1. Introduction
The agri-food sector is responsible for a high share of direct and indirect environmental impacts [
1]. Livestock production, including feed production, is a major contributor to the world’s total greenhouse gas (GHG) emissions [
2,
3,
4]. This is the result of energy use in food production (mostly fossil fuels consumed in machinery used for agricultural operations) [
5], fertilizer production [
6], carbon loss due to direct and indirect land transformation and occupation [
2,
7], and methane (CH
4) and nitrogen oxide (N
2O) emissions [
2]. Livestock is also one of the main drivers in biodiversity loss due to land use changes required for feed production [
4,
8].
Sown biodiverse permanent pastures rich in legumes (SBP) are a mixture of selected high-yield legumes and grasses that provide quality animal feed [
9,
10,
11]. These pastures have been developed since the 1960s in Portugal, with the main objective of increasing grassland productivity and sustainable stocking rates, by sowing mixes of up to 20 species/cultivars of legumes and grasses [
10,
12]. The main co-effect of increased productivity is the increase in soil organic matter (SOM), which translates into carbon storage and consequent sequestration from the atmosphere. SBP are estimated to have sequestrated 3.5 million tons of carbon dioxide (CO
2) as soil carbon between 1996 and 2008 in an area of 94,260 hectares in Portugal [
9,
13,
14]. Between 2008 and 2014, these pastures were additionally supported by the Portuguese Government through the Portuguese Carbon Fund (PCF), to assist the country in keeping with Portuguese goals for the Kyoto Protocol under the Agriculture, Forestry, and Other Land Uses activities of Article 3.4. The PCF supported the installation and maintenance of SBP using a system of service payments for carbon sequestration. As a consequence, 1095 farmers installed SBP in an area of more than 4% of the country’s agricultural land [
10]. The majority of the area of SBP (over 90%) was installed in the agricultural region of Alentejo; namely in highly important “Montado” areas, an agro-forestry landscape of high biodiversity value [
15].
The high yields of SBP are achieved because of (a) the biodiversity effect on productivity, (b) the selection of high-yield and locally adapted grass and legume species, and (c) technical support for precision management that often involves phosphate fertilization and pH correction using limestone [
10]. To farmers, the costs with management are counter-weighed by a decrease or elimination of the need for commercial concentrate feeds for ruminants, because SBP increased productivity when compared with their most common alternative land use system observed in farms before sowing—semi-natural pastures (SNP). Savings in concentrate feeds are an additional ecological and economic benefit for the country [
9].
Despite the savings in feed consumption, SBP require more agricultural operations and inputs for production, such as seeds production and machinery for installation, and regular applications of phosphate fertilizer and limestone for maintenance. So far, a plot-level assessment of GHG emissions has been carried out [
16], but no life cycle assessment (LCA) study has ever been carried out to also assess indirect emissions. The balance between increased material and energy use in SBP and reduction of concentrate thus remains indeterminate.
The objective of this study was to assess the environmental impact, in terms of global warming potential (GWP), of substituting commercial feeds with SBP for cattle production. We compared two animal feed scenarios: (1) SNP and commercial feed and (2) a combination of SNP and SBP. We used a cradle-to-farm-gate approach, considering the impact of all supply-chain and on-farm emissions and also including carbon sequestration in SBP.
4. Discussion
The results of this paper assess and compare the environmental impacts of two realistic pasture systems in Portugal as a way to assess the consequential effects of installing and maintaining SBP. The comparison was performed using an innovative approach that resorts to energy and nutritional equivalences between the systems first proposed for the specific comparison of pasture systems by Teixeira [
9]. The analysis was greatly simplified because of the fact that there is ample empirical evidence that one scenario does follow from the other one. In the 1000 farms of the PCF project, SBP were installed in former areas of SNP. Therefore, rather than performing a dedicated LCA for each system, which would be more burdensome and include redundant processes and flows, we followed a simpler and easier to interpret comparative approach. The consequential approach followed is interesting in this case because SBP cannot be assessed without including effects that are inextricably connected with their installation and added material and energy consumption, as well as feed avoided. So far, however, this analysis had stayed at the farm level and included only direct emissions. The approach followed here highlights these effects rather than diluting them within the entire carbon footprint of meat products for each of the systems.
The results show that phosphorus fertilizer and lime application are the main contribution to the total emissions assessed for SBP pastures (for installation, but also for maintenance). This issue was identified in Teixeira et al. [
10], but never duly quantified. Phosphorus fertilizer is not only an environmental issue in SBP, but also an economic issue, as identified by Almeida et al. [
44], as the fertilizer costs have been gradually increasing over the years. As future work, an economic comparison between systems should be made in order to complete the assessment. For example, through life cycle costing (LCC), as suggested by several authors [
45,
46,
47], with the purpose of emphasizing relationships and trade-offs between the economic and life cycle environmental performance.
Regarding the equivalence method, it is possible to conclude that energetic equivalence provides different emissions in comparison with crude protein and fiber equivalences. For example, when crude fiber is used to establish the equivalence of scenarios, the difference between scenarios is about 50% higher (low forage feed) than when using gross energy. This is because of the fact that SBP have higher nutritional quality and thus can replace more quantity of feed. For high forage feed, the variation between nutritional equivalences is significantly lower as silage dominates the composition of the feed and the variability of the indicators is smaller.
To assess the variability of the results when faced with different assumptions, we used sensitivity analysis whenever possible. One example is the required feed consumption for the desired growth level of the steer. This analysis revealed high sensitivity of results depending on the animal feed requirements considered (0.5% or 1.5% of live weight). The difference between scenarios increases with higher feed requirements because of the high unitary impacts of the concentrate.
There were, however, several assumptions made in the work that influence results and we were unable to include them in the (quantitative) sensitivity analysis. Reverting some of these assumptions would likely cause the difference in emissions between the scenarios to increase, while for others, it would decrease. Starting with assumptions that would make the relative benefit of sowing SBP larger, there is the fact that it is not entirely true that SBP cannot be grazed during the first year. Although it is a good management practice to avoid grazing during the establishment of the pasture to ensure a good seed bank for the following years, during the summer months, it is possible (and desirable) to graze in order to ensure good germination of grasses and legumes during autumn, as well as to control infesting plants. This means that animals do take some of their feed from SBP during the first year, and consequently there is some replacement of concentrate. This fact would improve the performance of the system with SBP. Additionally, relatively to commercial feeds, we assumed that all ingredients were produced locally (in the farm where the pastures are situated). This is a conservative approach that underestimates the impacts of feeds, mostly because it disregards feed transport, which can be an important phase because of consumption of energy and associated GHG emissions (can contribute up to 15% of total emissions of the feed) [
5,
42]. This approach could only amplify the differences between systems, as considering more transportation of ingredients would make the unitary impacts of the feed larger. These assumptions do not compromise the main conclusions regarding the comparison of scenarios.
Regarding assumptions that would lower the difference between scenarios, the most important one is that we did not consider any feed consumption in SBP. This was a simplification required because of the lack of data on the distribution of DM yield in SBP throughout the year. Good practices for SBP management state that at least during early spring, grazing should be light to ensure a healthy re-establishment of the seed bank. It is also possible that during the summer months, yields decrease as a result of the lack of rainfall—as these are rainfed pastures. If we assume a worst case scenario where there is no pasture intake during one month in early spring and two months in the summer, that would mean that for one quarter of the year, cattle would require feed. This means that all results presented here would be approximately reduced by one quarter. For example, the avoided impact per kg of meat would be approximately 13 (rather than 17) for LF feed and 5 (rather than 7) kg CO
2e/kg meat for HF feed. This would also not compromise the main conclusions from this study. Also relevant to note is that SBP are technically more demanding from farmers and their installation can fail. A critical problem is the lack of prevalence of legumes. However, the sample of farms used to collect data used in this assessment [
28] included some cases where the installation was less than fully successful. Therefore, the assessment in this work was already carried out using an average of all likely situations. Finally, we used two realistic feed formulations that enabled us to test the effects of silage, but there is a vast array of feeds used for beef cattle in Portugal. Feed formulations can and should also be optimized to achieve specific animal growth requirements and to better complement nutrition obtained from pastures, as well as to decrease GHG emissions [
48]. In this paper, we did not assess those potentially optimized feeds. We also used a commercial feed formulation rich in cereal grains. Some commercial feeds include a higher share of co-products (cereals husk, meals, and others). Those feeds can have lower impacts [
49], depending on the exact co-products. Here, we additionally considered that changing animals diets would not change live weight gain, which is a simplification that can change the results for avoided impact per kg of meat. Further, the feeds used were originally intended for fattening, and not for less than one-year-old steers or adult cows. Given the high energy provided by the feed to the animal, the fact that we consider this particular feed here means that the difference between scenarios could even be larger (as less area of SBP might be needed to replace feeds). In the future, results can be improved by considering the change in live weight gain due to changes in diets and the different commercial feeds consumed by different animal classes.
Other assumptions made are more unclear regarding the consequences for results. We assumed that ε = 2.33, which is equal to the ratio between stocking rates in SBP and SNP, according to Carneiro et al. [
28]. The definition of this parameter is a ratio between nutritional contents, but the lack of a representative breakdown of the species composition of SNP prevented us from using that definition literally. Therefore, we used stocking rates as a proxy. We also assumed, in the calculation of nutritional equivalencies between scenarios, that the proportion of the area of SBP was equal to the proportion between dry matter.
Here, we also used the IPCC AR5 GWP factors with CC feedback for impact assessment. The GWP factors with CC feedback are higher than the factors without carbon feedback (e.g., N2O with CC factor: 298; N2O without CC factor: 265). Nevertheless, the results are not considerably affected by this. If IPCC GWP without CC feedbacks had been used, the savings due to conversion to SBP would decrease by less than 6% (i.e., less than 0.1 t CO2e/ha). Finally, we did not consider N2O emissions from pastures litter. We assumed that the litter fraction is equal in both situations (SNP plus commercial feed and SBP plus SNP).
Future developments to this paper should be considered in the future. One example is the reassessment of the N
2O emissions from legumes. Here, we used a default value to estimate this flux. However, the factor was not dependent on the fraction of legumes (responsible for soil nitrogen fixation and soil N
2O emissions). Further, the IPCC [
38] does not even consider N
2O emissions from biological nitrogen fixation because of a lack of evidence that they contribute significantly to soil N
2O emissions. If we had disregarded those emissions, the avoided impact due to SBP installation would be even higher. Finally, an option to improve the estimation of livestock intake and animal growth may be the use of mass-balance approaches [
16].
Ultimately, we converted our reference flow into a mass unit (kilograms of meat as a final product) as it is the choice of most studies targeting meat production (among others, [
41,
50,
51,
52]). Alternatively, another common unit in other studies is one live animal (among others, [
53,
54]). This mass unit should, however, be adjusted for nutritional value, as performed by Poore and Nemecek [
42]. Meat from SBP and meat produced using concentrate feed likely diverge in terms of their quality, as meat from SBP is nutritionally superior [
10], as well as carcass yield, so new differences may arise if this differentiation is included in future LCAs of these systems.
Species variability in SBP is the most relevant uncertainty not addressed in this paper. Further improvements in the estimation of carbon sequestration can also be made, namely through the application of a geographical LCA, where the difference between scenarios is assessed for multiple sub-regions within Alentejo. The key parameter missing to perform such an evaluation is a regionalized estimation of the yield of SBP. A remote sensing approach is a viable alternative to estimate regionalized aboveground biomass productivity [
55,
56]. This approach also has limitations, such as, for example, the influence of tree cover. This effect is particularly relevant in these “Montado” agri-forestry regions due to high tree cover density (e.g., 170 trees/ha of
Quercus suber [
57]), 2 as trees also influence aboveground biomass productivity [
58]. Here, we used the process-based model RothC calibrated for SBP (root to shoot ratio, livestock intake fraction, livestock excretion during grazing, and easily decomposable plant material (DPM)/resistant plant material (RPM) ratio), which can consider some site-specific conditions if available—such as regionalized aboveground productivity and stocking rate. Regionalization would be welcome for other parameters of the study. For example, we used a single mixture of sown species, but mixtures are tailor-made for different site conditions (such as soil density and soil organic matter (SOM)). Even more importantly, the species found in the pasture some years after installation could deviate considerably from their proportion in the mixture of seeds used. This could considerably change the nutritional value of the pasture. For example, in the same group, average CP content in two distinguished clover species can range from 19% DM (for
Trifolium subterraneum) and 25% DM (for
Trifolium balansae). This difference can be even more significant between different plant groups [
29]. Further work is required to better understand species variation with location, management practices, and other aspects not identified yet. To do this, more field work is required, including multiple sites studied over a long period.
In this study, we only focused on the contribution of SBP for climate change mitigation. Nevertheless, in the future, other ecosystem services should be addressed, for example, leaching reduction or effects on biodiversity [
59]. There is evidence that SBP reduce nitrogen leaching when compared with SNP [
60], but not in a life cycle assessment approach. Besides placing more emphasis on the regionalization of the inventories and activity data, these additional impact categories should also be assessed using regionalized life cycle impact assessment methods [
31]. There are now multiple highly regionalized methods available, mainly focusing on land use [
61,
62] and biodiversity impacts [
63,
64].