Many practices have been advocated to mitigate emissions through the mechanisms cited above. Often a practice will affect more than one gas, by more than one mechanism, sometimes in opposite ways, so that the net benefit depends on the combined effects on all gases (Robertson & Grace 2004; Schils et al. 2005). In addition, the temporal pattern of influence may vary among practices or among gases for a given practice; some emissions are reduced indefinitely, other reductions are temporary (Marland et al. 2003a; Six et al. 2004). Where a practice affects radiative forcing through other mechanisms such as aerosols or albedo, those impacts also need to be considered (Marland et al. 2003b; Andreae et al. 2005). The impacts of various mitigation options considered are summarized in table 1. The most important options are discussed below.

For the livestock-based options, mitigation potentials (dairy cows, beef cattle, sheep, dairy buffalo and other buffalo) for reducing enteric methane emissions through improved feeding practices, specific agents and dietary additives, and longer term structural and management changes/animal breeding are shown in table 3. These estimates were derived using a model similar to that described in US-EPA (2006). The proportional reduction due to the application of each practice was estimated from reports in the scientific literature (see footnotes to table 3 for main references). These were adjusted for (i) the proportion of the animal's life where the practice was applicable, (ii) the technical adoption feasibility in a region, i.e. whether the farmers have the necessary knowledge, equipment, extension services, etc., to apply the practice (average dairy cow milk production in each region over the period 2000–2004 was used as an index of the level of technical efficiency in the region and to score a region's technical adoption feasibility), (iii) the proportion of animals in a region to which the measure can be applied (i.e. if the measure is already being applied to some animals as in the case of bST use in North America, it is considered to be applicable only to the proportion of animals not currently receiving the product), and (iv) non-additivity of simultaneous application of multiple measures. There is evidence in the literature that some measures are not additive when applied simultaneously, such as the use of dietary oils and ionophores, but this is probably not the case with most measures. However, we did account for the fact that once one measure is applied, the emissions base for the second measure is reduced and so on, and we also incorporated a further 20% reduction in mitigation potential to account for unknown non-additivity effects. Only measures considered feasible for a region were applied in that region (e.g. bST was not considered for European regions due to the ban on its use in the EU). It was assumed that total production of milk or meat was not affected by application of the practices, so that if a measure increased animal productivity, animal numbers were reduced in order to keep production constant.

As can be seen from tables 2 and ​and3,3, some of the mitigation measures operate predominantly on one GHG (e.g. dietary management of ruminants to reduce CH₄ emissions) while others have impacts on more than one (e.g. rice management). Some practices benefit more than one gas (e.g. set-aside/headland management) while others involve a trade-off between gases (e.g. restoration of organic soils). Table 2 also shows that the effectiveness of some mitigation practices differs between climate regions and can also differ within a climate region. A practice that is highly effective in reducing emissions at one site may be less effective or even counterproductive elsewhere. This means that there may be no universally applicable list of mitigation practices, but that any proposed practices will need to be tuned to individual agricultural systems present in specific climatic, edaphic and social settings.

All data were converted to real-area projections and the areas in square metre were converted to hectare. Cropland mitigation options were applied to the total crop area (minus those under rice cultivation, irrigation, set-aside or on organic soils or degraded soils, since other mitigation occurred on these lands), total mitigation was taken as the mean of the agronomy, nutrient management and tillage/residue management effects on 95% of the land, plus improved biosolid management on 5% of the land. Grazing land management was applied on all grassland, restoration of organic soils and degraded lands on the croplands occurring on these areas as calculated above, bioenergy on the land projected to be available for bioenergy production in 2030 by the Image v. 2.2 model (Strengers et al. 2004; Hoogwijk et al. 2005). Water management was applied only on the irrigable area identified in the FAO AEZ database, and agroforestry and set-aside only on projected surplus cropland in 2030. The total area of cropland and grassland for each region in 2030 for each SRES scenario is shown in table 4.

For emissions from livestock, total cattle, sheep and buffalo numbers in the various regions were obtained from FAOSTAT (2006). The cattle numbers for each region were broken down into numbers of dairy cattle and other cattle (owing to the different reduction potentials of both types) using FAOSTAT (2006). The biophysical emissions reduction potentials of the various practices were determined as described above. Estimated marginal costs of implementing each mitigation practice are shown in table 5.

In agriculture, there is a relationship between the amount paid for GHGs (i.e. the price of CO₂ equivalents) and the level of mitigation realized. The amount of mitigation achieved for a given carbon price can be used to define a marginal abatement curve (MAC) for each practice for each region. We used the MACs from US-EPA (2006) to define the level of implementation (economic potential) for each practice in each region, for carbon prices up to 20, up to 50 and up to 100 US$ t CO₂-eq.⁻¹ practices as described below:

At low prices, the dominant strategies are those consistent with existing production such as change in tillage practice, fertilizer application, diet formulation and manure management, while higher prices elicit land use changes that displace existing production, such as biofuels (and afforestation), and allow the use of more costly animal feed-based mitigation options. The portfolio of mitigation strategies also varies over time owing to (i) the limited ecological capacity of the sequestration related strategies (i.e. their approach to a new carbon equilibrium over time) and (ii) the limited market penetration potential of capital intensive strategies like biofuels (which are constrained by the rate of turnover in energy processing plants, prospects and costs of retrofits, and energy product growth; Lee et al. 2005). It is important to note that while the most prevalent cost-mitigation quantity schedules are for single strategies (i.e. the amount of sequestration obtained as prices increase; as in Antle et al. 2001), it is not valid to sum these to gain a total mitigation potential due to resource competition among strategies. For example, Schneider & McCarl (2006) show that at higher prices, adding individual strategies can yield a total mitigation estimate that is as much as five times too large.

The global technical mitigation potential from agriculture by 2030, considering all gases, is estimated to be approximately 5500–6000Mt CO₂-eq.yr⁻¹, with cumulative economic potentials of 1500–1600, 2500–2700 and 4000–4300Mt CO₂-eq.yr⁻¹ at carbon prices of up to 20, up to 50 and up to 100 US$ t CO₂-eq.⁻¹ (table 6). To put these figures in context, annual CO₂ emissions during the 1990s were approximately 29000Mt CO₂-eq.yr⁻¹, so agriculture could offset, at full biophysical potential, about 20% of total annual CO₂ emissions, with offsets of approximately 5, 9 and 14% at CO₂-eq. prices of up to 20, up to 50 and up to 100 US$ t CO₂-eq.⁻¹.

Of these total mitigation potentials, approximately 89% is from reduced soil emissions of CO₂, approximately 9% from mitigation of methane and approximately 2% from mitigation of soil N₂O emissions (figure 2). For each region, the biophysical potential is defined by the sum of the potential due to (i) improvements in cropland management (mean of cropland management, tillage practice, nutrient and manure management and water management) for the whole cropland area in 2030, (ii) improved grazing land management for the whole grassland area in 2030, (iii) reduction of soil GHG emissions under bioenergy cropping, (iv) improved rice management of the whole rice area, (v) restoration of native ecosystems on currently cultivated organic soils, (vi) restoration of all degraded lands, (vii) improved livestock management (mean of mitigation due to feeds/inocula/breeding and systems) and (viii) improved manure management. Figure 3 shows the total mitigation potential per region using the mean per-area estimates of potential for all practices and GHGs considered together.

The low, mean and high regional estimates of the biophysical mitigation potential are shown in figure 4. The low and high estimates about the mean (e.g. low and high estimates are approximately 400 and 10600Mt CO₂-eq.yr⁻¹, respectively about the mean estimate of 5500Mt CO₂-eq.yr⁻¹) are largely determined by uncertainty in the per-area estimate for the mitigation measure. For soil CO₂ emission reduction, this arises from the mixed linear effects model used to derive the mitigation potentials, accounting for approximately 89% of the total potential. It is important to note that the most appropriate agricultural mitigation response will vary at the regional level and different portfolios of strategies will be developed in different regions and in countries within a region.

Estimates in the IPCC Second Assessment Report (SAR; IPCC 1996) suggested that 400–800Mt C yr⁻¹ (equivalent to approximately 1400–2900Mt CO₂-eq.yr⁻¹) could be sequestered in global agricultural soils with a finite capacity saturating after 50–100 years. In addition, the SAR concluded that 300–1300Mt C (equivalent to approximately 1100–4800Mt CO₂-eq.yr⁻¹) from fossil fuels could be offset by using 10–15% of agricultural land to grow energy crops, with crop residues potentially contributing 100–200Mt C (equivalent to approximately 400–700Mt CO₂-eq.yr⁻¹) to fossil fuel offsets if recovered and burned. It was noted that burning residues for bioenergy might increase N₂O emissions but this effect was not quantified. The SAR concluded that CH₄ emissions from agriculture could be reduced by 15–56%, mainly through improved nutrition of ruminants and better management of paddy rice. It was also estimated that improvements in agricultural management could reduce N₂O emissions by 9–26%. The SAR noted that GHG mitigation techniques will not be adopted by land managers unless they improve profitability, but that some measures are adopted for reasons other than for climate mitigation. Options that both reduce GHG emissions and increase productivity are more likely to be adopted than those which only reduce emissions.

In the IPCC Third Assessment Report (TAR; IPCC 2001), estimates of agricultural mitigation potential by 2020 were 350–750Mt C yr⁻¹ (approximately 1300–2750Mt CO₂ yr⁻¹). It was noted that the range was mainly caused by large uncertainties about CH₄, N₂O and soil-related emissions of CO₂ and that most reductions will cost between US$ 0 and 100tC-eq.⁻¹ (approximately US$ 0–27t CO₂-eq.⁻¹) with limited opportunities for negative net direct cost options. The analysis of agriculture in the TAR included only conservation tillage, soil C sequestration, nitrogen fertilizer management, enteric methane reduction and rice paddy irrigation and fertilizers. The estimate for global mitigation potential was not broken down by region or practice.

These estimates, based on the best data currently available, are comparable with previous estimates, but give for the first time, an assessment of the agricultural mitigation potential for all gases, for all regions, at a range of potential carbon costs. The comparison of previous estimates of agricultural mitigation potential with comparable figures from this study is summarized in table 7. Given the differences in areas considered and the different assumptions made in previous studies, the estimates in this study are strikingly similar.

In addition to GHG emission reduction, agricultural land can provide feedstock for bioenergy production. Bioenergy to replace fossil fuels can be generated from agricultural feedstocks including by-products of agricultural production and dedicated energy crops. For residues from agriculture, the energy production and GHG mitigation potentials depend on yield/product ratios and the total agricultural land area, as well as type of production system. Less intensive management systems require reuse of residues for maintaining soil fertility. Intensively managed systems not only allow for higher usage rates of residues but also usually deploy crops with lower crop-to-residue ratios. Estimates of energy production potential from agricultural residues vary between 15 and 70EJyr⁻¹. The latter figure is based on the regional production of food (in 2003) multiplied by harvesting or processing factors and the assumed recoverability factors. These figures do not subtract the potential alternative use for agricultural residues. As indicated by Junginger et al. (2001), competing applications can reduce the net availability of agricultural residues for energy or materials significantly. In addition, the expectations about future availability of residues from agriculture vary widely among the studies. Dried dung can also be used as an energy feedstock. The total estimated contribution could be 5–55EJyr⁻¹ worldwide, with the range defined by current global use at the low end, to technical potential at the high end. Usage in the longer term is uncertain because dung is considered a ‘poor man's fuel’.

Organic wastes and residues together could supply 20–125EJyr⁻¹ by 2050, with organic wastes potentially having an important role. The potential fossil fuel offset for 2050 from agricultural organic wastes and residues when used for energy production, assuming that it replaces gas, its energy content is 20GJ t ⁻¹ of dry biomass (IPCC 2001) and 1 t of dry biomass used to generate electricity prevents 0.28t C from gas from being emitted to the atmosphere (Cannell 2003), is 1000–6000 Mt CO₂-eq.yr⁻¹. If we assume linear uptake, a rough estimate of the potential by 2030 is 600–4000 Mt CO₂-eq.yr⁻¹.

The energy production and GHG mitigation potentials of dedicated energy crops depend on land availability, considering that food demand has to be met, combined with nature protection, sustainable management of soils and water reserves and other sustainability criteria. Since future biomass resource availability for energy and materials depends on these factors, an accurate estimate is difficult to obtain. Berndes et al. (2003) reviewed 17 studies of future biomass availability and showed that no complete integrated assessment and scenario studies were available.

Energy cropping on current agricultural land could, with projected technological progress, deliver over 800EJyr⁻¹ without jeopardizing the world's food supply. Various studies have arrived at differing figures for the potential contribution of biomass to future global energy supplies ranging from below 100EJyr⁻¹ to above 400EJyr⁻¹ in 2050. A recent study (Sims et al. 2006), using lower per-area yield assumptions and bioenergy crop areas projected by the Image v. 2.2 model, suggests more modest potentials by 2025. The differences among studies are largely attributable to uncertainty in land availability and yield levels. The potential fossil fuel offset from dedicated energy crops by 2050, if assumed to supply 100–400EJyr⁻¹ by replacing gas, and assuming 20GJ t⁻¹ of dry biomass (IPCC 2001) and that 1 t of dry biomass used to generate electricity prevents 0.28tC from gas from being emitted to the atmosphere (Cannell 2003), is 5000–20000Mt CO₂-eq.yr⁻¹. If we assume linear uptake, a rough estimate of the potential by 2030 is 3000–12000Mt CO₂-eq.yr⁻¹.

Total GHG mitigation potential from agricultural bioenergy by 2030, including dedicated energy crops and agricultural wastes and residues is 4000–16000Mt CO₂-eq.yr⁻¹. The economic analysis presented above, using figures for bioenergy uptake from Lee et al. (2005), suggests that 4, 14 and 100% of the biophysical potential would be implemented at 0–20, 0–50, 0–100 US$ t CO₂-eq., respectively. Assuming that 16000Mt CO₂-eq.yr⁻¹ represents the total biophysical potential, economic mitigation potential of biomass energy from agriculture at 0–20, 0–50, 0–100 US$ t CO₂-eq. is estimated to be 640, 2240 and 16000Mt CO₂-eq.yr⁻¹ accounting for 30, 90–100 and 500% of all other agricultural GHG mitigation measures combined, respectively. The bioenergy mitigation potential is compared to other agricultural GHG mitigation options at a range of prices of CO₂-eq. in figure 5.

Like mitigation from bioenergy, where the mitigation effect is usually counted in the user sector, enhanced energy efficiency (i.e. through reduced fossil fuel) is also possible in the agricultural sector. Figure 6 shows the potential for energy savings by 2030 in different world regions, derived by summing estimates from individual countries. These were calculated as emission savings which were calculated as follows:

In many regions, non-climate policies, including macroeconomic, agricultural and environmental policies, have greatest impact on agricultural mitigation options. These are reviewed elsewhere (Smith et al. 2007). Some evidence suggests that, despite significant biophysical potential for GHG mitigation in agriculture, very little progress has been made since 1990 and little is expected by 2010. There are barriers to implementation which may not be overcome without policy/economic incentives (Smith et al. 2005a).