided they are compatible with agroecological principles that promote interactions between organisms and their integration into the local ecosystem.

Numerous studies have highlighted the benefits of mixed farming systems, which improve environmental efficiency by incorporating agroecological principles, such as closing nutrient cycles and promoting plant and animal biodiversity. In hot regions, other strategies include improving pasture quality, supplementing diets, and changing land use and ruminant breeds ( Thornton & Herrero, 2010). Natural grasslands are often used for grazing livestock. Farmers deal with the lack of feed by managing rations every day (e.g. using woody and herbaceous forages) and by moving livestock to new pastures (transhumance). In these systems, there are large seasonal variations of fodder quality and availability. Studies in the Sahel ( Assouma et al., 2018; Bois et al., 2020) have assessed the impact of this seasonality, demonstrating that fodder intake during the dry season results in increased methane production in vitro. Supplements can reduce methane emissions per kilogram of dry matter ingested, particularly when nitrogen intake is not limiting ( Gbenou et al., 2024b).

Agropastoral farming systems are examples of agroecological systems because they exploit the complementarity and synergy between spontaneous and cultivated plants and animal diversity (behaviour and performance). They generally have low environmental impacts ( Vigan et al., 2017). Although they are more vulnerable to hazards, they offer better adaptability and mitigation capacity by promoting mixed farming (or even poly-livestock farming), which has positive effects on carbon storage and manure utilisation. Analysing nutrient flows provides a better understanding of efficient processes and those associated with methane emissions ( Puech & Stark, 2023). This allows researchers to propose options that promote synergies between carbon and nitrogen. Indeed, livestock play a crucial role in the spatial distribution of organic matter, soil nutrients and carbon stocks. For example, in tropical ecosystems, around 50% of the dry matter in fodder is returned to the soil ( Schlecht et al., 2004). The carbon balance of a pastoral area can even be negative when all emissions are considered. In a semi-arid silvo-pastoral landscape ( Assouma et al., 2019), the amount of carbon fixed over the annual cycle (in soil organic matter, wood, and tree roots) exceeds the emissions caused by animals and their manure (methane fermentation from livestock and termites, GHG emissions from soil and water, and vegetation fires and motorised pumps).

Together, polyculture-livestock systems in harsh environments can provide inspiration for mitigation strategies in temperate regions, particularly those most susceptible to climate change.

The first challenge is to improve our understanding of, and develop new tools for managing, animal and plant diversity in different regions and climates. This involves calibrating existing models using specific data sets and/or developing context-specific models based on the local animal and plant resources.

A second challenge is developing appropriate indicators or metrics with which to assess the contribution of different livestock systems to sustainability. In this context, it is important to consider methane and other GHG emissions together with carbon storage and nutrient cycling (nitrogen and phosphorus).